35347 Edited by Robert Holzmann and Edward Palmer 2000 2050 Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes Pension Reform Pension Reform Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes Edited by Robert Holzmann and Edward Palmer THE WORLD BANK WASHINGTON, D.C. ©2006 The International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington, DC 20433, USA Telephone: 202-473-1000 www.worldbank.org E-mail: feedback@worldbank.org All rights reserved. 1 2 3 4 09 08 07 06 The findings, interpretations, and conclusions expressed herein are those of the author(s) and do not necessarily reflect the views of the Executive Directors of The International Bank for Reconstruction and Development / The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The bound- aries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Recon- struction and Development / The World Bank encourages dissemination of its work and will nor- mally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; telephone: 978-750-8400; fax: 978-750-4470; Internet: www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202- 522-2422; e-mail: pubrights@worldbank.org. ISBN-10: 0-8213-6038-8 ISBN-13: 978-0-8213-6038-5 eISBN-10: 0-8213-6166-X eISBN-13: 978-0-8213-6166-5 DOI: 10.1596 / 978-0-8213-6038-5 Library of Congress Cataloging-in-Publication Data Pension reform: issues and prospects for non-financial defined contribution (NDC) schemes / edited by Robert Holzmann, Edward Palmer. p. cm. Includes bibliographical references and index. ISBN 0-8213-6038-8 1. Defined contribution pension plans. 2. Pensions. I. Holzmann, Robert. II. Palmer, Edward. III. World Bank. HD7105.4.P465 2005 331.25'22--dc22 2005045767 Cover photos: Curt Carnemark/The World Bank; Chris Stowers/Panos Pictures. Contents Preface xi Anna Hedborg Acronyms and Abbreviations xiii 1 The Status of the NDC Discussion: Introduction and Overview 1 Robert Holzmann and Edward Palmer PART I. THE CONCEPT OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEM--VARIATIONS ON A THEME 2 What Is NDC? 17 Edward Palmer 3 What Are NDC Systems? What Do They Bring to Reform Strategies? 35 Axel H. Börsch-Supan 4 Non-Financial Defined Contribution Pensions: Mapping the Terrain 57 Nicholas Barr 5 Conceptualization of Non-Financial Defined Contribution Systems 71 Contributors: Assar Lindbeck, Peter Diamond, Salvador Valdés-Prieto PART II. CONCEPTUAL ISSUES OF DESIGN AND IMPLEMENTATION 6 Demographic Uncertainty and Evaluation of Sustainability of Pension Systems 95 Juha M. Alho, Jukka Lassila, and Tarmo Valkonen Discussant: Sergio Nisticň 7 The Rate of Return of Pay-As-You-Go Pension Systems: A More Exact Consumption-Loan Model of Interest 117 Ole Settergren and Boguslaw D. Mikula Discussant: Ronald Lee iii IV PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES 8 A Market Method to Endow NDC Systems with Automatic Financial Stability 149 Salvador Valdés-Prieto Discussant: Marek Góra 9 Conversion to NDCs--Issues and Models 169 Edward Palmer Discussants: Marek Góra, Elsa Fornero 10 NDCs: A Comparison of the French and German Point Systems 203 Florence Legros Discussant: Marek Góra 11 Toward a Coordinated Pension System in Europe: Rationale and Potential Structure 225 Robert Holzmann Discussant: Bernd Marin 12 NDC Pension Schemes in Middle- and Low-Income Countries 293 David Lindeman, David Robalino, and Michal Rutkowski Discussants: Monika Queisser, Elaine Fultz 13 How Much Do People Need to Know about Their Pensions and What Do They Know? 325 Annika Sundén Discussant: John Ameriks 14 Lashed to the Mast? The Politics of NDC Pension Reform 345 Sarah M. Brooks and R. Kent Weaver Discussants: Daniele Franco, Agneta Kruse PART III. LESSONS FROM COUNTRIES WITH NON-FINANCIAL DEFINED CONTRIBUTION SCHEMES 15 NDC Strategy in Latvia: Implementation and Prospects for the Future 397 Edward Palmer, Sandra Stabina, Ingemar Svensson, and Inta Vanovska , 16 The NDC System in Poland: Assessment after Five Years 425 Agnieszka Chlo´n-Domi´nczak and Marek Góra 17 The NDC Reform in Sweden: The 1994 Legislation to the Present 449 Bo Könberg, Edward Palmer, and Annika Sundén 18 NDCs in Italy: Unsatisfactory Present, Uncertain Future 467 Daniele Franco and Nicola Sartor 19 Implementing the NDC Theoretical Model: A Comparison of Italy and Sweden 493 Sandro Gronchi and Sergio Nisticň CONTENTS V PART IV. THE POTENTIAL OF NON-FINANCIAL DEFINED CONTRIBUTION SCHEMES IN OTHER COUNTRIES' REFORMS 20 Investigating the Introduction of NDCs in Austria 519 Bernhard Felderer, Reinhard Koman, and Ulrich Schuh 21 The NDC Reform in the Czech Republic 551 Agnieszka Chlo´n-Domi´nczak and Marek Mora 22 The German Public Pension System: How It Will Become an NDC System Look-Alike 573 Axel H. Börsch-Supan and Christina B. Wilke 23 The Spanish Pension System: Issues of Introducing NDCs 611 Carlos Vidal-Meliá and Inmaculada Domínguez-Fabián 24 Reforming Social Security in Japan: Is NDC the Answer? 639 Noriyuki Takayama Contributors 649 Index 657 FIGURES 5.1 A Taxonomy of Social Security Systems 72 6.1 Age Distribution: Actual and Three Scenarios 98 6.2 Median Relative Error of Fertility Forecast 100 6.3 Median Relative Error of Mortality Forecast 100 6.4 Replacement Rates and Contribution Rates in Lithuania in 2050, with = 0 (left) and = 1 (right) 104 6.5 Standard Deviation in 100 Simulations of Pension Contribution Rate in the Finnish Private Sector TEL System under Alternative Funding Rules 108 6.6 Pension Contributions in the Finnish Private Sector TEL System 109 7.1 Illustration of Equations 7.7 and 7.8 122 7.2 Turnover Duration in Sweden, 23 Annual Changes, percentages 1981­2003 123 9.1 Contribution Base Growth with Wage Per Capita Growth of 3 Percent and Labor Force Growth/Decline of 0.3 Percent 178 11.1 Pension Expenditure in EU and EUA Countries (plus Croatia), 2000 or latest (percentage of GDP) 226 11.2 Actuarial Fairness (the Swedish and Polish NDC System) vs. Subsidizing Early Exit and Penalizing Working Longer (the Austrian and Czech DB Systems), as compared to the German DB Intermediate, 2003 268 11.3 Swedish NDC Actuarial Neutrality vs. Austrian DB, Amount of Labor-depressing Adverse Redistribution, 2003 269 11.4a Europe Aging at Multiple Speeds, Lags and Peaks, 1995­2050: Subregional Deviations from All-European Regional Averages (percent) 270 11.4b When the Aging Process Will Reach Its Peak: Average Annual Increase of the Population Above Age 60 in Three Scenarios, by European Subregions (percent) 271 VI PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES 11.4c When the Aging Process Will Reach Its Peak: Average Annual Increase of the Population Above Age 60 in Three Scenarios, by country (percent) 272 11.5 Corporatist Pension Disharmonies: Cost of Minimum Contribution One Month of Retirement, Austria 2003 276 11.6 Variation in Public Expenditure for Disability Related Programs (percentage of GDP) 284 11.7 Disability Status of Disability Benefit Recipients 285 11.8 Country Differences in Age-specific Inflow Rates (ratio of age-specific inflow rates over age group 35­44, 1999) 288 12.1 Implementing a Complementary Noncontributory Pension 305 12A.1 Illustration of the Dynamics of the Average Wage, the Wage Bill, and Coverage Rates 314 12A.2 Primary Balance and Last-Year Replacement Rates (Stable Environment) 315 12A.3 Contributions Minus Expenditures (Volatility Environment) 316 12A.4 Income Replacement at Age 60 (Volatile Environment) 317 13.1 Self-Reported Knowledge about the Swedish Pension System 335 13.2 Share of Participants Needing Additional Information 337 15.1 Life Expectancy at Birth in Latvia 404 15.2 The Latvian Population 410 15.3 Economically Active Population (age 15 up to the minimum pension age) 411 15.4 System Dependency Ratio 412 15.5 Number of Old-Age Pensioners with Grants under the Old Law, Transition, and New Law 413 15.6 Old-Age Pension Expenditures As a Percent of the Contribution Base 414 15.8 Total Social Insurance Expenditures As a Percent of the Contribution Rate 415 15.7 Old-Age Pension Expenditures As a Percent of the Contribution Base 415 15.9 Total Social Insurance Costs As a Percent of the Contribution Base (FDC with a 10% contribution rate from 2010) 417 15.10 Total Social Insurance Costs As a Percent of the Contribution Base (FDC with a 6% contribution rate from 2008) 418 15.11 Total Social Insurance Costs As a Percent of the Contribution Base 419 15.12 Total Assets of the FDC Pension Scheme As a Percent of GDP 419 16.1 Value of Hypothetical Pension for the Initial Capital Calculation 429 16.2 State Budget Subsidies to Social Insurance Fund 433 16.3 Change in Pension Value Because of Increases in Life Expectancy 435 16.4 Demographic and System Dependency Rates, 2002­50 440 16.5 Number of Pensioners, 2002­50 440 16.6 Pensioners Distinguished by Participation in the Old or New System, 2002­50 441 16.7 Covered Workers: Total and Those with NDC and FDC Accounts, 2002­50 441 16.8 Non-Financial Scheme Contribution Revenues (Old System and NDC), 2002­50 (% of covered wage bill) 442 16.9 Pension System Revenue and Expenditure, 2002­50 (% of covered wage bill) 443 18.1 Effective Equilibrium Contributory Rates (legislated transition) r = 3.0%; g = 1.5% 472 20.1 Unused Capacity versus Tax Force to Retire 524 20.2a Replacement Rate, NDC Pillar 531 20.2b Total Replacement Rate, NDC Pillar with Fully Funded Pillar (5 percent contribution rate) 543 CONTENTS VII 20.3 Expenditure As Percent of GDP, NDB versus NDC Pillar 544 21.1 The Increase in the Elderly Dependency Ratio in EU Countries between 2004 and 2050 552 21.2 Share of Economic Active Populationa 553 21.3 Replacement Rates by Earnings Level in the Czech PAYG System 554 21.4 Simulation of NDC versus DB Pension in the Czech Republic (2004) 559 21.5 Functioning of the Minimum Pension Guarantee 563 21.6 An Alternative: Flat Benefit plus NDC Pension 563 21.7 Replacement Rates: Minimum Guarantee versus Flat Benefit 564 21.8 Simulated Value of Old-Age Pension: The Role of Initial and NDC Pension for Transition Cohorts 566 21.9 Employed and Pensioners: Increased Retirement Age Scenario 567 21.10 Average Pension as a Percent of Average Wage: Increased Retirement Age Scenario 567 21.11 Expenditures and Revenue of the Old-Age Pension System: Increased Retirement Scenario 568 21.12 Employed and Pensioners: Constant Retirement Age Scenario 569 21.13 Pensions As a Percent of Average Wage: Constant Retirement Age Scenario 569 21.14 Expenditures and Revenues of the Old-Age Pension System: Constant Retirement Age Scenario 570 22.1 Development of the Economic Dependency Ratio 576 22.2 Pathways to Retirement, 1960­2002 584 22.4 Distribution of Retirement Ages, 1970, 1975, 1980, and 1995 585 22.3 Average Retirement Age, 1960­1995 585 22.5 Retirement Age with and without "Actuarial" Adjustments (1992 and 1999 reforms) 589 22.6 Depth of Subsidies to Riester Pensions 596 22.7 Filling the Pension Gap 598 22.8 Composition of Retirement Income by Birth Cohort 599 22.9 Development of Pension Levels prior to and after the 2001 Reform 600 22.10 Contribution Rates prior to and after the 2001 Reform 601 22.11 The Effects of the Sustainability Factor on the Development of Contribution Rates 603 22.12 The Effects of the Sustainability Factor on Pension Levels 604 22.13 Total Pension Level Including Private Riester Pensions 605 22.14 Popularity of Pension Reform Options 606 23.1 Historical Evolution of the Various Arithmetic Rates in Spain, 1961­2004 623 24.1 Per Capita Income by Age in Japan 643 24.2 Drop Out Rate from Social Security Pensions (nonemployees) 643 TABLES 3.1 Dimensions of Pension Systems 41 6.1 Distribution of the Social Security Contribution Rate in Lithuania (percent) 102 6.2 Distribution of Policy Effects on the Social Security Contribution Rate (percent) 102 6.3 Joint Probabilities of the Contribution Rate Being Lower Than c* and the Replacement Rate Being Higher Than r* in 2050 105 6.4 Pension Contributions during Selected 10-Year Periods in the Finnish TEL System (fractiles of the predictive distribution) 107 VIII PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES 6.5 Probability of Excess Funding during 2005­69 110 7B.1 Effect of a Shift in Income Pattern on Cohort Contributions and Benefits 129 7B.2 Effect of a Shift in Income Pattern on Turnover Duration and Pension Liability 130 7B.3 Example 1: Income Statement, and Balance Sheet 131 7B.4 Effect of a Shift in Mortality on Cohort Benefits 133 7B.5 Effect of a Shift in Mortality Pattern on Turnover Duration and Pension Liability 133 7B.6 Example 2.1: Income Statement and Balance Sheet 135 9.1 Varying Results over Time of a DB Scheme that Resembles an NDC Scheme 172 9.2 Transition Models for Initial Capital and Introduction of NDC 182 9.3 Other Transition Issues 188 10.1 Schemes and Contributions in the Main French Pension Scheme 204 10.2 Life Expectancy at Age 60, by Socioprofessional Categories, in Years (the case of France) 211 11.1 Projections of Old-Age Dependency in EU and EUA Countries, 2000­50 (ratio of people aged over 64 to working age population, percent) 228 11.2 Public Pension Expenditure in EU and EUA Countries in 2000­50 (percentage of GDP) 229 11.3 Labor Force Participation, Male and Female, in EU and EUA countries, 1960, 1980, 2000, and 2050 (percent) 230 11.4 Pension Arrangements for Widows/Widowers and Divorcees in EU and EUA Countries around 2000 231 11.5 Changing Family Structures: Divorces and Marriages in EU and EUA Countries around 2000 (per 1,000 people) 236 11.6 Selected Work Arrangements in Europe, 1988 and 1998 (percent of total employment) 237 11.7 Scope of Funded Pensions in EU and EUA Countries around 2002 248 11.8 Scope and Form of Social Pensions in EU and EUA Countries around 2002 253 11.9 Same Lifetime Contributions, Highly Different Pension Entitlements for Atypically Employed Part-Timers and No-Career Regular Full-Time Employees, Austria, 2003 (euros) 273 12.1 Changes in Spending and Primary Balance in Countries with NDC Systems (percent of GDP) 298 12A.1 Optimal Retirement, Savings, and Labor Supply Decisions under Alternative Benefit Formulas 310 12A.2 Model Parameters 311 12A.4 Pension Schemes under Consideration 313 12A.3 Parameters Used in the Dynamics Simulations 313 13.1 Types of Pension Schemes 326 13.2 The Orange Envelope, 2003 331 13.3 Knowledge of Participants about the Swedish Pension System, 2003 336 14.1 Non-financial Pension Provisions As a Continuum from Defined Benefit to Defined Contribution 349 14.2 The NDC Pension Continuum in Practice 352 14.3 Potential Relationships between Structural Variables and NDC Choice 358 14.4 Country Characteristics and NDC Reform Outcomes 362 14.5 Potential Erosion in NDC Pension Systems 372 15.1 Overview of Some Key Economic Indicators for NDC Pensions 401 15.2 Key Demographic Pension Ratios 402 CONTENTS IX 16.1 Alternative Approaches to Pension Reform 427 16.2 Introduction of the New System (age groups) 428 16.3 Changes in Life Expectancy of Persons at Retirement Age, 1995­2002 431 16.4 Number of Covered Workers, 1998­2002 (thousands of persons) 431 16.5 The Collected Contribution Revenue and Covered Wage Bill As a percent of GDP 432 16.6 Simulations of Replacement Rates for Different Cohorts, both Men and Women, Depending on Retirement Age and Taxation of NDC Returns (percent of final wage) 434 16.7 Projected Effects in Poland Compared with Projections in Selected OECD Countries 436 16.8 Demographic Assumptions (baseline scenario) 437 16.9 Macroeconomic Assumptions 438 16.10 Projection's Results 439 17.1 Assets and Liabilities NDC 2001­4 (Swedish kronor, billions) 458 18.1 Mandatory Public Pension Scheme for Employees 471 18.2 Equilibrium Contributory Rates for Employees 473 18.3 Selected Indicators of the Italian Mandatory Pension Scheme for Employees Old-Age and Seniority Pensions for Employees--Average Seniority 474 18.4 Incentive to Postpone Retirement by One Year 480 19.1 Protection of Prereform Entitlements in Sweden and Italy 506 20.1 Status Quo Scenario 532 20.2 2000 Legal Situation Scenario 534 20.3 Low Labor Force Participation Scenario 536 20.4 Medium Labor Force Participation Scenario 538 20.5 High Labor Force Participation Scenario 540 21.1 Parametric Reforms in the Czech PAYG Pension Scheme (1990­2002) 555 21.2 Real Net Rate of Return (RNRR) of the Czech Second Pillar (in percent per year) 556 21.3 Costs and Benefits of the Switch from PAYG to Fully Funded Financing 557 21.4 Employment Rate in 2004 559 22.1 The German Public Pension System from Bismarck until Today 574 22.2 Old-Age Pensions (1972 legislation) 579 22.3 Adjustment of Public Pensions by Retirement Age 581 22.4 Overview of the Core Elements of the Riester Reform 590 22.5 Direct Savings Subsidies 592 22.6 Minimum Savings (euros/year) 592 22.7 Maximum Savings (euros/year) 592 22.8 Types of Occupational Pension Systems 595 23.1 Ratio of the Number of Contributors to the Number of Contributory Pensions 614 23.2 Population Projections for Spain 2002­50 Based on the 2001 Census by the National Institute of Statistics: Hypothesis I or Central Scenario 616 23.3 Transfer Components and Real Internal Rate of Return (IRR), Law 26/1985 (by system as a percentage of total pensions paid) 617 23.4 Transfer Components and Real Internal Rate of Return, Law 24/1997 (by system as a percentage of the total pensions paid) 618 23.5 Forecast Spending on Pensions for 2025 and 2050 619 23.6 Projections for the Spanish Contributory Pension System 2003­50, Central Scenario 621 X PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES 23.7 Ratio between the Initial Pension in Spain (current system) and the Initial Pension Using Calculation Rules from Notional Formulae Applied in Other Countries 623 23.8 IRR According to Number of Years Contributed, Men (Women) 624 23.9 Real Internal Rate of Return (IRR) and Replacement Rate (RR) for Retirement Age (Xr), with 35 Years of Contributions, Men and (Women) 624 23.10 Real Internal Rate of Return (IRR) and Replacement Rate (RR) for Retirement Age (Xr), with Different Numbers of Years Contributed, Men and (Women) 625 23.11 Formulae for Calculating the Initial Pension and Its Later Variation 627 23.12 Average Expected Replacement Rate (retirement age 65) 628 23.13 Average Internal Rate of Return (IRR) and Expected Deviation for Men (M) and Women (W) (retirement age 65) 629 23.14 Timetable for Adapting to Pension Reform in Spain 631 24.1 Balance Sheet of the KNH before Reform as of March 31, 2005 (yen, trillions) 642 24.2 Balance Sheet of the KNH after Reform as of March 31, 2005 (yen, trillions) 645 24.3 Balance Sheet of the KNH: An Alternative as of March 31, 2005 (yen, trillions) 646 BOXES 8.1 What Is Meant by Funding 151 8.2 Why the Rate of Return on Capital Must Be Larger than the Growth Rate 154 12.1 Gradations from "Flawed DB" to "Good DB" and NDC 296 16.1 Tasks and Managing Institutions 444 Preface THE PREVIOUS DECADE HAS BEEN A DECADE OF PENSION REFORM throughout the world. In high income countries the driving force has been the threat that current systems will become unaffordable in coming decades, with demographic developments presenting a major risk. In another setting, countries in the process of transition from a command to a market economy are confronted with the challenge of introducing a public pension system that will provide social security in old age but that also supports the fundamentals of a market economy. In the latter sense, it is important to examine carefully the experiences of developed market economies. Even in these countries, the driving force behind reform is demographic change and affordability. In a third setting, middle and lower-middle income countries are faced with the question of what system will best serve the interests of their specific country goals for the future. In all of these settings "NDC"--non-financial defined contribution--pension schemes have been on the agenda in discussions of possi- ble options. Sweden is one of the handful of countries to have implemented an NDC scheme in the 1990s. The Swedish NDC pension reform has its roots in the work of a Working Group on Pensions that began its work at the end of 1991. After about two years of public discussion of the central ideas of the reform, following the Working Group's publication of the pro- posed framework of the reform in 1992, the final proposal was published in February 1994, and passed as legislation by the Parliament in June 1994. After a short political pause, fol- lowing the election in the fall of 1994, the process of implementation began in 1996. This encompassed rewriting extensive benefit and tax legislation, including difficult transition legislation, as well as developing new information technology, a new annual statement for participants, and a new electronic information system for participants. The final concep- tual bolt was put in place with the construction of a mechanism to assure long-run balance between assets and liabilities, the automatic balancing mechanism. Account statements were sent out for the first time in 1999. Benefits have been paid out according to the new rules since 2001. Now, in 2005, all of the nuts and bolts are in place and the system has been operating over several years. Additionally, the political consensus behind the Swedish pension reform is as strong today, in 2005, as it was in 1994. From the mid-1990s, NDC came into its own as a concept and was implemented in four EU countries. Italy, Latvia, and Poland are the other three. Following these examples, NDC has become a reform option considered by many coun- tries. xi XII PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES It is now safe to say that the reforms in the four original NDC countries in Europe gen- erated considerable interest especially among European neighbors, but also world-wide. The interest generated in Europe is easy to understand since most of Europe has a pay-as- you-go tradition, and NDC constitutes a new way to "organize" a mandatory, universal pay-as-you-go pension system. And, the majority of European countries are discussing reform of their systems in one way or another. What's more, we now have some experi- ence of NDC schemes implemented during the 1990s. Against this background, it felt par- ticularly relevant for Sweden to host a conference devoted to discussing both the conceptual and institutional aspects of NDC. This would provide an opportunity for Swedish reformers to reflect over the design of the Swedish system in the presence of international pension experts, from both the academic and the policy communities. The goal was even more ambitious, however. The goal was to contribute to creating a synthe- sis of current knowledge on this new topic. This book is the realization of that goal. The Swedish National Social Insurance Board was particularly pleased to organize the conference jointly with the World Bank, which served to broaden the discussion to encom- pass considerations of importance for World Bank client countries. The conference was held during two days in late September 2003 on the island of Sandhamn, one of the outer- most islands in the Baltic Sea Archipelago outside Stockholm. The discussion of papers and ideas was spontaneous and challenging, with a whole room of experts from varying affiliations, regions of the world and intellectual "priors." My experience is that intellectual reciprocity is a rare outcome of seminars and confer- ences, and that staying with one's own hobby-horse is the normal modus operandi. This was certainly not the case on Sandhamn, however. Here, experts were not only speakers; they were listeners. Ideas were exchanged and developed in a positive, collegial atmosphere. This gave the feeling of something happening in the room, the feeling that at the end a new and shared understanding had been reached. I believe that all the participants felt that they left the island of Sandhamn with the feeling that they had participated in a con- versation that lifted their thoughts to a new level. Although we didn't call it NDC in the initial years in Sweden, I have personally been on the NDC journey for well over a decade. From late 1991 to June 1994 I represented the Social Democratic Party in the Government's Working Group on Pensions, and then became responsible for the reform process as Deputy Minister for Social Insurance, fol- lowing the election of 1994. In mid-1996, as Director General of the National Social Insur- ance Board, I became responsible for implementing the reform. Now, in 2005, over a decade after the historical decision of the Swedish Parliament, it is a pleasure for me to help provide a broad readership with the opportunity to share the fruits of the Sandhamn deliberations. And, I'm pleased to note that since the Sandhamn conference in autumn 2003, the contents of this book have developed considerably, not only with improved papers originally presented there, but with several new papers. I believe that this book will provide stimulating reading for pension experts and politicians representing both varied academic and country backgrounds. Anna Hedborg Acronyms and Abbreviations ABM automatic balancing mechanism AEI average earnings index AF adjustment factors ATP allmänna tilläggspension (earnings-related benefit in Sweden) CNAV régime général (basic general scheme in France) CWB bonds covered wage bill bonds DB defined benefit DB-PAYG defined-benefit pay-as-you-go DC defined contribution EP earnings points EPC Economic Policy Committee (European Union) FDB financial defined benefit FDC financial defined contribution FEDEA Fundación de Economía Aplicada (Spain) FFS fully funded schemes FP folkpension (flat-rate universal benefit in Sweden) FRD Demographic Reserve Fund (Poland) FUS Social Insurance Fund (Poland) GRV Gesetzlich Rentenversicherung (public retirement insurance in Germany) IC initial capital IFM integration to financial markets ILO International Labor Organization INPS Istituto Nazionale per la Previdenza Sociale (National Social Security Institute in Italy) IRA individual retirement account IRR internal rate of return ISSA International Social Security Association IT information technology KNH Kosei-Nenkin-Hoken (principal pension program for private-sector employees in Japan) LE life expectancy LMW Latvian Ministry of Welfare MEA Mannheim Research Institute for the Economics of Aging xiii XIV PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES MODPENS Modelo de Pensiones (Spain) NDB non-financial defined benefit NDC non-financial defined contribution (also, notional defined contribution) NIPSSR Japanese National Institute of Population and Social Security Research INE Instituto Nacional de Estadística (National Institute of Statistics in Spain) IRR internal rate of return OLG overlapping generations PAYG pay-as-you-go PDV present discounted value PPM Premiumpensionsmyndighet (Premium Pension Agency, the public clearinghouse for the new FDC scheme in Sweden) PROST Pension Reform Options Simulation Toolkit RPI retail price index SSA U.S. Social Security Administration TD turnover duration TFR total fertility rate TSSCI total social security contributions index VAT value-added tax ZUS Social Insurance Institution (Poland) Chapter 1 The Status of the NDC Discussion: Introduction and Overview Robert Holzmann and Edward Palmer* PENSION REFORM IS A WORLDWIDE ISSUE. There is hardly any country in the world where the reform of the existing pension system is not on the reform agenda. The drivers behind the need to reform differ somewhat between countries and across regions, but they share three communalities: first, short-term fiscal pressures typically dictate immediate action, while the looming, much larger problems resulting from an aging population and insufficient long-term resilience of systems remain inadequately addressed. Second, socioeconomic changes demand a rethinking of the basic ideas behind pension system design, some dat- ing back more than 100 years. Third, the challenges and opportunities of globalization require countries to pay more attention to the economic effects of pension schemes, includ- ing creating a larger risk pool that extends over occupation, branch, and sector, and enables easy portability of rights both within and across country borders.1 Against this background, in the 1990s a number of countries started comprehensive reforms of their pension systems, notably countries in Latin America and the transition economies of Central and Eastern Europe. These fundamental reforms were undertaken in an environment of difficulties in making parametric reform work, and they have aimed to at least put schemes on a financially sustainable basis. These systemic reforms have exhib- ited the potential as well as the limitations of moving toward comprehensive funded sys- tems. Although the potential benefits of funding are high, so are the transition costs and the requirements for setting up these financial schemes. The need to address the demo- graphic and economic reform pressures while avoiding creating additional burdens for future workers has generated international interest in a new genre of pension systems, non-financial or notional defined contribution schemes (NDCs).2 NDCs address the fiscal, political, social, and economic needs of reform while keeping the fiscal burden of reform low. Yet, do NDCs offer a panacea for pension reform or are they merely the latest fad? To address more systematically the question of what the NDCs bring to the pension reform table--their strengths and weaknesses--the World Bank and the Swedish National Social Insurance Board joined forces to organize an international conference, which was held October 2003 in Sandhamn, Sweden. The motivation for this conference was quite straightforward: The World Bank wants to know under what conditions and with what * Robert Holzmann is director, Social Protection Department, at the World Bank and Edward Palmer is professor of social insurance economics at Uppsala University and Head of the Division for Research at the Swedish Social Insurance Agency. 1 2 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES limitations this approach can be proposed to its international cliental. Although the expe- rience in Latvia and Poland is encouraging, the sample is still too small to draw general conclusions about implementing NDCs in transition countries. The Swedish reformers, who spearheaded this new reform approach, wanted to share their experience; but they wanted also to bring a group of experts together to discuss possible improvements in the design and implementation of this innovation. This conference volume presents the result of the deliberations of a number of leading economists, pension specialists, and policy makers from around the world. It contains 24 chapters plus a number of prepared comments. The chapters cover conceptual issues as well as the specific reform experiences of "pilot" countries. Most of the chapters were papers presented at the conference, but they have undergone major revisions since then. Some new chapters have been added to the proceedings after the conference to broaden further the contents of the book, making it a primer on the theoretical underpinnings of NDC and practical issues involved in its implementation. The structure of the remainder of the introduction is as follows: the first section outlines in somewhat more detail the background of the reform pressure and the promised advan- tages as well as drawbacks of NDC systems. The second section provides a short review of the individual chapters and their main findings. The final section sketches the outlook: the lessons so far, the potential of the NDC approach, and the main research gaps to be filled. Reform Pressures and the Promises of NDCs What are the main reform pressures on pension systems? What do NDCs offer to the toolkit of pension reform when viewed within a broader economic and political perspec- tive? What do the chapters contained in this volume say about what the NDC framework has to offer in addressing these questions? Our point of departure in answering these questions is to elaborate briefly on the pressures confronting pension reform as we enter into the twenty-first century. Fiscal pressures continue to be the trigger for the initiation of a pension reform. Reforms are inevitably driven by short-term budgetary pressure resulting from the increasing costs of what in the long-term are fiscally unsustainable public systems. The more important longer-term problems of sustainability due to population aging typically figure less prominently in the political debate and associated reform process. The short-term fiscal pressure for pension reform is always budgetary. The diagnosis is almost invariably that there are too many recipients and too few contributors. The under- lying problem in developed economies is system maturation resulting in increasing bene- fit levels mixed with decreasing de facto retirement ages, accentuated by population aging and the deterioration in demographic dependency ratios (the ratio of elderly to the work- ing-age population). Countries are frequently strapped down with high levels of expendi- tures on benefits, reflecting low and inflexible retirement ages set in times when life expectancy was considerably lower and in an economic environment that was character- ized by high economic growth and employment. Developing countries usually have the additional problem of a highly informal labor force, which prevents the translation of a ris- ing labor force into a low system dependency ratio (the ratio of beneficiaries to contribu- tors). On the contrary, the system dependency ratio is high as a result of closed systems (civil service) or strong incentives to stay in the informal economy, including design and implementation features of the pension scheme, or other obstacles to join the formal econ- omy. Stop-gap emergency action may move the system closer to short-term financial bal- ance, but this can be counterproductive by removing focus from design issues that need to THE STATUS OF THE NDC DISCUSSION: INTRODUCTION AND OVERVIEW 3 be addressed to deal with the underlying long-term imbalances and the need to provide a country with a lasting and transparent set of rules for the public system. The longer-term fiscal pressure is clearly linked to the aging of the population. All coun- tries are more or less affected by this pressure as birth rates fall toward or below the level needed to reproduce the population and as falling mortality rates generate substantial increases in the longevity of the older population. It is not a question of whether this prob- lem will become acute in a country, but when. The need to act and the broader political and economic implications of different courses of action have come into sharper focus during the last decade. Three issues are of particu- lar relevance for the design of pension systems: (1) Most of the future aging will take place in the developing countries where resources are scarcer. Careful design and cost-effective implementation are important for all countries, but the poorer countries are also less able to bear the fiscal burden of poor design. (2) Aging and population stagnation--and even decline--in the wealthier nations of the northern hemisphere will create arbitrage oppor- tunities with the much younger and growing population in the developing nations of the southern hemisphere. This calls for a pension system that allows easy cross-border porta- bility of acquired rights. (3) National interests are best served by managing population aging in a framework that supports efficient use of both capital and labor. Socioeconomic changes are a second main reason for reform. The public provision of retirement income for workers in the private sector originated in the late nineteenth cen- tury in the now developed countries--as a political response to urbanization and industri- alization. Schemes were based on the model of a working husband with a spouse responsible for the household and children, with a high probability of becoming a surviv- ing widow, a probability that increased with age. The first aim of pension systems was understandably to provide some relief to surviving children and widows. It is noteworthy that the minimum retirement age in Europe was originally set high in systems originating around 1900, typically 67­70 years, an age only a small share of male contributors could expect to reach. Typically, in the 1960s and 1970s, politicians capitalized on the shared interests of unions and management and lowered the pension age in many developed economies without paying attention to the long-term costs. This vogue proved to be short- lived, as all these countries began to tighten age and other eligibility requirements in the 1990s, largely with ad hoc adjustments in the rules. To date, there seems to be no known limit on the improvement in healthy years of life. This trend alone calls for a system design that provides incentives for workers to stay longer in the labor market as improvements in health and working conditions lead to longer healthy lives--improving the individual prerequisites for working when older, but also as the decline in fertility contributes to increase the demand. Toward the end of the twentieth century, two additional socioeconomic changes have altered the landscape: rising female labor force participation and changing family struc- tures. High female labor force participation, high rates of divorce, and the formation of new partner alliances call for a rethinking of benefits for widows--one of the raisons d'ętre of the nascent public schemes of a century ago. Although these changes are most pro- nounced in the postindustrial countries, they are increasingly a part of the picture in the developing world, too. Globalization, with the increasing integration of markets for goods and services, fac- tors of production, and knowledge, requires changes in the way public programs operate, including pensions. Such reforms are needed not only to reap the benefits of globalization but also to deal with the challenges it creates, including profound shocks resulting from technical innovations and shifts in the demand and supply of goods and factors of pro- 4 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES duction. This calls for pension systems that distort less decisions made on the labor mar- kets; that support labor mobility across occupations, sectors, and countries; that are con- ducive to financial market development; and that support lifelong learning. Against these diverse reform pressures, what are the promises of NDCs, and how are they put in perspective? But, before we get to these promises, very quickly: What is NDC? An NDC scheme is a pay-as-you-go scheme that by design mimics a financial defined con- tribution (FDC) scheme,3 and in doing so has many of the features of an FDC scheme. Indi- vidual contributions based on a fixed contribution rate are noted on individual accounts. Accounts are credited with a rate of return. Account values continue to accumulate as long as the participant works and pays contributions. There is no "full-benefit" age, but instead all new contributions enhance the size of the individual's annuity once he or she claims it, or after a minimum pension age. The annuity is calculated by dividing the account balance with an estimate of life expectancy at retirement. Compared with an FDC scheme, the two most important differences are, first, that the internal rate of return in an NDC scheme is a function of productivity growth, labor force growth, and factors linked with contribution and benefit payment streams as opposed to "the" financial market rate of return. Second, the only financial saving that can occur is in the form of a reserve--or buffer--fund, as opposed to the prefunded character of FDC. The NDC buffer fund is necessary to save the surplus paid by larger cohorts to be paid out when they become pensioners, but the fund also has other buffer functions depending on specific system design. Since the financial rate of return is generally expected to be higher than the rate of eco- nomic growth, NDC requires a higher level of contributions to achieve the same payout as an FDC scheme, and for this reason an NDC scheme is second best in economic terms. On the other hand, the appeal of NDC is that it does not involve the transition cost associated with introducing an FDC scheme where some form of pay-as-you-go arrangement already exists. NDC does not aspire to create financial saving, whereas FDC schemes may do this. Nevertheless, NDC can contribute indirectly to creating financial saving for retirement by clearly establishing the limit of the public pension commitment. NDC moves the focus to individual responsibility for providing for her- or himself in old age through work and accumulation of "funds" in individual accounts. NDC has two desirable design features as opposed to the counterfactual design of a defined benefit scheme. First, NDCs are fair in the sense that two persons from the same birth cohort who make the same contributions into the scheme in the same period can expect to receive the same pension rights for these. Second, a generic NDC is, in principle, financially stable. The system is geared to maintain long-run equivalence between assets and liabilities at an unchanged contribution rate. This is accomplished in part through the dependence of annuities on life expectancy, in part by the fact that the internal rate of return reflects growth of productivity--and hence real wages--but also changing demo- graphic conditions, the most important being declining fertility as propagated through the size of the labor force, and finally by designing the system to fulfill the long-term condi- tion that assets be at least as great as liabilities. As a result of these characteristics, a generic NDC promises to deal with the effects of the aging population more or less automatically and in the absence of political interven- tion. A generic NDC scheme maintains financial stability, with a given contribution rate, completely autonomously, based on legislated rules. It is accompanied by account infor- mation to participants, and a system financial statement can (should) be produced and made public annually. NDC provides no guarantee against political intervention, how- ever, although it makes the results of political intervention more transparent than a con- ventional pay-as-you-go defined benefit scheme does. In practice, however, the transition THE STATUS OF THE NDC DISCUSSION: INTRODUCTION AND OVERVIEW 5 from the previous regime to NDC may place some restrictions on the full-fledged imple- mentation of generic NDC. For example, there may be a "tax overhang" from the old sys- tem in the form of commitments politicians choose to honor, even though they might be inconsistent with the logic of the new NDC regime. There are features of NDC that promise to make it particularly attractive in terms of effects on the labor supply. First, NDC contributions can be perceived by the individual as premium payments into an insurance system--that is, NDC is like a career savings plan that is illiquid until retirement and then pays out in the form of an annuity. Second, NDC rules--together with tax rules that provide equal treatment of pensions and earnings--are neutral in this sense on the work­leisure decisions of older workers. Third, NDC accounts are "infinitely" devisable, which means that it is only practical considerations that deter- mine what percent of a full annuity an individual can claim at any time--after reaching the minimum pension age. Older workers can continue to work, but less, and can change their minds about how much, since an annuity can always be recalculated--and new con- tributions can be included in the calculation. Fourth, accounts can be shared between spouses or cohabiting couples. Fifth, NDC (like FDC), by basing benefits on individual accounts, makes labor mobility easy. In the NDC (and FDC) world it makes no difference for a pension in which occupation, branch, region, or country one works. Portability is possible by definition, and there are no benefit conditions that lock workers into specific employment situations. Finally, in economic environments where informality is prevalent, NDC rewards those who choose to participate in the formal economy. This being said, however, it is important to recall the importance of supplementing an NDC scheme with some form of "top-up," which in very poor countries with a high degree of informality may be a main benefit for some time. What NDC provides in this context is an explicit structure for integrating an emerging formal sector into an earnings-related mandatory pension scheme. In sum, NDC promises to address the issues confronting pension policy in the twenty- first century. How well it does this is the subject of this book. The Structure of the Book and an Overview of the Contributions The book is organized into two parts with two sections each. Part I presents the NDC con- cept and issues. Section 1 defines NDCs and focuses on conceptual questions. A set of broader conceptual, policy, and cross-country questions are the topics of section 2. Part II presents country experiences and prospects, with the chapters in section 3 reviewing the experience of countries that have actually introduced NDC schemes, while section 4 pre- sents chapters on selected countries where NDC is being considered as an option for reforming the current defined benefit (DB) system. Section 1 sets the stage for the book as it approaches the concept of NDC schemes from different angles as variations of a theme. The first chapter of this section defines generic NDC. It is intended to function as a "primer," providing a common ground and putting everybody on the same page. The second chapter addresses what NDCs can bring to pen- sion reform strategies, and the third chapter assesses NDCs from the perspective of wel- fare economics. Following this, there are three pieces that are written comments emanating from an ad hoc panel discussion organized at the conference in Sandhamn. It adds further light and also conceptual shades on the discussion about the nature of NDCs and their role in pension systems and reforms. Chapter 2, by Edward Palmer, sets the stage by defining a "generic NDC." He establishes and discusses a set of conditions characterizing a generic NDC scheme. As noted above, 6 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES NDC is like an individual saving scheme in which contributions on earnings are paid into an account, but NDC follows the principle of pay-as-you-go--that current contributions of workers finance the pensions of current retirees. An essential micro condition defining NDC is that the value of an individual account at any time is the present value of the par- ticipant's pension entitlement. The macro condition of NDC is that assets must be at least as great as liabilities, a condition that can be fulfilled in practice in a number of ways, which are discussed in the chapter. The chapter also discusses NDC in an intra- and inter- generational distributional context and examines the possibility of sharing accounts between spouses. Chapter 3, by Axel H. Börsch-Supan, presents first the basic ideas underlying NDC sys- tems and discusses their main advantages and disadvantages. He argues that an NDC sys- tem is mainly a political device. It makes parametric reform, badly needed to stabilize the pay-as-you-go (PAYG) pillars all over the world, easier because it exposes the trade-offs and clarifies concepts. It may also change the microeconomics of labor supply and savings. It does not, however, change the macroeconomics of PAYG systems and thus does not sub- stitute for the introduction of prefunded second and third pillars. NDC systems can be installed as individual account systems, as is done most prominently in Sweden. However, Börsch-Supan argues that they can also be mimicked by a set of rules in a conventional defined benefit PAYG system, showing that NDC systems are more a political than an eco- nomic device. The second part of his chapter claims that the German pension reform pro- posals made in late summer 2003 effectively introduced an NDC-like system without explicit NDC-type accounting. Chapter 4, by Nicholas Barr, offers an assessment of NDC pensions from the perspective of welfare economics, in terms of three questions: Is the particular feature an advantage? If so, is the advantage inherent in NDCs or could it be achieved by other arrangements? And is the advantage one of policy design or of political reality? The chapter argues that NDC pensions can facilitate desirable design features such as a flexible retirement age and auto- matic adjustment to rising life expectancy, enhance the ability to cope with risk and uncer- tainty, assist sustainability, enhance transparency, and reduce incentives to fraud. It also argues, however, that these advantages are not the sole property of NDCs, but could be achieved by other designs. The chapter goes on to discuss equivocal aspects of NDC pen- sions and their disadvantages. The final section offers three strategic conclusions: First, NDC pensions are not a theoretically dominant policy--they are a policy reform, not the policy reform; second, NDC schemes can take many guises--the case for NDCs is strongest as part of a wider pension system, rather than as a stand-alone design; third, NDC pensions do little to address the fundamental cause of unsustainability--the fact that earliest pensionable age is not related to life expectancy. The presence of Peter Diamond, Assar Lindbeck, and Salvador Valdés-Prieto at the confer- ence suggested the creation of a panel on the conceptualization and position of NDC among alternative pension systems. Their written contributions in chapter 5 emanate from this discussion and provide highly valuable insights into the conceptualization of pensions and the position of NDCs among alternative systems. All three start with the three-dimensional classification developed in Lindbeck and Persson (2003): (1) funded versus unfunded systems, (2) actuarial versus nonactuarial systems, and (3) DB (defined benefit) versus DC (defined contribution) systems. But they differ importantly on the interpretation of these characteristics and the position of NDCs in the spectrum of pen- sion system. Although all three give the NDC system pretty good overall grades, they see also virtues in alternative system designs and put up a warning flag about "over- selling" NDC. THE STATUS OF THE NDC DISCUSSION: INTRODUCTION AND OVERVIEW 7 Section 2 covers a broad spectrum of conceptual and policy questions related to reform needs and NDCs, including the role of demographics in pension systems, the internal rate of return in NDC, conversion from a PAYG-DB scheme to NDC, cross-country considera- tions such as the usefulness of NDC as a coordinated European pension system or the applicability of NDCs for middle- and low-income countries, and the politics of NDC pen- sion reforms. Chapter 6, the opening chapter of this section, by Juha M. Alho, Jukka Lassila, and Tarmo Valkonen approaches NDC by examining some of the demographic and economic "shocks" with which a generic NDC design would be confronted, but without applying all the tools of a truly generic NDC scheme. Hence, the chapter serves to illustrate the prob- lems confronting the pension system reformer. The chapter combines stochastic popula- tion simulations with economic models, providing us with a new way to think about the sustainability of public welfare systems. The chapter discusses the expected change in the age distribution of Europe and consequences for Europe's pension systems. In addition, using four country examples, the authors demonstrate how demographic risks can be shifted in time and between groups. Chapter 7 by Ole Settergren and Boguslaw D. Mikula presents a method for calculating the internal rate of return in PAYG pension systems in general. This generic method is most easily applicable in NDC schemes. They demonstrate that, in addition to the rate of growth of average covered earnings and the number of covered persons--as would be the case in a steady state--a third factor becomes important. This is the time structure of con- tributions and pension payments, which is labeled the turnover duration (TD)--the expected money-weighted average time from contributing to the system to receiving the pension. They show that at any given time assets can be valued as the product of the TD of contributions in the system and contributions actually paid in the period evaluated. Lia- bilities in the NDC scheme are the current value of all obligations. With the TD valuation of assets it is possible to estimate the (implicit) asset side of the balance sheet, making it possible to determine the ratio of assets in a given period over liabilities, that is, to deter- mine whether the pension system fulfills the condition of financial balance. Salvador Valdés-Prieto presents an alternative balancing mechanism to the Swedish auto- matic balancing mechanism (ABM) in chapter 8. When NDC design was adopted in the early 1990s it was known that, in addition to responding to the rates of growth of produc- tivity and the labor force, NDC liquidity would be affected by the timing of payments--by the age-earnings distribution of contributions and age distribution of payments. Yet it was not clear initially how this knowledge could be incorporated into the formulas. Swedish policy experts responded to this challenge by formulating how the internal rate of return in a PAYG system could be represented, and using this knowledge, created the ABM, which is used to secure financial balance. This chapter proposes an alternative to the ABM that also endows NDC plans with automatic financial stability in the short run. This method is labeled integration to financial markets (IFM). In contrast to the ABM, the IFM proposal is based on financial-market-determined discount rates and trade of uninsurable risks with market investors, that is, agents who are neither participants in the scheme nor taxpayers. Relative advantages and risks of the IFM and the ABM are discussed. Chapter 9 by Edward Palmer reviews the issues arising in converting a PAYG-DB scheme to an NDC scheme and presents models for conversion. The first one is based on the NDC formula and the second on the acquired DB rights. Adhering to the NDC formula requires (1) establishing how to approximate past earnings if there is not complete information; (2) what contribution rate should be set for past earnings if it is not to be the rate looking for- ward after the implementation of NDC; and (3) what the imputed rate of return on past 8 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES contributions should be. The chapter also discusses how countries introducing NDC have dealt with acquired rights (including special privileges where they have existed), and how other components of DB schemes for low-income recipients' noncontributory rights and disability and survivor benefits are handled in conversion to NDC. Speed of transition and the consequences of complete and immediate transition (as in Latvia) contrasting with slow transition focusing largely (Italy) or solely (the Kyrgyz Republic) on new entrants are discussed. Slow transitions protect acquired rights in the old system, but are counterpro- ductive in terms of the policy goals of the new NDC system. The balance of pros and cons, it is argued, favors rapid transition. Florence Legros discusses similarities and differences between NDC and the French and German point systems in chapter 10. She focuses on how these systems differ when there is an external shock (demographic, economic, and so on) and discusses the possible conse- quences of moving from the point system to NDC. The French point system--with no automatism in its indexing device--can be regulated each year according to the forecasts. Legros concludes that this may be the best way to react to changes in the economic and demographic environment--but to do this also requires, first, reliable and frequent fore- casts, and second, a total independence of the governing board of the scheme from the retiree and worker lobbies. She notes that although the second point can be handled with rules, as opposed to the present state of affairs, it is questionable whether the first point can be fulfilled. Germany has adopted a method to correct the excessive generosity of the scheme with what she calls a "return spring"--a mechanism in which the pension yield is lowered in relation to a desired contribution rate. This mechanism is reinforced by the "Rürup sustainability factor," which explicitly introduces the dependency ratio and accounts for life expectancy changes. She notes that with this strategy it is probable that the German scheme will move into surplus within some years, allowing for a reserve that might be needed for intergenerational transfers. She concludes by asking the question Why introduce NDC? Her answer is that FDC schemes promote individual responsibility while NDC maintains the principle of social cohesion in public pension schemes. In chapter 11, Robert Holzmann (1) reviews the reform needs of pension systems in the member countries of the European Union for fiscal, economic, and social reasons; (2) makes the case for a move toward a more coordinated pension system in Europe; and (3) sketches what such a system may look like and how it may come about. The central claim and content of the chapter is that a multipillar system, with an NDC system at its core and coordinated supplementary funded pensions and social pensions at its wings, is an ideal system to deal with diverse fiscal and social reform needs. The approach would permit introducing a harmonized structure while allowing for country-specific preferences about coverage and contribution rate. Although the chapter focuses on the European Union, the arguments and proposed structure for a more harmonized system with some autonomy for subnational entities are also valid for other countries or integration areas such as China or Mercosur. Chapter 12, by David Lindeman, David Robalino, and Michal Rutkowski, discusses the suit- ability of NDC for middle- and low-income countries. Much of the policy discussion to date so far has been focused on high-income countries such as Sweden and Italy. The authors' point of departure is the observation that middle- and low-income countries are characterized by less stable macroeconomic environments, a narrow and highly sensitive contributory base, a more uneven income distribution than in developed countries, and, in addition, they are often in the midst of a demographic transition. The chapter focuses on the general macroeconomic, demographic, and institutional variables that can affect the application and performance of the NDC concept. The chapter discusses the primary ben- THE STATUS OF THE NDC DISCUSSION: INTRODUCTION AND OVERVIEW 9 efits from adopting the NDC concept in middle- and low-income countries in terms of reducing economic distortions and improving financial sustainability of PAYG pension schemes. The chapter identifies initial conditions and design factors that are likely to influ- ence the realization of these benefits. It also addresses implementation issues, redistribu- tion and poverty, institutional capacity requirements, and the political economy of NDC schemes in the middle- and low-income setting. The point of departure of chapter 13 by Annika Sundén is that the shift to defined contri- bution schemes increases the need for general financial literacy. Both NDC and FDC schemes put more of the responsibility and need to plan for retirement on the individual participant. Studies of what people know indicate that they have poor knowledge about the conditions that determine their benefits, and that learning about pensions is difficult. This chapter discusses what people need to know and what surveys show they know, using data gathered for the new Swedish NDC scheme. The chapter concludes that information and education leading to participant understanding of the scheme and, generally, improved pension literacy is important, and that pension system design that makes understanding easy is important. The latter provides an argument against complicated rules, for example the rules for conversion to NDC that affect different cohorts or persons in the same cohort differently, and a need to create an understanding of the basic logic of NDC. Chapter 14 by Sarah M. Brooks and R. Kent Weaver focuses on the political aspects of NDC. It begins by listing the positive policy arguments for NDC compared with the coun- terfactual policies of PAYG-DB and FDC schemes. The authors argue that the impression of "property rights" that NDC conveys has the negative feature of creating a "lock-in" effect. In the words of the authors, proponents argue that the political advantage of NDC is that it "absolve[s] politicians of responsibility for potential future benefit reduction," for example due to an aging population. Put differently, in adopting NDC, governments "lash themselves to the mast." The chapter addresses two questions: First, is NDC reform more likely to get onto the reform agenda in countries with certain characteristics? Second, does NDC really have the potential to absolve politicians of their responsibility, or will NDC schemes eventually spark resistance that undercuts their intended effects? Section 3 contains a set of chapters that review the experience of countries that have actually introduced NDC schemes; more specifically these are countries that have reformed traditional non-financial--that is, PAYG--defined benefit (NDB) systems into an NDC system. The country set includes all the reform countries of a fully fledged NDC sys- tem: Italy, Latvia, Poland, and Sweden. Italy is even represented with two chapters. Edward Palmer, Sandra Stabina, Ingemar Svensson, and Inta Vanovska present and analyze , the Latvian NDC reform, which was implemented on January 1, 1996, in chapter 15. This was the first, and to date is the only, NDC reform covering all workers from the first year of implementation. The most difficult design issue was the conversion of rights acquired in the old scheme into NDC rights, because there were only historic service records and because the transition occurred during a period of general economic upheaval in the early years of transition. The pros and cons of the method chosen are discussed. The chapter also pre- sents financial calculations for the coming three-quarters of a century under a variety of extreme economic and demographic assumptions. Initially, the NDC contribution rate was 20 percent, but the overall rate of 20 percent is being successively redistributed to an FDC scheme with a rate of 10 percent by 2010. The chapter illustrates the impact of moving into the financial account scheme and reducing the NDC rate to 10 percent. The introduction of the FDC scheme requires a tax to finance previous NDC commitments made with NDC contribution rates above 10 percent. Alternative estimates are provided for a less ambi- tious financial account pillar with a contribution rate of 6 percent rather than the sched- 10 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES uled 10 percent (leaving 14 percent for the NDC scheme). The chapter also shows that the NDC scheme per se is financially robust under a large variety of assumptions, even with the pessimistic assumption that the working-age population will decline by 50 percent in the coming half decade. The authors stress that the system design needs to be completed by formally implementing the reserve fund and producing more comprehensive financial accounts, including the tax-financed debt to the NDC scheme accompanying the introduc- tion of the FDC scheme. Chapter 16 by Marek Góra and Agnieszka Chlo´n-Domi´nczak presents the new Polish pen- sion system, which was introduced under the banner of "Security through Diversity," on January 1, 1999. It automatically replaced all previous legislation on old-age pensions for the majority of the working population as for all persons under age 50 in the year the reform the new scheme was introduced. The chapter presents the design of the Polish NDC scheme in the context of the framework of the entire pension system implemented in 1999, and it assesses the early experiences with implementation and discusses the future of the NDC pension system in Poland. The chapter also presents financial calculations over the coming 50 years, showing that the system moves into balance from a point of depar- ture of substantial imbalance--and this is helped along by a gradual formalization of the economy that for a couple of decades counteracts a gradual fall in the working-age popu- lation. Though the general design of the Polish NDC scheme has remained unchanged, some elements of its original construction have become a subject of debate. One of the most important is the decision in the original legislation to allow women to retire at age 60, while men retire at age 65. In addition, the reform left farmers outside, which is likely to create a coverage problem for this segment of the population as they integrate into the for- mal economy. A major lesson to be drawn from the Polish experience is that the imple- mentation needs good administrative support if major problems are to be avoided. Chapter 17 by Bo Könberg, Edward Palmer, and Annika Sundén presents the Swedish NDC experience. The Swedish pension reform dates back to the summer of 1992 when the Par- liamentary Working Group on Pensions published a "sketch" containing almost all the essential elements of the reform. In June of 1994, the Swedish parliament passed the reform legislation that was to be the Swedish NDC scheme. Implementation began in 1995 and moved forward in stages. This chapter discusses how the Swedish reform came about, what it is, its public image, and remaining issues. A driving force behind the creation of the Swedish NDC scheme was the goal of creating financial stability while using contributions per capita as the rate of return on accounts. This led to the creation of the automatic bal- ancing mechanism (ABM). The authors discuss the fact that one of the remaining chal- lenges is to create a popular understanding that the ABM is simply a component of the rate of return on NDC accounts. The authors highlight the idea that the potential risk for the autonomy of the system is the very mechanism that was designed to create autonomous financial balance--the ABM. They conclude by reflecting that Sweden had already accumulated large reserves in the old system to meet the demographic burden of the post-war baby boomers, which will help to provide payments when both the boomer and next boomer generation--the children of the postwar boomers--retire. This puts Swe- den in a very favorable position compared with many countries considering introducing NDC. Calculations show that demography will stress finances in the 2040s, and this will be taken into consideration in solving the remaining problem of "balancing"--a method of determining when reserves are large enough to allow them to be distributed to the then- living participants. Finally, the authors conclude that what remain to be examined in greater depth are the implications of the construction of the balancing mechanism for the intergenerational distribution of resources. THE STATUS OF THE NDC DISCUSSION: INTRODUCTION AND OVERVIEW 11 The first case study of Italy is presented by Daniele Franco and Nicola Sartor in chapter 18. Since 1992, reform of the pension system has been at the core of the effort to consolidate Italian public finances. The reform process began in 1992 when a quarter of prospective public sector pension liabilities was abruptly cancelled. A second major reform in 1995 introduced NDC, but beginning with new entrants and with a pro rata formula for per- sons with fewer than 18 years of coverage. The authors provide and discuss empirical evi- dence on the results of the reform and conclude that the reform process is not yet finished. According to their analysis, further adjustments are required to: (1) accelerate the intro- duction of notional accounts; (2) avoid increases in the ratio of pension expenditure to GDP; and (3) strengthen the self-equilibrating mechanisms of the new pension regime, such as by moving to an annual revaluation of the life expectancy factor used in calculat- ing newly granted NDC annuities. In the view of these authors, present problems derive from the fact that the 1995 reform was implemented without an extensive analysis and an open public debate. The lengthy and incremental process associated with the "unfinished reform"--following the first stage in 1992--has generated uncertainty, and, according to the authors, may limit the microeconomic benefits of the approach embodied in NDC. New reform proposals discussed after 1995 do not explicitly call into question the 1995 reform, but, generally, they have not been consistent with its underlying philosophy. The authors conclude that the future of NDC in Italy is therefore uncertain. The second case study on Italy by Sandro Gronchi and Sergio Nisticň in chapter 19 com- pares the Italian and Swedish reforms. The point of departure is a theoretical framework that highlights the implications of the different ways in which Italy and Sweden imple- mented the NDC model. In particular, this chapter focuses in depth on the real growth norm factored into the life annuities in the Italian and Swedish NDC schemes--a feature no other country has implemented. The chapter is unique in providing an analysis of this feature. Like the Swedish annuity, the Italian annuity contains an economic growth fac- tor--in Italy's case, a real growth factor of 1.5 percent per annum over the life of the annu- ity (in Sweden it is 1.6 percent). The main conceptual defect of the "Italian-style" NDC scheme is that the indexation of pensions should reflect the difference between the sus- tainable rate of return (with GDP growth as a proxy) and the value of 1.5 percent used in calculating annuities, which is the case in Sweden. The absence of this correction jeopar- dizes the objectives of sustainability and fairness. Other shortcomings and inconsistencies of the Italian reform are discussed. The authors conclude that, although blueprints have been available since 1995, successive governments have not taken the necessary step to form a committee of experts with a mandate to "bridge the gap between the theoretical foundations of the NDC scheme and the countless details that implementation would inevitably bring." Section 4, the last section of the book, presents case studies for countries that are think- ing about introducing NDC systems, or, at least according to the authors, should have a close look at NDC approach as a promising option for reforming the existing PAYG sys- tem. As to be expected, some country considerations are more advanced than others. Five country studies are presented: Austria, the Czech Republic, Germany, Spain, and Japan. The characteristic of all country cases is the size of the inherited Bismarckian-type retire- ment system, which ranges between 9 percent (Japan) and 15 percent (Austria) of GDP; the piecemeal approach of past parametric reforms, which has failed so far to put the public pension schemes on a sustainable basis; and implications of introducing NDCs for benefit levels for a given retirement age. Austria competes with Italy, Poland, and Uruguay for the "world championship" in public pension expenditure. All countries have expenditures close to 15 percent of GDP, 12 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES but some contenders start to fall back in the race as they have reformed their system (such as Poland). So far Austria is leading the race. Recent reforms have helped to reduce the sustainability problem but will not solve it, as is discussed in chapter 20 by Bernhard Felderer, Reinhard Koman, and Ulrich Schuh. This study demonstrates the impact of different scenarios on expenditures, revenues, and replacement rates through 2050. They present new evidence that confirms the need for a more fundamental reform of the Austrian pen- sion system. In a second part, the chapter investigates the potential effects of moving from the current NDB to an NDC system. The results indicate that an NDC reform can put the public scheme on sustainable basis, but the arithmetic of pensions that the NDC approach forces to light exhibits the impact on replacement rates: For men, replacement rates would be reduced from some 80 percent (if unchanged but unsustainable) by some 20 and more percentage points (depending on the labor force scenario); for women, the replacement rates would be reduced from some 50 percent by similar dimensions. Based on these sobering calculations, the authors recommend introducing an additional fully funded, mandatory second pillar to a reformed NDC scheme to provide acceptable benefit levels for future retirees. The Czech Republic shares many features with its southern neighbor, including the his- torical origins of the pension system. According to UN population projections,4 population aging is projected to be even more pronounced than in previous projections and the Czech Republic--together with Greece, Italy, and Spain--is expected to have the highest share in the population of people 65 and older by 2050. The projected pronounced population aging in the coming decades makes a reform of the Czech pension system even more important to avoid a future financial crisis. Although various reforms have been under- taken in the course of 1990s, they are not sufficient to create financial stability, making more significant changes even more necessary. The authors of this case study (chapter 21)--Agnieszka Chlo´n-Domi´nczak and Marek Mora--review the reform options and conclude that a move toward a mandatory funded pillar does not seem a viable option, as public finances would not be able to bear the costs of transition financing. Also, following rela- tively poor performance of the voluntary pension funds, there is little support among the politicians and the society generally for introducing a funded pension system. On the other hand there is increasing political resistance to another round of parametric reforms. Against this background the introduction of an NDC scheme provides a framework that joins the microeconomic incentives of the funded system with PAYG financing that does not imply transition costs. The chapter analyzes NDC as a viable option for reforming the Czech pension system, with particular focus on advantages and disadvantages of such an option from the perspective of the Czech Republic. The financial simulations that comple- ment the analysis show that a mere move from the current two-tier PAYG system to an NDC scheme does not lead to financial sustainability. Such a move would have to go hand in hand with other changes, such as an increase (and equalization) of retirement ages for men and women and lower pension indexation. Germany still has a very generous public PAYG pension system. It is characterized by early effective retirement ages and very high effective replacement rates. Most workers receive virtually all of their retirement income from this public retirement insurance. Costs are almost 12 percent of GDP, more than 2.5 times as much as the U.S. social security sys- tem). The pressures exerted by population aging on this monolithic system, amplified by negative incentive effects, have induced a reform process that began in 1992 and is still ongoing. This process is the topic of chapter 22 by Axel H. Börsch-Supan and Christina B. Wilke, and they proceed in two parts. The first part describes the German pension system as it has shaped the labor market until about the year 2000. The second part describes the THE STATUS OF THE NDC DISCUSSION: INTRODUCTION AND OVERVIEW 13 three-staged reform process that will convert the exemplary and monolithic Bismarckian public insurance system after the year 2000 into a complex multipillar system in which the PAYG pillar will emulate an NDC system. The chapter assesses how far these reform steps will solve the pressing problems of a prototypical PAYG system of old age provision, it is hoped with lessons for other countries with similar problems. Spain is another European country with pronounced projected population aging as a result--as elsewhere--of increasing life expectancy, but mixed with one of the lowest fer- tility rates in the world. Chapter 23 by Carlos Vidal-Meliá and Inmaculada Domínguez Fabián has two goals: first, to provide an objective technical assessment of the current situation of the contributory pension system in Spain and its prospects for the future, and second, to look at the issues surrounding the introduction of a system of notional defined contribu- tion accounts. To this end the chapter sketches the basic elements upon which the current system is based, and highlights its main indicators before discussing some of its funda- mental problems. These problems include equity issues resulting from highly unequal implicit rates of return in the different pension schemes organized according to occupa- tional affiliation. The review of the set of forecasts of Spanish researchers and institutions all confirm the financial unsustainability of an unreformed system. When investigating the usefulness and applicability of the NDC approach for Spain (and its largest subsystem), the chapter has two interesting twists. It compares the replacement rates (and internal rates of return) of the current scheme with that of alternative NDC implementations using Spanish data. The reference benefit formulas are reforms in Sweden, Italy, Brazil, Poland, and Latvia. The result shows variations across the different benefit benchmarks, but across all benchmarks are clear signals that the current Spanish benefit formula is totally unsus- tainable because it is much too generous. As a second twist, the chapter takes an alterna- tive realization of potential notional interest rate options for contributions and pensions and suggests which formula or group of formulas would best fit the profile of contributor- beneficiary risk and what the transition process would be. Japan already has the oldest population in the world. Since 2001, the income statement of the principal pension program has turned into a deficit. Increasing drop out has been observed from the basic old-age income protection for nonsalaried workers. The generous pension program has to be reformed to be much more incentive-compatible. In this little note (chapter 24), Noriyuki Takayama claims that a reform of the Japanese PAYG system toward a Swedish-type NDC system would very much help Japan overcome a number of problems in its current schemes, including problems of incentive-compatibility. Japan faces the additional specific challenge of handling huge excess liabilities resulting from entitlements of past pension contributions, equivalent to 130 percent of GDP in 2004. If Japan were to move to an NDC system, the chapter suggests handling this legacy problem of excess liability through a budget mechanism such as an earmarked consumption tax or transfers from the general budget. Summing Up and Research Outlook As is evident from the summary of the contents of this book, the NDC conference held on the island of Sandhamn outside Stockholm resulted in a very rich conference producing important contributions on the topic of NDC. The contents of this book should be helpful for countries that want to investigate this approach to reform an NDB pillar that is not working, to readjust weights between unfunded and funded pillars, or to introduce an NDC scheme as the foundation, perhaps with a mandatory public and/or a private FDC scheme on top. NDC is worthy of a country's examination as a means of dealing with the 14 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES financial instability accompanying the aging of the population. It provides a framework for dealing with most of the demands on a modern pension system. In some environ- ments, it may be the best alternative where conditions are not ripe for the introduction of a financial account scheme. The rich conference outcome provides some answers but, as should be expected, also poses many more questions that require the attention of policy makers and researchers. In addition, the relatively short experience of the little more than a handful of countries that can claim to follow an NDC-type approach is simply too short to draw strong conclusions. Nevertheless, there was definitely a clear agreement among the diverse group of researchers and policy makers at the Sandhamn conference that all countries are invited to have a close look at the NDC approach as a viable option for dealing with the diverse reform pressures of their pension schemes. Against this background, we make no attempt to provide a detailed concluding sum- mary. This said, there are three main themes that run across all the chapters. These lead to some conclusions, while at the same time they suggest questions for future research. First, the conference outcome suggests that the conceptual promises of an NDC reform, in particular that of financial sustainability, can be fulfilled. Yet, as with other reform approaches, the challenge is in its implementation. More specifically, the experience of NDC reform countries suggests that NDC is not a foolproof approach, and it needs to be politically well managed. As in other reforms, communication with the stakeholders of the reform to get their buy in is crucial, and administrative preparedness is also crucial. The limited country experience also suggests that faster transition from the inherited DB sys- tem may have major advantages. Although the concept of the balancing mechanism intended to achieve financial sustainability seems to have been understood and embraced by experts, it may be difficult to implement, both for technical (data quality) and political reasons. In the one country where the automatic balancing mechanism has been imple- mented--Sweden--there is still no experience of the possible political repercussions of automatic balancing. But, perhaps most importantly, too little thinking has been given to the issue of "tax overhang," that is, the need for additional fiscal resources or reform mea- sures to deal with transitory deficit resulting from past commitments that are not covered by the fixed NDC contribution rate. Second, the short time period since the introduction of NDC (the longest period is nine years in Latvia) and interaction with other reform measures has not allowed for careful assessment of the economic effects of NDC reforms, and the distributive effects have not figured prominently in the political and academic discussions. As a result, too little is known, especially about the intergenerational distributive effects, and very little attention has been devoted to dealing with the spouse pension. We still have very little empirical evidence about the promised effects on labor supply, in particular the retirement decision, although the limited evidence from reform countries looks promising. As a result we do not know if individuals faced with a lower pension at early retirement age do sufficiently postpone their retirement decision or draw on a much lower pension, and how much this is influenced by overall system design issues (such as the minimum pension guarantee) or constraints imposed from outside the system. Last but not least, there was a shared feeling at the conference that NDC schemes must be assessed as part of a broader (multipillar) pension concept and not (only) viewed in iso- lation. An issue that needs to be given more attention is the question of how to integrate occupational schemes into the overall pension framework. This may be relatively easy for, say, pensions of civil servants, but it is likely to require more thinking for, say, farmers or persons involved in the informal or semiformal agricultural setting in developing THE STATUS OF THE NDC DISCUSSION: INTRODUCTION AND OVERVIEW 15 economies. As NDC systems apply the logic that benefits depend only on one's own con- tributions, and with life expectancy determining the size of the life benefit, for any given contribution rate and retirement age the benefit level will be lower--perhaps markedly lower--than under the old NDB scheme. Individuals may compensate with later retire- ment but they may also compensate with more individual savings--but in what form (mandated, voluntary), with what instruments (financial assets, annuities), under what form of supervision, and so on? Hence, opting for an NDC scheme also requires having a game plan for the (second or third) supplementary pillars, and there are many issues involved. What's more, as the NDC scheme has no distributive aspirations, the choice of overall framework also requires paying attention to the needs of basic income support to the needy elderly in a way that achieves the social objective while not unduly compromis- ing the expected behavioral effects of NDC. That we can pose these and other research questions at all, or at least much more clearly, is the result of the conference and of the chapters in the book. Enjoy the reading! Notes 1. See Holzmann et al. (2005). 2. "Non-financial defined contribution" and "notional defined contribution" are used interchangeably in this book and should be understood to have the same definition. 3. Note that we adopt the terminology presented in Góra and Palmer (2004) that distin- guished between defined benefit and defined contribution schemes on the one hand, and financial and non-financial schemes on the other. A financial scheme is a scheme the assets of which are held in financial market instruments, and a non-financial scheme is a scheme in which the assets are the fewer contributions of workers. This gives four distinct categories of pension schemes: Financial defined contribution (FDC), financial defined benefit (FDB), non-financial defined contribution (NDC) and non-financial defined benefit (NDB). FDC is what is called individual account schemes in some of the pension literature. Since NDC is also an individual account scheme, it seems more appropriate to distinguish between the financial and non-financial aspects of these in choosing the nomenclature. 4. See UN (1998). References Góra, M., and E. Palmer. 2004. "Shifting Perspectives in Pensions." IZA Discussion Paper 1369, The Institute for the Study of Labor, Bonn. Holzmann, R., R. Hinz, and World Bank team. 2005. Old Age Income Support in the 21st Cen- tury: An International Perspective on Pension Systems and Reform. Washington, DC: World Bank. Lindbeck, A., and M. Persson. 2003. "The Gains from Pension Reform." The Journal of Eco- nomic Literature 41(1): 74­112. UN (United Nations). 1998. World Population Prospect: The 1998 Revision. New York: UN Population Division. PART I THE CONCEPT OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEM-- VARIATIONS ON A THEME 15 Chapter 2 What Is NDC? Edward Palmer* IT IS ALWAYS A PRECARIOUS BUSINESS TO TRACE THE ORIGIN OF AN IDEA, but among the pub- lished works available to an international audience, Buchanan (1968) probably contains the first proposal that can be likened to what have emerged as notional--or non-financial defined--defined contribution (NDC) pension schemes.1 A proposal in Boskin, Kotlikoff, and Shoven (1988) is also on this track, and the point systems of Germany and France cer- tainly contain the embryo of an NDC system. Nevertheless, NDC as it is now known was not given a face until the arduous task of redesigning pension systems was taken on in a series of reforms in Europe in the 1990s. It is probably safe to say that NDC has emerged as a result of the interchange of ideas and discussions among pension experts and politicians working on pension reform issues in those countries where NDC has been implemented. Since the mid-1990s, the exchange of ideas about the possible contribution of NDC to the pension policy toolbox has been truly international, with both avid proponents and a healthy number of skeptics. To date, however, ten years after NDC legislation was passed in two countries--Sweden and Italy--no paper has attempted to bring all the strands together to provide a cohesive tech- nical framework. That is the ambition of this study. In the general discussion in the next section, this study touches briefly on the differ- ences in principle between NDC and alternatives for mandatory national pension schemes. The purpose of the study, however, is not to argue the comparative merits of alternative pension schemes, or to discuss what countries implementing NDC have done in practice. Instead, the aim of this study is to present what an NDC pension system is. The second section of the study briefly discusses efficiency and distribution with respect to NDC schemes. In the third section the generic NDC scheme is presented. The fourth section deals briefly with considerations in integrating social policy into the NDC framework and touches upon some other issues that can arise in constructing an NDC scheme. The study closes with a brief summary and some final remarks. * Edward Palmer is professor of social insurance economics at Uppsala University and head of the Division for Research at the Swedish Social Insurance Agency. The author is grateful for comments on this chapter generously provided by Robert Holzmann. 17 18 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES A General Overview of NDC A First Description of NDC An NDC scheme is a defined contribution, pay-as-you-go (PAYG) pension scheme. Contri- butions are defined in terms of a fixed contribution rate on individual earnings. These con- tributions are noted on an individual account. As opposed to a financial defined contribution (FDC) scheme, the contributions of participants noted on individual accounts are not "funded." More specifically, individual account money is not invested in financial market instruments. Compared with an FDC scheme, where individual account money is invested in financial market assets, by definition, the pay-as-you-go individual account DC scheme is a notional DC scheme. It can be argued that it is more precise to use the terminology "financial" and "non- financial"--that is, financial DC and non-financial DC--to distinguish these two types of defined contribution schemes, stressing the fact that money in accounts is invested in mar- ket assets in the FDC scheme, but not in the NDC scheme. Note that in the pension litera- ture financial DC schemes are generally referred to as "individual account schemes." Given that an NDC scheme is also an individual account scheme, the use of the term "indi- vidual account scheme" to describe financial account schemes seems ambiguous. Góra and Palmer (2004) argue that financial individual account schemes should be called FDC schemes, and notional--or non-financial--account schemes should be called NDC schemes, which is the terminology used throughout in this study. Before describing NDC in greater detail, two brief statements about the difference between NDC and FDC individual account schemes are helpful. First, they differ with respect to their (potential) contribution to national saving. An NDC scheme provides no direct contribution to saving, except through the possible mechanism of demographic sav- ing. An FDC scheme contributes to national saving during the build-up phase to maturity, but thereafter the potential effect on national saving will depend on the demography of the scheme participants. Furthermore, it is a well-discussed caveat in the economic litera- ture on pensions that the net overall effect of mandatory pension schemes on national sav- ing depends on private and government offsetting behavioral responses. Second, NDC and FDC differ with respect to the system rate of return. Participants in an FDC scheme earn a financial market rate of return, whereas participants in an NDC scheme earn an internal rate of return, which is determined by factors underlying the development of the economy. The next section discusses potential economic consequences of the differences between NDC and FDC schemes in greater depth. Now, what is an NDC scheme? A general description follows. In the next section, more precise criteria for a generic NDC scheme are developed. Basically, in an NDC scheme, participants--or employers on their behalf--pay contributions on earnings during their whole working career. Although there is a minimum age at which an annuity can be claimed, there is no "pension age." As long as people continue to receive earnings from work, these earnings generate contributions to individual accounts. This is true even if the individual is already drawing on a partial or full annuity after the minimum age to claim the annuity. Once again, each period's contributions are noted on an individual account. This notional individual account grows with new contributions and is credited with a rate of return. In the generic NDC scheme, the rate of return is the internal rate of return. The NDC benefit is a life annuity. It can be claimed at any time from the minimum retirement age. The generic NDC annuity embodies a rate of return based on the same internal rate of return that is credited accounts during the accumulation phase and, impor- tantly, cohort life expectancy at the time the annuity is claimed. Since newly granted annu- ities reflect life expectancy, in principle, NDC is an actuarially fair pension system. WHAT IS NDC? 19 Thus an NDC scheme distributes individual resources over the life cycle, but within the framework of a countrywide (universal) PAYG insurance scheme. In this sense it can be likened to an illiquid, individual cash balance scheme, with the important difference that it is an insurance scheme: that is, it redistributes the capital of the deceased to the survivors in the scheme. It fulfills the function of insuring against the individual risk of outliving the average participant. Because the NDC benefit is based on individual contributions from individual earnings, it may not provide sufficient coverage in old age for everyone because, for one reason or another, some individuals will not have had sufficient earnings before retirement to pro- vide a sufficient benefit in old age. This is not a problem inherent to NDC; rather, it is a characteristic of any earnings-related benefit. As opposed to many possible defined bene- fit (DB) formulas, the pension formula in an NDC scheme contains no built-in redistribu- tion of the system's revenues. Thus an NDC scheme must be supplemented with some form of low-income support scheme, in the form of minimum income or minimum benefit guarantee. NDC accounts can also be supplemented with noncontributory rights, e.g., for childcare years. What is essential is that these be financed with revenues exogenous to the NDC scheme, i.e., general tax revenue. Efficiency and Distribution Both NDC and defined benefit pay-as-you-go--that is, non-financial defined benefit-- schemes differ from financial schemes in terms of economic efficiency. The relevant dis- count rate for money set aside today to pay for consumption tomorrow is the financial market rate of return. In terms of this metric, a pension scheme that gives a risk-adjusted rate of return below the risk-adjusted financial market rate creates a tax wedge, and this means participants would attain a higher level of lifetime consumption with the counter- factual financial market rate of return. Rated in terms of efficiency, non-financial pension schemes, among these NDC schemes, are only potentially at least as efficient as financial pension schemes. They would be as efficient as financial schemes in a steady-state (closed) Golden Rule econ- omy2 but otherwise not necessarily so. If the rate of return on assets, r, traded in the financial market dominates the growth rate of the economy, + g, where is population (labor force) growth and g is productivity growth or, more exactly, the internal rate of return (IRR) of the NDC scheme (see the next section),3 then financial schemes domi- nate non-financial schemes in terms of efficiency. Nevertheless, although empirical evi- dence from the past century suggests this is the case, at least if the account portfolio includes equities,4 it need not always be the case in specific country settings and spe- cific periods. Although r > IRR means, ceteris paribus, that an FDC scheme dominates an NDC scheme, by linking individual benefits solely to individual contributions, an NDC scheme does not have the tax effect inherent in a DB-PAYG scheme. However, in the final analysis one should take into account the effects of the tax regime.5 For example, a country's taxes on earnings and capital may differ. In addition, although by definition the IRR of an NDC scheme depends on the develop- ment of the country's economy, in a comparison with an FDC counterfactual, it is impor- tant that two other considerations are kept in mind. First, an FDC scheme with a portfolio dominated or exclusively invested in home country debt is exposed to home country risk, just as NDC is. Second, an FDC scheme counterfactual, also holding only or mainly gov- ernment bonds, can be likened to an NDC scheme fully monetized with Buchanan bonds. However, if the government bond rate is greater than the IRR, this "advantage" will be paid for through taxes levied on the same countrywide insurance collective, that is, the 20 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES recipient of the higher return; although there are likely to be tax-distributional effects that distinguish these two approaches. The expected value of an NDC pension in any period t is exactly the amount in an indi- vidual account at that time. This amount, in turn, has, been determined by the individual's contributions and the rate of return on his or her account. This is just another way of say- ing that in a generic NDC scheme there is no built-in mechanism--either explicit or implicit--to redistribute income, as opposed to a DB scheme. A DB scheme involves at least some within-scheme redistribution, by definition. Furthermore, an NDC scheme differs from defined benefit pay-as-you-go (DB-PAYG) schemes in how it deals with the effects of changing demographic and economic factors. In an NDC scheme, the effects of economic and demographic fluctuations are accommodated endogenously, as the NDC internal rate of return adjusts account values during both the accumulation and the payout phases, and as annuities adjust to changes in birth cohort life expectancy at retirement. In a DB-PAYG scheme, benefits are fixed by "contract"--at least in principle--and sys- tem adjustments to exogenous economic and demographic shocks are accommodated through changes in the contribution rate. Obviously, this approach to dealing with exoge- nous economic and demographic risks can generate very different distributional outcomes from NDC, which adapts by adjusting accounts. Because an NDC scheme has a constant contribution rate over time, successive genera- tions of participants can expect to pay the same fixed percentage of their earnings into the scheme. If this property is chosen as a definition of intergenerational fairness, then an NDC scheme is intergenerationally fair. Each generation will pay a fixed percent of life- time earnings into the scheme and can expect to receive a stream of benefits, determined by these individual payments and the system rate of return. NDC schemes can nevertheless give varying outcomes across cohorts. This is because the value of individual benefits depends on the intertemporal rate of return, which will vary from period to period. This is true of any pension scheme, however. Individual outcomes are a function of either the market rate of return (financial schemes) or some form of index- ation (non-financial schemes), and the distribution of individual outcomes over time will be a function of either the development of the financial market or the economy, respec- tively, with the exact outcome depending on how the system rules link the scheme to either of these. Variation in the NDC intertemporal rate of return means that outcomes for given amounts of real valued contributions can vary over cohorts, and even over persons in the same birth cohort, depending on when they enter and exit the labor force. A generic NDC scheme distributes the resources of the scheme through the time dimen- sion of the IRR, to be defined more precisely in the next section. In addition, it is important to note that, to date, all country designs of NDC have included a transfer mechanism through the actual annuity formula employed, through the application of unisex life expectancy. Furthermore, taxes and transfers reenter the picture indirectly through the external (to the NDC scheme) tax and transfer system, since the government can transfer general tax revenues to the accounts of the NDC participants. It follows then that what the NDC scheme does in practice is to provide a setup for transparent accounting of the flows of money into the overall system and the sources of finance for these flows. NDC does not eliminate the political risk in this sense, but it pro- vides a transparent framework for weighing the pros and cons of political decisions on taxes and transfers regarding a nation's pension system. In sum, an NDC scheme is dominated by or is at least as good as an FDC scheme on the basis of efficiency. Whether a particular FDC counterfactual dominates a generic NDC WHAT IS NDC? 21 scheme in practice will depend on the investment portfolio of the FDC scheme and the tax- regime backdrop. In addition, dominance is not only dependent on these institutional con- siderations, it is also time-dependent, as the relation between the financial market rate of return and the IRR is dependent on the home economy--and indirectly, on the interna- tional economy, and the domestic and international financial markets. By definition, pay- as-you-go DB schemes include intra- and intergenerational tax distortions through internal redistribution not embedded in a generic NDC scheme--but NDC is not necessar- ily free of distributional effects, either. Also, by maintaining a constant contribution rate, an NDC scheme means all generations pay the same percent of earnings into the scheme. The value of these contributions depends, however, on the economy-determined rate of return, but this is also true of DB-PAYG schemes. Finally, for the same intragenerational distributional goals, an NDC scheme also dominates defined benefit schemes by making any form of distribution and the financial sources for this transparent. Generic NDC Design The central component of the NDC scheme is the individual lifetime account. This section describes how account values arise and how annuities are calculated. It begins with the concept of notional capital. Note that the entire discussion is in terms of real values. In practice, however, the system is inflation-neutral since it is set up in terms of nominal val- ues. NOTIONAL CAPITAL According to the NDC formula, for an individual, i, paying contributions on wages (or earnings), w, during an accounting period, t, with contribution rate, c, notional capital at the end of any period, T, is T Ki = (2.1) ,T c wi It , ,t t=1 where cwi,t denotes the individual's contributions in period t. Note that usually a ceiling will be set on earnings giving rights in the system. It is an index, calculated from an internal rate of return, t: T-1 It = (1+ t) t+1 IT =1. (2.2) The internal rate of return is discussed in greater detail below, but for now it is sufficient to say that it is the rate of return that maintains financial balance over time, for a given con- tribution rate. Some participants will not live long enough to claim a benefit, and this creates "inheri- tance capital," which in a closed system will be distributed to the account values of the survivors in the insurance collective. In practice, this can be done on a birth cohort basis, and in accordance with the individual's share of capital in the cohort's total capital. The inheritance capital is distributed in conjunction with some set age, such as the minimum age at which an annuity can be claimed. This adds another component to individual capi- tal at retirement, increasing the return on capital. 22 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES THE ANNUITY Total capital at retirement, beginning in period , is the sum of capital accredited the indi- vidual's account prior to the period of retirement. Total capital at the end of the period prior to retirement, K­1, divided by an annuity factor, G, gives the initial value of the annu- ity, P, for pensioner j, from birth cohort Pj, = Kj ,-1 GLE , LE ( ) (2.3) The annuity is a function of two factors. The first is cohort-based life expectancy, LE, for cohort at retirement. The second is the internal rate of return, , computed over the same period. The latter becomes an indexation formula in practice. The life expectancy factor. In principle, men and women would have separate LE factors. However, in countries where NDC schemes have been introduced, the annuity is calcu- lated using unisex life expectancy. The use of unisex life expectancy in calculating the annuity introduces an explicit redistribution from those who live a shorter life to those who live a longer life, for this reason, given current longevity patterns and a unisex LE fac- tor that entails a transfer from men to women. The rate of return and the life annuity. An NDC scheme provides a life annuity that encom- passes a real rate of return--determined by the internal rate of return, discussed below. The rate of return for the period over which the annuity is to be paid out is only known ex post, whereas the annuity is calculated ex ante. This means that in practice a computa- tional rule is required to determine how is to be entered into the annuity formula. One procedure is to include an ex ante value of in the formula. In practice this could be any value reasonably close to the expected internal rate of return. This approach requires a second rule for dealing with the deviation of the actual outcomes from the included ex ante in the formula used to compute the annuity. A convenient method is to adjust (for example, annually) for the difference between the rate of return included in the annuity, which can be called the norm, and the actual rate of return determined on an ex post basis (annually). If the actual return is less than the norm used in the annuity, a nega- tive correction is called for, and vice versa.6 In practice, use of formula 2.3 "frontloads" the annuity and, in economic terms, creates a higher annuity in the beginning of the total payout period compared with straight-for- ward yearly indexation, but at the expense of a lower annuity (compared with straight-for- ward annual indexation) in the latter segment of the payout period. This is consistent with assuming that the time preference of individuals is weighed in favor of more consumption now than in the future, and creating the annuity to reflect this assumption. The alternative to including the rate of return in the annuity is to index the annuity annu- ally, still with the internal rate of return. This form of real indexation gives higher con- sumption for each year into the future--as long as there is real growth--but at the expense of a lower initial value. This is because the individual lifetime pension-benefit stream available to distribute is fixed. This in turn means that if the annuity is higher from the beginning, by including a norm in the calculation of the initial benefit, then benefit-based resources available for consumption when young are also higher. However, since there is a fixed sum of money to distribute over the lifetime for the individual, resources will be lower (relatively to a contemporaneous wage earner's earnings) when older. WHAT IS NDC? 23 Note also that the individual account is an entity that exists until death. This is a conve- nient property, for two reasons. First, in practice, it is always possible to continue to work and pay contributions on earnings, even after a full annuity has been claimed. Additional work and contributions always enhance account values. Second, an annuity can always be converted into an account value and vice versa, using formula 2.3, at any time after retire- ment, using the expected payout period to make the conversion. Note also that it is possi- ble to grant partial annuities--of any percentage--since it is technically easy to convert annuities back into account values and vice versa. This feature of the NDC scheme makes it very suitable for combining work and pensions in any proportion, enabling gradual retirement from the labor force. In sum, the lifetime capital of an individual j from retirement is determined by his or her account at retirement, cohort life expectancy at retirement, and the internal rate of return, expressed as a norm based on the internal rate of return (with the deviation correc- tion factor) or in the form of annual indexation with the internal rate of return. What's more, total (available) individual lifetime benefit payments in the NDC scheme can be dis- tributed over the life cycle of the retiree, either weighting them more heavily toward the beginning of the retirement period or allowing them to increase gradually as the country's prosperity increases during the life annuity period. THE INTERNAL RATE OF RETURN AND FINANCIAL BALANCE The internal rate of return is the rate of return required to keep the scheme in financial bal- ance, where financial balance is defined as the state where the present value of overall sys- tem assets PV(At) equals the present value of total system liabilities PV(Lt): that is, PV(Lt) = PV(At). (2.4) System liabilities at any time t are the sum of all NDC commitments to all living partic- ipants, both workers and pensioners. Where the liability to worker i is Ki,t and the liability to pensioner j is Pj (where, for simplicity, denotes the date when members of cohort , became pensioners)7 total liabilities are PV(Lt) = Ki,t + Pj, ,t (2.5) at time t. The present value of assets is the present value of the stream of future contribu- tions for all workers in period t plus funded (technical) reserves. This can be expressed as PV(At ) = TD cw (2.6) i,t+Fundt i=1 where TD denotes the turnover duration of contribution assets. TD is a density function, developed and described in Settergren and Mikula (2006). In discrete terms, the density function in period t is the product of the earnings-weighted average number of years par- ticipants have worked based on the age-earnings profile in period t and the payment weighted average number of years payments have to be made in year t based on the pay- ment profile in period t. The turnover duration of the contribution asset is the average time a unit of money (such as a dollar or a euro) is in the system. In a discrete time frame, TD represents a static time cross-section density of money flows in the system, as seen at time t. With each new period t there is potentially a new density of money flows, thus gen- erating a new value of TD. 24 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES In the generic NDC model, both the accounts of workers during the accumulation period and the accounts of pensioners--converted into annuities--earn the same rate of return in a given period t. In a steady state, defined by a fixed workforce-age-wage-distribution and fixed age- mortality rates, an NDC rate of return based on the (instantaneous) rate of growth of the contribution base is sufficient to maintain financial equilibrium. The rate of return is deter- mined by the rate of growth of productivity, g, and the population, or more specifically in the actual pension context, the labor force, . In a steady state, the latter is determined by the rate of growth of the working-age population, and is what Samuelson (1958) called the biological rate of return. The rate of return of + g is not sufficient to maintain financial equilibrium8 under var- ious circumstances. Given any rate of economic growth--that is, + g--different distribu- tions of contribution and benefit payment flows will affect the NDC scheme's ability to maintain temporal balance between assets and liabilities. In insurance terms, the time money remains in the system before it has to be paid out affects the instantaneous liquid- ity of the system. This is shown in Settergren and Mikula (2006) in the context of a more- than-two generation model, since two-generation models are not sufficient to bring out this feature. If it were known in period t that a unit of money were to remain in the system some duration of time so that TDt is greater or less than TDt then the value of assets according -1, to expression 2.6 would change solely as a result of this, and this change would constitute a component of the internal rate of return = [PV(At)/PV(Lt)] ­ 1. (2.7) In the generic NDC scheme, this gives an internal real rate of return, = g + + , (2.8) also derived in Settergren and Mikula (2006). The component constitutes an adjustment of the well-known real growth criterion, g + , which can arise in nonsteady state. In prac- tice, an operative procedure has to be developed to determine the value of , based on an estimate of liabilities and assets (see below). Also in practice, will consist of a pure pay- ment timing component and a component that picks up what can be called system "noise": that is, design features that can lead to a less than "ideal" outcome. There is a market-based financial alternative to determining the internal rate of return. This would be for the government to issue Buchanan bonds to cover the liabilities of the system. Full market monetiziation of an NDC scheme through the issuance of Buchanan bonds can be criticized on two counts, however. First, if the wage sum, which using the terminology here is wi,t (from equation 2.1) is proportionate to national income, or approximately so over the long run--which would be the case if the share of profits in income were constant over time--then contributions, which constitute new account assets, would also be a constant proportion of national income and the tax base. This means that new Buchanan bonds earn a government decreed rate of return, equal to growth of income, t + gt, and that the ratio of government Buchanan bond liabilities to national income could grow at the rate t + gt for any given contribution rate, c, (and any given profit ratio). However, if the government were to set a rate of return on the Buchanan bonds of t + gt, this would be tantamount to running a nonmonetized NDC scheme with internal rate of return t + gt but with a deadweight loss required to cover WHAT IS NDC? 25 market issuance and transaction costs. This makes the nonmonetized NDC scheme more efficient.9 In addition, this procedure would not account for the possible effects on system finances of the factor . Second, if the bond (which over time would have to be a series of bonds) were to be a normal government bond(s), the rate of return would depend on the government's status as a borrower on the market. In such a set-up, if the rate of return on bonds is greater than the rate of growth, the money transferred to NDC accounts through this mechanism would be financed by a tax on the same persons who are the workers and pensioners in the NDC scheme. This not only introduces transaction costs, but also rate of return and a time-related distributional profile that depend on the government's creditworthiness and market evaluation of government policy. Properties of NDC NDC is identified through a set of properties, which are satisfied by setting up the system as described above. These properties are: Property 1. At any time the present value of an individual's lifetime benefit equals the individual's account balance. For each participant and at all times, the amount in the account, K, is the present or expected value of his or her benefit. The value of the account is determined by the indi- vidual's own contributions and the system's internal rate of return. This is a first property of an NDC scheme. It can be stated (where E denotes the expected value) as E(Pt) = Kt. (2.9) This is a property of a defined contribution system. Note that it has an important economic implication: Contributions constitute a payment for the individual's own pension, similar to premium payments to a financial account system. Fulfillment of this property is a nec- essary condition for a scheme to be called an NDC scheme. Property 2. To maintain a fixed contribution rate, total NDC system assets must equal or be greater than total liabilities. In the generic system, set out above, assets always equal liabilities. This gives a second property defining an NDC scheme At Lt. (2.10) Fulfillment of this property is necessary to maintain a constant long-term contribution rate. This is also a property that is consistent with a financial defined contribution scheme, where A = L. The process of individual account valuation distributes the returns among the participants in the generic NDC system, fulfilling the equality 2.10. In practice, design issues impose various constraints. As long as A > L, NDC schemes other than pure generic NDC schemes can also be called NDC schemes according to this definition. Property 3. The NDC benefit is constructed as a life annuity reflecting life expectancy at retirement. A key feature of an NDC scheme is that the annuity is based on life expectancy. A correct valuation of life expectancy leads to a correct valuation of the liabilities associated with annuity payments. Even if a lump-sum payment were theoretically possible, it is not a 26 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES desirable feature of a mandatory system that aspires to protect people in old age (and tax- payers) from the temptations of short-sightedness or the results of otherwise poor plan- ning or management of money. Property 4. Financial balance in the generic system requires that accounts be valued at the internal rate g + + . If in practice the rate of return is set lower, for example, by not factoring full real growth into the annuity, which may be the result of a policy restriction during the implementation phase10 than compared with the counterfactual of generic NDC, a surplus arises--and if it is used for purposes other than accrediting accounts, then it should be viewed as a tax. If the system is monetized with Buchanan bonds, then the monetization process gives rise to trans- action costs, which is also a tax, albeit a tax whose financial source is exogenous to the NDC system. In this case, it does not impede upon the function of the internal rate of return. RESERVE FUND All other things equal, if the NDC scheme is not monetized with Buchanan bonds, the occurrence of large and small fluctuating birth cohorts means that, ceteris paribus, contri- butions of large cohorts should be funded and paid out when the large cohorts become pensioners. If this is not done, there will be a tendency for the system to be out of financial equilibrium. For example, in a two-or-more-cohort world with a large and small cohorts, this would result in a situation where L > A when large cohorts are working, with the dif- ference between this outcome and A = L being the amount of the "missing fund." With automatic balancing (see discussion below) the automatic balance would be triggered to restore the condition A = L. Also, the rate of return on reserves must be at least equal to the internal rate of return. The reserve fund serves the purpose of collecting and disbursing the demographic bal- ance. It would be difficult, if not impossible, to do this without the institution of the reserve fund. More generally, the reserve fund is a general buffer fund. Its sources of funds and withdrawals will depend on the exact design of the system. A number of examples can be given of how design affects the reserves. One is that accounts can be augmented with noncontributory rights. These must have a source of financing, such as general tax revenues or a special contribution rate earmarked for this purpose, and these revenues should be transferred to the reserve fund. The financial source is the asset counterpart of the liability to the account holder. This money should also earn at least the rate of return accredited individual accounts of workers. A second example is a "tax overhang" from the old system, to be elaborated upon below, which should be quantified and funded--in the sense that it also needs to earn the internal rate of return. A third example of the use of the reserve fund is the Swedish reform principle, which uses g to index accounts and benefits, based on the principle that the value of benefits should increase at the rate of growth of the per capita covered wage. A liquidity band is defined between the value A/L = 1 and some value greater than unity. As long as A/L remains within the band, the surplus remains as liquidity in the system, and the system is "insulated" from changes in + in expression 2.8. Liquidity in excess of the band ceiling is distributed according to a rule, and the automatic balancing mechanism (see below) bal- ances the system for A < L. Within the band, no reserves are disbursed. In practice, reserves are needed for other reasons, as will be discussed in the next section. WHAT IS NDC? 27 AUTONOMY An NDC system is self-adjusting financially and can run without external intervention. The generic system is autonomous by construction, which is a result of fulfillling proper- ties 1 through 4, and maintaining a reserve fund. Of course, whether politicians will let the system run free from intervention is another question. Finally, no country has introduced the generic model as it is presented here. Design in implementation varies considerably among countries that have implemented varieties of NDC schemes. Where they generally differ is in the indexation rules chosen, and to date, only Sweden has legislated an ABM. There are many reasons for the deviations from the generic model. These include the transitional considerations of phasing in the indexation of new NDC benefits with the already existing stock of pension benefits at the time of implementation (Palmer 2006), political goals, and data and information requirements. Automatic Balancing Because system designs deviate from the generic model, in practice, equation 2.7 will encompass all design deviations from the generic model. Hence in practice, equation 2.7 yields bt = [PV(At)/PV(Lt)] ­ 1 (2.11) where bt = t if the model is generic, but otherwise not. The value bt is the factor with which the accounts of workers and annuities of pensioners needs to be adjusted to main- tain balance. This is the automatic balancing mechanism (ABM). The ABM mechanism is described as follows. If PV(At) < PV(Lt), regardless of the cause, account values of workers and annuities of pensioners will have to be given a lower rate of return to bring the system back into financial equilibrium. On the other hand, the system can distribute a higher rate of return to participants if PV(At) > PV(Lt). The ABM is used in the country in which it was conceived, Sweden.11 Other countries adopting NDC have relied on system design that can be expected to give A > L primarily as a result of not dis- tributing all possible internal returns. This is tantamount to imposing a tax on returns and using the revenues for other purposes.12 Some examples of the most common sources of generic design deviations follow. One is that ex post mortality rates can deviate from the ex ante values used to compute life expectancy for newly granted annuities. Another is that annuities may cost more (or less) than thought ex ante. For example, persons with higher than average incomes (and contri- butions) may live longer, and the annuity factor may not be designed to account for this, or if it is, it might not do the whole job. Another form of generic deviation would be the choice to use the per capita rate, g, as the rate of return, rather than a measure closer to the internal rate of return. This choice can be motivated by the goal of maintaining a constant ratio of an average pension to an average wage. Yet another form of generic deviation is the choice of the wage sum rate g + for indexation, instead of g + + . Note also that, ceteris paribus, the choice of g + for indexation would be the same as using g together with a balancing mechanism without the proposed Swedish liquidity interval. There are more examples of how financial imbalance can arise. A rate of return on reserves that deviates from the rate accredited individual accounts also creates imbalance. Potential imbalance is also caused by the fact that in practice indices used to compute the rate of return need to be based on historical data, and probably a smoothing mechanism, 28 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES which means they are always one or more periods behind. This may be no or only a small problem for random positive and negative fluctuations, but it can become a serious prob- lem for negative trends, especially in , if g is the chosen system rate of return. Finally, there may be a liability overhang after conversion from the old DB to the new NDC scheme, with unfinanced liabilities. How a tax overhang should be dealt with is discussed below. Valdés-Prieto (2006) presents an alternative to the ABM. He envisages market determina- tion of the balancing mechanism through the issuance of a limited number of bonds with property rights, which he labels integration to financial markets, or IFM. He argues that full coverage of NDC debt is not essential to its valuation, and that a limited issuance of debt would be sufficient to enable the NDC rate of return to be determined by the market. It is not evident, however, why this rate of return would be preferable to + g + . Computing Life Expectancy Some procedure must be adopted to compute life expectancy at retirement. At least three procedures can be employed in practice. These are described and discussed briefly in this section. One approach is to form a committee of demographic experts, charged with the respon- sibility of performing analyses that lead to official cohort projections. Their proceedings, analytical report, and minutes from deliberations would be published to assure that the process is as transparent as possible. Revisions can be made annually, but with relatively accurate long-term projections from the outset, only small, infrequent revisions would be needed. Revisions would always apply to nonretired cohorts and would be greater the far- ther from retirement age the cohort is. The transparent process would support system autonomy from political interventions in the projection process. A second approach is to base the estimates of birth-cohort life expectancy on known ex post cross-sectional survival data. This provides stronger protection from political inter- vention since it involves no process of judgment. However, inherent in this procedure is the almost certain risk that projections will lag behind reality, systematically creating higher than warranted annuities for older workers, ceteris paribus, shifting costs to future cohorts. This financial strain on the system is eventually rectified if the system design includes automatic balancing, but a distributional effect remains. Furthermore, all other things equal, this procedure yields an unfinanced deficit. A third alternative is to use current (ex post) cross-sectional data and regularly adjust the benefits of all retirees in accordance with new life expectancy tables as they are revised. Compared with trying to get the projection right from the beginning, as in the first alter- native, or burying the problem in the general error-correction mechanism, as in the sec- ond, the naked perpetual change model, although right on target, is nevertheless the least appealing for the policy maker. All other things equal, it means that the administrator will be constantly decreasing the value of the benefits of retirees of all ages as new life expectancy statistics become available. For some longer-lived pensioners, this process of benefit deflation could go on for over a quarter of a century. This method shares the ceteris paribus disadvantage of the second method, where the automatic balancing mechanism performs the task of balancing all possible causes of a deviation between assets and liabil- ities, including errors in projecting life expectancy. On balance, the first and third approaches explicitly fulfill the criterion of financial bal- ance, whereas the second and third provide greater autonomy, at least in principle, since they are based on actual outcomes, which presumably cannot be influenced by politicians. The first approach has a good chance of being autonomous, however, if its operation and proceedings are made transparent for the public. WHAT IS NDC? 29 Converting to NDC with a "Tax Overhang" from the Old DB Framework A country may wish to introduce NDC but is confronted with what we will call a "tax overhang" from the previous system. A tax overhang can be defined vis-ŕ-vis the counter- factual of having had an NDC scheme from the very outset, including a contribution rate fixed at some specific level. This is obviously a controversial statement, but the logic in it is that if the policy maker has decided that NDC is the best option, then it is consistent to view old pay-as-you-go commitments that cannot be financed by the contribution rate chosen for the NDC scheme as a tax overhang. An example of a tax overhang would be the failure to create a fund for a large birth cohort of persons for whom most of their working careers have lapsed at the time of con- version to NDC. Another example would be old-system commitments that are more gen- erous than those the NDC formula provides, and which the government decides to honor as a part of the transition to NDC. In principle, the tax overhang can be estimated and the government's commitment to honor it can be made transparent by including the debt to the system as a fund asset in equation 2.11, which defines total assets. To finance the overhang, the government could transfer money or issue bond(s) to be held by the pension fund, yielding the internal rate of return, and which would be mone- tized at a pace needed to finance tax-overhang commitments. Alternatively, the amount owed would be kept on the books, and the government would finance the amount owed (with the internal rate of return) as payment commitments occur, through a tax levied as the liabilities have to be honored. In practice, there is no difference between these two methods, although issuing a bond has the formal advantage of being a more steadfast commitment. NDC and Social Policy Considerations The Scale of the System The introduction of NDC does not mark the downfall of social policy. On the contrary, it provides a means for separating the goals of social policy from the goal of smoothing out individual consumption over the life cycle with the aid of a mandatory insurance scheme. A first consideration is that the scale of the NDC scheme--like any other mandatory pension scheme--should be set so as not to impose unjustifiably on individual preferences regarding the time preference of consumption. For example, in poorer societies, a large pension in the future paid at the expense of current consumption can easily conflict with the obvious time preference of persons living in poverty who would prefer higher con- sumption in the present. This suggests a lower scale for the system and a high minimum pension age. On the other hand, in richer societies, many people--but never all--can pre- fer a lower scale to leave room for individual choices for work and saving over the life cycle. A reasonable scale goal in a developed market economy is to provide an adequately dimensioned NDC benefit in old age for the normal worker retiring at or close to the min- imum retirement age, but leaving room for individual choice regarding consumption and leisure. The minimum retirement age needs to be set so that people cannot claim a benefit at an age that by definition gives an annuity that on average is too low, given that benefits are based on life expectancy from retirement. Generally speaking, the size of an annuity is determined by individual preferences between work and leisure. In an NDC scheme, individuals themselves determine the tim- ing and scope of their withdrawal from the labor market. If the tax rules provide equal treatment of pensions and earnings, an NDC scheme is neutral in the work-leisure deci- sions of older workers (at least above the guarantee level). 30 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Also, NDC accounts are "infinitely" devisable. This means that only practical consider- ations determine what percent of a full annuity an individual can claim at any time--after reaching the minimum pension age--and it is possible to combine work and a pension in any proportions and at any age. Finally, NDC (like FDC), by basing benefits on individual accounts, facilitates labor mobility between occupations, branches, sectors of the economy, and regions of the country, or for that matter, between countries. Minimum Pension Income--for the Elderly By definition, although an NDC scheme provides some benefit for all who work and con- tribute to the scheme, it will not necessarily provide a benefit that gives sufficient income for all to live on. There will always be some in the population whose genetic and human capital--or other individual characteristics or circumstances--will lead to an insufficient NDC benefit. Continuing with the discussion in the previous section, generally speaking, it is impor- tant to emphasize that the sufficiency of the NDC benefit depends on the minimum pen- sion age. Data on national pension systems reveal that people tend to exit the labor force at the minimum possible age, even though this may give them a low lifelong benefit. Part of the essence of the mandatory scheme is protect people from their own bad judgment in this sense. Also, the practice of setting lower pension ages for women, whose exit from the labor force may be timed to coincide with the exit of her (often older) spouse, itself can lead to poverty in old age for surviving women. Finally, it is reasonable to adjust (index) this age, once thoughtfully set, to changing longevity. Given a reasonably set minimum pension age, some form of minimum income floor (for example, a guarantee or flat rate) is unavoidable. For various reasons, there will always be a certain part of the older population that will not have a sufficient contribution-based pension to live on. Therefore an NDC scheme will always need to be supported at the "zero" level by some form of external transfer paid by general tax revenues. Finally, given some form of minimum income transfer floor for the worst-off in society, a threshold effect will occur at this floor level that may mean that additional contributions to the NDC scheme in old age will not have a counterpart in a benefit, in the sense that they do not bring the contributor above the guarantee level. Remaining in the formal labor force may nevertheless be worthwhile given individual preferences on work and leisure-- or necessary, because of dire circumstances. In countries where the informal market is an important feature of the economic landscape, people may take their minimum benefit and (continue to) work informally. This could be regarded as an informal counterpart to flexi- ble retirement in a formal market economy. Putting Distributional Policy into NDC It is possible to supplement NDC (and FDC) schemes with noncontributory rights. What is essential, to maintain financial stability, is to finance these as they are granted, since these represent liabilities that need to have an asset counterpart in the financial balance. Since all contributions to the NDC scheme on earnings give rights directly to the contributing par- ticipant, there is no room to finance noncontributory rights from individual NDC contri- butions. Instead, these will have to be paid from other sources of public revenues. The financing of noncontributory rights in the same accounting period in which they are granted keeps order in the system. The alternative would be for the accounting system to record a claim on the state budget (including the internal rate of interest), with the state making partial payments on this debt sometime in the future when it becomes time to honor the commitments. This creates a different payment distributional profile. In the worst case, a future government can decide not to honor these rights. WHAT IS NDC? 31 Finally, the principle of separating distributional aspects of policy from the pure NDC insurance system has the advantage of creating transparent distributional policy. All deci- sions to introduce a new distributional right must be specified clearly, including the source of finance identified. The Connection to Disability Benefits In most social insurance pension systems devised during the past century, disability bene- fits have been a part of the overall pension package. This is not true of an NDC scheme, which is an old age insurance scheme. Because of this, integration of disability into the NDC framework is an important issue. Before receiving disability, a worker will normally have had an account in the NDC system. In principle, the straightforward way to handle disability and NDC is to devise a rule for paying--that is, financing externally to the NDC scheme--contributions for the disabled during the remaining period up until the mini- mum pension age (or some other age at which a disability benefit is replaced by an old age benefit), and then to replace the disability pension with an old age benefit from the NDC scheme. The contribution rule chosen for financing the disability benefit will determine the coverage earned during the disability period. Sharing NDC Accounts between Spouses It is possible to share accounts between spouses in NDC. The principles of legal codes gov- erning the financial affairs of spouses could also be extended to NDC accounts. The dis- cussion that follows describes how this might work in practice. In doing so, it is important to distinguish between the accumulation period and the annuity period. During the accumulation period, NDC capital earned prior to marriage (or some form of legal cohabitation, depending on national legislation) could remain with the individual, whereas capital earned during a marriage would be shared equally between the spouses. This capital would also be retained upon divorce, and would remain the individual's own capital upon entering into a new marriage. Here two considerations are important. The first is whether the account division rule should be mandatory or voluntary, and the second is whether the default option should be joint accounts or single accounts. A main argument against mandatory sharing of accounts in a country where cohabitation is legally recognized is that persons uninterested in shar- ing accounts would choose not to marry, whereas some married people might even choose divorce plus cohabitation as a means to get around a mandatory system upon its introduc- tion. The latter means that a cohabitation rule would also have to be devised, or that this behavior would have to be accepted. Because of these possible behavioral responses, voluntary sharing of accounts can seem more appealing. Then the question is whether independent accounts should be the default, with a positive decision of both spouses required to enter into shared accounts, or whether shared accounts should be the default, requiring a joint decision to create sepa- rate accounts. Making joint accounts the default option certainly agrees well with other joint property arrangements for spouses. Joint accounts would always be to the advantage of the spouse with lower earnings and would help compensate for more time spent in non- market activities. The next question is, should a joint annuity for spouses (cohabitants) be mandatory, voluntary, or not permitted at all in NDC? The goal of a joint annuity is to provide a better annuity in old age for the surviving spouse, generally women. Would this be fulfillled by entering into joint annuities? The answer to the question is empirical. If unisex life expectancy is used in computing the individual annuity, it is possible to find examples where women would lose with a joint NDC annuity. 32 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES To illustrate this, assume that women have on average lifetime earnings that are 80 per- cent of those of men, that unisex life expectancy is used in calculating the annuity, and that women have an expected (average) 15 percent transfer from men due to the difference in longevity. Sharing accounts equally would lead to a benefit for a woman based on 90 per- cent of her spouse's life income. However, if the single benefit alternative means that the average woman would receive a transfer that raises her benefit by 15 percent, due to the difference in the average life expectancy of men and women, she would receive a benefit amounting to (0.8 × 1.15) = 92 percent of her spouse's income, which is superior to the ben- efit based on a straightforward splitting of pension rights. Generally speaking, shared accounts are to the disadvantage of women if their life earnings are close to or surpass those of men, but become advantageous as the lifetime earnings gap becomes large. This illustration suggests that what seems to be a simple question turns out not to be so simple. A mandatory system requiring joint annuities at retirement could actually work against the goal of providing better benefits for many women. Seen over the long run it could also become counterproductive, as the gender earnings gap becomes small. A vol- untary rule would provide an opportunity for spouses with relatively large lifetime earn- ings differences to share accounts, thereby providing an opportunity for the spouse with lower life earnings to secure a better benefit. A voluntary arrangement requires the con- sent of both spouses: that is, including the spouse with the higher life earnings. This sometimes may not give the desired result, but as a rule, it would be socially desirable. In sum, voluntary sharing of accounts, with sharing as the default, is an alternative in the accumulation period that reflects most country's legal arrangements for sharing of prop- erty between spouses; however, it requires special rules for cohabiting couples. Voluntary sharing of annuities is preferable to mandatory sharing of annuities, but the default should be chosen with respect to who the winners are expected to be. This is because the closer the earnings of women come to those of men, single annuities calculated with uni- sex life expectancy will be to the advantage of surviving women--who constitute the likely survivors. Final Remarks This study has established a generic conceptual framework for NDC pension schemes, an innovation that emerged as recently as the mid-1990s. Important design issues have been discussed. Still, it remains to continue to identify issues and improve our engineering as we gain experience. Notes 1. "Notional defined contribution" and "non-financial defined contribution" should be understood to have the same definition. 2. See Samuelson (1958). 3. This is an accepted tenant of economics. See, for example, Lindbeck and Persson (2003); Valdés-Prieto (2006). 4. See, for example, Burtless (2003). 5. See Diamond (2003). 6. To date, two countries, Italy and Sweden, have included a rate of return in the annu- ity. Both use per capita covered income as the norm, rather than the internal rate of return (see below). Sweden applies the type of correction procedure described here. Italy has no legislated correction procedure. WHAT IS NDC? 33 7. Cohort is implicitly assumed to either all become pensioners in or to be normally distributed around so that it would be a "money-weighted" average date of retirement. 8. See, for example, Palmer (1999) and Valdés-Prieto (2000). 9. See Góra and Palmer (2004). 10. See Palmer (2006). 11. The ABM was first presented in the international literature in Settergren (2001). 12. For example, Latvia is using this tax to help finance the introduction of a mandatory individual financial account scheme. References Boskin, M., L. J. Kotlikoff, and J. Shoven. 1988. A Proposal for Fundamental Social Security Reform in the 21st century. Lexington, Mass.: Lexington Books. Buchanan, J. 1968. "Social Insurance in a Growing Economy: A Proposal for Radical Reform." National Tax Journal December 21: 386­97. Burtless, G. 2003. Asset Accumulation and Retirement Income Under Individual Retirement Accounts. Evidence from Five Countries. Washington, DC: The Brookings Institute. Diamond, P. 2003. Taxation, Incomplete Markets, and Social Security. Munich Lectures in Eco- nomics--Center for Economic Studies. Cambridge, Mass. and London, England: MIT Press. Góra, M., and E. Palmer. 2004. "Shifting Perspectives in Pensions." Institute for Study of Labor (IZA) Discussion Paper 1369. Bonn, Germany. Iyer, S. 1999. Actuarial Mathematics of Social Security Pensions. Geneva: International Labour Office (ILO). Lindbeck, A., and M. Persson. 2003. "The Gains from Pension Reform." Journal of Economic Literature XLI (March): 74­112. Palmer, E. 1999. "Individual Decisions and Aggregate Stability in a NDC PAYG Account Scheme." Avaliable on: www.rfv.se/english. ------. 2000. "The Swedish Pension Reform Model: Framework and Issues." Social Pro- tection Paper 0012, Pension Reform Primer. Washington, DC: World Bank. ------. 2006. "Conversion to NDCs--Issues and Models." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 9. Washington, DC: World Bank. Samuelson, P. 1958. "An Exact Consumption-Loan Model of Interest With or Without the Social Contrivance of Money." The Journal of Political Economy 6 (December): 467­82. Settergren, O. 2001. "The Automatic Balance Mechanism of the Swedish Pension System: A Non-technical Introduction," Wirtschaftspolitishe Blätter 4/2001: 339­49. Settergren, O., and B. D. Mikula. 2006. "The Rate of Return of Pay-As-You-Go Pension Sys- tems: A More Exact Consumption-Loan Model of Interest." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 7. Washington, DC: World Bank. Valdés-Prieto, S. 2000. "The Financial Stability of Notional Account Pensions." Scandina- vian Journal of Economics 102 (3): 395­417. ------. 2006. "A Market Method to Endow NDC Systems with Automatic Financial Sta- bility." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 8. Washington, DC: World Bank. Chapter 3 What Are NDC Systems? What Do They Bring to Reform Strategies? Axel H. Börsch-Supan* THE PRESSURES EXERTED BY POPULATION AGING, amplified by the negative incentive effects that induce early retirement, make public pension systems unsustainable all over the world. The demographic pressures are strong in Europe because Europe's population is already relatively old. Demographic change is occurring particularly quickly in Asia. Japan shares both the European level and the Asian speed, an especially dangerous com- bination. Population aging is not sparing the more youthful United States, but the speed and level is slower. Aging also affects the developing countries and adds to their difficult economic and political problems. It comes as no surprise that pension reform is on the agenda in almost all countries of the globe. Pension reforms seem to follow a wave-like pattern. After a wave of reforms and reform attempts with a strong stress on prefunding, not the least precipitated by the World Bank's 1994 book, Averting the Old-Age Crisis: Policies to Protect the Old and Promote Growth, the newest wave of reforms fashion "notional defined contribution" (NDC) systems.1 This study's aim is to make a systematic assessment of such a reform strategy, in both economic and political terms. This study poses two questions: what are NDC systems, and what do NDC systems bring to pension reform? It may come as a surprise that the second question in the title has more quickly been answered (and often with more confidence) than the first question. This chapter goes back to square one. It begins with the first question. While there are large differences between a Beveridgian defined benefit (DB) pension scheme and an NDC system, and large differences between an NDC system and a prefunded defined contribu- * Axel H. Börsch-Supan is director of the Mannheim Research Institute for the Economics of Aging (MEA) and professor of macroeconomics and public policy at the University of Mannheim, Germany. This study takes as its point of departure and owes much in the way of intellectual debt to Disney (1999). This analysis has profited from Richard Disney's detailed comments, as well as those of Peter Diamond, Ed Palmer, Anette Reil-Held, and Christina Wilke, and from the lively discussion during and after the Sandhamn conference. Special thanks go to Nick Barr, Anna Hedborg, Robert Holzmann, Assar Lindbeck, Michal Rutkowski, and Salvador Valdés-Prieto. The German National Science Foundation (DFG) and the Gesamtverband der Deutschen Versicherungswirtschaft (GDV) provided additional financial support. The usual disclaimer applies. 35 36 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES tion (DC) system, the difference between Bismarckian DB pension schemes and NDC sys- tems is less clear-cut, in terms both of the underlying economic substance and the percep- tion in the political arena. This study thus provides a taxonomy of pension systems that helps structure answers to both questions. The analysis shows how blurred the distinction between conventional DB and NDC systems can become, and tries to condense the economic meaning of NDC. Our answer to the second question follows the same line. Properly designed NDC sys- tems can contain very powerful economic and political mechanisms that may facilitate pension reform, such as transparency and accountability. Cleverly designed DB systems, however, may often do the same job--and in some circumstances may be even better. Whether NDC systems bring new life into the pension debate is therefore a question of the historical path and the nature of the debate in each country. The second section sets the stage by stressing the most important challenges that NDC systems are supposed to master. The third section is conceptual and devoted to the first question in the chapter. It defines an NDC system on the mechanical level and then pro- ceeds to extract its economic and political core. The study contrasts this to DB and FDC schemes. The fourth section answers the second question. It discusses the pros and cons of an NDC system in terms of economics, the perception of workers and pensioners, and the political process. The fifth section provides a real-life example of a well-known conven- tional DB system, which almost perfectly mimics an NDC system. This is the German "point system" augmented by a so-called sustainability factor as proposed by the German government's reform commission for the "sustainability in financing the social insurance systems." The sixth section concludes and picks up the challenges described in the second section: which of those challenges can be mastered by NDC systems, and which challenges must be addressed by other economic and political mechanisms, notably prefunding? The Challenges The first and foremost challenge to pension systems around the globe is population aging, long ago heralded by often-cited publications such as OECD (1988) and World Bank (1994). The consequences of demographic change have now reached many countries in a perceptible way. Population aging is becoming one of the secular "mega trends" of the new century. Demographic change has two distinct components: a decline in fertility and an increase in longevity. The fertility decline is most pronounced in Europe and occurred as a histori- cally unparalleled and in many countries rather sudden sequence of the baby boom in the late 1950s and early 1960s, followed by the baby bust in the 1970s. In the developing coun- tries, fertility decline has been less sudden, but is still steady and incisive. The secular increase in life expectancy is largely due to the progress in medical technology. With an unchanged or even declining retirement age, the increase in longevity has led to a massive expansion of pension systems. Coping with these two demographic developments requires two strategies. To deal with the sudden decline in fertility due to the baby boom/baby bust transition, a reduc- tion of the pay-as-you-go (PAYG) replacement rate for the baby boomers and, simultane- ously, an increase of privately financed prefunded pensions substituting for part of the PAYG pensions seems unavoidable in those countries that already have very high payroll taxes. Increasing longevity, in turn, is most naturally addressed by an increase of the length of working life: that is, a higher retirement age. Demography is, however, not the only challenge to current pension systems. The Euro- pean economies--and not only they--face two deeply routed macroeconomic problems: WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 37 poor growth and high unemployment. The most pressing goal of economic policy is therefore to increase employment (and thereby growth), not the least in order to provide the eco- nomic base that should support the social security systems in general, and the pension sys- tem in particular. If at least a part of pension contributions is perceived as distortionary taxes, the demographic and macroeconomic challenges interact with each other through the wedge that pension contributions drive between total labor costs paid by employers and net earnings received by employees. Germany is a striking example of a country where population aging and macroeconomic weakness combine to create serious eco- nomic problems. Germany has the lowest economic growth rate among all European Union (EU) countries. At the same time, Germany has by far the highest labor costs (total hourly labor compensation) within the EU.2 Much of the thinking of current German pen- sion reform is guided by the conviction that the goal of more growth demands more employment, and more employment demands at least a stabilization of--or better yet, a reduction of--payroll taxes and tax-like contributions. Stabilizing or reducing contribution rates, however, requires cost cutting. This can be done by shifting the retirement age or by reducing pension benefits. Both reform strategies cut deeply into existing claims. It is an illusion to sell the necessary structural reform steps as a "win-win" situation. Neither a reduction of the replacement rate nor an increase of the retirement age is popular. To minimize negative coalitions against reform, employees should therefore be able to choose as flexibly as possible between the two unpopular options. Many people, particularly in Europe, view pension systems as social achievements that one must defend, no matter how much the demographic and macroeconomic environ- ment has changed. Hence a change of paradigm is necessary from thinking what one wishes to claim toward thinking what can be financed. This requires a change in the rhetoric of our pension systems; it also requires a transparent statement of the costs of the current systems that support this rhetoric. There are further political challenges. Confidence in the PAYG pension systems has been steadily declining during the past two decades (Börsch-Supan and Miegel 2001). This poses a huge political challenge since every reform must prove that things will eventually get better than without reform. Rhetoric alone will not bring confidence back, in particular because it has been discredited in the past. What is needed--not only for a change of par- adigm, but also to restore confidence--is transparency and a credible reform process. Adjustment processes should be gradual. Any kind of emergency operations undermine credibility. The recent history of German reform provides bad examples. These include the 2000 unexpected discretionary indexation rule change and the 2002 contribution rate hikes. These occurred after much of the liquidity reserve had been given up during the last business cycle upturn, exposing the pension system to sudden changes and thereby fur- ther undermining confidence in the PAYG system. Another challenge consists in the many bad microeconomic design features in existing PAYG systems that create negative incentive effects. DB systems that base benefits solely on the final year of earnings, for example, do not reward additional years of work and yield incentives for strongly backloaded earnings. Other examples are systems that pro- vide benefits that are not actuarially related to retirement age. There is strong evidence that the negative incentive effects exerted by such systems have shifted the effective retirement age to substantially earlier ages, and thus contributed to the financial pres- sures on PAYG systems.3 In Germany, early retirement makes up for almost 25 percent of the old-age pension budget, corresponding to almost 5 percentage points of the contribu- tion rate.4 Negative labor supply incentives are an important reason to make contributions closely reflect pension benefits (the "equivalence" or "insurance principle"). This will minimize 38 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES the tax-character of contributions. Ideally, a fully equivalent pension system with an inter- nal rate of return equal to a suitably risk-adjusted capital return has no tax-character at all. In turn, violating the equivalence principle, providing relatively low rates of return, and a lack of credibility all add to the tax-character of contributions and thus to labor market dis- tortions. In a similar vein, flexibility to choose the date for retirement entry and the aboli- tion of earnings tests minimize distortions in the choice between labor and leisure and should therefore also minimize tax-like distortions. In summary, the challenges are manifold. Population aging creates financial pressures on pension systems, requiring unpopular cuts in benefits and the duration of retirement. Macroeconomic growth and employment problems are amplified by pension systems if and when pension contributions are perceived as distortionary taxes. Reforms may lack political credibility in the same way as the underlying PAYG systems have lost their credi- bility. A host of bad design elements on the microeconomic level requires fixing to mini- mize negative incentive effects. Notional Defined Contribution Systems NDC systems are supposed to address these economic and political challenges. Propo- nents argue that NDC systems make a large step to solve all these problems in one big stroke. 5 Skeptics argue that they add little that is new but may distract from deeper reform.6 NDC systems were legislated in 1994 in Sweden and in 1995 in Italy, among other countries. In Sweden, the new system was introduced for all employees with a 15-year transition period. It will be fully implemented within a few years from now. Palmer (2000) provides a good description of the Swedish NDC system and its introduction. In Italy, the NDC system was introduced as part of the so-called Dini-Reform with a very long transition period. It will be relevant only for workers who are younger than the baby boom generation. Franco and Sartor (2006) provide a critical evaluation. While Sweden and Italy are the most cited examples of NDC systems, Latvia and Poland have actually served as trial grounds for these reforms.7 What are NDC systems? What are their mechanics, and what is the economic essence of NDC systems? The Mechanics of an NDC System NDC systems are accounting devices that treat a PAYG system like a DC system. Pension benefits are paid out of current contributions, as in a conventional PAYG system; however, the link between benefits and contributions is individualized and defined by the NDC accounting mechanism. Later this study will describe a "pure" NDC system as a special form of a "pure" PAYG system. In real life, almost all PAYG systems, and especially NDC systems, have a buffer fund, which make them "mixed" PAYG-funded systems. Like any other DC system, the system starts with the individual contributions to the pension system, which are credited to, and accumulated on, individual accounts kept by the pension system. The balance is fictitious (or "notional"), since no capital is accumu- lated. The accumulated sum represents the fictitious (or "notional") pension wealth. The balance earns interest at some rate of return. The magnitude of this return is a cen- tral parameter of the NDC system. Since no capital is accumulated and the claims on the balance are not traded, there is no market-mechanism to determine the rate of return.8 Viewed from a macroeconomic perspective, the "natural" rate of return for an NDC sys- tem is the implicit return of a PAYG system: that is, the growth rate of the contribution bill. However, some NDC systems--such as the Swedish system--have chosen rates of return, which are higher under current circumstances, such as the rate of wage growth. WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 39 When entering retirement, the notional pension wealth is converted into a lifelong pen- sion ("annuitized") according to actuarial rules. The annual pension benefit depends on three variables: · The notional pension wealth (proportionality guarantees equivalence) · The interest rate used to compute the annuity (using the implicit rate of return from the PAYG system guarantees equivalence within each birth cohort), and · Life expectancy at retirement (using up-to-date cohort-specific life tables guarantees actuarial sustainability). The two last elements are often combined and referred to as "annuitization divisors" or "G-values" in Sweden and Latvia.9 In Italy, these values have been tabulated. Benefits B are then B = PW/G, where NPW denotes the notional pension wealth. Including the implicit rate of return from a PAYG system and including the expected length of retirement in the benefit calculation links pension benefits to the demographic and employment parameters of the macroeconomic environment. This makes NDC sys- tems more sustainable than conventional DB systems in the sense that changes in the demographic and macroeconomic environment automatically lower benefits. Including the remaining life expectancy links benefits to the retirement age at the indi- vidual level. This makes NDC systems actuarially neutral (at the employed rate of inter- est). Since the present discounted value (PDV) of pension benefits is independent from the actual retirement age, the financial burden of the younger generation is fixed for each cohort and determined by the G-value: that is, the notional interest rate and the expected average duration of retirement. The actual properties of an NDC system depend on many detailed design features. They are discussed below. At this point, three features are worth noting. First and fore- most, the determination of the "notional" interest rate is central since it governs both the demographic and macroeconomic sustainability of the system and the microeconomic incentive effects. Second, it makes a big difference which life tables are used. Third, the extent to which retirees are protected from future shocks is an important parameter poten- tially conflicting with financial sustainability. NDC accounting systems do not change the mechanics of PAYG systems: that is, the necessity to adapt either the contributions or the replacement rate (or both) to changes in the demographic or macroeconomic environment. This is an important point. The current young generation pays the current old generation. The determination of the notional inter- est rate and the estimated remaining life expectancy amounts to the specification of the link between benefits (represented by some replacement rate) and contributions (repre- sented by some payroll-tax rate). By changing this link, the system can shift the burden of population aging between the younger and the older generation. A pure NDC system, however, is still financed purely PAYG, and thus cannot mimic a prefunded system in the sense that the financial burden of a cohorts' worth of pension benefits will be carried by that same cohort. The significance of this point is most clearly seen in the sudden transition from a thick baby boom to a thin baby bust generation. If the thick baby boom generation should finance a major part of their retirement income out of their own income, rather out the income of the much thinner baby bust generation, the baby boomers need to give up some 40 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES consumption early in life and transfer the corresponding resources to their postretirement period. This requires saving and the build-up of a real capital stock by the baby boomers. A notional capital stock cannot serve this purpose because the annuities computed from the national wealth accumulated by the baby boomers have to be financed by the contri- butions of the baby bust generation. The Economic Essence of an NDC System Stripped down to its economic essence, there are three crucial mechanisms that make a PAYG system an NDC system: 1. An accounting mechanism that credits all lifetime earnings 2. A mechanism linking the final balance with the demographic and macroeconomic environment 3. An actuarial rule converting the final balance into an annuity. To emphasize the last point made in the preceding subsection, a fourth element can be added, which distinguishes pure notional from fully funded DC plans: 4. Claims on future benefits are not collaterized with real capital but promises by a (almost always) government-related entity. The first mechanism is realized by crediting all lifetime contributions to an individual account, just like funded DC plans. In many respects, this parallels the German and French point systems, except that the unit of credit is currency (euros), not earnings points.10 It dif- fers substantially from the many pension systems in which only the X best years are cred- ited (at the extreme, only the earnings of the final year of work), and of course from Beveridgian systems that provide flat benefits. The second mechanism is realized in NDC systems by the notional interest rate--if cho- sen to be the internal rate of return of a PAYG system--that reflects demographic changes, employment changes, and productivity changes, since the contribution bill grows with the rate at which the number of contributors and their labor productivity increases. This mechanism parallels the indexation rules of conventional DB systems. Most pension sys- tems index their benefits at retirement to the current wage level. Most conventional PAYG systems add discretionary adaptations of the replacement rate to demographic changes; one of the main features of NDC systems is the direct and automatic linkage to demogra- phy once the notional interest includes the growth rate of the contribution base. In funded DC plans, the demographic and macroeconomic environments enter benefits through exactly the same mechanism as in NDC systems: namely through the rate of interest-- although the applicable rates of interest are generally different. The third mechanism is the essence of the B = NPW/G rule. Proportionality between B and NPW and an actuarially correct determination of the G-values makes an NDC system actuarially neutral and, within each cohort, also actuarially fair (see Disney 2003 for this taxonomy). Some conventional PAYG systems have actuarial adjustments, notably the U.S. social security system between age 62 and 65. Most funded DC plans are automati- cally actuarially neutral, since conversion to an annuity takes place at actual retirement. Many DB-type PAYG systems, however, have no or little linkage between annual benefits and retirement age.11 A Taxonomy of Pension Systems How close are NDC systems to funded DC systems? And how close are they to conven- tional PAYG-financed DB systems? The taxonomy in table 3.1 serves to clarify matters. It WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 41 distinguishes pension systems by four dimensions that are closely related to the four ele- ments listed in the preceding subsection. The many possible design features in real life, however, add additional complexity to these four dimensions. We do not claim that all dimensions are covered (such as voluntary vs. mandatory).12 One of the main features of table 3.1 is that all the dimensions provide a continuum of allocations within each dimension. NDC systems often enforce an extreme position along a dimension, but conventional DB systems may come close in some of the dimensions of table 3.1. In terms of crediting contributions, NDC systems do indeed take an extreme position: contributions are credited on a lifetime basis and earlier contributions get a higher weight according to the rate of interest. While the German point system also credits all lifetime contributions, the point system weights all contributions equally, independent of time. This corresponds to ignoring compound interest (see below). The French point system credits only the best 25 years. The U.S. social security system permits taking out the five worst years, which creates less labor supply disincentives than the French system. Many systems in developing countries use earnings only from the final year of work.13 In strictly Beveridgian systems such as the Dutch or British base pensions, credits do not play a role in the determination of benefits at all since benefits are flat.14 NDC systems feature a "natural" accrual of interest through the crediting mechanism. However, the conversion factor between benefits at retirement and the sum of contribu- tions over the working life can be interpreted as crediting all accrued interest at the time of retirement.15 While the latter mechanism ignores compound interest because it does not matter when contributions are paid over the working life, there is no other genuine differ- ence between NDC systems and a DB formula that (implicitly) credits the accrued interest at retirement. The determination of the rate of interest is another matter. In a funded sys- tem, the internal rate of return is r, the market rate of interest. In a PAYG system, it is n + g, where n is the rate of growth of the contribution base, and g is the growth rate of labor pro- ductivity. In Sweden, the government chose g as the notional rate of interest, leaving out a Table 3.1. Dimensions of Pension Systems 1. Credits for contributions Base: Life-long . . . . . . . . . Best X years . . . . . . . . . Final salary . . . . . . . . . Flat Weights: Early contributions earn interest . . . . . . . . . . . . . . . . . . Equal (point system) 2. Accrual of interest Rate: r (market) . . . . . . . . . . . . g (wages) . . . . . . . . . . . . n + g (Aaron-Samuelson) 3. Conversion to benefits Conversion: Linear (equivalence) . . . . . . . . . . . . . . . . . . . . . . . Concave (redistributive) Indexation: NDC: B = NPW/G . . . . . . . . . . . . . . . . . DB: B = f(credits, other; accrual rate) Actuarial: Neutral at retirement (at the margin) . . . . . . . . . . . . . . . . . . . . . . . Flat Risk: Benefits frozen at retirement . . . . . . . Indexation rules . . . . . . . Fully adjustable 4. Funding Extent: No fund at all . . . . . . . . . . . . . Reserve buffer . . . . . . . . . . . . . Fully funded Collateral: None . . . . . . . . . . Government bonds . . . . . . . . . . Commercial bonds/stocks Source: Author's compilations. 42 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES direct link of accruing interest to demography.16 In an aging population, n + g tends to be smaller than g since n is negative, and it is in most circumstances much smaller than r.17 The third dimension relates to the conversion of the accumulated notional wealth into benefits at and after retirement. There are four elements in that conversion. First, how are individual credits related to individual benefits? NDC systems use simple propor- tionality by applying the B = NPW/G rule. Some DB systems have the same propor- tionality, such as the point systems in France and Germany or systems with a simple accrual rate rule. The U.S. social security system credits additional contributions at a decreasing value; the system redistributes through a concave linkage function between contributions and benefits. In a strictly Beveridgian system, credits play no role at all, as mentioned before. This third dimension includes an important aspect of practical pension policy: namely, how strictly these rules are adhered to. The Swedish NDC and the German DB system attempt to avoid discretionary decisions. In both countries, the benefit rules are actually written into the law as mathematical formulae. So far, this attempt has been successful in Sweden, and also in Germany, with some notable exceptions between 1999 and 2001. Dis- cretionary deviations have taken place more often in the French point system. The Italian NDC system leaves ample room for discretionary adaptations to the political climate. The future has yet to show whether the political risk is smaller in NDC systems than in con- ventional DB systems. The second element in this third dimension is how benefits are related to the demo- graphic and macroeconomic environment. In NDC systems, this is expressed in the G-val- ues. Conventional DB systems usually have benefit indexation rules that link the benefits at retirement to the current wage or earnings level, and then continue to index benefits to prices, wages, or a combination of the two during the retirement period.18 They fail to include demographic factors directly, although there may be indirect linkages. Indexation to net wages, for instance, where net means net of taxes and contributions, entails an indi- rect linkage since rising contributions, precipitated by population aging, will also reduce benefits. There is nothing, however, to prevent DB formulae from including direct linkages to demography to increase the elasticity of benefits with respect to demographic changes. This study presents a concrete example below. The third element is the relation between retirement age and benefits. NDC systems are automatically actuarially neutral in the sense that the PDV of benefits is not affected by the actual retirement age if the PDV is computed at the internal rate of return of the NDC sys- tem. However, workers may use a different rate of interest when they are computing the PDV in order to make retirement decisions. This difference then creates a wedge between actuarial neutrality and the absence of labor supply disincentives. The latter is defined as the case in which the PDV of benefits is independent of the retirement age, if the PDV is computed using the personal discount rate of workers. The crucial question is then, what is the personal discount rate of workers? If it is the market rate of interest, then it is usually larger than the internal rate of return of the NDC system, especially in times of aging pop- ulations.19 This is an important point: NDC systems may be actuarially neutral, but they may still create substantial labor supply disincentives. The root cause for this distinction is the dif- ference in the discount rates that are applied to the actuarial adjustment. There are at least three candidates: (1) the internal rate of return, which is n + g in a PAYG system, indepen- dent of whether it is NDC or conventional DB; (2) the market rate of interest r, which is also the internal rate of return of a funded system; and (3) the rate of time preference of the employees who make the retirement decision. As noted, the market rate of interest r tends WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 43 to be higher than n + g. While one may argue that the average rate of time preference should be approximately equal to the market rate of interest, the empirical evidence shows that the workers' rate of time preference, guiding their retirement behavior, is even larger than r. Hence NDC systems may be more incentive-neutral, when they apply large discount rates, than DB systems, which apply very small adjustments of annual benefits to retire- ment age. For example, the adjustment in Germany is 3.6 percent per year phased in after 2001; see discussion below. NDC systems are certainly less distortionary than systems in which annual benefits are completely independent of retirement age (such as the German system before 2001). On the other hand, NDC systems may distort the retirement decision more than DB systems that apply rather large adjustments (such as the UK system, with adjustments of 9 percent per year, which is likely to be closer to the personal rate of time preference). NDC systems may provide a most "natural" way to compute retirement-age specific adjustments; but again, there is nothing intrinsic in this respect that distinguishes NDC from DB systems with actuarial neutral (or larger) adjustments such as the U.S. social security system at age 62 to 65. The fourth element is the indexation of benefits after retirement (different from the indexation, or anchoring, of benefits at retirement). Conventional DB systems index bene- fits after retirement to cost of living (United States, Italy's new system) or net wages (France, Germany) or some combination in between (Switzerland). In funded DC systems, this is represented by the type of annuity (real or nominal, or any other schedule). The Swedish NDC system freezes the benefits in real terms, corresponding to a cost-of-living indexation, but other NDC systems have other indexation mechanisms. Again, NDC sys- tems provide no special features in this respect. By now, it should be clear that NDC systems are accounting devices with properties that can be introduced in DB systems as well (although that may not be easy politically). They may serve to provide more transparency and credibility because some features come more "naturally" in an NDC system than through complex formulae in a DB system; how- ever, it is more rhetoric than economic substance that distinguishes NDC and DB systems. In fact, there are conditions under which NDC and DB schemes are mathematically equiv- alent, as Lindeman, Robalino, and Rutkowski show (2006). The discussion below presents the example of the new German DB system, which almost perfectly mimics an NDC system. Table 3.1 has shown the complexity by which pure PAYG systems may differ from one another, and the distinction between NDC sys- tems and "modern" DB systems is much smaller than between "modern" and "tradi- tional" DB systems. How is NDC different from a funded DC plan? This is addressed in the fourth dimen- sion of table 3.1. Most PAYG systems have some reserve fund, although the size of it varies greatly. The Swedish reserve fund amounts to some five years of expenditures, while the German system's reserve fund is minimal, with a few days worth of expenditures. The crucial question, however, is whether the accumulated balances are collateralized and which claims represent the collateral. NDC systems are "notional" in the sense that there is no collateral at all. Balances are claims against future taxpayers, and they are not backed by a financial instrument.20 Funded DC plans are usually understood as being collateral- ized against physical capital, mostly through financial instruments such as commercial bonds or stocks. We use the word "funded" only for these plans. Some authors also call those DC plans that are collateralized with government bonds "funded." We think that this is a misuse of the word "funded." Although benefit claims of such plans are mar- ketable and yield a well-defined rate of interest, they do not represent claims on physical 44 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES capital. NDC systems may permit a "natural" way to make the implicit debt of a PAYG system explicit by linking the NDC balances to government bonds, and the resulting insights about future benefits and contributions may create saving incentives on the microeconomic level.21 However, NDC systems and conventional DB systems share the crucial macroeconomic features of PAYG systems: NDC systems do not accumulate sav- ings in real assets with the potential beneficial side effects on the national saving rate, cap- ital market development, and growth.22 NDC systems are therefore no substitute for prefunding. Pros and Cons of Notional Defined Contribution Systems Now consider the chapter's second question: what do notional contribution systems bring to pension reform? What are their advantages and disadvantages? As the preceding sec- tions showed, NDC systems can almost perfectly be mimicked by conventional DB sys- tems. Thus this discussion focuses on the psychological and political aspects of a new rhetoric. This does not mean that this study ignores the microeconomic implications of transparency and credibility. This study will stress that they are important for pension reform. They are, however, sometimes quite subtle and indirect. Advantages An NDC system has many advantages: · It adapts itself automatically through an internal interest mechanism to the changed balance of contributors to pensioners (baby boom/baby bust problem) without the neces- sity to intervene in a discretionary way. · It adapts itself automatically to changed life expectancies (longevity problem) through the actuarial conversion of the notional pension wealth into a lifelong pension. Reductions for early retirement result automatically and are automatically adapted to the demographic situation. (This adaptation is incomplete when benefits are frozen at retirement and pensioners are effectively insulated against subsequent changes of longevity.) · It avoids arbitrariness of benefit indexation rules, adjustment factors, and so on, the change of which have undermined the credibility of many unfunded DB systems. · It strengthens the equivalence principle and for this reason minimizes the wedge between gross and net income, which results from the distortionary impact of taxes and tax-like contributions. · It adds transparency to the PAYG pillar by clearly identifying individual contributions and the resulting benefit claims, helping to regain credibility. · It strengthens the principle that pensions are based on lifelong earnings, and honors employees who enter the labor market early. · It allows transfer mechanisms to be easily identified as in-lieu contributions: notably tax-financed credits for higher and vocational education and similar credits for edu- cating children. · It creates a framework that can consistently be enlarged to a general "accounting sys- tem" of all PAYG subsystems. The advantages and disadvantages of joining subsys- tems (such as civil servants and the self-employed) will become immediately obvious in such an accounting system. · It produces a suitable framework for independent pensions of spouses. · It creates a homogeneous paradigm for the first, second, and third pillar of a multiple- pillar system; hence, it may increase the acceptance of the second and third pillar, WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 45 which are regarded as "alien" system components in those countries that used to have a monolithic first-pillar dominated pension system. · It permits a considerable amount of flexibility for employees in choosing their retire- ment age; makes the inflexible and politically problematic fixation of a "normal" retirement age superfluous; and exposes the trade-off between accumulated contri- butions and retirement age in an internally consistent fashion. · It permits easy portability of pension rights between jobs, occupations, and sectors. Disadvantages One of the main strengths of NDC systems, namely transparency, comes with some obvi- ous disadvantages in the political realm: · The financial situation of an unsustainable PAYG system becomes more obvious since workers "see" their declining benefits (while contribution rates are increasing) on their own accounts--thereby translating a general knowledge about the financial situation of the pension system into a personal concern. This is the flip side of the advantage of transparency: it may turn into a disadvantage because this may under- mine confidence even further. · If the contribution rate is fixed, the replacement rate becomes uncertain since it is dependent on the future development of earnings and demography. The replace- ment rate cannot serve any longer as a political instrument. This is of course the essence of a defined contribution system, including a notional one, but the uncer- tainty will become painfully visible in an NDC accounting system. · The system does not change the tension between declines in business cycle-related earnings and long-term spending commitments. Therefore, a genuine liquidity reserve is necessary. In countries where it has been depleted, it must be rebuilt, even if this means sacrificing lower contribution rates or installing higher ones, especially during business cycle upswings. Some countries, such as Sweden, have the historical gift of a buffer fund that is sufficiently large to mitigate even a substantial part of the demographic shock in the decades to come. · If the annuity is frozen at the beginning of retirement, a stabilizing feedback mecha- nism is missing if there is an unexpected rise in life expectancy. With a fixed contri- bution rate, the system will not automatically obey the annual budget restriction of a conventional PAYG system.23 In Sweden, a complex "automatic balancing mecha- nism" was introduced to overcome this problem.24 · Discretionary decisions are not absent; they are simply more hidden. They take place at the choice of life table, computation rules (such as the averaging) for the internal rate of return, the determination of a minimum retirement age, and so on. · The system does not change the fact that only prefunding can change which genera- tion pays for a given pension benefit. If one wants to have the workers of generation X at least partially pay for their own pension, rather than their children in generation X + 1, some extent of prefunding is necessary. An NDC system is no replacement for such partial funding. It is only an optimization of the PAYG pillar. Design Details This is not the place to discuss the many design issues that must be resolved in order to establish an NDC system.25 Only some especially important ones are discussed below: · Like any other PAYG system, a minimum age must be established when healthy work- ers may begin to receive a pension. Such a minimum retirement age is necessary unless one makes the participation in the PAYG system optional--which is not a sta- 46 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES ble solution, especially in times of population aging. None of the existing NDC sys- tems has an automatic adjustment of the minimum retirement age to longevity. Hence the existing systems respond only partially to demographic changes. · Specification and updating of life tables have immediate consequences for benefits. Most countries so far have adapted unisex life tables based on cross-sectional data.26 What is actuarially correct are cohort tables, with some projection for changes in life expectancy. Many countries do not have a reliable mechanism to compute such tables. An independent board of actuaries should supervise the construction of such tables, and the life tables should be consistent with those used in the private pension sector. · The internal interest of the contributions (growth rate of the contribution bill) must be smoothened over time. Several models are on hand. A moving average across a busi- ness cycle seems to be the natural choice. Peaks should be leveled off to flow into the liquidity reserve, used for the troughs of a cycle. Designing such smoothening mech- anisms is not trivial since the amplitude and duration of a business cycle is not known in advance, and there is no foolproof mechanism to distinguish trends from cycles. · Since the internal rate of return of a PAYG system is in general substantially lower than the workers' discount rates, incentive effects to retire early may still be large if later pension uptake is governed by actuarial neutrality. From a macroeconomic point of view, it is not obvious whether actuarial neutrality (using the internal rate of return of the PAYG system) or absence of incentive effects (using the workers' discount rate) is welfare maximizing. · There are many ways to design the annuities. They can be chosen to rise with inflation or with wages, or any other schedule; the initial level is adjusted accordingly, hold- ing PDV over expected duration constant. The freedom to choose is probably more important than potential problems with adverse selection. However, there is little experience around the world so far with such trade-offs. · Benefits are determined at the beginning of retirement, but the demographic and economic environment may unexpectedly change after such determination. If the growth rate of contributions, measured in nominal terms, is very low, pensions may fall in real terms, which is politically unattractive. Some adaptation to the new envi- ronment is probably optimal, but pensioners seem to value protection quite highly. The overall welfare-maximizing policy is not known so far. As emphasized earlier, if pensioners are fully protected from demographic and economic changes taking place after they entered retirement, NDC systems lose an important feedback mechanism, which may undermine long-run financial sustainability. If pensions are protected by a floor (for example, never to fall in real or nominal terms), the system can become financially unbalanced because expenditures and contributions are treated asymmet- rically: Expenditures cannot fall below a certain floor, while contributions can. · The rules for the transition period should follow the Swedish model, not the Italian one.27 The extremely lengthy Italian transition time does not permit the NDC sys- tem to address the most urgent financial problem: the strain on the pension bud- get that will occur when the baby boom generation retires. A more difficult question is the extent to which existing pensions should be protected. "Natural" transition paths can be constructed when benefits are defined pro rata in propor- tion to the time spent under current and new law. The transition can be faster if important design elements of an NDC system are already in place. In Germany, for instance, a relatively short transition period (from 2005 through 2020) would WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 47 be possible because the existing point system has an important feature of NDC systems already in place. A Blend of DB and NDC: The German Point System and the Sustainability Indexation This final section uses the proposed new German public pension system as an example of a DB system that quite closely mimics an NDC system. There are three main elements of the German public pension system: the point system of credits, the actuarial adjustments, and the benefit indexation formula. During the recent pension reform process,28 these ele- ments have been changed consistently in the direction of NDC-type pensions. The point system, described in detail further below, has been a feature of the German pension system since its conversion to a PAYG system in 1957. Actuarial adjustments were legislated by the 1992 reform; they have been phased in since 1997, with the bulk of adjustments in force by 2007. As a result, effective retirement ages are expected to increase by about two years within the next ten years.29 In 2001, the so-called Riester reform made a first step from a purely pay-as-you-go to a capital-funded pension sys- tem. It established upper limits to the contribution rate, discontinued the benefits indexation formula, and substantially lowered pension levels. At the same time, the Riester reform introduced state-subsidized supplementary private pensions ("Riester pensions") to fill the upcoming pension gap. However, these reforms did not suffice to stabilize public pension finances. In late fall of 2002, the government established a reform commission to achieve "sustainability in financing the social insurance sys- tems." A broad majority of this commission took the view that the upper limit of the contribution rate legally anchored by the Riester reform must be adhered to, and conse- quently changed the benefit indexation formula to follow an income-oriented policy. This change has been approved by the government and passed the Bundestag in first read- ing in fall 2003. Details of this new indexation mechanism are described below. This was the last step that in effect turned the German DB system into an NDC-type pension system. The German Point System The German public pension system computes benefits according to the following for- mula: Bt,i = PVt * EPi * AAi where Bt,i = Benefits of pensioner i in year t, PVt = Current pension value in year t, EPi = Number of individual earnings points collected by pensioner i until his retirement, and AAi = Actuarial adjustment, dependent on the retirement age of pensioner i. Benefits therefore have a simple structure: an individual component EPi * AAi determined by each person's earnings history and retirement age, which stays fixed for the entire retirement period, and an aggregate component PVt, which adjusts benefits over time equally for all pensioners. 48 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES EPi represents the "point system" and AAi is determined by actuarial accounting rules (see below). A typical worker who works for 40 years and who earns the average labor income in each of these 40 years receives 40 earnings points. If this worker retires at age 65, no actuarial adjustments take place (AA = 1). In the second half of 2002, the current pen- sion value PVt was 25.86 euros.30 Hence this typical worker receives a pension of 1,034.40 euros per month. A worker who has worked for 20 years at average earnings, or a worker who has worked for 40 years at 50 percent of average earnings, will receive half these pen- sion benefits, while workers who earn twice the average labor income for 40 years will receive twice as much as the 40-year average earner. The Actuarial Adjustments Before 1992, adjustment of benefits to retirement age was only implicit via additional earn- ings points. For a worker with 40 years of service at the average earnings level, an addi- tional year of service would therefore increase the annual pension benefit by 2.5 percent. There were no further actuarial adjustments. The 1992 reform changed this, and the changes are currently being phased in.31 Age 65 will then act as the "pivotal age" for bene- fit computations. Benefits will be reduced by 3.6 percent (maximum 10.8 percent) for each year of earlier retirement. The 1992 reform also introduced rewards for later retirement: for each year of postponement, benefits increase by 6 percent. There is some debate whether these percentage adjustments are actuarially neutral.32 The German Reform Commission took a cautious position by stating that "the adjustments are low, but probably sufficiently close to actuarial neutral."33 At conventional discount rates, they still exert considerable negative incentives to retire early.34 The Benefit Indexation Formula Each year, currently on July 1, the current pension value PVt is recalculated with the aid of the benefit indexation formula. Until recently, this benefit indexation formula was essen- tially a simple indexation rule, increasing pensions by the growth rate of net earnings: PVt = PVt *ANWt -1 -1ANWt -2 where PVt = Current pension value in year t, and ANW = Average earnings of all employees subject to compulsory insurance, net of taxes and social insurance contributions. Since the current pension value PVt has a direct influence on every individual pension, the benefit indexation formula is a critical determinant for the well-being of pensioners and the amount of money spent by the public pension scheme. However, the benefit indexation formula gives no direct reference to the demographics of the system nor to the number of employed persons, although there is a feedback through net earnings. Higher contributions dampen net earnings, and thus pension benefits, incorporating a kind of burden-sharing between generations. The limitations of this feedback mechanism are one reason for the unsustainability of the German pay-as-you-go system. Starting in 2005, the benefit indexation formula will be augmented by a "sustainabil- ity factor" that incorporates demography and employment into the German benefit indexation formula. Specifically, it also indexes benefits to the numbers of contributors and pensioners. The relative number of contributors to pensioners, the so-called system WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 49 dependency ratio (PQ), is the most important long-term determinant of pension financ- ing: PVt = PVt * ANWt -1*1- -1 -1 ANWt PQt PQt * -2 -2 +1 where PVt = Current pension value in year t, ANW = Average earnings net of public and private pension contributions,35 and PQ = System dependency ratio [pensioners/(contributors + unemployed)]. Incorporating the sustainability factor in the benefit indexation formula links annual increases in pensions to productivity growth and the growth of the contribution base. The weighting factor gives weight to each of these two determinants; it effectively spreads the financial burden among contributors and pensioners. If equals zero, the current ben- efit indexation formula would remain unchanged and the financial burden generated by a higher proportion of pensioners in the population would mainly be shouldered by the labor force. The condition equals one implies a purely receipts-oriented pension expen- diture policy. The German Reform Commission has set to 1/4 to target the contribution rates anchored in the Riester reform (contribution rate below 20 percent until 2020, and under 22 percent until 2030). The NDC Characteristics of the German Pension System While there is no perfect equivalence, the three main elements of the German public pen- sion system (the point system of credits, the actuarial adjustments, and the indexation to both earnings and system dependency) mimic the essential features of an NDC system.36 The point system includes all earnings over the lifetime; almost all redistributive features of the German old-age pension system take the form of earnings points credited without actual earnings (for example, years of unemployment, years of higher education, years of educating a child, in each case valued at an imputed earnings level). The equivalence between the point system and an NDC system is not perfect since all earnings points count equally in the German point system, while in an NDC system the earnings necessary to gain one such earnings point are valued more when earned earlier in life, due to com- pound interest.37 One deviation from equivalence is the fact that the actuarial adjustments in the German system are not directly linked to life expectancy. The actual adjustment rates are somewhat arbitrary, probably too small, and certainly controversial.38 One of the advantages of an NDC system is to automatically generate the adjustment rates by the annuitization mech- anism. The benefit indexation formula of the German DB system, with its indexation to earnings growth and changes in the system dependency ratio, approximates the effect of the accu- mulated interest in an NDC system, in which the internal rate of interest is the growth rate of the contribution bill (1 + n)*(1 + g). This is easy to see in the most stylized case when all contributions (normalized to one unit) are credited upfront. In this stylized NDC case, the notional pension wealth after T years is T*(1 + n)T*(1 + g)T. The pension benefit is therefore P = T*(1 + n)T*(1 + g)T/G where G denotes the annuity factor (or "G-value"). In the German DB system, this average worker earns T earnings points. During these T years, the average pension value PV will increase with the rate of wage growth (g) and the growth rate of the dependency ratio (n, 50 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES if the number of pensioners remains constant): PVT = PV0 *(1 + n)T*(1 + g)T. Hence, the pen- sion benefit is P = T*PV0*(1+n)T*(1+g)T, proportional to the NDC value. This stylized comparison may ignore many differences in detail. Nonconstant contribu- tions will have a differential impact on the two formulae; net wages are likely to grow at a different rate than gross earnings; the system dependency ratio is likely to shrink faster than the labor force. The principles, however, are the same. What a country prefers as a pension policy--an almost textbook-like NDC system as in Sweden or a demography- indexed DB system as in Germany--is probably more dependent on the historical path of a country and the specific circumstances of the political debate than on abstract economics. Conclusions: Can NDC Systems Master the Challenges? This summary starts with three claims about NDC systems that are not true. NDC systems are not automatically balancing: they do not automatically fulfill the PAYG budget constraint when economic parameters change. In particular, automatic balancing will not hold if annuities are frozen at retirement and the contribution rate is fixed since there is no feed- back mechanism if longevity of current pensioners increases unexpectedly. Second, an NDC system is not automatically sustainable unless the contribution rate is fixed and the rate of return equals the contribution bill (or the system follows an equivalent trajectory). Third, an NDC system is no substitute for prefunding. An NDC system does not change the basic PAYG mechanism in which the children pay for the pensions of their parents, and it does not create savings unless it generates a benefit cut, which in turn precipitates savings. However, if correctly designed, an NDC system will automatically respond to changes in the demographic and macroeconomic environment because benefits are indexed to longevity (due to the annuitization mechanism), fertility, and employment (through the notional rate of interest, if indexed to the contribution bill). Moreover, an NDC system has potentially important microeconomic effects. It will cre- ate a sense of actuarial fairness (because annual benefits are in line with lifetime contribu- tions) and actuarial neutrality (because the system creates automatic adjustments to retirement age). It exposes redistribution because any noncontributory credits appear clearly marked on the account statements. An NDC system changes the rhetoric of pension systems. It makes people think in terms of accounts rather than entitlements and thus may make the transition to partial funding psychologically easier. Moreover, by exposing the dwindling balance of first pillar pen- sions, it may actually create incentives to save in the second and third pillar. An NDC sys- tem makes workers and administrators think in terms of "pension wealth," which may ease portability both within a country and between countries. It enables interpersonal transfers (for example, between husband and spouse) and eases replacement of survivor pensions by independent pension claims. An NDC system also takes certain issues out of the political agenda, potentially easing reform. It minimizes the role of the "normal retirement age" and permits a more flexible choice between consumption (work longer) and leisure (get lower replacement rate). This flexibility finds its limitations in the conflict between actuarial neutrality and absence of labor supply disincentives and the necessity to establish an early retirement age. NDC sys- tems also create a new set of "parameters," another aspect of the new rhetoric, which may make reform more palatable. It permits redefinitions and readjustments and changes the focus of debate from parametric reform to the introduction of "a new system" (while this is not the case with respect to economic substance), thus enabling parameter change. This point is not without some irony, and even more so, because we claimed that NDC systems WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 51 strengthen credibility through transparency, while we now use it as a device to deflect emotional opposition by using a new rhetoric. The usage as a rhetorical and psychological device, however, should not be belittled, and insights among workers and pensioners precipitated by a new rhetoric may have real economic effects. By exposing the economics of a pay-as-you-go system, by visualizing the budget constraint of a pension system, and by making the trade-off between retirement age and replacement rate concrete, NDC systems may induce economic reactions, such as later retirement or higher saving rates. In addition, the new rhetoric may help to get a pension reform process going because it provides a framework to introduce actuarial adjustments (since they come "automati- cally"), a framework to diffuse the explosiveness of changes in the retirement age (since a flexible choice of retirement age minimizes opposition), and a framework to change inter- generational redistribution in a genuine sense if, and only if, NDC systems make workers save more. Coming back to the challenges posed in the second part of this study, NDC systems are well positioned to manage the challenge of longevity. They are also well set up to react to slow changes in fertility, if the internal rate of interest is properly defined as the growth rate of the contribution bill. Sudden changes of fertility, however, such as the sudden baby boom/baby bust transition, are not well manageable in NDC systems. Countries in which the younger generation is simply overwhelmed by the financial burden of pensions need prefunding, enabling the members of the older generation to carry part of the pension bur- den themselves. Design flaws of current DB systems (such as labor supply disincentives) are relatively well manageable in NDC systems; changing the rhetoric may be instrumental here. We have, however, already stressed the important distinction between absence of labor sup- ply effects and actuarial neutrality. NDC systems are only indirectly devices to foster growth, savings, and improvements in capital market performance. Since NDC systems are still PAYG devices, they do not alter the macroeconomic mechanisms at all; by changing the microeconomic incentives, at least by psychological means, they precipitate substantial real effects after all--through later retirement and higher savings. To alter the growth path of an economy, NDC systems therefore must be coupled with a strengthening of second and third pillar pensions. This combination of NDC with prefunding looks like one of the most fruitful paths of pension reform. Notes 1. "Notional defined contribution" and "non-financial defined contribution" should be understood to have the same definition. 2. See IW (2003). 3. See Gruber and Wise (1999); Börsch-Supan and Schnabel (1998, 1999); and Börsch- Supan (2000). 4. See Börsch-Supan, Kohnz, and Schnabel (2003); Börsch-Supan, Schnabel, Kohnz, and Mastrobuoni (2003). 5. See Palmer (2000); Chlo´n, Góra, and Rutkowski (1998). 6. See Disney (1999); Valdés-Prieto (2000) 7. See Rutkowski (1998); Góra and Rutkowski (1998); Fox and Palmer (1999); Chlo´n- Domi´nczak and Góra (2006). 8. Valdés-Prieto (2003) suggests a potential trading mechanism. 52 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES 9. In Sweden, the G-value is the remaining unisexual life expectancy at retirement age, reduced by the effect of compound interest during retirement. The assumed interest rate is 1.6 percent. This is modified if the growth of contributions is unusually low, by the "auto- matic balancing mechanism." For details. see Settergren (2001). 10. See Legros (2006). 11. See Gruber and Wise (1999). 12. See the textbooks by Homburg (1988) and Valdés-Prieto (1998). 13. See Börsch-Supan, Palacios, and Tumberello (1999). 14. This is strictly true only in citizenship- or residence-based flat benefit systems, such as in Australia. In the United Kingdom, a minimum number of credits determines eligibil- ity. 15. See Ruland (2000). 16. Demography enters directly through longevity-dependent annuities and indirectly through feedbacks that change g, such as through age-specific productivity. Sweden also has a rebalancing mechanism that will respond to demography once the current system fails (see Settergren 2001). 17. The theoretical relation between r and n+g in and out of steady state and its relation to dynamic efficiency fills volumes and is not the subject of this study. See Valdés-Prieto (1998). For some enlightening empirical results, see Schnabel (1998). 18. Note this distinction between the initial indexation at retirement for the flow of new entrants and the indexation of benefits after retirement for the stock of existing retirees. 19. Empirically measured personal discount rates tend be even larger than r, and by a substantial margin. See the survey by Frederick, Loewenstein, and O'Donoghue (2002). 20. This could be changed by converting the implicit claims on the contributions of future workers in explicit claims on future taxes backed by government bonds (see Valdés- Prieto 2003). 21. See Góra and Palmer (2003). 22. See Holzmann (1997); Schmidt-Hebbel (1998); Börsch-Supan and Winter (2001). 23. For a clearly stated mathematical exposition of this point, see Valdés-Prieto (2000). 24. A detailed description can be found in Settergreen (2001). 25. See Palmer (2003) for many of such details. 26. An exception is Latvia, which applies a cohort life-table. 27. The transition rules in Latvia and Poland are similar to the Swedish rules, while Kyr- gyzstan's transition rules are similar to Italy's. 28. Börsch-Supan and Wilke (2003) present a detailed description of the pension reform process. 29. Econometric estimates are provided in Berkel and Börsch-Supan (2003). 30. This value was determined by the goal to provide a 70 percent ratio between the average pension and the average earnings level (see discussion below). 31. See Berkel and Börsch-Supan (2003) for details of the transition process. 32. There is a controversial discussion about the correct actuarial adjustment rate in the German public pension system. See Ohsmann, Stolz, and Thiede (2003) vs. Börsch-Supan (2000). 33. Kommission für die Nachhaltigkeit in der Finanzierung der Sozialen Sicherungssys- teme (2003). 34. Börsch-Supan (2000); Börsch-Supan and Schnabel (1998, 1999); Börsch-Supan, Schn- abel, Kohnz, and Mastrobuoni (2003); Börsch-Supan, Kohnz, and Schnabel (2003). 35. The careful reader may note the difference in the ANW definitions. We choose to ignore these here since we want to focus on the NDC characteristics. The current formula does not net out taxes and social insurance contributions other than taxes, but does net out WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 53 contributions to second and third pillar pensions at an imputed rate. It also computes the system dependency ratio using full-time equivalents for workers and pensioners. For these computational details, see Börsch-Supan and Wilke (2003). 36. Ruland (2000) has succinctly expressed the relationship between earnings points and current pension value by regarding earnings points as "shares" in the "financial assets" held by the pension fund. The current pension value corresponds to the current "share price." 37. Note that the metric is earnings points. If the metric is euros, the relation is reverted. 38. See the review in Berkel and Börsch-Supan (2003). References Berkel, B., and A. Börsch-Supan. 2003. "Pension Reform in Germany: The Impact on Retirement Decisions." MEA Discussion Paper 31-03, Mannheim Research Institute for the Economics of Aging, Mannheim University, and NBER Working Paper 9913, National Bureau of Economic Research, Cambridge, Mass. Börsch-Supan, A. 2000. "Incentive Effects of Social Security on Labour Force Participation: Evidence in Germany and Across Europe." Journal of Public Economics 78: 25­49. Börsch-Supan, A., and M. Miegel, eds. 2001. Pension Reform in Six Countries. Heidelberg, New York, Tokyo: Springer. Börsch-Supan, A., and R. Schnabel. 1998. Social Security and Declining Labor Force Partic- ipation in Germany. American Economic Review 88 (2): 173­178. ------. 1999. "Social Security and Retirement in Germany." In International Social Security Comparions, ed. J. Gruber and D.A. Wise, Chicago: University of Chicago Press. Börsch-Supan, A., and C. Wilke. 2003. "The German Social Security System: How it Was and How it Will Be." MEA-Discussion Paper prepared for the MRRC-Conference on Social Security, May 2003, Washington, DC. Börsch-Supan, A. and J. Winter. 2001. "Population Aging, Savings Behavior, and Capital Markets." Working Paper 8561, National Bureau of Economic Research (NBER), Cam- bridge, Mass. Börsch-Supan, A., R. Palacios, and P. Tumberello. 1999. Pension Systems in the Middle East and North Africa: A Window of Opportunity. Pension Reform Primer. Washington, DC: World Bank. Börsch-Supan, A., S. Kohnz, and R. Schnabel. 2003. "Micro Modeling of Retirement Choices in Germany." In Incentive Effects of Public Pension Systems, ed. J. Gruber and D. Wise. Chicago: University of Chicago Press, in press. Börsch-Supan, A., R. Schnabel, S. Kohnz, and G. Mastrobuoni. 2003. "Budget Effects of Pension Reform in Germany." In Budget Effects of Reforming Public Pension Systems, ed. J. Gruber and D. Wise. Chicago: University of Chicago Press, in press. Chlo´n, A., M. Góra, and M. Rutkowski. 1999. Shaping Pension Reform in Poland: Security through Diversity. Social Protection Discussion Paper 9923, World Bank, Washington, DC. Chlo´n-Dominczak, A., and M. Góra. 2006. "The NDC System in Poland: Assessment after ´ Five Years." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribu- tion (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 16. Washington, DC: World Bank. Disney, R. 1999. "Notional Accounts as a Pension Reform Strategy: An Evaluation." Social Protection Discussion Paper 9928, World Bank, Washington, DC. ------. 2003. "Are Contributions to Public Pension Programmes a Tax?" Paper prepared for the Fourth Annual CeRP Conference, Turin, September 16. 54 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Fox, L., and E. Palmer. 1999. "Latvian Pension Reform." Social Protection Discussion Paper 9922, World Bank, Washington, DC. Franco, D., and N. Sartor. 2006. "NDCs in Italy: Unsatisfactory Present, Uncertain Future." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 18. Washington, DC: World Bank. Frederick, S., G. Loewenstein, and T. O'Donoghue. 2002. "Time Discounting and Time Preference: A Critical Review." Journal of Economic Literature 40: 351­401. Góra, M., and E. Palmer. 2003. "Shifting Perspectives in Pensions." Revised version of working paper published by CASE in Warsaw, 2002. National Social Insurance Board, Stockholm. Góra, M., and M. Rutkowski. 1998. The Quest for Pension Reform: Poland's Security through Diversity. Social Protection Discussion Paper 9815, World Bank, Washington, DC. Gruber, J., and D.A. Wise, eds. 1999. Social Security and Retirement Around the World. Chicago: University of Chicago Press. Holzmann, R. 1997. "Pension Reform, Financial Market Development, and Economic Growth: Preliminary Evidence from Chile." IMF Staff Papers 44 (2): 149­78. Homburg, S. 1988. Theorie der Alterssicherung. Berlin, Heidelberg, New York: Springer. IW (Institut der deutschen Wirtschaft). 2003. Deutschland in Zahlen. Köln: Deutscher Insti- tuts Verlag. Kommission für die Nachhaltigkeit in der Finanzierung der Sozialen Sicherungssysteme. 2003. Abschlußbericht. Bundesministerium für Gesundheit und Soziale Sicherheit. Berlin. http://www.bmgs.bund.de/deu/gra/themen/sicherheit/kommission/index .cfm. Legros, F. 2006. "NDCs: A Comparison of the French and the German Point Systems." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 10. Washington, DC: World Bank. Lindeman, D., D. Robalino, and M. Rutkowski. 2006. "NDC Pension Schemes in Middle- and Low-Income Countries." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 12. Wash- ington, DC: World Bank. OECD (Organisation for Economic Co-operation and Development). 1988. Ageing Popula- tions: Social Policy Implications. Paris: OECD. Ohsmann, S., U. Stolz, and R. Thiede. 2003. Rentenabschläge bei vorgezogenem Rentenbe- ginn: Welche Abschlagssätze sind "richtig"? Die Angestelltenversicherung Heft 4, 119­124. Palmer, E. 2000. "The Swedish Pension Reform Model: Framework and Issues, Social Pro- tection." Discussion Paper 0012, World Bank, Washington, DC. Palmer, E. 2003. "Conversion to NDC--Isssues and Models." Paper presented at World Bank and Swedish Social Insurance Agency Conference on NDC Pensions, Sandhamn, Sweden, September 28­30. Ruland, F. 2000. Neugestaltung der Rentenformel im bestehenden Umlageverfahren. Vortrag bei Speyerer Sozialrechtsgesprächen. www.vdr.de. Rutkowski, M. 1998. "A New Generation of Pension Reforms Conquers the East: A Taxon- omy in Transition Economies." Transition 9 (4, August): 54­76. Schmidt-Hebbel, K. 1998. "Does Pension Reform Really Spur Productivity, Saving, and Growth." Working Paper 33, Central Bank of Chile, Santiago. Schnabel, R. 1998. "Rates of Return of the German Pay-As-You-Go Pension System." Finanzarchiv 55 (3): 374­99. WHAT ARE NDC SYSTEMS? WHAT DO THEY BRING TO REFORM STRATEGIES? 55 Settergren, O. 2001. "The Automatic Balance Mechanism of the Swedish Pension System." Wirtschaftspolitische Blätter 2001/4. Valdés-Prieto, S. ed. 1998. The Economics of Pensions: Principals, Policies, and International Experience. Cambridge: Cambridge University Press. ------. 2000. "The Financial Stability of Notional Account Pensions." Scandinavian Journal of Economics, 102, 385­387. World Bank. 1994. Averting the Old-Age Crisis: Policies to Protect the Old and Promote Growth. New York: Oxford University Press. Chapter 4 Non-Financial Defined Contribution Pensions: Mapping the Terrain Nicholas Barr* The Backdrop This chapter offers an assessment of notional defined contribution (NDC) pensions from the perspective of welfare economics, with a brief discussion of management and imple- mentation.1 The opening section sets out the objectives of pension schemes, the idea of NDCs, and the simple economics of pensions. The second section assesses NDC pensions in terms of policy design. Is the aspect under discussion an advantage? If so, is the advan- tage specific to NDCs, and is it one of policy design or political reality? The third section briefly discusses the institutional prerequisites necessary if a country is to introduce NDC pensions effectively. The final section offers three conclusions. NDC pensions are not a theoretically dominant policy, but one with pros and cons that need to be compared with those of other designs. The case for NDCs is strongest as part of a wider pension system, rather than as a stand-alone construct. The NDC approach does little to address the cen- tral problem of pension finance--the age at which people are first eligible to claim their pension. Objectives of Pension Schemes From the viewpoint of the individual, pensions have two purposes: · Consumption smoothing over the life cycle, and · Insurance, notably in respect of the longevity risk. Public policy has additional objectives: · Poverty relief. This is necessary if someone is poor for his or her lifetime as a whole and, as a practical matter, is needed to address transient poverty. · Distributional objectives. Governments may also have broader distributional objec- tives. They frequently wish to protect the pension rights of people with caring responsibilities and they may wish to subsidize the consumption smoothing of peo- ple whose earnings are above the poverty line, but not by much. * Nicholas Barr is professor of public economics at the London School of Economics and the author of numerous books and articles, including The Economics of the Welfare State (OUP, 4th edition, 2004). 57 58 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES These four elements are the primary objectives of pensions. The analysis, particularly in the second section, considers the extent to which different types of NDC arrangements do, or do not, contribute to their achievement. There is also an important constraint--sustain- ability--that recurs in the discussion below. NDC Pensions The basic idea of NDC pensions is to separate the state pay-as-you-go (PAYG) scheme into two components: a strictly actuarial element (the NDC pension), operating on a PAYG basis but mimicking a funded defined contribution scheme; and a redistributive element financed from general taxation.2 The actuarial element is calculated in the following way: · A contribution of x percent of a person's earnings is credited to a notional individual account: that is, the state "pretends" that there is an accumulation of financial assets. · The cumulative contents of the account are credited periodically with a notional interest rate. · At retirement, the notional account is converted into an annuity. Thus NDC pensions mimic conventional (funded) defined contribution schemes by paying an income stream whose present value over the person's expected remaining life- time equals his/her accumulation at retirement. QUESTIONS ABOUT THE DESIGN OF STATE PENSIONS Policy makers face a series of issues. How large should the state pension be; how redis- tributive from richer to poorer; and should benefits be defined contribution or defined benefit (defined and discussed shortly)? Further, if a new pension system, such as an NDC arrangement, is introduced, how will the transition be financed? Moreover, there is a range of questions specific to the design of NDC pensions: Question 1. What is the minimum pension? Is it paid in addition to the NDC pension, or does it take the form of a guarantee that is paid only if the NDC pension falls below a predetermined minimum level? If the former, is the minimum flat rate or with an earnings-related element? Is it minimal or larger? Question 2. Is there a maximum NDC pension? Question 3. Is minimum pensionable age unconstrained, with full actuarial adjustment of the pension to a person's age at retirement? Or is there a legally defined minimum age, with actuarial adjustment for retirement at a later age? And/or is there a minimum age that rises according to some functional relationship with life expectancy? Question 4. Is the accrual rate during working life based on earnings growth per worker, and hence unaffected by unemployment, or on earnings growth in aggregate, and hence lower in years when unemployment is higher? Separately, is the pension formula adjusted for life expectancy? Several NDC schemes have an accrual rate equal to Rate of growth of = productivity growth + employment growth. the contributions base Most schemes, being new, include an adjustment for life expectancy. NON-FINANCIAL DEFINED CONTRIBUTION PENSIONS: MAPPING THE TERRAIN 59 Question 5. Is the annuity, once in payment, adjusted annually in line with changes in consumer prices, or with real wage growth? DEFINED CONTRIBUTION AND DEFINED BENEFIT PENSIONS: A BRIEF COMPARISON In a defined contribution (DC) scheme, a person's pension is an annuity whose size, given life expectancy and the rate of return on pension assets, is determined only by the size of his lifetime pension accumulation. Thus the approach leaves the individual to face the risk that his pension portfolio might perform badly. Under a defined benefit (DB) scheme, often run at the firm or industry level, a person's pension is based on his wage and length of service. Thus his annuity is wage-indexed during working life, and the risk of varying rates of return to pension assets falls on the employer, and hence on some combination of the industry's current workers (through effects on wage rates), its shareholders and the taxpayer (through effects on profits), its customers (through effects on prices), and/or its past or future workers, if the company uses surpluses from some periods to boost pen- sions in others. The debate between DC and DB pensions is often posed as one between polar oppo- sites, a strictly actuarial DC scheme being compared with a final salary DB scheme. The reality is more subtle, as Diamond (2002, 55­57) points out. Suppose a person's earnings in a particular year are 70 percent of average earnings that year; call that variable x. Call the average value of x over n years,X, which is thus a measure of the person's earnings each year indexed by the rate of wage growth. ThenX is the earnings base on which a person's pension in a DB scheme is determined. If n relates to earnings in the last year before retire- ment, the scheme is a final-salary DB scheme. In contrast, if n spans an entire working life, the scheme is a DB scheme in which pensions are based on lifetime contributions com- pounded each year by the rate of wage growth. In a funded DC scheme, annual contribu- tions are compounded by the return on assets (for short, the interest rate), again over a person's working life. If the rate of interest and the rate of wage growth are similar, the dif- ference between DC and DB is minor; and the difference is even smaller between a lifetime DB scheme and an NDC scheme with an accrual rate equal to wage growth. In the limit, suppose that a DB scheme bases benefits on a person's entire working life; has an accrual rate that is age-related (that is, contributions in early years have a heavier weight, as with compound interest); and offers an annuity rate that is announced only at the time that a person retires. In that case, DB and DC converge. Thus DB schemes can be very different, and hence have very different economic impacts; the same is true of NDC regimes. The analysis that follows tries to be clear which type of scheme is being compared with which. The Simple Economics of Pensions A final piece of background before turning to specific assessment of NDC pensions is the simple economics of pensions, which can be confusing because the literature tends to focus on financial aspects such as analysis of portfolios of financial assets. This discussion tries to simplify matters by concentrating on the essential economic issues: namely, the production and consumption of goods and services. There are two (and only two) ways of seeking security in old age.3 It is possible, first, to store current production by storing part of current output for future use. Though this is the only way that an individual such as Robinson Crusoe could guarantee consumption in retirement, the method in practice has major inefficiencies. It is costly; it does not deal with uncertainty-- for example, how one's tastes or constraints might change; and it cannot be applied to services deriving from human capital, medical services being a particularly 60 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES important example. With few exceptions, organizing pensions by storing current produc- tion on a large scale is therefore a nonstarter. The alternative is for individuals to exchange current production for a claim on future production. There are two broad ways in which a worker might do this: by saving part of his wages each week, he could build up a pile of money that he would exchange for goods produced by younger people after his retirement; or he could obtain a promise--from his children, or from government--that he would be given goods produced by others after his retirement. The two most common ways of organizing pensions broadly parallel these two sorts of claim on future output. Funded schemes are based on accumulations of financial assets, PAYG schemes, including NDC schemes, on promises. Given the deficiencies of storing current production, the only way forward is through claims on future production. Thus the central variable is the level of output after a worker has retired. The point is fundamental: pensioners are not interested in money (colored bits of paper with portraits of national heroes on them), but in consumption--food, heating, medical services, seats at football matches, and so on. Money is irrelevant unless the pro- duction is there for pensioners to buy. The discussion thus far suggests a series of propositions against which an NDC (or any other) pension scheme should be assessed. · From the point of view of sustainability, the central variable is the level of national output, not the specific method by which pensions are financed. Since NDC pensions per se do not increase output, the main contribution to sustainability is if they facili- tate lower pensions, later retirement, and/or an increase in contributions. · The design of the state scheme matters. If the state scheme is unsustainable, the only solution is to fix the state scheme. Thus a move to NDC pensions may make it possi- ble to reduce pensions to sustainable levels and/or to make it possible to remove or reduce special benefits for particular groups. · Insurance, consumption smoothing, and poverty relief are all important. A pure NDC pension is concerned only with insurance and consumption smoothing, with implications, which this study explores, for the broader design of a pension system. Assessing NDC Pensions In assessing NDCs, it is useful to distinguish different questions: · Is the particular feature under discussion an advantage? · Is the advantage inherent in the NDC design or could it be achieved by another arrangement? · Is the advantage one of policy design or of political reality? A final question, discussed in the concluding section, is whether NDC pensions are a dominant policy. Advantages This section assesses a series of advantages that are claimed for NDC pensions. ASSIST CLEAR THINKING It is useful to distinguish three elements of pensions that are often conflated: public or pri- vate; PAYG or funded; DC or DB. The NDC approach reminds us that public + PAYG + DC is a possible option: in other words, that PAYG does not automatically mean DB. This reminder is helpful. NON-FINANCIAL DEFINED CONTRIBUTION PENSIONS: MAPPING THE TERRAIN 61 FACILITATE DESIRABLE FEATURES OF PENSION DESIGN NDC pensions facilitate a number of desirable design features. A flexible retirement age is welfare-improving because it increases the range of individual choice over consumption smoothing. This advantage can, however, be a feature of other types of schemes. In terms of the retirement decision, what is needed is an actuarial rela- tionship between contributions and pensions at the margin, but not necessarily across the entire contributions record.4 A flexible combination of work and retirement, again, increases individual choice between both work and leisure and over the time path of income. Again, however, this is possible with other pension arrangements, such as a state scheme offering defined benefits from the age of 65 but with actuarial adjustment for delayed retirement and options for com- bining work with receipt of pension. Automatic adjustment to rising life expectancy. Given the sustained increase in life expectancy, this feature of NDC pensions is essential for long-run viability. But it could equally be a feature of other pension arrangements: for example, if the age at which the pension is first payable rises according to an explicit relation with life expectancy. If NDC pensions have any advantage in this respect, it is that the politics of adjustment might be easier, rather than a design feature possible only with NDCs. ENHANCE THE ABILITY TO COPE WITH RISK AND UNCERTAINTY Risk and uncertainty lower the welfare of risk-averse individuals (proof: the amount that people spend on voluntary insurance). Consumption smoothing is thus more efficient if people can be protected from excessive risk and uncertainty. The distinction is important: with risk, the probability of the occurrence of the insured event is known; with uncer- tainty, it is not. Risks can be covered by actuarial insurance. With uncertainty, in contrast, ignorance of the underlying probability distribution makes it difficult or impossible to assess an actuarial premium; hence uncertainties are generally covered badly, if at all, by actuarial insurance. In the case of pensions, the estimates of life expectancy have a suffi- ciently small variance that annuities are possible. With inflation, in contrast, the variance of future rates is so high that fully inflation-proofed private pensions are hard to come by and expensive. In short, it is no accident that it is possible to buy life insurance but not inflation insurance.5 What risks and uncertainties can hamper consumption smoothing? All pension schemes face macroeconomic shocks, demographic shocks, and political risks. Private, funded schemes face additional risks: · Management risk. This can arise through incompetence or fraud, which imperfectly informed consumers cannot monitor effectively. · Investment risk. Pension accumulations held in the stock market are vulnerable to stock market fluctuations. At its extreme, if a person is required to retire on the day of his sixty-fifth birthday, there is a lottery element in his pension accumulation. · Annuities market risk. For a given accumulation, the value of an annuity depends on remaining life expectancy and on the rate of return the insurance company can expect over those years. Both variables face both risk and significant uncertainties. NDC pensions avoid some of the risks that private pensions face, notably management risk and investment risk. They may also reduce annuities market risk, if only because with a single, nationwide annuities pool, the law of large numbers will reduce the variance facing the insurer (that is, the state). This is an unambiguous advantage. However, the advantage is inherent in state-run PAYG schemes generally, rather than NDC schemes specifically. 62 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Less stringent demands on private sector institutional capacity. Private pensions make con- siderable institutional demands on both the public and the private sector. The latter will be absent in poorer countries. Even where it is present, private pensions may not be the most welfare-enhancing use for scarce private sector skills, which might better be used in build- ing up productive capacity. NDC pensions make no demands on the private sector, though that advantage is inherent in all state pensions. As discussed in the third section, however, NDC pensions make heavier demands on public institutional capacity than state schemes with a less tightly defined relation between contributions and benefits. Capacity to cope with uncertainty, not just risk. With social insurance, the contract is not fully specified. Precisely for that reason, social insurance can adjust to changing conditions and unforeseen contingencies. Atkinson (1995, 210) points out that "the set of contingen- cies over which people formed probabilities years ago may have excluded the breakdown of the extended family, or the development of modern medicine, simply because they were inconceivable." Thus social insurance, in sharp contrast with actuarial insurance, can address not only risk but also uncertainty. Thus NDC pensions have the potential to ameliorate uncertainty in ways that private schemes do not. The ability to pay fully indexed pensions once a person has retired is one example; the capacity to protect the pension rights of people with caring responsibilities (which is not an insurable risk) is another. This is a significant advantage--again, however, one that resides in social insurance generally, rather than the NDC design specifically. ASSIST SUSTAINABILITY If an NDC scheme is genuinely actuarial, then future expenditure is by definition equal to revenues, so that the scheme--again by definition--is sustainable. This feature, however, is not exclusive to NDCs. Consider a balanced PAYG scheme, where: sWL = PN (4.1) where s = the PAYG social security contribution rate, W = the average real wage, L = the number of workers, P = the average real pension, and N = the number of pensioners. If the social security act specifies a pension formula in which P = sWL/N (4.2) again, expenditure = revenue by definition. Thus sustainability is not specific to the NDC design. It may be, however, that the poli- tics are easier with NDCs. ENHANCE TRANSPARENCY The argument is that NDC pensions have explicit rules and therefore that the system is transparent in two ways: individuals know the basis on which their pension will be calcu- lated; and any attempts by government to alter the scheme are visible. These features are important, but not exclusive to NDCs. The U.K. system prior to 1975 was highly transpar- NON-FINANCIAL DEFINED CONTRIBUTION PENSIONS: MAPPING THE TERRAIN 63 ent, with a flat rate contribution for all workers giving entitlement to a flat rate benefit. Another example of transparency is a PAYG scheme with defined benefits, but with retire- ment age explicitly related to life expectancy, significantly reducing the need for other parametric change. A final salary scheme is also transparent to the recipient, and govern- ment attempts to change the benefit are very visible. REDUCE INCENTIVES FOR FRAUD In an NDC scheme, like all PAYG schemes, the only pot of money is the current year's con- tributions: that is, the flow of contributions, not the stock. Thus there are few assets for the state or private actors to pillage. Separately, if the state wants to increase the taxation of pensions, it can do so only on benefits in payment, not on the fund, since there is no fund. Both features, once more, are inherent in PAYG generally rather than in NDC design specifically. NDC schemes thus have advantages, but almost all of them are features of state pen- sions generally, rather than exclusive to the specific design of NDCs. Separately, any advantage of principle may impose a heavy requirement in terms of government capacity, a topic to which the third section returns. Equivocal Aspects Some features of NDCs can be regarded either as advantages or disadvantages, depending on views about theory, empirical facts, or values. NONDISTORTIONARY In discussing the impact of pensions on labor market decisions, it is helpful to distinguish two statements: badly designed pensions cause labor market distortions; and fully actuar- ial pensions minimize such distortions. This chapter argues that the first statement is true, but the second does not follow. Badly designed pensions undoubtedly cause labor market distortions in terms of both retirement decisions and responses earlier in life. If the concern is the retirement decision, 6 pensions should be related at the margin to individual contributions. The argument is important. It is open to policy makers to have a pension formula that is redistributive in the sense that worker A, with twice the earnings of worker B over his working life, gets a pension which is higher than B's, but less than twice as high. However, if either A or B retires early, his pension would be actuarially reduced relative to the pension he would have received at age 65. In contrast, earlier labor market decisions depend not only on the marginal relationship between contributions and benefits, but also on the effect of an increase in earnings on the total pension package. In this case, it is necessary to consider a fuller actuarial relationship between contributions and benefits. How do these arguments apply to fully actuarial pensions like NDCs? Badly designed state pensions cause major distortions. However, state schemes, whether NDC or DB, avoid one important distortion--the labor immobility problem caused by private DB schemes. In addition, DB schemes with long averaging periods are less distortionary than with a short period and, as discussed earlier, can be very similar to a DC scheme. More fundamentally, the next section argues that fully actuarial benefits are not optimal in a sec- ond-best world. A second reason why a fully actuarial design might not be optimal is that minimizing distortions is only part of the story. The argument implicitly assumes that all that matters is labor supply. But it can be argued that what really matters is economic welfare. It may be, for example, that a defined benefit scheme reduces labor supply at the margin; but if 64 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES the loss of utility resulting from lower output is more than offset by the utility gain result- ing from greater certainty about consumption smoothing, then defined benefit arrange- ments may be welfare-improving, notwithstanding reduced labor supply. At a minimum, the welfare gains from greater certainly should be set against any costs of reduced labor supply. Thus the argument that NDC pensions minimize distortions is far from definitive. If the argument is true, then it is also true of other schemes in which contributions bear an actu- arial relationship to contributions, for instance a scheme with flat rate contributions and flat rate benefits, such as in the United Kingdom between 1948 and 1975. The desirability, or otherwise, of actuarial benefits is taken up in the next section. EQUITABLE The argument that actuarial benefits are equitable rests on the belief that redistribution should apply only to poverty relief and to credits in specific instances such as caring for small children. A contrary view is that the state pension should include redistributive assistance for consumption smoothing as well as for poverty relief. It can also be argued that, though NDC pensions help to cope with risk and uncertainty, they continue to face the individual with significant risks associated with the variability of earnings; other approaches, for example, DB, share risks more broadly, as discussed in the first section. Thus NDC pensions do not have a unique claim to equity. They are inequitable if policy makers or the electorate believe that social insurance has a redistributive role broader than poverty relief and/or if policy makers want risks to be shared more broadly than is possi- ble with actuarially based benefits. TIES THE HANDS OF GOVERNMENT The proposition is that NDC pensions, being actuarially based, constrain the govern- ment's freedom of action. Two sets of questions arise. Does the NDC design really tie the government's hands? In theory, the contract is fixed; but government could change the contract. Second, is tying the government's hands wel- fare-improving? At the core of this question lie two further sets of questions. The first is an empirical issue about the competence and motivation of government, about which people may take different views, and about which conclusions might differ for different countries. Some writers are sceptical about government, arguing that politicians award concessions to special interests in exchange for short-run political support, leaving the costs of those con- cessions to future taxpayers, at a time when the politicians who have granted them have long since retired. The contradictory argument is that a disadvantage of NDCs is that they reduce policy flexibility by adopting a fully specified contract, and thus forgo options for enhancing consumption smoothing by reducing the uncertainty faced by the individual. A second question concerns the trade-off between the certainty of a supposedly rigid scheme versus the greater options for risk-sharing that can occur with a more flexible scheme. If tying the hands of government is thought an advantage, is this possible only with NDC pensions? In principle the answer is no: NDC schemes are based on a social security law just like other PAYG schemes. It is true, however, that it might be harder politically to change NDC pensions. Disadvantages Alongside their advantages and equivocal aspects, NDCs have two significant sets of dis- advantages: they are not efficient, and they are suboptimal in welfare terms. NON-FINANCIAL DEFINED CONTRIBUTION PENSIONS: MAPPING THE TERRAIN 65 INEFFICIENT A central objective of pensions is to allow each person to make efficient choices about the time path of his or her consumption. Such a system of consumption smoothing should minimize distortions. On the face of it, this suggests that a strictly actuarial system would be efficient. Indeed, Góra and Palmer (2003, 15­16) write: In the NDC and FDC [funded defined contribution] framework there is no redis- tributive ambition, other than redistribution over the individual's own lifecycle from working years to years of retirement. Instead, the government's redistribu- tive policy . . . is financed through explicit taxes from general revenues. . . . In this way, insurance and its source of financing and social policy and its means of financing are kept separate, enhancing transparency. The approach gives rise to a number of queries. First, why is it efficient to have both first- and second-tier pensions organized on a DC basis? More fundamentally, though a strictly actuarial scheme may be efficient in a first-best world, policy design needs to cope with serious market imperfections. People can be myopic and/or imperfectly informed, giving a justification for compul- sion. The problem is nontrivial, and means that the simple assumption of rational utility maximization may not hold. New (1999) makes the useful distinction between an infor- mation problem and an information-processing problem. An information problem can be resolved by providing the necessary information, such as the capacity of different comput- ers, after which the individual can make his or her own choices. With an information-pro- cessing problem, in contrast, the problem is too complex for agents to make rational choices, even if the necessary information is provided. The problem can arise where the time horizon is long, as with pensions; or where the good or service involves complex probabilities, including, for example, life expectancy (the failure in this case is an inability to process probabilities); or where the information is inherently complex, as with compli- cated pension products. A second problem is missing markets. The market for indexed contracts, for example, is thin, to say the least. Progressive taxation is a third deviation from first-best. Diamond argues that in the comparison between defined contribution and defined benefit schemes, "there is no sim- ple dominance of one over the other in the presence of other labor market distortions" (2002, 57). Assuming that the rate of interest exceeds the rate of wage growth over the longer term, he continues: Indeed, with a progressive annual income tax and age-earnings profiles that are generally increasing in real terms, the marginal income tax rate is rising with age, on average. Thus, a well-designed DB system may well have better labor market outcomes since the overall tax burden, income tax plus net tax from social security, will vary less over the life-cycle. That is, income taxes are lower on the young and net social security taxes are higher. Therefore, without a detailed calculation, one cannot reach an efficiency conclusion. In any case the difference is likely to be much smaller than the difference between DB systems with long and short averaging periods. Formulating the issue as an optimal taxation problem would make it clear that in a sec- ond-best world, a scheme that is strictly actuarial is not, in general, efficient. 66 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES SUBOPTIMAL IN WELFARE TERMS Consumption smoothing is one objective of pensions but, as discussed at the start of the chapter, there are others, including reducing the risk people face (implicit in both the con- sumption smoothing and insurance objectives), poverty relief, and distributional objec- tives (which may include subsidizing consumption smoothing by people only slightly above the poverty line). A strict adherence to actuarial benefits may provide consumption smoothing, but sets aside the others. Proponents of NDC pensions then argue that the NDC pension provides consumption smoothing, and other instruments provide poverty relief and distributional goals. But-- going back to a point I learned so many years ago as a graduate student--if there are three targets, three instruments are needed to achieve them. The optimal solution, however, is normally not a single one: one relationship between each instrument and a particular tar- get. The NDC arguments are tidy in this respect and, on that account, rather appealing. But that does not make them right--as an optimal tax formulation would make clear. Prerequisites for Implementation The previous section asked whether NDC pensions are desirable. This section considers very briefly the parallel question: are they feasible? An initial question for policy makers concerns the level and distribution of income. If the country is poor, the poverty line, which determines the minimum pension, is relatively close to average earnings. Hence there is little gain from an earnings-related pension in general, and NDC pensions in particular. Thus a prerequisite for NDC pensions is suffi- cient disparity in the earnings distribution to make consumption smoothing relevant. A second central issue is that contributory pensions in general, and NDC pensions in particular, require considerable government capacity. The government needs to have suffi- cient economic capacity to maintain macro stability, sufficient political capacity to make long- term pension promises credible, and sufficient basic institutional capacity to collect contributions, to account for them in each year, and to cumulate records across years. The last of these conditions is particularly important for NDC arrangements, where every cent of every contribution counts toward a person's final pension. For NDC pensions, govern- ment also needs advanced institutional capacity for monitoring changes in life expectancy and for maintaining the long-run balance of the scheme, for example, the capacity to run a reserve fund effectively. Merely to state these requirements makes it clear that NDC arrangements impose particularly heavy demands on public sector capacity. Where this capacity is absent, NDC pensions should be regarded as future option rather than current policy. Conclusion Góra and Palmer (2003) talk about the need to "create new concepts" (p. 2) and about the "design of a new vehicle for efficient accumulation over the life cycle" (p. 27). Palmer's work has mapped out the idea of NDCs--in terms both of policy and implementation-- much more fully than previously. This is a considerable advance. NDCs remind us that state PAYG pensions can be as much or as little actuarial as we want: in other words, social insurance is not necessarily redistributive. Thus the approach is important because it reminds us of an important but often forgotten truth, but is not itself new. As I wrote in 1987 (and others had doubtless written before), "[Redistribution] is not inevitable, since a PAYG scheme could be organized to pay actuarial benefits" (Barr 1987, 222, emphasis in original). NON-FINANCIAL DEFINED CONTRIBUTION PENSIONS: MAPPING THE TERRAIN 67 Conclusion 1: NDCs Are Not a Dominant Policy Put another way, NDCs are a design for state pensions, not the design. Except in a world that is first-best and where policy makers are indifferent about distributional matters, a strictly actuarial relationship is not an optimum. That is stated as a proposition in theory. It implies that we cannot say that a strictly actuarial relationship is always and necessarily the best way to design pensions; indeed, it will generally not be optimal. Put another way, the theoretical conclusion leaves it open to people to take different views about pension design. Thus it is entirely sensible, coherent, and defensible to advo- cate NDC pensions. But since they are not a theoretically dominant policy, there are other sensible, coherent, and defensible policies: for example, a pension design that includes redistribution not just for poverty relief but also for consumption smoothing. Sweden, with its NDC system, offers an important example. But so do other countries, such as Aus- tralia (noncontributory, income-tested, first-tier pensions plus mandatory DC second-tier pensions), the Netherlands (tax-funded citizens' pensions plus mandatory, largely occupa- tional second-tier pensions), and the United States (contributory DB first-tier pensions plus voluntary DC third-tier accumulations). In short, there is room for different views about preferred pension design. On what basis should different policies be assessed? Much of the issue depends on the answers to the following questions: · Question 1. Is policy flexibility an advantage or a disadvantage? This is the old rules- versus-discretion debate. The answer depends on empirical views about the effec- tiveness and probity of government, and will therefore vary from person to person and by country. · Question 2. Is a wholly actuarial system (for example, NDC first tier + funded DC sec- ond tier) efficient? As discussed earlier, the answer is generally no; but the extent of welfare loss will depend, among other things, on the extent of risk aversion in the population (the welfare gains from greater certainty being higher the greater the degree of risk aversion). · Question 3. Are actuarial benefits equitable or not? This ultimately is a value judg- ment about whether redistribution is or is not properly limited to poverty relief. · Question 4. Would NDC be more sustainable than a defined benefit scheme? Note that we are comparing current defined benefit schemes as they are, with lots of bar- nacles, with a perfect, pristine NDC scheme. The answer is probably more political than economic. · Question 5. Is the scheme cost-effective? The answer will depend on objective factors such as the level of income in a country, and on empirical judgments about whether or not the relevant supporting institutions are sufficiently developed. Conclusion 2: It Depends on What You Mean by NDC NDCs can take many guises. Two polar cases are particularly relevant. · Case 1. The pension system is NDC plus a minimal guarantee. Such a system comes close to being strictly actuarial, and hence offers insurance in respect of the longevity risk and consumption smoothing, but only minimal poverty relief and vertical redis- tribution. This is a corner solution, and hence can be criticized for being inefficient and also, depending on one's perspective, inequitable. · Case 2. The pension system has two elements: a tax-funded pension that can either be flat rate or with an earnings-related component, and an NDC element. The latter may include tax-funded credits, for example, to recognize caring activities. This arrange- 68 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES ment offers poverty relief, insurance, and consumption smoothing. If the tax-funded element has an earnings-related component, there is a redistributive element in con- sumption smoothing. The latter construct contains a richer array of policy options. But in this case the NDC pension is not the first tier, but the second. It amounts to a pension system with a tax- funded first tier and an NDC second tier. NDC is no longer the pension, but an element in a wider system. Finally, as discussed earlier, a state-run DB scheme with accrual over a full working life, an age-related accrual rate, and annuities determined ex post is formally identical to an NDC scheme based on earnings growth per worker. Conclusion 3: NDC Pensions Do Little to Address the Central Funding Issue The root of long-term unsustainability is that in virtually all countries pension schemes incorporate a retirement age of 60 or 65, which remains largely fixed as life expectancy rises. Rising life expectancy is a source of joy; the problem is having a fixed age at which the pension first becomes payable. NDC pensions address the problem in a formal sense by reducing the accrual rate. But unless people retire later, this approach on its own risks pensioner poverty. That is, sus- tainability is in conflict with sound social policy. In the absence of any constraints, the endogenous variable is not the minimum pensionable age but the size of the pension. In a world of rationality and perfect information, this would face each person with an actuarial budget constraint against which to make his or her optimal choice about when to retire. But if people have a personal discount rate higher than the interest rate used for actuarial adjustment, they will tend to retire as soon as possible, with progressively larger actuarial adjustments as life expectancy increases. In the limit, this pulls everyone down to the min- imum pension. One of the conclusions to emerge from Gruber and Wise (1998, 2002) is that many people retire as early as they are allowed. "The collective evidence from all countries combined shows that statutory social-security eligibility ages contribute importantly to early departure from the labor force" (Gruber and Wise 1998, 161). Thus a minimum age at which a person may first receive a pension is an important element in pension design. Given these arguments, my own view is that a minimum pensionable age that rises over time is an essential ingredient in the policy maker's armory. A more comprehensive solution has five elements: · Policy makers should set an initial pensionable age at a point that makes it fiscally feasible to provide a genuinely adequate state pension. In the absence of a normative theory, a pragmatic approach would be to work out the maximum fiscal envelope for pensions, and the minimum genuinely adequate pension. Together, these determine the maximum number of pensioners that can be supported. That figure, combined with the age distribution, determines the initial pensionable age. · Deviations from that pensionable age should be roughly actuarial. · Over time, the minimum pensionable age should increase in line with rising life expectancy in a way that is rational and transparent, so that people know a long time in advance when (in broad terms) they will be able to retire. · Labor market reform should introduce flexibility to allow people to move from full- time work toward full retirement along a phased path of their choosing. The design of pensions will need to support these choices. · Government should strengthen its efforts to increase public understanding of the simple economics of pensions. NON-FINANCIAL DEFINED CONTRIBUTION PENSIONS: MAPPING THE TERRAIN 69 Notes 1. "Notional defined contribution" and "non-financial defined contribution" should be understood to have the same definition. 2. Pay-as-you-go pensions are paid (usually by the state) out of current tax revenues. With funded schemes, pensions are paid from a fund built over a period of years from the contributions of their members. 3. See Barr (2001), chapter 6. 4. The question of whether pensions should be actuarial at the margin or across a per- son's entire contributions record is discussed in subsequent sections. 5. See Barr (2004), chapter 9. 6. See Gruber and Wise (1998, 2002). References Atkinson, A. B. 1995. Incomes and the Welfare State: Essays on Britain and Europe. Cambridge: Cambridge University Press. Barr, N. 1987. The Economics of the Welfare State. London: Weidenfeld; and Stanford, Calif.: Stanford University Press. ------. 2001. The Welfare State as Piggy Bank: Information, Risk, Uncertainty and the Role of the State. Oxford and New York: Oxford University Press. ------. 2004. The Economics of the Welfare State (4th ed.). Oxford: Oxford University Press; and Stanford, Calif.: Stanford University Press. Diamond, P. 2002. Social Security Reform. Oxford and New York: Oxford University Press. Góra, M., and E. Palmer. 2003. "Shifting Perspectives in Pensions." Paper prepared for the "Conference on NDC Pensions," Sandhamn, Sweden, September 28­30. http://www.rfv.se/konferens/docs/shifting_perspectives_in_pensions.pdf. Gruber, J., and D. Wise. 1998. "Social Security and Retirement: An International Compari- son." American Economic Review 88 (2, May): 158­63. ------. 2002. "Social Security Programs and Retirement Around the World: Micro Estima- tion." Working Paper W9407, National Bureau of Economic Research, Cambridge, Mass. New, B. 1999. "Paternalism and Public Policy." Economics and Philosophy 15: 63­83. Chapter 5 Conceptualization of Non-Financial Defined Contribution Systems Assar Lindbeck* WHEN COMPARING ALTERNATIVE PENSION SYSTEMS, it is useful to rely on a three-dimen- sional classification: funded versus unfunded systems, actuarial versus nonactuarial sys- tems, and defined benefit (DB) versus defined contribution (DC) systems (see Lindbeck and Persson 2003). The basic distinction between funded and unfunded (PAYG) pensions is that pension benefits in funded pension systems are financed by the return in financial markets on ear- lier accumulated pension funds, but in unfunded (PAYG) systems they are financed by the current flow of contributions (taxes) from the active population. In the context of the box (trapezoid) in figure 5.1, where funding is depicted on the vertical axis and actuarial fair- ness on the horizontal axis, variations in the degree of funding are illustrated as vertical movements. It is also useful to distinguish between "broad funding," when the build-up of pension funds is associated with increased national saving, and "narrow funding," when this is not the case. The second dimension of pension systems, actuarial versus nonactuarial arrangements, refers to the link between the individual's own contributions and his or her future pension benefits. The strength of this link may be characterized as an expression of the degree of actuarial fairness. A pension system is completely nonactuarial if there is no link at all. By contrast, a link is "actuarially fair" if the capital value of the individual's expected pension benefits is equal to the capital value of his or her own contribution--also on the margin (that is, when the individual varies his or her hours of work during his lifetime). This is the only type of pension system in which there is no labor market distortion. Variations in the degree of actuarial fairness, and hence in the labor market distortion, are schematically depicted as horizontal movements in the figure. Thus not only the funding dimension but also the actuarial dimension is a continuous variable. In principle, all pension systems covered by the two dimensions depicted in figure 5.1 could be either DC or DB. This aspect of pension systems is then a third (orthogonal) dimension (not depicted in the figure). I define a DC system as one where the contribution rate is fixed, which means that the pension benefits must be (endogenously) adjusted from time to time to ensure that the pension system remains financially viable. In a DB system, * Assar Lindbeck is professor of international economics at the Institute for International Economic Studies (IIES), Stockholm University, and at the Research Institute of Industrial Economics (IUI). 71 72 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 5.1. A Taxonomy of Social Security Systems III IV degree of funding I II degree of actuarial fairness Source: Lindbeck and Persson (2003). by contrast, the individual is promised either a lump-sum pension benefit or a specific relation between earnings and subsequent pension benefits (often expressed as a promised replacement rate). The contribution rate then must be (endogenously) adjusted from time to time to ensure financial balance. The taxonomy may be illustrated by considering the extreme points I­IV in the figure. A tax-financed lump-sum pension benefit, equal for all pensioners, is the prototype of a non- actuarial unfunded system (position I in the figure). In the context of a simple overlapping generations model, where the individual's lifetime is divided into two periods--work and retirement--the average return on previously paid contributions is then equal to the growth rate of the tax base, G, while the marginal return is zero. Position III in the figure instead depicts a nonactuarial, fully funded system, where the average return is the return in financial markets, R, and the marginal return is zero. In position II, with the maximum degree of actuarial fairness for a PAYG system, both the average and the marginal return is G. Finally, position IV depicts an actuarially fair fully funded system where both the average and the marginal return is R. Since the average and marginal returns in position II differ from the corresponding returns in position IV (the former are usually lower), I call a system in the former position "quasi-actuarial" rather than actuarially fair. In the context of this classification of pension systems, a "notional defined contribution" (NDC) system is simply a quasi-actuarial system (position II in the figure) of DC type: that is, a system with an exogenous contribution rate. The individual's pension wealth may then be registered in an abstract ("notional") account--rather than in a factual financial account as in fully funded, actuarially fair systems.1 Although the entire contribution rate, , constitutes a marginal tax wedge in a com- pletely nonactuarial system (positions I and III in figure 5.1, as well as in between them), the wedge falls when the system becomes more actuarial (that is, by a move to the right in the figure). For instance, a horizontal move from position III to position IV reduces the marginal tax wedge from to zero. The consequences for the marginal tax wedge of a move from I to II are slightly more complex. While the marginal tax wedge in the case of a CONCEPTUALIZATION OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEMS 73 lump-sum pension (position I) is , in a quasi-actuarial system (position II), it may be writ- ten as follows in a two-period overlapping generations model (Lindbeck and Persson 2003): (R-G)/(1+ R) 1-(1+G)/(1+ R) [ ] (5.1) The equation says that the individual's income loss from having to pay the contribution rate in a quasi-actuarial system is equal to (the discounted value) of the difference between the market return, R, and the return in the PAYG system, G, multiplied by the contribution rate, . (The individual is forced to save the fraction of his earnings at the return G rather than at the market rate R.)2 Thus a shift from a completely nonactuarial PAYG system to a quasi-actuarial system implies a reduction in the marginal tax wedge from to the expression in equation 5.1. For instance, if the contribution rate is 20 percent, such a shift may cut the marginal tax wedge by approximately half--with realistic assumptions about the number of years in work and retirement, respectively, and with reasonable assumptions about G and R.3 This would amount to a nontrivial reduction in the tax distortion of work. Since there is no income effect in this case (the average return is G both before and after the reform), the individual is encouraged to choose more hours of work, a longer working life, more investment in human capital, more effort to be pro- moted, more geographical mobility, and so on--provided he understands that the mar- ginal tax wedge has been reduced considerably. This schematic calculation has assumed that the original pension system was com- pletely nonactuarial. If, instead, the original PAYG system has some actuarial elements--if it is located somewhere between positions I and II in figure 5.1--the reduction in the mar- ginal tax wedge would, of course, be smaller. Indeed, simply basing the pension benefits in a DB system on more years of earnings than the current scheme may reduce the mar- ginal tax wedge. In the limiting case, when the pension benefit is based on an individual's lifetime earnings, the implicit marginal tax wedge could, in fact, be the same as in the expression in equation 5.1, provided yearly earnings are indexed by the same factor as the return factor in the NDC system. Thus an earnings-based DB system may mimic the work incentives in an NDC system. The reason for claiming that the DB system then mimics an NDC system, rather than the other way around, is that the latter type of system is by its very nature quasi-actuarial, while the degree of actuarial fairness varies in DB systems. Indeed, in many countries, the degree of actuarial fairness in DB systems is quite modest, since such systems are designed to ensure income protection (hence to guarantee an accus- tomed income level) and to redistribute income across income classes. Earnings-based DB systems may also mimic other typical features of NDC systems. One example is automatic reductions in pension benefits in proportion to increased aver- age life expectancy for each cohort. Such adjustments are also an inherent ("natural") fea- ture of NDC systems, which are designed to include strong actuarial elements. Although the same mechanism can be introduced in earnings-based DB systems, this would be less self-evident since actuarial features are not emphasized in such systems. Another difference between DB and NDC systems concerns the international transferabil- ity of pension benefits. Such transferability is also an obvious element in NDC systems, since an individual's pension wealth consists of personal accounts based on contributions paid earlier by the individual. Thus the wealth in a notional account is in a natural way regarded as the individual's property. Such transferability is a less obvious feature of DB systems, where pension benefits have traditionally been tied to citizenship or residence in a specific country. Moreover, international complications are bound to arise if interna- 74 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES tional transferability is allowed in today's DB pension systems. The reason is that such sys- tems are often tied to earnings over a limited number of years. Without intergovernment coordination of pensions systems, some individuals may lose their pension benefits when they move to another country (because of too few years of work to be eligible for pension benefits). In other cases, the sum of pension benefits from different countries may add up to a larger amount than an individual would have accumulated if he or she had lived in one country during his or her entire working life. (Only if all years of earnings would count, as is automatically the case in NDC systems, could this problem be avoided in an earnings-based DB system.) DB and NDC systems may also differ with respect to their financial viability. A crucial issue in this context is whether it is politically easier, or more difficult, to adjust the pen- sion benefits in NDC systems than to adjust the contribution rates in DB systems. Histori- cally, it has certainly turned out to be politically feasible to raise the contribution rates gradually over the years in DB systems--from a few percent of earnings to 15 to 25 percent in most developed countries. The situation may be different today in the sense that voters and politicians may regard contribution rates as so high that further increases are not politically feasible. Indeed, ambitions to avoid further increases in contribution rates are an important background for recent attempts to freeze the contribution rates in DB sys- tems. Again, this would mean that a typical feature of NDC systems is introduced into a DB system. How financially stable then are NDC systems compared with DB systems? This depends, of course, on exactly how the systems are designed. In principle, to be financially stable, pension benefits in an NDC system should ideally be gradually adjusted through- out an individual's retirement period in response to variations in the growth rate of the tax base of contemporary workers and expected remaining longevity of retired individuals. This would obviously require a variable rather than a fixed annuity. (The need for such adjustments of annuities could be reduced, however, if the systems had buffer funds that smooth the benefit payments over time, although this would make the pension system less actuarial.)4 The distributional effects also tend to differ between DB and NDC systems. In principle, an NDC system reflects no particular distributional ambitions, either within or across gen- erations, since the system is supposed to be quasi-actuarial. DB systems have no such dis- tributional limitations, and they have traditionally been regarded as legitimate and useful tools for both inter- and intragenerational redistribution. Like other PAYG systems, earnings-based DB systems redistribute income in favor of the first generation of pension beneficiaries, which also implies subsidization of the labor supply (another example of a labor market distortion). Intragenerational redistributions are more ambiguous. For instance, although floors and ceilings on benefits (without corresponding arrangements in the case of contribution rates) tend to redistribute income in favor of low-income groups, rules according to which pension benefits are based on earnings during the best x or last y years usually favor relatively affluent individuals. The net effect on the redistribution of lifetime income among income classes within generations usually seems to be rather mod- est in real world systems. Although "pure" NDC systems are void of redistributional ambitions, real world NDC systems are often combined with policy measures designed to bring about intragenera- tional redistribution. Obvious examples are specific tax-financed pension benefits for mil- itary service, childcare, sickness, and unemployment. In other words, redistributional elements are often added to pure NDC systems, so that the distributional consequences become more like DB systems. In this special sense, NDC systems may mimic the distrib- CONCEPTUALIZATION OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEMS 75 utional features of DB systems. If politicians are anxious to maintain the "purity" of an NDC system itself, such tax financing may be kept outside the NDC system. Nevertheless, such complementary arrangements have to be considered when analyzing the conse- quences of shifting from DB to NDC systems. DB and NDC systems also have different implications for intergenerational risk sharing. Mature DB and NDC systems distribute income risk differently in response to distur- bances of various types. In principle, in DB systems, the "burden" of unexpected changes in the growth rate of aggregate earnings (and hence the tax base) is borne by active gener- ations, in the form of variations in contribution rates. In NDC systems, by contrast, this risk is shared between generations, since pension benefits will change for both pensioners and workers (in the future). Current employees, of course, also experience a change in earnings. The distribution of income risk also differs in response to changes in life expectancy. If pension annuities in NDC systems are adjusted gradually (for instance, every year) during the retirement period (a variable annuity), pensioners bear the entire income risk. In NDC systems in the real world, however, the adjustment often takes place only at the time of retirement. If no further adjustment occurs during an individual's retirement period in response to changes in remaining longevity (a fixed annuity), such disturbances may impair the financial viability of an NDC pension system. Subsequent generations then have to bear this risk in one way or another. In the real world, politicians often try to cre- ate financial viability by combining reduced benefits and increased contributions, presum- ably to force active generations and pensioners to share the risk. To summarize: Even if DB systems may mimic typical features of NDC systems (and in some cases, vice versa), the systems differ in their general philosophy. Some features "fit" better in NDC systems than in DB systems--and vice versa. The systems also have differ- ent distributional consequences, within as well as across generations, and they tend to dis- tribute income risk differently. Moreover, they differ with respect to the ease of allowing pension benefits to be transferable across countries. It is also likely that the "property rights" of pension benefits are more robust politically in NDC systems. Notes 1. "Notional defined contribution" and "non-financial defined contribution" should be understood to have the same definition. 2. However, the implicit tax wedge will be higher in early than in late working life, since in the former case the contributions are locked in (at a low yield) over a longer period than in the latter case; see Lindbeck and Persson (2003, p. 85). 3. The calculation is based on the assumption that the value of the contributions accumulates, on average, for 32 years, which means that equation 5.1 may be rewritten [1 ­ (1 + G1)32/(1 + R1)32]. I also assume that G = .02 and R = .04. 4. The Swedish system is unnecessarily unstable because the return is tied to average real wages rather than to the wage sum, and changes in life expectancy do not result in changes in pensions during the course of the retirement period. This is the background for the introduction of a special "balancing mechanism" in the Swedish system. Reference Lindbeck, A., and M. Persson. 2003. "The Gains from Pension Reform." Journal of Economic Literature 41 (1): 74­112. Conceptualization of Non-Financial Defined Contribution Systems Peter Diamond* IN THEIR PAPER IN THE JOURNAL OF ECONOMIC LITERATURE, Assar Lindbeck and Mats Pers- son (2003) provide a three-dimensional classification of pension systems. One dimension is seen in the contrast between defined contribution (DC) and defined benefit (DB) sys- tems based on adjustment methods to financial realizations. DC systems adjust benefits, while DB systems adjust revenues. This distinction is really a continuum in that one can adjust a combination of the two. This could be done as part of automatic adjustment, as has been proposed for the United States by Peter Diamond and Peter Orszag (2004), who proposed that roughly half the automatic adjustment for the impact of life expectancy increases on social security finances be accomplished by benefit reductions and roughly half by payroll tax rate increases. Or a combined approach can be used in the course of leg- islation, as in the 1983 reform of U.S. social security (see Light 1985). When it is done by legislation, then the picture can become even more complicated, in that benefits for some can be increased while the general level of benefits is decreased. In addition to arguing that this dimension be considered a continuum, I wonder if it might not be better to use the phrase "adjustments to stochastic realizations," recognizing that pure DC and pure DB systems are just two points in this continuum. Lindbeck and Persson's second dimension is the degree of funding, which is a continu- ous variable as well. This dimension is also more complex in that there is the important distinction made by the source of the funding. Analysts are very aware of the difference between assets that are politically committed to paying for benefits and assets that also have been accumulated in a way that contributes to national savings. Thus there is further complexity in this dimension, as well. They refer to their third dimension as actuarial--the extent to which there is a tight link between paying taxes and getting benefits. This is clearly tied to labor market incentives and is also more complex than they describe. One can think of a system that is a combina- tion of a flat benefit and a benefit proportional to the accumulation of taxes paid. Then the relative sizes of the two portions indicate how distortive the labor incentives are (dis- tortive in the sense that this would interfere with the fundamental welfare theorem if there were no other violations of the conditions needed for the theorem to hold). This example makes it clear that, as with the other two dimensions, there is no sense that "more actuar- ial" is necessarily better (since income distribution matters as well as efficiency), just as there is no sense in which "more funded" is necessarily better or further along on the DB, DC dimension (in one direction or the other). But there are other ways in which the determination of benefits can differ from a DC system (which may not be distortive in the sense I used the term above). There is the issue of the weights on different years of earnings in determining benefits--the differ- * Peter Diamond is an institute professor at Massachusetts Institute of Technology. 76 CONCEPTUALIZATION OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEMS 77 ence between accumulating at a market interest rate in a DC system and an internal rate of return in an NDC system. Actually this comparison needs to be adjusted if it is to be orthogonal to the other dimensions. That is, the comparison ought to be performed holding constant the present value budget constraint for a given cohort. In this case, the NDC approach is expected to weight earlier years less than the market interest rate does and weight later years more than the market interest rate does. Use of the market interest rate involves no distortion if the cohort breaks even on taxes and benefits, there is no redistribution within the cohort, and there are no other labor market distortions. But these conditions do not hold generally. Indeed, the use of progressive annual income taxes would make the NDC approach less distortive since it would tend to bal- ance the rise in marginal income taxes that comes with the usual age-earnings profile. That is, with an upward sloping age-earnings profile and progressive annual income taxes, the sum of explicit income tax and implicit social security tax can well be smoother with NDC than with DC. It is common to observe that the redistribution to earlier cohorts distorts the labor supply of later ones. In a two-period overlapping generations (OLG) model where the first generation gets a surprise benefit after retirement and all later cohorts pay for it, this is the full story. In practice, early generations are treated more generously over an extended period, thereby affecting labor supply of both recipients and payers of inter- generational redistribution. That is, early cohorts get their labor supplies subsidized, while later ones get them taxed. Given a presumption of a preference for relatively stable tax rates, this is suggestive of a distortion, but a more complex one than in the simple two-period model. The same issue arises in systems that use a shorter averaging period--the last n years rather than all years, for example. Then, with a break-even comparison, there is taxation of earnings in earlier years that do not count for benefits, and subsidization in later years, which produce so much in benefits as to more than offset the taxes paid. The story becomes even more complex in a setting of individual uncertainty and the use of some measure of highest earnings rather than last earnings. That is, different benefit rules, com- bined with different stochastic structures on earnings possibilities, will yield different degrees of uncertainty about retirement benefits. I am not convinced that actuarial is a good term for this dimension since it is an intervention in the labor market that affects effi- ciency, individual insurance, and redistribution. Such a term is not used in considering the degree of progressivity of the income tax and it is not clear it is helpful to do so here. This is not to suggest a disagreement with Lindbeck and Persson's identification of labor mar- ket incentives as a very important third dimension when classifying systems--just that this dimension, like the other ones, is itself multidimensional, not a single point in a one- dimensional scale. Perhaps "labor market incentives" is a better phrase than "degree actu- arial." Thus I would rename their three dimensions, with new names of "adjustments for sto- chastic realizations," "degree of funding," and "labor market incentives." Renaming is essentially agreeing with the value in this tripartite way of approaching the effects of social security designs. In this setting, a pure NDC does all its adjustments on the side of benefits and none on the side of taxes, has limited funding through its buffer stock of assets, and has good labor market incentives. Where each of these three choices is, relative to optimality for some par- ticular country's initial position, is a hard question to answer. There is no basis for claim- ing a general optimality for any of the positions of a pure NDC system in any of the three dimensions. 78 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES To put an NDC system into context, let us briefly review how other systems work. If there is no system, an individual who is doing life cycle optimization saves, with different savings rates at different times; invests in some combination of assets; and at some time purchases an annuity of some form, with at least some of the accumulation. (Rolling pur- chase of annuities would be better insurance if available at equivalent pricing.) Such a per- son adjusts the level of savings over time in response to both the realizations of returns on assets and the realized earnings levels. A mandatory DC plan preserves the individual character of budget balance and the reliance on market pricing of assets and annuities and the bearing of the risk in both asset returns and earnings trajectories. A mandatory DC system does not typically attempt to adjust the savings rate to realizations. The uniform savings loses out on both the liquidity needs behind an ex ante plan of varying savings rates and the ability to adapt savings to experience. But there is room for varying savings if the mandatory rate is not too high-- below a savings level adequate for financing all of appropriate retirement income. Redis- tribution can be combined with this, either through a separate arrangement (such as minimum incomes) or within the system by transferring between accounts. A corporate DB plan typically relates benefits to a history of earnings and uses projected needs to determine assets to be accumulated. If there is government regulation of financ- ing, it does not apply to an individual but to the plan as a whole. Contribution rates would be continuously adjusted if there was a serious attempt to preserve full funding. In prac- tice, corporations adjust both benefit formulae and contribution rates in response to real- izations of both corporate earnings and pension system experience. Moreover, the wage levels themselves are among the candidates for responding to the risks in both pension experience and corporate earnings, subject of course to labor market responses. A mandatory national DB often differs from corporate DBs in the formula chosen for relating benefits to the history of earnings, although it need not (some corporate plans use the entire history of earnings in determining benefits). It also can differ from a regulated corporate plan in the target level of funding. Put differently, the risk implications of the level of funding are different for corporate and national plans since conditions leading to corporate bankruptcy are different from conditions resulting in countries repudiating benefits. An NDC is a hybrid with two creative innovations. One is that benefits depend on taxes paid, not earnings. The second is that the NDC plan is discussed in terms of a DC vocabu- lary, not a DB vocabulary. In the context of an unchanging tax rate, the first innovation is of little significance. The second innovation must have been helpful in achieving political consensus for reform in Sweden, but Axel Börsch-Supan, (2006) has argued that it would not have been helpful in Germany. If followed closely, a pure NDC has one less degree of freedom than does a similarly constructed DB. An NDC is supposed to provide benefits for different cohorts that have a present discounted value (PDV) that equals the value of the account, using the internal rate of return (IRR) for a discount rate. A DB system could adjust benefits for successive cohorts that followed a similar rule for relating relative benefits to relative life expectan- cies. But it has a degree of flexibility in setting the relationship between benefits and earnings. In practice, the Swedish NDC used a degree of flexibility in choosing to use period mortality tables rather than cohort ones, either projected or adjusted based on experience, as does a (CREF) annuity pioneered by TIAA-CREF in providing annuities for university employees. Instead, the Swedish system does its adjustment in two ways. One is the level of assets to provide the system at the start. The other is the automatic adjustment mechanism. CONCEPTUALIZATION OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEMS 79 In contrasting a well-designed DB or NDC system with a DC system, we see the poten- tial in the DB to improve social welfare by redistributing income and providing insurance for earnings through a progressive benefit formula. (Differently designed redistribution is potentially present in both DC and NDC, but is more in keeping with the approach of a DB.) We see the potential in both the DB and the NDC to provide more within-cohort risk sharing by relying less on rates of return, (rates of return also being earned on individually held assets). We see differences among the three in the weighting of earnings in different years in the determination of benefits. There is no sense in which an NDC is better than a well-designed DB. Instead I think of it as a way to get a DB system that is better designed than many current or former DB systems. There is wide agreement on several properties of a good system. A country should have one system--not separate systems for separate groups, with political power affecting the relative treatment of different workers. Benefits should be based on at least a large fraction of a career. A system should preserve projected balance, either through fully automatic adjustments or some combination of some automatic adjustments and periodic legisla- tion.1 A system also needs to have a reliable process for projecting the future workings of the system--both its financial position and its fulfillment of its social insurance goals. And not too much of the cost of reaching balance should be shifted onto generations in the dis- tant future. Does an NDC help countries not meeting these conditions meet them? It may, but it need not. It has been claimed that it is a virtue of the Swedish system that there is no reliance on forecasts. I think this is not necessarily a virtue. In a fully privatized system, the market engages in projections when deciding how to price annuities and when committing to rates of return on long-lived investment options that have given rates of return. I see noth- ing inherently problematic in using projections in determining a balance between benefits of different cohorts. I also note that with a private market system with sensible workers, workers would be adjusting their savings rates to realizations of their experience in financ- ing retirement incomes. Moreover, the adjustment rules that do not use an explicit forecast can be seen as merely relying on a naďve forecast. It should be noted that the value of one kronor in an NDC account is not equal to the value of one kronor in a funded DC account. Since the NDC kronor is earning a lower rate of return than the DC kronor, it is worth less. Thus the claim that workers know the value of their accounts is wrong. It is good for workers to be informed about anticipated monthly benefits. Since there is speculation that workers overvalue lump sums relative to the flows they can finance, more information is definitely useful. Moreover, the depen- dence of the value of the accumulation in an NDC account on future legislated returns means that accounts with the same accumulation would have different values in countries with different anticipated growth rates. Thus wider use of NDC does not provide ready transfers between countries without detailed actuarial calculations of assets that would need to be transferred to accompany a transfer of liabilities. An NDC system faces a choice between how it allocates risk to different participants at different times and how likely it is to have a need for an adjustment. For example, recog- nizing the higher risk aversion of retirees than of workers, it makes sense to have benefits in force not fully subject to the fluctuations in taxable earnings. In sum, an NDC system is likely to be pretty good--serving its social insurance goals well. It leaves open several choices about design, choices that should be based on the con- sequences of choice, not some notion of an ideal NDC in light of NDC philosophy. The choices in the design of benefits may be particularly important--single or joint-life annu- ities, choice of indices for adjusting benefits, and the time shape of benefits more generally. 80 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES These choices should reflect both the impact on retirees and the labor market incentives. On a break-even basis, steeper benefits that start lower may be particularly useful for dis- couraging retirements if they are thought to be occurring too early. Related is the choice of whether to have a retirement test for a few years after initial eligibility for retirement ben- efits. Such a test affects the time shape of consumption beyond the early entitlement age as well as retirement decisions. Although an NDC is likely to be a pretty good system, it does not make sense to oversell it, claiming excessive virtues relative to alternatives. Note 1. For discussion of different ways of achieving balance--of allocating the risk associ- ated with realizations of economic equilibria--see Diamond (2004). References Börsch-Supan, A. 2006. "What Are NDC Systems? What Do They Bring to Reform Strate- gies?" In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 3. Washington, DC: World Bank. Diamond, P. A. 2004. "Social Security." American Economic Review 94 (1): 1­24. Diamond, P. A., and P. R. Orszag. 2004. Saving Social Security: A Balanced Approach. Wash- ington: Brookings Institution Press. Light, P. C. 1985. Artful Work: The Politics of Social Security Reform. New York: Random House. Lindbeck, A., and M. Persson. 2003. "The Gains from Pension Reform." Journal of Economic Literature XLI (1, March): 74­112. Conceptualization of Non-Financial Defined Contribution Systems Salvador Valdés-Prieto* THE FIRST SECTION OF THESE COMMENTS SHOWS where the notional defined contribution (NDC)1 design stands regarding funding method and benefit formulae. The second sec- tion shows that most NDC plans allocate the outcomes of risk in a way that is fundamen- tally different from fully funded defined contribution (DC) plans (which are mutual funds), and in this sense most NDC plans are not really defined contribution. The third section argues that the NDC design should not be allowed to have a role in first-pillar and in third-pillar policies. The final section argues that notional accounts (NA) plans can and should provide some liquidity for their members, illustrating that the room for improve- ments of NDC plans is significant. Mapping the NDC Family The three main dimensions of pension plan design are the degree of financial funding of the plan, the type of benefit formula offered to members, and the method used by the plan to allocate the outcomes of aggregate plan risk.2 It is generally agreed that NDC combines a benefit formula that uses "actuarial" formulae, with a low degree of financial funding: that is, it approaches pure pay-as-you-go finance. Adoption of the NDC design seems driven by the wish to attenuate initially very high labor distortions, and by the inability of the adopting country to overcome the transition cost toward partially or fully funded alternatives. These issues relate to the first two of the main design dimensions, not to the third. Degree of Funding Consider the degree of "financial" funding of the plan, which is the same as the "narrow funding" mentioned by Assar Lindbeck in this roundtable. This degree of funding is obtained by dividing the market value of the pension fund by the expected discounted value of accrued liabilities at the same date. What economic assets are part of the pension fund? Possible assets include not only financial securities, but also land, real estate, and even precious metals owned by the plan. On the other hand, the present expected value of government subsidies and other donations to the plan do not belong in the pension fund because those policies can be modified and those donations can stop. It follows that a pen- sion fund is the set of payment promises in favor of the plan that are protected by property rights, defined in turn by protections spelled out in the constitution (Valdés-Prieto 2002). * Salvador Valdés-Prieto is professor of economics at Catholic University of Chile and researcher at the Centro de Estudios Públicos, Santiago. Robert Holzmann provided excellent comments on the final draft. Reactions at the conference by Axel H. Börsch-Supan and Elsa Fornero were also very useful in preparing this written version. 81 82 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES There is another aspect of funding, referring to economywide savings. The degree of "ultimate" or "broad" funding is the degree to which individual wealth is backed by phys- ical assets and net foreign assets. The degree of broad funding is a function of the size of the public debt and of publicly financed infrastructure, and not only of the funding status of the national pension plan. In fact, intergenerational transfers in health subsidies, in edu- cation, and in the stock of knowledge can back substantial amounts of individual wealth,3 influencing the degree of broad funding. Which of these two is most relevant? It is not appropriate to characterize a pension plan by the features of the economy in which it operates. Therefore, the degree of broad fund- ing of the economy cannot be one of the defining characteristics of a pension plan.4 Con- sider an economy where credit card securitization is introduced, allowing an expansion of the stock of consumer credit outstanding and a transitory increase in the flow of consumer expenditures. As the economy's stock of national saving has fallen, the degree of broad funding of the economy has fallen. However, if a fully funded pension plan buys these securitized bonds, any observer will ascertain that the plan itself remains fully funded. As shown by Bernheim and Shoven (1988), there is no link between aggregate savings and changes in the funding status of a pension plan. In conclusion, what matters for labeling the funding status of a pension plan is its degree of "financial" funding. A pension plan's impact on aggregate national saving may be positive, but it depends heavily on how much other components of savings adjust when the pension plan changes its degree of narrow funding. Raising the degree of "financial" funding of a plan does have two desirable consequences, even if national saving is unaffected. First, the plan gets the additional protection of property rights over the future cash flows produced by the assets. Second, a higher share of plan assets can be traded in financial markets, and this allows the plan to gain from trade, benefiting members. In contrast, raising the size of national sav- ings may or may not be desirable after the transition cost is taken into account. Benefit Formulae Benefit formulae are functions that link the amount of benefits received by an individual member during the passive phase of life with that same member's circumstances during the active phase of life. There are two main classes of benefit formulae. The "actuarial" class is based on the amount contributed when active, uses individual account balances, interest is credited to the accounts and uses annuity conversion factors to transform the balance in an annuity. This class includes deferred annuities purchased by installments and variable annuities. The other main class of benefit formulae is built around "years of service" and "revalued career average salary." There are other benefit formulae outside these two classes, such as the Brazilian "fator previdenciario" introduced in 1999. Under some parameter configurations, a years-of-service benefit formula can be almost equal to an actuarial benefit formula, but they always differ in one aspect: the response of benefits to increases in contribution rates is positive for actuarial formulae and is zero for years of service formulae. It was suggested in the conference that NDC plans introduce no tax wedge on labor decisions because they use an actuarial benefit formula. I disagree, because a benefit for- mula may be actuarial but can use nonmarket parameters. Examples of parameters are the interest rate credited to individual accounts and the annuity conversion factors that set the amount of the pension. The link between individual contributions and the expected pre- sent value of additional benefits may be far from the analogous link offered by financial market prices, despite the mathematical shape common to all actuarial formulae. Consider NDC plans. The fact that they use pay-as-you-go (PAYG) finance fixes the direction of the deviation between the link provided by the financial market and the link CONCEPTUALIZATION OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEMS 83 provided by the benefit formula, at least for steady states. Consider a situation where GDP and the covered wage bill grow at a constant rate g, where the internal rate of return (IRR) offered by the NDC plan's benefit formula is g, and where the real rate of return on a bal- anced portfolio of investments and insurance contracts available in the financial market is r, after adjusting for risk differences. Asset market equilibrium requires that r > g in the long run. If not, it would be possible for any infinitely lived agent whose income grows in proportion to the economy to borrow at rate r and roll over its debt indefinitely without ever paying interest, and the total debt would still fall over time relative to assets (Tirole 1985). In addition, the empirical evidence supports the hypothesis that r > g in most actual economies most of the time (Abel et al. 1989). It follows that r > IRR. Therefore, the wedge introduced by an actuarial benefit formula with rate g is a tax, not a subsidy, in the long run. Two important labor decisions that are distorted by this hidden tax in NDC plans are: · Retirement at earlier ages--the fact that IRR < r creates an economic incentive to start a NDC pension as early as allowed, and · Joining covered employment at older ages--the fact that IRR < r gives an incentive to the young to prolong their studies and ramblings and to engage in uncovered work for a longer period. This may prevent some of them from getting crucial on-the-job training at the most appropriate age. There are two other good reasons to avoid attaching efficiency, equity, and insurance implications to pension plans that use "actuarial" benefits formulae, as some authors did in the conference. One is that the plan's incentives may be irrelevant in the wider economic context in which the individual operates. The other is that the plan's incentives might be highly correlated with the incentives provided by other government policies that affect the individual. To see this, consider the efficiency of the labor market decisions made by active work- ers, such as hours, effort, and participation. These decisions depend on net taxes on earn- ings. These net taxes, in turn, depend on both the marginal and the average link between contributions and the risk-adjusted expected present value of benefits in this particular pension plan. However, labor incentives also depend on other pension plans in which the individual may participate (say an occupational plan), on the income tax schedule, and on the marginal and average link between contributions and the expected value of benefits offered by mandatory health insurance and unemployment insurance. If a pension reform raises the marginal link between benefits and contributions in the pension plan, the impact on labor choice may be small or zero, if the other programs maintain or increase their tax wedges. Now consider a case where the current pension plan is fully funded, but it replaced an older PAYG financed plan that was solvent, that is, financially balanced. Assume that the contribution rate to the new funded plan was cut to take advantage of the higher return (r > g). Assume also that the transition cost was financed by issuing new public debt and by introducing a new explicit and permanent tax on earnings to finance interest payments on that new public debt, equal to (r ­ g) × (New Debt). This new tax collects revenue that is just enough to prevent the ratio of new debt to GDP from exploding in the long run (Auer- bach and Kotlikoff 1987). Then changes in the tax wedge of the mandatory plan--cur- rently zero--are fully negatively correlated with the total explicit tax rate on earnings. In fact, the overall tax rate on covered earnings is the same as it was under the old PAYG- financed plan. The impact of the wider economic context should also moderate statements about the impact of pension reforms on insurance and income redistribution. Consider a compari- son between a defined benefit (DB) plan that offers implicit insurance for an unexpectedly 84 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES short working life through a less-than-actuarial reduction to early pensions, such as the U.S. social security program, and a simple NDC plan. It is standard practice to attach insurance advantages to the DB plan over the NDC plan, without checking if the personal tax system in the second country includes implicit insurance for a short working life. This insurance will be offered if the income tax schedule on pension income is progressive, because those who start their pension younger--say, due to an unexpectedly short work- ing life--will receive a smaller pension and therefore will pay lower tax rates, achieving a less-than-actuarial cut on after-tax early pensions. Moreover, if the country with the DB plan exempts pensions from personal income taxes, and the second country does not, total insurance for a short working life may be larger in the second country, despite a plan that does not offer that type of insurance. From the two features of NDC--an actuarial benefit formula and a low degree of fund- ing--it follows that NDC is a label that covers a wide set of pension designs. For instance, if the notional interest rate credited in individual accounts during the active phase of life is the growth rate of average wages minus 1.6 percent per year, as in Sweden, the financial behavior of the plan and its implications for members are bound to be quite different from those of another NDC plan where the notional interest rate is 6 percent real per year, as was proposed in Brazil during the debate leading to the "fator previdenciario" reform of 1999.5 The use of actuarial benefit formulae still allows for substantial differences in para- meters, which may create large differences in the labor market incentives, insurance, and redistributional impacts. NDC and Redistribution It is useful to distinguish between three types of redistributive policies. The first supports the poor, including the elderly poor, with resources from the rest of society. The second redistributes income among the employed middle classes, say from professionals with higher earnings toward menial workers with lower earnings. The third type taxes the owners of capital to support the working classes. Consider the following policy of the second type: concentrating the redistribution among the middle classes in the tax system versus dispersing such redistribution among a variety of pension, health, housing, education, and unemployment insurance programs, in addition to the tax system. If concentration is chosen, then this type of redistribution can be designed in more encompassing ways than if these redistributive policies are dispersed and embedded in several sectoral policies. When the trade-off between redistribution and efficiency is presented globally to public opinion, transparency increases, because the debate is freed from the complicated aspects of sectoral policies that raise the cost of com- munication. Some may believe that policies should be made less transparent to enable experts to escape the manipulations introduced by politicians seeking reelection through pork rather than through the production of desirable public policies. This can be the case in some coun- tries and episodes. However, experts also have an interest in gaining access to the broad picture when they design policies that redistribute income within the middle classes. Concentration implies that redistribution within the middle classes ceases to be one of the purposes of pension policy because that goal is served by overall tax policy. Of course, redistribution continues to be important, but it is pursued more comprehensively. A major advantage of adopting NDC is that it facilitates concentration, allowing an increase in the transparency of the overall redistribution policy. The overall degree of redistribution also implies a degree of partial insurance of dispos- able income, because taxes fall when income falls. If pension amounts reflect lifetime income better than annual income, then let fiscal policy impose a different set of income CONCEPTUALIZATION OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEMS 85 tax rates on pension income than on other sources of annual income. The same logic is used in most countries to set special tax rates on the capital gains that occur when a family sells their house. Another advantage of concentration is that tax rates on pension income can be designed to take into account income from third-pillar pension contracts and from capital. Method Used to Allocate the Outcome of Aggregate Plan Risks Casual observers tend to believe that the method used by a plan to allocate the outcomes of aggregate risks is set by the benefit formula. For example, it is thought that actuarial benefit formulae come together with the DC method used by mutual funds. It is also believed that years-of-service benefit formulae always allocate aggregate risks with the DB model. Two examples show that this association is invalid. Consider first an annuity con- tract. As the life insurance company guarantees a certain benefit amount regardless of shocks to investment returns and to the life tables, this contract is appropriately described as a defined benefit. However, its benefit formula is actuarial. In the opposite extreme, con- sider a plan that uses a years-of-service benefit formula to set the initial pension for each member, but indexes subsequent pension payments to the actual observed growth rate of contribution revenue, in a PAYG context. This plan uses a method to allocate aggregate risks that transfers them in full to older members, so it is not DB.6 The method for allocating aggregate plan risks is heavily influenced by other aspects of the plan, too. These other aspects can be rules for adjusting plan parameters (if such rules exist), the rules for the use of credit lines from sponsors (such as the treasury), and the rules for trading the uninsurable component of aggregate risk in the financial markets. Risk can be allocated by discretion as well. Discretion is what happens when the allocation of aggregate financial losses and gains is decided ex post by a designated set of people-- say, members of parliament, members of a board of elders, or a minister of finance. Such authorities may also limit the pension plan's access to credit. In practice, successive layers of rules for adjusting parameters blend gracefully into discretion. These other aspects turn the risk allocation method into a design dimension that is effectively independent from the benefit formula, as emphasized by Lindbeck and Persson (2003).7 Risk may be created, destroyed, or transferred. Badly designed rules and discretion can create risk. Consider a hypothetical NDC plan whose factor for converting account bal- ances to annuities is a fixed function of the returns earned by a relatively small buffer fund that is partly invested in equities. That rule is a bet that creates risk, making the plan oper- ate as a casino, which would be Pareto-dominated by other rules unless some members are risk lovers. Now consider risk creation by a plan whose parameters are set by parliament according to the electoral needs of a changing majority. This plan may turn into a political casino. A partial-equilibrium model of optimal life-cycle consumption calibrated with data for Germany and the United States finds that political risk creates a welfare loss equivalent to between 1.7 percent and 3.6 percent of the pension amount (Holst 2005). These examples show that the set of possible risk allocation methods that may be used by a pension plan is wider than the conventional distinction between defined benefit and defined contribution. This fact raises two challenges: the first is to classify the options in a useful way, and the second is to develop welfare criteria to identify the optimal option in a given environment. Classifying the Options One of the most useful classifications of risk allocation methods was presented above: rule-based versus discretionary. Another common distinction is between DC and DB. 86 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES These distinctions are independent, so we have a two-way classification of risk allocation methods. In one axis is the degree to which aggregate risk is allocated by rules versus dis- cretion. The other axis contains pure DB, pure DC, and combinations such as portfolios that change risk in response to the member's age. Consider the precise meaning of pure DC. The origin of the expression defined contribu- tion is legal: the obligations of the sponsor are limited to the obligation of paying the agreed contribution, and are defined by this limit. When considering the economic meaning of DC, it cannot be that the contribution rate is fixed over time because when that rate has been changed in fully funded DC plans, those plans have not lost their DC character.8 DC cannot mean that the contribution rate is not adjusted in response to realizations, because the mutual fund, which is the original DC design, does not change its DC nature when a member adjusts her contribution rate in response to realizations.9 I offer the following eco- nomic definitions (Valdés-Prieto 2002, p. 717):10 DC: When the risk allocation method transmits all the plan's aggregate financial risk to current members only, and it does so in proportion to the capital value of the accrued rights to pension benefits owned by each member. This definition of DC is a summary description of the standard mutual fund. This is the natural benchmark because of its widespread use in financial markets. A mutual fund allo- cates risk by defining the price of each share as the market value of assets divided by the number of shares (the value of liabilities is this price times the number of shares). The number of shares varies with inflows and outflows valued at yesterday's share price.11 The proposed definition of DC allows some leeway because the method for calculating capital value is not specified. It is quite different to discount future expected benefits at market interest rates than to discount them at the internal rate of return of the plan, which may be quite lower,12 because the pro-rata shares for allocating the aggregate outcomes of risk are distributed among plan members in a different way. DB: When the risk allocation method does not transmit any portion of the plan's aggre- gate financial risk to pensioned members. Shielding pensioners from aggregate risk implies that risk must be allocated to other plan members, or to a bond issuer (who pays a fixed amount in most states of nature), or to a sponsor (as in annuities, where the life insurance company takes the risk), or to tax- payers (as in discretionary DB, where parliament can pass the risk to taxpayers and bene- ficiaries of government transfers). Optimal policy can be analyzed after the options have been outlined. Automatic finan- cial stability in the short run is a valuable property for a pension plan because it prevents insolvency. International experience shows that discretion has failed to provide automatic financial stability in the pension policy area, in contrast to the case of central banks and monetary policy. This evidence suggests that rules are better than discretion in pension policy. Which is the optimal rule? Economic theory shows that to maximize welfare, a plan should trade risk with the rest of the economy at market prices, and the uninsurable risk that remains should be shared among members in proportion to individual risk toler- ance.13 Empirically, the predicted risk tolerance of a member is a function of the volatility of labor earnings, the presence of other wealth, access to insurance and credit lines, and age. Pure DB (as defined above) cannot be socially optimal, because it is unlikely that all pensioners have zero risk tolerance.14 Even among aged members, risk tolerance is likely to be higher than zero for a subset of members. In the same way, a DC plan with a single CONCEPTUALIZATION OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEMS 87 portfolio is not optimal either. In contrast, a DC plan where members are allowed to choose among a set of balanced portfolios with different risk levels appears closer to opti- mality, because adaptation to individual risk tolerance is allowed. Moreover, a portfolio designed to serve those who plan to start a pension in a given date (say in 2020) can pro- vide a valuable service to its members: change the level of risk slowly as they age, moving toward portfolios more heavily invested in bonds. Bond portfolios are DB as well because the issuers of the bonds absorb 100 percent of the underlying economic risks. Implications for NDC Plans An NDC plan that indexes pensions to price inflation is DB because pensioners are exempt from sharing in the plan's aggregate financial risk. In contrast, when an NDC plan indexes pensions to GDP growth (as in Italy) or to the growth rate of average covered wages (as in Sweden), the plan does not guarantee a fixed standard of living to pensioners, and there- fore is neither DC nor DB. Those designs allow pensioners to share in the economy's growth or in the growth of average wages. However, they do so regardless of whether the plan is experiencing an aggregate actuarial surplus or deficit in the present. For example, if unemployment rises due to a drop in labor demand for menial jobs, the Swedish rule registers that average wages rise while slower employment growth may push plan assets below liabilities. The ensuing increase to pensions (triggered by the increase in average wages) raises liabilities, exacerbating the financial imbalance. This NDC rule may be foist- ing risks on members when aggregate risk is zero, a situation that may be described as risk creation or wagering. In contrast, if the origin of the faster growth in average covered wages is an across-the- board increase in labor productivity, then this rule can be defended on the grounds of intergenerational risk sharing. Policy makers should check how rules respond to a wide variety of shocks, to avoid risk creation. Now consider rules for active members. Most NDC plans are less financially stable than fully funded DC plans because all deviations from steady state growth create actuarial surpluses and deficits (see Settergren and Mikula 2006, and a proof in Valdés-Prieto 2000). Moreover, reliance on credit lines from the treasury means that these plans create risk for the treasury, or require taxation of past generations to endow the plan with a buffer fund. If parliament makes the final decision but organized members respond with political activism, the plan may come to be seen by some as a political casino. The creation of risk for active members is a trick that a truly DC rule, as the one used by mutual funds, is unable to pull. A truly DC rule transmits all aggregate risk to current plan members, without creating risk. Thus, the NDC design is fundamentally different from DC, in the sense defined above. For this reason, I subsequently refer to the Italian, Polish, and Swedish plans as "notional accounts" (NA), rather than NDC. The Swedish plan's original rules were to credit notional capital of active workers at the growth rate of contribution revenue, and to adjust pensions in payment by the growth rate in average wages minus 1.6 percent per year. However, since May 2001, a different rule applies: When the plan's liabilities exceed assets, the notional interest rate credited to active members and the rate of adjustment of pensions in payment are both cut, to equal the growth rate of average wages minus the growth rate in liabilities plus the growth rate in assets (Settergren 2001). This mechanism might endow the Swedish NDC plan with automatic financial balance in the short run and might prevent the creation of risk for the treasury. However, a further twist was introduced in the 2001 reform. Even if subsequent liabilities are surpassed by assets, the lower notional interest rate and the lower rate of indexation of pensions remain in force, until the original ratio of average pensions to aver- age wages is restored. No symmetric rule for the case when assets exceed liabilities is in 88 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES place. This asymmetry may produce permanent cash surpluses, which would strengthen the treasury at the expense of plan members. This discussion suggests that sponsors of NA plans have much work to do improving risk allocation methods. This task requires the development of a more comprehensive framework to define optimality. The Role of NA in Pillars One and Three Let us discuss first the role of the NA design in third-pillar policy. I define third-pillar policies as those that introduce fiscal or regulatory incentives, but not mandates, to induce improvident people in the middle classes to save more for old age on a volun- tary basis.15 PAYG finance allows workers to invest in an implicit security whose dividend is a share of the human capital returns (earnings) of future generations of covered workers.16 Assume that this security should have pride of place in most people's long-term portfo- lios.17 It follows that any financial services company should consider offering an NA plan to clients who save for old age induced by fiscal incentives. For example, banks and mutual funds that offer IRAs and 401(k) plans may want to offer an NA plan on a volun- tary basis. However, third-pillar plans that are NA are likely to enjoy an undeserved marketing bonanza in their introductory phase. This is because PAYG finance offers net subsidies to the initial generation. But when PAYG finance matures, it must create a tax on members to remain financially independent. It will be difficult to find new members in the long run, in the absence of a mandate to future workers to join the plan. An example is the pyramid pension plan that appeared in Paraguay in 1985, which grew for at least 10 years thanks to the subsidies offered to the initial members.18 Therefore, investor protection criteria sug- gest that financial firms should be banned from offering voluntary NA plans. For a related reason, the state should beware of mandating an NA plan on a class of firms or on a sector of economic activity. Although a new NA plan can pay subsidies to the initial generation, which will help attract workers into this sector, in the long run the tax hidden in mature PAYG finance is likely to scare them away. In that case the plan will default on promised pension amounts. At that point, the losing members are likely to remember the sectoral mandate and demand compensation from the government. Now consider the role of NA in first-pillar policy, which I define as the set of govern- ment programs that provide support to the old poor. The aim of such programs is to pro- vide more subsidies to those people who are poorer--and who are precisely those who contribute less. Given the fact that NA plans use actuarial benefit formulae, which pay less benefits to those who contribute less, it is obvious that NA plans contradict the aim of first- pillar policy. Poor pensioners are likely to be relatively more risk-averse, because of their closeness to perceived subsistence consumption levels. The fact that NA plans are not DB means that the old poor are allocated an excessive share of the plan's aggregate financial risk. More- over, an NA first-pillar program may subject poor pensioners to the bets embedded in the risk allocation method. It follows that NA should not be used in first-pillar programs. NA Plans and Partial Liquidity for Mandatory Saving Mandatory plans where member wealth is illiquid create efficiency costs when the mem- ber values liquidity. Only those members who own substantial voluntary liquid assets are exceptions. Even provident members respond to illiquidity by modifying their labor CONCEPTUALIZATION OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEMS 89 choices in order to minimize this perceived cost. Mandatory plans that freeze accrued pen- sion rights into an illiquid claim are also likely to be inequitable. The reason is that work- ers who own negligible amounts of precautionary savings are likely to be the poorest. The illiquidity of pension claims may force these poorer workers to sacrifice too much con- sumption in order to build up a stock of precautionary savings, or to depend too much on the emergency loans provided by close kin or loan sharks. These problems besiege all mandatory plans, including fully funded, NA, and traditional designs. An interesting claim made in this conference is that in an NA plan, benefits are as illiquid as in fully and partially funded plans. Pension plan liquidity refers to giving members the option of getting immediate access to a limited portion of their claims on old age benefits in response to an emergency, ideally defined by members themselves. Partial liquidity holds the promise of mitigating the efficiency costs and the inequities caused by imposing a mandate to save at a flat rate, disregarding individual liquidity needs. Partial liquidity is a real possibility in fully and partially funded plans, which can sell some assets to give credit to members. As international experience shows, recovery of liq- uidity loans can be ensured by possessing a number of pension payments starting at the date of pension issue, delaying effective pension age.19 The amount of such loans should be capped with methods that create incentives to repay.20 Now consider an NA plan, financed with the PAYG method. Can it offer the option of providing partial liquidity to its members? If members are allowed to make partial with- drawals, the plan has to finance the aggregate net flow by issuing debt in the financial market. To break even, the plan must charge members a market interest rate (plus admin- istrative costs).21 However, this option is not available to all NA plans: the plan must enjoy enough financial stability to be creditworthy. Thus, NA plans that wish to offer partial liquidity to their members face one more requirement than those faced by fully funded plans. They must be creditworthy, despite being financed with the PAYG method and being subject to some political discretion.22 Summing up, the NA design is very attractive for initial conditions characterized by highly inefficient benefit formulae and fiscal constraints that prevent a transition to full funding. However, much remains to be done to improve this promising design. Notes 1. "Notional defined contribution" and "non-financial defined contribution" should be understood to have the same definition. 2. This three-way classification, used by many, including Lindbeck and Persson (2003), is more useful than the older two-way classification that merged together the benefit for- mula and the method for allocating the outcome of aggregate risks. 3. See Lee (1994). 4. In this we differ from Lindbeck and Persson's taxonomy (2003, p. 75). 5. That proposal was rejected because 6 percent real was far below market interest rates in Brazil at the time. 6. There is no suggestion that any of these risk allocation rules is optimal. 7. One may also consider defining DC as a combination of a certain benefit formula (actuarial) and some risk allocation property. That approach would fail to build upon the orthogonality between benefit formulae and the risk allocation method, and would fail to manage the different ways in which a plan may deal with risk. 8. On this point, I disagree with Assar Lindbeck, in this roundtable. 9. On this I disagree with Peter Diamond, in this roundtable. 90 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES 10. This definition of DC fulfills the principle enunciated by Settergren (2001, p. 4, foot- note 9): "In a DC plan uninsurable risks must be assumed by the pension level, rather than by the contribution rate." 11. The use of daily prices to value the accrued rights of members transmits any short- term volatility contained in market prices to the value of accrued rights. To prevent the transitory component of that volatility from creating unwarranted worries among mem- bers with limited financial experience, it may be wise to report accrued values using aver- ages over reasonable periods. 12. For example, Sweden discounts liabilities and assets at the growth rate of average covered wages, which is below market interest rates in the long run. 13. See the summary of "Syndicate Theory" provided by Kreps in his textbook, A Course in Microeconomic Theory (1990, pp. 169­74). 14. Merton (1983) stated this point more than 20 years ago. 15. Provided that the fiscal incentives are more generous than those given to savings not oriented to cover old-age needs. The fiscal incentive may be flat rate, progressive (as in the Czech Republic's 1995 program), or regressive (as in deductions from income tax in the United States). 16. The word security implies that the owner is protected by property rights. This is usu- ally not the case for claims on a share of the future earnings of workers. 17. For example, see Merton (1983) and Dutta, Kapur, and Orszag (2000). 18. See "Voluntary participation in a pay as you go plan," in Valdés-Prieto (2002, box 10.2, p. 516­7). 19. This repossession method is used successfully in the Philippines by the Employees Provident Fund (EPF). The critical requirement for economic efficiency is that the cost of nonpayment remains with the individual, as if the loan were deducted from the individual account balance, not from a collectively owned fund. One requirement for political sus- tainability of liquidity loans is that repossession does not reduce the pension amount, and this justifies a preference for raising the effective pension age. See a summary of interna- tional experience in Valdés-Prieto (2002, chapter 4, section 4.4). 20. A recent proposal is to cap the loans outstanding at, say, 70 percent of the sum of contributions made during the last 36 months. This works like airline frequent-flyer miles. It ensures that 100 percent of the contribution is saved for old age after 3 years and 30 per- cent is saved immediately. As the option of drawing a liquidity loan is valuable, many are unlikely to remain fully indebted most of the time. See Beyer and Valdés-Prieto (2004) for a recent proposal for Chile. 21. In PAYG-financed plans in steady state, this interest rate will be above the IRR offered by the plan to its members. 22. An NA plan can become creditworthy thanks to a credit line from the treasury, but in this case the scheme should be described as engaging the treasury in consumer finance. The interesting case is where partial liquidity is provided without guarantees from the treasury. References Abel, A., G. Mankiw, L. Summers, and R. Zeckhauser. 1989. "Assesing Dynamic Effi- ciency." The Review of Economic Studies 56 (1): 1­20. Auerbach, A. and L. Kotlikoff. 1987. Dynamic Fiscal Policy. New York: Cambridge Univer- sity Press. CONCEPTUALIZATION OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEMS 91 Bernheim, D., and J. Shoven. 1988. "Pension Funding and Saving." In Pensions in the U.S. Economy, ed. Z. Bodie, J. Shoven, and D. Wise, 85­114. Cambridge, Mass.: National Bureau of Economic Research, University of Chicago Press. Beyer, H., and S. Valdés-Prieto. 2004. "Análisis de la situación de trabajadores con baja densidad de cotizaciones y propuestas para corregir esta situación." Report to Servicio Nacional de la Mujer, Santiago, Chile, March. Dutta, J., S. Kapur, and M. Orszag. 2000. "A Portfolio Approach to the Optimal Funding of Pensions." Economic Letters 69: 201­6. Holst, R. 2005. "Policy Risk: Some Evidence, its Relevance and Welfare Costs in Retirement Programs." Dept. of Economics, University of Chicago, March. Kreps, D. 1990. A Course in Microeconomic Theory. Princeton, N.J.: Princeton University Press. Lee, R. 1994. "Population Age Structure, Intergenerational Transfer and Wealth: A New Approach, with Application to the U.S." Journal of Human Resources 29 (4): 1027­63. Lindbeck, A., and M. Persson. 2003. "The Gains from Pension Reform." Journal of Economic Literature XLI (March): 74­112. Merton, R. 1983. "On the Role of Social Security as a Means for Efficient Risk Sharing in an Economy where Human Capital is not Tradeable." In Financial Aspects of the U.S. Pen- sion System, ed. Z. Bodie and J. Shoven. Cambridge, Mass.: National Bureau of Eco- nomic Research, University of Chicago Press. Settergren, O. 2001. "The Automatic Balance Mechanism of the Swedish Pension System: A Non-Technical Introduction." Riksförsäkringsverket (August). www.rfv.se/English. Settergren, O., and B. D. Mikula. 2006. "The Rate of Return of Pay-As-You-Go Pension Sys- tems: A More Exact Consumption-Loan Model of Interest." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 7. Washington, DC: World Bank. Tirole, J. 1985. "Asset Bubbles and Overlapping Generations." Econometrica 53 (6): 1499­1527. Valdés-Prieto, S. 2000. "The Financial Stability of Notional Account Pensions." Scandina- vian Journal of Economics 102 (3): 395­417. ------. 2002. Políticas y Mercados de Pensiones: Un Texto Universitario para América Latina. Santiago, Chile: Ediciones Universidad Católica. mriverv1@puc.cl PART II CONCEPTUAL ISSUES OF DESIGN AND IMPLEMENTATION 93 Chapter 6 Demographic Uncertainty and Evaluation of Sustainability of Pension Systems Juha M. Alho, Jukka Lassila, and Tarmo Valkonen* POPULATION AGING IN EUROPE IS MORE COMPLEX than is generally recognized. The uncer- tainty in demographic projections is also larger than is usually assumed. This makes the assessment of long-term characteristics of pension systems challenging. Since different pension policy instruments react in different ways to demographic developments across cohorts and over time, the risk characteristics of the existing policies and proposed alter- natives should be investigated and "crash-tested" under a wide range of realistic alterna- tives. Because of the complexities of changing age structures, neither individuals nor firms nor administrators of pension systems can easily see what will happen to pensions and contributions if unlucky demographics materialize. Thus it is difficult for anyone to design and apply their risk strategies efficiently. "Crash testing" provides one way of addressing the problem. The broader conclusion is that for a pension strategy to be sustainable, it should explicitly state what actions are to be taken if the population and the economy do not evolve as expected. For example, while the notional defined contribution (NDC) sys- tem1 in Sweden provides an exceptional degree of transparency in this respect, the chal- lenge of quantifying and communicating the risk characteristics of the system to the population at large remains. This study combines stochastic population simulations with an overlapping genera- tions (OLG) model that assumes perfect foresight from the agents. We view this as a first step toward a more comprehensive model, in which future uncertainty is handled in a more advanced manner. Yet even the current models provide novel insights. New ques- tions can be formalized and new strategies entertained. We suggest areas for future research that are relevant for pension policy design. * Juha M. Alho is professor of statistics at the University of Joensuu, Finland; Jukka Lassila is research director and Tarmo Valkonen is head of the Unit in Public Finance Research at the Research Institute of the Finnish Economy. We thank Edward Palmer and Sergio Nisticň for comments. We also acknowledge the financial support of the European Commission for two underlying research projects: DEMWEL "Demographic Uncertainty and the Sustainability of Social Welfare Systems" (QLK6-CT-2002-02500), and UPE for "Uncertain Population of Europe" (HPSE-CT-2001- 00095). 95 96 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Uncertainty Caused by Lack of Sustainability A well-known definition of the sustainability of fiscal policies is the view of the Organisa- tion of Economic Co-operation and Development (OECD): "Sustainability is basically about good housekeeping. It is essentially about whether, based on the policy currently on books, a government is headed towards excessive debt accumulation."2 More precisely: "Fiscal policy can be thought of as a set of rules, as well as an inherited level of debt. And a sustainable fiscal policy can be defined as a policy such that the ratio of debt to GNP even- tually converges back to its initial level."3 The forecasts for spending and transfers are taken as given. This is close in spirit to generational accounting. In the case of pension systems, the OECD view requires that the current contribution rate suffice to finance benefits and that the possible fund stays at a reasonable level compared with the size of the system. However, other aspects merit attention. In partic- ular, a contributor may consider a system sustainable even if rules are changed or pen- sioners suffer but the contribution rate does not change too much. A pensioner may consider the system sustainable as long as the pension is as expected, even if this requires changing the rules and the contributors suffer. In either case the risk is that, because of future developments, the system puts an unexpected strain on one party or another. If the likelihoods of the various risks were known, the agents could prepare for them in a rational manner. Reducing the scope of the unexpected seems to be a natural aspect of sustainability. In game theory, a strategy is a set of rules, defined for the present and the future, stating what action will be taken under all states of the world.4 In the case of pensions, a sustain- able policy strategy is a set of rules such that both the contributors and the pensioners know, beforehand, what will be done in any reasonable future circumstance, and they accept the future actions, or at least cannot force a change of the system. The difference between a policy and a policy strategy is a very practical one. Consider- ations about sustainable pension policies are usually based on one set of base assumptions about key factors such as future demographics, productivity, and interest rates. In con- trast, considerations about sustainable pension strategies must be based on a large number of possible states of the world that cover a realistic range of economic and demographic developments. The whole Swedish NDC system may be thought of as a step toward implementing a strat- egy rather than merely being a set of policy instruments. It is "designed to be financially sta- ble, that is, regardless of demographic or economic development it will be able to finance its obligations with a fixed contribution rate and fixed rules for calculating benefits."5 Adjust- ment mechanisms have been defined; even a "brake" has been established, to be activated if things go badly.6 Whether the strategy is sustainable depends on its operating characteristics. Will the contributors and pensioners accept the outcomes that come of the new legislation, under different future circumstances? The time horizon may also be important. Valdés-Prieto (2000) suggests that long-term financial stability is irrelevant if the rules allow imbalances that continue long enough so that the political process is likely to intervene. In practice, one cannot know with certainty how future workers and pensioners will react to the system. The designers of new policies tend to emphasize the most likely future developments, when they argue for the reform. Opponents try to imagine circumstances in which one or another aspect of the system would cause it to crash. Our proposal is to provide realistic descriptions of the future contingencies in a probabilistic fashion. Con- ventional "high" and "low" scenarios that have been used for this purpose have had little or no effect on policy recommendations. Without any probabilities attached to the alterna- DEMOGRAPHIC UNCERTAINTY AND EVALUATION OF SUSTAINABILITY OF PENSION SYSTEMS 97 tives, their importance is suspect and the results are difficult to interpret. Analyses based on expected or "most likely" assumptions have dominated. For an individual contributor or a pensioner, sustainability is related both to trustwor- thiness and predictability. Can the level of future contributions and future pensions be known with sufficient accuracy, so that a choice can effectively be made between supple- mentary savings or consumption (or leisure time)? To have operational counterparts, this study will approach "trust" by defining thresholds for a change in contribution rates and replacement rates. These thresholds form a politically viable region within which the rates can change without leading to system reform.7 "Predictability" requires a known or esti- mated distribution of outcomes in the viable region. In the Swedish case, for example, it would be desirable to produce estimates of the likelihood of having to use the "brake." What is said above concerning an individual can be said for a firm, as well. In decisions concerning which country to invest in, it helps if the firm has a realistic view about indirect labor costs in the future. Unrestricted pension contribution rates can be a major source of risk. Aspects of Future Fertility and Mortality in Europe Age Structure and Negative Growth The total fertility rate of year t is defined as the expected number of children a newborn baby girl is expected to have, under the fertility and mortality regimes of year t. About 105 boys are born for every 100 girls, so if all women would survive to the end of child-bearing ages, a total fertility rate of 2.05 would suffice for internal population renewal. Allowing for mortality, a somewhat higher value, about 2.07, is sufficient in Europe. In year 2000, Iceland had a total fertility of 2.07. The remaining European countries have below replace- ment fertility. Belgium, France, Luxembourg, the Netherlands, the Nordic countries, Por- tugal, and the United Kingdom had a total fertility in the range [1.51, 1.89]. Austria, Germany, Greece, Italy, Spain, Bulgaria, Hungary, Poland, and Romania form a low fertil- ity group with fertility in the range [1.24, 1.36]. It is well known that these are exception- ally low values, in historical perspective. What has received less attention is the implication of fertility on the age distribution. In Finland, the total fertility rate in 2000 was 1.7. If this level were to persist, the popu- lation would start to decline at the rate of about 0.63 percent per year. Figure 6.1 tries to put this into a perspective. The solid curve corresponds to the age distribution of a closed sta- tionary population (in which births equal deaths) whose mortality equals that of the Finnish women of the late 1990s. The dashed curve is based on the same survival proba- bilities, but with births exceeding deaths by a constant ratio. The ratio has been chosen so that the resulting rate of increase (= 0.0065) equals the average growth of the Finnish pop- ulation during the 20th century. The dotted line has the corresponding age distribution when the rate of decline is negative (= ­0.0065). This happens to be almost exactly the asymptotic rate of decline implied by the current Finnish fertility. The declining stable population has a much older age distribution than the stationary one, let alone the growing population from which we derive our understanding of the world. This form of population aging derives from fertility alone, since we are keeping mortality fixed. In general, convergence to stability might take over a century, so stable populations have not received much attention in past years. However, the solid line with squares in fig- ure 6.1 gives the current age distribution of Finland. In about 10 to 15 years' time, the Finnish age distribution will be quite close to the asymptotic stable age distribution that 98 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 6.1. Age Distribution: Actual and Three Scenarios 0.020 0.015 stationarity 0.010 stable with rate of density increase 0.0065 stable with rate of 0.005 increase ­0.0065 actual in Finland in 2000 0 20 40 60 80 100 age Source: Authors' calculations. would result from constant schedules of fertility and mortality. The conclusion is that bar- ring large changes in fertility or migration, the Finnish population will resemble its stable equivalent quite soon. A similar conclusion probably holds for most European countries, but with variations in timing, and with due allowance for differential migration. Although much of current aging discussion involves future prospects of mortality, the major determinant of population aging is fertility, both via the current low level that leads to negative growth, and via the large baby boom cohorts that will begin to retire soon. The decrease in mortality does have an important effect on the sustainability of the pen- sion systems. Vaupel and Oeppen (2002) have presented evidence of the development of the "best practice life expectancy." This is the life expectancy of the country that at any given time has the longest life expectancy. Vaupel and Oeppen show that for women the curve goes almost linearly from the value of 45 years, observed in Sweden in 1840, to 85 years, observed in Japan in 2000. Accordingly, the best practice life expectancy has improved by approximately 0.25 years annually for 160 years. In individual countries the development has been quite erratic, however. In Finland, for example, female life expectancy increased by 24 years (81 years ­ 57 years) from 1930 to 2000, or by 0.34 years annually. During the first 40 years, the increase was 0.45 years annu- ally. During the latter 30 years, the increase was 0.20 years annually. Finland is not alone in this respect. Female data from 19 industrialized countries (Austria, Australia, Belgium, Canada, Denmark, Finland, France, Iceland, Ireland, Japan, Luxembourg, the Nether- lands, New Zealand, Norway, Sweden, Switzerland, the United Kingdom, and the United States) from 1950 to 2000 show that Japan's improvement is much higher than that of the other countries. During the first half of the 50-year period the average improvement of the remaining 18 countries was 0.23 years annually, but during the latter half it was 0.18 years annually, and in the 1990s it was only 0.15. It is difficult to say how the development should be interpreted. One can argue that as the leading country demonstrates that the improvements are possible, this will lead to pol- icy responses in other countries to catch up. However this may be, the example of Den- mark shows that such a response may take a long time. Fifty years ago Denmark was DEMOGRAPHIC UNCERTAINTY AND EVALUATION OF SUSTAINABILITY OF PENSION SYSTEMS 99 almost at the best practice line, but now it is six years behind. Pessimistic forecasters would suggest that improvements in European mortality are becoming smaller. Even if this view agrees with the most recent data, we note that a diminishing returns hypothesis has been repeatedly advanced in the past, but in retrospect, it has been the major source of error in mortality forecasts. A compromise between the optimistic and pessimistic views is to assume the continua- tion of past trends, but to quantify the variations about the declining trend. This can be particularly important for individuals making their career plans. Decisions about how much and how long to work may depend on how the level of pensions is viewed relative to the wages one might earn and the savings one might have. If, in addition, the pensions are lowered--assuming a fixed retirement age--as a response to improvements in life expectancy (as is the case in Finland), decisions about saving may become more important than before. Quantifying Uncertainty Past forecasts can be analyzed to assess how accurate they have been. The difficulty in doing this in practice is that the number of past forecasts is small for all European coun- tries. Reliable estimates are hard to obtain. Statistical time-series models can also be used, but it is difficult to find a compromise between overfitting on the one hand, and ignoring substance knowledge on the other. An intermediate way of assessing the uncer- tainty of forecasting is to consider the so-called naive or baseline forecasts. There is evi- dence that forecasts in the United States for total fertility have essentially assumed the current value to persist indefinitely.8 This has later been observed in many other (post- demographic transition) countries, as well. If total fertility were a random walk (or more generally, a martingale), then such a forecast would be optimal. Indeed, the autocorrela- tions of first differences of the total fertility suggest that a random walk provides a rough approximation. For countries with long data series available, one can determine how large errors such baseline forecasts would have had in the past, had they been systematically made every year (Alho 1990). In considering past errors, this study concentrated on the absolute value of the relative error and used the median to describe central value because this automati- cally eliminates the effects of outliers that were caused by wars or famines. Figure 6.2 (from Alho 2003) shows estimates from the Netherlands, Denmark, Norway, Finland, Ice- land, and Sweden (here listed from the largest to the smallest in terms of forecast error). The differences between the countries are considerable, but the order of magnitude is sim- ilar. This is to be expected due to the high autocorrelation of the errors. To appreciate the order of magnitude, note that 0.10 in the log-scale corresponds to an expected median error of 10 percent. Under a normal (Gaussian) model of relative error, this corresponds to a standard deviation of 0.15, or 15 percent. A similar analysis was carried out for mortality. Long data series from nine European countries (Austria, Denmark, France, Italy, the Netherlands, Norway, Sweden, Switzer- land, and the United Kingdom), in ages 50­54, 55­59, . . . , 90­94, were analyzed much the same way as fertility data. The difference was that in this case, the baseline forecast assumed that the decline observed during the most recent 15 years would continue indef- initely. Such a forecast would be optimal, if the actual development would be a random walk with a drift, and 15 years worth of data were available. Lee and Carter (1992) have shown that models of this type have considerable empirical support. The data were aggre- gated over age groups for each country. Figure 6.3 shows the relative error. Comparing with figure 6.2, we find, surprisingly, that the relative error one can expect in age-specific mor- tality forecasts is almost equal to that of total fertility. We have no explanation for the close 100 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 6.2. Median Relative Error of Fertility Forecast 0.6 0.5 0.4 error 0.3 median 0.2 median 0.1 0 10 20 30 40 50 lead time Source: Authors' calculations. Note: The figure shows the median relative error of fertility forecast as a function of lead time for six countries with long data series, and their smoothed median (circle). agreement of the two estimates. One should note, however, that the probability of survival can be forecast much more accurately. If one makes a large relative error in forecasting a mortality rate that is of the order of 1 percent, then the relative error in the number of sur- vivors is about one-hundredth of that. Estimates such as those displayed in figures 6.2 and 6.3 can be coupled with point fore- casts of age-specific demographic rates, and translated into statistical models that can be Figure 6.3. Median Relative Error of Mortality Forecast 0.45 0.40 0.35 0.30 error 0.25 0.20 median median0.15 0.10 0.05 0 10 20 30 40 50 lead time Source: Authors' calculations. Note: Relative error is forecast as a function of lead time for nine countries with long data series, and their smoothed average. DEMOGRAPHIC UNCERTAINTY AND EVALUATION OF SUSTAINABILITY OF PENSION SYSTEMS 101 used in simulation. A number of additional parameters for correlations across age and over time are needed in this. In our practical work, we have used the program PEP (Pro- gram for Error Propagation) written at the University of Joensuu.9 Demographic Uncertainty, Pensions, and Public Finances In the past, the volatility of demographics has often been underestimated, and economic analyses have concentrated on other uncertainties that are essential in population aging research. Research results obtained thus far unequivocally show that demographic uncer- tainty must not be neglected. Its magnitude should make us humble when recommending policies to avoid bad and unsustainable outcomes caused by aging. However, this does not mean that inactivity in policies is recommended. Quite the contrary, uncertainty is an extra reason for activity. As Auerbach and Hassett (2001) point out, for a risk-averse popu- lation the cost of future outcomes worse than expected outweigh the benefits of outcomes better than expected. Wise aging policies should be especially robust against a worse demographic future than the expected one. The discussion below presents examples of how stochastic population simulations have been used with economic models to analyze pensions in Europe. We look at projections of pension outcomes and the effects of single policy instruments. We hope to address explicit pension policy strategies, consisting of contingent use of several instruments, in future studies. Although our main goal is to throw light on the sustainability of pension systems, it is useful to keep in mind that there are other social systems that compete for the same eco- nomic assets. The costs of health care, schools, and social services are financed essentially by pay-as-you-go (PAYG) systems that aggravate the fiscal burden of the economically active population, in addition to pensions, for some potential future population paths; for other paths, they may bring relief.10 This chapter will not discuss these aspects further. Economic Calculations Subject to Exogenous Demographic Uncertainty The following two tables summarize the effects of the current policy, and four alternative pension policies, to contribution rates in Lithuania.11 The study used an OLG model for Lithuania.12 The demographic population paths were generated as described in Alho (2001). They were treated as being exogenous to the economic system. The results have been obtained by simulating the population 100 times and solving the OLG model in each case. A number of technical issues had to be solved to be able to carry this out. For exam- ple, arranging the population paths in a suitable manner helped in achieving convergence faster. The results were stored so that their statistical characteristics can be examined using standard statistical programs. The old-age pension system in Lithuania consists of the basic pension and the earnings- related supplementary component. The pension system is purely pay-as-you-go. Contri- butions are collected from the wage bill. The basic pension is almost flat, and it depends only to a minor extent on a person's insurance period. The supplementary pension com- ponent depends on the number of years the insured person has worked according to social insurance records and income. Pension benefit is fully indexed to average wage income. Table 6.1 presents statistical summaries of the predictive distribution of the social secu- rity contribution rate, under each of the four policies. In Lithuania, social security contri- butions consist of a 31 percent employer contribution on the wage bill, plus a 3 percent employee contribution on individual wages. These are used to finance pensions (old-age, disability, and survivorship), short-term benefits (sickness and maternity), as well as unemployment and health insurance partially. The contribution rate of 28 percent in 2000 102 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Table 6.1. Distribution of the Social Security Contribution Rate in Lithuania (percent) 2030 2050 Policy measure 2000 Median 50% 80% Median 50% 80% Current system 28 29.41 28.49 27.46 35.70 33.31 30.19 30.40 31.39 38.05 42.69 Wage-bill indexation 28 28.43 27.64 26.64 32.10 30.39 28.29 29.42 30.15 34.25 36.55 Longevity adjustment 28 29.16 28.34 27.42 34.84 32.90 29.96 30.19 31.20 37.17 42.01 Retirement age increase 28 25.05 24.20 23.31 29.83 28.06 25.24 25.78 26.84 31.77 35.51 Source: Authors' calculations. Note: The table shows the median and 50 percent and 80 percent predictive limits. is the employer rate where the health insurance is excluded. In the future the contribution rate is assumed to endogenously adjust so that the benefits are financed each period. Table 6.2 presents summaries of the predictive distribution of policy effects: that is, the differ- ence between the contribution rate under an alternative policy and the current policy. The width of the 80 percent predictive interval for the contribution rate is 4 percentage points in 2030 and 12 percentage points in 2050. Ranges of this sort are not atypical. As shown in the previous section, demographic uncertainty is of the same order of magnitude in different countries, and the closer a pension system is to a PAYG system the closer it replicates the underlying demographics. These estimates suggest what can be expected about the accuracy of our aging projections--and this is just the demographic component. Tables 6.1 and 6.2 show that pension policies have effects on both the location of the dis- tribution of outcomes and its scale. Indexation of benefits to total wages, instead of aver- age earnings, cuts the expected contribution rate by 3.6 percentage points in 2050 and narrows its 80 percent predictive range from 12 to 6 percentage points. This instrument is analyzed more closely in the discussion below on wage-bill indexation. Longevity adjust- ment of future pension benefits was applied only partially, and has very small effects on outcomes, perhaps because of problems related to the quality of old-age mortality statis- Table 6.2. Distribution of Policy Effects on the Social Security Contribution Rate (percent) 2030 2050 Policy measure Median 50% 80% Median 50% 80% Wage-bill indexation ­0.84 ­0.38 0.22 ­3.00 ­1.74 ­1.17 ­1.62 ­2.35 ­5.38 ­7.46 Longevity adjustment ­0.18 ­0.12 ­0.04 ­0.47 ­0.13 0.25 ­0.29 ­0.39 ­0.82 ­1.19 Retirement age increase ­4.38 ­4.18 ­4.07 ­5.64 ­5.22 ­4.96 ­4.64 ­4.98 ­6.48 ­7.23 Source: Authors' calculations. Note: The table shows the median and 50 percent and 80 percent predictive limits. DEMOGRAPHIC UNCERTAINTY AND EVALUATION OF SUSTAINABILITY OF PENSION SYSTEMS 103 tics in Lithuania. The next section considers the likely effects of the longevity adjustment in Finland. Increasing the retirement age also effects both the location and the scale of the predictive distribution of outcomes, but it is not directly linked to demographics, and this study will not discuss it in more detail. Predictive Distribution of Longevity Adjustment An increase in life expectancy can put a strain on the finances of a defined benefit pension system. In a PAYG system this means increasing the contribution levels of the current workers. In anticipation of future gains in life expectancy, a law was passed in Finland that automatically adjusts pensions if life expectancy changes. The aim is to preserve the expected present value of future pensions. Define p(x) as the conditional probability of surviving to age 62 + x given survival to age 62. Let 0.02 be the discount rate. Suppose a pension is paid continuously, at the rate of one euro per year. Then the Finnish law stipulates that the expected net value of the pension that forms the basis of life expectancy adjustments is = p(x)e-0 .02xdx. (6.1) 0 In practice, estimates of expected net values are computed based on past data. Consider the cohort of individuals who become 62 years old during a calendar year t = 2009. To cal- culate the expected present value for year t, denote it by x(t), mortality data from the five- year period [t ­ 6, t ­ 2) is used to calculate p(x). Thus there is no element of forecasting in the calculation, but the expected net value does not correspond to the actual cohorts of pensioners either. This aspect has been analyzed more closely in Lassila and Valkonen (2003), showing that the use of forecasts may mitigate the adjustment factor. The life expectancy adjustment is then defined as A(t) = (2009)/(t). Or, the pensions of those who became 62 years old in year t are multiplied by A(t). If mortality decreases from the year 2009 onward, A(t) < 1, so pensions would be cut. Since the future level of mortality cannot be known with certainty, the values of A(t) cannot be known accurately at the present time. However, in the interest of showing what one might expect, we can provide a probabilistic description of how the A(t)'s are likely to behave, since a predictive distribution of future mortality is available. Without going into details, we note that a number of technical issues must be resolved in any such calculation. For example, since a unique adjustment factor is used for males and females, a combined measure of mortality is used. This depends on the shares of women and men, in ages x 62, in the future. Various approximations were used to obtain the results shown here. The practical calculations were carried out via stochastic simulation using the program Minitab. The median, the first and third quartiles, and the first and ninth deciles, for the predictive distribution of the adjustment factors in 2030 and 2050, are as follows:13 Year d1 Q1 Md Q3 d9 2030 0.86 0.88 0.92 0.95 0.99 2050 0.78 0.81 0.87 0.92 0.98 We expect the adjustment factor to decline to about 0.87 in 2050, with an 80 percent pre- diction interval [0.78, 0.98]. These intervals are valid provided that the volatility of the 104 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES trends of mortality during the next 50 years does not exceed the volatility of mortality dur- ing 1900­94. In view of the discussion above, an optimist in mortality reduction who believes in a reversal of the recent slowdown might use the first decile (0.78 for year 2050) as a benchmark to consider how to adjust the predictive distribution to better match his or her beliefs. Wage-Bill Indexation Studies of the type presented in the previous two sections can be criticized because they do not incorporate the notion of sustainability or system reform in any way. The current (or alternative) rules are assumed to apply, irrespective of how the population or the econ- omy develops. Addressing this issue in full generality is vastly beyond current analytical capabilities. Yet some aspects of a more realistic analysis can be easily introduced in the existing modeling framework via thresholds. In Lithuania, the earnings-related pension benefit is fully indexed to average wage income. Alho et al. (2003) studied an alternative indexation, where the benefit follows the weighted average of the average wage income (with a weight 1 ­ ) and the total wage bill in the economy (with a weight ). The rationale of wage-bill indexation is that it would provide automatic relief to the working population should, as expected, the share of pen- sioners in the population increase. In the spirit of the previous two sections, figure 6.4 shows how the situation would change by 2050, in the polar cases of = 0.0 (left-hand side) and = 1.0 (right-hand side). The scatter plots are based on 300 population paths, each producing a dot on the (c, r) plane. First, we see that increasing tilts the scatter cloud from a horizontal position downward. Second, with current indexation, there are some extremely high contribution rates by 2050. A high degree of wage-bill indexation effectively cuts down the most extreme contribution rates, but even at = 1.0 the contribution rate can reach values above 0.4 with nonnegligible probability. Third, the control of contribution rates is accomplished by accepting a lower replacement rate. Figure 6.4. Replacement Rates and Contribution Rates in Lithuania in 2050, with = 0 (left) and = 1 (right) Wage indexation Wage bill indexation 0.34 0.34 0.32 0.32 0.30 0.30 0.28 0.28 0.26 0.26 0.24 0.24 pension/wage0.22 pension/wage 0.22 0.20 0.20 0.18 0.18 0.2 0.3 0.4 0.5 0.6 0.7 0.2 0.3 0.4 0.5 0.6 0.7 contribution rate contribution rate Source: Alho et al. (2003). DEMOGRAPHIC UNCERTAINTY AND EVALUATION OF SUSTAINABILITY OF PENSION SYSTEMS 105 As pointed out by Edward Palmer, a crucial assumption in these calculations is that the retirement age is kept fixed, even though life expectancies vary between population paths. Assuming, for example, a constant retirement/work ratio would certainly produce smaller declines in replacement rates, and further analysis should take this into account. At present this cannot be done properly because, as noted in connection with table 6.2, variation in longevity appeared modest, perhaps due to quality problems of old-age mor- tality data in Lithuania. In an effort to address the sustainability issue, Alho et al. (2003) approximate the com- plex political process by postulating bounds for the contribution rate, c, and the replace- ment rate, r, that cannot be violated. They assume that there is an upper bound c* > 0 such that values c > c* would not be considered politically acceptable for the working popula- tion, and institutional arrangements would be changed instead. Similarly, they assume that there is a lower bound r* > 0 such that replacement rates r < r* would lead to a system reform. The set {(c, r) ( c c*, r r*} is the viable region of the policy. Neither c* nor r* can be known with certainty. Alho et al. (2003) consider a range of val- ues for both bounds, in an effort to get a feeling of how likely it is that the indexation rule might survive in future political process. Using these bounds, they try to find a suitable degree of indexation. Table 6.3 replicates some of their results for one value of c*. The limit c* = 0.38 represents a 10 percentage point increase in the contribution rate. Many countries face the prospect of such increases in their payroll taxes, and it is exactly these projections that have led the experts and decision makers to pay attention to the aging problem and seek ways to avoid such increases. The rightmost column of table 6.3 shows that the replacement rate is below 33 percent in all alternatives. The values in bold italics are maximum probabilities of staying in the viable region that can be achieved by choosing optimally, for the specific combinations of c* and r*. For example, should one think that the replacement rate has to be at least 30 percent and the contribution rate must not rise more than 10 percentage points from the current level (or it stays at or below 38 percent), then it is best to stick with the current indexation system where = 0. With this policy, the probability of staying in the viable region is 63 percent. If one would accept a replacement rate as low as 29 percent, choosing = 0.2 would produce a probability of 68 percent of staying in this viable region, whereas sticking with the current policy ( = 0) would have lower probability of being sustainable, or 63 percent. Table 6.3. Joint Probabilities of the Contribution Rate Being Lower Than c* and the Replacement Rate Being Higher Than r* in 2050 c* = 0.38 r* = 0.20 0.25 0.26 0.27 0.28 0.29 0.30 0.33 0 0.63 0.63 0.63 0.63 0.63 0.63 0.63 0 0.2 0.70 0.70 0.70 0.70 0.70 0.68 0.44 0 0.4 0.77 0.77 0.77 0.77 0.72 0.52 0.22 0 0.6 0.83 0.83 0.81 0.71 0.56 0.32 0.13 0 0.8 0.86 0.83 0.72 0.58 0.39 0.23 0.09 0 1 0.90 0.74 0.62 0.47 0.30 0.17 0.07 0 Source: Alho et al. (2003). Note: The table is for c* = 0.38 and selected values of r* and for different values of the indexing parameter . 106 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Other measures for operationalizing sustainability can be developed based on the min- imal r and maximal c over the forecast horizon (2001 to 2050), for example. Yet even the simple analysis that concentrates on year 2050 adds an aspect of realism to the study of sustainability that is absent from statistical summaries such as the expected values, vari- ances, or fractiles. Fertility-Dependent Prefunding The following example is based on the partially funded defined benefit (DB) private sector pension system (TEL) in Finland. In this statutory system, funding does not affect pension benefits at all; it affects only the timing of contributions. This cohort-specific forced saving makes current workers partially pay their own future pensions. It is an open issue whether this increases total saving in the economy or whether people counter it by saving less pri- vately. What it definitely does is to smooth the changes in pension contribution rate that are due to demographic developments. That is the main aim of funding in Finland; the issue is how efficiently this aim is fulfilled. Lassila and Valkonen (2001) showed that to reduce the effect of the expected aging of the population on the contribution rate, increasing the level of prefunding is a sensible policy. Yet there is a clear danger of prefunding too much, in the sense that current work- ers pay unnecessarily high contributions and future workers will face lower contributions. Utilizing fertility data in setting the funding level is a promising approach to increase funding and avoid excesses. The discussion that follows elaborates on this idea. Prefunding in a DB system reduces the risk caused by changing fertility on contribution rates. The risk reduction is obtained, in part, by introducing new risks through uncertain investment performance. Since the funding rate is far from full (no target is set for the funding rate in the Finnish system, but currently it is roughly one-quarter), the funding rules can potentially be improved by considering the future size of the working cohorts. The current prefunding rule is as follows. Every year t, new pension rights accrue at a rate k for each worker. A share a of the present value of the accrued right, to workers aged i = 23, . . . , 54, is put in the funds. The present value is calculated from age 65 to a maxi- mum age, denoted here by M. Let g(t,i) be the labor income of the individual in age i dur- ing year t. For prefunding purposes, the magnitude of this right is evaluated ignoring all future changes due to wage or price developments. An interest rate r is administratively set. Suppose the proportion S(i,j,t) of those in age i at t is expected to be alive at age j. Then, the following amount is prefunded for the worker in age i during period t, M h(i,t)= akg (i,t)S(i, j,t)/(1+ r)j . -i (6.2) j=65 We propose to amend the rule so that, for each funding cohort (those aged 23 to 54), the share funded also depends on the size of the cohort at birth B(t) relative to the size of the later born cohorts. The idea is that we can estimate from the size of recently born cohorts the size of the work force in those future periods when the funding cohort is retired. Or we propose to multiply h(i,t) by i-1 b(i,t)= B(t-i)/ w (j,i)B(t- j -1), (6.3) j=0 where w(j,i) 0 add up to 1 for each i. The weights are calculated so that they approximate the shares of the various cohorts in the working-age population, when the funding cohort DEMOGRAPHIC UNCERTAINTY AND EVALUATION OF SUSTAINABILITY OF PENSION SYSTEMS 107 (those in age i at t) has retired. This fertility effect on funding varies between cohorts, and for each cohort it varies in time. If the funding cohort is bigger than the younger cohorts are, b exceeds unity and thus funding is increased. If fertility increases and younger cohorts are bigger, funding declines compared with current rules. Under current pension rules the contribution rate is expected to stay close its current level of 21.5 percent over the next 10 years, and then rise to about 30 percent. Uncertainty increases with the time horizon, and in the 2060s the 80 percent predictive interval is 12 percentage points wide. These estimates were obtained by simulating the population 100 times, and solving the Finnish OLG model in each case. The demographic population paths were generated as described in Alho (2002). A fertility-dependent funding rule would narrow the predictive intervals from the 2050s onward. The cost of this is that the intervals before the 2050s become slightly larger. Doubling the current funding degree would restrict the future variability sooner, but it would also increase the variability in the near future. Thus funding in general shifts demo- graphic risks in time. Figure 6.5 shows this in the form of standard deviations, calculated each period from the contribution rates in each of the 100 population paths. With a fertility-dependent funding rule, the median contribution would be first slightly higher and later somewhat lower than with current rules, but the distributions would be Table 6.4. Pension Contributions during Selected 10-Year Periods in the Finnish TEL System (fractiles of the predictive distribution) 2005­14 Prefunding rule d1 Q1 Md Q3 d9 Current rules 21.,4 21.6 21.9 22.0 22.2 Fertility-dependent funding 22.2 22.5 22.8 23.0 23.3 Funding doubled 24.8 25.1 25.6 25.9 26.1 2030­39 d1 Q1 Md Q3 d9 Current rules 27.7 28.6 30.0 30.7 31.9 Fertility-dependent funding 27.0 28.4 29.9 31.2 32.1 Funding doubled 28.5 29.4 30.6 31.2 32.3 2050­59 d1 Q1 Md Q3 d9 Current rules 26.8 28.4 30.4 32.7 34.7 Fertility-dependent funding 26.2 28.0 29.7 32.2 33.5 Funding doubled 25.0 26.2 27.8 29.6 31.3 2060­69 d1 Q1 Md Q3 d9 Current rules 26.4 28.5 31.3 35.1 38.4 Fertility-dependent funding 26.2 28.3 30.2 33.0 34.7 Funding doubled 24.2 25.9 28.3 31.2 33.9 Source: Authors' calculations. 108 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 6.5. Standard Deviation in 100 Simulations of Pension Contribution Rate in the Finnish Private Sector TEL System under Alternative Funding Rules 7 6 5 4 percent3 2 current rules 1 doubled funding fertility-dependent funding 0 2000 2010 2020 2030 2040 2050 2060 2070 Source: Authors' calculations. markedly different. The 80 percent predictive interval in 2060­69 would be about 26 to 35 percent instead of 26 to 38 percent. The slight increase in variability in some earlier periods seems a price worth considering for reducing the third quintile value by 2 percentage points and 9th decile value by close to 4 points in 2060­69. Figure 6.6 compares the contri- bution rates under a fertility-dependent rule with those under the current rule in each of the 100 population paths. Funding too little and funding too much are not symmetric cases. Increasing funding requires unpopular decisions of increasing contributions, and smaller increases are easier to accept. Thus in the future, looking back to the present time, we would probably be hap- pier for not prefunding too much. This provides another tool for comparing different funding rules. The simulations suggest that excess funding has taken place if somewhere in the future the contribution rate declines significantly. Different threshold values for "significant" can be used. Table 6.5 shows the share of cases in the 100 population paths where excess funding was observed, with five different threshold values. The unit period in calculations is five years, so qualifying for "excess" requires that there is a pair of five- year periods, not necessarily successive but both within the total period of 2005­69, where the contribution rate in the earlier period exceeds the latter period's contribution rate by at least the threshold value. Table 6.5 shows that small decreases in the contribution rate have a probability of one- third with current pension and funding rules. With fertility-dependent funding, that prob- ability will be slightly larger. But looking at decreases over 2 percentage points, or 3 or 5, note that with fertility-dependent rules, the probability of overshooting in funding would be reduced markedly compared with current rules. A permanent increase in the funding degree does not seem a good choice because it does not adjust to demographics. The gains are there from 2050 onward, but they are partly in the form of very low contributions. The price to pay, higher contributions in the near future, is high. Excess funding would become much more likely. DEMOGRAPHIC UNCERTAINTY AND EVALUATION OF SUSTAINABILITY OF PENSION SYSTEMS 109 Figure 6.6. Pension Contributions in the Finnish Private Sector TEL System Fertility-dependent funding Fertility-dependent funding 65 65 2010­14 2020­24 55 55 45 45 35 35 25 25 15 15 15 25 35 45 55 65 15 25 35 45 55 65 65 65 2030­34 2040­44 55 55 45 45 35 35 25 25 15 15 15 25 35 45 55 65 15 25 35 45 55 65 65 65 2050­54 2060­64 55 55 45 45 35 35 25 25 15 15 15 25 35 45 55 65 15 25 35 45 55 65 65 65 2070­74 2080­84 55 55 45 45 35 35 25 25 15 15 15 25 35 45 55 65 15 25 35 45 55 65 Current rules Current rules Source: Authors' calculations. Note: Each of the 100 dots represents one population path. 110 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Table 6.5. Probability of Excess Funding during 2005­69 Threshold value 1 2 3 5 10 Current rules 35 17 13 5 0 Fertility-dependent funding 44 12 5 0 0 Funding doubled 94 77 56 21 0 Source: Authors' calculations. Note: Share of cases where the pension contribution rate decline, in some subperiod of 2005­69, exceeds the threshold value. Threshold expressed as percentage points. Future Directions The need for sustainable financing strategies is not confined to pensions. Publicly financed health care and long-term care are another area where the tools described above can be used.14 These areas are correlated: demographic paths that are costly from the pension point of view are likely to be costly in terms of health care. The pension examples above seem relevant also for NDC systems, although both Lithuania and Finland have DB pension systems. Longevity adjustment and wage-bill indexation are by nature NDC instruments,15 and funding, at least in buffer form, is required to keep contributions fixed and prevent the necessity of using brake-type arrangements continuously. The Swedish NDC concept has strong strategy features, especially the aim of keeping contributions fixed in all states of the world. It seems like a major improvement compared to the many current PAYG defined benefit systems. Yet the system has not been crash- tested, and it is conceivable that it may not be sustainable in all realistic circumstances. There may also be other systems with comparable risk characteristics. We view the combination of stochastic population simulations and a numerical OLG model as a first step toward a more comprehensive model. Stochastic population simula- tions for all EU countries and some other European countries are being produced in the European Union's fifth framework research project, Uncertain Population of Europe (UPE). In another EU project, Demographic Uncertainty and the Sustainability of Social Welfare Systems (DEMWEL), several European research institutes are working together to create models where future uncertainty is handled in a more advanced manner. But even with current models many aspects of uncertainty in the economic consequences of popu- lation aging can be explored, and that is also the aim in DEMWEL. In future work, it might be useful to extend the notions of sustainability and sustainable strategies into a probabilistic direction, complemented with a "viability" concept with soft or unknown limits of acceptability. A theoretical challenge is to achieve a better concor- dance of the demographic and economic models. Alternative descriptions of how uncer- tainty is taken into account in actual practice, when it is not clear what the relevant decision horizon might be, is one aspect of such work. Creating pension strategies more sustainable than the current ones is important. As Dis- ney (1999) argues, if the future paths turn out to be unsustainable, there are stark choices left: to adjust other public finances or to change the rules ex post. Stochastic simulations with models combining the economic and demographic ingredients of the pension system can be used to crash-test the current systems and reveal the circumstances where their potential weaknesses become crucial. Similarly, the simulations will help to design alter- DEMOGRAPHIC UNCERTAINTY AND EVALUATION OF SUSTAINABILITY OF PENSION SYSTEMS 111 native, and possibly very complicated, pension policy strategies, and test their conse- quences both at the system level and the individual level. Notes 1. "Notional defined contribution" and "non-financial defined contribution" should be understood to have the same definition. 2. See Blanchard et al. (1990, p. 8). 3. See Blanchard et al. (1990, p. 11). 4. See, for example, Rasmusen (1989, p. 17). 5. See Settergren (2001, p.1). 6. See Könberg, Palmer, and Sundén (2006). 7. For more on viability theory, see Aubin (1991). 8. See, for example, Lee (1974). 9. For a description, see Alho (1998) or visit http://www.joensuu.fi/statistics/juha.html. 10. See Alho and Vanne (2001). 11. See Alho et al. (2002). 12. See Jensen and Lassila (2002). 13. See Alho (2003). 14. See Lassila and Valkonen (2004). 15. See Palmer (2000). References Alho, J. M. 1990. "Stochastic Methods in Population Forecasting." International Journal of Forecasting 6 (4): 521­30. ------. 1998. "A Stochastic Forecast of the Population of Finland." Reviews 1998/4. Statis- tics Finland, Helsinki. ------. 2001. "Stochastic Forecast of the Lithuanian Population 2001­2050." Unpublished report, September 25. http://www.etla.fi/lithuania. ------. 2002. "The Population of Finland in 2050 and Beyond." Discussion Paper 826. ETLA (The Research Institute of the Finnish Economy), Helsinki. ------. 2003. "Predictive Distribution of Adjustment for Life Expectancy Change." Work- ing Paper 3, Finnish Centre for Pensions, Helsinki. Alho, J. M., and R. Vanne. 2001. "On Predictive Distributions of Public Net Liabilities." Paper presented at "International Meeting on Age Structure Transitions and Policy Dynamics: The Allocation of Public and Private Resources Across Generations," Taipei, December 6­8. Alho, J. M., S. E. H. Jensen, J. Lassila, and T. Valkonen. 2003. "Controlling the Effects of Demographic Risks: The Role of Pension Indexation Schemes." Discussion Paper 2003-12, Centre for Economic and Business Research, Copenhagen. Alho, J. M., S. E. H. Jensen, J. Lassila, R. Lazutka, A. Morkuniene, and T. Valkonen. 2002. Ż "The Economic Effects of Population Ageing and Demographic Uncertainty in Lithua- nia: Summary, Conclusions and Recommendations." Research report, downloadable from www.etla.fi/lithuania. Aubin, J. 1991. Viability Theory. Basel: Birkhäuser. Auerbach, A. J., and K. Hassett. 2001. "Uncertainty and the Design of Long-Run Fiscal Pol- icy." In Demographic Change and Fiscal Policy, ed. A. J. Auerbach, and R. D. Lee. Cam- bridge: Cambridge University Press. 112 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Auerbach, A. J., and L. J. Kotlikoff. 1987. Dynamic Fiscal Policy. Cambridge: Cambridge University Press. Blanchard, O., J. C. Chouraqui, R. P. Hagemann, and N. Sartor. 1990. "The Sustainability of Fiscal Policy: New Answers to an Old Question." OECD Economic Studies 15 (Autumn): 7­36. Disney, R. 1999. "Notional Accounts as a Pension Reform Strategy: An Evaluation." Social Protection Discussion Paper 9928, World Bank, Washington, DC. Jensen, S. E. H., and J. Lassila. 2002. "Reforming Social Security in a Transition Economy: The Case of Lithuania." Journal of Policy Reform 5 (1): 17­36. Könberg, B., E. Palmer, and A. Sundén. 2006. "The NDC Reform in Sweden: The 1994 Leg- islation to the Present." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 17. Washington, DC: World Bank. Lassila, J., and T. Valkonen. 2001. "Pension Prefunding, Ageing, and Demographic Uncer- tainty." International Tax and Public Finance 8 (4): 573­93. ------. 2003. "Ageing, Demographic Risks and Pension Reform." In Social Security and Pension Reform, ed. Martin Weale, Occasional Paper No. 56, NIESR (The National Insti- tute of Economic and Social Research), London. ------. 2004. "Prefunding Expenditure on Health and Long-term Care under Demo- graphic Uncertainty." Geneva Papers on Risk and Insurance: Issues and Practice 29 (4): 620­39. Lee, R. D. 1974. "Forecasting Births in Post-transition Populations: Stochastic Renewal with Serially Correlated Fertility." Journal of the American Statistical Association 69: 607­17. Lee, R., and L. Carter. 1992. "Modeling and Forecasting U.S. mortality." Journal of the Amer- ican Statistical Association 87 (14): 659­71. Lindbeck, A., and M. Persson. 2003. "The Gains from Pension Reform." Journal of Economic Literature Vol. XLI (March): 74­112. Palmer, E. 2000. "The Swedish Pension Reform Model: Framework and Issues." Social Pro- tection Paper 12, World Bank, Washington, DC. Palmer, E. 2002. "Swedish Pension Reform: How Did It Evolve, and What Does It Mean for the Future?" In Social Security Pension Reform in Europe, ed. M. Feldstein and H. Siebert. NBER conference volume. Chicago: University of Chicago Press. Rasmusen, E. 1989. Games and Information. Oxford: Blackwell. Settergren, O. 2001. "The Automatic Balance Mechanism of the Swedish Pension System: A Non-Technical Introduction." Riksförsäkringsverket (August). Downloadable at http://www.rfv.se/english/publi/index.htm. Valdés-Prieto, S. 2000. "The Financial Stability of Notional Account Pensions." Scandina- vian Journal of Economics 102 (3): 395­417. Vaupel J. W., and J. Oeppen. 2002. "Broken Limits to Life Expectancy." Science 296: 1029­31. Discussion of "Demographic Uncertainty and Evaluation of Sustainability of Pension Systems" Sergio Nisticň* THE CHAPTER BY ALHO, LASSILA, AND VALKONEN succeeds in showing how stochastic pop- ulation simulations could assist the main agents (individuals, firms, policy makers) in dealing with the lack of reliable information about the strains demography is exerting on pension systems. More specifically, the aim of the chapter is to show that it is possible to enforce sustainability of defined benefit (DB) pay-as-you-go systems by providing all agents with a "sustainable policy strategy...[that is,] a set of rules such that both the con- tributors and the pensioners know, beforehand, what will be done in any reasonable future circumstance, and they accept the future actions." The authors rightly emphasize the importance for agents to know "the likelihood of the various risks" because "without any probabilities attached to the alternatives, their importance is suspect and the results are difficult to interpret." This is why stochastic demographic simulations in conjunction with appropriate economic models can certainly help policy makers evaluate the likely impact of alternative policy measures on the key parameters of those pension systems on which demography puts a strain. Although in a pure DB scheme the burden of adjustment entirely falls on contribution rates--in that if benefits are defined there are no means to contain pension expenditure-- in actual (spurious) DB schemes, all parameters can undergo possible revisions: accrual rates (through longevity adjustments), indexation of already awarded pensions, and retirement ages. Alho, Lassila, and Valkonen provide some very interesting stochastic exercises showing the comparative impact on contribution rates in Lithuania of continuing the current policy as opposed to the three alternative parametric reforms. They convinc- ingly suggest that a widespread diffusion of this kind of information can help agents fore- see what policy will actually be enforced in a context in which the "rules of the game" allow for various policy measures. However, it should be added that NDC schemes are precisely designed to provide indi- viduals with that set of rules the authors argue will stem from good demographic infor- mation. As discussed by Gronchi and Nisticň (2006), the NDC scheme is endowed with a sort of automatic pilot that intervenes promptly on expenditure, avoiding the usually pro- tracted waiting period necessary before governments resign themselves to the necessity of sustaining the electoral cost of altering the award parameters and unions assume the responsibility of agreeing to such changes. The NDC's intervention is not only prompt but also, more importantly, perfectly predictable. This contrasts with DB schemes, wherein the outcome of the political process leading to parametric reforms is not only delayed but also uncertain. * Sergio Nisticň is professor of economics at the University of Cassino (Italy) where he teaches Microeconomics, History of Economic Thought, and Social Insurance Economics. 113 114 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Take as an example the argument the authors develop in the section of their chapter focusing on alternative possible adjustments of indexation for pensions already awarded. The authors rightly emphasize that the domain within which indexation will ultimately fall is "constrained" by the contribution rate, which cannot exceed a given ceiling, and the replacement rate, which cannot fall below a given floor. In this respect, the authors show that, within DB schemes, good forecasting of possible demographic scenarios becomes crucial for the identification of a sustainable indexation: that is, for an indexation compat- ible with admissible levels of contribution and replacement rates. Again, the NDC adjust- ment mechanisms solve this dilemma through an automatic, predictable, and fully transparent mechanism. Once the contribution rate is freely chosen, the policy maker has a "lever" to solve the trade-off between the first pension annuity (the replacement rate) and the indexation of future installments. The lever is the return on pensioners' account bal- ances that can be "prepaid" ("imputed," in Swedish terminology): that is, devoted to increasing the first installment. Indexation of subsequent installments will, each year, equal the sustainable return minus the prepaid return. (This is why in chapter 19 the pre- paid return is referred to as the "deviation rate" between the sustainable return and the indexation.) This "endogenous" indexation, which Italy has awkwardly failed to adopt, ensures the sustainability of the NDC scheme. The lower the prepaid return (the replace- ment rate), the better indexation will perform in adverse economic and demographic sce- narios. This is why Sweden has probably prepaid an excessively high return (1.6 percent). The difference between the DB and NDC adjustment mechanisms also has important consequences on the distribution of the burden of adjustment among generations. The authors argue that within a DB scheme "to reduce the effect of the expected aging of the population on the contribution rate, increasing the level of prefunding is a sensible pol- icy." Also on this respect, good demographic forecasting is essential to convince active generations not to shift the entire burden of adjustment on future generations and to pay a contribution rate that is above its equilibrium level. It is hard to ask current workers to fund the system on the basis of "weak" predictions: that is, when the risk of overshooting is not negligible. Funding is generally possible only in the face of good demography: that is, by keeping the contribution rate unaltered when it could be lowered. Again, the authors hit the target when they emphasize that information is essential for fine tuning the policy mix within a DB setting to guarantee, at the same time, both sustainability and fairness (in this case, intergenerational). On the other hand, it is common to misinterpret the ability of NDC systems to be self-sustainable for any given contribution rate, as if pensioners only were asked to bear the burden of balancing the system. According to this view, NDC schemes would suffer from an intergenerational unfairness, which is symmetric to that characterizing pure DB schemes, wherein active workers only bear the burden of adjust- ment. However, one should consider that the prompt intervention of the "automatic pilot" has the twofold form of slower indexation of old pensions and reduction of the contribu- tion balances being formed (thus containing future pension awards), so that the sacrifice is imposed at the same time on pensioners and active workers. Good demographic information is obviously also essential to NDC schemes, though not as a direct support to the provision of alternative policy measures, but rather to pro- vide individuals with some important pieces of information needed to respond to demo- graphic trends. The flexibility in retirement age is one of the most important features of NDC schemes. Postponing retirement age is the main instrument at individuals' disposal to offset the effects of a bad demographic scenario. Good predictions about longevity, when collected, filtered, and synthesized by signals (such as the expected levels of the DEMOGRAPHIC UNCERTAINTY AND EVALUATION OF SUSTAINABILITY OF PENSION SYSTEMS 115 annuitization divisors and of the sustainable return) that are easy to be perceived and interpreted by individuals will play a fundamental role in this respect. The authors conclude by asserting that the NDC system "has not been crash-tested and it is conceivable that it may not be sustainable either in all realistic circumstances." This is true, but no pension scheme can survive in any demographic scenario that is seriously negative. However, when something can be done for public pension schemes to survive, NDC does it in a prompt and fair way; and good demographic forecasting will help. Reference Gronchi, S., and S. Nisticň. 2006. "Implementing the NDC Theoretical Model: A Compari- son of Italy and Sweden." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 19. Wash- ington, DC: World Bank. Chapter 7 The Rate of Return of Pay-As-You-Go Pension Systems: A More Exact Consumption-Loan Model of Interest Ole Settergren and Boguslaw D. Mikula* Paul Samuelson's well-known article, "An Exact Consumption-Loan Model of Interest With or Without the Social Contrivance of Money," published in 1958, has been interpreted as showing that the rate of return of pay-as-you-go (PAYG) pension systems--that is, unfunded pension schemes--is the growth in the contribution, or tax, base of the system. In the absence of technological progress and with a constant number of hours worked per person, the growth in the contribution base is equal to the growth in population, or what Samuelson calls the "biological interest rate." Several researchers have pointed out that the two-age overlapping generation (OLG) model used by Samuelson cannot explain the dynamics of the equilibrium interest rate in a world of more than two-age OLGs. Arthur and McNicoll (1978) and Willis (1988) have demonstrated that in a more than two-age OLG model, changes in the differential between the ages at which the average income is earned and consumed is a critical factor in deter- mining equilibrium interest rates. Likewise, Keyfitz (1985, 1988), as well as Lee in numer- ous works (1980, 1988a, 1988b, 1994a, 1994b, 2000), showed that the amount consumed at some or all ages is affected by changes in this age differential. However, statements that the rate of return on PAYG financing is equal to the growth in the contribution base are surprisingly common.1 Rarely are such claims accompanied by the necessary qualification that they are valid only in a two-age OLG model or in the equally unrealistic case of an economy and demography characterized by a steady state.2 The common assumption that the rate of return of PAYG pension systems is equal to the growth in the contribution base is rarely an efficient simplification. Recent experience in * Ole Settergren is the director of the Pensions Department at the Swedish Social Insurance Agency and Boguslaw D. Mikula is a researcher in the Division for Analysis at the same agency. We are grateful to Sergio Nisticň and Edward Palmer for their valuable comments. Special recognition goes to Hans Olsson. Without his early support, experience, and hard work preceding the decision by parliament in 1998, when it settled for an asymmetric design of the indexation of the new system's pension benefits and implicit pension liabil- ity, the ideas presented here would probably not have survived their infancy and become Swedish legislation. With the permission of the World Bank, this chapter was published in the Journal of Pension Economics and Finance, volume 4, issue 2, July 2005. 117 118 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Sweden indicates how poor that assumption can be. The increase in life expectancy in Sweden from 1980 to 2001 made the income-weighted average age of retirees increase from 72 to 75 years, while the income-weighted average age of contributors to the system remained relatively stable at 43 years.3 Thus the increase in life expectancy made the dif- ferential between the average age at which contribution was paid to the system and the average age at which pension was paid from it grow from about 29 to 32 years. This repre- sents close to a 12 percent increase, which added roughly 0.4 percentage units to the annual real rate of return on contributions to the PAYG system during this period. As the average annual growth in the contribution base over the same period was only 0.3 percent, the common simplification that the rate of return equals the growth in the contribution base revealed less than half of the true return. Thus one consequence--and perhaps a counterintuitive one--of increases in life expectancy is that it adds to the rate of return of PAYG pension systems. This indicates an even more serious drawback with the simplified view than its low precision. The assump- tion that the rate of return of PAYG financing is the growth rate of the contribution base hides a structure vital to understanding the financial dynamic of PAYG pension systems. Possibly the simplification is so frequent because it is assumed that, without it, the analy- sis becomes prohibitively complex for making statements about the system's cross-section internal rate of return (IRR). The aim of this study is to demonstrate that there is a simple method for estimating the cross-section internal rate of return on contributions to PAYG pensions systems, even as the two-age OLG restriction, and the steady state restriction, is canceled. The method entails a procedure for valuing the contribution flow of PAYG financing and identifies the complete set of factors that decides the cross-section IRR. The procedure makes it possible to apply the algorithm of double-entry bookkeeping in PAYG pension systems. The method presented here for calculating the rate of return of PAYG pension financing, including the valuation of the contribution flow and the use of double-entry bookkeeping in a PAYG context, is a result of the research undertaken to deal with some conflicting ambitions of the new Swedish pension system.4 This method for solving, or rather manag- ing, this conflict was reached in ignorance of the research cited above by Arthur and McNicoll, Willis, Keyfitz, and Lee.5 In this text, the phrase cross-section IRR is used to indicate a measure distinct from the more familiar longitudinal IRR measure, which informs the rate of return that equates the value of the time-specific contributions with the benefits to an individual or a group of individuals. The cross-section IRR is the return on the pension system's lia- bilities that keeps the net present value of the pension system unaltered during a period of arbitrary length. However, to derive the cross-section IRR, a continuous time model is used. The expression cross-section internal rate of return is abbreviated below to rate of return, while we sometimes use the abbreviation IRR. We also use the terms contribution base, contribution rate, and contributions, where some would prefer tax base, tax rate, and taxes. In the next section, a method for estimating the value of the contribution flow to PAYG pension systems is presented. In the following section, this method is used to obtain a for- mula for calculating the cross-section IRR on contributions to such systems. The final sec- tion concludes by commenting on the results. In annex B, the methods used to value the contribution flow and the definition of the cross-section IRR are illustrated by means of simple numerical examples. Some readers will probably find it helpful to read the numer- ical examples before they read the second and third sections. THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 119 The Value of the Contribution Flow PAYG financing implies that the flow of future contributions is used to finance a pension liability that has already accrued.6 It is probably a matter of personal preference whether one considers that a PAYG system, by definition, has a deficit equal to this liability, or accepts that its net present value is zero if contributions match pension payments. Here the latter view is taken and financial balance is defined as Assets ­ Liabilities = 0. (7.1a) This standard definition of financial balance is unconventional for PAYG pension systems. The commonly performed projections of cash flows to and from PAYG pension systems used to evaluate their financial situation have traditionally not been presented in the form of assets and liabilities, as the methods used does not allow this.7 As already indicated, it seems reasonable to consider that a PAYG pension system whose contributions and bene- fits match have a zero net present value and subsequently to conclude that its liability is matched by an implicit asset, below referred to as a contribution asset. In another context, Lee (1994 and later) calls a corresponding concept transfer wealth. Often PAYG systems are considered to be defined by the absence of any funded assets. In practice, however, there is normally a transaction account, and sometimes there are sub- stantial funded assets. Systems without any funded assets are only a special case of the general description that follows. Hence equation 7.1a can be re-expressed as CA(t) + F(t) ­ PL(t) = 0, (7.1b) where CA = contribution asset, F = buffer fund, PL = pension liability, and t = time. In a steady state, contributions will equal pension benefits, thus CA(tss) = PL(tss), and F(tss) = 0. For each income and mortality pattern and set of pension-system rules, there is a unique value for the pension liability. Equations 7.2 through 7.4 give an expression for this value in steady state. In the case of a stable population--that is, a population with constant mortality rates and constant population growth--the age distribution of the population can be expressed as N(x) = N(0) l(x) e­ , x (7.2) where N(x) = number of persons of age x, x = age, = the rate of fertility-driven8 population growth, and l(x) = life table survival function. In this system, the indexation of benefits can have any relation to the growth in average wage. Thus the pension benefit may vary in size relative to this average wage at different 120 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES ages. If, for example, pensions are indexed by the change in consumer prices, and average wages grow at a faster rate, the average pension benefit per birth cohort will be lower for older cohorts than for younger ones. The distribution of pensions within a cohort is ignored, since it has no relevance for the system-level rate of return. The pension liability, V, is defined as the present value of future pension benefits to all persons to whom the system has a liability at the time of evaluation minus the present value of future contributions by the same individuals, m m V = population(x) PV[pensions(u)- contributions(u)]du dx , (7.3) 0 x where m = maximum age, and x, u = age; both are variables of integration. Discounting payments to and from the pension scheme by the growth in the contribu- tion base, the pension liability can be re-expressed as: ageu m m averagepension, V = N(0)l(x)e- x l(u) e- (u-x) u (7.4) l(x) 0 population, agex x discountinng k W epensionpayments R(u) - cW W(u) du dx , contributions survivorrate where W(x) = wage pattern--that is, the average wage for age group x, as a ratio of the average wage for all age groups, W = average wage in monetary units per unit of time, c = required contribution for a financially stable PAYG pension system, = the rate of pension indexation relative to the rate of average wage growth, R(x) = number of retirees as a ratio of the number of individuals in age group x, and k = constant determining the pension level (equals the replacement rate if = 0). The rate of discount is the product of the growth in average wages times the rate of pop- ulation growth. As both wages and benefits grow with the average wage growth, the growth in average wage falls out of the equation and leaves the population growth rate as the effective discount rate, . It would be inappropriate to use a market rate of return on capital as a discount rate. The return on capital has no impact on the financial balance of a PAYG pension system, disregarding its effect on the buffer fund if there is one. For a stable population with stable income patterns, the contributions, C, are generated by the size of the population by age, N(x); the wage pattern, W(x); the average wage,W; and the required contribution rate for a financially stable system, c. m C = N(x)cWW(x)dx. (7.5) 0 THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 121 In a steady state, the contribution rate that satisfies the financial-stability criteria of equation 7.1 is also the contribution rate that makes contributions equate with pension payments in every period. Thus c is calculated as m m N (0)l(x)e- x k W e R(x) dx = N(0)l(x)e- u x cW W(x) dx , 0 0 pensionppayments contributions me-(-)x l(x) R(x) dx c = k 0 . (7.6) m e-x l(x)W(x) dx 0 It is possible to get a form of measurement of the pension liability, in steady state, that is independent both of the size of the contribution base and of the contribution rate simply by dividing the pension liability by contributions paid per time unit. Thus equation 7.4 is divided by equation 7.5, where equation 7.6 is substituted for c. Rearranging and integrat- ing by parts, this simplifies to m -)x m x V = 0 xe-( l(x) R(x) dx - 0 xe- l(x)W(x) dx . (7.7) C m m x 0e-( -)x l(x) R(x) dx 0 e- l(x)W(x) dx averageageof retirees averageageof contribbutors The intermediate steps of the simplification are presented in annex C. Equation 7.7 informs the conceivably intuitively reasonable fact that in steady state, the liability divided by contributions is equal to the time differece between the average age of retirees (the first term of the right-hand side), and the average age of contributors (the sec- ond term of the right-hand side). Both ages are money-weighted. However, this is not evi- dent from the expression, as the average wage is a part of contributions, C. This leaves equation 7.7 with only the age patterns. The age difference between the average contribu- tor and retiree is a measure of the duration of the pension liability. We call it turnover dura- tion (TD). V = Ar - Ac =TD (7.8) C where Ar = money-weighted average age of retiree, and Ac = money-weighted average age of contributor. The top half of figure 7.1 illustrates the age structure for average wage, W(x); a certain life-table, l(x); a population growth trend, ; and rules for indexation of pensions, . It also illustrates the retirement pattern, R(x). The bottom half of figure 7.1 illustrates the result- ing age structure for the contribution base and the pension payments. The resulting age- differential between the money-weighted average age of retirees, Ar and contributors, Ac is also shown. Hence, for a stable population with stable income patterns, the factors determining the size of the pension liability can be separated into a volume component that is contributions 122 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 7.1. Illustration of Equations 7.7 and 7.8 W(x) l(x) R(x) e - .x e- ( - ).x 0 10 20 30 40 50 60 70 80 90 100 Ac < 0 & > 0 TD Ar e - .x.l(x).R(x) e- ( - ).x.l (x).R(x) 0 10 20 30 40 50 60 70 80 90 100 Source: Authors. and a structural component that is the turnover duration. Turnover duration is a useful con- cept that sums up the factors that determine the scaleless size of the pension liability: scale- less refers to disregarding the size of contribution rate and contribution base. The present value of the pension liability for a stable population with stable income patterns is expressed in years of contributions: V =TD V =TDC. (7.9a) C The separation of the steady state pension liability into a volume and a structural com- ponent also has a temporal aspect. Except in a steady state, there will be no definite value for the turnover duration; however, the current economic and demographic patterns can be used to measure the expected turnover duration. It is expected in the same sense as the common measure of life expectancy; that is, it uses current observations to calculate a value that will turn out correct ex post only if observed patterns remain constant. The probability that any generation will live according to any published life table is virtually zero. Nevertheless, life tables are relevant and useful. Repeated estimations of the expected turnover duration9 will reflect the changes in the financially relevant patterns and thus yield new estimates of the contribution asset, which are infinitely unlikely to pro- duce the ex post correct figure. This procedure of repetitive revaluation of the contribution asset is not so different from the recurring revaluation of funded assets by the market.10 For these reasons we find it appropriate to define the value of the contribution flow as the current turnover duration times the current contributions: CA(t) = TD(t) C(t). (7.9b) THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 123 The turnover duration indicates the size of the pension liability that the present con- tribution flow can finance, given the present income and mortality patterns and the population growth rate. As economic and demographic patterns change, the new value of the contribution flow can be estimated. The inverse of the turnover duration is a computable discount rate for the contribution flow, a measure of the current internal time preference of the PAYG pension scheme. This time preference is a function of the system design with respect to the rules that govern the indexing of pensions, the income and mortality patterns of the insured population, and the population growth trend. Annex C provides rough estimates of the turnover duration for 41 countries from Settergren and Mikula (2001). The country-specific turnover duration varies from 31 to 35 years; thus, with the internal time preferences of the hypothetical pension system in the estimate, the discount rates for contributions vary between approximately 2.8 and 3.2 percent. These rates are interestingly close to the frequently assumed real interest rate of about 3 percent. The usefulness of the turnover duration for valuing the contribution flow is critically dependent on its volatility. In many countries, perhaps most, the volatility of the turnover duration can be anticipated to be moderate to low. The stem-and-leaf exhibit in figure 7.2 presents estimates of the annual percentage change in the turnover duration in Sweden in the period 1981­2003.11 The average increase was 0.4 percent, most of it attributable to the increase in life expectancy; the average, money-weighted age of contributors remained closely around age 43, with no clear trend. The maximum one-year increase in turnover duration was 2.1 percent; the maximum annual decrease was 0.5 percent. More than half, 12, of the annual changes were between zero and 0.5 percent, and the standard deviation of the 23 observations was 0.6. In the next section, the above method for estimating the value of the contribution flow is used to derive an expression for the rate of return of PAYG pension systems, and the application of double-entry bookkeeping for such systems is outlined. Figure 7.2. Turnover Duration in Sweden, 23 Annual Changes, percentages 1981­2003 2 1 1 8 1 11 0 865 0 4443220000 ­0 234 ­0 5 The stem-and-leaf diagram is read as follows: ... ... 1 11 = 1.1% and 1.1% 0 865 = 0.8%, 0.6% and 0.5% etc. Source: Riksförsäkringsverket (2003, 2004). 124 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES The PAYG System's Rate of Return Financial balance can be ensured by adjusting the size of the pension liability: that is, by adjusting the value of present and or future benefits, or by adjusting the contribution rate--the size of the contribution asset as defined by equation 7.9b--or by doing both. Irre- spective of the type of tuning, the financial-balance requirement of equation 7.1b, a net present value of zero, applies. To continue the derivation of the IRR, equation 7.1b is rephrased to TD C + F ­ PL = 0, (7.10) where F = buffer fund, and PL = pension liability. Equation 7.10 implies that both negative and positive funded assets are allowed, and in some situations necessary, to comply with the definition of financial balance.12 Annex B presents numerical examples that illustrate this point. The rate of return of the pension liability that yields a net present value of zero is, by definition, the rate of return on contributions to the system. The formula for the rate of return of a PAYG pension system follows from differentiating equation 7.10 with respect to time d(TDC + F - PL) = TD ddtC + - d PL = 0. (7.11) dt dTD C + ddtF dt dt The change in pension liability is a function of the rate of return of the liability and of the difference between payment of contributions and disbursements of pensions, as shown in equation 7.12 d PL = PLIRR +(C- P) , (7.12) dt where IRR = internal rate of return, and P = pension payments in monetary units per unit of time. The IRR may turn up in two different forms: implicit and explicit. The implicit rate of return is a function of the impact of changes in mortality on the pension liability, and of any divergence between new pension obligations and contributions paid. In addition, changes in the rules of the system will normally alter the value of the pension liability, pro- ducing an implicit effect on the IRR. The explicit rate of return is the result of any explicit rules for indexing the liability--that is, the benefits to present and future retirees. The net difference in payments to and from the pension system is captured by the buffer fund, if there is one. In addition, the value of the fund is changed by the return on its assets, as shown in equation 7.13 dF = Fr +(C- P), (7.13) dt THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 125 where r = rate of return on the buffer fund. Depending on its sign and magnitude, the return on the buffer fund may increase or decrease the rate of return of the PAYG pension system. Equation 7.11 can then be re- expressed as TD ddt + C dTD C + Fr- PLIRR = 0 . (7.14) dt Finally, the IRR, separated into its components, is TD ddt C dTD C IRR = + dt + Fr . (7.15) PL PL PL i ii iiii Thus the rate of return of a PAYG system is a function of: i Changes in contributions This component consists of Samuelson's biologi- cal interest rate, changes in labor force participa- tion, average wage growth, and changes in the contribution rate. ii Changes in turnover duration This component consists of changes in income and mortality patterns and in the fertility-driven growth rate of the population.13 iii Buffer fund return (interest) This component consists of the return (interest) on any assets (debt) in the system. The part of the IRR that is caused by mortality changes and any divergence between new pension obligations and contributions paid, or by changes in the system rules, can be considered an implicit indexation of the pension liability. The IRR reduced by the rate of implicit indexation is the rate of available indexation of the pension liability. Thus, Rate of available indexation = i + ii + iii ­ rate of implicit indexation. (7.16) In practice, the indexation rules, or the adjustment of the contribution rate or other sys- tem rules, are not necessarily such that they distribute all the available indexation each time period. The applied indexation will differ temporally from what is available. The dif- ference is the net income or loss to the system during the measured time period. Rate of available indexation ­ rate of explicit indexation = system net income. (7.17) The accrued value of such net income or losses gives the opening surplus or deficit for the next period. How does the expression for the cross-section IRR of PAYG pension systems in equation 7.15 relate to the longitudinal IRR on contributions? For individuals, this calculation becomes possible at the time of death. For a birth cohort, it is possible when everyone in 126 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES the cohort has died; for the pensions system, it is possible when it has been closed down. Such delays of information are, indeed, impractical. Both participants and policy makers want regular information on financial position and development. In order to produce this information, cross-section rate of return measures must be adopted, for which there is only imperfect information. In the business world, the problem is similar. The true rate of return can be calculated only when all payments to and from the business entity have been made. As business stakeholders need regular information on the rate of return, accounting principles are developed for an ongoing business. It is well known that accounting mea- sures of the rate of return--basically business net income--are subject to some degree of arbitrariness. Therefore, the preferred choice of method is a matter for debate. For PAYG pension systems, it is possible to envisage other accounting procedures than the one described here, and other measures will normally yield a different rate of return for a specific period. By our method, the contribution flow is valued according to the turnover duration with cross-section observations at the time, while the pension liability is estimated with an actuarial projection that may or may not imply changes in future turnover duration. Such differences will have an impact on the trajectory of the measured rate of return, but not on the aggregate rate of return as the system approaches a hypothet- ical steady state. Conclusions The rate of return on contributions to PAYG pension systems is not only a function of the growth in the contribution base of the system. It is also a function of changes in income and mortality patterns and in the trend of population growth. These three factors cause changes in the average age at which contributions are paid and pensions received: that is, they cause changes in turnover duration. Further, if there is a buffer fund in the system, the return on that fund will influence the rate of return on contributions. The rate of return can be implicitly distributed through the effects on the pension liability from mortality changes and also by differences between contributions paid and new pension liabilities. The difference between the rate of return and the implicitly distributed return is the rate of available indexation--the explicit indexation of the pension liability, which must be made to keep the systems net present value unaltered. The turnover duration provides an estimate of the discount rate for the contribution flow to systems, which finances obligations with a zero prefunding requirement, that is, PAYG systems. This makes it possible to apply a form of double-entry bookkeeping. With the double-entry algorithm, the financial position of these schemes can be reported by means of a balance sheet, as summarized in equation 7.10, and changes in the financial position can be reported by means of an income statement, summarized in equation 7.17.14 We would argue that extending the field of double-entry bookkeeping to PAYG pensions systems has the potential of improving the quality and transparency, and thus the under- standability, of financial information on these important transaction systems over the dif- ferent measures of actuarial balance used today. Disentangling the components of the rate of return also adds options for the design of PAYG pension systems. In particular, the forms of indexing pensions can be given a more efficient design.15 THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 127 Annex 7A. A Bibliography of the Legislative History of the Income Index and the Automatic Balance Mechanism Settergren, O. With the assistance of H. Olsson. 1997. Kapitel 5 Inkomstindex "Inkomst- grundad ĺlderspension--finansiella frĺgor m.m.," Ds 1997:67. Socialdepartementet, Stockholm. (Chapter 5 The Income Index in "Income-Related Old-Age Pension-- Financial Issues etc." Ds 1997:67, Ministry of Health and Social Affairs, Stockholm.) ------. 1998. Chapter 16 Inkomstindex, "Regeringens proposition 1997/98:151 Inkomst- grundad ĺlderspension, m.m.," Riksdagen, Stockholm. (Chapter 16 The Income Index in "The Government Bill 1997/98:151 Income-Related Old-Age Pension etc." Swedish Parliament, Stockholm.) ------. 1999. With the assistance of H. Olsson and D. Sundén. "Automatisk balansering av ĺlderspensionssystemet--regler för avsteg frĺn inkomstindexeringen inom ĺlderspen- sionssystemet," Ds 1999:43. Socialdepartementet, Stockholm. (Automatic Balancing of the Old-Age Pension System--Rules for Deviating from the Income Indexation within the Old-Age Pension System." Ds 1999:43, Ministry of Health and Social Affairs, Stockholm.) ------. 2000. With actuarial modeling and calculations by B. D. Mikula, N. Holmgren, and J. Leander. "Automatisk balansering av ĺlderspensionssystemet--redovisning av regeringens beräkningsuppdrag," RFV Analyserar 2000:1 Riksförsäkringsverket, Stockholm ("Automatic Balancing of the Old-Age Pension System--An Account of the Calculation Assignment from the Government." RFV Analysis 2000:1). ------. 2001. "Regeringens proposition 2000/01:70 Automatisk balansering av ĺlderspen- sionssystemet," Riksdagen, Stockholm. (The Government Bill 2000/01:70 Automatic Balancing of the Old-Age Pension System." Swedish Parliament, Stockholm.) 128 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Annex 7B. Numerical Illustrations in an Overlapping Generation Model A three-age overlapping generation (OLG) model is used to illustrate the impact of changes in the average ages at which income is earned and consumed. Three is the mini- mum number of ages needed for changing the differential between the ages at which the average income is earned and consumed. This age differential is called turnover duration, and is formally derived in equations 7.1­7.9. To demonstrate the effects of changes in mor- tality on the rate of return, the model is extended from three to four ages. In the model, the life of an individual is divided into three (four) periods of equal length. All individuals work for exactly two periods, at ages 1 and 2, and they are all retired for the entire third (and fourth) period, age 3 (and 4). All are born on the first day of each period; all birth cohorts are of equal size; there is no fertility-driven population growth, no migration, and no preretirement mortality. And everyone in retirement dies on the last day of their final period. There is no technological progress. Under these assump- tions, the contribution base for the pension system is constant. All financial transactions are made at the end of each period. To avoid the complication that changes in contribution rate have on the internal rate of return (IRR) (see equation 7.15) the examples are con- structed so that the system in all examples can finance its pension payments with an unchanged contribution rate--25 percent--for every period in all examples. The effects on IRR from shifts in income and mortality patterns are described for certain alternative pension-system rules. The reason for this is to illustrate that: · The system's cross-section IRR is not a function of system design1 · The distribution of the IRR over cohorts, the "longitudinal IRR,"2 is a function of sys- tem design · The timing of cash flows is a function of system design, even when designs are equally financially stable in the sense that they all produce a zero net present value as defined in equation 7.10. Although the numerical examples are straightforward, the somewhat complex feature of OLG, in combination with the detailed account of the effects of the shifts in income and mortality patterns, may make it tedious to work through the examples. However, this effort can be well invested as the examples, once grasped, clearly reveal structures that are vital for understanding important aspects of PAYG financing. Example 1. A Shift in Income Pattern Summary of what the example illustrates. In the following example, the income pattern shifts--the income of older workers increases relative to that of younger workers, so that the income-weighted average age of contributors increases. The example shows how this change decreases the turnover duration and leads to a negative IRR. The effects of the neg- ative IRR are illustrated for a pension system where the rules are such that this specific shock will result in a rate of implicit indexation equal to the negative IRR. In example 1.1 below, the effects of the same shift in income pattern are illustrated for a system where the rules are such that this specific shock will result in a rate of implicit indexation equal to zero. Thus in example 1.1, to maintain a zero net present value, the negative IRR must be distributed through explicit indexation equal to the IRR. The subsequent effects on the cash flow, buffer fund, and so on of the system are illustrated with an income statement and a balance sheet. The shift in income pattern. Up until and including period 1, the wage is 48 for the older working cohort and also 48 for the younger. In the period 2, the income pattern is THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 129 changed.3 From then on, the wage is 72 for the older cohort and 24 for the younger. Thus the wage sum, equal to the contribution base, is constantly 96. The average wage for work- ers in general also remains unaltered; only the distribution of the average wage between the age groups has shifted. The rules of the pension system. The pension system is designed to pay a benefit that is 50 percent of the gross average wage of all wage earners--admittedly an awkward rule, but here it serves our purpose. The effects of the shift in income pattern. Table 7B.1 illustrates that this system will result in contributions of 24 that perfectly match pension benefits of 24 before and after the shift in income pattern. Cohort B, the only cohort whose lifetime income is altered by the change in income pattern, will receive a pension of 24, whereas it paid contributions of 30, the sum of 25 percent of wages 48 and 72, respectively. As the pension received is only 24, this cohort will receive 6 less than they paid, a periodically compounded rate of return of roughly ­15 percent. The computation is 0.25 × 48 × r2 + 0.25 × 72 × r = 24 r ­ 1 ­15%. The effect of the change in income pattern on the system's cross-section rate of return is the monetary effect, ­6, relative to the systems pension liability4 of 36. Thus the cross-section rate of return is ­1/6. Table 7B.1 shows that the cross-section rate of return is equal to the rel- ative decrease in the money-weighted average difference in time between the payment of contributions and the collection of benefits, that is, the decrease in turnover duration from 1.5 to 1.25. From table 7B.2 it is also clear that the change in turnover duration makes the con- Table 7B.1. Effect of a Shift in Income Pattern on Cohort Contributions and Benefits Period 0 1 2 3 4 Cohort total Cohort Contributions Pensions A 48 48 24 24 24 B 48 72 24 30 24 C 24 72 24 24 24 D 24 72 24 .. Wage sum .. 96 96 96 .. Period Contribution 25% 25% 25% 25% 25% total rate Contributions .. 24 24 24 .. Pensions .. .. 24 24 24 Source: Authors. Note: The central box shows wage sums in normal type and pensions in bold face, per period, for each cohort. 130 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Table 7B.2. Effect of a Shift in Income Pattern on Turnover Duration and Pension Liability Before shift After shift Relative change Average age of retiree,AR 3a 3a -- Average age of contributor,AC 1.5b 1.75c +1/6 Turnover duration, TD, (AR ­AC) 1.5 1.25 ­1/6 Contribution asset, TD × contributions 36 30 ­1/6 Pension liability, PL 36d 30e ­1/6 IRR (Monetary loss / PL)f ­6/36 ­1/6 Source: Authors. Note: In explanations b, c, d, and e, contributions are shown in normal type, pensions in bold face, ages in italics. For explanation of pension liability, brackets [ ] are used to group money flows from and to the same cohort. Figures relating to cohorts are presented in order from oldest to youngest. a. All pensions are paid at age 3. b. (48 × 2 + 48 × 1) / (48 + 48). c. (72 × 2 + 24 × 1) / (72 + 24). d. [24] + [24 ­ 12]. e. [24] + [24 ­ 18]. f. The monetary loss occurs "at" the shift, indicated by the placement of the number between the "before shift" and "after shift." tribution asset, which is calculated as the contribution flow times the turnover duration, decrease and that this decrease is equal to the monetary loss incurred by cohort B. As a combined effect of the shift in income pattern and the rules of this pension system, the pension liability decreases to the same extent that the value of the contribution flow is reduced by the shorter turnover duration. Before the shift the pension liability was 36; after the shift the pension liability is 30. Owing to this implicit negative indexation of the pension liability, the net present value of the system is consistently zero throughout the shift. The shift in income pattern in combination with the rule that says that pensions are 50 percent of average income of all wage earners implicitly distributes the negative IRR to cohort B. However, the negative IRR itself was not a consequence of the system's rules, as will be illustrated in the following example. Example 1.1 Same Shift in Income Pattern, Different Pension System Rules The rules of the pension system. The same shift in income pattern is now applied in a pen- sion system designed as a so-called notional defined contribution (NDC) system. The rules of such a system imply that each cohort is to be repaid an amount equal to their contribu- tions indexed at some rate, positive or negative. Initially, the indexing rules of the system are assumed to provide that notional pension capital and pensions are to be revalued at the growth rate of the contribution base, which in the example is zero for every period. The effects of the shift in income pattern. Up until and including cohort A and period 2, this system will yield an identical result as the first set of rules: zero cross-section and longitu- dinal IRRs. But when cohort B retires, it will have accumulated a notional pension capital of 30, equal to what it has paid in contributions. As the flow of contributions is constant at 24, the system can only repay cohort B their notional pension capital by incurring a deficit of 6--a figure familiar from example 1. This deficit is caused by the same reduction in turnover duration as in example 1. However, in the NDC system the same negative IRR causes a cash deficit since the rate of (implicit) indexation is zero, while in Example 1 the rate was ­1/6, thus distributing the negative IRR. THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 131 The shift in income pattern does not immediately reduce the pension liability of the notionally defined contribution system; this liability remains at 36 in period 2,5 while the shorter turnover duration--just as in Example 1--has decreased the value of the contribu- tion flow to 30. To be financially stable, the notional defined contribution pension system must explicitly distribute the negative IRR by reducing the pension liability. One way to accomplish this is to index the system's total pension liability by the "rate of available indexation," (see equation 7.16). Table 7B.3 shows the development of the income state- ment and balance sheet of the NDC system, which applies explicit indexation at the avail- able rate, here equal to the IRR. Indexing cohort B's and C's notional pension capital of 30 and 6,6 respectively, by the available rate of 5/6 reduces it to 25 and 5, respectively. Thus the total pension liability of Table 7B.3. Example 1: Income Statement, and Balance Sheet Period 1 2a 2 3 4 Income statement Contributions 24 24 24 24 24 Pensions ­24 ­24 ­24 ­25 ­23 = net cash flow (a) 0 0 0 ­1 1 Change in contribution asset (b) 0 ­6c ­6c 0 0 New accrued pension liabilityb ­24 ­24 ­24 ­24 24 Paid-off pension liabilityb (= paid pension benefits) 24 24 24 25 23 Indexation of liabilityb 0 0 6 0 0 = change in pension liability (c) 0 0 6 1 ­1 Net income/ ­loss, (a) + (b) + (c) 0 ­6 0 0 0 Balance sheet Buffer fund 0 0 0 ­1 0 Contribution asset 36 30 30 30 30 = total assets (d) 36 30 30 29 30 Pension liability, age 3 0 0 0 0 0 Pension liability, age 2 24 30 25d 23e 24 Pension liability, age 1 12 6 5e 6 6 = total liability (e) 36 36 30 29 30 Net present value of system (d) ­ (e) 0 ­6 0 0 0 Source: Authors. a. Values before indexation with the available rate of return. b. A negative number denotes an increase in the pension liability and thus a cost. A positive number denotes a decrease in the pension liability and thus an income. c. ­0.25 × 24 = ­6 {change in TD × [contributions (t) + contributions (t­1)]/2}. d. 12 + 18 × 5/6 = 25 [cohort B's period 1 contribution + cohort B's period 2 contribution × IRR]. e. 6 × 5/6 + 18 = 23 [cohort C's period 2 contribution × IRR + cohort C's period 3 contribution]. 132 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES the system is reduced from 36 to 30--equal to the new shorter turnover duration of the system (1.25) times the contribution flow (24). This implies that the system has regained its zero net present value. Nonetheless, the shift in income pattern and the negative indexa- tion of the pension liability will affect the system's cash flows. Period 3 pension payments to cohort B will be 25. As system income is 24 every period this will result in a deficit of 1. In period 4, pension payments to cohort C will be 23 (5 + 18); thus a cash flow surplus of 1 will arise and close the deficit.7 The systems total assets period 3 are 29, that is the sum of the buffer fund period 3 is ­1 and the contribution asset is 30. The total assets are equal to the system's pension liability, and the system's net present value is consistently zero. Example 2. A Shift in Mortality Pattern Summary of what the example illustrates. In example 2 the mortality pattern changes--life expectancy shifts upward--so that the money-weighted average age of retirees increases. The example shows how this change increases the turnover duration and results in a pos- itive IRR. The effects of the positive IRR are illustrated for a pension system where the rules are such that the rate of implicit indexation equals the positive IRR. In example 2.1 the effects of the same shift in mortality pattern are illustrated for a system where the rules are such that the rate of implicit indexation is zero. Thus in example 2.1, to maintain a zero net present value, the positive IRR must be distributed through explicit indexation equal to the IRR. The subsequent effects on the cash flows, buffer fund, and so on of the system are illustrated with an income statement and a balance sheet. The shift in mortality pattern. The shift in mortality occurs--simply, though unrealisti- cally--through a one-time increase in life span. After one period of retirement, no retiree in cohort B dies; instead, after the third period, all retirees in the cohort continue to live for exactly one more period. Subsequent cohorts also live for exactly two periods as retirees. The rules of the pension system. In the example, we keep the contribution rate fixed at 25 percent. Thus average pension benefit must be halved after the first cohort with a longer life expectancy received its first pension payment. Cohort B's pension is thus 24 in their first period as retirees and 12 in their second. Cohort C, the second cohort with a longer life span, will receive a pension of 12 in each period, as will subsequent cohorts. The effects of the shift in mortality pattern. Table 7B.4 illustrate that the system will result in contributions of 24 that perfectly match pension benefits of 24 before and after the shift in mortality. However, cohort B, the first to benefit from the longer life span, will receive a total pension of 36, whereas it paid only 24 in contributions, for a periodically com- pounded rate of return of approximately 25 percent. The computation is 0.25 × 48 × r3 + 0.25 × 48 × r2 = 24 × r + 12 r ­ 1 25%. The effect of the change in mortality pattern on the system's cross-section rate of return is cohort B's monetary gain, 12, relative to the systems pension liability 36. Thus the cross- section rate of return is 1/3. Table 7B.5 shows that the cross-section rate of return is equal to the relative increase in the money-weighted average difference in time between the pay- ment of contributions and the collection of benefits--that is, the increase in turnover dura- tion from 1.5 to 2. The reason for the positive return is the longer time span between the average wage-weighted age of contributors and the average benefit-weighted age of retirees resulting from the shift in mortality pattern, the increase in turnover duration. With the longer turnover duration, the value of the contribution flow increases from 36 to 48. The system is financially balanced throughout the shift since the pension liability increases to the same extent as the value of the contribution flow. The positive return of 12 is implicitly distributed to the cohort whose initial pension was calculated on the basis of THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 133 Table 7B.4. Effect of a Shift in Mortality on Cohort Benefits Period ­1 0 1 2 3 4 Cohort total Cohort Contributions Pensions A 48 48 24 24 24 B 48 48 24 12 24 36 C 48 48 12 12 24 24 D 48 48 12 24 ... Wage sum .. 96 96 96 .. .. Period Contribution 25% 25% 25% 25% 25% 25% total rate Contributions .. 24 24 24 .. .. Pensions .. .. 24 24 24 24 Source: Authors. Note: The central box shows wage sums in normal type and pensions in bold face, per period, for each cohort. Table 7B.5. Effect of a Shift in Mortality Pattern on Turnover Duration and Pension Liability Before shift After shift Relative change Average age of retiree,AR 3a 3.5f + 1/6 Average age of contributor,AC 1.5b 1.5b -- AR ­AC, turnover duration, TD 1.5 2 + 1/3 Contribution asset, TD × contributions 36 48 + 1/3 Pension liability, PL 36d 48g + 1/3 IRR (monetary gain / PL) 12/36 + 1/3 Source: Authors. Note: See table 7B.2 for explanation of the use of normal type, bold-face, and italics. a., b., d: see table 7B.2. f. (12 × 4 + 12 × 3) / (12 + 12). g. [12] + [12 + 12] + [12 + 12 ­ 12]. the previous life expectancy. This can also be illustrated by placing the numbers in the example into equation 7.16 [rate of available indexation] = [i] + [ii] + [iii] ­ [rate of implicit indexation]. 0 = 0 + 1/3 + 0 ­ 1/3 The positive return resulting from an increase in life expectancy is due neither to the design of the pension system nor to the imperfect knowledge of life expectancy assumed 134 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES in the example. If cohort B's life span had been known ex ante and the benefit had been reduced to 12 already in cohort B's first period of retirement, there would have been a sur- plus of 12 in period 2. Equation 7.16 would then read as follows [rate of available indexation] = [i] + [ii] + [iii] ­ [rate of implicit indexation]. 1/3 = 0 + 1/3 + 0 ­ 0 If the available indexation is not used to increase the pension liability, the identity requirement for financial stability--a zero net present value--is not met since an undis- tributed surplus then arises. Example 2.1 illustrates the effects of one rule for distributing this surplus. Example 2.1 Same Shift in Mortality Pattern, Different Pension System Rules The rules of the pension system. We now again assume an NDC system. This NDC system indexes its liability by the available rate of return. In the example, this return will equal the IRR since we assume perfect information on life expectancy. In such a system and with this information, the surplus of 12, representing a rate of available indexation of 1/3, will be distributed through indexation of the pension liability in period 2. The effects of the shift in mortality pattern. Before the shift, pension payments and contri- butions will both be stable at 24. Periods 3, 4, and 5, pension payments will be 16, 30, and 26, respectively. Pension payments will be back at 24 as from period 6. Periods 3 and 4, the buffer fund, will thus stand at 8 and 2, respectively, and be back at zero as from period 5. (Readers are encouraged to verify these calculations.) The positive fund is necessary to balance the pension liability, which will temporarily exceed the contribution asset by the same magnitude as the value of the fund. Assuming revaluation of the pension liability at the rate of available indexation and, more realistically, imperfect information on life expectancy, the flow of payments will be different. Still, the system will maintain a zero net present value at all times and in a steady state end up with a zero buffer fund. Summary of What the Examples Show For financially stable PAYG pension systems, the examples showed that the cross-section IRR, regardless of system design and ability to forecast life span, is affected identically by changes in income and mortality patterns. We also learned that the distribution of the IRR among cohorts depends on system design and ability to forecast mortality. Furthermore, the principle of double-entry bookkeeping in PAYG pension systems has been illustrated.8 Notes 1. However, if the rate of return on buffer fund assets or the interest rate on a buffer fund deficit differs from the growth rate in the contribution base, the design of the PAYG system as one with or without buffer fund will have a (small) impact on the system's cross- section IRR. 2. See the first section of this chapter for a definition of cross-section and longitudinal IRR. 3. Income pattern is defined as the ratio of the average wage for each age group to the average wage for all age groups. A stable income pattern exists when this ratio is constant over time for all age groups. 4. Pension liability, or PL, is defined in equation 7.3 as the present value of future pension benefits to all persons to whom the system has a liability at the time of evaluation, minus THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 135 Table 7B.6 Example 2.1: Income Statement and Balance Sheet Period 1 2a 2 3 4 5 Income statement Contributions 24 24 24 24 24 24 Pensions ­24 ­24 ­24 ­16g ­30h ­26i =net cash flow (a) 0 0 0 8 ­6 ­2 Change in contribution asset (b) 0 12c 12c 0 0 0 New accrued pension liabilityb ­24 ­24 ­24 ­24 ­24 ­24 Paid-off pension liabilityb(= paid pension benefits) 24 24 24 16 30 26 Cost of / income from indexation of liabilityb 0 0 ­12 0 0 0 = change in pension liability (c) 0 12 ­12 ­8 6 2 Net income /­loss (a) + (b) + (c) 0 12 0 0 0 0 Balance sheet Buffer fund 0 0 0 8 2 0 Contribution asset 36 48 48 48 48 48 = total assets (d) 36 48 48 56 50 48 Pension liability, age 3 0 0 0 16 14 12 Pension liability, age 2 24 24 32d 28f 24 24 Pension liability, age 1 12 12 16e 12 12 12 = total liability (e) 36 36 48 56 50 48 Net present value of system (d) ­ (e) 0 12 0 0 0 0 Source: Authors. Note: NDC system and indexing at the available rate, in the example equal to the IRR. a., b., see table 7B.3. c. 0.5 × 24 = 12 {change in TD × [contributions (t) + contributions (t ­ 1)]/2}. d. (12 + 12) × 4/3 = 32 [(cohort B's period 1 contribution + cohort B's period 2 contribution) × IRR]. e. 12 × 4/3 = 16 [cohort C's period 2 contribution × IRR]. f. (12 × 4/3) + 12 = 28 [(cohort C's period 2 contribution × IRR) + cohort C's period 3 contribution]. g. 32 / 2 = 16 (cohort B's notional pension capital period 2 / life expectancy). h. 16 + (28 / 2) = 30 (cohort B's pension period 4 + cohort C's pension period 4). i. 14 + (24 / 2) = 26 (cohort C's pension period 5 + cohort D's pension period 5). the present value of future contributions by the same individuals. As there is neither pop- ulation growth nor technological progress, the contribution base will be constant; thus, the discount rate will be zero. 5. Pension liability to cohort B is 30, and to cohort C it is 6. Only after cohort B has passed through the system will the total pension liability drop to the new sustainable level of 30--disregarding the deficit of 6 caused by the shift in income pattern. 136 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES 6. The total wage of cohort C period 2 is 24; with the contribution rate of 25 percent this will make cohort C's contribution and notional capital equal 6 this period. 7. The return on the buffer fund assets is assumed to equal the growth in contribution base which is zero. 8. The accounting standard used is a simplified version of the format developed and used since 2001 for the Swedish public pension system. THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 137 Annex 7C. Intermediate Steps before Equation 7.7 me-(-)a N(0)Wk m ml(u)e- u u 0 dudx 0 x a V e- l(a)W(a)da = R(u)e - W(u) m 0 l(a) R(a)da ... C me-(-)a l(a) R(a)da N(0)Wk m l(x)e- x W(x) 0 dx 0 me- l(a)W(a)da a 0 The expression above can easily be reduced by elementary algebraic manipulations. However, to make it simpler, the following substitutions will be useful: FFW((aa))==ee- -(-)a R l(a) R(a) a l(a)W(a) m m m l(u)e- u u 0 FW (a)da FR(a)da du dx 0 x m ... = R(u)e - W(u) 0 m m FR(a)da FW (x) 0 dx 0 m FW (a)da 0 m ml(u)e- u u m FW (a)da-W(u) m = 0 x R(u)e 0 0FR(a)dda du dx mFR(a)da m FW (x)dx 0 0 m m FW (a)da ml(u)e- u R(u)e du -u m FR(a)da ml(u)e-u = 0 0 x 0 x W(u)du dx m FR(a)dda m FW (x)dx 0 0 m m FW (a)da mFR(u)du- m FR(a)da m = 0 0 x 0 xFW (u)dudx m m FR(a)da FW (x)dx 0 0 mFW (a)da m m mFR(a)da m m = 0 0 xFR(u)dudx - 0 0 x FW (u)dudx m m FR(a)da FW (x)dx 0 0 m m m m = 0 x FR(u)dudx - 0 x FW (u)dudx... m m FR(x)dx FW (x)dx 0 0 138 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES For the next step, the following identity is needed: = m m m f (u)dudx = mx f (x)dx 0 x 0 proof: m m f (u)dudx = x m m 0 x x f (u)du0 - 0m x - f (x) dx ( ) mm f (u)du- 0 0m f (u)du + 0m x f (x)dx = 0 + 0 +0mx f (x)dx . Thus m m m -)x m ... = V = xFR(x)dx xFW (x)dx xe-( l(x) R(x) dx xe-x l(x)W(x)dx 0 - 0 = 0 - 0 C m m m m FR(x)dx FW (x)dx e-(-)x l(x) R(x) dx e-x l(x)W(x)dx 0 0 0 0 QED. THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 139 Annex 7D. Rough Estimate of Turnover Duration in 41 Countries Individuals who are not in the labor force and are 55 years or older are assumed to receive benefits from the pension system that on average amount to 50 percent of the average wage. Pensions are assumed to be indexed by the growth in the average wage, thus =0. Country Year Estimated population growth, ­ 1 (percent) Turnover duration (years) Tajikistan 1991 3.8 35.3 Argentina 1990 1.5 34.1 Spain 1990 0.2 34.1 New Zealand 1990 0.8 34.0 Australia 1994 0.3 34.0 Kyrgyz Republic 1995 2.8 33.7 Israel 1994 1.8 33.6 Portugal 1992 0.1 33.3 Canada 1992 ­0.1 33.3 Chile 1997 1.5 33.2 Romania 1992 0.5 33.2 Italy 1994 ­0.3 33.2 United States 1995 0.2 33.1 Austria 1996 ­0.4 33.1 Belgium 1994 ­0.1 33.1 Ireland 1990 0.9 33.0 France 1995 0.2 33.0 United Kingdom 1996 ­0.1 32.9 Hungary 1996 ­0.1 32.8 Greece 1995 ­0.1 32.8 Kazakhstan 1996 1.4 32.7 Slovak Republic 1995 0.8 32.7 Denmark 1994 ­0.4 32.7 Sweden 1996 ­0.3 32.7 Netherlands 1995 ­0.3 32.7 Latvia 1996 ­0.1 32.6 Norway 1996 0.1 32.6 Armenia 1993 1.6 32.5 Czech Republic 1996 ­0.1 32.1 Slovenia 1993 ­0.2 32.0 Estonia 1996 0.0 31.9 Belarus 1996 0.1 31.9 Poland 1996 0.4 31.8 Russian Federation 1995 ­0.1 31.7 Germany 1994 ­0.8 31.7 Bulgaria 1993 ­0.2 31.6 Japan 1990 ­0.4 31.6 Finland 1996 ­0.4 31.6 Korea, Rep. of 1991 0.9 31.5 Ukraine 1993 ­0.1 31.3 Moldova 1994 0.8 31.2 Source: Authors' calculations, based on UN and ILO statistics. For details, see Settergren and Mikula (2001). 140 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Notes 1. While many examples could be cited to illustrate this point, here are just two of them: "As Paul Samuelson showed 40 years ago, the real rate of return in a mature PAYG system is equal to the sum of the rate of growth in the labor force and the rate of growth in pro- ductivity" (Orszag and Stiglitz 1999, p. 15). "The rate of return in a notional system can only be the rate of growth of the tax base that results from rising real wages and increasing numbers of employees (Samuelson 1958)" (Feldstein 2002, p. 7). 2. In the context of pensions, a steady state is defined as a situation where the average wage at each age, relative to the average wage for all ages, is constant over time and where the number of retirees at each age, relative to the total number of retirees, is constant over time: that is, where mortality rates are constant. Thus the definition of steady state is con- sistent with population growth (or decline) if the change rate remains constant over time. 3. See Riksförsäkringsverket (2004). To be more precise, the average ages refer to the expected average ages. The expected ages will correspond with actual average ages only if fertility-driven population growth, income, and mortality patterns are stable: that is, if they are in a steady state. 4. See the legislative history of the indexation of the Swedish pension system and Setter- gren (2001, 2002). 5. This ignorance is clear from the legislative history of the Swedish pension reform (annex A), as well as Settergren (2001). It is evident that we were not alone in being unaware of the works, or of their implications, that "explore the interface of richer demo- graphic models and the overlapping generation models of economists" (Lee, 1994a). An example is Valdés-Prieto (2000), who observed that changes in income and mortality pat- terns influence the financial balance of a "notional defined contribution" PAYG pension scheme. However, Valdés-Prieto does not explain his observations by the effects that changes in income and mortality patterns have on the money-weighted age difference between the average ages when income is earned and consumed. 6. Pension liability is defined as the present value of future benefits to all persons to whom the system has a liability at the time of evaluation, minus the present value of future contributions by the same individuals (equation 7.3), sometimes also referred to as the implicit pension liability (see also Iyer 1999). The practical problems of measuring pension liability in PAYG schemes are often substantial. Depending on the system design, and the quality and availability of data, the estimate of pension liability may indeed be so uncer- tain that it is practically useless. This chapter does not deal with these important practical obstacles for applying the method suggested for estimating IRR and applying double- entry bookkeeping. 7. An example of the conventional presentation of financial status in a PAYG pension system is the analysis in the Annual Report of the Board of Trustees in the Federal Old-age and Survivors Insurance and Disability Insurance Trust Funds (2003). 8. The expression could be extended also to incorporate the effects of migration on the expected contribution-weighted average age of contributors. See Settergren and Mikula (2001) for such an extended interpretation of . 9. Below we will not use the full expression expected turnover duration to indicate that the turnover duration is measured outside of a steady state, but will refer only to turnover duration. 10. An obvious difference between repetitive re-evaluation of the contribution assets and recurring re-evaluation of funded assets is that funded assets are tradable, which make their prices much less "implicit." However, their valuation is inevitably hypothetical to some degree as long as they are not sold off. THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 141 11. The data for the estimate are the annual individual earnings and benefit records for all persons covered by the national pension scheme in Sweden from 1981 through 2003. 12. To get a zero buffer fund in steady state, either the steady state rate of return on the buffer fund or the interest rate paid on a deficit must equal the growth in the contribution base or the valuation of the fund must reflect an assumption of a return on capital different from the growth in the contribution base. 13. Note that since the turnover duration is affected, however mildly, by changes in the fertility-driven growth rate, , the IRR may differ from the contribution base growth even in the unrealistic case of constant mortality and income patterns. This points out the short- comings in the two-age OLG model: it cannot represent the relevant geometry of the prob- lem. 14. In practical applications equation 7.17 should be extended to accommodate the pos- sibility of an opening surplus or deficit--that is, a difference between assets (buffer fund assets and contribution asset) and liabilities. 15. Whether or not the claim that double-entry accounting provides better information than traditional measures of actuarial balance is correct can, perhaps, be judged from the Annual Reports of the Swedish Pension System, which have been published every year since the 2001 report, published in 2002. The design of the index in the new Swedish pub- lic PAYG pension system indicates that separating the components of the IRR adds new options for designing the indexation of pensions. See the legislative history of the income index and automatic balance mechanism in annex A and Settergren (2001, 2002). References Aaron, H. J. 1966. "The Social Insurance Paradox." Canadian Journal of Economics 32 (August): 371­74. Arthur, W. B., and G. McNicoll. 1978. "Samuelson, Population and Intergenerational Transfers." International Economic Review 10 (1): 241­46. Buchanan, J. 1968. "Social Insurance in a Growing Economy: A Proposal for Radical Reform." National Tax Journal 21: 386­95. Disney, R. 1999. "Notional Accounts as a Pension Reform Strategy: An Evaluation." Social Protection Discussion Paper 9928, World Bank, Washington, DC. Federal Old-Age and Survivors Insurance and Disability Insurance Trust Funds, Board of Trustees. 2003. 2003 Annual Report of the Board of Trustees of the Federal Old-Age and Sur- vivors Insurance and Disability Insurance Trust Funds. Washington, DC: U.S. Govern- ment Printing Office. Feldstein, M. 2002. "Introduction, an American Perspective." In Social Security Pension Reform in Europe, ed. M. Feldstein and H. Siebert, 1­8. National Bureau of Economic Research. Chicago: The University of Chicago Press. Iyer, S. 1999. Actuarial Mathematics of Social Security Pensions. Geneva: International Labor Organization. Keyfitz, N. 1985. Applied Mathematical Demography. Berlin and Heidelberg: Springer Verlag. ------. 1988. "The Demographics of Unfunded Pensions." European Journal of Population 1: 5­30. Lee, R. D. 1980. "Age Structure, Intergenerational Transfers and Economic Growth: An Overview." Revue Economic 31 (6): 1129­56. ------. 1988a. "Induced Population Growth and Induced Technological Progress: Their Interaction in the Acceleration Phase." Mathematical Population Studies 1 (3): 265­88. ------. 1988b. "Intergenerational Flows of Time and Goods: Consequences of Slowing Population Growth." Journal of Political Economy 96 (31): 618­51. 142 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES ------. 1994a. "The Formal Demography of Aging, Transfers, and the Economic Life Cycle." In The Demography of Aging, ed. L. Martin and S. Preston, 8­49. Washington, DC: National Academy Press. ------. 1994b. "Population Age Structure, Intergenerational Transfer, and Wealth: A New Approach, with Application to the U.S." Journal of Human Resources 29 (4): 1027­63. Lee, R. D., and H. Yamagata. 2003. "Sustainable Social Security: What Would It Cost?" National Tax Journal 56 (1, part 1): 27­43. Orsag, P., and J. Stiglitz. 1999. "Rethinking Pension Reform: Ten Myths About Social Secu- rity Systems." Paper presented at World Bank conference "New Ideas About Old Age Security," Washington, September 14­15. Riksförsäkringsverket. 2002. The Swedish Pension System Annual Report 2001. Stock- holm. ------. 2003. The Swedish Pension System Annual Report 2002. Stockholm. ------. 2004. The Swedish Pension System Annual Report 2003. Stockholm. Samuelson, P. 1958. "An Exact Consumption-Loan Model of Interest With or Without the Social Contrivance of Money." The Journal of Political Economy 6 (December): 467­82. ------. 1975a. "The Optimum Growth Rate for Population." International Economic Review 16 (3): 531­38. ------. 1975b. The Optimum Growth Rate for Population: Agreement and Evaluations. International Economic Review 17 (2): 516­525. Settergren, O. 2001. "The Automatic Balance Mechanism of the Swedish Pension System-- A Non-technical Introduction." Wirtschaftspolitishe Blätter 4: 339­49. ------. 2002. "Financial and Inter-Generational Balance? An Introduction to How the Swedish Pension System Manages Conflicting Ambitions." Scandinavian Insurance Quarterly 2: 99­114. Settergren, O., and B. D. Mikula. 2001. "Financial Balance & Inter-Generational Fairness in Pay-As-You-Go Pension Systems--Empirical Illustrations in an Alternative Frame- work." Paper prepared for presentation at the meeting of the Old-age, Invalidity and Survivors' Benefits Technical Commission held during the General Assembly of the International Social Security Association in Stockholm, September. Valdés-Prieto, S. 2000. "The Financial Stability of Notional Accounts Pensions." The Scandi- navian Journal of Economics 102 (3): 395­417. Willis, R. J. 1988. "Economics of Changing Age Distribution in Developed Countries." In Life Cycles, Institutions, and Population Growth: A Theory of the Equilibrium Interest Rate in an Overlapping Generations Model. International Studies in Demography, ed. R. Lee, 106­138, Oxford: Oxford University Press. Discussion of "The Rate of Return of Pay-As-You-Go Pension Systems: A More Exact Consumption-Loan Model of Interest" Ronald Lee* WHAT RATE OF RETURN IS EARNED BY PARTICIPANTS in a pay-as-you-go (PAYG) pension sys- tem? We all know that in steady state, the rate of return equals the population growth rate plus the rate of productivity growth, or equivalently the rate of growth of the tax or contri- bution base, or the rate of growth of the economy. But many of us casually replace the qualifying condition "steady state" with the condition "mature," and conclude that the pension programs of most industrial nations should yield this rate of return since most cover the great majority of workers and are therefore mature. Not so, the authors of this chapter tell us. Mature pension systems depart from the steady state characterization for many reasons, including changes in population age distri- butions, ages at retirement, ages at beginning work, mortality, rates of productivity growth, and the age structure of wages. Mortality has been declining for more than a cen- tury in the industrial nations, and most analysts expect these declines to continue. Even if most other aspects of the pension program were in steady state, this one factor would mean that the effective rate of return earned by each generation in a strict PAYG system would exceed the growth rate of the contribution base. Each generation of workers would pay taxes to cover the retirement costs of generations whose retirements are cut short by mortality at earlier ages than will be true for the generations of workers when they them- selves retire several decades later. For this reason, their rate of return will exceed the growth of the contribution base. Of course, to fund this lengthening retirement and higher rate of return, the contribution rate must be rising over time. The effect of declining mor- tality on the generational rate of return is quite noticeable in the U.S. social security sys- tem, adding perhaps 0.3 to 0.5 percent per year to the implicit rate of return, a little realized fact. The rate of return earned by any particular generation can always be calculated ex post, once its last members have died. But the last members of a generation will not have died until it is well over 100 years old, and by then the question will be of only historical inter- est. For practical purposes, we may want a measure of the system's rate of return based on its circumstances and performance today. The chapter addresses this problem by constructing a measure of the implicit rate of return to a PAYG pension system. The authors first develop a measure for the steady state context, deriving and interpreting an accounting identity. This part of the chapter builds * Ronald Lee is director of the Center on the Economics and Demography of Aging, Uni- versity of California at Berkeley. Research on which this comment was based was supported by a grant from the National Institute of Aging, R37-AG11761. 143 144 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES on an existing literature describing an accounting framework for intergenerational trans- fers in general. The accounting identity states that the implicit debt of the system, V, or equivalently the present value of the system's current net liabilities, is equal to the annual flow of contributions, C, times the difference between the average age at which a pension payment is received, minus the average at which a contribution (payroll tax) is paid. The authors call this age difference the turnover duration, or TD. The TD synthesizes into a sin- gle number a great deal of information about the system's rules, the age distribution of the population, age patterns of labor supply and earnings, and survival. Note that the TD will be greater when pension benefits are indexed to rise with wage growth following retire- ment than when they are indexed only to inflation, as in the United States. TD would be shorter for a NDC program in which the beneficiary received a lump-sum payment on retirement, and then converted this to an annuity through the private sector. In these and many other ways, the TD will reflect the details of the particular pension plan. Annex D of the chapter presents rough estimates of the TD for 41 countries, based mainly on their labor force participation rates. These 41 countries, with very different demographic situations, retirement ages, and age profiles of earnings, have remarkably similar TDs, all falling between 31 and 35 years. I find this very surprising, but perhaps the similarity reflects the strong simplifying assumptions used for the calculation. The num- ber given for the United States, 33 years, is right on target, which I know from my own detailed calculations. We can try out this simple equation for the United States, using the age difference of 33 years. Social security benefit payments (for pensions, survivors, and disability insurance, OASDI) in 2004 amounted to 4.33 percent of GDP, and in a pure PAYG system, this would also be the share of payroll tax payments.1 The cross-sectional estimate of net pension obligations, or implicit debt for current participants, is therefore 33*4.33 percent of GDP, or 1.43 times GDP. This comes out to US$16.5 trillion, from which the trust fund of US$1.5 trillion should be subtracted to get US$15 trillion.This is considerably greater than the US$11.2 trillion reported by the actuaries in the Trustees Report (2005: Table IV.B8) as the unfunded obligation for past and current participants. However, the actuaries use a much higher discount rate than the population growth rate which is implicitly used in this cross- sectional calculation and which is probably responsible for most of the discrepancy (see Lee 1994 for a similar estimate for the United States). The authors suggest that the turnover duration can be interpreted as a measure of the time preference of the pension program, and that its inverse is a measure of the discount rate for contributions. Based on the range of values described above, the implied discount rate would be between 2.8 and 3.2 percent per year. I found this idea puzzling, and would be interested to see the authors develop it further and explain it more clearly. The results in the first part are interesting and conceptually useful, but they are of limited practical use since actual systems are not in steady state. The results reveal the parameters upon which financial stability is dependent, within the framework of a continuous financial balance sheet. What is unique is that this method of presentation of the framework follows a traditional "business" financial balance, while it introduces the economist's or economic demographer's way of describing this balance. To use this framework for actual, nonsteady state situations, we have to imagine stopping time at two intervals and using a comparative static comparison between them. This is the approach developed by the authors in the sec- ond part of the chapter. If we consider a simple special case of their more general equation, in which the trust fund is zero, then the steady state identity can be written PL = TD*C, where PL is pension liability (which I have been calling the implicit debt owed to current participants), TD is the turnover duration just discussed, and C is the annual flow of contri- THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 145 butions to the system. They argue that the rate of return on contributions to the system (in a cross-sectional sense) is the rate of return at which the change in PL is zero over time. I would have liked to see more discussion of this point. In a growing economy, fixing the size of the PL would mean that pension wealth would decline relative to GNP, which does not seem right. In any event, setting equal to zero the time derivative of the equation just given, their result implies that the IRR of the system equals the exponential rate of growth of contri- butions, C, plus the exponential rate of change of the turnover duration, TD (for the special case in which there is no trust fund balance). In the steady state case, the rate of growth of contributions will just equal the rate of population growth plus the rate of growth of productivity, while the rate of change of TD would be zero. This gives the well-known steady state rate of return for a PAYG system. However, out of steady state, the growth rate of C will also depend on demographic fluc- tuations rippling through the labor force, changes in the age structure of earnings, changes in retirement age, and changes in the age at starting work, in addition to productivity growth and general population growth. Similarly, out of steady state, TD will vary due to changes in many of the factors just mentioned, plus changes in survival in old age. For example, if life expectancy is rising so that retirees continue to receive pensions at higher ages, then the average age of receiving pensions will increase and so will TD, and therefore the IRR received through the pension system will rise as well. If life expectancy ceases to rise, then TD will stop rising, and the rate of return will drop back down again. TD will also be changing in other ways due to rising life expectancy, depending on the details of the program. If benefits are indexed to mortality, then younger generations of retirees will receive lower benefit flows, altering TD, for example. To take another example, if the peak of the cross-sectional age-earnings profile shifts to older ages reflecting an increasingly educated labor force, that would raise the age of paying contributions, leading to a reduction in TD and a lower rate of return. All these year-to-year changes are reflected in the suggested calculation, in addition to the standard rate of growth of contributions. The cross-sectional rate of return measure is affected only so long as these changes are occurring. Once the changes stop, the rate of return will also tend to go back to its steady state level, which is unaffected by the level of the TD. This general approach, and its strengths and hazards, are familiar to demographers who confront similar issues when measuring fertility or mortality. There is a straightfor- ward measure that can be calculated at the end of the reproductive years, completed cohort fertility, and another that can be calculated after all in a generation have died, the average age at death or cohort life expectancy. But for obvious reasons, we would like a measure of fertility and mortality that describes the current situation. Demographers con- struct synthetic cohort measures, or period measures, that summarize the current situation by imagining that a generation lives its entire life experiencing the age-specific rates of a given year. In the case of fertility, this yields the most commonly used measure, the period total fertility rate (TFR). For mortality it yields another most commonly used measure, period life expectancy. However, there has been intense controversy about these measures, particularly the TFR, because it provides a distorted indication of the completed fertility of any cohort when the timing of fertility in a woman's life cycle is changing--the "tempo" effect--as opposed to the "quantum" of completed cohort fertility. In Europe, the mean age at first birth has been rising by about 0.2 years of age per calendar year for several decades, which depresses the TFR by about 20 percent or nearly 0.4 births relative to the likely eventual completed cohort fertility. Many demographers have suggested adjust- ments to the period TFR to make it a better indicator of cohort completed fertility. 146 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES In the case of the proposed cross-sectional measure of the rate of return to pension par- ticipation, similar issues are bound to arise, but have not yet been deeply explored. How- ever, this chapter makes a very promising and enlightening start on addressing this problem in its second part. We can think of the change over time in the turnover duration as introducing a kind of tempo distortion. When the turnover duration is increasing, for example, we would expect that the current rate of benefit payments would be temporarily reduced, and the reverse when the duration is contracting. The rate of change in the con- tribution rate will likewise be affected by transitory changes in the age of starting or end- ing work, which come on top of the effects of productivity growth and in a sense distort it. These effects are correct for the cross section, but may be misleading when we try to draw inferences from the cross-sectional rate of return about the longitudinal or long-run rate of return. In an NDC system, changes across generations in age at retirement and in life expectancy would presumably have no effect on the generational rate of return, since NDC is actuarially fair. But such changes would affect the timing of the payment of contri- butions and the timing and level of benefits, and these timing changes would be reflected in a changing cross-sectional TD, and thus in the cross-sectional rate of return. The trick is to choose an accounting method, whether cross-sectional or longitudinal, that treats the different generations fairly in a nonsteady state system, avoiding the kinds of inequities that are illustrated clearly in the very helpful annex B. The chapter is relatively brief, even including its useful appendices, and many details are not spelled out. Discussion toward the end suggests that the cross-sectional pension liability that they have in mind, PL(t), is actually based on an actuarial projection into the future. I found this confusing because there is also a cross-sectional measure of PL, based on the contributions from and benefit payments to the synthetic cohort, and I don't under- stand why this is not used instead. Similarly, the measure of transfer duration is referred to as "expected" TD toward the end of the chapter, which suggests again a longitudinal and projected aspect, whereas there is a purely cross-sectional measure of TD that does not involve expectations. Because of these details, I do not fully understand the calculations or their underlying logic. Doubtless most of these issues could be resolved through discus- sion with the authors. In other places, I had difficulty following because of unfamiliarity with some of the terms and phrases used. This was particularly true of the important discussion toward the end of the "rate of available indexation," "rate of implicit revaluation," "rate of explicit indexation," and "system net income." I am not clear about the relation of these terms and phrases to the IRR, the main topic of this paper. I believe, however, that these terms refer to a kind of residual correction to the cross-sectional estimate of the rate of return. The method described in this study appears to be a powerful tool for assessing the per- formance of PAYG pension systems. However, questions remain, and we will need to see a more complete exposition of this approach, and perhaps a more complete analysis, before fully understanding what is being proposed for the rate of return out of steady state. Proper evaluation of the proposed procedure can come only after a more complete under- standing of it by the research community. This is quite possibly just a matter of fuller expo- sition. In my view, the authors are too modest about their contribution on this topic. While the literature they cite, including my own contributions, is certainly highly relevant, most of it takes a comparative steady state approach and does not deal with nonsteady situations, and none of it addresses the specific problem of how to measure the rate of return to a non- steady state PAYG pension system. The work in this chapter is an important application THE RATE OF RETURN OF PAY-AS-YOU-GO PENSION SYSTEMS 147 that will, I hope, lead to a series of articles by the authors and by others that further develop these ideas and probe their limitations and uses. Note 1. See Board of Trustees (2005). References Board of Trustees. 2005. The 2005 Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Disability Insurance Trust Funds. Washington, DC: United States Government Printing Office. Lee, R. 1994. "The Formal Demography of Population Aging, Transfers, and the Economic Life Cycle." In The Demography of Aging, ed. L. Martin and S. Preston, 8­49. Washing- ton, DC: National Academy Press. Chapter 8 A Market Method to Endow NDC Systems with Automatic Financial Stability Salvador Valdés-Prieto* PENSION PLANS THAT ARE FINANCED WITH THE PAY-AS-YOU-GO (PAYG) METHOD, but that have an actuarial benefit formula that includes individual accounts, have been spreading in Europe since the end of World War II. These plans have been recently named notional defined contribution (NDC).1 A more apt name is notional accounts (NA), since such plans can also offer defined benefits, as shown below. An important property of a pension plan is its degree of automatic financial stability in the short run. It is normally considered valuable to insulate the government's finances from the shocks that impinge on the pension plan, and to insulate the pension plan from the shocks that affect the treasury because it destroys the risk created by random parlia- mentary delays in parameter adjustment to shocks--that is, it minimizes political risk. Insulation in this sense is not an absolute concept. Fiscal policies can always be used ex post by parliament either to subsidize or to tax pension income. Insulation adds value because it gives parliament the opportunity to debate possible fiscal policies in response to shocks, free from the pressure coming from the insolvency of the pension plan. A major precondition for mutual insulation is that the parameters of the plan are such that, if no more shocks arrive in the future, the plan would be able to meet its payment promises for the indefinite future. Call this characteristic initial financial independence, or initial solvency. Many pay-as-you-go-financed plans are currently insolvent in this sense, and for them, the issue of financial stability is simply a far-off aspiration. For initially solvent plans, the issue of financial stability is renewed over time, as unin- surable shocks arrive and add their aggregate financial impact. These shocks can be demo- graphic (longevity and fertility), economic (changes in average taxable income of covered workers, variations in employment), and financial (changes in stock and bond prices, which affect the fund portion in partially funded plans). The aggregate financial impact is the difference between the impact on plan liabilities and the impact on plan assets. The benefit formula of the plan, whatever it is, controls the liability side and normally specifies * Salvador Valdés-Prieto is professor of economics at Catholic University of Chile and researcher at the Centro de Estudios Públicos, Santiago. This chapter is related to a quite different earlier paper circulated at the Swedish Social Security Fund­World Bank Conference on Notional Defined Contribution Pension Systems, September 29­30, 2003, Sandhamn, Sweden. Comments are welcome at svaldes@faceapuc.cl. 149 150 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES adjustments to the benefits of the individual member. For example, if taxable earnings and contributions drop, promised pension amounts drop in all earnings-related plans. Adding these adjustments across all individual members yields the aggregate financial conse- quence of the shock on the liability of the pension plan. Independently, the shock has an impact on the assets of the plan. The difference between both impacts is the aggregate financial impact of the shock. This impact may be relatively large in the short run.2 An example is an increase in unemployment in an initially solvent defined benefit (DB) plan financed with the pay-as-you-go (PAYG) method. This chapter discusses alternative ways in which a plan adjusts to aggregate financial shocks. In general, adjustment depends on four sets of rules, which operate in successive layers: (1) the benefit formula of the plan, which links individual contributions with indi- vidual benefit amounts; (2) the rules for adjusting parameters of the plan; (3) the rules for the use of eventual buffer funds and credit lines offered by a sponsor; and (4) the rules governing the exercise of discretion ex post by a set of people, say members of parliament, members of a public board governing the plan, civil servants who control deeper technical parameters, and finance ministers who retain the right to limit the plan's access to credit lines or to buffer funds. The financial stability of a pension plan can be described as "auto- matic" when layers (3) and (4) are avoided in almost all states of nature (except war and the like) because layers (1) and (2) are able to cope with most shocks by themselves. The literature has proved that NA plans that use any of a set of simple benefit formulae that were adopted in Europe in the 1990s do not enjoy automatic financial stability in the short run in the face of demographic and economic shocks.3 In an NA plan, the benefit for- mula comprises the individual account of active members, the formula used to set the notional interest rate credited to those accounts, the factor to convert the account balance at retirement into the initial pension, and the formula for indexing pensions in payment. Consider the case where the instantaneous notional interest rate credited to individual accounts is the instantaneous growth rate of contribution revenue, where pension pay- ments are indexed to the growth rate of contribution revenue, and where the conversion factor at pension age is an adequate function of this same growth rate and expected longevity. Even in this case, which is more favorable to automatic financial stability than others, a permanent drop in the population growth rate triggers a long wave of cash deficits that lasts up to 80 years.4 These deficits are due to discrepancies in the timing of the expenditure and revenue changes triggered by this shock. The discrepancy is significant: for a permanent drop in the population growth rate of just 0.5 percent per year, in a plan where the contribution revenue is 10 percent of GDP, the present discounted value of these cash deficits discounted at 3 percent real and located in the year of maximum deficit is about 9 percent of GDP. In contrast, rule-based and contract-based plans such as defined contribution (DC) plans (mutual funds) and defined benefit annuities offered by life insurance companies always achieve automatic financial stability in the short run merely by using layers (1) and (2). Consider the mutual fund formula, invented in Boston in 1924, that underlies DC plans. Each individual member has a number of shares of the mutual fund, and its price is adjusted daily in the following way: the new price is the new value of assets (at forward- looking market prices) divided by the number of shares outstanding. It may be argued that rule-based and contract-based plans are irrelevant, because they require a costly tran- sition to full funding (see definition in box 8.1). This is incorrect, as shown by Valdés-Pri- eto (2005), because the transition can be costless and immediate. In 2001, the Swedish authorities devised an alternative path to reach automatic financial stability in their NA plan. The original rules were to credit notional capital of active work- ers at the growth rate of contribution revenue, and to adjust pensions in payment by the A MARKET METHOD TO ENDOW NDC SYSTEMS WITH AUTOMATIC FINANCIAL STABILITY 151 Box 8.1. What Is Meant by Funding The issue of whether a pension plan is "funded" or not is controversial because there are three meanings of the term. In the fiscal meaning, the pension fund is the financial securities, real estate, and other rights owned by the plan. In contrast, the present expected value of govern- ment subsidies to the plan is not a part of the pension fund, because the law can be modified and these donations can stop. These donations can be explicit subsidies, as when notional plans receive a regular budget transfer, or can be implicit subsidies such as the net lifetime tax implicit in mature PAYG finance as described in the main text. Therefore this critical revenue source is not part of the pension fund either. Generalizing, a pension fund is defined as the set of payment promises in favor of the plan that are protected by property rights.5 The degree of funding in the financial sense is a ratio obtained by dividing the market value of the pension fund by the expected discounted value of accrued liabilities at the same date. In its fiscal meaning, funding is the degree to which changes in benefit liabilities bring about changes in the net fiscal debt, or affect the riskiness of the payment profile of the net public debt. For this purpose, the "net public debt" is the standard public debt, minus publicly owned assets, plus the implicit fiscal debt to insolvent health and old-age benefit programs. The fiscally costly "transition costs" are caused by restructuring of fiscal funding, not by changes in the degree of financial funding. There is another meaning of funding, which refers to the impact of the plan on the volume of economywide savings. This "macro," "ultimate," or "broad" funding is not a feature of the plan itself but of the wider macroeconomic context. The degree of broad funding is a function of private savings, national debt policy, the degree of Ricardian equivalence, and the presence of other state-sponsored transfer programs such as health subsidies for the old, student loans, and the pension plan. Broad funding is a characteristic of the economy, not the plan. As pension plans should not be labeled on the basis of something not related to them, the meaning of funding should be restricted to the financial definition (the first one). The IFM reform discussed in this chapter does not affect the economy's degree of funding. However, it increases the plan's degree of financial funding because it grants the plan property right pro- tection over the revenue of the tax implicit in the contributions extracted from future contrib- utors who are forced to remain in or join the plan. growth rate in average wages minus 1.6 percent per year. However, since May 2001, a dif- ferent and highly creative rule--called the automatic balance mechanism (ABM)--was adopted. When the plan's liabilities exceed assets, the notional interest rate credited to active members and the rate of adjustment of pensions in payment are both cut, to equal the growth rate of average wages minus the growth rate in liabilities plus the growth rate in assets.6 The ABM seems to endow the Swedish plan with automatic financial stability. This chapter discusses an alternative response, which also endows NA plans with auto- matic financial stability in the short run, by adopting a rule-based method to allocate aggregate risk. I call it "integration to financial markets" (IFM) for reasons to be explained below. In contrast to the ABM, the IFM method is based on discount rates determined by the financial and capital markets, and thus relies on objective projections. This also allows the IFM to be decentralized, reducing the risk of political and bureaucratic interference, an aspect that might increase the confidence of members in the plan. It is also important to stress that as the IFM relies on asset values, it adjusts using rational expectations rather than slow-moving averages. In addition, the IFM method allows the plan to trade part of the uninsurable risks that it faces with other investors and internationally, creating scope for substantial economic gains. 152 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES In most NA plans, the value of the pension fund is negligible as a proportion of liabili- ties. Even in the case of the Swedish plan, which is endowed with a large buffer fund, the pension fund was about 10 percent of liabilities in 2001, under current accounting rules.7 Therefore, most of the assets of a NA pension plan are the present expected value of the net lifetime tax on labor earnings that future members will have to bear. This tax revenue is not owned by the pension fund in the legal sense because the law can be modified to stop this cash flow. For example, a future law may allow Swedish workers to divert part of those contributions to the PPM (a financial DC plan). Access of the NA plan to most of the economic assets that back its promises is not protected by property rights. Consider endowing NA plans with property rights over the net lifetime tax revenue that future covered workers will yield, followed by securitization of these cash flows and by trade of at least a few of these new securities in the financial markets. The new securi- ties are called covered wage bill securities, or CWB securities (see Valdés-Prieto 2005). This chapter compares the resulting pension plan with the ABM system. It stresses that the market prices that emerge for these new securities will be forward-looking, and that these prices contain the discount rates needed to ascertain the assets and liabilities of the plan. Therefore, feeding these discount rates back into the benefit formula of the NA plan-- specifically, making the notional interest rate equal to the rate of return on CWB securities observed in the financial market and adopting the mutual fund rule to determine the bal- ance of the individual accounts--endows the plan with automatic financial stability in the short run. This stability is different from the one produced by the ABM, because it is based on forward-looking prices. The chapter proceeds as follows. The next section summarizes the proposal to create CWB securities, to endow the NA plan with them, to trade some of them, and to insert the resulting market prices into the NA benefit formula. The third section compares this method to achieve automatic financial stability with the automatic balance mechanism adopted by Sweden in May 2001 in its pure form. The final section offers concluding com- ments. Endowing a Solvent NA Plan with Automatic Financial Stability This section summarizes the proposal to modify an NA plan that is initially solvent to endow it with automatic financial stability. These steps are developed and justified in detail by Valdés-Prieto (2005). The Economic Asset That Backs PAYG-Financed Pension Promises Every pension plan has a liability given by the expected present value of its commitments to pay pensions to its current members. The plan does not yet owe anything to future members, since they have not contributed. The size of the existing liability is set by the cur- rent benefit formula--which may be based on individual accounts or years of service--by the history of taxable salaries or contributions of current members, by pension ages, and by the formula that indexes pensions in payment.8 The size of this liability does not depend on the financing method used by the plan. A solvent pension plan is defined as one that does not require financial support from its sponsor, nor will pay a profit to it, in present expected value. An initially solvent plan financed with "pure" PAYG does not own financial assets or real estate in any measure comparable to its liabilities. It may seem that this plan has no assets but has a large liabil- ity, so that a large negative net worth would be needed to respect the balance sheet iden- tity. But the cash flow in this plan is balanced at zero by the assumption of initial solvency. This implies that the net worth of the plan for its sponsor is zero, not negative. The answer A MARKET METHOD TO ENDOW NDC SYSTEMS WITH AUTOMATIC FINANCIAL STABILITY 153 to this riddle is that the pension plan holds a special economic asset, in addition to what- ever pension fund it may own. Lee (1994) has argued that the present value of expected future transfers to be received, minus transfers to be made, is a form of wealth that is held by individuals in modern economies--and by institutions such as pension plans. He also points out that when a transfer system can obligate people as yet unborn to make future transfers to members of the current population, aggregate "transfer wealth" may be positive for the current popu- lation, and negative for the yet unborn generations. Consider the difference between the present value of contributions and the present value of benefits promised to each future generation of members. For example, consider an overlapping-generations (OLG) economy populated by agents whose lives last two periods, young and old, with probability 1, and where the old do not work. Each worker active at date t earns covered labor income in amount yt. The number of workers that are active in date t is Nt. Assume that plan parameters are in a steady state where contribu- tions are paid at rate O percent of labor income, while in old age benefits are received at a level such that percent of the earnings of the previous period are "replaced." The real discount rate after inflation, given by alternative investments of the plan in the financial market or in real estate, is r percent per period. The difference between the contributions and the present value of benefits promised in exchange to that same generation is, in the aggregate and as of young age: Tt 0 ytNt - 1 ytNt + r= 0 - 1 ytNt + r ytNt , (8.1) where Tt is the "lifetime hidden tax" on that generation of members. Tt is not zero unless = O (1 + r): that is, unless r = (/O) ­ 1. The internal rate of return (IRR) of pure PAYG finance in steady state is the growth rate of the contribution base, which is also the growth rate of the economy (GDP): say, g percent real, given by the sum of population growth and labor-augmenting productivity growth. In hypothetical "golden rule" economies, the real interest rate earned by physical capital is r = g.9 Box 8.2 argues that g must be smaller than the real interest rate available in the financial markets before taxes (and earned by pension funds invested in financial instruments or in physical assets) after discounting for risk. It turns out that for two-period lives and pure PAYG finance, the IRR for members of the plan is IRR = (/O) ­ 1. In a steady state where the plans keeps a constant share of the overall economy, IRR = g. Therefore, the condition r > g (see box 8.2) implies that the dis- count rate for plan liabilities must be r > IRR. Given this, the tax Tt defined by equation 8.1 must be positive (see proof in Sinn 1999). One intuition for a positive tax Tt is that it is the counterpart of the gift granted to the initial old generation when PAYG finance was intro- duced, which is positive. This result generalizes to out-of-steady-state situations, to the partially funded case and to the uncertainty case, but refers now to the average tax paid by all generations (see proof in Valdés-Prieto 2005). However, the tax can be negative for a few generations, depending on the plan's rule for adjusting to shocks, as shown by Setter- gren and Mikula (2006). Define the PAYG asset as the expected present value of the hidden taxes that the plan expects to collect from its future members, who have not yet joined the plan, for the infi- nite future. This is just another way to write the "expanded" balance sheet for an ongoing pension plan. The expanded balance sheet adds to the asset side the expected present dis- counted value of the contributions that all future generations will pay, and adds to the accrued liabilities the expected present discounted value of benefits for all future genera- 154 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Box 8.2. Why the Rate of Return on Capital Must Be Larger Than the Growth Rate There is abundant empirical evidence in favor of this hypothesis for developed countries (Abel et al. 1989). Moreover, there is by now a well-developed theory that explains why this inequality must be a feature of any market economy. In the presence of assets that are infinitely durable, such as land, whose ownership is widely dispersed in many societies, it has been shown that r > g in macroeconomic equilib- rium (Scheinkman 1980; Richter 1993). One simple way to explain this reasoning is as follows: If land is in fixed supply and yields a rent of 1 unit of output per year, and the economy grows at rate g, and the share in consumption of land-based goods is constant, then the price of a unit of land output must grow at rate g. The present value of this rent, discounted at the real inter- est rate r, is $1/(r ­ g). Therefore, if r = g as in golden rule economies, the value of land would be infinite. As land ownership attracts part of the portfolio of households, they would cease to hold other physical capital (machinery). However, a fall in the supply of other physical capital would raise its return, until r > g. Confirming this result from another direction, Tirole (1985) showed that the inequality r > g also follows from the need to prevent arbitrage by entities that are infinitely lived and whose income grows at rate g. Examples of such entities are governments that can issue debt, dynas- ties that earn a wage bill that grows at rate g and may issue debt, pension plans, and health plans for the old. The following "slow Ponzi scheme" is possible for each such entity when r = g: issue $1 million of financial debt at interest rate r = g, and when interest and principal comes due, refi- nance both at rate r = g. Although the size of the new debt grows exponentially at rate g, the ratio of the new debt to current income is fixed at a constant number, because current income also grows at rate g. Therefore, the entity remains solvent in the sense of ability to repay all its debts on time forever. Thus the entity can distribute $1 million in dividends to its owners now, without demanding any compensating cuts in dividends in the future. Put another way, when r = g in the long run, such entities are not subject to a budget constraint, and even less so if r < g. As owners still attach a positive marginal value to extra wealth, the supply of debt explodes to infinity when r falls so low as to equal g. This increase in the demand for credit must raise interest rates until r > g. The inequality r > g also implies that the economy is dynamically efficient in the sense that it would not be better off by destroying physical capital. Given these facts, golden rule economies where the real interest rate earned by physical capital is r = g are only hypothetical. tions. If this latter liability is written as a negative asset, the difference between these two expected present discounted values appears as the PAYG asset.10 As the average hidden lifetime tax is positive, the PAYG asset is positive too (given ini- tial solvency). Therefore, this is an asset for the plan, not a liability. This is a real asset for the plan, which collects real revenue, as long as it is reasonable to expect that future con- tributors will continue paying the net lifetime tax embedded in their contributions. The PAYG asset is used by the plan to back its accrued liabilities to current members. Valdés-Prieto (2002, chapter 8) proved that if the plan is initially solvent, the size of the PAYG asset will be just enough to fill up the asset side of the balance sheet of the plan. This proof has been extended to the uncertainty case (Valdés-Prieto 2005). Therefore, the PAYG asset can be calculated with a simple rule: take liabilities and subtract the pension fund, if partially funded. However, this is not true if the plan is not solvent initially. When projections show that the present value of cash flow is negative--that is, when the plan is initially insolvent-- A MARKET METHOD TO ENDOW NDC SYSTEMS WITH AUTOMATIC FINANCIAL STABILITY 155 the PAYG asset is not large enough to cover the plan's liabilities at current values for the parameters of the benefit formula. Insolvency does not imply that the pension plan is illiquid. Its cash flow may be positive for the next 10 or 20 years. Insolvency means that financial independence is unsustainable in the longer run if current parameter values are not modified. Let us review some implications. Young and future members are taxed because the rate of return on saving through this plan is below the rate of return offered by investment in the financial market or in physical capital. The ratio = (Tt/ytNt) is the rate at which the hidden tax is levied on covered labor earnings.11 The hidden tax rate always meets the condition < O for workers who are forward-looking or "provident" regarding old age, because each one sees that he will recover a portion of contributions as pension benefits. Relabel the Cash Flows Consider a reform that comprises several steps. First, the tax hidden in PAYG finance is made explicit. One way to do this is to split the initial mandatory contribution rate 0 into a new (smaller) contribution rate ' and a new residual payroll tax on covered earnings, levied at rate , where is equal to the estimated rate of the hidden tax in the initial situa- tion. In our model of two-period lives for a plan that is initially solvent, this implies 0 = '+ , where 0 - 1 + r> 0. (8.2) Second, a law endows the plan with the revenue collected by the new residual payroll tax on covered earnings. However, this law can be changed by another law in the future, and in that event, the plan will not have a right to claim any compensation for expropria- tion. Thus the plan does not yet enjoy a property right over this tax revenue. These two steps merely relabel existing cash flows. The sum of income flows received by the pension plan does not change, the take-home earnings of workers do not change, and firms' labor costs do not change. The only difference so far is that the IRR earned by plan members on their new contributions (just ') rises to r. Meanwhile, the IRR earned on the residual payroll tax component drops to zero. This expresses in yet another way the budget identities summarized by Lindbeck and Persson (2003). Make the PAYG Cash Flow Tradable In this step, a law grants the plan property rights, in the standard legal sense, over the cash flows indicated previously by Tt. It was already pointed out that a future law could repeal this law, totally or partially. As long as this remains possible, the plan will not have a "real" property right over the PAYG asset. The standard approach to this problem--which affects all private property rights as well--is for the courts to enforce a "property right." Most judicial systems (courts or con- stitutions) specify that in case of expropriation of something covered by property rights, the previous owner is entitled to compensation at market value, paid by the state. This is the current legal status of government debt securities and the legal status of all private prop- erty, since it is always subject to the possibility that a new tax in the future may cut a large part of its value to the owner. To protect plan ownership over the revenue Tt, the analogous approach is proposed: The law that grants the plan the revenue Tt must declare that it is owned by the plan, and must add that if some future government wishes to withdraw part or all of the residual payroll tax in the future, or if that government taxes or otherwise ceases to pay the tax col- 156 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES lection to the plan in the future, then the plan will be considered an expropriated owner, just like any other owner. This clause must be written in a way that is clear for the local courts. Of course, the courts must be independent from the executive power for property rights to exist. The objective of spelling out such details is to minimize the risk that proce- dural objections by a cash-strapped state may devalue the plan's property right over the hidden tax revenue in a significant amount. It turns out that the resulting commitment does not reduce the freedom of future politicians to modify the rate and the base of taxes on covered earnings, and thus engage in welfare-improving tax policy in the future. Parliament keeps the freedom to replace the residual payroll tax at rate by other substitute revenue sources, such as an increase in value-added tax (VAT). This can be done by using the extra revenue from VAT to finance a permanent subsidy to covered workers that defrays the cost of the residual pay- roll tax for workers. The only requirement imposed so far is that any new policy that imposes losses on the plan by diverting its cash flow must compensate the plan. Of course, the plan may remain publicly managed at this stage (but see Valdés-Prieto 2000 for the longer term). Securitize the New Cash Flows Owned by the Plan Securitization is defined as a process that transforms an illiquid asset into a set of securities that are standardized at small enough sizes to make them tradable, which can be sold to investors in the financial markets.12 Consider transforming the PAYG asset into a tradable financial asset, by creating CWB securities. The dividend of CWB securities is Tt. The financial design of CWB securities can be uniform, consisting of a simple pro rata share of the revenue from the residual payroll tax to be received in the future. Alternatively, these shares can take different shapes over time and over states of nature, provided that the sum of all CWB dividends in each and every contingency and time period add up to total tax revenue available at that contin- gency and time period. In a standard case, the volume of CWB securities is a vast proportion of GDP (if this securitization is applied to all the earnings-related NA plan). For example, if Tt is a rev- enue of 3 percent of GDP, the real interest rate is about 3 percent per year, and the expected growth rate of covered earnings is 1 percent per year, then the value of all CWB securities is 3 percent/(0.03 ­ 0.01) = 150 percent of GDP. Note that the market discount rate on CWB securities must be larger than the growth rate of the covered wage bill, because the divi- dend Tt grows at rate g. If the market discount rate were equal to g, then the market value of a CWB security would be infinite. CWB securities differ from public debt in a fundamental way: CWB securities are not owed by the government. They are owed by future contributors to the pension plan, despite the fact that the state does force them to join and pay the residual payroll tax. In addition, CWB securities have a payment profile that is like equity, in the sense that it varies directly with fluctuations in fertility, labor force participation, and real earning trends. In contrast, public debt makes a fixed nominal (or CPI-indexed) payment. The owners of CWB securities take 100 percent of the loss when dividend growth falls below expectations, and also reap the gains when the dividend Tt rises faster than expected. The amount of CWB to be sold in the market by the pension institution can take two very different values. On the one hand, some small amount must be sold to obtain a mar- ket valuation of CWB securities. Achieving sufficient trades for this purpose requires a modest sale of CWB securities compared with the total available. A schedule of weekly A MARKET METHOD TO ENDOW NDC SYSTEMS WITH AUTOMATIC FINANCIAL STABILITY 157 sales of CWB securities adding up to, say, 0.3 percent of GDP per year for the first 10 years may be enough to free the market valuation from liquidity penalties. On the other hand, the pension institution may attempt to abandon its inherited portfo- lio and move toward the risk-return frontier, creating substantial economic gains. This requires exchange of much more substantial amounts of CWB securities, say 30 percent of the total, for corporate securities, both national and foreign.13 Valdés-Prieto (2005) argues that it will not be in the interest of pensioners for the plan to exchange much more than this because many are likely to prefer the safety of CWB securities. Adapt the Benefit Formula of the NA Plan In order to endow the NA plan with automatic financial stability, its benefit formula must adopt a rule that ensures that the current value of liabilities tracks the market value of plan assets on a daily basis. One rule-based benefit formula expresses the notional account balance of each individ- ual member as a number of shares in the assets of the plan. The price of each share is set daily by dividing the value of plan assets by the number of shares outstanding, just as mutual funds do. Then the notional interest rate credited to active members' account bal- ances is chosen: the rate of change of the price of shares. The resulting rule for allocating aggregate uninsurable risk is "defined contribution." Rule-based defined benefits can also be arranged, as follows. The plan buys fixed annu- ities from a life insurance company, which in turn would back those promises by purchas- ing a separate portfolio made up of long-term fixed-income bonds. The pension institution would purchase these annuities with the proceeds of sales of CWB securities in the same amount as the account balance of the retiring members. In addition, the owners of the life insurance company must put up some capital as a guarantee of their promise to take losses if the portfolio falls in value. (If the portfolio rises in value, the owners take the gains.) This insurance function could be performed by the pension institution itself if it had some cap- ital, but this would have to be supplied by the state. Of course, the supply of an adequate volume of fixed-income bonds at reasonable prices can be a problem. This problem is solved if the CWB securities are created in two different classes, the first one paying a dividend indexed to average wages and the other paying as dividend the remaining cash from Tt, (which tracks demographic and participa- tion risks). Life insurance companies would be happy to buy the wage-indexed CWB secu- rities and offer wage-indexed variable annuities to retiring members. The second class of CWB securities, with returns linked to demography and participation, would be held by the more risk-tolerant investors, including younger workers. In both cases, the NA plan will have been endowed with rules that provide automatic financial stability in the short run. This follows from the use of either the mutual fund for- mula (defined contribution) or the hiring of a sponsor that guarantees a wage-indexed annuity (defined benefits). Let us explain the term integration to financial markets: Although the basic purpose of rule-based benefits is to destroy political risk by attaining automatic financial stability in the short run, a side benefit is that members become able to direct the funds in their individual account toward the portfolios available in the financial markets. Active members are likely to differ in their degree of risk tolerance. For example, those near retirement may seek a high degree of certainty about the interest rate that will be credited in the future to their individual account balance. The plan should respond by offering several balanced portfolios with different levels of uninsurable risk to active members, and also should offer to purchase deferred annuities for those members who 158 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES fear a rise in annuity prices for the year in which they plan to start a fixed annuity. At least one of these portfolios should be a fixed-income one, such as the Thrift Savings Plan offered by the U.S. government, which has 3.2 million members.14 To create each portfolio, the plan would mix CWB securities with equities and fixed-income, long-term bonds in the financial market (purchased with the proceeds of sales of some CWB securities) in the same amount as the account balance of the members that choose the associated portfolio. Comparison between this Proposal and the Swedish ABM Consider as a starting point a country with a two-pillar pension policy. The first pillar comprises solidarity programs that help the old poor. The second pillar is a set of manda- tory savings and insurance plans that pay earnings-related pensions. The aim of these plans is to help the improvident or myopic members of the middle class save for old age more than they would have saved on their own. An example is the two-plan second-pillar created in Poland in 1997. Assume that one of these plans is an NA plan, financed by (almost) pure PAYG finance. The other plan is a fully funded defined contribution (DC) plan. At least two methods to endow the NA plan (in the second pillar) with automatic finan- cial stability can be considered. The first is the ABM method, which was adopted by Swe- den in May 2001. This mechanism relies on accounting measures of plan assets and plan liabilities. The "pure" version of the ABM method works as described below. Assume that at some point the plan's measured liabilities exceed assets. Then, the notional interest rate credited in this period to active members and the rate of adjustment of pensions in payment are both set to equal the original notional interest rate, minus the growth rate in liabilities from the previous period, plus the growth rate in assets from the previous period. The conversion factor that turns an account balance into an annuity is not modified.15 The second way to reach automatic financial stability is the rule-based IFM method, described in the previous section. Summarizing, the IFM method relabels contri- bution flows into a new contribution rate ' and a residual payroll tax at rate , such that o = ' + ; endows the NA plan with property rights over the revenue of this new tax on earnings; creates new securities (CWB securities) out of this cash flow; trades a small share of the new securities in the financial market to obtain market valuation; and crucially, adopts a rule-based benefit formula. There are many similarities between the two methods. For example, both adjust the value of liabilities automatically toward asset values, using either a rule-based benefit for- mula (IFM) or an adjustment to the notional interest rate and indexation rule (ABM). Another similarity is that neither tries to increase national savings or to reduce taxes on labor. On the other hand, the following differences between these two approaches are apparent and interesting: 1. The IFM method values the plan's assets at prices and implicit discount rates deter- mined objectively by transactions of CWB securities in the financial markets. There, many different participants, presumably guided by the profit motive, compete to make accurate projections of dividends and apply discount rates that take into account the cost of risks that impinge on these cash flows, including their correlations with other cash flows available in the local and world economy. In con- trast, the ABM method leaves valuation of the plan's assets to a single team of civil servants. This team is supervised by the political leadership, by the press, and by A MARKET METHOD TO ENDOW NDC SYSTEMS WITH AUTOMATIC FINANCIAL STABILITY 159 public opinion, including interested academics. However, the tightness of this vigi- lance is bound to be quite different from the one produced by multiple participants in the financial markets. 2. The discount rate used by the ABM method to value assets is the growth rate of the economy, g. This is clear from Settergren (2001) and Settergren and Mikula (2006). They follow the tradition started by Arthur and McNicoll (1978) and summarized by Lee (1994), which analyzes asset valuation in hypothetical golden rule economies where the real interest rate earned by physical capital is r = g. In contrast, the IFM method values CWB securities at a discount rate actually set by the financial markets. As explained in box 8.2, this discount rate must be above the growth rate of GDP to avoid an infinite price for CWB securities. This difference in discount rates implies that the capital values of account balances, overall liabilities, and over- all plan assets will be different between NA plans that adopt the IFM method rather than the ABM. This implies that the ABM method contains implicit taxes and subsi- dies between members that do not arise in the IFM. These taxes and subsidies have an impact on both equity and efficiency. For example, since high-earners live longer on average, they benefit on average from an implicit subsidy because the duration of their pension rights is longer and therefore the liabilities owed to them are over- valued in the ABM, compared with market discount rates. 3. The difference between discount rates also affects intergenerational distribution. Adopting the ABM method implies assigning a higher pro rata share of assets to plan members who hold longer-lived rights to benefits at any given point in time: that is, a higher pro rata share is assigned to the younger (active) members than to older (pensioned) members. 4. Valuation can produce surprises in the ABM. Valuation in the ABM is as follows: If there were just $1 in lifetime tax revenue paid by the initial generation, and growing at rate g, then the present value of all these lifetime taxes would be $1/(g ­ g), infinite! On the other hand, when r = g, the tax Tt is zero, so the numerator in this expression is also zero. Both effects cancel each other in the limit as the discount rate falls toward g, yielding a finite asset for the ABM method. This limiting amount is the product of the current contribution revenue (a flow) and the average difference between the timing of contributions and the timing of liabilities (usually about 30­40 years). This valuation formula for assets is not valid out of steady states, so its application in an uncertain world can create surprises. For example, assume that the unemployment rate rises because of a fall in international demand, which is only transitory. The ABM valuation formula takes the fall in revenue as permanent. The asset is estimated to fall in the same percentage in which revenue fell (the average difference in timing is essentially constant). This translates into a large cut in the notional interest rate credited to active members and a drop in pen- sions. When the shock reverses, the notional interest rate rises and the drop in pen- sions is recovered. In contrast, the IFM method takes the change as transitory and the price of CWB securities is almost unaffected. The account balances are not affected. In this example, the ABM valuation method creates risk. 5. Valuation of assets in the ABM is not affected, but the valuation of liabilities is affected when the rate of economic growth changes. Consider a case where the projected economic growth rate falls slightly but permanently. Because market interest rates are fixed by international capital mobility, and because the dividend payout of a CWB security grows less rapidly, the value of CWB securities falls, 160 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES say, by 2 percent. Consider the impact on active members (workers). In the IFM method, the price of shares is cut by 2 percent in the following day and individual account balances are cut in the same proportion. Newly issued pensions will be 2 percent smaller. In the ABM method, the valuation of assets is not affected (it does not depend on the growth rate), but the value of accrued lia- bilities rises. The net impact is an actuarial deficit, so the notional interest rate will be cut and pension amounts must fall gradually. The time-phasing of this fall is different from the one in the IFM method. There, a person about to retire whose pension rights are fully invested in CWB securities takes a 2 percent cut in pensions and future generations buy CWB securities at reduced prices so are not affected. In the ABM, the same person would take a very limited immediate cut, given by a drop in the notional interest earned in the last year. That person would also take additional cuts as each year's rate of pension indexing is below the original path, as long as she lives. A share of the loss would be passed on to future generations, because those generations would be credited a smaller notional interest rate despite the fact that market interest rates have not changed (by assumption). 6. In an uncertain world, the discount rate assigned by the financial market to CWB securities is stochastic as well. If this discount rate changes, the ABM method leaves account balances and pensions unaffected because financial prices do not enter its valuation formulae. In contrast, the IFM method appears to transmit this risk to active members (assuming average wages are not affected, pensioners with wage- indexed annuities will not be affected either). For example, if the discount rate on CWB securities rises, active members will suffer a cut in their account balances. However, the dividend payout of CWB securities is not affected by changes in the discount rate, and pensions are financed mostly from this payout.16 This creates a sort of immunization: a member just about to retire would not be affected by a rise in the discount rate if he wanted an annuity indexed to average wages. If the insur- ance company offers such annuities and backs it with the first tranche of CWB secu- rities (which are perpetuities), it would earn a higher interest on this portfolio after the rise in discount rates. Thus, annuity prices should drop in the same proportion as the drop in the member's account balance, allowing the member to purchase an annuity of the same size as before the shock. The difference, however, is that in the IFM method the contributions made after the rise in discount rates would earn a higher return, while in the ABM method they would still earn the growth rate g, which did not change. 7. Regarding the allocation of the plan's aggregate uninsurable financial risk among plan members, in the ABM method all members get a single level of uninsurable risks. In contrast, the IFM method allows the plan to offer several balanced portfolios with different levels of uninsurable risk, allowing each member to self- select according to his or her risk tolerance. The ABM method must offer a single portfolio to all its members. Therefore, the IFM method can do better by recommending a portfolio to each member as a function of the determinants of the member's predicted risk tolerance, such as the presence of other wealth, earnings variability, and access to insurance and credit lines, in addition to age. 8. Adopting the ABM method does not allow the plan to trade uninsurable risks with other participants in the financial markets, such as other institutional investors, owners of firms, and foreigners. The IFM method allows this. This option is A MARKET METHOD TO ENDOW NDC SYSTEMS WITH AUTOMATIC FINANCIAL STABILITY 161 especially valuable for smaller countries, which place a greater value on international risk diversification. 9. Distributional impact of changes in corporate risk premia. Consider a case where some members have chosen a balanced portfolio where corporate securities (bonds and equities) have a substantial weight, and their risk premia rise but the discount rate on CWB securities remains constant. In the IFM method, only the more risk-tol- erant members who chose the balanced portfolios with corporate securities are affected. They should respond to the losses by increasing their saving rate and by working more hours per year, if still active, and by cutting lifetime consumption. To compare this outcome with the one occurring under the ABM method, assume that the plan is partially funded and holds such corporate securities in the same total amount. In this case, the rise in risk premia cuts the value of the plan's fund, and this translates into a cut to the notional interest rate for all active members and into a cut in the indexation rate for all pensions in payment. The more risk-averse mem- bers are not spared. Concluding Comments It has been shown that an NA plan can be endowed with automatic financial stability in the short run. This feature is appreciated because it helps insulate the government budget from demographic pressures and in the future would insulate the pension budget from fis- cal pressures. However, the literature proved that the initial generation of benefit formulae did not achieve automatic financial stability in the short run in the face of demographic and economic shocks. Sweden responded to this challenge by creating an ABM, which was legislated in May 2001. This study discusses an alternative response, which also endows NA plans with auto- matic financial stability in the short run. The method is presented in more detail in Valdés- Prieto (2005) and labeled here integration to financial markets (IFM). In contrast to the ABM, which relies on asset valuations made by a single group of civil servants supervised by politicians and the press, this proposal bases asset valuation on market-determined discount rates set by multiple investors, presumably guided by the profit motive. In this dimension, the difference boils down to private versus public management. Both have well-known advantages and disadvantages, the magnitudes of which differ across countries. Another important difference is that the discount rate used by the ABM method to value assets is the growth rate of the economy, g. This is appropriate for asset valuation in hypothetical golden rule economies where the real interest rate earned by physical capital is r = g. In contrast, the IFM method values the PAYG asset at the discount rate actually set by the financial markets. These market discount rates come from real economies where infinitely lived assets pay a return that is higher than g, and provide the efficient incen- tives. A critical difference is that the IFM method allows the pension institution to abandon its corner portfolio and move toward the risk-return frontier, creating substantial economic gains. The participation in the financial market also allows the offer to individual mem- bers of a set of, say, five balanced portfolios, permitting a better adaptation of the pension plan to individual circumstances. This would also create immediate social gains. It is apparent that the choice between the ABM and IFM methods merits attention from policy makers. This choice has many facets. The weights that should be granted to each 162 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES one are likely to differ across economic and political settings, and therefore across coun- tries. It is possible that advanced nations that have an independent judiciary, well-regu- lated financial markets, and are well integrated to international capital markets can gain relatively more by having their pension plan engage in international risk diversification through the IFM method. Notes 1. "Notional defined contribution" and "non-financial defined contribution" should be understood to have the same definition. 2. It is important to be precise about the meaning of "short" run. For our purposes in this chapter, the "short" run is an election cycle. This is the maximum time available for adjustment before politicians respond to an aggregate deficit in the plan with legal reforms, taxes, subsidies, and other discretionary measures taken ex post. In most well- functioning democracies, political response can be faster. 3. See Valdés-Prieto (2000). The paper by Arthur and McNicoll (1978) does not discuss the issue of financial stability. On the contrary, that paper restricts itself not only to the case of steady state growth but, within that, it considers only the case where the growth path meets the "golden rule" condition (marginal productivity of capital equal to the growth rate of the economy), which implies dynamic inefficiency. 4. The original paper by Valdés-Prieto (2000) made the mistake of stating that in that event, there would be a cash surplus rather than a cash deficit (table 1 in that paper). Pro- fessor Friedrich Breyer spotted this mistake. His correction and my response are available in the Scandinavian Journal of Economics (see Breyer 2004 and Valdés-Prieto 2004). 5. See Valdés Prieto (2002, pp. 391­3). 6. See Settergren (2001). 7. See Settergren (2001). 8. With lives spanning two or more active periods, some current plan members have not yet completed their career at any given reform date. To measure the plan's liability to them, it is necessary to determine the "recognition bond" that the plan's charter grants to members who leave the plan at mid-career. This recognition might equal past contribu- tions plus interest, or, alternatively, might equal the present value of future expected bene- fits for a full career minus future expected contributions. Which approach is used is important out of steady state. 9. Such economies were labeled "golden" by an earlier literature that compared steady states without taking into account the cost of moving from one steady state to another. Rate g can be positive, zero, or negative. 10. The PAYG asset differs from a Buchanan bond--a type of public debt that pays a rate of return equal to the growth rate of national income--in two fundamental ways (see also Góra and Palmer 2004). The Buchanan bond is a debt of the state, serviced by the treasury, and its holders are protected by property rights against expropriation. In contrast, the PAYG asset is a debt of future members of the pension plan (the present value of the lifetime taxes they will pay), not the state. Morever, its holder, the pension institution, is not protected by property rights because if future legislation allows contributors to quit the plan (and, say, contribute instead to a mutual fund), or cuts the coverage of contributions, the pension institution does not have the right to claim compensation in the courts. In addition, there is a financial differ- ence: the rate of return of the PAYG asset differs from the rate of growth of national income almost surely due to variations in average taxable wages, coverage, benefits, and longevity. A MARKET METHOD TO ENDOW NDC SYSTEMS WITH AUTOMATIC FINANCIAL STABILITY 163 11. This definition of assumes that the tax rate hidden in PAYG finance is applied to earnings, when the member is active. 12. See Kendall and Fishman (1996). For example, securitization can transform assets such as the accounts receivables of a telephone company into securities that can be traded in a stock exchange. 13. Asset exchanges between the pension institution and foreign residents have no impact on the exchange rate because international capital inflows and outflows balance exactly. 14. For details of the TSP plan, see www.tsp.gov. The TSP plan is available for comple- mentary tax-favored savings for employees at the federal, state, and municipal levels in the United States, and offers five balanced portfolios, including international and national equities and bonds. 15. The factor used by the current ABM method to convert individual account balances into annuities is not sensitive to the extent by which liabilities exceed assets because it always uses a discount rate of 1.6 percent. This small divergence could be overcome in a more advanced version of the ABM method. 16. See Valdés-Prieto (2005). References Abel, A., G. Mankiw, L. Summers, and R. Zeckhauser. 1989. "Assesing Dynamic Effi- ciency." The Review of Economic Studies 56 (1), 185, January: 1­20. Arthur, W. B., and G. McNicoll. 1978. "Samuelson, Population and Intergenerational Transfers." International Economic Review 10 (1): 241­6. Breyer, F. 2004. "Comment on S. Valdes-Prieto, `The Financial Stability of Notional Account Pensions.'" Scandinavian Journal of Economics 106 (2): 387­88. Góra, M., and E. Palmer. 2004. "Shifting Perspectives in Pensions." Discussion Paper 1369, IZA Institute for the Study of Labor, Bonn (October), http://www.iza.org. Kendall, L., and M. Fishman. 1996. A Primer on Securitization. Cambridge, Mass.: MIT Press. Lee, R. 1994. "Population Age Structure, Intergenerational Transfer and Wealth: A New Approach, with Application to the U.S." Journal of Human Resources 29 (4): 1027­63. Lindbeck, A., and M. Persson. 2003. "The Gains from Pension Reform." Journal of Economic Literature XLI (March): 74­112. Richter, W. 1993. "Intergenerational Risk Sharing and Social Security in an Economy with Land." Journal of Economics (Zeitschrift für Nationalökonomie), Suppl. 7: 91­110, Springer Verlag. Scheinkman, J. 1980. "Notes on Asset Trading in an Overlapping Generations Model." Dept. of Economics, University of Chicago, unpublished. Settergren, O. 2001. "The Automatic Balance Mechanism of the Swedish Pension System: A Non-Technical Introduction." Riksförsäkringsverket, August. www.rfv.se/English. Settergren, O., and B. D. Mikula. 2006. "The Rate of Return of Pay-As-You-Go Pension Sys- tems: A More Exact Consumption-Loan Model of Interest." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 7. Washington, DC: World Bank. Sinn, H. W. 1999. "Why a Funded Pension is Useful and Why It is Not." CESifo Working Paper 195, Institute for Economic Research, Munich. (Also printed in International Tax and Public Finance 2000 (7): 389­410.) 164 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Tirole, J. 1985. "Asset Bubbles and Overlapping Generations." Econometrica 53 (6): 1499­1527. Valdés-Prieto, S. 2000. "The Financial Stability of Notional Account Pensions." Scandina- vian Journal of Economics 102 (3): 395­417. ------. 2002. Políticas y Mercados de Pensiones. Santiago, Chile: Ediciones Universidad Católica. ------. 2004. "Response to Breyer." Scandinavian Journal of Economics 106 (2): 387­90. ------. 2005. "Pay-As-You-Go Securities." Economic Policy 42 (April): 215­51. Discussion of "A Market Method to Endow NDC Systems with Automatic Financial Stability" Marek Góra* THE CHAPTER ADDRESSES A VERY IMPORTANT ISSUE, namely short-run stability of unfunded pension plans. The focus of analysis is on non-financial defined contribution (NDC) sys- tems. The author provides a set of arguments to support the view that pension systems that are not backed with financial assets are potentially unstable. More precisely, even if they are automatically solvent over the long term, they do not have the property of short- run stability. The arguments provided in the chapter are strong and lead to clear conclu- sions. The author suggests creating a mechanism called integration to financial markets (IFM) that will impose more financial stability in NDC systems. The idea--similar to various types of buffer fund existing in countries running NDC--is interesting, and worth further discussion and possibly implementation.1 Valdés-Prieto's chapter is not only technical, but it is also a very inspiring background for a broader discussion on pension systems and their reforms. The discussion starting from this chapter can lead to various specific issues. I shall address only some of them. The arguments supporting the key thesis of the chapter, namely the possible lack of sta- bility of NDC, can also be applied to pension systems using financial markets (financial defined contribution, FDC) that are not pure savings plans but are saving plans in the activity period and are turned into insurance after retirement. So arguments should take into account the method of annuitization in NDC versus annuitization in FDC. Full oblig- atory annuitization of account values irrespective of their type is natural if pension sys- tems are a part of social security. This leads to a very important point. Does implementing individual accounts mean pri- vatization of the pension system (whatever this means) or creating public-private partner- ship based on public (social) goals and private management (using financial markets) of the system? We can assume that defined contribution (DC) is a voluntary savings plan or a forced private savings plan that is treated as a voluntary one, even if it is mandatory. How- ever, we can also assume that a DC regime applied to a pension system is just a method leading to stabilization in economy. The chapter assumes the first, while for instance the design of the new Polish pension system--both the NDC and FDC part--assumes the lat- ter. So different understanding of the issues discussed in the chapter may stem from deep differences in assumptions. Defining key concepts helps the discussion. Valdés-Prieto provides a clear definition of funding. The term is often broadly used, referring to all types of savings in the economy, including ones based on political decisions. By contrast, Valdés-Prieto prefers to use the narrow definition of funding that is based on property rights. The difference between the two definitions and its consequences lead to an interesting multifold problem. * Marek Góra is professor of economics at the Warsaw School of Economics. 165 166 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Property rights are needed to protect pension rights against political manipulation. They are needed for two reasons. The first is to reduce the tax wedge; the second is to con- tribute to economic stability. The latter needs explanation. Traditional mandatory pension contributions, although perceived as "robbery" (leading to an increased tax wedge), are in fact "charity," since participants typically receive more, or even much more, than the pre- sent value of contributions paid. Implementation of individual accounts leads to actuarial neutrality. This means reduction of pension rights: that is, the amount of pension if compared with prior law is reduced. The reduction is socially just if the welfare of each subsequent generation is equally valued. The political manipulation that societies need to be protected against is an attempt to finance inflated pension rights at the expense of lower remuneration of production factors. The successful implementation of the new pension system in Poland was based-- among other factors--on the common feeling that the old system cheated people. In fact, it paid out too much. The new one (NDC + FDC) will generate much lower replacement rates. So both effects will be achieved: namely, reduction of the part of GDP allocated to the entire retired generation (GDPR), and reduction of the contribution of the reduced GDPR on the tax wedge. Promoting property rights within NDC does not need to be restricted to IFM. Upon first thought, the most straightforward idea that comes to mind would be to give the NDC rights the same legal protection as they would have if they were FDC rights, based on gov- ernment bonds with explicit property rights. From an economic viewpoint, this would be natural. In both cases, namely NDC and FDC based on government bonds, running the system is just rolling debt. It is easier to understand the similarity when the institution administering NDC is split into two parts. The first collects contributions and passes them to the budget; the other pays out benefits, receiving money from the budget. FDC based on government bonds can be described in the same way. The only difference-- and it is an important one--is the property right. So why not turn NDC liabilities into legally pro- tected rights? The only thing preventing that is accounting procedures. A country that drew the logical conclusion from the economic similarity of NDC and FDC would show huge deficits. In Europe, this would violate the Stability and Growth Pact. In the whole world, this would be badly perceived by financial markets--although economically neu- tral. So NDC is used as a clever trick to avoid the accounting problem, rather than to achieve full DC purity. Valdés-Prieto's definition of funding in the narrow sense provokes an important ques- tion. Is a funded system possible if it covers entire population? The answer "yes" is possi- ble only if government debt is assumed to be economically identical with private debt/investment. From the economic viewpoint, this is not necessarily true. Discussing this issue goes beyond the scope of my comment. The discussion should take into account not only FDC but also NDC. Commenting on Valdés-Prieto's chapter is difficult because it needs a much broader discussion that can be offered here. So I shall address only one additional issue: the argu- ment that the market rate of return, rF, is larger than the economic growth, g. There are both theoretical and empirical arguments to support that view. Irrespective of that discus- sion, it is interesting to discuss the consequences of rF > g. The pension system is an institutional framework for intergenerational exchange. Irre- spective of whether participation is voluntary or mandatory, current GDP is divided between the working and the retired generation. This is done with or without the interme- diation of financial markets. The latter method has a lot of advantages but eventually leads to the same qualitative result. Proportions of the division are subject to the market or implicit rate of return. Actually, both can be--and often are--"too" high. Traditional pen- A MARKET METHOD TO ENDOW NDC SYSTEMS WITH AUTOMATIC FINANCIAL STABILITY 167 sion systems must increase contributions to pay out pensions inflated by promises that are too high (implicit rate). We can also imagine a funded pension system based on govern- ment bonds that are overvalued by markets (the reason for this does not matter). Individ- ual accounts receive high rates of return. However, the result will be higher taxes. This is the problem of fiscal policy, but the pension system--as a large share of nations' economies--can accelerate the problem. The pension system, especially the one that covers the entire population, is a large part of the economy. What really matters is the stability of the entire economy. The stability of a pension system can contribute to economic stability--or just the opposite. A poorly designed pension system that generates ever-increasing costs can destabilize the economy. The key goal for pension reform is to stop the increase of GDPR/GDP (the pension system consuming too large a portion of GDP).2 From that viewpoint, NDC can be as effective as FDC, even if one of them or both are not perfectly stable themselves. Actually, NDC can do the job quicker since the rate of return it generates, rN, is close to if not equal to the GDP growth rate.3 Both NDC and FDC can lead to reintroducing intergenerational equilibrium, which means keeping GDPR/GDP constant over the long term. Equilibrium could be reached even without involving financial markets. However, it would be more difficult because for ordinary people, investing in financial markets is based more on intuition than on a grasp of macroeconomic fundamentals. So if their con- tributions are managed in financial markets in a way similar to their other financial resources, they will feel more comfortable and accept the concept of individual accounts. Without financial markets, the reform would be less understandable, and hence less acceptable for the public. Financial markets play a "demonstration" role. However, NDC has a very good property: namely, rN = g. This prevents inflation of pen- sion rights, which will have to be financed out of real GDP--irrespective of the type of the system. Here we could suggest extending the definition of funding formulated in the chapter. The very narrow definition of funding would refer to this part of funding, in the narrow sense that is not based on government bonds. In spite of the privileged legal status (property rights) of the government bonds, tax revenues are needed to turn them into income-financing consumption. The IFM (buffer fund)--being a short-run concept--can be less strict. Valdés-Prieto's narrow definition of funding is especially appropriate for designing the IFM financing. NDC could replace FDC in the part that is based on rolling government bonds. At the same time, NDC should be limited to the level that cannot or should not be spent on pri- vate investment instruments. FDC can be more efficient because it is much more rooted in thinking of economists, policy makers, and general public. So coming back to Valdés- Prieto's idea, making NDC more like FDC is really a good one. However, one of the pre- conditions for property rights within NDC would require some modernization of procedures applied to national accounts (the calculation of the deficit and debt in the econ- omy). This modernization is needed anyway, since without it, pension reforms will not be manageable, irrespective of their designs. So it probably will be possible to implement the idea of endowing NDC with at least some property rights. Notes 1. In the new Polish system the buffer fund has been designed in a similar way to the concept of IFM. The fund is invested in financial markets. 2. The scale of the share of GDPR to GDP does not matter for aggregate demand, but it t does for supply. 3. Technically, this can be another rate converging to g. Chapter 9 Conversion to NDCs-- Issues and Models Edward Palmer* NON-FINANCIAL (NOTIONAL) DEFINED CONTRIBUTION (NDC) SCHEMES began to take shape on the drawing boards in the 1990s, and were implemented for the first time in practice in a handful of countries from the mid-1990s. The conceptual development of NDC has gone hand in hand with the emergence of implementation issues and other technical challenges. The implementation of the first NDC schemes involved addressing a large number of design issues that had never been confronted within the traditional pay-as-you-go (PAYG) context. The initial decade of work on design of NDC schemes helped to clarify some of these issues. One of the most important design issues in introducing an NDC scheme is the conver- sion of rights acquired from the previous system--normally a defined benefit PAYG sys- tem--into NDC account values. This is the topic of this study. The study provides a systematic framework for thinking about conversion from PAYG defined benefit to NDC and identifies and discusses issues that arise in implementation. The study also examines what countries have done in practice and discusses the implications of the methods cho- sen in terms of the more generic, theoretical framework. To begin with, briefly, what is an NDC pension scheme?1 An NDC pension scheme can be likened to a lifetime savings plan where individuals regularly contribute a percentage of their income until retirement and then draw on their account throughout their retire- ment period. An NDC scheme is more than a savings scheme, however, since the capital of the deceased becomes available for distribution among the survivors, as in any typical pension insurance scheme. In an NDC scheme, an amount equal to the contributions paid by or on behalf of the participant is credited to an individual account in each accounting period. In addition, the account is credited periodically with a rate of return, based on what the system can afford--the internal rate of return (IRR). The account value constitutes a claim on future resources. Based on the account value and cohort life expectancy at retirement, from some stipulated minimum age, the NDC benefit is calculated as a life annuity. As opposed to a financial account scheme, where the assets are all held in financial mar- ket instruments, the assets in an NDC scheme are the future contributions of workers: that * Edward Palmer is professor of social insurance economics at Uppsala University and head of the Division for Research at the Swedish Social Insurance Agency. The author is grateful to Agneta Kruse for comments on an earlier draft of this study. 169 170 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES is, financing is on a PAYG basis. As opposed to a defined benefit (DB) scheme, where the contribution rate adjusts to accommodate the scale of overall benefit payments, in an NDC scheme the contribution rate is fixed and the rate of return credited accounts is that which upholds long-term financial stability, which is a function of the development of the contri- bution base. Since the assets are not held in financial market instruments, the scheme is non-financial. For this reason, it is more appropriate to call NDC a non-financial defined con- tribution scheme. Personal account schemes, whether they are non-financial (NDC) or financial defined contribution schemes (FDC) schemes,2 require a full working life to introduce. However, since accounts are not financial in the NDC framework, it is possible to convert rights acquired from previous years of coverage in a pay-as-you-go (PAYG)--non-financial, defined benefit (NDB)--scheme directly into NDC rights. This means that in practice-- given a conversion rule--an NDC scheme can be introduced fully at any time. This is in fact one of the appealing features of converting from a PAYG-DB scheme to NDC, com- pared with the counterfactual of introducing a financial account scheme. It follows that an issue that must be dealt with in conversion from a PAYG-DB scheme to an NDC scheme is the issue of how rights acquired in the old system are to be recog- nized. This raises the related questions of what are acquired rights and what constitutes fair treatment of these. The point of departure in this study is a discussion of the question of fairness in the context of converting a public PAYG pension system from a defined ben- efit (NDB) to defined contribution (NDC) scheme. In the 1990s, a handful of countries--as diverse as Italy, the Kyrgyz Republic, Poland, and Sweden--began gradual transitions from mandatory PAYG-DB schemes to NDC schemes for the working population. Latvia was the first to make a total conversion for all workers, which it did in 1996. Many more countries are now considering implementing NDC.3 This study draws upon the experiences of the original five NDC countries in dis- cussing how countries have dealt with acquired rights in conversion to NDC. The remainder of this study is organized as follows. The next section discusses acquired rights and fairness. Two principles of fairness are presented and discussed in this context. The third section presents and discusses the merits of several approaches for calculating initial capital, and relates these to the principles of fairness. The fourth section discusses issues in determining the speed of conversion. The fifth section presents examples of how countries have made the conversion in practice. The sixth section discusses other impor- tant related issues that can arise in conversion from a PAYG-DB scheme to NDC, namely disability and survivor benefits. The last section presents some final remarks. Transition to NDC--What Is Fair Treatment of Acquired Rights? Some Introductory Remarks By way of background, there is a whole body of literature on the more general topic of the distribution and redistribution of resources between individuals. In the literature, var- ious principles are developed and employed to determine the fairness of a redistributional policy for individuals or groups of individuals, as well as the effect on overall social wel- fare. As is well known, what is fair depends on the definition of fairness, which is a rather circular but unavoidable approach. Perhaps the most frequently employed definition of fairness is the Kaldor-Hicks crite- rion, in which a redistribution of resources between individuals is acceptable if total wel- fare is at least as great with the redistribution and where the winners could compensate the losers. This implies that if NDC dominates NDB alternatives in terms of potential CONVERSION TO NDCS--ISSUES AND MODELS 171 effects on economic growth, conversion to NDC can fulfill the Kaldor-Hicks criterion, even if the redistribution accompanying the conversion to NDC creates losers. In the context of this discussion, if a country has a PAYG pension scheme--of any sort-- a redistribution of resources has already taken place in accordance with the rules and design of the particular scheme. All PAYG-DB schemes create a transfer of resources between individuals, the exact outcome of which, however, may be known only ex post, when individuals have retired and benefits have been calculated according to the rules of the scheme. If in a given scheme, the present value of the expected stream of an individ- ual's benefits is greater than the present value of his or her expected contributions, he or she is the recipient of a positive transfer to lifetime resources. According to standard economic consumer theory, individuals experiencing an exoge- nously determined increase in individual lifetime income will be inclined to spend more time in leisure over the life cycle--including earlier retirement--and to save less and con- sume more prior to retirement. NDC aspires to be neutral in this sense. The underlying principle of an NDC scheme is that the amount on the individual's account reflects only his or her own contributions and an IRR granted the accounts of all participants.4 In this sense, then, an NDC scheme can yield a higher level of national wel- fare than NDB schemes that result in less individual labor supply and saving. There is a built-in "conflict" when NDC replaces some set of PAYG-DB rules, since the implication is that individual advantages gained under the DB scheme will be lost as a result of the transition. Of course, the nearer NDC the actual DB scheme is that is to be replaced, the smaller this conflict will be. Generally speaking, the winners under the DB scheme are the potential losers in a transition to NDC, whereas the DB losers are likely to be the winners.5 To judge the overall effect of a reform introducing NDC, one must also take into consid- eration the low-income guarantee that must accompany an NDC scheme, and which by definition will function as a progressive tax. There are various ways of designing the guar- antee, but these will not be discussed in this study. Instead, the discussion that follows assumes that the floor in the pension system consists of some form of guarantee, which has the clear objective of transferring public revenues to the needy elderly, but no further reference is made to a guarantee. Comparison of Two Principles The two principles for conversion of rights in a PAYG-DB scheme to an NDC scheme that will be examined here are the acquired rights principle and the contribution principle. There is a definition of fairness implicit in each: Principle 1. Acquired rights principle. A fair transition to a new system preserves acquired rights. Principle 2. Contribution principle. A fair transition to a new system gives rights based on individual contributions already paid, including a relevant rate of return.6 This is the NDC principle. These two principles also represent the two possible generic models for converting rights in a PAYG-DB scheme to NDC. They are analyzed and discussed in this section. Application of the two principles in converting rights from the old DB scheme will lead to very different results. This is illustrated with the help of an example presented in table 9.1. In the example, a DB scheme that in many ways resembles an NDC scheme is chosen to show how different the outcomes are, despite the fact that the design of the DB scheme is close to an NDC scheme.7 172 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Table 9.1. Varying Results over Time of a DB Scheme that Resembles an NDC Scheme Year of retirement Year 1 (1960) Year 40 (2000) Year 80 (2040) Life expectancy (P) from age 60 = years with benefit 17 years 21 years 25 years Years of work (L) 40 years 40 years 40 years Replacement rate (average benefit/average wage, (b/w) 0.5 0.5 0.5 Contribution rate 21.25 26.25 31.25 Source: Author's calculations. Note: See the text for details. In the example, the defined benefit to be replaced by an NDC benefit is based on indi- vidual contributions. Forty years of coverage are required to claim a full benefit in the DB scheme, and a benefit can be claimed at age 60. The contribution rate is computed as b P/w L, where b is the benefit, w the wage, L is years of work, and P is life expectancy as a pensioner. The ratio b/w is the individual replacement rate. With 40 covered years, a full benefit in the DB scheme replaces 50 percent of income from retirement at age 60. The individual in the example is an average worker who receives the average wage. For simplicity, all individuals are assumed to earn the average wage, which is set to unity in the absence of wage growth and which would grow with the rate of technical progress with growth. The only factor that changes in the example is life expectancy P at age 60, which is assumed to increase at the rate of one year for every ten years of time that lapse. By definition, the DB scheme generates increasing costs for all future birth cohorts of workers because the retirement age is fixed at 60 with forty years of work, whereas the benefit period is increasing with increasing pensioner longevity at the rate of one year for every ten years that lapse. As a result, the ratio of years of work to years of retirement is declining (for the whole country), which means that younger birth cohorts of workers must pay a higher contribution rate to support a fixed benefit here assumed to equal 50 percent of the wage. In the beginning, the contribution rate needed to finance this DB scheme is 21.25 per- cent. Eighty years later it is 31.25 percent, due solely to the additional costs created by increasing longevity.8 This can be thought of as a DB scheme implemented in 1960, and where costs have already gone up to 26.25 percent by the year 2000--with further expected increases to 31.25 percent in 2040. Since the example is constructed assuming a fixed population and constant labor force participation patterns, the cost increase is driven solely by the aging population through the increase in the dependency ratio. To illustrate the difference between the two transition principles, assume that policy makers decide to discontinue the DB scheme in year 20 and convert acquired rights in the old scheme into NDC rights. The difference in the outcomes of the two principles can then be illustrated for a person with 20 years of coverage in the "old" DB scheme. ACQUIRED RIGHTS PRINCIPLE The acquired right (the life stream of benefits) the individual can expect from the DB scheme cannot be known exactly until he or she--or the relevant birth cohort--has died. There are two obvious reasons for this. The first is that the benefit the individual is entitled CONVERSION TO NDCS--ISSUES AND MODELS 173 to under the DB scheme will typically require x years for a full benefit, in the example 40. More generally, the size of the benefit may be determined by some number of years y, which is less than or equal to x, or on final salary. The second is that longevity determines the payout period, and, typically, defined benefit schemes do not take projected cohort longevity into account in determining the size of the benefit. There are two logical possibilities for calculating the acquired right at the time of con- version to NDC: Alternative 1. Wait until the worker retires to compute the acquired right, using the "best possible" cohort projection of life expectancy at age 60 for the birth cohort.9 In this example, the individual has worked 20 years of the necessary 40, and his or her acquired right is equivalent to half a full right at retirement in year 40 (for example, the year 2000, if the starting point is 1960). This gives a benefit of 5.25,10 based on a period of 21 years of pension payments from retirement at age 60. Alternative 2. Compute the acquired right at the time of conversion. This alternative uses the life expectancy estimate at the time the conversion from the DB to the NDC scheme is to be made. In the example, this gives 19 years of pension pay- ments instead of 21, and an acquired right of 4.75.11 CONTRIBUTION PRINCIPLE This principle honors contribution payments actually made prior to the time of conversion. By design, the example in table 9.1 avoids the important question of how past contribu- tions should be indexed. However, since by assumption in this illustration. there is neither positive nor negative labor force growth, nor are there any changes in the density of con- tribution or pension payments, this is no problem.12 According to the contribution princi- ple, initial capital in the new NDC scheme equals what individuals have paid in contributions. Given the assumptions of the example, these are valued in terms of the cur- rent wage at the time of valuation.13 Initial capital is calculated as IC= Ct, where Ct denotes contributions paid during t years. The wage is set equal to unity in the example, but in practice account values would be indexed to the development of the average contribution wage, giving a present value equivalent to a current wage. Using the contribution principle, initial capital is 4.5, which is lower than in the acquired right calculation. THE DIFFERENCE BETWEEN PRINCIPLES The DB scheme embodies a tax on future generations generated by steady increase in longevity. The acquired right principle honors this tax. The question is whether this is "fair," since it honors a transfer of resources that has little to do with a transfer from the rich to the poor--a more acceptable criterion for redistribution--but rather involves a gen- eral transfer from younger to older cohorts. In addition, it is inconsistent with the decision to introduce NDC, which embodies a decision to abolish the tax on future generations aris- ing due to increasing life expectancy.14 In terms of honoring the DB "contract" reigning prior to the introduction of NDC, Prin- ciple 1, the acquired rights principle, can be said to be fair. According to this measuring stick, Principle 2 is not. Of course this reasoning presumes there is reason to believe that the first contract would have been honored. The obvious question is whether future work- ers will honor the contract. In other words, are workers in 2040 willing to pay the contri- bution rate of more than 31 percent needed to continue to honor the "contract" (which they were not a part of drawing up)? As this example has been designed, the terms of the 174 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES contract imply that transfers will increasingly subsidize the leisure of successive younger generations--it has nothing to do per se with a transfer of resources from the rich to the poor. So, where does the tolerance of future younger generations for transferring resources to subsidize the leisure of the previous generation (with increasingly better health) stop? Ex ante, decision makers can only make a qualified guess about the willingness to pay of future generations. Furthermore, for the terms of the contract to be forever valid, all future generations must believe that workers coming after them will be willing to pay more pro- portionally than they have to continue to honor this fixed DB contract for them. What is almost certain is that eventually coming generations will whittle away--with ad hoc changes now and then--the terms of the contract. In sum, straightforward application of the NDC principle for conversion is neutral vis-ŕ- vis future generations. It contains no implicit intergenerational tax and abolishes the intra and intergenerational taxes typically embedded in the reigning DB commitment. On the other hand, application of the acquired-rights principle adheres to "the contract," but in doing so preserves intra and intergenerational taxes inherent in the DB rules. The follow- ing sections discuss how the NDC principle can be applied in practice, drawing on the experiences of countries implementing NDC as examples. We return to the acquired rights principle in a later section in a discussion of the implementation of reforms, using the Pol- ish model as an example of a possible implementation technique. Calculating Initial Capital in Applying the NDC Framework This section begins by describing a generic model of initial capital based on the NDC prin- ciple--the contribution principle discussed in the previous section. This model is used to discuss the pros and cons of various implementation strategies. The subsections then go step-by-step through many of the issues arising in applying this model. A Generic Model of Initial Capital The point of departure is a generic formula for initial capital (IC) based on past individual wages, w, a contribution rate, c, applied to these to derive imputed contributions and an index, It, encompassing the rate of return on contributions made prior to time t and indexed up to this time. The time at which initial capital is calculated and credited to the individual account is T. The generic formula is T ICi = ,T c IC IC IC wi I ,t t, (9.1) t=1 and T-1 It = (1+t) IC t+1 IT =1. IC (9.2) t is the rate of return in period t used to establish a value for accounts in the accounting period t. Total capital at retirement, KT , is thus the sum of initial capital, accredited individual * accounts for pension contributions made prior to the introduction of NDC, and NDC capital earned thereafter. The benefit calculated at retirement is based on this capital, distributing it over a full life in accordance with cohort-based life expectancy at retire- ment. CONVERSION TO NDCS--ISSUES AND MODELS 175 There are three variables that need to be considered in determining initial capital: the contribution rate, cIC, past earnings, wIC, and the rate of return, . These are discussed below one at a time. Application of the Generic NDC Model: Earnings Histories The models presented in this section for the determination of initial capital are based on either actual records of individual earnings, or on countrywide macro statistics on earn- ings that enable creation of an acceptable proxy on which to base initial capital. They all embody an assumption that rights earned prior to the introduction of the NDC scheme are valued in terms of the growth of wages per capita. The internal rate of return in an NDC scheme is a function of more than the wage per capita, however. The consequences of using the per capita wage as the internal rate of return in computing initial capital are dis- cussed in a separate section below. The following models are possible to employ in determining initial capital: Model 1. Use actual contributions paid historically. If this model is chosen, then what remains is to determine the valuation of these contributions: that is, the indexation factor. The logic of this model is that workers get entitlements in rela- tion to what they have already paid. The drawback is that this is likely to give a low level of coverage in a situation where the contribution rate has increased considerably over a short time. The closer the historical contribution rate is to the current one, the easier it becomes to implement this model. Model 2. Use actual individual earnings histories combined with the forward-looking contribution rate to be employed in the new NDC scheme.15 If earnings histories are available, these can be used directly to compute initial capital, together with the current contribution rate and indexation principle for historical earnings. The same rate can be used both ex poste to compute initial capital and ex ante to compute future notional capital. Model 3. Use of individual service years. Assume that there is no reliable historical information on the exact values of individual earnings or contributions at the time implementation is being considered, but that there are reliable records of years in which individuals have worked and paid contributions. This situation was typical in the countries of the former Soviet Union and Central and Eastern European transition countries. These years are called service years.16 Records of these years can be used to compute initial capital. Use of individual service years in computing initial capital, per se, takes individual labor force participation into account, and, indirectly, age-gender participation, as well. What remains is to choose an acceptable method of determining the individual wage for years with participation. Model 3a. Combine data on individual service years with the known average wage for all covered participants at the time of implementation. The simplest procedure is to use the average covered wage for all participants the year prior to the introduc- tion of the NDC scheme. This has the advantage of starting the scheme out on a scale in line with the actual development of average covered earnings. Model 3b. Combine data on individual service years with an average--historical or cur- rent--age- and gender-related wage distribution. This model can be designed to give the same macro result as Model 3a. Whether it gives more precision depends on the actual situation. 176 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Model 3c. Combine data on individual service years with the individual's own average wage based on an available series of actual outcomes immediately prior to and/or follow- ing the implementation of the NDC scheme. This model attempts to reflect the past distribution of individual earnings and contributions based on the individual's own current status. This is fair to the extent that current individual earnings rela- tive to other workers reflect the past pattern of relative earnings.17 Note, also, that it discriminates in favor of persons in the formal sector. If applied in coun- tries where informality is a property of the labor market, using actual outcomes following implementation puts a premium on reporting and paying contribu- tions on earnings after the conversion. In sum, if previous contributions or earnings are not used to compute initial capital, due either to inadequate information or for reasons of policy, or if there are no or only inade- quate historical individual earnings data available, the earnings component in the calcula- tion of initial capital can be approximated using one of the above models, and an appropriately chosen contribution rate and rate of return. In the end, since different mod- els involve different potential advantages and deficiencies, policy makers will have to defend the arguments favoring their chosen model. The political decision should be based on analysis of available data, supported with a view of what is "fair"--given the political- economic environment. Finally, regardless of the model chosen, if some rights granted under the conversion rule are not financed within the generic NDC framework, a "tax hangover" is created. This tax hangover should be identified and a means of financing it externally should be speci- fied explicitly--as a component of the conversion. Application of the Generic Model: The Contribution Rate The contribution rate determines the scale of the system. There are two approaches to determining the contribution rate. The first and most straightforward approach is to employ a model to determine initial capital that is based on actual contributions paid, that is, model 1 above. This approach may not be desirable, however, since it may not reflect the scale of the system desired by the policy maker. Use of actual contribution histories links initial capital to what people have actually contributed, but this may not be a desirable policy. Why? If the contribution rate has been increasing over time historically, this procedure would be expected to yield lower benefits for older workers precisely because they paid contributions that were lower than what they would have had they paid according to the contribution principle. In other words, it may be a desirable policy to set a higher contribution rate for determining initial capital and to calculate--and earmark finances--for the tax overhang generated by this decision. The example in table 9.1 above can help illustrate how the policy maker can approach the problem of determining the size of the contribution rate and, by doing this, the scale of the system. In the example, the contribution rate needed to finance the DB scheme increased from 21.25 to 26.25 percent, solely because of an increase in the life expectancy of the retired over a period of 40 years: for example, between the years 1960 and 2000. The first important point is that, in the example, the country is currently paying benefits based on a contribution rate of 26.25 percent. The contribution rate could be fixed "for- ever" at the rate that is required to finance the DB system at the time the NDC scheme is to be introduced. In the example, 26.25 percent. An obvious drawback of this approach is that the policy maker may consider the current rate to be too high, in which case a lower con- version rate could--or should--be set. CONVERSION TO NDCS--ISSUES AND MODELS 177 If the forward-looking NDC contribution rate is to be set below the existing DB rate then a tax overhang will arise. It will no longer be possible to finance the existing benefits with a lower contribution rate. The tax overhang will require a source of tax financing external to the NDC scheme. In practice, this could continue to be a contribution rate that is higher than that attributed NDC accounts, which would decline with time as the system converges to the NDC rate. Note that the lowering of the contribution rate with the introduction of the NDC scheme might be a part of a larger policy package: for example, increasing the minimum age at which a pension can be claimed. This leads to an increase in the average (macro) replacement rate, which can help to moderate the effect of introducing the NDC life expectancy adjustment of benefits. In sum, application of the same contribution rate in the computation of both initial cap- ital and future contributions yields internal time consistency. For political or other reasons this may not be desirable or possible, however. If the same contribution rate is not applied to both past and future earnings, everything else equal, under-financing or over-financing may occur, depending on the choice made--but also depending on the rate of return to be applied in the calculation, which is the topic of the next section. The extent of under- financing--the tax hangover--should be estimated and a means of financing identified. This is required to maintain system financial stability within the NDC scheme. Both under-financing and over-financing lead to a redistribution of resources over birth cohorts of the population, the result of which can be studied empirically in the process of making the policy decision. Application of the Generic Model: The Rate of Return All the models for computing earnings presented above are based on individual earnings. The rate of growth of the average covered wage is a logical choice for indexing benefits up to present values in computing initial capital because it adjusts rights in accordance with actual growth of the average covered wage. The rate of growth of the average covered wage is only one of the components of the IRR, however. What are the arguments for and against taking other components into account? An important component of the IRR, together with the average wage, is the rate of growth of the covered population. Together, the rate of growth of the average covered wage and the number of covered participants determine the growth of the contribution base, and, hence, the funds available to finance NDC benefits. Figure 9.1 illustrates the development of the contribution base with a per capita (covered) wage growth of 3 per- cent and a growth/decline in the covered population of 0.3 percent per year, respectively. The latter constitutes a strong, although not entirely unrealistic, "band" around the rate of growth of the real wage rate. With steady growth, the contribution base, and ceteris paribus, the 20-year rate of return is 10 percent higher than the rate that would be ascribed accounts with the earnings models discussed above, and with a decline in the covered population 10 percent lower. The positive labor force growth scenario suggests that the affordable scale of the scheme is higher than that given by the per capita earnings model. If the NDC scheme had been running from the outset, instead of the DB scheme, ceteris paribus, per capita wage indexation would have generated a system surplus. A difficulty arises, however, with a decline in the labor force. The NDC scheme was not running during this time, however. Instead, there was a DB scheme that at the point of introduction of the NDC scheme cost exactly the amount of the sum of all individual contributions paid at the point of introduc- tion. This suggests that in terms of covering acquired rights in the old DB scheme, which is 178 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 9.1. Contribution Base Growth with Wage Per Capita Growth of 3 Percent and Labor Force Growth/Decline of 0.3 Percent 15 10 5 0 percent - 5 - 10 - 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 year Source: Author's calculations. the goal of creating initial capital for the NDC scheme, the combination of the actual con- tribution rate at the time of introduction and indexation of historical rights with the aver- age wage index will cover rights acquired in the old scheme up to the point of introduction of the NDC scheme. Ceteris paribus, use of the average contribution wage to value acquired rights at the point of introduction of the NDC scheme can be viewed as starting the system with one large increment value of first period earnings, given the contribution rate. This means that past earnings are given their current average value. This is also the value of individual ini- tial capital that corresponds to the NDC principle. Whether initial capital computed according to the above formula exactly covers the acquired DB rights depends in practice on the parameters chosen to make the conversion. Acquired Rights "Revisited" in the Context of Demography The demographic history of a country is itself an important issue in the implementation of NDC. If the working birth cohorts nearest retirement are larger than younger birth cohorts, then the system is starting with a demographically generated deficit. This is also a tax overhang from the old system, unless a reserve has already been built up in the old sys- tem that covers this overhang. What can be done about this? In principle, one can conceive of different time profiles for the payment of a tax to finance a demographic overhang. These would affect different generations differently, beginning with the current pensioners and continuing through different generations of current or future workers. Ceteris paribus, the fairest tax model--it could be argued--is to impute the tax and charge it to the "big" cohort of workers, whose total rights have not been funded in advance. After all, this was a generation whose past consumption was "too high" given what they should have paid in the NDC framework to acquire NDC rights. This could be done by reducing their initial capital accordingly. If there is no historically accumulated reserve and if the tax is not levied specifically on the "big" cohort in the form of lowering their initial capital, then there is an implicit tax CONVERSION TO NDCS--ISSUES AND MODELS 179 accumulated from previous times, the payment of which has to occur in the future. Alter- natively, if the NDC system is equipped with an automatic balancing mechanism,18 then imbalance between system assets and liabilities will eventually occur and cause internal adjustment within the NDC pension system. In sum, there is a time--and hence genera- tional--difference associated with different approaches.19 Demographic Funding The conversion of rights from the "old" PAYG DB scheme to NDC account values is a nec- essary condition for creating an NDC scheme, but it may not be sufficient to fulfill the con- dition of long-term financial stability. The main reason for this is that NDC also requires that a financial reserve be built up for relatively large cohorts of workers to help finance their pensions. The reserve fund is a practical institution that acts as the system's piggy bank. It ensures that there is "money in the bank" when it is eventually needed to meet the commitments to relatively large cohorts. It also ensures that the "right" generation pays for its own commitments. A Tax Overhang from the Old System The unfinanced demographic bubble passed on from the old system is an example of a tax overhang, as has already been discussed. There are others. Two others that typically arise in the DB setting are associated with financing old-age commitments for the disabled and financing survivor benefits. A third tax overhang that can arise is the one created by the need to phase out special privileges. In the NDB framework, liabilities to future old-age pensioners "presently" covered by disability or survivor benefits are covered by the wages of future workers: that is, those who are in the work force when the disabled and survivors become old-age pensioners. However, survivor benefits should be kept outside the NDC scheme; these benefits are discussed in a separate section below.20 A tax overhang will arise with the introduction of NDC if, before the NDC scheme is introduced, there is no financial transfer into the NDC scheme to cover old-age pension rights of persons on disability. This tax overhang is just the overhang that existed anyway and would have to be paid once persons on disability benefits became old-age pensioners. In the new NDC scheme, the old-age pension accounts of the disabled have to be accred- ited with capital financed by a transfer into the system. The overhang rights earned prior to the introduction of NDC will exist until the old commitments are phased out. In many countries, various groups of workers have been granted special privileges within the defined benefit context. These must be financed, but as in the case of disability, since they are DB benefits, the intention was not to provide financing until sometime into the future, when benefits were claimed. These will either be phased out or made external, perhaps through occupational arrangements, with the introduction of the NDC scheme. This study returns to survivor, disability, and special benefit rights in the context of transi- tion in a separate section below. Speed of Conversion, Acquired Rights, and Fairness The simplest form of transition from a PAYG-DB scheme to an NDC scheme is to start with new entrants into the labor force. The "new entrant model" takes decades until it is fully implemented, however, and for this reason a faster transition will usually be desirable. This section begins with a discussion of the new entrant model and concludes with a more gen- eral discussion of the pros and cons of different speeds of transition into NDC. 180 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Conversion Beginning with New Entrants into the Labor Force If NDC covers only new entrants into the labor market, then no estimate of initial capital is needed. By definition, the rights acquired in the "old system" are preserved, since only new entrants are covered by the new rules. For this reason, at first glance this form of implementation looks fair. Upon further consideration, there are potential drawbacks with this approach, depending on how it is formulated. For the sake of illustration, assume two identical twins, A and B. Assume also that they both have unit wages over their entire lifetimes, so that their benefits depend only on years of work, t, and the benefit rules in the old DB scheme and the new NDC scheme. Assume that the DB rule is such that with 35 years of contributions or more, it gives the same ben- efit as the lifetime-earnings-based NDC scheme does for a full earnings career from nor- mal entrance age 20. Now assume that twin A enters the labor force at age 29 and twin B at age 30, but after the introduction of NDC with the new entrant rule. Twin A is covered by the DB scheme and twin B is covered by the NDC scheme. Comparing the two schemes, all other things equal, twin B's benefit will be about 78 percent of twin A's if twin B retires at age 65. This is because in the NDC scheme, the ben- efit is based on 35 of 45 possible years of contributions from age 20 to age 65, whereas 35 years were sufficient in the DB scheme for a full benefit. This sort of discrepancy will arise as long as the new entrant rule is not set at the first possible age from which pension rights can be acquired. Also, the NDC life expectancy factor may lead to an even greater differ- ence. What can be said about fairness in this case? If fairness is defined as straightforward adherence to contracts, the results of the above application of the new entrant rule are fair, by definition. Twin A continues to be covered by the contract of the old system, while Twin B is covered by the NDC contract. However, if what we mean by fair is that the size of the benefit should be linked to the contributions, the unfair outcome is a result of the generos- ity of the DB scheme--not the conversion to the NDC scheme. In sum, for the new entrant rule to be fair in terms of treating twins equally, it must be implemented so that it applies from the youngest age at which pension rights can be earned. This gives the twins the same information and same benefit criteria from the start, and an equal opportunity to take this information into consideration in formulating life plans for work and leisure. In terms of the broader question of fairness, couched in terms of the "contractual right" to tax future birth cohorts of workers embodied in the DB rule, it can be argued that twin B is, in fact, being treated fairly, and the introduction of NDC dis- continues the unfair "advantage" of the DB legislation that favors twin A. Speed of Conversion to NDC: General Considerations The above example illustrates a more general problem associated with the speed of con- version. To begin with, equality of opportunity requires that the rules of a regime apply to all, whereas the very fact of change means that different agents will have to operate under different rule sets. In this sense, any regime change is unfair, regardless of what it entails-- and regardless of whether it happens in an existing NDB scheme or in conversion to NDC. On the other hand, in this discussion, regime change is associated with a move from a regime that passes a tax forward to future generations of workers, which can be consid- ered unfair to a regime that does not. In terms of the twins' example above, generations covered by the NDC scheme will not be creating and passing on new taxes, whereas as long as the DB scheme operates, a tax is created. One can argue that the higher the tax in the old regime, the more unjustifiable the redistribution of individual resources is under the old system and the more important it is for the conversion to occur and to be speedy. CONVERSION TO NDCS--ISSUES AND MODELS 181 In sum, a slow transition prolongs the life of a tax that itself may be considered unfair to pass on to future workers, and an immediate transition puts an end to the creation of unfi- nanced rights--or identifies a tax source to pay for these. On the other hand, the introduc- tion of NDC will be accompanied by a redistribution of pension rights, given the replaced DB scheme as the counterfactual. The further from the contribution principle the DB scheme is, the greater this redistribution is likely to be. Speed of Conversion and the Replacement Rate IMMEDIATE TRANSITION An immediate transition has the advantage of breaking clearly with the past and marking the beginning of a new system. It reduces the potential prolongation of the "twin effect" discussed above, while making it possible to reap the advantages of NDC from the outset. However, the distribution of rights will change. An immediate transition is likely to emphasize this otherwise generic NDC effect compared with the DB counterfactual. The dimension of the effect will depend on the "distance" between the NDC and counterfac- tual scheme. GRADUAL TRANSITION A gradual transition can take one of several forms. One form would be to introduce an immediate conversion, but beginning with a birth cohort that at the time of implementa- tion is x years from the normal (minimum) retirement age. However, if the overall effect of the transition on older cohorts is expected to be small, there is really no reason to wait to begin to compute benefits according to the new rules. In fact, waiting may simply lead to a pronounced difference between the two future birth cohorts retiring immediately before and the other after the date set for the conversion. The larger the expected immediate change in individual outcomes due to the introduc- tion of NDC, the more justifiable it may be to construct a gradual conversion. This sug- gests another form of gradual transition, which takes a weighted average of a benefit computed according to the old and new rules for a "transition generation."21 The advantage of a "long" gradual transition is that it reduces the effect on "losers": generally speaking, persons with shorter working careers inherited from the NDB regime.22 Of course, it also reduces the potential positive effect for "winners": persons with long contribution careers. Nevertheless, the gradual transition inadequately addresses the most justifiable reason for having a long transition: the fact that in terms of giving people equal opportunities, the transition should cover workers beginning with the first age at which pension rights can be earned, as discussed above. The disadvantage of the "long" transition is the fact that it is long, and prolongs the period during which the old system creates a tax to be passed on to future generations--and generates unjustifiable financial costs. In sum, the arguments for a fast transition appear, on balance, to outweigh the arguments against it. Country Examples The implementation of NDC introduces at least two sets of problems. The first is the deci- sion of how to deal with acquired rights under the old regime, which has thus far been the topic of this chapter. The second is how to phase in the NDC scheme smoothly, given that at the outset the entire collective of existing pensioners is still subject to the rules of the old system. The rate of return on NDC annuities is especially problematic in practice, since the method of indexation applied in the scheme being replaced may not coincide with what is 182 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES desirable within the framework of NDC, whereas a good policy can be considered to give both old system and new system pensioners the same rate of indexation of pension bene- fits. A third design issue has to do with special rights. These issues are discussed below, drawing on the experiences of countries that introduced NDC in the 1990s as illustrations of possible approaches. Tables 9.2 and 9.3 provide a summary overview of how countries have dealt with various issues. Acquired Rights and Transition Table 9.2 provides an overview of how the five countries that introduced NDC in the 1990s dealt with the issue of acquired rights. Two of the countries, Poland and Sweden, created initial capital and converted relatively rapidly, while one country, Latvia, created initial capital and converted immediately to NDC. Two other countries, Italy and the Kyrgyz Republic, have not created initial capital and have chosen a slow pace of transition. These groups are discussed separately. COUNTRIES THAT DID NOT CONVERT ACQUIRED RIGHTS INTO INITIAL NDC CAPITAL Both the Kyrgyz Republic and Italy have pro rata models to combine rights in the old and new regimes. For this reason, transition into full NDC is much longer in Italy and the Kyr- gyz Republic than in the other three countries. In the Kyrgyz Republic, everyone has had an NDC account based on contributions made since 1996--but not for previous years. In Italy, persons with fewer than 18 years of coverage before 1996 are covered by NDC; per- sons with 18 or more years are not. Notional accounts were first established in 1996. Italy. The Italian reform breaks with the NDC principle in that both employees and the self-employed are accredited with capital based on a contribution rate that is higher than Table 9.2. Transition Models for Initial Capital and Introduction of NDC Calculation of initial capital Rate of Contribution Country Earnings (WIC) return () rate (cIC) Transition model Italy No initial capital. The conditions of -- -- Persons with the pre-NDC DB formula were tight- fewer than 18 ened in 1992. With the introduction of years of contri- NDC, persons with 18 or more years butions in 1995 of contributions in 1995 remained in receive a benefit the old system, with the benefit rule based on a pro prior to 1992. rata formula, combining the old and new schemes. New entrants covered only by NDC. Kyrgyz No initial capital. Total benefit during -- -- Earnings 1996 Republic the transition to full NDC: Benefit = and after are DB transition benefit (based on service covered for the years ­ max 30 prior to 1996 and best entire population. yearly earnings in 1991­95) + NDC (based on earnings from 1996) + uni- versal flat rate (12% of average wage). CONVERSION TO NDCS--ISSUES AND MODELS 183 Table 9.2. (continued) Calculation of initial capital Rate of Contribution Country Earnings (WIC) return () rate (cIC) Transition model Latvia Initial capital calculated as recorded Average 20% Initial capital is service years and for persons retiring covered calculated for all in earnings workers claiming 1996: average wage of the covered a benefit in 1996 population 1995 or after. Gener- 1997: the individual's wage in 1996 if ally, this means over the average for all individuals in NDC capital is 1996, or guarantee calculated for 1998: the average of individual's wage women born 1996­97 if more than the average for 1940 and later all individuals, or guarantee and men born 1999: the average of individual's 1936 and later wages 1996­98 if over the average for (given the pen- all individuals, or guarantee. sion age in 1996 of 56 for women Guarantee wage for initial capital and 60 for men). 1997­99: If 30+ covered years, the overall average wage for the covered population in the relevant period. 2000­200X: the average of the individ- ual's covered wage 1996­99, or 40% of average of all covered wages for all years from 1996 up to the year a pen- sion claim is made. From 200X, straightforward rules (X is still to be determined). Poland Initial capital = present capital value -- -- NDC applies to of acquired rights from the old system all employees calculated as of December 31, 1998, the and non- day before implementation of the new agricultural scheme on January 1, 1999. New rights self-employed are based on a contribution rate of workers born 12.22%. 1949 and after. Sweden Initial capital based on individual Average 18.5% for Persons born earnings histories from 1960, for covered earnings 1938: 4/20 of everyone born 1938 and after. Contri- earnings through benefit with new bution rate of 18.5% through 1994; 1995. and 16/20 with 16.5% for 1995­97; and 16% beginning old rules, etc. 1998. Persons born 1954 + later, full NDC. Source: Author's compilations. Note: -- denotes that initial capital was not computed in the introduction of NDC. 184 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES the one they actually pay (table 9.2). This means the system is creating liabilities that are not covered by contemporary assets. Thus, as it is currently set up, the system can not become internally self-financing, even after persons with pre-1992 benefits have left the scene. Instead, a tax has been created for payment in the future. In addition, using GDP for the rate of return is only an approximation of the IRR that the scheme can pay. There is no mechanism that ensures that assets are at least as great as liabilities, the way there is in Sweden. In Italy, pension payments were about 16 percent of GDP in 1999 and are likely to rise to around 17.5 percent of GDP by 2015.23 These figures are high by international standards. The Italian NDC scheme appears to stabilize outlays in the coming half century, but at a high level. Italy's slow transition into NDC is a prime reason for this. The slow transition means that a large percentage of the labor force around 2000 will retire with non-NDC benefits. For example, Franco (2002) notes that about 40 percent of the labor force in 2000 will retire with benefits calculated according to the pre-1992 pension formula. In addition, the pronounced difference in treatment of persons who in 1992 and 1995 had small differ- ences in contributory records raises a question of equity. This concern is consistent with the analysis presented in the first half of this study. What's more, Franco concludes that the length of the transition period and other aspects of the reform may significantly reduce its microeconomic effects. The Kyrgyz Republic. The Kyrgyz Republic's scheme is for new entrants and is incom- pletely designed, especially regarding the rate of return used. With the slow transition, it will take until 2039 before the first person retires with a benefit based on the NDC formula, given entry at age 21 in 1997 and retirement at age 63. It will take until around 2060 for the entire population to be covered by NDC. With only partial wage indexation of notional capital accounts, replacement rates will decline. Furthermore, benefits are to be indexed ad hoc. There is no legislated rule for indexation. With only sporadic indexation of benefits, the share of pensions in GDP is likely to decline from an already low level of just over 4 percent.24 In sum, the Kyrgyz NDC reform will take a half century to emerge, and more work is required to bring it closer to true NDC, where the rate of return is a crucial para- meter to design correctly. COUNTRIES THAT CONVERTED ACQUIRED RIGHTS INTO INITIAL CAPITAL The slow-transition models of Italy and the Krygyz Republic can be contrasted with the conversion in Latvia, wherein January 1996 the NDC rules completely replaced rights acquired in the existing DB scheme for the entire work force at that time. In Sweden, with a slower rate of transition, all newly granted benefits are based 100 percent on NDC from 2017, and over 50 percent determined by NDC from 2007 (with an assumed retirement age of 63). The models chosen for the three countries converting acquired rights into initial capital are very different. Sweden based initial capital on earnings histories from 1960. Latvia and Poland had service books from the pre-1991 regimes with good records of years of service (including "non-contributory" periods); however, these did not provide any useable infor- mation on earnings. Hence, Latvia and Poland had to devise rules accommodating this reality to compute initial capital. Sweden. Sweden had computerized records of earnings from 1960. These were used to cre- ate notional accounts from earnings histories, using a contribution rate close to the one required to pay old-age commitments in the DB system that was being replaced at the time CONVERSION TO NDCS--ISSUES AND MODELS 185 of reform.25 In computing initial capital, Sweden used an index based on the covered wage per capita to create account values, and a contribution rate of 18.5 percent applied for his- torical accounts from 1960 though 1994. The Swedish reform also encompassed the intro- duction of a mandatory financial account (FDC) scheme, with contributions set aside for this in 1995. The contribution rate to the FDC scheme was increased to 2 percent and then 2.5 percent, and the NDC rate was the difference between the total mandatory rate of 18.5 percent and the amount set off for the FDC scheme. The long-run NDC contribution is 16 percent. Sweden also increased the minimum pension age from 60 to 61, but most workers wait longer in practice to claim an old-age benefit. Under the old system, people viewed 65 as the normal pension age, owing to a considerable reduction in a benefit claimed early from both the public scheme and occupational supplements.26 Latvia. Initial capital in Latvia is computed according to the general formula: IC = indi- vidual covered earnings × individual service years × 0.20. The latter is the contribution rate used to compute both initial capital and capital from 1996 with the implementation of the NDC scheme. This rate was lower than the contribution rate needed to cover actual old- age pension payments at the time, and was set so consciously to bring down the level of commitments. Earnings in the formula for initial capital are determined by the product of recorded service years and an individual-based covered wage. The original rule used to determine the individual's covered wage was: the average covered wage for all participants in 1995 for persons retiring in 1996; the individual's own average covered wage in 1996 for retirees in 1997; the individual's own average covered wage in 1996­97 for retirees in 1998; the individual's own average covered wage in 1996­98 for retirees in 1999; and the individ- ual's average covered wage in 1996­99 for persons claiming a benefit in 2000 and later. The choice to use the individual's own wage after 1995 as the basis for conversion of pre- reform service years to initial capital was aimed at creating a strong incentive to report income--and penalize evasion.27 This method had its drawbacks, however, as is discussed extensively in Palmer et al. (2006). Briefly, the timing and length of the period used to com- pute this wage proved to be a problem for persons who genuinely had a poor footing in the labor market in the 1990s.28 For this reason, a transition guarantee and additional tran- sitional rules were adopted for computing initial capital. These were adopted in parlia- ment in 1996 and will be in force through 2010. The Latvian transition-guarantee rule applies to persons with at least 30 years of service (or who were registered as unemployed during 1996­99). According to the guarantee, ini- tial capital is based on the best of either the individual's earnings according to the main law, or an amount based on the average wage for all participants in the years 1996­99.29 For all others, initial capital is computed as the best of either the computation using the individual's own earnings or an amount based on 40 percent of the average covered wage, which is approximately the level of the minimum wage. Hence, in practice, the minimum guaranteed amount of initial capital based on service years prior to 1996 is that which cor- responds to the minimum wage. As time passes, the importance of initial capital will decline. Total notional capital will depend more and more on actual individual earnings and contributions, and less on pre-1996 service years and the transition rule. Latvia also increased the minimum pension age (gradually) from 55 for women and 60 for men to 62 for men in 2003 and for women in 2008. Poland. Poland chose to compute the capital value of commitments earned under the old scheme "as if" these were to be paid out on December 31, 1998, and used these values for 186 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES initial capital.30 However, initial capital is calculated to deliver the same pension benefit as the old system formula (adjusted for age and contribution years), if everyone had retired December 31, 1998. The contribution rate used is 24 percent, which was approximately the cost of old-age pensions at the time of the reform. Poland chose this path for three reasons.31 First, no individual contribution or earnings data were available. Second, recalculating the accrued rights to the initial capital allows smoothing the transition. Third, this procedure provided no incentives or disincentives in the choice given workers between participation solely in the NDC scheme or both the NDC and FDC schemes in the new system, as workers received exactly the same pension rights for past years, regardless of their choice.32 The Rate of Return In principle, the NDC rate of return is the rate of return that equates system assets with sys- tem liabilities over the long run. Assets grow with the growth of the contribution base. The valuation of assets will depend, in addition, on changes in the income and mortality pat- terns of the working population, because these--together with the money-weighted length of the payout period--determine the duration of time a unit of money is in the system, called the turnover time.33 The contribution base changes with the rates of growth in the cov- ered per capita wage and the number of covered persons. Note that system reserve funds (for example arising due to demographic swings or transfers from the state budget for non- contributory rights) are also system assets. The IRR embodies all these components. Sweden uses the per capita wage as the rate of return for notional capital and annuities. This means that if one of the other factors determining the value of assets (for example, the rate of growth of the covered population, the turnover duration as determined by changes in earnings and mortality patterns, or the return on the reserve fund) or liabilities (poor life expectancy projection) causes the present value of liabilities to surpass the present value of liabilities, then the automatic balancing mechanism is triggered--and the rate of return is decreased--until balance is restored. Likewise, if the ratio of the present value of assets to liabilities surpasses a threshold value, there will be grounds to distribute the "surplus" in the system. In this sense, the Swedish NDC scheme can be said to resemble a traditional financial insurance scheme. Participants in the Swedish NDC scheme receive a rate of return on accounts deter- mined by the nominal covered wage per participant, possibly adjusted downward for "negative" balancing, or upward for "positive" balancing. Negative balancing arises when the value of system assets is less than the value of liabilities at the end of an accounting period, and--in principle--the opposite occurs when a surplus is large enough to be dis- tributed, according to a rule. The Swedish NDC annuity is calculated from the outset with an expected rate of wage growth of 1.6 percent, called the norm. Indexation of an annuity includes the rate of inflation plus or minus the difference between the outcome for wage growth and the norm. Hence, the annuity is indexed solely to inflation if growth follows the norm and if system assets are greater than liabilities--but under the level justifying the distribution of a surplus. Even the benefits of persons who retired under the old rules are adjusted with deviations from the norm, in order to give both old-system and NDC pen- sions the same indexation. The use of the norm made it possible to grant a higher initial annuity, but at the expense of a declining relation between a fixed real-valued annuity in relation to the average wage of contributors. This is equivalent to the result obtained when a defined benefit is price indexed, all other things equal. The alternative to this procedure would have been straightforward wage indexation, but from a lower initial annuity value. CONVERSION TO NDCS--ISSUES AND MODELS 187 Italy's scheme also embodies a norm in the calculation of the annuity,34 using a value of 1.5 percent, and based on the growth of GDP. GDP growth is a proxy for the growth of the system's contribution base and using GDP growth implies an assumption of a fixed ratio of the wage sum (and, hence, profits) to GDP. It also implies that the GDP measure of the wage sum is a sufficient measure of covered earnings, which in economies with a large degree of informality will not be the case. System design in the other original NDC countries--the Kyrgyz Republic, Latvia, and Poland--differ from the Swedish model in various ways. First, for example, notional capi- tal in Latvia earns a rate of return determined by the rate of growth of contribution wages (contributions), while Poland uses 75 percent of contribution wages and the Kyrgyz Republic 75 percent of the average covered wage. The lower-than-100-percent rates of return have been set purposely by politicians to hold down pension costs.35 An alternative would have been to set a lower contribution rate for notional capital, which would have established a lower scale for the scheme--instead of "taxing" the return--by setting index- ation below 100 percent. The latter yields unnecessarily low pensions, for example, com- pared with a financial scheme where there is no such tax. Clearly, a tax of this nature will undermine support for the NDC scheme. Second, none of these countries has full wage (or wage sum) indexation of benefits because at the point of conversion to NDC, practically all benefits were determined by the previous DB scheme, and were price indexed. No country has found it plausible to give price indexation to one group of pensioners and wage indexation to another. For this rea- son, wage indexation is also a transition issue. A country can afford to do this as the NDC scheme increases its share in total benefit payments. Third, if, ceteris paribus, the growth path of the labor force--and covered participants-- is negative, a need for "negative" balancing will accompany the decline if indexation is based on the covered per capita wage rather than the development of the contribution wage sum, that is, the combined effect of the development of the per capita labor force. In this respect, the effect of balancing will be the same as applying a wage sum index, all other things equal. All other things are not equal, however, and in practice balancing will depend also on other factors affecting the value of system assets and liabilities in the accounting period, such as the value of a reserve fund. Furthermore, the balancing mecha- nism designed for Sweden requires sophisticated data, which are still not available in countries like Latvia, the Kyrgyz Republic, or Poland. Perhaps when data are available, these countries will choose to convert to this more sophisticated procedure. Finally, none of the other countries had the large-scale reserve fund that Sweden had at the point of introduction of NDC. This means that if negative labor force growth is likely from the time of introduction,36 the scales are tipped in the direction of continuous nega- tive balancing due to the effect of negative labor force growth on the contribution base-- and hence contribution assets. Wage-sum indexation of notional capital performs this balancing function in practice in Latvia and Poland, although both countries have chosen not to apply this indexation mechanism to pensions. Special Rights In principle, special rights, if they exist, need external financing. Special rights can even occur in the NDC system, but they require external financing to maintain the financial sta- bility of the system. Special rights will be tax-financed either over the budget, or by neces- sitating a higher contribution rate than that accredited to accounts. Italy is a special case. It introduced unfinanced special rights (see column one of table 9.3) when it introduced NDC. These will create taxes on future workers. 188 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Table 9.3. Other Transition Issues Occurrence of "unfinanced" Rate of return on Rate of return on Country special rights notional capital annuities Reserve fund Italy The contribution rate for GDP index GDP index No employees is 32%, but 33% is used to calculate account values; for the self- employed, the contribution rate is 15%, but 20% is used to calculate account values. Kyrgyz Special rights from the old 75% of wage Indexation by ad No Republic regime were reduced by growth hoc political about 70% in legislation decision from 1997. Latvia Account values for persons Covered wage Inflation adjust- Not at time of with special rights under sum ment 1996­2002, implementation, the old regime are adjusted and inflation but reserves are according to specific adjustment plus projected to turn formulas, and all special 25% of the real positive from rights are phased out wage sum from ~2005 under the new NDC law. 2002, and 50% Beginning in 2003, "special real wage sum rights increments" are from 2011 financed from general state budget revenues. Poland Farmers are excluded from 75% of the cov- Inflation adjust- A contribution the NDC scheme, but are ered wage sum ment plus at least rate was covered by a separate 20% of real wage earmarked to scheme, approximately growth, the latter build up 95% subsidized by general by ad hoc politi- reserves, but tax revenues. Bridging pen- cal decision general budget sions were created for per- deficits have sons with special rights, thus far prohib- and are financed with state ited this in prac- budget revenues. tice Sweden No special rights in either Covered contri- Covered contri- Large reserve the old or new NDC butions per butions per fund from the scheme. participant + participant + outset automatic automatic balancing balancing Source: Author's compilations. Reserve Funds and Other Considerations As already noted, Sweden is the only country introducing NDC that came into the transi- tion with a reserve fund. This fund is especially important for demographic reasons, for, among other things, the large birth cohorts from the 1940s will retire around 2010­15, using some of the reserves. Sweden had a reserve fund in the old PAYG scheme that was sufficient at the time of transition to pay approximately three years of current benefit pay- CONVERSION TO NDCS--ISSUES AND MODELS 189 ments (even after transferring about 40 percent of the fund to the general budget as a part of the conversion from the old to the new regime). The eventual demographic overhang from the old system is a part of the overall liability of the PAYG scheme, encompassing old and new NDC benefits, while the inherited fund is a part of the overall assets. Italy has no reserves, and current calculations show that, instead, the contribution rate will have to be increased (see the discussion above) to cover increased benefit outlays. The Kyrgyz Republic should develop reserves in the next decade, given that it has no pension indexation rule and the minimum pension age progressively increases to 63. On the other hand, these reserves will increase the scope for better indexation of and all PAYG benefits, including gradually rising NDC benefits, continuing the transition process. Poland intro- duced an earmarked contribution rate to contribute to the accumulation of a reserve fund. Because a large deficit in the state budget arose from around the time the NDC scheme was introduced, to date, the government has not allowed these reserves to develop. Latvia will develop positive reserves in the overall PAYG system from around 2005, cal- culations indicate.37 These will be needed around 35 years later to finance the Latvian baby boomers, born around the late 1980s. Latvia also introduced a mandatory financial account scheme in 2002, with a contribu- tion rate of 2 percent in 2002­6, decreasing the contribution rate for the NDC scheme to 18 percent.38 A further exchange of up to 10 percent each for the NDC and FDC schemes is scheduled in the legislation. All other things equal, since many workers have notional cap- ital based on a rate of 20 percent, this decease in the NDC contribution will create a finan- cial deficit that has to be tax-financed. However, calculations show that because of the increase in the minimum pension age for women from 55 and for men from 60 to 62, con- siderable room is created for introducing the mandatory financial scheme39 without creat- ing a serious transitional financial deficit. Whether a serious financial deficit arises depends on the scale of the FDC scheme together with the degree of indexation of NDC benefits. A contribution rate to the mandatory financial scheme that surpasses 6­7 percent creates a tax on the transition generation that increases with the scale of the FDC scheme,40 whereas a lower contribution rate to the FDC scheme leaves more room for eventually increasing the scale of wage indexation in the direction of 100 percent, and finishing the NDC transition in this respect. Other Important Issues in the Overall Conversion to NDC Disability Benefits In the DB framework replaced by NDC, the old-age commitments to recipients of disabil- ity benefits are paid by workers when the disabled have reached and have passed the age of retirement. In this context, costs associated with today's decisions to grant disability benefits carry a commitment to pay benefits in old age. NDC does not work this way. Instead, what is consistent with the NDC framework is the idea that contributions should be paid into the old-age scheme to finance disability benefits on a current account basis. In principle, this means money should be transferred to NDC accounts to cover the old-age pension rights of recipients of disability benefits on a regular--for example yearly--basis. Two models of dealing with disabled persons have been employed in countries con- verting to NDC. One has been for the disability system to make contributions (through state general revenues) to the accounts of disabled persons, based on imputed earnings during the time of disability. The other is to base the transfer to the accounts of the dis- abled persons on some general rule applied to all, without taking individual foregone earnings careers into account. The first is more in line with the insurance idea, and has been applied in Sweden, for example. 190 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES A related issue is whether the disability recipient should be allowed to continue to receive a disability benefit on a lifelong basis, to receive the best of either the disability or old-age benefit, or should be restricted to taking the old-age benefit from a specific age. Different models are currently employed in different countries. If people are allowed to keep the best of a disability benefit or an old-age benefit based on notional capital accounts, in principle, there should be financial transfers arranged from the disability to the old-age scheme(s). If the disability benefit is higher than the old- age benefit, and the claimant is allowed to retain the higher of the two, this is equivalent to granting a special right and the extra amount should be financed by a transfer from the disability scheme into the old-age scheme(s). A third issue involves the timing of the conversion from disability to an old-age benefit. In principle, this should be at the minimum age to claim an old-age benefit. This is the case in Latvia, for example, but not in Sweden. In Sweden, the conversion occurs at age 65. This means that it is normally to the advantage of individuals to be on disability before they turn 65, rather than to claim an actuarially fair old-age benefit from age 61. Survivor Benefits The death of workers in an NDC scheme leaves an "inheritance gain": that is, an undis- tributed amount of money. This can be distributed to surviving workers in the old-age scheme, on a birth cohort basis and at the minimum age for claiming a benefit. This creates a better benefit for all who survive to the age of retirement. The alternative is to use this surplus for some other purpose. One example would be a benefit for a surviving caregiver with children. Another example would be to create a traditional survivor benefit. There is a question of principle regarding creating a survivor benefit along traditional lines, however. A survivor benefit for a spouse (not caring for children) puts single persons and couples on different statuses and, more generally, can be argued to be inappropriate in a county where the goal is gender equality. A surviving spouse if working is no different from any single person working, and a surviving spouse not working is, in principle, no different than any other person who is not working. For this reason, it is not clear why a working surviving spouse should receive the inheritance gain, rather than the whole col- lective of working participants in the pension scheme. One might argue for some form of short-term adjustment benefit to help make the transition from married to single-- although even here, the problem is hardly different from that arising in a divorce. Generally speaking, one can only argue that is justifiable to compensate time spent at home caring for younger children. This time can be compensated for using noncontributory childcare rights. Any other financial support provided to survivors with children should be arranged outside the pension system. In sum, it is difficult to argue for a traditional survivor benefit where gender equality is an explicit goal. The traditional spouse benefit for working-age persons is from a time when women were not expected to participate in market work. Finally, it is also difficult to find a fair rule for splitting NDC accounts between spouses. If the argument against a gen- eral (traditional) survivor benefit is accepted, then what remains is to calculate acquired rights for survivors under the old regime and to phase out this benefit. Of course, the alter- native of creating a traditional survivor benefit using the inheritance gains will always be a possible choice for a country that does not accept the above argument. Final Remarks This study has examined issues occurring in the transition to NDC. A major conclusion is that Sweden is an outlier in the process of conversion to NDC, and because of this has been CONVERSION TO NDCS--ISSUES AND MODELS 191 able to fulfill all conditions for NDC. Perhaps the most advantageous feature of the pre- reform landscape was the existence of a large reserve fund. The Swedish reform also had the advantage of a relatively fast and total transition to NDC. As opposed to Sweden, Italy has moved forward in a piecemeal fashion, and because of this, discussion about reform is an ongoing feature of the political landscape. As in Sweden, the reforms in Latvia and Poland have been overall reforms, although transition issues remain concerning the formulation of the rate of return and the creation of reserves. This is not the result of oversight, but a function of the realities of transition, some political and some due to circumstances. The most important restriction on the rate of return on annuities in the NDC scheme is the indexation formula for current pensioners, whose pensions were granted according to the old regime. However, it is possible to achieve a better return as the NDC scheme grows. The introduction of NDC in the Kyrgyz Republic is in a nascent phase. It will be so for some time to come because it only applies to earnings from 1996. By contrast, Latvia, Poland, and Sweden began by computing values of initial capital for current workers. Nevertheless, the model employed in the Kyrgyz Republic is one of many possible models of introducing NDC. That said, it is important to point out that the Kyrgyz model needs further development if participants are to receive a good replacement rate. Even Poland's NDC scheme is suffering from the decision to give notional accounts 75 percent rather than 100 percent of a possible rate of return: that is, the decision to "tax" accounts. These experiences show that it is important to set the NDC framework--the contribu- tion rate and rate of return--so that the resultant income replacement rate is reasonable, given other mandatory schemes. For example, Latvia, Poland, and Sweden have all intro- duced mandatory financial account schemes to supplement NDC. It is the overall pack- age--the expected replacement rate from the mandatory schemes, including the variation in rates of return--that is important. Finally, even the room left above the ceiling in the mandatory scheme will determine the room left for development of individual, private coverage. In sum, in system design it is important to leave room for private initiative, while at the same time satisfying a need to provide good mandatory coverage for the aver- age wage earner. Notes 1. See Palmer (2006), for a more thorough definition of an NDC scheme. 2. This study employs the terminology developed in Góra and Palmer (2004) to distin- guish between financial and non-financial schemes, depending on whether they are financed through purchases of financial market assets or through the future contributions of workers, and are defined benefit as opposed to defined contribution schemes. This leads to four possible schemes: Financial defined contribution (FDC), which in much of the liter- ature is loosely called "individual account" or "privatized" schemes; non-financial defined contribution schemes (NDC); financial defined benefit schemes (FDB); and non- financial defined benefit schemes (NDB), loosely called "public" schemes or pay-as-you- go schemes. The point of this dichotomy is to focus on the importance of distinguishing clearly between NDC and NDB as two forms of non-financial schemes, with very different economic logic, and between FDC and FDB as two forms of financial schemes, with poten- tially very different contractual logic. 3. See Felderer, Koman, and Schuh (2006) for Austria; Chlon-Dominczak and Móra ´ ´ (2006) for the Czech Republic; Börsch-Supan and Wilke (2006) for Germany; Domínguez- Fabián and Vidal-Meliá (2006), and Takayama (2006) for Japan. Also, in 2002, Russia intro- duced a version of NDC. 192 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES 4. As is also generally recognized, NDC is still, however, only second best in terms of economic efficiency. 5. The same discounting factor applies to both a traditional PAYG-DB and an NDC scheme, so that in a comparison, this factor can be disregarded. 6. The IRR is the rate of return the system can afford, given the present value of contri- butions and benefit liabilities. This concept will be elaborated on in the context of conver- sion below. 7. Account values and benefits can be viewed as arising in a steady state world with or without technical growth. In the latter case, accounts and benefits are indexed with aver- age wage growth. An alternative assumption that would yield the results in table 9.1 is that of a point system, where each year individuals get accredited with a point that is the ratio of their own covered wage to the average covered wage. Individuals earn the average wage and account values are valued at retirement by the average wage of all contributing participants at the time the valuation is made. Benefits continue to be indexed by the aver- age covered wage. If the size and composition of the labor force are unchanged, financial balance "simply" requires forever increasing the contribution rate in line with what is required to finance the system with increasing life expectancy. 8. An obvious "fix" for this DB scheme would be to introduce a life expectancy indexa- tion of the full benefit pension age. This would bring it closer to the NDC scheme. To exactly create NDC from this scheme, one will still have to go to lifetime accounts, together with an annuity based on life expectancy at retirement and indexation with the internal rate of return. Demographers and actuaries have notoriously underestimated the future development of life expectancy in their projections. 10. The full benefit equals 0.5 times the unit wage and is paid out over 21 years of retire- ment: that is, B1 = 0.5 × (0.5 × 1 × 21) = 5.25. 11. That is, B1* = 0.5 × (0.5 × 1 × 19) = 4.75. 12. The next section of this chapter deals with the more complicated question of index- ation of past rights. The more general framework of NDC is important in this context. 13. This would be exactly what the accounts in a point system indexed to the growth of the contribution wage would yield. The difference between the NDC and point system, given that the point system also employs cohort employees life expectancy at retirement to calculate the life annuity, would then rest upon whether (how) the macro financial condi- tion--that system assets should be at least as great as system liabilities for a given contri- bution rate--is fulfilled in the point system. 14. Note that there is an opportunity for policy makers to use the tax implicit in the DB scheme for another purpose, for example, to retarget resources to needy elderly instead of providing a general "longevity dividend" for everyone. 15. This is the model employed in Sweden and is discussed below in the section on what countries have done. 16. Workbooks with recorded service years were kept in the former Soviet Union and Central and Eastern Europe transition countries prior to the fall of communism and the emergence of independent states. 17. This is the approach adopted by Latvia in 1996. Exactly how this approach was designed and the pros and cons of applying it are discussed at length in Palmer et al. (2006). 18. Readers not familiar with this concept are referred to Settergren (2003), Settergren and Mikula (2006) or Palmer (2006) to understand its construction and role in the NDC framework. Basically, given an estimate of system assets, A, and liabilities, L, it adjusts CONVERSION TO NDCS--ISSUES AND MODELS 193 account values automatically when assets fall short of liabilities and can be used to increase liabilities (indexation) when assets are greater than liabilities. 19. For the big cohort the question is whether they should bear the whole burden through a reduction in initial capital, or share the burden when they become older and retire with younger generations of workers. 20. There is a separate issue as to whether notional accounts could (should) be split between spouses. In considering this issue, one has to consider whether it should be mandatory or voluntary: if it is voluntary (with both parties agreeing), at what point a decision has to be made (at the time of marriage, declared cohabitation, or another time, and whether a decision can be reversible), and what the consequences are at separation prior to retirement due to death or divorce. Many problems arise, and to date, no country has legislated a mechanism for sharing accounts. 21. See the example of Sweden below. 22. Note, however, that countries introducing NDC systems have created special rights for parents (mothers) for years of child care immediately following childbirth, which are financed externally. 23. See Franco (2002). 24. Based on data from IMF (2003) and the Kyrgyz Pension Fund. 25. Childcare credits were almost exclusively accredited mothers and were also imputed retroactively on the basis of historical records. 26. This is the age at which a full occupational supplement to the old DB scheme applied. Following the reform of the public system, the majority of occupational schemes have also converted to defined contribution, but claiming a benefit before age 65 gives a much poorer outcome. In addition, the minimum guarantee can be claimed first at age 65, which has the effect of keeping healthy low-income workers in the labor force until this age. 27. An alternative, the use of the overall average wage for all contributors, would have favored low-income earners and penalized high-income earners. In this sense, it would also have been unfair. Unfortunately, some people were in a position to gain from these transition rules by negotiating with their employers to declare higher wages during this short period. 28. See Vanovska (2004). 29. Note, however, that this guarantee also covers persons who consciously evaded mak- ing full contribution payments. This is an inadvertent consequence of having such a rule. 30. See Chlon-Dominczak and Góra (2006) for a more in-depth discussion. ´ ´ 31. See Chlon-Dominczak and Góra (2006). ´ ´ 32. This can be compared with the reform in Hungary, where people traded rights in the old system to participate in the new financial account system. 33. See Settergren and Mikula (2006). 34. See Gronchi and Nisticó (2006). 35. See Chlon-Dominczak (2002) for the experience in Poland. ´ ´ 36. The Kyrgyz Republic had a period fertility rate 2.1 or higher in the 1990s. However, there is a net outflow of workers to Russia and other neighboring countries, according to national statistics. Latvia's fertility rate is expected to remain well below 2.1 per woman over the coming half century, combined with a net outflow of working age persons (Lat- vian Ministry of Welfare 2003). 37. See Palmer et al. (2006). 38. The financial account scheme is mandatory for persons under 30 years of age July 1, 2002 and is voluntary for others up to 50 years of age in 2002. Participation was relatively 194 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES small at the outset because of little voluntary interest. The contribution rate to the financial account (FDC) scheme is scheduled to increase again in 2007 to 4 percent, and possibly up to 10 percent thereafter. 39. See Latvian Ministry of Welfare (2003) and Palmer et al. (2006). 40. See Latvian Ministry of Welfare (2003). References Börsch-Supan, A. and C. B.Wilke. 2006. "The German Public Pension System: How It Will Become an NDC Look-Alike." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 22. Wash- ington, DC: World Bank. Chlon-Dominczak, A. 2002. "The Polish Pension Reform of 1999." In Pension Reform in ´ ´ Central and Eastern Europe Volume 1, ed. E. Fultz. Budapest: International Labour Organisation (ILO). Chlon-Dominczak, A., and M. Góra. 2006. "The NDC System in Poland: Assessment after ´ ´ Five Years." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribu- tion (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 16. Washington, DC: World Bank. Chlon, A., M. Góra, and M. Rutkowski. 1999. "Shaping Pension Reform in Poland: Secu- ´ rity Through Diversity." Social Protection Discussion Paper 9923. World Bank, Wash- ington, DC. Chlon-Dominczak, A., and M. Mora. 2006. "The NDC Reform in the Czech Republic." In ´ ´ Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 21. Washington, DC: World Bank. Domínguez-Fabián, I. and C. Vidal-Meliá. 2005. "The Spanish Pension System: Issues of Introducing NDCs." In Pension Reform: Issues and Prospects for Non-Defined Contribution (NDC) Schemes, eds. R. Holzmann and E. Palmer, chapter 23. Washington, DC: World Bank. Felderer, B., R. Koman, and U. Schuh. 2006. "Investigating the Introduction of NDCs in Austria." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 20. Washington, DC: World Bank. Franco, D. 2002. "Italy: A Never-Ending Pension Reform." In Social Security Pension Reform in Europe, ed. Martin Feldstein and Horst Siebert. Chicago: University of Chicago Press. Góra, M., and E. Palmer. 2004. "Shifting Perspectives in Pensions." Institute for Study of Labor (IZA) Discussion Paper 1369. Bonn, Germany. Gronchi, S. and S. Nisticň. 2006. "Implementing the NDC Theoretical Model: A Compari- son of Italy and Sweden." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 19. Wash- ington, DC: World Bank. IMF (International Monetary Fund). 2003. Kyrgyz Repubic: Selected Issues and Statistical Appendix. IMF Country Report No. 03/53. IMF, Washington, DC. Kotlikoff, L. J. 2004. "Generational Policy." In The Handbook of Public Economics, Second Edition, ed. Alan J Auerbach and Martin S. Feldstein. Amsterdam: North Holland. Latvian Ministry of Welfare. 2003. The State Social Insurance System in Latvia, Financial Analysis 2002­2070. Riga: Latvian Ministry of Welfare. CONVERSION TO NDCS--ISSUES AND MODELS 195 Lindbeck, A., and M. Persson. 2003. "The Gains from Pension Reform." The Journal of Eco- nomic Literature 41: 74­112. Palmer, E. 1999. "Individual Decisions and Aggregate Stability in the NDC System." World Bank, Washington DC. Processed. ------. 2000. "The Swedish Pension Reform Model: Framework and Issues." Social Pro- tection Discussion Paper 0012, World Bank, Washington, DC. ------. 2002. "Swedish Pension Reform: How Did It Evolve and What Does It Mean for the Future?" In Social Security Pension Reform in Europe, ed. Martin Feldstein and Horst Siebert. Chicago: University of Chicago Press. ------. 2006. "What Is NDC?" In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 2. Wash- ington, DC: World Bank. Palmer, E., S. Stabina, I. Svensson, and I. Vanovska. 2006. "NDC Strategy in Latvia: Imple- mentation and Prospects for the Future." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 15. Washington, DC: World Bank. Samuelson, P. 1958. "An Exact Consumption-Loan Model of Interest with or without the Social Contrivance of Money." Journal of Political Economy 68 (6): 467­82. Settergren, O. 2001. "The Automatic Balance Mechanism of the Swedish Pension System-- A Non-technical Introduction." Wirtschaftspolitishe Blätter 4/2001: 339­49. Settergren, O., and B. Mikula. 2006. "The Rate of Return of Pay-As-You-Go Pension Sys- tems: A More Exact Consumption-Loan Model of Interest." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 7. Washington, DC: World Bank. Takayama, N. 2006. "Reforming Social Security in Japan: Is NDC the Answer?" In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 24. Washington, DC: World Bank. Valdés-Prieto, S. 2000. "The Financial Stability of Notional Account Pensions." Scandina- vian Journal of Economics 102 (3): 395­417. Vanovska, I. 2004. "Pension Reform in Latvia: Achievements and Challenges." Reforming Public Pensions. Sharing the Experiences of Transition and OECD Countries. Paris, OECD. Discussion of "Conversion to NDCs--Issues and Models" Marek Góra* ALTHOUGH ONLY A FEW COUNTRIES (LATVIA, POLAND, AND SWEDEN) have fully introduced non-financial defined contribution (NDC) schemes,1 and a few others have introduced them to an extent, the NDC concept is increasingly considered in discussions and reform proposals all over Europe and elsewhere. This is the case for a number of reasons, includ- ing relatively easy conversion to NDC from defined benefit (DB) pay-as-you-go (PAYG) plans. It is true that introducing NDC is easier than introducing financial defined contri- bution (FDC). This does not mean, however, that NDC does not need a deliberate approach. A number of options can be applied both at the general and technical level. Edward Palmer's study is probably the most comprehensive, critical presentation and dis- cussion of options available in the course of conversion to NDC. In his chapter, Palmer also compares the experience of countries that have introduced NDC to their pension sys- tems. He highlights the pros and cons of various ways of conversion to NDC. The chapter can serve not only as an economic analysis but also as a "manual" for those who consider conversion to NDC as an option for other countries. Palmer provides the reader with a very clear view on NDC and its implementation. In this particular study, however, Palmer does not address the economic nature of NDC. In this regard, the following questions are pertinent. How does conversion to NDC differ from conversion to FDC? More generally, what is the key difference between the two options: namely, introducing individual retirement accounts in a non-financial way (with the returns stemming directly from real economy growth) or in a financial way (with the returns stemming from real economy growth via financial markets)? The questions are beyond the scope of Palmer's chapter. However, the problem of whether to convert all or part of DB-PAYG into NDC (or FDC, or a combination of the two) is crucial for pension reform. Two factors are probably the most important for the decision. First is the "fairness" of the change of the regime (whatever "fairness" means). The second is the "transition cost." How do these two issues differ, depending on whether NDC or FDC is to be imple- mented? The chapter broadly discusses the fairness of conversion. However, whatever the defin- ition of fairness used, it applies to conversion to NDC and to FDC in exactly the same way. This message is important for the reader, especially for a policy maker involved in public debate about the reform options. It makes sense to formulate this conclusion explicitly in the "manual." The chapter does not address the issue of transition costs. However, that issue is clearly one that should be discussed in a broader context of conversion from the traditional approach (DB-PAYG) to a modern approach (NDC, FDC). Palmer's chapter is a very good starting point for such a discussion. The study provides readers with a number of indirect * Marek Góra is a professor at the Warsaw School of Economics. 196 CONVERSION TO NDCS--ISSUES AND MODELS 197 hints--that the difference between conversion to NDC and to FDC stems from accounting standards, to a large extent. Actually, NDC is a clever "trick" to solve a fundamental eco- nomic problem, namely converging to intergenerational equilibrium, without generating problems related to accounting conventions. The guidelines Eurostat, the statistical agency for the European Union (EU), has recently formulated for EU countries on how to account flows through FDC-type individual accounts formally do not apply to NDC. They are not, which is bad news (economic inconsequence) and good news (fewer short-term problems for reformers) at the same time. I do think that both types of individual accounts differ from each other. That is why Poland has followed a well-balanced approach and com- bined NDC and FDC, and has chosen the largest share of FDC in the entire old-age pen- sion system among countries that have also introduced NDC. The difference, however, has little to do with the way government's deficit and debt are accounted. This is the message that matters for implementation of pension reforms. Returning to Palmer's chapter, the study compares implementation approaches in five countries (Italy, the Kyrgyz Republic, Latvia, Poland, and Sweden) and concludes that the Swedish approach is the most sophisticated. It is difficult to disagree with this conclusion. For many reasons, including less-developed data processing and unavailability of reserves left by the previous system, the Swedish approach to balancing the newly introduced NDC was not applicable in countries such as Latvia and Poland. However, is the less- sophisticated approach really worse? Should it, as Palmer suggests, be changed for the more sophisticated one? The simple approach, precisely because it is simpler, is easier for the participants to understand. It is also more transparent. Isn't the simple (straightfor- ward) approach more suitable for NDC? Note 1. "Non-financial defined contribution" is alternatively called "notional defined contri- bution" should be understood to have the same definition. Discussion of "Conversion to NDCs--Issues and Models" Elsa Fornero* AS DEBATE AND EMPIRICAL WORK ON PENSION SCHEMES PROGRESS, transition phases stick out as the really important element at the level of economic policy. This is true for two rea- sons. First, transition phases last so long that they outlive politicians, and it is hard for politicians to set themselves goals that will deliver benefits they will hardly be able to see. Second, during transition phases, a highly critical redistribution element is unavoidable. So, how can these phases be controlled? Are rules possible or must politicians be left with- out guidance and able to act on whim? Economists are normally very good at designing steady state solutions; it goes to Edward Palmer's great credit that he is not afraid to grap- ple with the far more messy and less gratifying issue of transition. The transition that he is interested in is the one from a defined benefit (DB) to a defined contribution (DC) pay-as-you-go (PAYG) system, the so-called notional defined contribu- tion (NDC)1 or Bismarckian scheme, characterized by pension formulae that, as a general rule, correlate benefits with payroll taxes paid over the whole working career and with retire- ment age. Palmer introduces readers to a very detailed menu of transitional issues, which he dis- cusses according to two criteria: recognition of acquired rights and acknowledgment for contri- butions paid. Before commenting briefly on his points, I will observe that his presentation implies that at least two preliminary questions have been answered. Is a NDC scheme preferable to a DB one? Doesn't it impose too much risk on workers? The implied answers are "yes" to the first question, and "no" to the second. Although I broadly agree with these answers, I do not think they are obvious. In my view, and maybe also in Palmer's, a DC scheme is preferable to a DB one because--at a properly defined internal rate of return--it ensures both a stable financial equilibrium and a "fair" treatment, within and between cohorts. Of course, one feature is the mirror of the other, provided the definition of what is "fair" rests on the concept of actuarial fairness as the basic principle for calculating benefits. This means that lifetime payroll taxes are "notionally" accumulated, earn a return equal to (an average of) the growth rates of either the wage bill or GDP, and are converted at retirement into an annuity according to mortal- ity tables. This method has many advantages, but is far from optimal. The possibility of achieving an almost automatic financial equilibrium reduces the need for politically determined cor- rections. The uniformity of treatment avoids the maze of schemes characteristic of many PAYG systems in Europe, with various privileges accorded to more influential categories. A flexible contribution rate and a flexible retirement age--which are more easily accom- * Elsa Fornero is professor of economics at the University of Turin and director of CeRP (Center for Research on Pensions and Welfare Systems, Turin). 198 CONVERSION TO NDCS--ISSUES AND MODELS 199 modated in a DC scheme than in a DB one--are more in line with a life cycle perspective than a constant payroll tax and a mandatory age. The direct link between age of retirement and the level of benefits avoids the bias toward early retirement normally associated with DB formulae. On the other hand, although the definition of what is "fair" is always subjective, the definition I just referred to implies that it is fair to offer each worker the rate of return cor- responding to the rate of growth of aggregate earnings. It could be objected that this is too restrictive a definition, and that fairness should imply some kind of explicit redistribution toward the less fortunate members of society (in terms of working career, earnings, and hardship). The experience of many countries shows, however, that a DB scheme may redistribute very heavily but also very unjustly, with DB rules used to favor steep earnings careers as opposed to flat ones, workers in "rich" industries as opposed to workers in "poor" ones, the relatively young retirees as opposed to the older ones, the self-employed as opposed to the employees, and public as opposed to private employees. In all these comparisons, redistribution is very often not from rich to poor, but the other way round. It would therefore seem that one of the main advantages of a DC scheme is to eliminate all these kinds of misdirected (and often hidden) redistributive flows. Its superiority derives then as much from tight economic reasoning (financial equilibrium and a good incentive structure) as it does from its ability to withstand politicians' attempts to use it for their own interests due to the greater transparency of its rules. Moreover, the separation of insurance from redistribution need not be complete, as some kind of redistribution, if desired, can be easily superimposed on a DC scheme. A guaranteed floor, a rate of return decreasing with income, survivors' and disability bene- fits, and the like may all be adapted to a DC scheme with an appropriate correction to the basic rate of return. This somewhat blurs the general picture, but leads to much clearer and more just results in terms of fairness. Now to the question about risk. The coverage offered by a DC scheme depends, roughly speaking, on the rate of growth of the economy. The welfare of pensioners is therefore linked to the welfare of the rest of the community. Is this unfair? The alternative solution-- to isolate pensioners by offering them a predetermined rate of return--implies shifting the risk to the working generation. In the latter case, if growth is slow, workers suffer twice, both because their income is static and because they must pass a higher fraction of it to the pensioners. If, instead, income growth is rapid, pensioners perceive a greater gap vis-ŕ-vis the young. True, the demographic risk still rests on the workers, unless the state is ready to transfer (part of) it to taxpayers--which again shows that the goal of a complete separa- tion between insurance and redistribution is illusory. The separation is even more difficult to achieve in the transition phase, which brings us to the core of Palmer's chapter. This is because there are costs in the transition even to a notional fund; they are higher when the initial conditions involve a very large "implicit debt" and even higher when building a real fund is also an objective of the new design. The political process that brings about the change in the design is thus very difficult because it normally requires a much larger redistribution both between and within gener- ations than the one implicit in the previous steady state.2 Abstracting from political difficulties, Palmer examines the key decision points of the implementation process, presents possible solutions, and evaluates them according to the two principles mentioned above--acquired rights and contributions paid. In more detail, he discusses how to calculate the "initial capital" to be recognized when the system is applied to all workers (irrespective of their seniority in the past system). This entails a three-step process: reconstructing the individuals' earnings histories, assigning contribu- 200 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES tion rates, and crediting a rate of return. He also examines the speed of transition, with polit- ical options ranging from the slowest phasing in (new entrants only) to an immediate application of the new system, passing through a gradual solution, as when a pro rata mechanism is applied, with the new rules valid only for future seniority (the Italian case). He also considers the treatment of supplementary benefits (survivor and disability benefits), as well as issues such as pension indexation mechanisms and ways to increase the effec- tive retirement age, even if they seem more pertinent to the question of reform design than of its implementation. The chapter examines three models for the reconstruction of earnings under the hypothesis--particularly relevant for Eastern European countries--that individual records are either not available or missing: the nationwide average wage; the age- and gender-related average wage; and a rough estimate of the individuals' earnings profile, given the earnings of a few reference years at the time of the implementation. All three models involve some degree of redistribution. Palmer notes that "policy makers will have to defend the arguments favoring their chosen model. The political decision should be based on analysis of available data, supported with a view of what is "fair"--given the political-economic environment." Of course, by choosing solutions that are more favorable to current retirees and workers already in the system, the implicit debt that was involved in the past system is simply rolled over, and may even be increased, which means a more prolonged transition. Palmer is very accurate in identifying the various "tax overhangs" that can derive from the (necessarily imperfect) practical implementation of the NDC principle, also showing how to "factor" them in the entitlements initially recognized or in the return to be awarded, in order to maintain it as nearly as possible to the "true" internal rate of return. This can diverge for various reasons from what is formally awarded, as when a demographic cycle temporarily changes the underlying trend, making it more difficult to identify the contribution by population growth to the system's return, or when, for political reasons, it is decided that the first pension is calculated by assuming a positive fixed rate of return, which nicely increases the first pension, but may not correspond to the internal rate, requiring subsequent corrections. In all cases, Palmer's preferred solu- tions are those that appear to be more coherent with the steady state, perhaps not sur- prisingly once one accepts his premises, discussed above, for the very introduction of the NDC system. Given this, I agree with most of his propositions, which are quite sensible in the adopted framework. However, I have two main reservations. The first is about the possi- bility that the two principles he refers to (acquired rights and contributions paid) are not enough for guiding the practical implementation of the reform. The second is about the treatment of families, as opposed to individuals, in pension systems. I will now consider these two caveats in turn. Logical consistency would require that, if the new system to be implemented is judged better than the previous one, the conversion should be undertaken as quickly as possible and, in any case, by choosing at any point of the decision tree the option that appears to be more in line with the adopted NDC principle. In the extreme, this would mean an imme- diate application of the DC formula to all cohorts, and a strict adherence to Palmer's sec- ond principle, the "contributions paid" criterion. In practice, there are at least two kinds of limitations to this radical scenario: group pressures to maintain "acquired rights" and a shared view of "social justice," with particular reference toward the worst-off individuals of society. However, if one of the main advantages of the NDC system rests in the abolition CONVERSION TO NDCS--ISSUES AND MODELS 201 of previous privileges, the "acquired rights" criterion will very likely freeze them for the current generation and impose a highly unbalanced transition, with costs to be borne mainly by the poor, as well as the young. On the other hand, the "contributions paid" cri- terion will distribute more evenly across generations but will leave the poor equally or even more exposed. In this context, just as the NDC system is normally integrated by a minimum pension provision (possibly financed through general taxation), an external criterion could be required to guide the transition, such as the "Rawlsian principle" of sustainable maxi- mum.3 To give this discussion a more concrete content, consider the case of Eastern Euro- pean countries, many of which have already implemented an NDC system, or are about to do so. Given that typically these countries do not have the earning histories, in calculating the "initial capitals" to be recognized for past seniority, they could certainly refer to Palmer's criteria, with the risk, from a social point of view, of postponing the previous structure of privileges or excessively penalizing the poor according to whether the first (acquired rights) or the second (contributions paid) criterion is chosen. In these historical circumstances, the fact that the older generations had little or no choice in the previous political regime deserves perhaps more consideration than the two principles advocated by Palmer could allow. The second caveat concerns the treatment of families, as opposed to individuals. Under the heading of survivors' benefits, Palmer seems to suggest the opportunity of using dif- ferential conversion factors for women and men according to their specific longevity, thus eliminating the implicit subsidy to women that derives by the use of unisex mortality tables. He also argues, on the basis of the "gender equality" principle, against maintaining survivor benefits within the NDC scheme. In an ideal world, where gender equality in the job market and in the family were effectively realized, he would be right in both claims. There would be no point in objecting to the use of gender-specific mortality rates, just as nobody objects to insurance companies developing and selling products for the individual and pricing them according to their cost, and thus differentiating private pensions with respect to the different mortality patterns that characterize genders, cohorts, and profes- sions. Along the same reasoning, survivors' benefits for spouses, even when properly priced, would probably be obsolete. The reality is different. On the one hand, the fact that in many countries survivors' ben- efits have greatly helped to prevent poverty among widows is an undoubted achievement of the traditional social security systems. Moreover, families are still typically made up of a working man, a woman whose working life outside the family is shorter and more dis- continuous, and one or more children. The question therefore is whether the family instead of the individual should be the proper unit of reference for a public pension sys- tem, even of the NDC type.4 There are surely some insurance opportunities within the family that would be lost in a system centered on the individual. Indeed, typically the pub- lic system offers survivors' coverage without either charging higher contributions or reducing the benefits paid to the male as long as he survives. Disregarding the less typical situations, in this case social security, as compared with an actuarially fair system, taxes both single men and single women in favor of families, in the sense that these supplemen- tary benefits are not paid for on an individual basis. Is this fair? The answer is probably yes, if bearing and rearing children is considered a socially relevant duty. A favorable treatment by social security is an offset (and only a very partial one) to the financial and non-financial costs involved. Aspects of this risk sharing within the family should be maintained, in my view, in the pension system, whether or not it is NDC. 202 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Notes 1. "Notional defined contribution" and "non-financial defined contribution" should be understood to have the same definition. 2. See Diamond (2001). 3. See Schokkaert and Van Parijs (2003). 4. See Diamond (2004). References Diamond, P. 2001. "Towards an Optimal Social Security Design." CeRP Working Paper 2, Center for Research on Pensions and Welfare Policies, Turin. http://cerp.unito.it ------. 2004. "Social Security." American Economic Review 94 (1): 1­24. Schokkaert, E., and P. Van Parijs. 2003. "Debate on Social Justice and Pension Reform." Journal of European Social Policy 13 (3, August): 246­63. Chapter 10 NDCs: A Comparison of the French and German Point Systems Florence Legros* IN RECENT YEARS, MANY COUNTRIES HAVE TAKEN STEPS to increase the age at which individ- uals receive benefits from their retirement programs. This is supposed to shift the incen- tives in the right direction: that is, to keep workers in the workforce longer than they have been in recent decades. Notional defined contribution (NDC) pension plans have been considered one of the best means to increase the effective retirement age because they take into account the individuals' whole career, from the first day to the last. From this point of view, they are closer to actuarial fairness than traditional pay-as-you-go (PAYG) defined benefits (DB) pension plans.1 This is probably one reason why NDC has received increas- ing attention during the last years, even if the ability to increase the labor force participa- tion rates for older workers has probably been oversold. Other reasons why NDC has received attention recently are qualities such as trans- parency, flexibility, adequacy of provided pensions, and robustness: that is, the ability to resist various economic and demographic shocks. This study focuses on the feature of robustness and assesses it by analyzing the Swedish system (as a benchmark) and comparing it with the German and French schemes that often have been claimed to be NDC variants.2 To this end, the chapter is organized as follows. The next section briefly describes the French and the German pension schemes. It then considers key points of NDC and actuar- ial fairness: how pure NDC guarantees actuarial fairness both at the margin and on aver- age, and how this is reflected in the role of the life expectancy in the parameters of NDC pension schemes. In the third section, the study focuses on the conceptual equivalence among the French, German, and Swedish systems and highlights the differences in their design and in their evolution policies. This leads to a discussion of how the reactions of these systems differ when there is an external shock (demographic, economic, or other). The chapter concludes with a discussion about the possibility (and opportunity) of mov- ing from a point scheme toward an NDC scheme (whether pure, with a minimum pension * Florence Legros is professor of economics at the University Dauphine in Paris. I thank Robert Holzmann for helpful discussions and valuable comments. An acknowledgment is also due to Jean-Louis Guérin for the fruitful discussions we had during our co-authorship of a paper dealing with actuarial fairness (Guérin and Legros 2003). Any remaining errors are fully mine. 203 204 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES scheme, or some combination). The conclusion considers the redistributive impacts of such a transition between generations and inside a generation. A Quick Description of French and German Pension Schemes French and German pension schemes have a component that uses points: that is, part of these schemes provides a pension that depends on the individuals' whole career. France This section focuses on the main French pension scheme: the compulsory pension scheme that provides pensions to employees in private enterprises, including white-collar work- ers, about 61 percent of the working population. There are plenty of other schemes. Some run like the basic French pension scheme, the régime général discussed below. Some run like the private sector pension scheme. Previous schemes include both a scheme that is close to the régime général and a scheme with points. For the private sector, French pensions rely on two pillars. The basic general scheme (régime général, or CNAV) offers benefits corresponding to the share of the wages below the social security ceiling (TRanche A, or TRA) equal to 2,516 per month, for the year 2005). The complementary schemes include AGIRC (the Association Générale des Institutions de Retraite des Cadres) for executives, for the fraction of their wages over TRA; and ARRCO Table 10.1. Schemes and Contributions in the Main French Pension Scheme Considered wage share Below TRA Up to three times TRA Up to eight times TRA Population Nonexecutives CNAV Workers: 6.55%a Firms: 8.2% a Total: 14.75% a CNAV: + Firms: 1.6% ARRCO ARRCO Workers: 2.4% Workers: 6.4% Firms: 3.6% Firms: 9.6% Calling rate: 125%b Calling rate: 125%b Effective rate: 7.5% a Effective rate: 20% a Executives CNAV Id. nonexecutives ARRCO AGIRC AGIRC Id. nonexecutives Workers: 6% Workers: 6% Firms: 10% Firms: 10% Calling rate: 125%b Calling rate: 125%b Effective rate: 20% a Effective rate: 20% Source: Author's compilations. a. Contribution rates are applicable to the gross wage in the defined interval (below the TRA). b. The calling rate is a multiplicative factor that is applied to the contractual contribution rate to increase the schemes' resources (below the TRA, workers are supposed to contribute up to 2.4 percent of their gross wage-- contractual rate--but pay in reality 125 percent of 2.4 percent--that is, 3 percent--while they accumulate their rights proportionally to the contractual rate). NDCS: A COMPARISON OF THE FRENCH AND GERMAN POINT SYSTEMS 205 (the Association des Régimes de Retraite Complémentaires), for the other workers, based on their total wage, and for executives, for wages below TRA. These plans are summarized in table 10.1. At the beginning of 2003, the CNAV pension, Pcnav, was computed as follows Pcnav = w * * Min[1, (T/37.5)], where w is the yearly gross reference wage average of the 25 best wages for the generation born after 1948.3 The rate of pension, , is = 0.50 * [1 ­ 0.10 * Min[40 ­ T; 65 ­ A)], where T is the number of contributing years and A is the retirement age. The last reform (approved on August 21, 2003) will progressively affect the formula as follows. The full rate contributing period (40 years, currently) will increase according to the gain in life expectancy (one quarter a year between 2009 and 2012, to increase to 42 years in 2012). The discount rate, which is currently 0.10, will decrease step by step to reach 0.05 in 2013. The T/37.5 term's denominator (currently 37.5, corresponding to the full rate contributing period before the previous reform in 1993) will increase by two quar- ters a year to reach T/40 in 2008 ; at this period it will be equal to the full rate contributing period (40 years). After 2008, it will increase by one quarter a year and reach 42 years in 2012. If the corresponding full rate is denoted by T', the contributing period in the above formula becomes Pcnav = w * * Min[1, (T/T')], and = 0.50 * [1 ­ 0.05 * Min[T' ­ T; 65 ­ A)]. Note the existence of a minimum pension provided by CNAV. It reaches 6,935 per year for a single person and 12,440 for a cohabiting or married couple. The complementary pension schemes are fully contributory schemes. The first pensions after retirement, Pcomp, are computed according to a system of points, which are the pen- sion basic units. These are "sold" when people retire at the age of A (after a career of length T) at the price that prevails during the year C + A (C being the cohort generation, that is, the birth year). Thus the value of the point, VP, a virtual price at which individuals trans- form their virtual capital--corresponding to the sum of the points that are accumulated during the working period--into a yearly pension, is expressed as VPC+A.4 Pcomp,C+A= points *VPC +A T During any year t, during the working period, the flow of accrued points can be com- puted as follows for nonexecutives pointst = pointsARRCO,t points = [ARRCO ,TRA* w TRA + ARRCO ,TRB* TRA < w TRB]/PPARRCO, for executives points = [ARRCO ,TRA* w TRA]/PPARRCO + [AGIRC * TRA < w TRC]/PPAGIRC, 206 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES where ARRCO ,TRA, ARRCO ,TRB , and AGIRC are the "contractual contribution rates" or "facial contribution rates" of the regimes (respectively, 6 and 16 percent in 2005); TRB and TRC are equal to three and eight times TRA (7,548 and 19,456 per month, respectively, in 2005). PPARRCO and PPAGIRC are the "purchasing" price of a point. Note that during the retirement period, while basic pensions (CNAV) are revalued at the inflation rate, complementary pensions are revalued at the point's value, VPt. One of the ways to handle the resources of the schemes is a coefficient, the "calling rate," that applied to the facial contribution rate, increases the schemes' resources. For example, the flow of income of ARRCO is given by ResourceARRCO = [ARRCO ,TRA* coeff] * w TRA + [ARRCO * coeff] * TRA < w TRB. With the current coeff equal to 125 percent, that leads to the effective contribution rates of 7.5 and 20 percent. This coefficient (the calling rate) is an important means to manage the scheme, like the price and the value of the point. This rate was lower than one during the first decades of existence of the system and now increases more or less regularly. Simplifying matters,5 the French complementary pensions by points are computed according to the following formula for an individual i who retires at age A after a contri- bution period lasting from t0 to A ­ 1: A PC+A = i t-1 *wt-1 i *VPC+A . (10.1) t=t0+1 PPt-1 This is the pension that the pensioner gets when aged of A + C, while the resources of the scheme in t are Rt = i t*wt *coefft . i Note that although pensions are indexed on prices after A, PP and VP are changed reg- ularly by the boards of the schemes according to the forecasts about the pension schemes. Also note that AGIRC and ARRCO are private associations linking together sectoral pen- sion schemes. This is because when the schemes were created, each economic sector had its own pension scheme. French complementary pension schemes are fully contributive and proportional schemes (according to their Bismarckian origin), compared with the basic scheme that only partially relies on the career and that includes a flat part. They are managed by trade unionists with the help of retirement specialists, including actuaries and financial investors. They were created in 1945, and were designed by insurers. They became compulsory in 1972. The accounts are held by individuals. The management of the executives' pension scheme, AGIRC, serves as an example of the piloting of such schemes. Since the first oil shock in 1974, a decrease in the pensions' yield has been scheduled. The analysis of AGIRC data shows the prevailing, discrete, and progressive mechanism based on the price of the point (PP). This is because as soon as PP increases more quickly than the contributors' average wage, the number of points earned decreases. In 1970, a share contribution of 13.9 percent of the taxable wage provided 1,000 points, while the same contribution rate would have provided only 850 points at the end of the 1990s.6 The trade-off between the price increase, the value decrease, and the calling rate management is a result of the burden-sharing between retirees and contributors. Until 1976, the AGIRC point value was the result of an automatic adjustment by compar- ing for the next 10 years the expected contributions' flows to the expected pensions' flows NDCS: A COMPARISON OF THE FRENCH AND GERMAN POINT SYSTEMS 207 (and taking into account the buffer fund balance). 7 In 1976, the point value was forecast to drop sharply and the pensions' schemes' board chose to favor the retirees and to employ the calling rate. Germany The German pension scheme (described, for example, in Queisser 1996 and Verničre 2001) seems simpler since it relies on only a single point scheme that covers not only the employ- ees of the private enterprises but also certain self-employed and some other specific parts of the population, some 85 percent of the active population. In fact, as it includes an early retirement pension scheme, the qualifying conditions are quite complex. Note that this is not the case in France, where, by law, early retirement does not exist. "Old workers"--that is, workers just below the retirement age--are exempted from searching for a job that they would not find anyway and do draw pensions that are generally funded either by unemployment schemes or by disability insurance. This "tool" has been used greatly by firms to decrease their wage bill when necessary. Compared with the French pension schemes, the contributory ceiling in the German system is rather low. As noted, in France, workers must contribute until their wage reaches eight times the social security ceiling (20,128): that is, 8.5 times the average wage. In Ger- many, the contributory ceiling is 1.8 times the average wage. This has implications that will be discussed in the next section of this chapter. There is no minimum contributory wage in Germany, but very low pensions are scaled up if people have contributed for at least 35 years. In this case, the personal points can be multiplied by 1.5, up to a maximum of 75 percent of the value of contribution for average earnings of all insured persons. Since the 2001 reform8 there has been a minimum pension, which helps very low income pensioners. On the other hand, there are noncontributory additional rights for part-time working women and for children's care. Like the French pension point schemes, each working period provides pension rights according to the wage earned by the contributor. The gross pension is given by the follow- ing formula A i PC+A = i ww t-1*i *VPC+A . (10.2) t=t0 t-1 The notations are the same as for France in equation 10.4:wt is the gross average wage in t, is the entry factor corresponding grosso modo to the French rate of pension, and wt is the i part of the wage under the ceiling. = 1 ­ [0.003(780 ­ A(12)], where A(12) means age A in months. This means that there is a discount equal to 0.003 per month in case the pension is drawn before age 65. Most of the adjustments rely on the value of the point VP. The 2001 reform deeply changed the parameters in the indexation formula, VPt =VPt * wt * wt-1 x- t -µt -1 -1 -1 . (10.3) -2 x- t -µt -2 -2 The indexation coefficient, x, was supposed to change. Between 2001 and 2010, x = 1, meaning a full indexation on net wages. After 2010, x = 0.9, meaning partial indexing. 208 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES The variable µ is the contribution rate to the voluntary additional private pension scheme sponsored by the fiscal chapter of the reform. What is called here net wage is in fact the wage net of all contributions dedicated to pension schemes (but containing the per- sonal income tax). In the forecasting exercises, the value of µ was supposed to increase between 2001 and 2008, from 0.5 to 4 percent, and remain stable after 2008. It corresponds to the value that binds the fiscal constraint. This reform has quickly been judged as inefficient.9 The so-called Rürup report led in October 2003 to another indexing formula, which will hold until 2010 VPt =VPt * wt * wt -1 x- t -µt -1 -1 -1 *SFt . (10.4) -2 x- t -µt -2 -2 SF is the sustainability factor, defined as st-2 *Nt-3 SFt = 1- *00.25 +1 st-3 *Nt-4Nt-2 Nt-3 where st ­2* Nt /Nt ­3 ­2is the sociodemographic dependency ratio: that is, the ratio of the retirees (in t ­ 2, those who were active in t ­ 3 and living in t ­ 2) to those active in t ­ 2. That means that--in the long run, with a constant demographic structure--the pen- sions will increase like the gross wages. In the transition period (the demographic shock), the pensions will increase like net wages minus the 25 percent of the demographic drift (that is, 25 percent of the demographic drift will be "paid" by retirees). As in the French scheme, life expectancy does not appear explicitly in the pensions' for- mula. It is implicitly introduced by two means: by the scheme equilibrium that links the contribution rate to the dependency ratio; and by means of the dependency ratio, which strongly depends on the retirees life expectancy. If it is assumed that the fiscal resources of the pension scheme (ecotax, the German energy tax program dedicated to limit the industrial consumption of polluting energy, 32 percent of the scheme resources) remain constant, and/or that, in the long run, this tax is a share of the wage bill, which is a constant share of the national income, resources can be derived as Rt = i t*wt . i NDC and Actuarial Fairness In both countries, many individuals used to quit their job five years before the legal pen- sionable age. To provide incentives to delay this retirement, actuarial fairness at the mar- gin is a concept that must be explored. The following section shows that even if one ignores the value of leisure, actuarial fairness at the margin is very difficult to implement and can lead myopic individuals to poverty. It is why actuarial fairness on average is a preferable concept. It provides every generation a pension strictly equivalent to its contri- butions; it spontaneously leads to NDC. Unfortunately, the second part of this section NDCS: A COMPARISON OF THE FRENCH AND GERMAN POINT SYSTEMS 209 shows that this collective concept can be unfair overall for individuals with low life expectancies. Actuarial Fairness at the Margin: An Individual Concept Actuarial fairness at the margin works along those lines: early retirement is swapped against a proportional reduction of the pensions' benefits during the whole retirement period. If the agent decides to postpone retirement (after a legal minimum retirement age, for example) and does not retire at date t ­ 1 and age A, but waits until date t and age A + 1, he or she will pay contributions for the longer time and will benefit from a higher pension. On the other hand, if the worker retires early, benefits will be reduced by a fraction d dur- ing the entire retirement (this reduction is called "decrual"). By retiring early, the agent will save contributions and benefit from a longer period of leisure. That actuarially fair decrual in pension is measured in such a way that the choice of the agent does not threaten the budgetary equilibrium of the system. If the decrual rate is set at a higher value, early retirement will improve the financial situation of the system; if it is less, the opposite applies. Whatever the case, the decrual will have an impact on agents' behavior. In the first case, if the decrual is high, it is costly for an agent to leave early and few workers are likely to make that choice. In the second, if the decrual is low, early retire- ment benefits the pensioners and the attractiveness of that formula will help degrade the budgetary balance of the system. It should be noted that the decrual/accrual rate depends on individual career's profile: if two agents are considered, each having earned the same rights to pension benefits, but with different wages for the last year of activity, the value of the decrual must be higher for the one having the largest wage. Intuitively, that agent must be rewarded by a higher pre- mium if he or she delays the retirement decision. With the same logic, it can be inferred that if pensions are indexed on gross wages, the accrual/decrual has to be higher than if they are indexed on prices. (For a formal demon- stration, see the annex.) From a very simple model (see the annex), where is the individual discount rate, r, the interest rate, and A t­1, the income equivalent of the leisure associated with leaving activity one year early,10 the following results can be derived: · If = r and wAt = A ­1 t­1, then the worker is indifferent between ceasing work early or delaying retirement. · If = r but wAt < A ­1 t­1, then the worker draws a high satisfaction from leisure, and he or she will opt for an early retirement, with the actuarially fair decrual in benefits. · If wAt A ­1< t­1and < r, then because his or her discount rate is high, the individual prefers to retire early, as he or she does not give much weight to the loss in income induced by an early retirement. · If wAt A ­1 > t­1and < r, then the individual opts for a postponed retirement, but if wAt > A ­1 t­1and > r, then the individual's preference for the present is so high that the financial gain induced by actuarial fairness at the margin does not mean much to him or her, and the agent retires early. So the efficiency of the system's parameterization strongly depends on individual pref- erences: most importantly on the taste for leisure and on the preference for the present. On a general note, it can be assumed that there is a high probability that high taste for leisure and preference for the present are true for aging workers. Many reasons can be given for that point: residual life expectancy is lower; the number of healthy years left is 210 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES reduced; the higher uncertainty prevailing about life expectancy increases preference for the present. The value of the actuarially fair decrual rate changes with the indexing rule. (This holds as well for an indexing on the net wages.) It rises with the interest rate and the contribu- tion rate, and decreases in line with life expectancy, the inflation rate, technical progress, and the replacement rate. It is worth noting that with realistic data and even without giv- ing any value to the leisure, the decrual will have high values: 6.6 percent with indexed pensions on prices and 7.1 percent in the case of indexed pensions on gross wages. As can be seen, for a standard Bismarckian, PAYG pension system, actuarial fairness at the margin implies a high value of the decrual, far from the values given by a linear calculus. (A replacement rate of 60 percent after 40 years of work gives an accrual rate of 1.5 percent.) Thus it is necessary to have a bend in the accrual curve of benefits. In other words, the con- cept of actuarial fairness at the margin can apply only to a part of the age distribution. In concrete terms, because of the indexing rule on the net wages (even in case of partial adjustment), Germany should have a more steeply sloping curve than France. This implies also that because productivity gains change every year, the benefits accrual curve has to change as well. Now consider the case of a system displaying both Beveridgian and Bismarckian ele- ments. Every pensioner earns a pension compounded of two parts. One is proportional to his career (as previously) and one is a lump sum fixed independently of his career profile, not subject to the decrual. Such a system cannot be actuarially fair to every member simul- taneously, as the benefits decrual rate is now a function of the wage and so depends on the wage profile of the individual. Because such minima exist in both the German and the French systems, actuarial fairness for the whole system is only a desideratum and cannot be reached. The existence of a contributory ceiling is another point that moves the schemes away from actuarial fairness at the margin. This is particularly the case in Germany, assuming that a worker has a nonnegligible probability of attaining this wage level by the end of his working life and that postponing retirement has a very weak return. That means that when this is the case--that is, when the contributory ceiling is rather weak--the accrual is nearly zero. Another consideration is the cost of actuarial fairness at the margin. If leisure is valued highly, the value of the decrual (or the accrual) should be higher than the values given pre- viously. That means that there is an infinite number of decruals/accruals, which provides actuarial fairness. A financial level with a null value of leisure insures the financial equilib- rium of the scheme. An individual level, with a bigger value of leisure, insures personal optimality. This is because, first, it is generally assumed that there is a decreasing relation- ship between life expectancy and value of leisure;11 second, there are non-Bismarckian ele- ments in the pension scheme that imply a decreasing relationship between the individual wage and decrual. For these reasons, the optimal decrual for low incomes will be higher than the optimal decrual for high incomes. Thus if the government's wish is to retain high and low incomes on the labor market (which pools the longevity risks, as can be seen in table 10.2), the decrual should be quite high but very costly. Actuarial fairness at the margin is costly and rather difficult to implement as soon as the life expectancy has to be taken into account, as correlated with the value of leisure. To combine individual freedom, incentives, and financial equilibrium for all the ages, actuarial fairness on average seems to be a better concept than actuarial fairness at the margin. In this case, the acquisition of rights is a continuous unbroken straight line, which is the condition for uniqueness of the PAYG yield. This relationship is supposed to lead to NDCS: A COMPARISON OF THE FRENCH AND GERMAN POINT SYSTEMS 211 Table 10.2. Life Expectancy at Age 60, by Socioprofessional Categories, in Years (the case of France) Category Men Women Executives, self-employed 22.5 26.0 Intermediate profession (technicians, etc.) 19.5 25.0 Artisan, shopkeepers, firm managers 19.5 25.0 Employees 19.0 24.0 Workers 17.0 23.0 Source: Mesrine (1999). pure actuarial fairness. It has certain advantages but its labor incentive character is less distinct than in the case of actuarial fairness at the margin. NDC: Actuarial Fairness on Average In an actuarially fair system on average, the discounted sum of contributions for every individual must be equal to the discounted sum of benefits. In addition, the discount rate is the same for every member of a cohort. As can be seen, the presence of Beveridgian ele- ments in a pension system prevents the existence of such characteristics. In the same way, the existence of specific contributory advantages or of a ceiling on benefits is outside that logic. Those systems are left aside in the following analysis. As soon as the pension sys- tem allows for full freedom in retirement age, the respect of actuarial fairness on average necessarily implies the respect of actuarial fairness at the margin. In the steady state, with invariant demographic and economic structures, if all workers earn back what they con- tributed, the financial balance of the system is ensured. But the respect of actuarial fair- ness on average does not imply budgetary balance in case of a demographic or economic shock. Two kinds of pension systems can aim to attain that notion of actuarial fairness on aver- age: the fully funded schemes and the notional accounts or the system by points. The lat- ter differ from pure fully funded schemes by their financing (the vast majority of such systems are PAYG systems), but also by the virtual nature of the acquired benefits. Every contributor has a personal account that records all of his or her pension contributions over his or her active lifetime. The hoarding of those contributions constitutes virtual capital, increased according to a specific rule of indexing. On the date of retirement, that virtual capital stock is converted into annuities according to a transformation coefficient that depends on the liquidation age and retirement life expectancy. In sharp contrast to the notion of actuarial fairness at the margin, fairness on average takes into account the whole career profile of the agent, without focusing on the last years of activity. Of course that concept must be considered at the collective level, given the uncertainty prevailing on individual life expectancy. The most important defining parameter of such a system must be the rate of discount (or rate of return) used to compare past paid contribu- tions and future earned benefits. With the same notation as above, and taking into account the average life expectancy of the cohort in question, the internal rate of return of the system can so be defined as 1+ t =Pit+1 *st+1, twt 212 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES where st +1is the portion of period t + 1 that is to be lived by the average retired individual s i t+1 st = i +1 . Nt The value of that rate, as well as the contribution rate, are assumed to be the same for all the agents at the given date. Of course, for an individual, one is able to compute his or her own personal internal return rate 1+ it = Pit+1 *sit+1 . twit As soon as the system is of a PAYG nature, the earnings at date t of the system can be writ- ten as Rt = w i (10.5) t t i and the benefits paid amount to Et = P i i (10.6) t * st , i with i denoting the individuals (active or pensioners).12 The budgetary balance of such a system depends on the indexing rule, which sets the internal rate of return and the evolu- tion of pension benefits. That rule is generally given by the evolution of an economic para- meter. Consider the conditions in which an ex ante equilibrium can be reached by the system, denoting gt the indexing rate of the pensions between t and t + 1 Pt = Pt (1+ gt); i i +1 Et = i i t+1 +1 st+1 ( 1+ gt ) p st t t i s = sst (1+ gt)Et ; i i w t+1 Rt = t+1 wt+1 i t +1 i +1 t t i wt = i Rt i wt t w t i as in a PAYG pension scheme, Et = Rt, i Et = Rt Et = st w t +1 +1 +1 +1 st (1+ gt) t t i (10.7) i Rt+1 . +1 w t+1 i As soon as the pension scheme contribution rate does not vary (which is generally the purpose of the switching from a "usual" PAYG toward an NDC PAYG), the indexing rule will try to provide an automatic stabilizing device to the pension scheme. At first glance, there is no NDC or point pension scheme providing this kind of automatic stabilizer, with NDCS: A COMPARISON OF THE FRENCH AND GERMAN POINT SYSTEMS 213 the single exception of Sweden. Italy uses the total wage bill as an indexing device, France relies on inflation, and Germany on net wages. There is a huge difference, on the one hand, between a simple comparison of the ability of different factors to achieve the stability of the scheme and, on the other hand, the expec- tation about the ability of the system to stabilize ex ante. In other words, even if the stabi- lizer is adequate, it can be inefficient or even incapable of playing its role. NDC and Resistance to Various Shocks As a consequence of the above developments, actuarial fairness on average, which is very close to pure NDC, can be a source of unfairness. Moreover, it has no automatic stabilizing device. In other words, if the stabilization of this type of scheme is discretionary (for exam- ple, decreasing pension benefits to avoid imbalances) the condition of actuarial fairness is broken. This implies a perfect forecasting--that is, no discretionary adjustments once the unbalance cannot be avoided but a perfect planning of future variables. As shown, there is a need for a "zero pillar" (Holzmann 2006) to provide some redistri- bution toward the poorest. A pure NDC--actuarially fair on average--is then far from being ideal. This chapter considers two actual cases, the French and German systems, to which, as a benchmark, is added the Swedish system, often presented as an ideal NDC. Consider first the Swedish system: a PAYG scheme with virtual funding. The basis of such a scheme is "virtual" capital, K, which arises through the accumula- tion of pension rights through contributions. This capital is re-evaluated each period at the rate r, while A is the age when retiring, and C is the birth date, which identifies the cohort C. KCt can be written as follows A-1 Kt = Kt (1+ rt ) C C C -1 -1 or KA = w (1+ r)t . t t=t0 Note that to be closer to the effective scheme, the contributing period can be divided into subperiods t. The virtual capital, KCt, comes from the past flows of contributions dur- ing the entire contributing period. At age A, an individual can draw a pension, which is supposed to be re-evaluated during the pension period (whose length is the cohort's life expectancy at retirement). This--initial--actualized pension, PCA, is the capital divided by sCA, the life expectancy of cohort C when age A, which appears in this type of scheme as a central management parameter. This yields C (1+ g)D-A* PA = C KA , C C after retirement. sCA and Pt = (1+ g)Pt +1 In the Swedish pension scheme, r = 1.6 percent. This is a parameter of a "yield in advance" given to the contributors, representing a benchmark of economic growth. Dur- ing the pension period, the index g is close to the nominal growth rate of the per capita GDP, from which this yield in advance is derived. Without any loss, is it later assumed that r = 0 and g equals the nominal growth rate of the per capita GDP. With the same simple mathematical manipulation as above, it is rather easy to lead to iC Et = Rt Et = st w t +1 +1 +1 +1 st (1+ gt) t t i +1w iC Rt+1 , +1 t+1 i 214 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES where it is not only the period that is in equilibrium but the cohort as well. Thus the equi- librium is both in time and in space. The C + 1 factor means that the cohorts have aged by one unit of time, say, one year. The two-period model provides helpful information--and also an optical illusion: in fact, a perfect NDC scheme would require a yearly adjustment to the evolution in the life expectancy, including an adjustment for all pensions that have already been drawn, and not only for pensions of new pensioners. In addition, the fact that the index g is the nominal growth rate of the per capita GDP implies that there is an "automatic adjustment" only when the share of the wages in the GDP is constant and if the active population remains constant--which is precisely the problem in case of fall in the fertility rate. In the former case, a high GDP growth with a slack wage growth would be a source of financial unbalances, since the PAYG resources are wages and social income. The same type of calculation in the French case provides different insights. Without any loss in the argument, let us denote the contribution rate as (instead of * coeff, which means to replace the contractual rate multiplied by the calling rate by a single rate). In this case, equations 10.5 and 10.6 still hold. In addition to equation 10.1 and taking into account the indexing rule to the value of the point, the previous equation becomes A Et = t-1 *wt-1 j *VPt * si t i t=t0+1 PPt-1 and then t i t w i t Rt+1 = Et+1 Et+1 =st+1 VPt+1 * PPt w t * i t * i * Rt . st VPt i-1 i+1 +1 t t-1 w t-1 t+1w t+1 PPt-1 i t i t VPt+1 A PPt represents the adjustment parameters of the scheme. VPt t=t0 t-1 PPt-1 Some remarks are in order: · First, there are three parameters: the contribution rate, the price of the point, and the value of the point. The transformation of one wage unit into n units of pensions depends on the relative evolution of the three parameters. · The parameters can be changed every year, but the burden will not fall on the same persons. Changing the contribution rate, the price of the point, or both will affect both the contributors and the pensioners (such as a change in the contribution rate in a "normal" PAYG pension scheme); however, the "old" contributors will be affected during a shorter period. Changing the value of the point will affect both the young retirees (those who just retired during the last years) and the contributors, whose pension will be changed by this decrease or increase (this can be compared with a change in the replacement rate). · These changes are deterministic. There is no explicit rule, but they rely on the long- run forecasts done by actuaries. · The fact that the changes in these parameters can be annual and that they are more or less deterministic means that it is rather easy to adjust them, especially if some fore- NDCS: A COMPARISON OF THE FRENCH AND GERMAN POINT SYSTEMS 215 casting errors were made. The bigger the errors, the greater the changes and the more difficult the adjustments are to implement. In the past, the adjustments have been made in the correct time intervals. As a result, the French complementary pension schemes AGIRC and ARRCO enlarged by 40 billion in provisions. · The above result shows the need for yearly adjustments. The first term represents the change in life expectancy between two cohorts and may be rather weak. The terms w i t i t w i-1 t-1 i t and t w i t i t+1w i+1 t+1 , i respectively, represent the change in the contribution base between the two cohorts and the opposite of the change in the pension rights between the two cohorts. In the German pension scheme, two equations remain A i PC+A = i ww t-1*i *VPC+A , and t=t0 t-1 st *Nt -2 -3 VPt =VPt-1 * wt * wt -1x- t -1 -2 * 0.25 +1 , -2x- t -2 * 1- st *Nt -3 -4Nt Nt -3 without any change in the reasoning (provided we take into account the fact that the con- tributions to the additional pension fund are exogenous, that the sensitivity of a fully funded scheme [FFS]) to the demographic shocks is the same as the sensitivity of a PAYG to the same shock, or both. This leads to consider instead of ( + µ). That implies the following st-1 *Nt-2 Pi = Pt *i wt *x- t t+1 wt-1 x- t-1*1- st *Nt -2 -3Nt Nt-1 -2 *0.25 +1 . (1+gt ) In other words, this yields the same NDC scheme as the one described in the second part of this chapter. It leads to a corroboration i Et = Rt Et = st w t +1 +1 +1 +1 st (1+ gt) tt i (10.8) i Rt+1 , +1w t+1 i 216 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES with st-1 (1+ gt) =wt-1* x- t-1 wt-2 x- t-2* 1- st (1+ nt-2) * -2(1+ nt ) -3 0.25 +1 . On the surface, this analysis leads to a notional interest rate, which is the growth rate of the net average wage. In fact, this formula deserves a rather long comment. First, gt depends on t, which depends on the PAYG equilibrium in t, which mainly comes from gt , yield- ­1 ingPt. Intuitively, there is a huge difference between the French and the German pension schemes. Any change in VP in the French pension scheme is a burden for contributors (as future pensioners--and not for those who have already retired whose pension has not been changed). By contrast, a change in VP in the German pension scheme is a burden shared between pensioners and contributors, since any change in t will imply a change in gt. In other words, if the increase in the point value VP between t and t + 1 is too great, the PAYG is unbalanced and is rebalanced by an increase in t +1(a burden for the contribu- tors), which decreases gt +1 (a burden for both contributors and retirees, that is, for current and future retirees). This mechanism will later be called the return spring mechanism. It cor- responds to the left-hand side of the equation. Second, the Rürup formula13 explicitly introduces the demographic ratio, which becomes quite easily a ratio depending on life expectancy, s, and labor force growth rate, n. How does that change if we compare the new formula (Rürup's) with the Riester formula? First, it explicitly introduces demography: until now, the burden was shared between con- tributors and retirees, as described above. Now, the retirees pay 25 percent of the burden. This will limit the increase in g, and thus in . A second point is linked with the way the retirement pensions schemes are funded in Germany. With the previous formula, it would have been possible to increase the fiscal part of the funding, leaving constant the contribu- tion rates with no change in the point growth rate. Now, because exogenous data hold, this full "fiscalization" of the deficit is not possible. The problem now is to explain two matters: · How does the return spring work? · How is the balance in PAYG achieved--specifically, what are the links to financial stability of the NDC scheme? How Does the Return Spring Work? Consider the left-hand side of the equation, yielding (1 + gt). Roughly speaking, the return spring mechanism is the following t-1 t 1+ gt VPt+1 Et+1 t+1 gt+1 . In its current formulation, it has two limits: Limit 1. Manipulating the previous equations, and taking into account the PAYG equilibrium (with the fiscal resources denoted T): NDCS: A COMPARISON OF THE FRENCH AND GERMAN POINT SYSTEMS 217 t *wt * Nt + Tt =Pt * st * Nt , wherePt is the average pension in t. This implies ­1 wt *x- Pt *st /(1+nt-1)+Tt 1+ gt = Nt . wt-1 *x- Pt-1 *st-1/(1+nt )+Tt-1 -2 Nt -1 This shows that the notional interest rate is the rate of increase in the average wage, a rate of increase that is adjusted in two ways: · First, it is diminished by a share of the average pension. This share is the ratio of the probability of surviving when retired to the increase in the labor force, n. In other words, the "reimbursed" share of the average pension increases when the life expectancy increases and it decreases when the labor force increases.14 Each time the average pension, or the life expectancy, or both increases (com- pared with the increase in the labor force), the return spring plays its part and the indexing is less favorable. · Second, it is increased by the per capita fiscal part of the pensions scheme resources. This plays the exact opposite part compared with the average pen- sion: the taxation allows a "neutral" increase in the indexing rate. Limit 2. Again, the main problem is hidden in the use of the two-period model. The life expectancy of the retired persons is given as st. In fact, it is a weighted average of all the life expectancies of all the cohorts of retirees s. In other words, this inertia sentences the scheme to deficit and thus to an increase in the contri- bution rate. This outcome helps partially answer the second question: How Is the Balance in PAYG Achieved: That Is, How Does It Bear on the Financial Stability of the NDC Pension Scheme? As noted, the notional interest rate depends partly on past life expectancies while ensuring that the financial equilibrium of the NDC would introduce the future life expectancies as well as the expected increase in the total amount of contributions (that is, the wage bill multiplied by the contribution rate). That would imply taking into account a change in the wage bill, which obviously does not appear in the indexing rate formula. With respect to this remark, assumptions about the automatic stability are in conflict with a change--that means a change either in the labor force or in the wages. Their realism is therefore ques- tionable. Conclusion: Nobody Is Perfect One can clearly see the potential contradiction between the aims of actuarial fairness (insurance logic) and redistributive goals, since life expectancy strongly differs between socioprofessional groups. Taking into account the average life expectancy of one popula- tion, the NDC obviously favors financial equilibrium over social equilibrium, as life expectancy is highly correlated with income. On the other hand, for a social reason, all these schemes have introduced either mini- mum pensions (as in Sweden and Germany) or are "only" complementary schemes. For 218 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES example, in France, there is a basic scheme offering the possibility of a minimum pension. In Sweden, there is double protection against poverty in old age for unemployed or socially subsidized people. In addition to a minimum pension provided by the scheme, the social income (unemployment, disability subsidies, or other sources) provides contri- butions to the pension schemes and then provides supplementary rights. In the end, these arrangements are not pure NDC schemes. Even though these schemes are supposed to offer a high degree of freedom for individ- uals to choose their own retirement age, they cannot escape fixing a minimum retirement age. With absolute freedom, the agents would not be very protected against the bias in their decision making arising from uncertainties or unrestrained optimism concerning life expectancy--situations all the more dangerous if pensions are no longer indexed to wages and/or if the pensions are low according to an early chosen retirement age. In addition, the application of NDC to a real economy under uncertainty is problematic. That uncertainty is a major problem in the context of pensions is well known. Various reports are full of illustrations of such uncertainty. In the German case, for example, fore- casted contribution rates through 2004 can vary by 15 percent depending on the assump- tions about active population,15 birth rates,16 and life expectancies.17 Of course, the schemes can be periodically regulated according to changes in the fore- casts. In the French case, this is done yearly. This requires two things: the first are reliable and frequent forecasts (which are problematic, according to Lassila and Valkonen 2002). The second is total independence of the boards from retirees and even wage earners' lobbies. The German return spring, strengthened by the Rürup sustainability factor, is another mechanism in which the pension yield is lowered by the current contribution rate, which is a way to shift the burden on both the active and the retired population. In both cases, the periodic changes in the parameters of the schemes obviously break the actuarial fairness. A precautionary strategy would be to use the less favorable assumptions in order to adapt the parameters of the schemes. With such a strategy, it is probable that the scheme will get into surplus within some years. In this case, these surpluses would be helpful by becoming a buffer fund, which has been demonstrated to be a useful tool for bringing about intergenerational equity. To sum up, a pure NDC plan would be actuarially fair on average. An ideal NDC plan would provide redistribution, incentives to work longer, transparency, and automatic sta- bilizing. The chapter pointed out the lack of distibutive properties of a pure NDC, the lack of actuarial fairness of the considered schemes, the need for a zero pillar and for a precau- tionary buffer fund, and the fact that the economic and demographic uncertainties lead to break the actuarial fairness--that is, any revision in the parameters induce a different indi- vidual balance. Thus an ideal NDC cannot be a pure NDC. Then why promote NDC? If pensions schemes are made to promote individual respon- sibility, the preferred schemes are FFS. If, in addition, the purpose is to promote social responsibility, a favorite choice is NDC schemes. NDCS: A COMPARISON OF THE FRENCH AND GERMAN POINT SYSTEMS 219 Annex A: Actuarial Fairness at the Margin In the following simple model, the contribution rate is . If an agent who would be allowed to retire decides to postpone his retirement decision by one year, he renounces retiring at date t ­ 1 and age A, but waits until date t and age A + 1, he will pay his contributions, wt --that is, the contribution rate multiplied by his wage w in t ­ 1, and will benefit A ­1 from a full pension, with liquidation rate p. On the other hand, if he retires early, his bene- fits will be reduced by a fraction d during his entire retirement. That may be summed-up by the following equation N N -wt +A 1 A+j 1 A+j -1 ( ( 1- d)pt+j-1, (10A.1) j=1 1+ r)j .pt+j-1= j=01+ r)j ..( where d denotes the actuarially fair decrual in benefits and N is the last age of pension earnings while j is the time index. To keep things simple, omit the uncertainty prevailing over life expectancy. Taking it into account would lead to multiplying both sides of the equation by survival probabili- ties. That operation would not modify the accounting reasoning. However, risk-averse agents factor life expectancy uncertainty into their choices and behavior: whatever the decrual/accrual rate in pensions benefits associated with early/late retirement, an increase of this uncertainty has the effect of moving forward the retirement date. The left-hand side of equation 10A.1 is the sum of discounted pensions benefits at rate r received from date t by an agent if he retires at age A + 1. The right-hand side is the sum of discounted pension benefits received from date t ­ 1 by the agent if he retires at age A, reduced by a fraction d. The choices made by the workers depend on their individual preferences. Let R(1) and R(2) stand for income equivalents of the welfare of agents in situation 1 (delayed retire- ment) or 2 (early retirement) N R(1) = wt .(1- )+ A 1 A+j -1 ( j=1 1+ )j .pt+j-1 , (10A.2) with the discount rate of the individual considered, N R(2) =(p-d.p)t + A +)j .(p-d 1 A+j A -1 (1 .p)t+j-1 + t-1, (10A.3) j=1 with A t­1the income equivalent of the leisure associated with retiring one year earlier. Let stand for the inflation rate, and , the productivity gains of the economy--gains supposed to pass on to real wages. Nominal wages then increase at the rate + . As done in Artus (2000), pensions benefits are assumed to be indexed at the rate + x, a rate that covers all potential indexing rules. If x = 0, pensions are indexed on inflation; if x = 1, they are indexed on gross wages. This upgrading rate is generally also the rate applied to activ- ity wages for calculating pension benefits. The replacement rate of gross wage, noted , has the following value ptA+1=.wt .(1+ + x.). A -1 220 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Equations 10A.2 and 10A.3 become N-1 R(1) = wt .(1- )+ pt A A+1 1+ + x.)j -1 +j-1( j=0 (1+ )j = wt-1.(1- )+.wt-1 A A N j j=1 1+ + x. /10A.2'/ 1+ R(2) =.(1-d)wt .11A A -1 + + x.. + + N j j=0 1+ + x. 1+ + t-1. /10A.3'/ In addition, the actuarial fairness condition /1/ can be rewritten N j -wt +.wt A A A 1+ + x. -1 -1 N j . /1'/ j=1 1+ + x. 1+ r =.(1-d).wt -1 1+ + j=0 1+ + x. 1+ r Along with equations 10A.2' and 10A.3', it leads to the following, under the following assumptions: · If = r and wAt = A ­1 t­1 , then R(1) = R(2) and the worker perceives no difference between the early or delayed retirement. Either option, under the assumption of an actuarially fair scale at the margin, provides the same satisfaction. · If = r but wAt < A ­1 t­1 , then R(1) < R(2). In that case, the worker draws a high satisfac- tion from leisure and will opt for an early retirement, with the actuarially fair decrual in benefits. · If wAt < A ­1 t­1, what are the determinants of the retirement decision? Equations 10A.2' and 10A.3' yield equation 10A.4 N j R(1) - R(2) = -.wt +.wt . A A A 1+ + x. -1 -1 N j j=1 1+ + x. 1+ -. 1-d wt ( ) -11+ + j=01+ + x. 1+ N j N = -.wt +.wt . A A A A -1 -1 =1 1+ r . (10A.4) j=1 1+ + x. 1+ - -.wt +.wt . j -1 -1 1+ + x. j As soon as < , the contribution rate is less than the replacement rate, which is always the case, and then R(1) > R(2). In other words, when individuals have a high time discount rate, they prefer to retire early, as they do not give much weight to the loss in income induced by an early retirement. · If wAt A ­1 t­1, actuarial fairness at the margin gives + x R( - R( (1+)j (1+r)j 1) 2)= wt - ( A A A 1 1 -1 t-1 )+w t-1 1- 1-d( )1+1 N j-1 + (1++x ) . j=1 - If wAt > A A ­1 t­1and < r, then R(1) > R(2). But if wAt ­1 > t­1and < r, then is R(1) < R(2) as the second part of the right-hand side is always negative and lower in absolute value than wAt . In other words, the individual's preference for the present is such that the ­1 financial gain induced by actuarial fairness at the margin does not mean much to him or her, and the agent retires early. NDCS: A COMPARISON OF THE FRENCH AND GERMAN POINT SYSTEMS 221 Notes 1. "Notional defined contribution" and "non-financial defined contribution" should be understood to have the same definition. 2. See Valdés-Prieto (2000) and World Economic Forum (2004). 3. The 1993 reform changed this point: previously the reference wage was computed as the average of the 10 best indexed wages. The number of years increased by one a year since 1993. The reform will fully hold for the 1948 cohort. 4. The first of July, 2005, the value of the point is 1.1104 euros for an ARRCO point and 0.3940 for an AGIRC point. The purchasing prices are, respectively, 12.6600 (ARRCO) and 4.4163 (AGIRC). 5. These simplications include ignoring the "comp" index and considering a weighed average for the contribution rate as for the price and value of the point. 6. See Hamayon (1995). 7. Both schemes have permanent buffer funds. In 2004, these amounted to 30 billion euros for ARRCO (to cover a 28 billion euro annual payment of pensions) and 10 billion euros for AGIRC (to cover a 14 billion euro annual payment of pensions). 8. See Veil (2003). 9. For example, in the long run--with constant contribution rates for both PAYG and funded schemes--the formula leads to a gross wages indexing rule. 10. See Guérin and Legros (2003), 11. There is obviously a link between (the internal rate of return of the system) and the life expectancy; it is included in the relationship between and the value of leisure. 12. This analysis follows Valdés-Prieto (2000), but adds the consideration of life expectancy. 13. The mechanism of "return spring" is thus a sort of a balancing mechanism, reacting to potential disequilibria by feedback. 14. The Rurup report (summarized in Börsch-Supan 2003) also suggests an increase in the pensionable age; here we discuss only the point mechanism. 15. For example, the French Council for Retirement Schemes' forecasts for 2040 are based on an assumption in which the active population was to increase by 10 million, while the forecasts of the Centre d'Etudes Prospectives et d'Informations Internationales (CEPII) rely on an increase of only 1 million. 16. That will affect the entire active population. In this area, divergences in statistical estimations worldwide are enormous. 17. See Konrad and Wagner (2000). References Artus, P. 2000. "Quelques résultats de base, préalables ŕ la réflexion sur les retraites." Flash CDC-Marchés 2000-17 (January letter). Börsch-Supan, A. 2003. "The Rurup Commission's Pension Reform Recommendations." MEA (Mannheim research institute for the Economic of Ageing) Newsletter 6 (June): 1­4. Bundesministerium für Gesundheit und Sociale Sicherung. 2003. Achieving Financial Sus- tainability for the Social Security Systems. Report by the Commission, A319, Berlin: Edi- tion of the Bundesministerium für Gesundheit und Sociale Sicherung. COR (Conseil d'orientation des retraites). 2001. Retraites: renouveler le contrat social entre les générations. Premier rapport du conseil d'orientation des retraites. Paris: La Documen- tation Française. Guérin, J.-L., and F. Legros. 2003. " Neutralité actuarielle: un concept élégant mais délicat ŕ mettre en oeuvre." Revue d'économie financičre (68): 79­90. 222 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Hamayon, S. 1995. "Perspectives du systčme de retraite en France." In Les retraites, genčse, acteurs, enjeux, ed. B. Cochemé and F. Legros. Paris: Armand Colin. Holzmann, R. 2006. "Toward a Coordinated Pension System in Europe: Rationale and Potential Structure." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 11. Washington, DC: World Bank. Konrad, K., and G. Wagner. 2000. "Reform of the Public Pension System in Germany." Dis- cussion Paper 200, DIW (Deutsches Institut für Wirtschaftsforschung), Berlin. Lassila, J., and T. Valkonen. 2002. "Retirement Age Policies and Demographic Uncertainty in Lithuania: A Dynamic CGE Analysis." Paper prepared for the workshop sponsored by ENEPRI (European Network of Economic Policy Research Institutes) on Welfare and the Labour Market, Marseilles, October 25­26. Mesrine, A. 1999. "Les différences de mortalité par milieu social restent fortes." Données Sociales, INSEE, Paris, 228­35. Queisser, M. 1996. "Pensions in Germany." Policy Research Working Paper 1664, World Bank, Washington, DC. Valdés-Prieto, S. 2000. "The Financial Stability of Notional Account Pensions." Scandina- vian Journal of Economics 102 (3): 395­417. Veil, M. 2003. "La réforme des retraites, premiers pas, débats, interrogations." Chronique internationale de l'IRES 82 (May): 3­16. Verničre, L. 2001. "Allemagne, la réforme 2002 du systčme de retraite." Question retraites, document de travail CDC (Caisse des dépôts et consignations, Paris) 2001-41. World Economic Forum. 2004. "Living Happily Ever After: The Economic Implications of Ageing Societies." A report of the World Economic Forum Pension Readiness Initia- tive developed in partnership with Watson Wyatt Worldwide, Watson Wyatt ed., Geneva. Discussion of "NDCs: A Comparison of the French and German Point Systems" Marek Góra* THE CHAPTER BY FLORENCE LEGROS presents a very interesting overview and comparison of the French and the German pension systems. Legros focuses on vital elements of the systems and mechanisms behind them. The chapter provides the reader with arguments to confirm that the two crucial features of both systems analyzed are the complexity of reg- ulations ruling participation in the system and the system itself, and exposure to discre- tional adjustments. These features are not exclusive to the French and the German systems. They are found in the majority of systems, especially in continental European pension systems (I call them traditional systems). These features did not appear by chance. These are the most crucial features of traditional pension systems. Without detailed regulations and without discre- tional decisions, traditional pension systems--including point systems--will crash. The growing popularity of notional defined contribution (NDC) is leading to an overuse of its name. In many cases, including Legros's chapter, NDC is understood in a very broad sense: namely, as any pension arrangement that assumes some linkage between contributions and benefits. In her chapter, Legros discusses practical cases of NDC arrangements that have been implemented. The discussion leads her to the conclu- sion that "a pure NDC does not exist." The conclusion is probably correct, but the same type of conclusion can be reached regarding other types of pension arrangements. How- ever, this type of approach locks us into a very conservative conceptual framework. In phenomena like pension systems, the aim is not to seek purity. I would rather suggest seeking concepts that can help us understand the nature of the systems. NDC combines features of different pension arrangements, which opens the door for possibly fruitful research that can contribute strongly to the development of pension economics. NDC has features that are opposite to the ones highlighted in the case of the French and the German point systems. NDC is very easy to describe and understand, and is transpar- ent. The shortest possible description of participation in NDC is the following: At the moment of retirement, participants receive from the system (in the form of an annuity) the present value of contributions they have paid throughout the working period. The discount factor is the GDP growth rate, or is close to it. The above is an ex ante arrangement that cannot only be understood but also internalized by the participants. NDC has built-in automatic stabilizers. The system adjusts revenues and expenditure without discretional intervention from outside (that is, it is an endogenous arrangement, * Marek Góra is a professor at the Warsaw School of Economics. 223 224 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES or ENDO). In practical application, this can be less than 100 percent effective; however, it is close to 100 percent. This does not ensure complete resistance to politicians' attempts to manipulate the system. Such an arrangement cannot exist in reality. This point also applies to financial defined contribution (FDC) plans. Nonetheless, NDC is much less exposed to political manipulation than any system that cannot function without intervention from outside. Both the French and German pension systems require interventions (or exoge- nous arrangement, EXO). Is automatic adjustment a desirable feature of a pension system? Interventions from outside can at best reach the solution that individual accounts (ENDO) lead to automati- cally. The challenge is counteracting the ever-growing burden put on each successive gen- eration of workers, who are increasingly discouraged in their labor supply and productivity growth. A good pension system should lead to a stable division of GDP between generations. The shares are subject to public choice. Once chosen, they should be kept stable. Individual accounts of both types--NDC and FDC--are endogenous arrange- ments, which lead to intergenerational equilibrium. Being in equilibrium, a pension sys- tem is economically neutral. This means remuneration of production factors is not affected by declining proportions of the retired and working generations, which leads cetreris paribus to the strongest possible growth of welfare of each generation being first a work- ing one and then a retired one. Traditional EXO arrangements would not lead to this out- come unless proper decisions are taken. However, they are notoriously postponed. The growth in pension systems' debts, which slow down economic growth, clearly prove that. In Legros's chapter, as is common in pension literature, NDC is reduced to a tricky way to balance defined benefit (DB) pay-as-you-go (PAYG) schemes. We are hostages of termi- nology inherited from social policy and finance. From the economic viewpoint, NDC is neither DB nor PAYG. Instead, in a very broad sense NDC is similar to a universal FDC scheme that spends the entire sum of contributions paid into it in buying government bonds and keeps them until maturity, and receiving a rate of return equal to economic growth rate. Discussing this issue in detail goes beyond the scope of this short discussion. Being so similar, NDC and FDC fit each other very well. Latvia, Poland, and Sweden not only implemented NDC in their pension systems but also terminated EXO arrange- ments entirely. They have systems entirely based on individual retirement accounts: NDC and FDC. This is the revolution in European pensions, even if some technical elements that are not "pure" remain. Chapter 11 Toward a Coordinated Pension System in Europe: Rationale and Potential Structure Robert Holzmann* THE NEED FOR A RAPID AND COMPREHENSIVE REFORM of the pension systems in most current and future member countries of the European Union (EU) is increasingly acknowledged by pension scholars and politicians. While a few countries have recently undertaken major reforms to make their pension systems financially sustainable, in the majority of European countries the reform efforts are still insufficient. Although national efforts can now draw support on intensified EU cooperation based on the Open Method for Coordination, this method takes the diversity of European pension design as a given, and much of the reform debate is still limited to fiscal issues at national levels. There is little discussion about a reform need beyond fiscal consideration. There is no discussion (anymore) about a reform move toward a more coordinated pension system within the European Union, and how such a system may look and come about. That is the topic of this chapter. To this end, it pro- gresses in four sections. The second section reviews the reform needs of the pension sys- tems for fiscal, social, and economic reasons. The third section makes the case for a move toward a more coordinated pension system in Europe. The fourth section sketches how such a system may look and come about. The central claim of the chapter is that a multipil- lar system, with a non-financial (or notional) defined contribution (NDC) system at its core and coordinated supplementary funded pensions and social pensions at its wings, is an ideal approach to deal with diverse fiscal, social, and economic reform needs. The approach would also introduce a harmonized structure while allowing for country-specific prefer- ences with regard to coverage and contribution rate. Such a reform approach may lead to a Pan-European reform movement as a number of countries have already or plan to intro- duce NDCs, and others may easily convert their point system into an NDC structure. * Robert Holzmann is director, Social Protection Department, at the World Bank. Revised study prepared for the joint Watson Wyatt & Deutsches Institut fuer Wirtschaftsforschung (WW-DIW) Lecture on Issues in Pension Reform, Berlin, September 26, 2003, and the NDC Conference in Sandhamn, Sweden, September 28­30, 2003. The draft and the revisions have benefited from valuable comments and suggestions by lecture and conference participants, in particular Bernd Marin and Edward Palmer, a pre- sentation at the EU Commission in Brussels on October 31, 2003, discussions with World Bank staff, and able research support by Kripa Iyer. This paper has not undergone the review accorded to official World Bank publications. The findings, interpretations, and conclusions reflected herein are those of the authors and do not necessarily reflect the views of the International Bank for Reconstruction and Development/The World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. 225 226 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES The Need for Pension Reform in EU and EUA Countries There are three main reasons1 why EU countries and the new EU accession countries (EUA) in Central, Eastern, and Southern Europe need rapid and comprehensive reforms of their national pension systems:2 First, the current high expenditure level and related bud- getary pressure will only worsen given the projected further aging of populations. The national systems need to be reformed to handle aging in a manner consistent with indi- vidual preferences and macroeconomic constraints. Second, ongoing socioeconomic changes are rendering current retirement income provisions inadequate at the social and economic level. Third, globalization creates opportunities and challenges, and to deal with them effectively requires, inter alia, benefit and tax regimes that improve the functioning of factor markets. The expenditure level for public pensions in most Western European countries is well above that of other highly industrial and postindustrial countries at a similar income level. The average of public pension expenditures as a percentage of gross domestic product (GDP) for the 15 EU countries in 2000 amounted to 10.4 percent (this is a low estimate because it includes only the expenditure under the projection exercise of the Economic Policy Committee 2001). The Organisation for Economic Co-operation and Development (OECD) estimate is about 1.3 percentage points higher (OECD 2002). The average for the non-European and affluent OECD countries--Australia, Canada, Japan, New Zealand, the Republic of Korea, and the United States--in 2000 was about 5.3 percent: that is, roughly half. In the EU, only Ireland (4.6 percent) and the United Kingdom (5.5 percent) have sim- ilar levels. This difference is also shared by the accession countries in Central and Eastern Europe. Except Romania (5.1 percent), all others have expenditure shares close to the EU Figure 11.1. Pension Expenditure in EU and EUA Countries (plus Croatia), 2000 or latest (percentage of GDP) 16 14 12 10 8 GDP 6 of 4 2 0 percentage Italy Austria Bulgaria CyprusDenmark FinlandGermanyHungary Latvia Malta Spain Belgium CroatiaRepublic Estonia France GreeceIreland Sweden Lithuania PolandRomania Slovenia PortugalRepublic Kingdom Luxembourg Netherlands Czech Slovak United Source: EPC 2001; Palacios and Pallares-Miralles 2000, updated; World Bank 2003b. Note: Croatia data from World Bank 2003b. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 227 average (and in Croatia, Poland, and Slovenia, well above) and hence much higher than non-European OECD countries, despite an income level of one-quarter and less. Poland's public pension expenditures, at close to 15 percent of GDP, rival that of Austria and Italy for the world championship (see figure 11.1). The gap between these expenditure levels and those in non-European OECD countries is only partially explained by differences in population age structure. Rather, it reflects differences in the public/private mix of provi- sions and in the benefit levels and the effective retirement age in the public systems. The replacement rate is generally much higher as public (largely unfunded) pensions are little supplemented by private and funded arrangements (except in Denmark, Ireland, the Netherlands, and the United Kingdom). The effective retirement age is typically low as a result of disincentives to work longer in current public schemes, special options for early retirement, and past labor market policy that deliberately attempted to keep the unem- ployment rate low by allowing older workers to exit prematurely. Yet the demographic component in pension expenditure is going to increase under unreformed systems as aging in Europe accelerates. In Europe, the total fertility rate has been below replacement level (approximately 2.1) since the 1970s in the West and since the 1980s in the East, and there are few signs of a rebound from the current low levels. On the other hand, life expectancy is likely to increase during the next 50 years by 4.2 years for women and 5 years for men. As a result, for the EU15, the old-age dependency ratio is projected to increase from 27.7 percent (2000) to 53.4 percent (by 2050) (see table 11.1), based on rather optimistic assumptions with regard to total fertility rate (assumed to rise again to 1.8 in most countries) and life expectancy (assumed to rise less than in the past). The projections for the EUA countries are very similar (United Nations 2002); actually the projected pace of aging in the EUA is faster. Based on this projected change in the old-age dependency ratio in the East and the West, and in a no-reform scenario, expenditures would roughly double. Of course, such a radical expenditure increase would not necessarily materialize because some reform measures have already been enacted, and system dependency ratios (beneficiaries to contributors) may not deteriorate to the same extent as old-age depen- dency ratios. Greater labor force participation by women is likely and that of both older (55+) men and women may increase as well. This, at least, is the scenario put forth by the Economic Policy Committee of the EU, and the country projections for the period 2000 to 2050 (EPC 2001; see table 11.2).3 As a result, the average EU public pension expenditures (captured under this exercise) are projected to increase "only" from 10.4 percent of GDP in 2000 to a peak of 13.6 percent around 2040 (with a projected fall from 5.5 to 4.4 percent for the United Kingdom, but almost a doubling for Spain from 12.6 to 24.8 percent). This mod- erate projected 30 percent increase of the average expenditure level (compared with a pure demographically induced increase of some 70 percent) is estimated as a result of lower benefit ratios (average benefits compared to GDP per capita) and higher employment ratios (employment to population aged 15 to 64). However, this modest increase in EU average public pension expenditure levels will require major changes in the pension schemes and their incentives for enhanced labor market participation and delayed retire- ment decisions. Put differently, a further major increase in pension expenditure can be pre- vented only if major reforms take place. No similar and coordinated projection exercise has been undertaken for the new EU member states, but existing projections clearly paint a two-class picture (EPC 2003): In countries that have undertaken major reforms--such as Hungary and Poland--the expen- diture share remains largely unchanged (and a similar path can be conjectured for reformed systems in Estonia and Latvia). In countries where a major reform is still out- 228 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Table 11.1. Projections of Old-Age Dependency in EU and EUA Countries, 2000­50 (ratio of people aged over 64 to working age population, percent) Country 2000 2010 2020 2030 2040 2050 Austria 25 29 32 44 55 55 Belgium 28 29 36 46 51 50 Denmark 24 27 34 39 45 42 Finland 25 28 39 47 47 48 France 27 28 36 44 50 51 Germany 26 33 36 47 55 53 Greece 28 32 36 42 51 59 Ireland 19 19 25 30 36 44 Italy 29 34 40 49 64 67 Luxembourg 23 26 31 40 45 42 Netherlands 22 25 33 42 48 45 Portugal 25 27 30 35 43 49 Spain 27 29 33 42 56 66 Sweden 30 31 38 43 47 46 United Kingdom 26 27 32 40 47 46 EU average 27 30 35 44 52 53 Bulgaria 24 24 29 34 41 53 Cyprus 18 20 26 32 34 39 Czech R. 20 22 32 38 47 59 Estonia 23 25 30 36 42 57 Hungary 21 23 29 33 40 50 Latvia 23 26 29 37 44 56 Lithuania 21 24 26 35 40 43 Malta 18 22 32 39 40 46 Poland 18 18 26 33 37 50 Romania 20 20 24 26 36 45 Slovak R. 16 17 23 30 36 47 Slovenia 20 24 32 44 53 64 EUA average 20 22 28 35 41 51 Sources: EU countries--EPC (2001); EUA countries--UN (2002). standing, the expenditure share in percentage of GDP is projected to increase dramatically: to almost double in Cyprus and the Czech Republic, and to increase further from an already high level in Slovenia. World Bank internal projections are largely consistent with this picture. Even if the budgetary and demographically induced pressures did not exist, there still would be a major need for most European countries to reform their public pension sys- tems to better align them with socioeconomic changes. Three changes stand out: increas- ing female labor force participation; high divorce rates and changing family structures; and the rise in atypical employment. Furthermore, rising life expectancy and other changes also call for a rethinking of the design of disability benefits. In the EU countries, the labor force participation of women has increased substantially over recent decades. In the former centrally planned countries, it was very high, but it decreased during the transition period of the 1990s (see table 11.3). The decrease for women followed that of men and was in some countries even less pronounced (World Bank 2003a). Although there are differences among EU countries (for example, in Italy, TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 229 Table 11.2. Public Pension Expenditure in EU and EUA Countries in 2000­50 (percentage of GDP) Country 2000 2010 2020 2030 2040 2050 Austria 14.5 14.9 16.0 18.1 18.3 17.0 Belgium 10.0 9.9 11.4 13.3 13.7 13.3 Denmark1 10.5 12.5 13.8 14.5 14.0 13.3 Finland 11.3 11.6 12.9 14.9 16.0 15.9 France 12.1 13.1 15.0 16.0 15.8 -- Germany 11.8 11.2 12.6 15.5 16.6 16.9 Greece 12.6 12.6 15.4 19.6 23.8 24.8 Ireland2 4.6 5.0 6.7 7.6 8.3 9.0 Italy 13.8 13.9 14.8 15.7 15.7 14.1 Luxembourg 7.4 7.5 8.2 9.2 9.5 9.3 Netherlands 7.9 9.1 11.1 13.1 14.1 13.6 Portugal 9.8 11.8 13.1 13.6 13.8 13.2 Spain 9.4 8.9 9.9 12.6 16.0 17.3 Sweden 9.0 9.6 10.7 11.4 11.4 10.7 United Kingdom 5.5 5.1 4.9 5.2 5.0 4.4 EU 10.4 10.4 11.5 13.0 13.6 13.3 Cyprus 8.0 -- -- 11.9 -- 14.8 Czech Republic4 7.8 -- -- -- -- 14.6 Estonia 6.9 -- -- -- -- -- Hungary4 6.0 -- -- -- -- 7.2 Latvia3 9.8 -- -- -- -- -- Lithuania 5.3 -- -- 6.0 -- 7.0 Malta 5.4 -- -- -- -- -- Poland 10.8 -- -- 9.6 -- 9.7 Slovakia3 7.9 -- -- -- -- -- Slovenia 13.2 -- -- 19.7 -- 18.1 Bulgaria 9.1 -- -- -- -- -- Romania 6.4 -- -- 7.8 -- 8.2 EUA 8.05 -- -- 11.0 -- 11.4 Source: EPC (2001), except as indicated. Note: For most EU member states, these projections include most public replacement income for persons aged 55 and over. -- = not available. 1. For Denmark, the results include the semi-funded labor market pension (ATP). 2. Results for Ireland are as percentage of GNP, not GDP. 3. Source is Gesellschaft fur Versicherungswissenschaft und -gestaltung e. V. (which in turn draws on national statistics). 4. Source is OECD 2002. female labor force participation in the age group 15­64 in 2000 stood at a low 46 per- cent, in contrast to Denmark where a 77 percent female participation rate is close to that of men), a further increase is projected for all countries. The EU average for women in the age group 15­54 is projected to increase from 63 to 76 percent, whereas that for men is projected to remain largely constant at around 85 percent. So far this change in female labor force participation is little reflected in countries' pension benefit structures (see table 11.4). The benefit rules still largely reflect the traditional image of a working hus- band and a child-caring housewife who needs a widow's pension for her protection in Table 11.3. Labor Force Participation, Male and Female, in EU and EUA countries, 1960, 1980, 2000, and 2050 (percent) Male Female 15­64 65+ 15­64 65+ Country 1960 1980 2000 2050 1960 1980 2000 2050 1960 1980 2000 2050 1960 1980 2000 2050 Austria 90.1 84.9 74.4 79.3 15.0 4.5 2.0 6.0 53.0 54.4 57.7 67.8 7.0 2.6 1.0 5.0 Belgium 85.9 79.7 71.7 71.9 9.5 4.6 1.4 1.3 30.5 41.2 58.6 67.8 3.3 1.3 0.5 0.6 Denmark 92.3 88.3 85.1 81.8 32.6 15.4 9.4 8.1 42.8 71.3 77.3 80.5 8.0 5.2 2.7 2.4 Finland 87.3 79.3 74.8 73.9 31.7 6.8 4.0 2.5 55.5 69.4 73.0 74.7 12.0 3.0 1.4 1.0 France 88.9 81.5 75.6 75.1 26.0 5.8 2.1 1.7 43.6 55.1 62.2 70.0 10.2 2.9 1.2 1.0 Germany 91.0 83.2 80.7 80.1 24.0 8.9 4.5 2.4 50.4 51.9 64.7 71.3 8.0 4.2 1.7 1.1 Greece 90.1 83.5 76.7 76.6 45.0 27.0 9.6 7.9 26.3 31.8 46.7 67.0 8.7 6.1 3.7 3.2 Ireland3 90.9 85.0 87.8 87.3 54.0 26.8 13.6 11.7 31.1 34.7 56.4 75.8 15.0 6.0 2.4 2.0 Italy 89.2 79.0 73.0 76.1 27.5 12.4 5.5 3.7 30.4 38.4 46.4 66.9 5.6 3.5 1.5 1.4 Luxembourg1 88.7 82.3 113.8 148.4 -- -- -- -- 30.8 39.0 74.3 115.0 -- -- -- -- Netherlands 90.9 77.6 77.4 76.2 19.9 4.8 1.0 1.0 24.9 36.1 55.2 70.9 2.5 1.0 1.0 1.0 Portugal 93.5 87.1 87.5 87.2 62.9 29.7 16.7 14.3 18.4 52.4 66.4 81.5 11.0 8.4 7.1 6.5 Spain3 92.8 86.4 83.6 85.5 56.6 12.3 2.8 2.8 20.3 32.9 54.7 75.2 9.4 4.1 1.1 1.1 Sweden 88.8 87.9 81.3 83.3 27.6 10.4 6.8 7.2 38.0 75.3 76.5 82.6 4.5 2.6 3.5 3.9 United Kingdom3 94.6 89.2 87.6 85.9 26.6 11.0 6.8 5.8 43.6 57.0 69.9 75.5 5.4 4.1 2.7 2.4 EU 90.3 83.7 82.1 84.6 32.8 12.9 6.2 5.5 36.0 49.4 62.7 76.2 7.9 3.9 2.3 2.3 Bulgaria 88.4 82.7 77.2 77.2 38.3 18.8 10.1 8.6 68.9 70.4 71.4 68.8 8.5 3.9 3.0 2.5 Cyprus 91.7 88.6 88.0 86.1 53.0 35.7 20.5 0.2 42.0 46.7 56.9 59.1 17.6 11.8 7.8 6.3 Czech R. 86.5 84.8 83.0 80.6 24.4 18.8 11.7 10.3 61.6 75.0 75.0 71.7 9.2 7.1 4.9 4.5 Estonia 87.2 85.4 81.7 81.5 20.5 17.5 23.0 22.4 67.3 79.2 74.0 74.0 6.8 9.5 13.3 13.4 Hungary 91.7 84.8 78.7 76.2 57.0 3.8 0.9 0.9 46.9 62.0 61.1 60.5 20.0 3.0 0.2 0.2 Latvia 84.8 84.8 82.2 83.0 24.3 22.4 20.2 19.2 64.3 77.9 74.2 75.4 12.8 12.3 11.3 10.8 Lithuania 83.3 83.0 81.2 81.7 32.0 19.4 12.3 11.0 61.3 74.8 70.8 71.7 9.5 7.8 6.5 5.9 Malta 88.0 85.7 78.8 76.3 27.3 14.3 5.0 4.0 17.2 22.5 30.2 34.6 0.0 0.0 0.0 0.0 Poland 89.8 84.2 77.9 77.4 57.5 30.0 24.1 21.3 62.1 67.7 66.2 66.4 30.0 17.5 15.3 13.8 Romania 93.0 83.6 76.8 76.7 62.6 11.4 4.9 3.9 72.4 69.0 61.2 61.4 30.0 8.9 4.2 3.5 Slovak R. 86.5 83.5 82.1 81.8 30.9 19.8 11.0 9.6 47.4 69.3 74.6 72.7 7.7 4.7 4.2 3.6 Slovenia 89.9 81.9 76.0 74.1 57.1 19.0 11.8 10.2 44.3 67.0 66.5 64.9 13.5 10.0 8.6 8.0 EUA2 88.4 84.4 80.3 79.4 40.4 19.2 13.0 10.1 54.6 65.1 65.2 65.1 13.8 8.0 6.6 6.0 Sources: EPC (2001); OECD (2003); ILO (2003); UN (2002). Note: -- = not available. 1. Estimates for Luxembourg assumes increase in cross-border workers, which explains the high rate. 2. Projections for EUA countries are for the year 2010. 3. Population aged 20­64. Table 11.4. Pension Arrangements for Widows/Widowers and Divorcees in EU and EUA Countries around 2000 Widow/widower's benefit Divorcee's benefit Country Elgibility Benefits Eligibility Benefits Austria Deceased met insurance or con- Up to 60% of deceased spouse's -- -- tribution requirements for dis- pension, income tested. Rates below ability pension or was a 60% may be increased depending pensioner. on beneficiary's income. Belgium Those aged 45+, or disabled, or 80% of deceased spouse's pension. Special pension at age 37.5% of former caring for a child. Should have Minimum 9,102.11 euros/year if 60. spouse's earnings dur- been married for at least 1 year worker was fully insured; if not, ing period of marriage at the time of spouse's death. then reduced. If widow(er) receives less pension earned in Conditions are waived if child other pension, receives survivor own right during the born out of marriage or in case pension only for 12 months and same years. of accidental death. total pension benefits may not exceed 110% of own pension. -- Bulgaria Deceased had 5 years of Minimum pension for each survivor -- service, or 3 years if aged 20­25 is 90% of social pension. One or was a pensioner. survivor, 50% of deceased's pension; if 2, 75%; if 3 or more, then 100%. -- Cyprus Conditions same as for old-age Same as old-age pension + 60% -- pension, lump sum paid if con- supplementary pension. Widow ditions not met. Payable to may substitute husband's coverage widow or dependent disabled record for her own for period prior widower. to his death. -- Czech Republic Deceased met pension Basic amount of 1,310 koruny plus -- conditions or was a pensioner. 50% of deceased's pension, payable to all widow(er)s for 1 year, thereafter only to widow(er)s aged 55(58), or any age if disabled or car- ing for disabled/dependent child or disabled parent. Table 11.4. (continued) Widow/widower's benefit Divorcee's benefit Country Elgibility Benefits Eligibility Benefits Denmark Survivor pension eliminated as of 1984. Lump sum paid to widow(er) and -- -- children under 18 of deceased; amount depends on pension of the deceased. Estonia Widow(er) not capable of gainful activity; One survivor, 40% of deceased's pension -- -- deceased had 1­14 years of coverage entitlement; 2 survivors, 70%; 3 or more depending on age. 100%. Finland Under age 65 if caring for a child; if child- Universal pension awarded for first 6 Eligible for survivor's -- less then at least age 50 at time of months after spouse's death, thereafter pension if not spouse's death and must have been mar- becomes income-tested. remarried; pension ried for at least 5 years, residing in proportionately Finland. divided if more than one surviving spouse. France At least 55 years and married for 2 years. 54% deceased spouse's pension, income- Former spouse eligible 54% deceased Conditions are waived if child from mar- tested, payable for 2 years. If beneficiary for survivor's pension. spouse's pension. riage or if widow(er) and deceased were is age 50, payment extended until 55. Amount split between disabled. Personal income must be less widow(er) and former than 13,874 euros/year and must not spouse according to have remarried. length of marriage. Germany Deceased had 5 years of coverage, or was 100% of deceased's pension for first 3 -- -- a pensioner. months; 55% if aged 45+, disabled, or caring for a child; otherwise 25%. Greece Eligible for survivor's pension for 3 years, Full pension paid if disabled. Those who -- -- those above 40 continue to receive it pro- work or receive other pension get 50% of vided they do not work or receive any normal survivor pension. When other pension. survivors reach age 65 they are paid full pension, if receiving other pension at 65+ then they get 70% of normal pension. Hungary Deceased was pensioner or met 50% of insured's pension paid to -- -- requirements for pension at death. widow(er) who at the time of death was 55(60), disabled, or caring for 2 children, paid to other widow(er)s for 1 year only. Ireland Annual average of at least 39 weeks paid Contributory pension: up to 123.30 -- -- or credited in last 3 or 5 fiscal years prior euros/week (144.80 euros if aged 66+), to date spouse died or attained age 66, at noncontributory pension: up to 118.80 least 24 weeks for minimum pension. euros/week (134.00 euros if age 66+). Italy Deceased was a pensioner or had 5 years 60% of insured's pension, 80% if 1 child, Separated spouse eligi- -- of contribution of which 3 years were in 100% if 2 or more children; lump sum ble for survivor's bene- the last 5 years. paid if conditions for survivors pension fit. not met; must have paid at least 1 year's contribution in last 5 years. Latvia Deceased was insured or was a pensioner. 50% of insured's pension, 75% if 2 -- -- survivors, 90% for 3 or more. Lithuania Deceased must have been a pensioner or 20% of deceased's benefit, 25% for each -- -- had adequate coverage for disability pen- child, total may not exceed 80% of sion at the time of death. Widow(er) who deceased's pension. has reached old age or is disabled is eligible. Luxembourg Insured had 12 months of coverage in 3 100% of insured's basic old age pension Divorced spouse eligi- Amount depends on years prior to death or was a pensioner. plus 75% of increment earned by ble. years of marriage, insured, payable without regard to per- not on personal sonal income. income. Malta Deceased paid 156 weeks of contribution Benefit varies depending on whether con- -- -- with annual average of 50 weeks, paid or tributions were made before or after Janu- credited, reduced pension awarded for ary 22, 1979. Earnings-related benefit that less coverage, earned income of widow(er) can be as much as 70.72 liri/week are must not exceed minimum wage. Widows 5/9th yearly average of best 3 consecutive under age 60 with children under 16 qual- years of last 10 years before husband's ify regardless of income. death or retirement. Upon remarriage, widow forfeits benefit from previous mar- riage and receives lump sum equal to 52 weeks pension. Table 11.4. (continued) Widow/widower's benefit Divorcee's benefit Country Elgibility Benefits Eligibility Benefits Netherlands Residents eligible. Payable to Income-tested for those born before -- -- widow(er)/unmarried permanent 1950, those 45% disabled, 932.38 partner. euros/month for those caring for child under 18, benefit reduced by survivor's income from employment. No benefits if income > 2,002.54 euros/month. Poland Deceased was a pensioner or met One survivor, 85% of deceased's pen- -- -- employment requirements for old- sion; 2 survivors, 90%; 3 or more, 95%. age pension or disability benefits. Portugal Deceased met pension 60% of insured's pension. Payable for 5 -- -- requirements or was a pensioner. years only unless beneficiary over 35, disabled, or caring for a child. Romania Insured met pension requirements Limited benefit paid for 6 months to -- -- or was a pensioner at the time of low-income spouse caring for child death. Widows must fulfill certain under age 7 who does not meet eligi- age conditions and also duration of bility conditions, 50% of deceased's marriage requirements. No prior old-age pension; 2 survivors, 75%; 3 or requirements if death was by work more, 100%. accident, occupational disease, or tuberculosis. Slovak Republic Deceased met pension 60% of insured's pension payable to -- -- requirements or was a pensioner. widows for 12 months, thereafter only to widows aged 50; aged 45 if she has reared 2 or more children; aged 40 if husband died in occupational accident; any age if disabled, caring for a child, having cared for 3 or more children; widowers pension 1,977 koruny/month. Slovenia Deceased met pension (old age or 70% of insured's pension; 2 survivors, -- -- disability) requirements or was a 80%; 3 survivors, 90%; 4 or more, pensioner and had 5 years of cov- 100%. erage and contribution; widow(er) must be at least 52(53) as of 2003. Spain Deceased had 500 days of contri- 46% of either the deceased's or Ex-spouse not eligible for -- bution in the last 5 years, was survivor's benefit base, whichever is old-age pension once pensioner at time of death, or had higher, for income below a particular remarried unless 61+ at 15 years of contribution. level--50%, 70% if there are time of marriage, 65% dependents. disabled or survivor pension is 75% of pensioner's total income. Sweden Residents eligible. Deceased must Benefit payable for 6 months if -- -- be credited with pension points married or cohabiting for at least 5 for at least 3 years or have 3 years years, under certain conditions. coverage. Payable for as long as living with child under 12. Special pension paid if unemployment or illness prevents self- support. United Kingdom Deceased met coverage Weekly allowance to those above age -- -- requirements or was a pensioner. 45 without dependent children payable for 52 weeks after death of spouse. Amount depends on age at widowhood. Widow aged 18­59 with dependent children gets weekly allowance of 53.05 pounds plus 31.45­32.25 pounds for each child minus amount of other benefits/income. Source: United States Social Security Adminstration (2002). 236 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES old age. Only a few countries, such as Denmark and Sweden, have fully moved toward independent pension rights and eliminated the traditional widow's pension (Denmark in 1984 and Sweden in 1991). As a result, there is often underprovisioning for young widows with children, and overprovisioning for widows with their own pensions--the latter group now includes widowers. To ensure gender neutrality, survivor's pensions in many countries have been extended to male spouses and the budgetary conse- quences are increasingly attempted to be curtailed by ceilings and tapers. Furthermore, eligibility for survivor's pensions is complicated by the rising divorce rate. In a number of countries the divorce rates are more than 50 percent of the rates of marriage per 1,000 inhabitants (see table 11.5), which means that in many countries more than 50 percent of marriages will not survive, including second or third marriages. Coun- tries with a more conservative divorce behavior, such as Italy and Ireland, can be expected to catch up quickly. But only a few countries have moved in the direction of establishing Table 11.5. Changing Family Structures: Divorces and Marriages in EU and EUA Countries around 2000 (per 1,000 people) Country Divorces (per 1000 people) Marriages (per 1000 people) Ireland 0.7 5.1 Italy 0.7 4.9 Greece 0.9 5.4 Spain 1.0 5.2 Portugal 1.8 5.7 France 2.0 5.1 Luxembourg 2.3 4.5 Netherlands 2.3 5.1 Germany 2.4 4.7 Sweden 2.4 4.0 Austria 2.5 4.2 Finland 2.6 4.8 United Kingdom 2.6 5.1 Denmark 2.7 6.6 Belgium 2.9 4.2 EU average 1.9 5.1 Bulgaria 1.3 -- Cyprus 1.7 -- Czech Republic 2.9 -- Estonia 3.1 -- Hungary 2.4 -- Latvia 2.6 -- Lithuania 2.9 -- Poland 1.1 -- Romania 1.4 -- Slovakia 1.7 -- Slovenia 1.1 -- Malta -- -- EUA 2.0 -- Sources: EU countries--EU (2003); U.K. Office of National Statistics, Inc. (2001). EUA countries--Americans for Divorce Reform (2003); UN (2001); Council of Europe (2001). Note: -- = not available. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 237 independent rights for spouses (and even less for partners): that is, the individualization of pension rights. In many countries, benefit traps for women still exist: that is, incentives against rejoining the labor market or remarrying when eligibility for a survivor's pension has been achieved. Another and more recent development concerns the rise in atypical employment: that is, the reduction in full-time salaried employment and the increase in part-time employment, self-employment, and temporary employment (see table 11.6). This development may be ascribed to globalization and competitive pressure that makes full-time employment and a life career with the same employer less dominant than it used to be; it may also be linked to more self-selected flexibility in the labor market (including the choice of retirement pro- visions). Data for OECD countries suggest that coverage under public pension schemes is decreasing (Holzmann 2003). Whatever the reason, these atypically employed people do not fare well under some pension schemes, which are based on the full-time employment fiction. In many current systems, the atypically employed fare extremely well, which lim- its their incentives to contribute on a continued basis. Again, this situation calls for reforms (and a stricter contribution-benefit relationships). Socioeconomic changes also call for a review and redesign of disability benefits, includ- ing the delinking of design, delivery, and financing of old-age benefits. At the start of the Bismarckian-type pension scheme, disability benefits were much more important for indi- viduals than old-age benefits as only one in six workers could expect to reach the advanced retirement age of 70. Old-age pensions then can be conceptualized as generalized or cate- Table 11.6. Selected Work Arrangements in Europe, 1988 and 1998 (percent of total employment) Total Self-employment employment (including Part-time Temporary (000s) family workers) employment employment1 Country 1988 1998 1988 1998 1988 1998 1988 1998 Austria -- 3,626 -- 13.8 -- 15.8 -- 6.8 Belgium 3,483 3,857 18.0 17.4 9.8 15.7 4.5 6.4 Denmark 2,683 2,679 11.0 9.7 23.7 22.3 10.2 9.1 Finland -- 2,179 -- 14.6 -- 11.7 -- 15.1 France 21,503 22,469 16.2 12.5 12.0 17.3 6.6 12.2 Germany 26,999 35,537 11.5 11.0 13.2 18.3 10.1 10.9 Greece 3,651 3,967 49.5 43.4 5.5 6.0 8.8 7.4 Ireland 1,090 1,496 25.3 20.2 8.0 16.7 6.8 6.1 Italy 21,085 20,357 29.5 28.7 5.6 7.4 4.1 6.1 Luxembourg 152 171 11.2 9.4 6.6 9.4 3.3 2.4 Netherlands 5,903 7,402 12.1 11.6 30.3 38.8 7.7 11.2 Portugal 4,427 4,764 30.9 28.2 6.5 11.1 12.6 12.4 Spain 11,709 13,161 29.1 23 5.4 8.1 15.8 25.3 Sweden -- 3,946 -- 11.4 -- 23.9 -- 11.4 United Kingdom 25,660 26,883 12.7 12.5 21.9 24.9 5.2 6.1 EU 128,345 152,494 19.1 16.6 13.2 17.4 7.8 10.6 Source: Holzmann (2003). Note: -- = not available. 1. Dependent employees, including apprentices, trainees, and research assistants. 238 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES gorical disability pensions: that is, insuring much the same risk. Nowadays an old-age pen- sion is a life-annuity paid with accumulated funds or acquired rights and insures against the uncertainty of death. Conceptually, it is totally delinked from failing work capacity. But the original design of disability pensions and the close link to old-age pensions still prevail in many of the European pension systems, and the mixed design has also helped the use of disability pensions as a form of unemployment benefits in many countries. Furthermore, sport and car accidents instead of general incapacity have become a major reason for the granting of disability benefits, in particular at younger ages. As a final argument, disability benefits--insurance-based and means-tested--need to be reviewed and integrated into the design of an overall work/benefit package for the disabled (OECD 2003). Last but not least, globalization--understood as high and increasing integration of mar- kets for goods and services, factors of production, and knowledge--calls for changes in the way public programs operate, including in the area of pension provision. Such reforms are needed not only to reap the benefits of globalization but also to deal with challenges that include profound shocks resulting from technical innovations and shifts in the demand and supply of goods and factors. This calls, inter alia, for more flexibility across labor markets, improved financial markets, and lifelong learning. A main conjecture about the fate of nations and their economic performance in a global- ized world is their capacity to deal with shocks, in particular those that require the existing economic structure to adjust. It is claimed that the more flexible and adjustable an econ- omy is in reacting to such shocks, the better it will fare. Such flexibility comprises mobility of individuals across professions, including between the public and the private sector. In most European countries, such mobility is hampered by separate pension schemes between both sectors that limit if not eliminate any movement between them. Moreover, separate schemes render the application of some reform measures difficult or counterpro- ductive. For example, increasing the retirement age for all primary school teachers to, say 67, may not be in the best interest of all concerned, but it is feasible if a teacher can move easily to a related or different profession. The integration of countries into the world economy is importantly linked with their own financial sector development. A developed domestic financial market is a main ingredi- ent for full capital account convertibility, including the capacity to diversify pension assets internationally.4 International diversification is, perhaps, the only free lunch in the world, and it promises major welfare effects as long as national and international rates on return of retirement assets (beyond shares) are little correlated. This requires, however, that some minimum domestic financial market exists. Forcing individuals to hold most or all of their pension assets in illiquid pay-as-you-go (PAYG) assets is not an optimal strategy for deal- ing with diverse risks to which individuals are exposed and is clearly not welfare enhanc- ing. Restricting a country solely to PAYG is truly an example of taking the risk of putting all pension eggs into one basket. Pension reforms that include introducing or strengthen- ing a funded pillar allow such a risk diversification and at the same time can importantly contribute to development of the domestic financial market. Well-developed domestic financial markets are a critical pillar of a market-based economy as they mobilize interme- diate savings, allocate and price risk, absorb external financial shocks, and foster good governance through market-based incentives. The level of financial market development is positively linked to output level and quite likely also to economic growth paths.5 Such effects are crucial for the EUA countries and are likely to be important for various current EU member states as well. Last but not least, to handle aging through prolonged labor market participation, to provide labor market flexibility in a socially acceptable manner, and to contribute to knowledge and skill formation as a major ingredient for economic growth, a pension sys- TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 239 tem that supports lifelong learning is required. Too many pension schemes today are still based on the strict separation of education, work, and retirement leisure. But a modern economy and the need for lifelong learning require a pension scheme in which the mixing of the three activities is encouraged and not impeded--for example, going back to school after years of work, bringing forward (retirement) leisure, or taking up work again after retirement (say, from ages 70 to 72). Such flexibility is discouraged in most current pension schemes. To deal with aging, socioeconomic changes, and globalization, a reform approach is required that moves toward a more actuarial system structure that better links contributions and benefits and includes more individualization to handle professional and family mobility and also some funding to allow more individual decision and choices. The approach must go beyond a parametric adjustment of existing schemes. For most old EU member countries, this contrasts with the adopted reform approach so far, while new EU member countries have shown more inclination to adopt a paradigmatic shift in pension provision,6 yet with stronger differentiation in system design compared with Latin America.7 Reforms in the 1990s and early 2000s in the EU countries were essentially of a paramet- ric nature--with Sweden and partly Italy as the only exceptions. The reform package typ- ically included a combination of the following elements: reduction or elimination of early retirement provisions, an increase in the retirement age or related indirect measures to this effect, reduction in the annual accrual factor, further changes in indexation, and introduc- tion or enhanced support of a funded voluntary pillar. Only a few countries started toward more harmonized national systems (for example, Austria and partially France). Most countries ignored the nonfiscal reform needs except, perhaps, for reasons of political economy.8 While essentially all these reforms move in the right direction, even from a fis- cal point of view more is needed and rapidly.9 The Need for a Better-Coordinated Pension System in an Integrated Europe While there is increasing support for national pension reforms in old and new EU member countries, and despite agreement with some or perhaps all of the arguments advanced above, there is little understanding of and support for a Pan-European approach that should lead to a coordinated pension structure. Pension systems are considered--like other parts of social policy programs--as a national agenda item with little indication that member countries see a necessity for more coordination, and even less harmonization. Astonishingly, neither does the Commission of the European Union, which in many other areas often sees the need for such coordination, or even harmonization, and pushes accordingly. The EU approach of "Open Coordination" of the reform efforts by member countries is viewed as a benchmarking device, not a harmonizing one.10 This section argues that a major impetus for a Pan-European pension reform approach resides in European economic integration, and the objective of common markets for goods, services, and factors of production under a common currency--the euro. This objective has implications for the provision of retirement income: budgetary implications, the need for more labor market flexibility, and the need for enhanced labor supply in an aging population. The concept of a stable common currency in Europe is linked with the Maastricht fiscal criteria to keep the fiscal deficit below 3 percent and public debt below 60 percent of GDP. Although the selection of the criteria may be questioned,11 the objective is sound: to avoid excessive and opportunistic fiscal expansion by some member countries at the detriment 240 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES of the internal and external value of the euro. To comply with the related growth and sta- bility pact, the 12 "Euroland" members engage to achieve a structural budget deficit of zero percent (to allow for fiscal expansion when cyclically needed). But many countries will not be able to achieve a zero budget deficit in a sustainable manner unless the pen- sion system is reformed and the explicit or implicit transfers from the budget are cur- tailed. In Austria, as an extreme example, the pension-related deficit amounts to almost 5 percent of GDP. And all current and future member countries are exposed to enhanced fiscal pressure of population aging in the main public programs--pensions and health-- in addition to the not yet fully grasped expenditure pressure in long-term care programs or infrastructure. Room for budgetary expansion (and contraction) is needed in a common currency area because exchange rate and interest rate policy are lost and few other instruments are avail- able to deal with asymmetric shocks hitting some member states and not others. Given the limited effectiveness of fiscal policy in an integrated economic area resulting from high leakages to other regions or compensating private sector savings, the other main policy instrument has to come into play: labor market flexibility through wage flexibility and migration. Empirical evidence for the United States suggests that although wage adjustment dur- ing regional crises is important, the main adjustment mechanism is migration from (tem- porarily) contracting to expanding regions.12 This can be contrasted with the European experience, in which both wage flexibility and migration have had little importance;13 actually the international and inter-regional mobility in Europe during recent decades has been very low.14 For Europe, both adjustment mechanisms are likely to remain less impor- tant than in the United States because of more rigid labor markets, and cultural and lin- guistic barriers; the last two restrictions also translate into a larger loss of social capital when moving.15 However, both wage adjustment and migration mechanisms need to be strengthened if sluggish adjustments after demand or supply shocks are to be avoided, along with their economic and social consequences. A particular recent drastic example of the consequences of delayed structural adjust- ment and lack of mobility in resource reallocation under a common currency-type arrangement is Argentina. The introduction of the currency board with the national cur- rency pegged to the U.S. dollar was motivated by many episodes of hyperinflation and the expectation that the tight monetary corset would help to push through reforms in the goods and factor markets. But these reforms (including reforms in the labor market) did not come through as expected, which left the country very vulnerable when shocks hit the world economy and neighboring countries. One important mechanism to support a common currency and adjustments after shocks is a pension system that does not lock persons into sectors and countries, but instead supports full labor mobility across professions and states--a requirement that is far from reality in the European Union. In many European countries, different pension rules for public and private sector workers impede mobility between the sectors. Mobility between states exists notionally for public schemes (less in reality), but full portability for corporate and voluntary funded systems is still under construction. As a result, the Euro- pean Union does not have a coordinated--and even less a harmonized--pension system, which characterizes other economically integrated areas under a common currency (such as Australia, Brazil, Canada, Switzerland, and the United States). These federations or con- federations exhibit many differences at state or provincial levels (including income taxes or short-term social benefits), but they have one thing in common--a public retirement income scheme across states. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 241 A third main argument for a more coordinated Pan-European pension system resides in the need for more labor market integration that goes beyond the requested labor market flex- ibility. A strand of international economics suggests that free trade in goods and services or alternatively free capital flows may be sufficient to lead to equalized factor prices and maximize welfare. However, in the real world of externalities and imperfect competition, quite likely the performance of all markets, including the labor market, needs to be improved and integrated more strongly to maximize welfare.16 Full integration of the European labor market requires full portability of pension rights between countries.17 Finally, the long-term external value of the euro is likely to be determined or at least co- determined by the growth expectation of Europe (compared with the United States or other currency areas). Current-period balances or imbalances in flows of goods and ser- vices or even the net-asset positions of countries are increasingly conjectured to lose their importance in determining the relative price of a currency under globalization. Productiv- ity growth can only compensate partially for the effects on GDP growth of projected pop- ulation decline in the EU15 (13 percent between 2000 and 2050, compared with an expected increase in the United States of 50 percent or more; see Holzmann and Muenz 2004)--and higher productivity requires mechanisms to reallocate workers from shrinking to expanding sectors and regions. If falling population and aging are not better compen- sated for through increased labor supply resulting from higher labor market participation, delayed retirement, and increased external migration, the impact on GDP growth will be substantial. The weakness of the euro until recently (compared with the U.S. dollar) may be explained by expectations of the financial markets about the relative growth of these two currency areas. Enhanced labor force participation and delayed retirement, however, require major changes in age management practices in work places and labor markets, as well as appropriately reformed retirement income schemes. Clearly, although a Pan-European pension system would help remove current con- straints on labor mobility, in and by itself it is not sufficient. It would help as it reduces the transaction costs for people wanting to move between member states. These costs can be very high and, in consequence, mobility very low, as suggested by migration research. But uncoordinated pension systems are not the only source of transaction costs. Other national social programs need to be adjusted in order to enhance mobility--most importantly, health care financing, and in particular the private/supplementary health care programs. And there are nonmonetary costs as a result of culture and language barriers. The latter will be gradually reduced as younger people and the more educated population are increasingly more proficient in other European languages or use English as lingua franca. Open borders with more travel, more inter-European marriages, and the emergence of a European identity will also reduce actual or perceived cultural barriers. Potential Structure of Pan-European Pension System and Transition Issues What structure could or should a more coordinated Pan-European pension system have? And if an appropriate steady-state system were to emerge from the discussion, what are the transition issues the approach would encounter? And how could they be solved? This section suggests answers to these questions. Issues of the political economy and how to get there will be addressed in the concluding remarks in the last section. This section starts out by outlining the general and specific main objectives a Pan-European pension system should have, before reviewing which of the main three options fits best. The proposed Pan-European system consists of a mandatory first pillar NDC plan, a (voluntary) funded 242 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES pillar with occupational and individual retirement plans, and a basic (or zero) pillar of social or noncontributory pensions providing minimum income support for the very vul- nerable elderly. All elements are discussed in turn, with main emphasis on the NDC pillar. The proposed structure has highly attractive features against a Pan-European objective, but is also suggested to be an extremely powerful reform option for the many ailing pen- sion schemes in Europe and beyond. Demands on a Reformed and Coordinated Pan-European Pension System What objectives should such a reformed system fulfill? A presentation of these desiderata should allow for a transparent and objective discussion and an easy comparison with alternative reform proposals. Two sets of objectives are suggested: generic objectives that all modern pension systems worldwide should fulfill, and specific objectives that result from the EU background. The generic objectives are the ones developed and proposed by the World Bank in a recent policy position report, and two levels of goals--primary and secondary--are distin- guished.18 The primary goal of a pension system should be to provide adequate, affordable, sus- tainable, and robust old-age income, while seeking to implement welfare optimizing schemes in a manner appropriate to the individual country. · An adequate system is one that provides benefits to the full breadth of the population that are sufficient to alleviate old-age poverty on a country-specific absolute level, in addition to providing a reliable means to smooth lifetime consumption for the vast majority of the population. · An affordable system is one that is within the financing capacity of individuals and the society: one that will not displace other social or economic imperatives or lead to untenable fiscal consequences. · Sustainable refers to the financial soundness of a pension system and its capacity to be maintained over a foreseeable horizon under a broad set of reasonable assumptions. · Robust refers to the capacity to withstand major shocks, including those coming from economic, demographic, and political risks. The secondary goal of mandated pension provisions (and their reform) is to create pos- itive output effects by minimizing negative impacts, such as on labor markets, while lever- aging positive impacts, such as on financial market development. This secondary goal is important since all retirement incomes--whether funded or unfunded--are essentially financed out of the country's output. The centrality of output for pension systems19 for delivering on the primary goals makes it imperative that the design and implementation of pension systems are checked for their economic output level and growth effects. The suggested specific objectives of a Pan-European pension system, to be used as cri- teria for selection and choice, are: mobility, national preferences, solidarity, and feasible transition. · First, the system should allow for easy, most unrestricted mobility between professions, sectors, and regions and also between stages of the life cycle (school, work, and leisure) and family structures. · Second, the system should be consistent with the (European) concept of solidarity, understood as a mechanism of risk sharing among and between generations, redistribution of income from the life-time rich to life-time poor, and open risk coverage. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 243 · Third, the system should allow for national preferences of target levels of (mandated) benefits or contributions, and the redistributive allocation of resources toward the poor or specific groups or activities. · Finally, the proposed future system should involve a feasible system transition from the current national systems for the largest possible number of member countries. Potential Structures of a Pan-European Pension System There are three main options for a future Pan-European pension system that aims to fulfill the objectives set out above: a basic pension plus a mandated fully funded pillar (Bev- eridge for all); an unfunded defined benefit system plus voluntary fully funded pensions (Bismarck for all); and a basic or noncontributory pillar plus an NDC pillar plus a volun- tary (or mandated) funded pillar. The first option--a basic pension in the form of demogrant or means-tested social pen- sion plus a mandated fully funded pillar providing defined contribution (DC) benefits-- would be consistent with all objectives, except most importantly the one on easy transition. Such a system may be structured in such a way as to target all primary and sec- ondary goals, and if well done it may achieve these goals pretty well. Namely, such a sys- tem can ensure mobility, allow for national preferences (for example by country-specific levels of basic pensions and contribution rates for the funded pillar), and can be structured to ensure solidarity: for example, through a central public pension fund that pays one rate of return (hence pooling of risks across individuals) and through explicit budget transfers to individual accounts to deal with low income or periods of unemployment (as in Mex- ico). A main obstacle is (easy) transition. Abstracting from political problems to find con- sensus for such an Anglo-Saxon approach in continental Europe, the main obstacle is fiscal. It is well known that such an approach makes the implicit debt that pension promises constitute explicit, and the level of this implicit debt is in the range of 200­300 percent for most European countries.20 Repayment of such an amount is beyond political and economic reach, and for a broad range of assumptions not Pareto-improving.21 Although a repayment of the debt may not be necessary to achieve the social policy objec- tives, it can be doubted that international markets are willing to live with such an explicit debt level of the European Union without consequences for interest rate and exchange rate of the euro. Under the second option, a future pension system would expand the dominant Bismarckian approach of an unfunded and publicly managed defined benefit (DB) system to the whole European Union. Supported by social pensions and voluntary funded pen- sions, such an approach can also achieve many but not all objectives. Well structured, it can achieve all the primary goals, and very well structured it may even support the sec- ondary goals of a pension scheme. But as experience with such systems throughout the world indicates, it will be difficult to make such structural reforms happen (and agreed at European level). With regard to the specific EU objectives, an inconsistency between the mobility goal and national preferences emerges. For example, with different accrual rates or additions for, say, childcare under another identical DB structure, it would be difficult but not totally impossible to move from one profession or member country to the next, but the administrative efforts to emulate such a mobility would be gigantic while not fully successful. Last but not least, the transition would first require a consensus on a DB struc- ture (and there are many), and a second consensus on complicated rules of transitions. The third option, the proposed structure of a (mandated) first pillar NDC plan, a (vol- untary or mandated, if so desired) funded pillar with occupational and individual retire- ment plans, and a basic pillar of social/noncontributory pensions that provides minimum 244 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES income support for the very vulnerable elderly, is claimed to fulfill all objectives--generic and specific, primary, and secondary.22 Of course, there is room for design and implemen- tation specificities to make a future structure fit very well or less well. The following sub- sections outline the basic structures and design elements to make it fit well. The Crucial (First) Pillar: Non-Financial or Notional Defined Contribution Plan To motivate the choice of NDC as the crucial pillar of a future Pan-European pension sys- tem, this subsection progresses in three parts: outlining the basic structure of an NDC sys- tem, highlighting its capacity to deal with system objectives and reform needs, and presenting the ease of transition for most (but not all) EU member countries.23 BASIC STRUCTURE OF IDEAL NDC24 One main attraction of an NDC system is the simplicity of its basic structure if one follows the rule book: that is, if it is seen as a system that makes the algebraic and economic logic and constraints of an (unfunded) pension system explicit. Simply put, an NDC system consists of an individual account system to which contributions by individuals (and their employers) are earmarked, notional interests paid, and at retirement the accumulated (notional) amount used to determine the level of annuity based on the residual life expectancy (and the notional interest rate). As a result, the system should be quasi-actuar- ially fair at the margin and on average.25 Crucial elements for design and implementation are: · The choice of a notional interest rate consistent with internal rate of return of a PAYG scheme: that is, growth rate of aggregate (covered) wage sum. Per capita rates of wage or GDP growth or contribution revenue will not do the trick if the contribution rate is constant, but the discussion about the (most) appropriate notional interest rate choice is far from over.26 · The choice of remaining life expectancy. Politically determined underestimation (for example, by taking the cross-section life expectancies instead of estimated cohort life expectancies) to deliver higher annuities will also jeopardize financial sustainability. · The indexation of benefits. Although indexation beyond price adjustments is feasi- ble, in principle, keeping benefits constant in real terms allows higher initial benefits. A temporary underindexation compared with a steady state also allows building up a reserve fund.27 · Although a reserve fund is not strictly needed in an NDC system to guarantee bal- ancing the pension budget in every period--that is, to make it fully immune against economic and demographic risks28--it avoids extreme fluctuation in benefit levels.29 Other important basic design elements, which are discussed below, concern the mini- mum eligibility age to own pension and to minimum pension, if any; the introduction of redistributive elements; and transition rules to new NDC benefits. This and other design elements are discussed in more detail in Palmer (2006a; 2006b). DEALING WITH SYSTEM OBJECTIVES AND REFORM NEEDS An NDC pillar (together with a well-designed basic plus voluntary pillar) is able to achieve all reform needs outlined in the second and third sections, and to fulfill all system objectives. The discussion that follows concentrates on a subset for reasons of space and importance: financial sustainability, changing family structure and establishing own pen- sion rights, mobility across professions and across states, and national preferences and sol- idarity. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 245 Achieving financial sustainability, in particular under conditions of an aging popula- tion, is one of the trademarks of an NDC system, albeit it is not fully automatic. As life expectancy increases, individuals receive a lower pension benefit for a given retirement age, for which they can compensate by extending their labor force participation (or addi- tional individual savings). Hence the system encourages behavior that deals with aging in a consistent and balanced manner: namely, splitting the increase in life expectancy between more work and more retirement leisure. Earlier or later retirement for a given age is sanctioned (rewarded) by quasi-actuarial decrements (increments) consistent with a PAYG scheme. But financial stability cannot be achieved automatically in all periods30 without the use of a balancing approach.31 A reserve fund is needed to fund the contribu- tions of relatively large cohorts, as well as external financing to cover possible noncontrib- utory risks. It is also a general buffer that helps stabilization of a number of key variables. Dealing with increasing female labor force participation, changing family structures, and rising divorces is easy under an NDC system, as it allows individualization of pension rights, together with considerations of fairness and efficiency. For example, marriage and separations over the life cycle can be easily handled by splitting the accumulated (notional) amounts (contributions and interests) of the time together. But even if the mar- riage lasts until retirement, one can imagine a splitting of benefits at retirement (as unisex survival probabilities may be applied anyhow). Also survivorship can be handled in an easy manner: for example, widows/ers with very young children receive a generous tran- sitory pension until, say, the children enter school, and the split accumulations from prior marriage help build her (or his) own pension account and eliminate any pension benefit trap. Since in most European countries accumulated financial and physical assets during marriage are split at divorce, it would be inconsistent not to split the accumulated pension rights. Mobility across professions can easily and quickly be established, as an NDC plan allows immediate harmonization of pension schemes with few technical problems. Take civil servants pensions to be integrated into a national NDC pillar. For those already retired, nothing changes. For those with accumulated pension rights, these rights can be estimated with high precision, transformed into a present value, and credited to an indi- vidual (notional) account. The next month (or year) this individual gets credited the uni- fied contributions and notional interests as everybody else. As a result, for those very close to retirement, little change in the pension amount takes place, while for those with only a few years of work record, the new system dominates by far. Quite likely such a reform will need to be accompanied by a review of the overall compensation package of the public sector, leading to changes in earnings profile or, perhaps, introduction of supplementary but funded pensions of DC type. The mobility across EU member countries can also be made very easy under an NDC plan. Albeit the accumulated amounts are only notional, they are very precise and allow an easy aggregation across countries with two main approaches. Under a transfer approach, a worker moving from, say, Germany to France would take his accumulated amount along (that is, the German social security scheme would need to make a cash transfer to the French social security scheme). The pension would be calculated and disbursed in the country when the worker stops his or her activity and applies for a pension. From a national point of view, only the balance for all labor market migrants (to and from the coun- try) need to be transferred, which is likely to be modest. Under the alternative preservation approach, each worker would keep his or her account and continue to receive national notional interests until retirement. Then the individual would receive partial pensions from as many countries as he or she has worked in. The second approach seems more transaction-cost intensive (unless administrated under a Pan-European clearinghouse) and 246 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES may create a problem in case minimum pensions are granted (unless delivered in each country through top-ups and based on residency). Of course, social arbitrage is not excluded under the first approach, as individuals may be tempted to move before retire- ment to a country with high minimum pension, low remaining life expectancy, and low income tax rates. But incentives for social arbitrage will always exist in case of national preferences and different depth of national solidarity across member countries, and NDCs cum social pen- sions/top-ups allow for national preferences. For example, one country may prefer a fru- gal mandated pension for its residents and prescribe a low NDC contribution rate (say 10 percent) and expect more voluntary contributions to well-regulated funded schemes (say also 10 percent). Another country may prefer a high target replacement rate and mandate a higher contribution rate accordingly (say 20 percent), but expect few people to contribute to a funded pillar. Individuals moving between these two countries would not fare too dif- ferently. The NDC approach exhibits national solidarity through its pooled rate of return approach--one single notional interest rate--and the sharing of economic and demo- graphic risks. The second element of solidarity--redistribution--can also be easily intro- duced in NDC systems, but requires direct payments from the budget at the time of granting. For example, low-income workers can be provided a copayment to their contri- bution or for periods of recognized unemployment. The contributions to the NDC system are paid in cash by the unemployment benefits system. DEALING WITH TRANSITION ISSUES ACROSS MEMBER COUNTRIES The previous subsection has already highlighted that a transition across earnings-related and unfunded pension regimes within a country is technically but not necessarily politi- cally easy. The same applies to countries that start from different systems. In the discus- sion that follows, such transition issues are discussed by country groupings. Coordinating among the existing NDC countries. Four EU countries have already intro- duced NDC systems: Italy (1995/96), Latvia (1995/96), Poland (1998/99), and Sweden (1994/99). The first date refers to the year of legislation and the second date to the first year of payment into NDC. Although these countries share the broad system design of NDC, there are major differences in some design and implementation elements.32 For example, the countries use different notional interest rates, different ways to determine the residual life expectancy, or different transition rules to the new system. This raises two general issues: to what extent must or should a Pan-European NDC system have the same system design and implementation features (and hence be fully harmonized, except, say the contribution rate levied); and to what extent must or should the transition rules be harmonized? For example, using different notional interest rates is primarily an issue of financial sus- tainability for the national scheme. Assuming that the choice of the rate of aggregated wage growth provides sustainability but the per capita average wage growth is too high, a country that chooses the latter would need to balance through other means (such as annual benefit indexation) or find additional budgetary resources. A priori there is no rea- son why such national preferences should not be granted. There are more arguments for some harmonization of transition from the old to the new system. For example, Italy and Sweden will only gradually phase in the NDC system over the next decades, while Latvia has moved all workers in one stroke to the new system. If mobility across professions and countries is the main goal of a Pan-European reform, it is the latter approach that is needed--an approach that, however, allows for the expression of national preferences, in particular concerning the generosity of the transition rules at the detriment of financial sustainability. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 247 Transitioning quasi-NDC countries. Two countries have unfunded DB systems that almost mimic NDC systems and hence should be easy to transit: Germany and France. A DB system that uses lifetime income revalued with national wage growth and actuarially determined annuities based on yearly revisions is algebraically similar (but not equiva- lent) to an NDC system.33 In reality, sizable differences do exist,34 which may not prevent a transition toward a common NDC design but would make such a transition not different from that of other earnings-related schemes. Transitioning other Bismarckian systems. The transitioning of the many other current and future EU countries with a typical unfunded and earnings-related social insurance scheme for old age is, in principle, very simple and equivalent to transitioning civil servants bene- fits to NDC (discussed above)--calculate the acquired pension rights and transform them into the present value: that is, a lump sum amount to be credited to the individual account. The alternative approach would be to use past contribution records and past notional interest rates to determine the initial amount. In an actuarially fair scheme, the result would be the same. Under current conditions in a number of countries, the first approach may be cheaper for governments, as it will capitalize on the recent reforms that have reduced the present value of pensions (via increase in retirement age, change in indexa- tion, and so on).35 Hence, for fiscal reasons, a substantive parametric reform prior to a move toward NDC makes sense. This will be the case for Austria, which just did such a parametric reform and which is discussing a move toward NDC/individual accounts. An NDC reform is also under discussion in Hungary and the Czech Republic, and proposed by researchers in countries such as Belgium, Germany, Greece, Portugal, and Spain (see, for example, Vidal-Meliá and Domínguez-Fabián (2006). Transitioning the European outliers. Although Bismarckian-type systems by far domi- nate the European scene by the number of population covered, four main countries have a more Beveridge-type system, for which a transition toward NDC would consti- tute a main policy change. Ireland has a flat-rate contributory and noncontributory sys- tem. The United Kingdom has a flat-rate contributory plus an earnings-related system (SERPS), with opting-out options to private sector arrangements for the latter. Denmark and the Netherlands have universal pensions, which are flat in Denmark, and prorata with regard to residency in the Netherlands (see EPC 2001). The new EU member coun- tries in Central and Eastern Europe have inherited a pension system that is typically earnings-related. This was not changed during the economic transition (except the reforms moving toward a multipillar structure; see annex to Holzmann, MacKellar, and Rutkowski 2003). If a transition/nontransition were to be envisaged, what would be the approach? For a typical universal and basic system plus a quasi-mandated funded scheme, such as in Den- mark, one solution to achieve some coordination with regard to mobility would consist in providing a buy-in option to the universal pension as well as funded scheme by transfers of an accumulated NDC amount, or the reverse when migrating from Denmark. The Funded--Second or Third--Pillar in a Pan-European Pension System With a well-designed Pan-European NDC scheme that allows for national preferences, what is the role of a funded pillar, what structure should it have, and what needs to be done to make it work well? All current and future EU member countries already have funded pillars at different levels of importance and sophistication. These will need some adjustment and coordination to achieve the objectives of a Pan-European pension system (table 11.7).36 Table 11.7. Scope of Funded Pensions in EU and EUA Countries around 2002 Mandated Share of Funded pension as Funded second Contribution covered % of retirement pension assets Country pillar Description rate LF as % income3 as % of GDP Austria no n.a. n.a. n.a. 3.0 2.6 Belgium no n.a. n.a. n.a. 0.5 4.8 Bulgaria2 yes Supplementary mandatory pension funds; not 2% payroll 48.4 Close to nil Close to nil less than 50­100 leva for farmers and 200 leva for self-employed; maximum monthly income 1,000 leva, current contribution 2% but planned increase to 5%; no reserves. Cyprus no Supplementary earnings-related n.a. n.a. Modest Modest contributions/benefits. Voluntary coverage for formerly covered persons and for Cypriots working abroad for Cypriot employers. Employer contributes 6.3% (voluntarily covered 10%), employee 6.3%, and state 4%. Czech R. no n.a. n.a. n.a. Low 3.4 Denmark yes Privately administered defined contribution n.a. 82.0 16.0 21.5 scheme; civil service pension scheme (defined benefit) for public sector employees. Estonia yes Employer contributes 4%, employee 2% to 6% payroll 60.0 Close to nil 0.13 funded system; no ceilings. Pension fund man- agement companies maintain individual accounts and must make quarterly contributions to a guarantee fund. Finland no n.a. n.a. n.a. 38.6 n.a. France no n.a. n.a. n.a. Low 5.6 Germany no n.a. n.a. n.a. 13.0 3.3 Greece no n.a. n.a. n.a. Low 11.9 Hungary yes Contribution to grow to 8% by 2004; employee 6% payroll 45.0 Low 5 contribution ceiling 250% average wage in 2003; no ceilings on employer contribution, maintained as individual accounts, 0.4% of contributions go toward guarantee fund. Ireland no n.a. n.a. n.a. High High Italy no n.a. n.a. n.a. 4.2 3.2 Latvia yes Current contribution 2% but rate expected to 2% payroll 72.0 Close to nil 0.4 increase to 9%; maximum income from which contributions are paid is 18,400 lats. Lithuania no n.a. n.a. n.a. Close to nil Close to nil Luxembourg no n.a. n.a. n.a. Low Low Malta no n.a. n.a. n.a. Low Low Netherlands1 yes Not mandatory but schemes set by industrial n.a. 91.0 19.0 85.6 agreements; 95% of schemes are defined bene- fit; occupational pensions integrated with pub- lic pension schemes. Poland yes DC individual account schemes in which 7.3% of total 70.0 Low 3.0 employees choose the fund. Employees social security contribute half and not less than minimum contribution wage, maximum for employers and employees 250% average wage (annually); guarantee fund is 0.1% pension assets; backed up with state budget guarantee. Portugal no n.a. n.a. n.a. Low 12.0 Romania no Partially legislated then questioned; second pil- 8% payroll 75.0 Close to nil Close to nil lar decided on principle; adoption depends on future fiscal condition. Slovak R. no n.a. n.a. n.a. Close to nil 1.0 Table 11.7. (continued) Mandated Share of Funded pension as Funded second Contribution covered % of retirement pension assets Country pillar Description rate LF as % income3 as % of GDP Slovenia no n.a. n.a. n.a. Close to nil 0.0 Spain no n.a. n.a. n.a. Low 2.1 Sweden yes Premium Pension Authority maintains the 2.5% payroll 100.0 Moderate 32.6 individual accounts of the system; workers choose from several hundred privately managed funds for investment of their capital. United Kingdom yes Mandatory pension component covers defined 17.5%­40% High High 83.7 benefit and defined contribution schemes; earnings-- some components run by state, some by varies with employers and some by financial services com- age panies. Sources: OECD (2000); World Bank (2003c); Luxembourg Income Study (2003); ISSA and INPRS (2003); Blommestein (2000); Whitehouse (2000, 2001); Palmer (2000); Denmark Ministry of Social Affairs (2002); Holzmann et al. (2003); Chlon-Dominczak (2003). ´ ´ Note: n.a. = not applicable. 1. The second pillar in the Netherlands is quasi-mandatory, based on collective labor contracts. Data on pension as a percent of retirement income not available so capital income as % of retirement income has been used. 2. For Bulgaria, the share of covered labor force column gives data on proportion of participants in funded systems as a percent of total contributors. 3. Includes total population, as specific data for age group 65+ is not available. In the qualitative and author-based assessment, "close to nil" refers to > 1%; low to 1%­ 5%, moderate to 5%­15%, and high to < 15% of funded pension income in retirement income of current population. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 251 The role of a funded pillar is essentially fourfold. The first main purpose is consump- tion smoothing beyond NDC benefits. Although an NDC system can provide generous replacement rates if the contribution is sufficiently high, as a mandated, general scheme, it should not do so. A very high mandated contribution rate under an NDC scheme would resemble a labor tax rate with all the known negative social and economic effects, in par- ticular for credit constrained individuals;37 albeit the incidence effects on wage levels seem to be lower if the reciprocity between contributions and benefits is stronger.38 An actuari- ally fair funded pillar allows better consumption smoothing according to individual pref- erences and has less distortionary effects on individual labor supply and savings decisions. The second main purpose is to support retirement flexibility in an aging society. NDC as a quasi-actuarial scheme encourages later retirement with high decrements for early leavers. To compensate for future lower pensions at early age, individuals need to plan to stay longer in the labor market or to save more under a funded pillar. The alternative of voluntary NDC contribution to finance an earlier retirement is possible but must be weighted against the third main purpose--risk diversification. As funded and unfunded pension pillars have a different exposure to economic, demographic, and political risks, and as their rates of return are little correlated, diversifying pension benefits from two dif- ferent pillars is welfare enhancing. It is often claimed that risks will increase in an aging and globalizing world that is subject to technological and many other changes, making risk diversification even more important (see, for example, Bovenberg 2003). Last but not least, funded pillars are important to support Pan-European mobility and beyond. In the proposed more coordinated but not harmonized Pan-European pension system, differences would still exist. Their mobility-reducing effects, however, can be lim- ited with a strong (voluntary or mandated) funded pillar. Furthermore, labor mobility with the rest of the world is also bound to increase, with Europeans working some part of their lives abroad, and migrants from developing countries working part of their lives in Europe. Again, a strong funded pillar that can easily be taken back home would make life for migrant workers, and host and sending countries, so much easier. A number of choices need to be made to achieve a good Pan-European structure of a funded pillar. First, the issue of a mandated or voluntary pillar, a corporate (second), or an individual (third) pillar.39 Mandating the second pillar at the explicit detriment of the first NDC pillar raises the issue of transition costs, and the assessment by many pension econ- omists is likely to be that it is not worth the effort. In addition, it can be argued that the economic rationale for mandating a high replacement rate is decreasing because of reduced myopia of individuals and better financial retirement instruments. What can and should be considered is to transform existing and mandated severance payments, which exist in all EU member states, into funded unemployment benefit cum retirement benefit accounts, as some countries have started to do.40 Hence I would argue that (newly) funded pillars should, in principle, be voluntary and the regulation should allow for both corpo- rate and individual pensions in a well-designed but simple manner. Second, the issue of DB or DC plans emerges. While as individuals we are likely to pre- fer a DB plan best in the form of the final salary-scheme type, economic rationale and recent trends tend to speak in favor of DC schemes. It is the least distortionary scheme with regard to individual labor supply decisions, including retirement, and it provides the required mobility across professions and states. Third, simplicity and transparency of the approach will be of importance: that is, the structure of the retirement products should be simple and there should be at least one set of instruments that is standardized across the European Union. The suggested instru- 252 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES ments are some kind of individual or personal retirement account, as well as some corpo- rate pension account offered by the employer as they exist with a relatively simple struc- ture in, say, the United States and Canada. Complicated structures ŕ la Germany, which try to achieve too many objectives at the same time, should be avoided. The mandated annu- itization of the accumulated retirement saving is not suggested, at least as long as the NDC account allows the financing of a minimum pension. Finally, funded pillars as part of a Pan-European pension scheme also have coordina- tion requirements at the level of regulation, supervision, and taxation that are likely to be difficult to fulfill. At the level of regulation and supervision, the question of mutual recog- nition versus more centralized approaches emerges. At the level of taxation, the issues of consistency of taxation (income versus consumption-type taxation, and in the latter case whether it is back-loaded or front-loaded) and recognition of tax deduction for contribu- tion to funded pillars across Europe emerge. Although progress has been made toward harmonization of tax treatment by EU directives, the launch of new infringement proce- dures against Belgium, France, Italy, Portugal, and Spain, and pushing forward existing cases against Denmark, signal that more needs to be done. The Pension Directive that emerged in 2003 after 10 years of preparation and discussion seemingly needs time to be digested by financial market institutions and multinational enterprises before a judgment can be made.41 The Social Pension Pillar: A Strengthened Social or Noncontributory Pension in EU Member Countries All current and future EU member states have some income provisions for the elderly poor, at least in the form of general social assistance, but increasingly also in the form of a (partially or fully) means-tested social pension, and a few in the form of a universal demogrant (table 11.8). It is strongly suggested that a Pan-European pension system will need to strengthen the social pillar (or zero or noncontributory pillar), which deals with the vulnerable elderly in Europe, for reasons of social objectives and system consistency. The main argument for a strengthened social pension pillar is twofold. First, having under the new structure a quasi-actuarial NDC system as the first pillar and actuarial funded second and third pillars tends to increase the efficiency in the labor market but reduces the redistribution of income toward the poor. Shifting from a nonactuarial to an actuarial system can result in Pareto improvement but will require (keeping or introduc- ing) a minimum benefit.42 Second, income support for the very vulnerable elderly to pre- vent old-age poverty is part of the adequacy objectives of any pension system. A strengthened social pillar can be motivated by the increase in vulnerability of the elderly as aging progresses, and by the solidarity objectives of the European Union. With incom- plete and perhaps falling coverage under earnings-related schemes, one can conjecture that poverty incidence will increase as the increase in life expectancy continues.43 With regard to how such a strengthened social pension pillar should be structured, three main issues emerge: Should there be a minimum pension in the NDC system in addi- tion to a social pension pillar? How is this related to the social pension? And what eligibil- ity criteria and level should be applied? First, there are a few good arguments for a minimum pension under the NDC system. Most importantly, it strengthens incentives for formal labor force participation. However, in order not to contradict the neutrality objec- tive of the NDC structure with regard to the individual retirement decision, eligibility needs to be restricted. For example, while allowing individuals to retire from the age of, say, 60 onward, it may be required to have a minimum accumulated notional amount equivalent to 100+ percent of the minimum pension or else the need to reach the standard Table 11.8. Scope and Form of Social Pensions in EU and EUA Countries around 2002 Percent Social Nationality/ share of assistance residency elderly expenditure Country General Elgibility requirements Benefits (65+)i as % of GDP Comments Austria General General assistance Must be residents, Income-tested 6.7 0.2 Supplements assistance, covers those unable EU nationals or allowance for minimum supplementary to maintain mini- recognized maintains pension level in pensions, mum standard of refugees; some minimum level of all schemes. minimum living and age > 19. provinces require pension. Social pension of 630.92 Older people (above Austrian national- assistance for euros for an indi- retirement age) ity. those without vidual. whose insurance coverage under pensions are below earnings- minimum qualify related pension. for supplements. Belgium All citizens in need, General assistance Minimum pension n.a. 0.7 n.a. General age >18 qualify for for those of 9,253.11 assistance, guar- general assistance. registered; some euros/year for a anteed income for Older people restrictions on for- single person fully old, minimum (women age 60, eigners. insured. Means- pension. men 65) who can- Guaranteed tested allowance of not maintain mini- income for Belgian 7,022.70 euros/year mum standard of or EU citizens plus for a single person. living eligible for residents of 5 guaranteed income years before claim scheme. or 10 years during lifetime. Bulgaria n.a. n.a. Flat rate of 44 leva/ n.a. n.a. n.a. Social pension. month. Table 11.8. (continued) Percent Social Nationality/ share of assistance residency elderly expenditure Country General Elgibility requirements Benefits (65+)i as % of GDP Comments Cyprus Social pension. Those 65+ and not 20 years of Lump sum n.a. n.a. n.a. entitled to pension residency after age payment of 15% of or similar payment 40 or 35 years after total earnings. from other sources. age 18. Social pension Lump sum is 133.63 payment to those pounds/month. aged 68 who do not meet contribution conditions for pension. Czech Republic Minimum n.a. n.a. 2,080 0.2 n.a. n.a. pension. koruny/month. Denmark Noncontributory People with low Residents of Den- Income tested sup- n.a. 1.4 n.a. supplementary pensions rights. mark. EU citizens plement of 4,406 pensions scheme. Payable at age 67. and recognized kroner/month. refugees given temporary help for 3 years until they become residents. Estonia n.a. n.a. n.a. n.a. 2.6 n.a. n.a. Finland Living allowance. Those who have no Residents n.a. n.a. 1.1 n.a. other source of registered by income; minimum municipality. age 18. France General People ineligible French and EU Minimum pension n.a. 2.0 n.a. assistance, bene- for other benefits nationals. calculated at 50%, fits for elderly and age > 25. Bene- not less than plus supplements fits for elderly for 6,307.62 euros/year. to guarantee min- people aged 65+ Coverage for 150 imum income, with low pension quarters. Minimum minimum income or no reduction depends pension. pension. on length of coverage. Germany General Those with insuffi- Residents. Restric- General assistance n.a. 2.3 Includes assistance, basic cient income to tions for non- is means tested. supplementary security benefit. meet needs. Secu- Germans Basic security bene- benefits for old rity benefits for including fit includes age. those 65+ (even if refugees. payment for hous- not eligible for old- ing and health care. age pension) and those 18+ with per- manent reduction in earnings capac- ity (not eligible if held responsible for own situation). Greece Assistance to old Older people aged Citizens who are Minimum pension n.a. 0.1 Benefits to and needy, mini- 65+ without permanent of 360 euros/month older people mum pension, adequate social residents; refugees plus 26.99 euros for without dependent's sup- coverage and those and asylum seek- nonworking wife or medical care plements. in need with no ers with permit to dependent disabled and minimum social security cov- stay. husband and 17.98 pension. Lump erage. euros for each sum paid to child. economically weak. Hungary n.a. n.a. n.a. n.a. n.a. n.a. n.a. Table 11.8. (continued) Percent Social Nationality/ share of assistance residency elderly expenditure Country General Elgibility requirements Benefits (65+)i as % of GDP Comments Ireland Supplementary Older people 66+ Residents. Restric- Up to 134 euros/ 8.7 5.1 n.a. allowance, old with limited tions on refugees week depending on age noncontribu- means; people with and asylum means test plus 88.5 tory pension. exceptional needs. seekers. euros for each adult dependent and euros 16.8 for each child. Italy Social assistance, All living indepen- Residence in Minimum pension n.a. 1.3 Social allowance social pension, dently eligible for municipality, legal is 392.69 euros/ scheme replaced social allowance. assistance. Social residents of Italy, month. Social social pension in pension for those EU citizens. allowance is 1996. No new 65+. Older people 3,775.83 euros/year. claimants for not eligible for For those 70+ with social pension social pension income < 6,714 since 1996. receive social euros/year may allowance receive up to 516.46 (minimum euros/month. pension). Latvia Minimum n.a. n.a. 30 lats/month. n.a. n.a. n.a. pension. Lithuania Basic pension. n.a. n.a. 110% of poverty n.a. n.a. n.a. level. Luxembourg Income support All above age 30; Resident for 10 out n.a. n.a. 0.5 n.a. benefit, minimum minimum pension of last 20 years and pension. coverage for at registered with least 20 years. local authority. Malta n.a. n.a. n.a. n.a. n.a. n.a. n.a. Netherlands General All above age 18. Residents. Non- Supplementary n.a. 2.2 n.a. assistance, citizens covered allowance reduced income tested only if special by 2% for each supplementary agreements exist. unexcused year of allowance for old. noncontribution. Poland Minimum n.a. n.a. Minimum pension n.a. n.a. n.a. pension. is 530.26 zlotys/ month. Portugal Guaranteed mini- Guaranteed income Nationals and EU Social pension is n.a. 0.5 n.a. mum income, for those in citizens; six month 138.27 euros/ social pension, economic need. residency required month. social supplement Social pension for for stateless to pension. older people (65+) persons and not covered by any refugees. other social security scheme. Social supplement to pensioners whose contributions insuf- ficient to generate minimum pension. Romania n.a. n.a. n.a. n.a. n.a. n.a. n.a. Slovak Republic Minimum n.a. n.a. 550 koruny/month n.a. n.a. n.a. pension. Slovenia n.a. n.a. n.a. n.a. n.a. n.a. n.a. Spain Minimum income Minimum income One year Minimum pension 1.6 1.1 n.a. scheme, social scheme for low residency require- is 385.50 euros/ pension. income working ment for minimum month (for those age households; income; 10 years aged 65), reduced social pension for residency includ- minimum pension those 65+ without ing 2 years preced- for those <65. insurance pension. ing claim for social pension. Table 11.8. (continued) Percent Social Nationality/ share of assistance residency elderly expenditure Country General Elgibility requirements Benefits (65+)i as % of GDP Comments Sweden Social welfare People with no Residents. n.a. n.a. 1.2 n.a. allowance, guar- other means of sup- antee pension. port; also serves as a supplement to people claiming social security benefits. United Kingdom Income support All excluding Residents only Depends on age, n.a. 4.2 Noncontri- benefit. unemployed. unless under EU income, and butory means- Income must be regulations or circumstances; tested social below certain level. refugee; 92.15 pounds/ assistance. Not payable if sav- restrictions apply week minus other ings are over 8,000 depending on income for a single pounds or if work- immigration person. ing more than 16 status. hours per week. Sources: Gillion et al. (2000); United States Social Security Adminstration (2002); ISSA (2003). Note: n.a. = not applicable. 1. Social assistance recipients as a proportion of total aged population. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 259 retirement age of, say, 67 (which is increased with a rise in life expectancy). Second, coor- dinating a minimum NDC pension with a social pension with regard to labor market incentives requires different eligibility criteria (such as some kind of means- or affluence- testing), different amounts, different eligibility ages, or some combination of these. Finally, eligibility for a social pension may have to be conditioned on higher ages (say 70 onward), but means-testing may be kept light, for example in the form of affluence-testing, which excludes people having access to other pension provisions and financial assets. How much national preferences such a social pension pillar would be able to exhibit without exces- sively inhibiting the incentive structure of a proposed Pan-European pension system is open for discussion and requires more research. Concluding Remarks This chapter examined why a more coordinated Pan-European pension system is needed and which potential structure could achieve this best. Both social and economic needs call for pension reform that is more radical and that cuts across member states. The economic needs are closely linked with the common economic area and currency. The suggested structure for the current and future EU member states is a multipillar system, with a NDC system at its core and supplementary funded pensions and social pensions at its wings. Such an approach would fulfill all generic and EU-specific demands on a Pan-European pension system, including the room for national preferences. Besides the why of a Pan-European approach and which structure it may have, what remains to be sketched is how such a system reform could come about. One could imagine three main avenues. The first would be an approach initiated and led by the EU Commission. This is possible but not likely. First, there is no intention by the member states to empower the Commission with such a reform request. Social policy continues to be seen as a national agenda item subject to the subsidiarity principle and hence not open for "centralization" by the Com- mission. Second, there are no visible efforts by the Commission to take such a lead, as the necessity for a more rapid and comprehensive reform does not seem to be seen. Last but not least, the recently introduced method of open coordination as a peer review process to accelerate reforms in the member countries has its merits, but is unlikely to lead to rapid national reforms--and even less to create a Pan-European reform vision. The second approach would be a competitive approach across EU countries. One of the existing or reformed pension systems would gradually be adopted by other countries as they see advantages with regard to social and economic policy goals. This is also possible, and a bit more likely, but not sufficiently rapid. Even if carried out, the outcome might be subopti- mal. First, the advantages of reformed systems emerge and get documented only with lapses of time, which may be measured in decades--and this may prove too late. Second, imitation of system reforms are and will be taking place (for example the inspiration of the Polish reform by the Latvian NDC reform, which in turn was inspired by the Swedish reform, or the possible introduction of individual accounts in Austria and Hungary, inspired by those reforms in the north). But imitation by other countries is likely to be restricted. Third, even if all countries were to follow a lead example under competitive pressure, this may not ensure sufficient consistency of approaches across countries to pro- vide the needed mobility of the workforce in Europe. Last but not least, and "to the extent that social policy is meant to redress market failures or to implement solidarity transfers, competition among systems will not lead to efficient outcomes when the elements of the relevant equation span the borders of policymaking constituency."44 By definition, collec- 260 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES tive action is needed to eliminate inefficient or unfair economic interactions; hence one can argue that bringing back competition at the inter-constituency level defeats both pur- poses.45 The third possibility is a cross-country approach led by governments. Issues of pension reform have started to be addressed by government officials, such as the Economic Policy Committee (EPC) of the European Union, which represents high-level officials from min- istries of finance and economy of EU member countries.46 EPC has, so far, been largely concerned with the fiscal consequences of aging. These concerns may be enhanced by the broader stability issues, including the need for cross-European labor mobility. To foster the points for a better-coordinated Pan-European pension system is quite likely the tasks of academics and research institutions, examined and supported by the EPC or similar core groups, and at some moment in the future espoused by a charismatic European politician as reform champion. Perhaps this will happen after the first main asymmetric shock hits Euroland. Notes 1. This and the next section draw partly on Holzmann, MacKellar, and Rutkowski (2003). 2. For a similar list of nondemographic arguments for pension reform, see Bovenberg (2003). 3. Other projections by academics and national research institutes are typically less opti- mistic and predict a much larger increase in expenditure under current service scenarios. See, for example, Rother, Catenaro, and Schwab (2003). 4. See Karacadag, Sundararajan, and Elliot (2003). 5. See Beck, Levine, and Loayza (2000) and Levine (2003). 6. See Holzmann, MacKellar, and Rutkowski (2003). 7. See Mueller (2003). 8. See Natali and Rodes (2003). 9. To deal with the fiscal issues resulting from aging, various recent reforms propose adjustments in annual pension indexation. For example, the recent Rürup Commission Report for Germany led to adjusting pensions in line with the shifts in the ratio of contrib- utors to retirees, and the recent Austrian reform envisages capping indexation by the amount the median voter receives. Balancing the fiscal accounts with reduced indexation instead of a lower initial pensions and price indexation thereafter is questionable for three main reasons: First, it introduces a high level of uncertainty for individuals, as the future real pension level cannot be determined, but once it is known the capacity to react may be nil. Second, in view of the unsettled issue of financing long-term care for the elderly, the financial needs of the elderly may increase but not be reduced. Last but not least, the reform is not credible, as politicians may not be able to withstand future pressures for changes in indexation. 10. See Holzmann, MacKellar, and Rutkowksi (2003). 11. See Holzmann, Hervé, and Demmel (1996). 12. See Blanchard and Katz (1992). 13. See Decressin and Fatŕs (1993). 14. See Braunerhjelm et al. (2000). 15. See Esping-Andersen (2001). 16. See Nicoletti et al. (2001). 17. On the recent debate about the need to harmonize or not to harmonize labor market policies in euro countries, see Calmfors (1998). TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 261 18. See Holzmann and Hinz (2005). 19. See Barr (2000). 20. There are various estimates for the implicit debt of European pension systems (see Holzmann, Palacios, and Zviniene 2001), but a simple rule of thumb may be sufficient, according to which the level of implicit debt is roughly 20 to 30 times steady-state annual pension expenditure. The average level of EU spending is more than 10 percent of GDP. 21. See Lindbeck and Perrson (2003). 22. Few other papers so far outline the basic structure of a more coordinated European social policy, even less a pension system. One recent exception is Bertola et al. (2001), which proposes contingent insurance provisions with three core elements: a minimum contribution rate, a close contribution-benefit link, and no penalization when moving. 23. This study is not the first one that proposes an NDC-type structure for a Pan-Euro- pean pension system. The idea has popped up in various papers and presentations (including by the author) and references include Feldstein (2001) and Gora (2003). Yet this study provides, perhaps, the most comprehensive treatment so far. NDCs for low- and middle-income countries have found very little attention. For a first exploratory paper, see Lindeman, Robalino, and Rutkowski (2006). 24. What constitutes an ideal NDC system and how it fares compared with other bene- fit options (such as nonfinancial defined benefit or financial defined contribution schemes), or how it performs in reality and under political stress, is still very much open to discussion and constitute the very reason for the Sandhamn conference of September 28­30, 2003. For an attempt at defining an ideal NDC system, see Palmer (2006a). 25. The applied discount rate is the rate of aggregate wage growth that is below the (risk-adjusted) interest rate in a dynamically efficient economy. The latter applies to a fully funded DC system, which is considered actuarially fair. Unfunded DC systems--that is, NDC systems--come close but are only quasi-actuarial. 26. See Palmer (2006b) and Settergren and Mikula (2006). 27. The quasi-actuarially fair annuity is determined by remaining life expectancy and notional interest rate. If productivity growth is above (negative) population/labor force growth, the growth rate of aggregate wages is still positive. Hence keeping pension bene- fits constant instead of indexing with positive notional interest rate provides a little sur- plus for reserve building, and additional indexation once a steady-state reserve fund is reached. 28. See Palmer (2006a). 29. See Knell (2004). 30. See Valdés-Prieto (2000). 31. See Settergren and Mikula (2006). 32. See Palmer (2006b). 33. See Valdés-Prieto (2000) and Settergren and Mikula (2006). 34. See Legros (2006). 35. The second (bottom-up) approach may be cheaper for countries that increased con- tribution rates from low levels and have not undertaken a benefit-cutting reform. 36. For details on supplementary and complementary funded pension arrangements in Europe and beyond, see ISSA (2003a; 2003b). 37. See Lindbeck and Persson (2003). 38. See Ooghe, Schokkaert, and Flechet (2003). 39. Please watch out: In the European terminology second pillar refers to corporate pen- sions (whether mandated or voluntary) and third pillar to individual pensions (whether mandated or voluntary). In the Anglo-Saxon terminology (and beyond) used by the World Bank, the second pillar refers to mandated and funded pensions (whether corporate or indi- 262 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES vidual), and the third pillar to voluntary and funded provisions (whether corporate or indi- vidual). In this chapter the European terminology is used. 40. On this topic of severance payments and their reform, a conference was held in Lax- enburg, near Vienna, on November 7 and 8, 2003. The conference was jointly organized by the World Bank, Washington, DC, and the Ludwig Boltzmann Institute for Economic Analysis, Vienna, and was hosted by the International Institute for Applied System Analy- ses, Luxembourg. For the many interesting papers visit www.worldbank.org/SP or http://members.vienna.at/libecon/boltzanalyse. The conference volume is scheduled to be published as Holzmann and Vodopivec (2005/6). 41. See IPE (2003). 42. See Lindbeck and Persson (2003). 43. Data for European OECD countries suggest that while poverty incidence tends to be the highest among those in the 65+ age group, the poverty incidence in this group fell most markedly between the mid-1980s and 1990s (and for the Czech Republic and Hungary, in the early to late 1990s). See Förster (2003). 44. See Bertola et al. (2001). 45. See Sinn (2003). 46. See, for example, EPC (2001). References Americans for Divorce Reform, Inc. 2003. Divorce Statistics 2002. http://www.divorcere- form.org Barr, N. 2000. "Reforming Pensions: Myths, Truths, and Policy Choices." IMF Working Paper WP/00/139, International Monetary Fund, Washington, DC. Beck, T., R. Levine, and N. Loayza. 2000. "Finance and the Sources of Growth." Journal of Financial Economics 58 (1­2): 261­300. Bertola, G., J. Jimeno, R. Marimon, and C. Pissarides. 2001. "EU Welfare Systems and Labor Markets: Diverse in the Past, Integrated in the Future?" In Welfare and Employ- ment in a United Europe, ed. G. Bertola, T. Boeri, and G. Nicoletti, 23­122. Cambridge, Mass. and London: MIT Press. Blanchard, J. O., and L. Katz. 1992. "Regional Evolutions." Brookings Papers on Economic Activity 1: 1­75. Blommestein, H. 2000. "Ageing, Pension Reform, and Financial Market Implications in the OECD Area." Center for Research on Pensions and Welfare Policies. Börsch-Supan, A. 1999. "Gesetzliche Alterssicherung Reformerfahrungen im Ausland. Ein systematischer Vergleich aus sechs Ländern." Deutsches Institut fur Altersvorsorge GmbH, Cologne. Bovenberg, A. L. 2003. "Financing Retirement in the European Union." International Tax and Public Finance 10: 713­34. Braunerhjelm, P., R. Faini, V. Norman, F. Ruane, and P. Seabright. 2000. Integration and Regions of Europe: How the Right Policies Can Prevent Polarization. London: Centre for Economic Policy Research. Calmfors, L. 1998. "Macroeconomic Policy, Wage Setting and Employment: What Differ- ence Does the EMU Make?" Institute for International Economic Studies, Seminar Paper 657, Stockholm. Chlon-Dominczak, A. 2003. "Evaluation of Reform Experiences in Eastern Europe." In ´ ´ International Federation of Pension Fund Administrators, ed., Pension Reforms: Results and Challenges, 145­237. Santiago, Chile. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 263 Council of Europe. 2001. Recent Demographic Developments in Europe 2001. Strasbourg: Council of Europe. Decressin, J., and A. Fatŕs. 1993. "Regional Labor Market Dynamics in Europe and Impli- cations for EMU." INSEAD, Paris. Denmark Ministry of Social Affairs. 2002. "The Danish Social Policy." http://www.sm.dk. EPC (Economic Policy Committee). 2001. "Budgetary Challenges Posed by Ageing Popu- lations." Brussels, http://europa.eu.int/comm/economy_finance/epc_en.htm. ------. 2001. "Budgetary Challenges Posed by Ageing Populations: The Impact of Public Spending on Pensions, Health and Long-Term Care for the Elderly and Possible Indi- cators of Long-Term Financial Sustainability of Public Finances." EPC/ECFIN/ 655/01-EN final (October 24, 2001). European Union, Brussels. ------. 2003. "Key Structural Challenges in the Acceding Countries: The Integration of the Acceding Countries into the Community's Economic Policy Coordination Process." EPC/ECFIN/114/03 final (April 29, 2003). European Union, Brussels. Esping-Andersen, G. 2001. "Comments." In Welfare and Employment in a United Europe, ed. G. Bertola, T. Boeri, and G. Nicoletti, 127­43. Cambridge, Mass., and London: MIT Press. EU (European Union). 2003. Eurostat Yearbook 2003. Luxembourg: European Union. http://europa.eu.int/comm/eurostat. Feldstein, M. 2001. "The Future of Social Security Pensions in Europe." NBER Working Paper Series WP 8487 (September), National Bureau of Economic Research, Harvard, Cambridge, Mass. Förster, M. 2003. "Income Inequalities, Poverty and Effects of Social Transfer Policies in Traditional OECD Countries and Central Eastern Europe: Patterns, Trends and Dri- ving Forces in the 1990s." Ph.D. thesis, chapter 4.3.3, University of Ličge. Gesellschaft für Versicherungswissen-schaft und -gestaltung e.V. 2003. "Proceedings of the 2002 Workshop on the Social Protection Systems of the Candidate Countries." Brus- sels. http://www.gvg-koeln.de/deutsch/projekte.html. Gillion, C., J. Turner, C. Bailey, and D. Latulippe, ed. 2000. Social Security Pensions: Develop- ment and Reform. Geneva: International Labour Organization. Gora, M. 2003. "The Quest for Modern Pension System." In Structural Changes in Europe, ed. G. Tumpel-Gugerell and P. Mooslechner. United Kingdom: Edward Elgar. Holzmann, R. 2003. "A Provocative Note on Coverage in Public Pension Schemes." In The Three Pillars of Wisdom? A Reader on Globalization, World Bank Pension Models and Welfare Society, ed. A. Tausch, 85­99. Happaugue, NY: Nova Science. Holzmann, R., Y. Hervé, and R. Demmel. 1996. "The Maastricht Fiscal Criteria: Required but Ineffective?" Empirica 23 (1): 25­58. Holzmann, R., and R. Hinz. 2005. Old-Age Income Support in the 21st Century: An Interna- tional Perspective on Pension Systems and Reform. Washington, DC: World Bank. Holzmann, R., L. MacKellar, and M. Rutkowski. 2003. "Accelerating the European Reform Agenda: Need, Progress, and Conceptual Underpinnings." In Pension Reform in Europe: Process and Progress, Directions in Development Series, ed. R. Holzmann, M. Orenstein, and M. Rutkowski, 1­45. Washington, DC: World Bank. Holzmann, R., and R. Münz. 2004. "Challenges and Opportunities of International Migra- tion for the EU, its Member States, Neighboring Countries and Regions: A Policy Note." Institute for Future Studies, Stockholm. Holzmann, R., R. Palacios, and A. Zviniene. 2001. "Reporting the Implicit Pension Debt in Middle and Low Income Countries." Journal of Pension Management 6 (4): 355­384. Holzmann, R., and M. Vodopivec, ed. 2005/6. Improving Severance Pay: An International Perspective. Washington, DC: World Bank. 264 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES ILO (International Labour Organization). 2003. Laborsta: Economically Active Population Estimates and Projections 1950­2010. Geneva. http://laborsta.ilo.org. IPE (Investment & Pension Europe). 2003. "European Directive Debate." September Issue, Supplement. ISSA (International Social Security Association). 2003a. Complementary and Private Pensions Throughout the World 2003. Geneva: ISSA. ------. 2003b. "Trends in Social Security." International Social Security Association, Geneva. ISSA and INPRS (International Social Security Association, International Network of Pen- sion Regulators and Supervisors). 2003. Complementary and Private Pensions 2003. Geneva: ISSA and INPRS. Karacadag, C., V. Sundararajan, and J. Elliot. 2003. "Managing Risk in Financial Market Development: The Role of Sequencing." IMF Working Paper No. 03/116, International Monetary Fund, Washington, DC. Knell, M. 2004. "On the Design of Sustainable and Fair PAYG Pension Systems when Cohort Sizes Change." Vienna: Austrian National Bank. Legros, F. 2006. "NDCs: A Comparison of the French and German Point Systems." In Pen- sion Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 10. Washington, DC: World Bank. Levine, R. 2003. "Finance and Growth--Theory, Evidence, and Mechanism." University of Minnesota and National Bureau of Economic Research, Harvard, Cambridge, Mass. Lindbeck, A., and M. Persson. 2003. "The Gains from Pension Reform." Journal of Economic Literature XLI (March): 74­112. Lindeman, D., D. Robalino, and M. Rutkowski. 2006. "NDC Pension Schemes in Middle- and Low-Income Countries." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 12. Wash- ington, DC: World Bank. Luxembourg Income Study. 2003. Project Data Base, Luxembourg, http://www .lisproject.org/techdoc.htm. Mueller, K. 2003. Privatising Old-Age Security--Latin America and Eastern Europe Compared. Cheltenham, UK and Northampton, Mass.: Edward Elgar. Natali, D., and M. Rodes. 2003. "The `New Politics' of the Bismarckian Welfare State: Pen- sion Reforms in Continental Europe." European University Institute, Florence. Nicoletti, G., R. Haffner, S. Nickell, S. Scarpetta, and G. Zoega. 2001."European Integra- tion, Liberalization, and Labor Market Performance." In G. Bertola, Welfare and Employment in a United Europe, ed. T. Boeri and G. Nicoletti, 147­235, Cambridge, Mass. and London: MIT Press. Ooghe, E., E. Schokkaert, and J. Flechet. 2003. "The Incidence of Social Security Contribu- tions: An Empirical Analysis." Empirica 30 (2): 81­106. OECD (Organisation for Economic Co-operation and Development). 2000. Institutional Investors Statistical Yearbook: 2000 Edition. Paris: OECD. ------. 2002. "Fiscal Implications of Ageing: Projections of Age-Related Spending." Eco- nomics Department Working Paper 305, OECD, Paris. ------. 2003. Transforming Disability into Ability: Policies to Promote Work and Income Security for Disabled People. Paris: OECD. ------. 2003. OECD Statistics Portal, Paris, http://www.oecd.org/statsportal. Palacios, R., and M. Pallares-Miralles. 2000. "International Patterns of Pension Provision." World Bank, Washington DC, with updates. Palmer, E. 2000. "The Swedish Pension Reform Model: Framework and Issues." Social Pro- tection Discussion Paper 0012, World Bank, Washington, DC. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 265 ------. 2006a. "What Is NDC?" In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 2. Wash- ington, DC: World Bank. ------. 2006b. "Conversion to NDCs--Issues and Models." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 9. Washington, DC: World Bank. Rother, P. C., M. Catenaro, and G. Schwab. 2003. "Aging and Pensions in the Euro Area-- Survey and Projection Results." Social Protection Discussion Paper 0307, World Bank, Washington, DC. Settergren, O., and B. D. Mikula. 2006. "The Rate of Return of Pay-As-You-Go Pension Sys- tems: A More Exact Loan-Consumption Model of Interest." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 7. Washington, DC: World Bank. Sinn, H.-W. 2003. The New Systems Competition. Malden, Mass.: Blackwell. Uebelmesser, S. 2003. "Harmonization of Old-Age Security within the European Union." CESifo Working Paper No. 1108, CESifo (Center for Economic Studies & Institut fuer Wirtschaftsforschung), Munich. United Kingdom Office of National Statistics. 2001. "Marriage and Divorce Rates: EU Comparison 2001." http://www.statistics.gov.uk/STATBASE. United Nations. 2001. UN Demographic Yearbook 2001. New York: United Nations. ------. 2002. World Population Prospects: The 2002 Revision. New York: United Nations. United States Social Security Administration. 2002. "Social Security Programs throughout the World--Europe 2002." http://www.ssa.gov. Valdés-Prieto, S. 2000. "The Financial Stability of Notional Account Pensions." Scandina- vian Journal of Economics 102 (3): 395­417. Vidal-Meliá, C., and I. Domínguez-Fabián. 2006. "The Spanish Pension System: Issues of Introducing NDCs." In Pension Reform: Issues and Prospects for Non-Financial Defined Contribution (NDC) Schemes, ed. R. Holzmann and E. Palmer, chapter 23. Washington, DC: World Bank. Whitehouse, E. 2000. "Pension Reform, Financial Literacy and Public Information: A Case Study of the United Kingdom." Social Protection Discussion Paper 0004, The World Bank, Washington, DC. ------. 2001. "Pension Systems in 15 Countries Compared: The Value of Entitlements." Centre for Pensions and Superannuation, London. Williamson, J., and M. Williams. 2003. "The Notional Defined Contribution Model: An Assessment of Strengths and Limitations of a New Approach to Old Age Security." World Paper 2003-18, Center for Retirement Research at Boston College, Boston. World Bank. 2003a. "The Labor Market Performance of ECA Countries--Recent Develop- ments." HDNSP Labor Market Unit, World Bank, Washington, DC. ------. 2003b. World Bank Labor Markets Dataset, Washington, DC. ------. 2003c. World Bank Pensions Dataset, Washington, DC. A Magic All-European Pension Reform Formula: Selective Comments Bernd Marin* ROBERT HOLZMANN HAS PROVIDED AN INTERESTING AND INNOVATIVE APPROACH to rapid and comprehensive pension reform in Europe. He shifts the debate beyond the conven- tional focus on fiscal affordability at the national level toward broader economic and social adjustment needs. And he proposes NDC as an "ideal" approach not just for dealing with a great variety of reform requirements but also for inducing pension harmonization across Europe while allowing for continuous country-specific preferences "and to lead to a polit- ical reform movement" toward NDC. This is a strong claim and he puts forward strong and sophisticated arguments in favor, some but not all of them convincing to me. The rationale for much wider domestic reform needs throughout Europe as well as for a move toward a more coordinated Pan-European pension reform are more persuasive than the proposals for its potential structure and transition strategy. While I share the central claim of the paper, which is new and sound, this occasionally is done so for other reasons--start- ing from other working hypotheses and then coming to other policy conclusions--than the ones put forward by the author, as in the case of atypical employment. At times, addi- tional or other empirical evidence is suggested to strengthen the points, as with exempli- fying pension barriers to mobility between the public and private sector. While NDC may be generally close to an "ideal" pension framework, whether the "ideal" NDC as proposed in the Holzmann design is truly "ideal" is still to be explored. Partly, parameters are not yet fully specified, partly risk reinforcement may be more probable than the risk diversifi- cation claimed for the Holzmann design of combining pillars. Furthermore, the guaran- teed minimum social pension should rather not be conceptualized as a "zero pillar." Crucial dimensions such as disability pensions may be missed. One could rather underline the primacy of political and not just economic desirability, feasibility, and sustainability of NDC. And finally one would stress some of its underestimated comparative advantages, such as its functions of fairness standard, anticorruption device and, as a core component of any pension constitution, its differentiation of welfare trigger, as well as its superior risk management (Gora and Palmer 2003). Robert Holzmann foresees an increased need, rather than a diminishing one, for rapid and comprehensive pension reform in both current European Union member-states and future accession countries, due to worsening budgetary pressures, socioeconomic changes, and the impact of globalization, all related to societal aging. But expenditure lev- els are less a reflection of population age structures and aging dynamics than the * Bernd Marin is executive director of the European Centre for Social Welfare Policy and Research. I wish to thank Michael Fuchs for skillful and valuable support, Silvia Fässler for good graphical assistance, and Robert Holzmann for critical rejoinders to my comments. All weaknesses and good suggestions untapped or ideas not yet elaborated are, of course, my responsibility. 266 TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 267 public/private mix of provision, benefit generosity, and the actual retirement age, which is typically low due to disincentives to work. In the future, further increases in longevity, which are likely, together with even modest rises in fertility, which are still below the replacement level, will continue to make for rapid population aging and corresponding increasing old-age dependency ratios. Even if system dependency ratios deteriorate less than old-age dependency ratios, due to reforms and to increased labor force participation of women and middle-aged workers, pension expenditures will rise till around 2040. With reforms, the expenditure increase may be "only" 30 percent, as against the demographi- cally required 70 percent, or a "rough doubling" of expenditures "in a no-reform sce- nario." Under all circumstances, "a further major increase in pension expenditure can be prevented only if major reforms take place." While nobody may object to this reasoning and its conclusions, additional forms of empirical evidence supplied to support the cases in point are suggested. Holzmann offers data on public pension expenditure in terms of percent of GDP and projections of old-age dependency till 2050, depicting the great--and increasing--variety within European Union and accession countries. He takes the design flaws of most existing pension schemes for granted: not in need of further documentation. But as the main goal of his chapter is to argue in favor of a reformed NDC system to replace current defined benefit (DB) systems and to push toward a coordinated pension system in Europe, I would like to strengthen his case by providing supplementary calculations. Deeds Defying Words--Reform Flaws Visible through NDC The approximation offered in this discussion is a comparison of four smaller European countries. Two (Sweden and Poland) adopted NDC schemes when introducing benefit cuts to discourage early labor market exit; two (Austria and the Czech Republic) did not. One big country (Germany) with a reformed DB system holds an intermediate position. The evidence supplied shows quite clearly that all of them have moved in the right direc- tion of increased benefit reductions for pre-retirement--but only NDC pension systems (by using the wage growth rate as notional interest rate) are "quasi-actuarially fair" and neutral to individual preferences. Existing DB arrangements, in contrast, actually continue to strongly subsidize early exit and to heavily penalize working longer (see figure 11.2). In this way, the governmental rhetoric of praising delayed retirement is undercut by counter-productive measures in the political economy. Words are defied by deeds: only fools willing to sacrifice a painful amount of money--or unable to make the elementary calculation of tangible advantages--will not escape the Austrian and Czech labor market at their earliest possible opportunity and convenience. Figure 11.3 shows that benefit rules in these non-NDC systems are detrimental to declared public objectives. The distortion factor is at least 1:2 in Austria and the Czech Republic, and between a third and a half in Germany. In the smaller DB countries, regular voluntary early retirees (that is, not those persons experiencing ill health, disability, unemployment, or another disadvantage) are exempt from more than half of the actuarial losses to be incurred by them (in special cate- gories up to 80 percent). Those working longer than expected, on the other hand, would lose much more than half and up to around six-sevenths (Czech Republic) of the savings generated to the insurance collective. No surprise, therefore, to find hardly anybody working up until the legal retirement age in such a country and to find less than 3 percent of the working population working to the official working age of 65. Without NDC standards of actuarial neutrality and fairness, neither the amount of distortions and of hidden taxes for younger working generations, nor the fact, so puzzling to policy-makers, as to why their well-intended and supposedly 268 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 11.2. Actuarial Fairness (the Swedish and Polish NDC System) vs. Subsidizing Early Exit and Penalizing Working Longer (the Austrian and Czech DB Systems), as compared to the German DB Intermediate, 2003 160 150 Polish NDC 140 130 120 Czech prop. NDC 110 German DB 100 percent 90 Austrian DB 80 70 Czech DB Swedish NDC 60 50 60 61 62 63 64 65 66 67 68 69 70 age Source: Author. "harsh" reform measures generate opposite effects than the ones intended, could be explained. Recall the reasons for selecting those four smaller countries for paired compar- isons. The Czech Republic has the single highest aging dynamics among the future EU-25 countries until 2050. Austria had confronted the single highest pension expenditures worldwide already by 2003 and "as an extreme example, the pension-related deficit amounts to almost 5 percent of GDP." Sweden ages more than a decade earlier than Aus- tria, whereas Poland and the Czech Republic age almost two decades later--but much more rapidly and drastically than Austria and Sweden. The four countries, therefore, lend themselves to several paired comparisons. Although all those overall tendencies can be seen from the tables provided by Robert Holzmann, I have transformed data on population aging until 2050 into figures displaying the time dimension in a comparative manner. They depict "l'Europe a plusieurs vitesses" of aging and reform needs, lags, and peaks. Which country will reach its peak of collective aging when, and how does this compare to European averages and sub-regional aging patterns? (See fig- ures 11.4a, 11.4b, and 11.4c.) Which are the avant-garde countries and which are the laggards hit latest by developments and therefore able to learn from the best, brightest and fastest? Reform Needs Beyond Demography and Budgets But "even if the budgetary and demographically induced pressures did not exist," Robert Holzmann reasons, "there still would be a major need for most European countries to reform their public pension systems to better align them with socioeconomic changes." Holzmann cites three such changes--"increasing female labor force participation; high divorce rates and changing family structures; and the rise in atypical employment"--and TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 269 Figure 11.3. Swedish NDC Actuarial Neutrality vs. Austrian DB, Amount of Labor-depressing Adverse Redistribution, 2003 160 157 150 143 140 130 130 120 119 121 109 116.8 110 112.6 100 108.4 104.2 percent100 95.8 90 91.6 87.4 92 80 85 84 70 78 72 60 50 61 62 63 64 65 66 67 68 69 70 age Source: Marin (2005); author's estimates. Note: The line across the bars indicates the steepness of the bonus/malus function and its deviation from actuar- ial neutrality. The light gray difference indicates the amount of subsidies for early retirement, which, perversely, grows with ever earlier retirement in Austria (e.g. 15 percent cuts in Austria compared to 28 percent in Sweden when retiring at age 61). The black color shows the amount of taxing away collective benefits from late retirement (e.g. 12.6 percent instead of 30 percent increments when retiring at age 68, or 21 percent instead of 57 percent higher monthly payments when retiring at age 70). Thus, socially desirable behavior of retiring later is sanctioned or punished in Austria. Harmful early exit behavior, which is strongly supported in Austria but not in Sweden, has led to a striking difference in retirement behavior in the two countries (more than 97 percent of Austrians retire before age 65). analyzes them. I once again agree with both the diagnosis and the remedies suggested, such as "the individualization of pension rights" to avoid benefit traps for women. But I sometimes do so for other reasons--starting from other working hypotheses and then arriving at other policy conclusions--than the ones put forward by the author. The Atypically Employed as Winners in Non-NDC-Schemes? Let me take Holzmann's assertion that "these atypically employed people do not fare well under some pension schemes, which are based on the full-employment fiction. . . . This sit- uation calls for reform (and a stricter contribution-benefit relationship)." This is true, no doubt, as it is general enough--but, therefore, the opposite proposition is equally true, and probably even more frequently true: namely, the atypically employed often fare very well under current pension schemes, despite, or even because, these schemes are based on the fiction of full employment. Just one outstanding example are full-time pension entitle- ments for predominantly or long-term part-time employees. In table 11.9, I provide a truly 270 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 11.4a. Europe Aging at Multiple Speeds, Lags and Peaks, 1995­2050: Subregional Deviations from All-European Regional Averages (percent) Northern Europe Southern Europe 9.0 9.0 6.0 6.0 6.0 3.1 3.0 3.0 1995 2050 1995 2050 0.0 0.0 ­3.0 ­3.0 ­2.5 ­6.0 ­4.7 ­6.0 ­9.0 ­9.0 Western Europe Eastern Europe 9.0 9.0 6.0 4.2 6.0 2.0 3.0 3.0 1995 2050 1995 0.0 0.0 2050 ­3.0 ­2.5 ­3.0 ­4.1 ­6.0 ­6.0 ­9.0 ­9.0 Source: Prinz and Lutz (1994a, 1994b). Note: Y-axis values greater than zero are considered above average; those below zero are considered below average. shocking example of "injustice" and "perverse redistribution" from long-term full-time workers without career development toward part-time employees with good careers (such as female academics). Despite much less than a lifetime working, these part-timers receive between around two and two and a half times higher a lifetime pension in real terms than the simple worker--with identical lifetime contributions. Such inequities of equal contributions and highly different benefits (and vice versa) within the private market sector may become even more drastic across the private and the public sector boundaries (as will be seen soon), as long as "best years" pension formulas prevail. All systems without a lifetime calculation-base lend themselves to such erratic redistributions in all directions, including frequent perverse redistribution from low- income to wealthy people, which are more able to fiddle the system by superior knowl- edge and ease of more flexible and well-dosed, opportunistically adjusted labor supply. Vesting periods as entitlements thresholds and other devices may also make for similar advantages. They are often used by self-employed persons and their family members, part-time farmers, railway workers, civil servants, and other special corporatist interest groups, providing them with benefits far above of those of people with equal contribu- tions--and even further above their own contributions. Thus the "stricter contribution-benefit relationship" called for by the author (which, of course, I agree with), would actually make few people better and many people much worse off than they are today--for good reasons of fairness. This is because current distor- TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 271 Figure 11.4b. When the Aging Process Will Reach Its Peak: Average Annual Increase of the Population Above Age 60 in Three Scenarios, by European Subregions (percent) Europe 1.98 2 1 0 ­1 1995 2025 2050 ­2 Northern Europe Southern Europe 2.2 2.0 2 2 1 1 0 0 ­1 1995 2010 2050 ­1 1995 2025 2050 ­2 ­2 Western Europe Eastern Europe 2.0 2.7 2 2 1 1 0 0 ­1 1995 2015 2050 1995 2025 2050 ­1 ­2 ­2 Source: Prinz and Lutz (1994a, 1994b). tions from defined contribution standards are probably non-accidental, and well-struc- tured by vested interests of atypical employees who typically stand to gain from existing pensions rules. Many "atypical" employees, in particular middle-class preferential part- timers, benefit from them at the expense of low-income people working long hours. In what one could call incomplete contributory Bismarckian insurance systems, DB pension awards are actually "based on the full-employment fiction." But--according to so my counter hypothesis to the one forwarded by Robert Holzmann and many other scholars-- this may more often make for their upgrading closer to full-employment or to non-con- tributory guarantee pension standards. In contrast, their downscaling to a minimum contribution equivalent base--or to indirect pensions rights only, derived through mar- riage, widowhood and other family status dependencies--is less probable and frequent. And non-contributory systems generally take care of flex-workers through other provi- sions, moving their pension entitlements above their contribution base. 272 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 11.4c. When the Aging Process Will Reach Its Peak: Average Annual Increase of the Population Above Age 60 in Three Scenarios, by country (percent) Austria Bulgaria Czech Rep. and Slovak Rep. 2.41 2.62 2 2 1.57 2 1 1 1 0 0 0 ­1 ­1 ­1 ­2 ­2 ­2 1995 2025 2050 1995 2050 1995 2015 2050 Denmark Finland 3.33 Germany 2.41 2.41 2 2 2 1 1 1 0 0 0 ­1 ­1 ­1 ­2 ­2 ­2 1995 2010 2050 1995 2010 2050 2000 2050 Greece Hungary Italy 2.52 2.13 2 2 2 1.96 1 1 1 0 0 0 ­1 ­1 ­1 ­2 ­2 ­2 1995 2050 1995 2015 2050 1995 2025 2050 Netherlands Norway Poland 3.19 3.29 2.46 2 2 2 1 1 1 0 0 0 1995 2015 2050 ­1 ­1 ­1 ­2 ­2 ­2 1995 2010 2050 1995 2010 2050 1995 2015 2050 Sweden 2.09 2 1 0 ­1 ­2 1995 2010 2050 Source: Prinz and Lutz (1994a, 1994b). TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 273 Table 11.9. Same Lifetime Contributions, Highly Different Pension Entitlements for Atypically Employed Part-Timers and No-Career Regular Full-Time Employees, Austria, 2003 (euros) Part-timer, Part-timer, Full-time employee, best years at best years at constantly below start of career end of career average income Lifetime contributions, indexed by: without any indexation, nominal 50,977 135,182 83,290 notional IRR median income 171,525 175,716 159,884 notional IRR covered wage sum 221,421 187,257 187,823 Nominal lifetime pension, residual life expectancy 24,25 years, calculated by: "best 15 years" 551,088 802,744 328,102 lifetime "40 years" 222,274 298,647 237,824 status quo: losses capped at 10% 495,979 722,470 295,291 median income 40 years 324,731 332,665 302,692 wage sum 40 years 419,195 354,516 355,587 Lifetime contributions, indexed by: without any indexation, nominal 54,026 135,182 91,743 notional IRR median income 175,716 175,716 175,713 notional IRR covered wage sum 225,944 187,257 207,383 Discounted lifetime pension, residual life expectancy 24, 25 years, discount 3% p.a., calculated by: "best 15 years" 399,206 581,504 261,166 lifetime "40 years" 166,111 216,339 190,026 status quo: losses capped at 10% 359,285 523,354 235,050 median income 40 years 240,982 240,981 240,978 wage sum 40 years 309,866 256,810 284,410 Lifetime contributions, indexed by: without any indexation, nominal 50,977 143,754 97,825 notional IRR median income 171,525 202,512 187,563 notional IRR covered wage sum 221,421 221,423 221,418 Discounted lifetime pension, residual life expectancy 24,25 years, discount 3% p.a., calculated by: "best 15 years" 399,206 581,504 278,102 lifetime "40 years" 161,015 242,015 202,774 status quo: losses capped at 10% 359,285 523,354 250,292 median income 40 years 235,234 277,730 257,228 wage sum 40 years 303,663 303,666 303,659 Source: Calculations by M. Fuchs, European Centre. Note: The table shows how well atypically employed part-timers can do under current DB pension schemes compared to no-career regular full-time employees. The 9 boldfaced rows in the table indicate both the lifetime contributions indexed by notional IRR median income and notional IRR covered wage sum as well as the nomi- nal and discounted lifetime pensions according to the "status quo: losses capped at 10%," an extended "best 15 plus a few years" combined with undervalued, "unfair" notional IRRs and crippling caps vs. a lifetime calcula- tion base (40 years) with fair IRRs indexed by the median income or the wage sum. Whereas the first option (status quo after the 2003 reform) shows highly unequal (up to 1:2.4) pension entitlements with identical contri- butions, lifetime calculations with income or wage-sum-based notional IRRs display identical entitlements for identical contributions. 274 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Other Non-Fiscal Reforms Needs Demanding NDC Both the non-individualization of social rights to health insurance, social insurance, and pensions, as well as all existing deviations of current DB pension schemes from NDC stan- dards of actuarial neutrality and fairness, are costly to society at large and increase public expenditures, Moreover, they tend to benefit those who are not most in need of support and targeted for special assistance, but those who are best able to seek the rents implicit in such incomplete arrangements, while taxing those outside the circle of the happy few privileged beneficiaries. NDC allows us to measure existing arrangements serving partic- ular sectional interests at the expense of others against universalistic standards most broadly accepted as reasonable, equitable, and fair. Another tendency requiring "changes in the way public programs operate, including in the area of pension provision" is increasing international market integration or globaliza- tion. It adds further non-fiscal reform needs to the already mentioned socioeconomic changes ignored by most countries that have introduced parametric reform packages dur- ing the last decade, exclusively addressed at balancing short- or mid-term fiscal require- ments. But open economies will not do well in a globalizing world with social security and public pension systems that "limit if not eliminate" labor mobility between sectors, occu- pations, and countries. They will not reap the benefits of globalization with pension rules that impede the improvement of financial markets, including a development of portable liquid pension assets from fully funded pillars. And they will not do well with pension arrangements that block life-long learning indispensable for knowledge and skill forma- tion, labor market flexibility, and prolonged activity in the workforce. Socioeconomic changes, globalization, and societal aging require a reform approach "that must go beyond a parametric adjustment of existing schemes," Holzmann argues, "toward a more actuarial system structure that better links contributions and benefits, more individualization to handle professional and family mobility, and also some funding to allow more individual decision and choices." At this point of reasoning, Holzmann ele- gantly interweaves the view that more pension reform or "adjustments" are needed with his second core idea: that even with more pension reform, more European coordination is needed. He finally weaves it into the third core assumption or leitmotiv: namely, that NDC is "ideal" to make it happen, a cornerstone in the welfare architecture of a "Pan-European pension system." The Claim for NDC as "Pan-European" Pension System The most innovative aspect of Holzmann's study is his argument in favor of NDC as responding to "the need for a better coordinated pension system in an integrated Europe." His "Pan-European approach" is not the first, but probably the most encompassing, treat- ment so far of the proposition to design NDC-type institutions to promote the emergence of an all-European pension system. He rightly deplores the fact that "there is little under- standing of and support for a Pan-European approach that should lead to a coordinated pension structure" and wonders why the Commission of the European Union (in contrast to other policy field) does little to overcome the perception of pensions as a strictly national agenda. Can European economic integration truly advance without at least some development toward an all-European pension reform approach? Budget requirements under a Maastricht fiscal regime, and enhanced labor market flex- ibility, mobility, and labor supply in aging societies, all demand some convergence in the area of pensions, which crucially affects overall labor supply and employment levels and TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 275 consumes up to half of all social expenditures. While migration and regional mobility will remain lower in Europe for a multitude of reasons, mobility is blocked not just across countries and regions, but also between sectors within countries and across professions within the same regions. In contrast to "other economically integrated areas under a com- mon currency (such as Australia, Brazil, Canada, Switzerland, and the United States) . . . the European Union does not have a coordinated--even less a harmonized-- pension system." While other states or provinces differ in many things "including income taxes or short-term social benefits . . . they have one thing in common--a public retirement income scheme across states." Even worse, European countries frequently have occupa- tionally fragmented pension systems within the same countries, preventing mobility across professions and between the public and the private sector, even within nearby areas. Barriers to Mobility There is nothing to criticize or add to Holzmann's reasoning about barriers to mobility, except perhaps some empirical evidence illustrating the almost incredible extent of barri- ers to mobility between public and private sector. In a corporatist country like Austria, public sector civil servants receive on average 264 percent of the median retirement income in the private sector (European Centre 2004). Within the same education bracket, public sector workers get up to 44 percent higher lifetime income than private sector employees; the gap decreases as educational attainment increases (Synthesis 2003). Due to a more favorable pension formula, civil servants get around 50 percent higher return on their retirement income--or notional interest rate, in NDC language--for the very same contributions during active life (Marin and Prinz 1999). To be even more specific: A female civil servant born in 1945 and retiring at age 56.5 in 2002 receives between 46 and 49 percent of her overall lifetime income as old-age pension. With primary education only, her lifetime income (1,926,190 euros) is more than double that of a private sector worker/employee (838,266 euros). Her retirement income (884,318 euros vs. 272,760 euros) is 3.24 times greater. With secondary education, the relationship of lifetime income is 2,408,151 euros, compared to 1,094,097 euros. The retirement income 1,175,108 euros vs. 370,124 euros, or 3.17 times higher (Synthesis 2003). In many cases, more than 50 percent--or the greater part--of overall lifetime income comes in retirement! Net income of retired civil servants on the regional (Länder) or munic- ipal level is rarely below 100 percent of their last (not their average) active income before retirement--and that is around 130 percent of average or calculation base active salary during their working life! In addition, the gap in retirement income between civil servants in the Länder or municipalities and that on the federal level in some of the regions may increase up to 43 percent over that of civil servants nationwide by the year 2025 (Marin and Fuchs 2003, table 3/B). And the minimal pension contribution varies up to around 10 times or 1,000 percent between different occupational groups for what may turn out to be the very same monthly retirement income entitlement (Marin and Prinz 1999; see also fig- ure 11.5). Under such circumstances, how could professional mobility ever take place, except in a unilateral direction toward the public sector? The most telling symptom in this context may be that even among academics, who profit least from these arrangements (and male academics hardly at all), more than 70 percent of university graduates seek public employ- ment as civil servants. Correspondingly, the outflow rates from public sector functions with permanent status (Pragmatisierung) is virtually nil--except toward those ever more 276 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Figure 11.5. Corporatist Pension Disharmonies: Cost of Minimum Contribution One Month of Retirement, Austria 2003 farmers--helping parents 41.39 farmers--with joint work management 41.39 farmers--helping children 44.86 "new" self-employed persons 80.67 traders (members of chamber of commerce), first three years 80.67 blue collar employees, white collar employees, freelancers 82.30 farmers and work managers 82.78 lawyers (excluding flat rate payment), Salzburg 133.67 traders (members of chamber of commerce) 160.92 liberal professions/self employed persons (FSVG) 214.56 lawyers (excluding flat rate payment), Upper Austria 248.10 notaries 331.94 0 50 100 150 200 250 300 350 minimum monthly pension contribution in euros Source: Hauptverband, AK department of statistics. frequent cases of special early retirement arrangements in ever younger age cohorts down to 45 to 55 ("Lehrermodell" for all, "Bundesbediensteten Sozialplangesetz," Berufsunfähigkeit- spensionen in "ausgeglieder-ten Einrichtungen"). Despite annual decrements of 4 percent (capped at 20 percent)--calculated from a standard retirement age 5 years below that of the private sector--the absolute monthly retirement income usually is still far above of that of an average private sector employee working until the age of 65. It thus provides a frugal early retirement basic income over around 30 to 40 years of further life expectancy. In addition, varying by education but through all education attainment levels, the term of the pensions is several years longer in the public sector. Based on this generous basic income, pensioners may start a second career or undertake rather informal income-gather- ing activities, but rarely do, as entrepreneurship is neither promoted nor needed for a comfortable third age over decades: more than a quarter of a century on average. The European Claims for NDC, Continued Robert Holzmann checks the potential structure of a Pan-European pension system against the (primary and secondary) goals developed by him and his team at the World Bank TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 277 (Holzmann, Orenstein, and Rutkowski 2003). A good pension system should "provide ade- quate, affordable, sustainable, and robust old-age income, while seeking to implement wel- fare maximizing schemes in a manner appropriate to the individual country." In doing so, it should create economic growth and minimize negative impacts on labor markets and other areas. As "specific objectives of a Pan-European pension system," he also suggests the four criteria of "mobility, national preferences, solidarity, and feasible transition." These ideal demands on a reformed and coordinated Pan-European pension system are then confronted with three main options for the "potential structures": "a basic pen- sion plus a mandated fully funded pillar (Beveridge for all); an unfounded defined bene- fit system plus voluntary fully funded pensions (Bismarck for all); and a basic of noncontributory pillar plus an NDC pillar plus a voluntary (or mandated) funded pillar." Holzmann then discusses the main arguments in favor of each option and the difficulties in implementation, the choses the third as superior. The arguments are all in all convinc- ing--apart from some reservations that I note below. NDC is designed as the crucial or first pillar, able to deal optimally with all system objectives and reform needs men- tioned--from financial sustainability to all socioeconomic requirements, including divorce, survivorship, mobility across professions and frontiers, and transition issues across member-country groupings. Holzmann considers coordination among the existing NDC countries, Italy, Latvia, Poland, and Sweden, which have adopted major differences in design and implementa- tion elements, including transition rules, within the same overall NDC scheme. Germany and France are considered "quasi-NDC countries" most easy to transit. Other Bismarckian countries (Austria, Belgium, Czech Republic, Greece, Hungary, Portugal, Spain) may need more time and reform impetus. The greatest difficulty would be transitioning the Euro- pean outliers with universalistic systems (Denmark, Ireland, the Netherlands, and the United Kingdom). Coordinated portability from/to other European systems may have to be achieved through buy-in options and transfers of the accumulated NDC amount. In debating the transition issues in introducing such a potential structure of a Pan-Euro- pean pension system, Robert Holzmann has no illusions, either about the speed or about the actors involved. "An approach initiated and led by the EU Commission" he considers "possible but not likely"; I would argue that is it impossible given the current interpreta- tion of the subsidiarity principle with regard to pensions. This may also explain that "there are no visible efforts by the Commission to take such a lead," even if "the necessity for a more rapid and a more comprehensive reform" is seen. Also, the method of open coordi- nation "is unlikely to lead to rapid national reforms--and even less to create a Pan-Euro- pean reform vision." He does not expect a great deal from country competition as various countries adopt reformed pensions systems it is "again possible, and a bit more likely, but not sufficiently rapid. Even if carried out, the outcome might be sub-optimal." Apart from the great time lags in a copy-cat world where countries learn from one another through imitation, institutional transfers will probably remain restricted--and may themselves not yet "ensure sufficient consistency of approaches across countries to provide the needed mobility of the workforce in Europe." Consequently, the only somewhat realistic perspective for Holzmann is a "cross-coun- try approach led by governments," for instance, through the EU's Economic Policy Com- mittee, should it take a broader view on aging issues. But the coalition to promote "a better coordinated, Pan-European pension system is quite likely the task of academics and research institutions, examined and supported by the EPC or similar core groups, and at some moment in the future espoused by a charismatic European politician as a reform champion. Perhaps this will happen after the first main asymmetric shock hits euroland." 278 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES I wish one could be more optimistic on that last point than the author. But perhaps Robert Holzmann's expectations already express the maximum feasible optimism in a Europe simultaneously growing together and drifting apart in many social policy areas, including pensions. If "No System is Politically Foolproof," Not Even NDC, Is the "Ideal" NDC Truly "Ideal"? Whether the very design of the "ideal NDC" in the Holzmann model--and the mix of pil- lars as such--is actually ideal must remain unanswered for the first and probably for the second query. The postulated mix of the model proposes the "structure of a (mandated) first pillar NDC plan, a (voluntary or mandated, if so desired) funded pillar with occupa- tional and individual retirement plans, and a basic pillar of social/non-contributory pen- sions that provides minimum income support for the very vulnerable elderly." This, in my view, is preferable to the older World Bank perspective of suggesting a combination of mandatory DB-PAYG with a mandatory fully funded DC (corporate or individual) private pillar and a voluntary fully funded DC (corporate or individual) private pillar. But there are also disadvantages, such as potentially fewer synergies in risk spreading, as will be seen in the next section. And whether the better mix proposed now is the best or "ideal" one is not easy to prove. The first step would be to demonstrate the crucial first NDC pil- lar as optimally designed. Generally, NDC may be close to an "ideal" pension framework, and thus the "basic structure of ideal NDC" as the core component of the overall combination is designed optimally almost by definition. But specifically, Robert Holzmann leaves actually open almost all parameters to be specified for an "ideal NDC." He neither specifies his choice of an ideal notional interest rate ("the discussion about the (most) appropriate . . . choice is far from over"), nor that of the remaining life expectancy, the indexation of benefits, the reserve fund to be created, or the amount of redistribution and the transition rules to new NDC benefits. Given the fact that all four European countries that have introduced NDC systems (Italy, Latvia, Poland, and Sweden) use different notional interest rates, different ways to determine the residual life expectancy, and different transition rules, some well-reasoned specificity in parameter choice for an "ideal" system would have been expected. But the only specific choice Holzmann makes is advocating a minimum eligibility age to a zero pillar pension that "may have to be conditioned on higher ages (say 70 onward)." He carefully outlines the issues at stake with any one choice taken, and the non-trivial "technicalities" involved in opting, for instance, for aggregated wage growth rates or for per capita wage growth. The same applies to choices between cross-section life expectan- cies and estimated cohort expectancies, between price indexation and revaluation beyond price adjustments, or between a top-down and a bottom-up approach in transitioning from DB-PAYG to NDC. Holzmann weighs the pros and cons, the contexts, and follow-up problems of any single choice taken without opting for a preferred one. This is legitimate and academic prudence, but may be somewhat disappointing to readers expecting policy conclusions from a policy expert directing an intergovernmental economic and social pol- icy institution. This even more so as we can see a risk of gradual erosion of NDC rules over time, from Latin America to Latvia. If no system is politically foolproof--not even NDC is a panacea, as Holzmann convincingly demonstrates--we may expect some more specific suggestions about how precisely to avoid choices that may turn out to be more foolish or less rational than intended. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 279 Risk Reinforcement Instead of Risk Diversification by the Holzmann Design of the "Ideal" NDC-Mix? As an "ideal" mix of pension systems around the core NDC pillar, Holzmann proposes combining this basic unfunded pillar with a supplementary funded scheme and with a social pension, unfunded as well. Whereas the crucial first NDC pillar and the social pen- sion pillar or non-contributory pension (as a kind of fall-back device "for the elderly poor") are both PAYG and mandated, the funded pillar--the second or third--is now pro- posed to be voluntary. Still, it will have to play an important role in a future "more coordi- nated but not harmonized Pan-European pension system," where national "differences would still exist," but where "their mobility-reducing effects . . . can be limited with a strong (voluntary or mandated) funded pillar." Apart from facilitating "Pan-European mobility," the funded pillar is meant to allow further "consumption smoothing beyond NDC benefits" and "according to individual preferences" without distorting labor supply and saving decisions; "to support retirement flexibility in an aging society"; and to acheive "risk diversification." Holzmann argues that "as funded and unfunded pillars have a different exposure to economic, demographic, and political risks, and as their rates of return are little corre- lated, diversifying pension benefits from two pillars is welfare enhancing." No doubt that it is welfare enhancing for globetrotting professionals like those assembled in Sandhamn: currently constrained by a conspicuous and hardly understandable lack of pension portability, great uncertainties about future entitlements, and significant retire- ment income losses from mobility. It thus, probably serves the great majority of inter- nationally mobile experts at the Sandhamn conference very well. Actually, many of us, despite being specialists in international pension issues, cannot more than very vaguely if at all envisage the kind and amount of retirement income to be expected from a diver- sity of institutions in a variety of countries. Whether similar welfare gains from com- bining NDC with a funded DC pillar can be expected by broader middle classes and lower-income strata is still to be seen. Again, Holzmann's argument holds in principle. But there usually is a much stronger complementarity of respective strengths and weaknesses between a tradi- tional unfunded DB and a fully funded DC system (the previous World Bank priority mix) than between a funded DC and a formally "unfunded" but quasi-funded NDC scheme. As the last combination basically introduces the logic, structures, and processes of private funded DC schemes within public PAYGO systems, synergies of supplementing the other system get lost. By implication, at least two groups of persons may find themselves in more difficult situations than before. Both of them belong to the large section of the population losing through a "quasi-actuarially fair" NDC system because they before have unduly benefited from DB schemes far above their contribu- tion base. The first group is those in principle able to compensate for the expected losses in pension income. They may do so by investing significantly in a voluntary second or third pillar of a funded pension: for instance, by converting their severance payment claims into an old-age provision. But the new funded pillar is meant to be DC as well, simple in design (to avoid disappointing take-up rates such as with the German Altersvermögensgesetz or Riester-Rente), and voluntary: three design elements with which I agree. But being voluntary also implies that people free to choose may choose as well not to take care of their supplementary old-age provision--or not take care of it sufficiently. 280 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Empirical evidence suggests that this is exactly what we have to expect, at least for some mid-term transition period. The majority of people are not fully aware of their "pension income gap." Those who are do only about a third of the savings they themselves consider necessary. The Sandhamn conference has provided additional evidence of incomplete infor- mation, lack of interest, severely limited rationality--and subjective preferences for DB schemes. Annika Sundén (chapter 13, this volume) shows that even for the comparatively well-informed participants in the Swedish system, less than half of them had looked at the benefit projection. They have less self-reported "good knowledge" in 2003 than they had in 2001 when NDC was started, and they show a relapse in most basic knowledge ("all years count") from 50 percent in the year 2000 to 38 percent in 2003. Low-income and younger par- ticipants are less likely to look at information at all. In the United States, workers generally lack knowledge about social security benefits, and those who depend most on them are least informed. In Michael Orszag's unpublished presentation to the Sandhamn conference, on which this book is based, he shows that people are generally more satisfied with quality of information in DB systems than in DC systems. Thus, combining two DC schemes will obvi- ously reinforce subjective feelings of uncertainty and information deficiency. As a consequence, even those who make provisions and voluntarily contribute to a funded DC scheme may at times find themselves with significant and unexpected real losses; and will always find themselves with the double uncertainty of two DC schemes and somewhat unpredictable annuities. Again, Mike Orszag has calculated the "loss in retirement income" in Europe and the United States from 2000 to 2003 as ranging from more than 15 to 25 percent for those who invested in half equity and from more than 40 to more than 50 percent for those who invested in all equity for the funded pillar. The index of all active funds in the Swedish PPM has gone down between the starting period in spring 2001 to 88 percent (spring 2002) and to 63 percent in spring 2003. The Swedish PPM is the Premium Pension system, with a contribution rate of 2.5 percent which accrues a capital-market-determined rate of return. The PPM complements the unfunded (first) pil- lar with a contribution rate of 16 percent in addition to a minimum "guarantee pension" which provides a basic safety net for those aged 65 and above. The corresponding share of Swedes opting for actively managing their premium pension capital has declined from 67 percent initially to over 14 percent last year to 8 percent in the first half of 2003, the third year of falling stock markets (Casey 2003). While fluctuations of NDC and funded DC schemes may be independent from each other, co-variation of volatility and thus risk reinforcement cannot be excluded, either. There are many scenarios where risks accumulate or even interact instead of cancelling one another smoothly. For many people, not only getting much lower pension but getting much less predictable retirement income from more sources than before may be the most probable outcome. This may still be meaningful from an overall welfare perspective, but it certainly is in sharp contrast to the hyper-stability of DB final salary pension rights for, say, civil servants in many countries today. It will therefore inevitably be perceived as deterio- ration over time, even if the final result may be equal to or even better than the status quo ante. In Germany, for instance, the Gesetzliche Rentenversicherung plus 4 percent savings qua Riester-Rente may generate a higher gross replacement income with anything higher than the projected 4 percent annual rate of return on financial markets--nominally (Börsch-Supan and Wilke, 2003, figure 13). But it will take time to find out what is actually going to happen--and how people actu- ally perceive what is going to happen under uncertainty. Welfare and old-age security must not only be provided, they must also seen to be provided, to be accepted as "good enough," if not "best" a practice or "ideal" pension arrangement. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 281 A Guaranteed Minimum Social Pension is Needed, Not a "Zero-Pillar" This applies even more so to persons on a minimum social assistance or old-age provision for the elderly poor, who may have been fewer before reforms, or not so poor under a no- reform scenario. In OECD countries, old-age poverty fell for decades until about the mid- 1990s, although it was still higher than average among the population above 65, and in particular the population over 75 (Förster 2004). Reforms will most probably reverse the past trend "of the increasing material well-being of the elderly population" and this turn- around has already been observed (Förster 2004). In an accession country like Hungary, for instance, around 150,000 more elderly poor in need of social assistance are expected as a consequence of reform, which is yet far from introducing an NDC scheme (see Marin, Stefanits, and Tarcali 2001). Holzmann recognizes clearly that "having under the new structure a quasi-actuarial NDC system as first pillar and actuarial funded second and third pillars tends to increase the efficiency in the labor market but reduces the redistribution of income toward the poor" and therefore requires minimum benefits. "Income support for the very vulnerable elderly to prevent old-age poverty is part of the adequacy objectives of any pensions sys- tem.' Consequently, he unequivocally calls for "a strengthened social or non-contributory pension in EU member-countries" necessary to counter-balance increasing "vulnerability of the elderly as aging progresses. He argues that "with incomplete and perhaps falling coverage under earnings-related schemes, one can conjecture that poverty incidence will increase as the increase in life expectancy continues." Let us not forget that more than the increase in residual life expectancy, it was and is the increase in survival rates to pension age that determines the pension load. Those survival rates rose from around 15 percent at Bismarck times in the 1870s to over about two-thirds in the post-war period to 90 percent today. Again, who would not share both the diagnoses provided and the normative and policy conclusions drawn in accordance with "the solidarity objectives of the European Union"? Two main questions remain. First, "how such a strengthened social pension pillar should be structured"? Second, as I added in my original comment on the Holzmann paper, "why should the guaranteed minimum social pension be conceptualized as--and initially even called--a `zero-pillar'"? Holzmann's response to the second query was to fully accept the objection "and to do away with the, perhaps, questionable notion" but not "with the proposed concept." He thus changed "the language around the zero pillar" into "social pension pillar" or "non-contributory pension," which signals rather a tribute to political etiquette or correctness--and less a rethinking of the real implications of semantic choices. Before returning to this issue, I accept that his answers to the first query, his own, are all specific enough and satisfying. He opts for a minimum pension under the NDC system in addition to a social pillar in order to "strengthen incentives for formal labor force partici- pation." But he sees that this also requires eligibility restrictions "in order not to contradict the neutrality objective of the NDC structure with regard to the individual retirement deci- sion." Again, I tend to partly agree with the philosophy as well as with the measures pro- posed. "For example, while allowing individuals to retire from the age of, say, 60 onward, it may have to be required to have a minimum accumulated notional amount equivalent to 100+ percent of the minimum pension or else the need to reach the standard retirement age of, say, 67 (which is increased with a rise in life expectancy). Second, coordinating a minimum NDC pension with a social pension with regard to labor market incentives requires different eligibility criteria (such as some kind of means or affluence testing of the 282 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES social pillar), different amounts, or different eligibility ages--or some combination of these. Finally, eligibility for a social pension may have to be conditioned on higher ages (say 70 onward), but means-testing may be kept light . . ." Still, Holzmann himself seems to be skeptic about the persuasiveness of his overall pro- posal: "How much national preferences such a social pension pillar would be able to exhibit . . . is open for discussion. . . ." This obviously needs more debate--and research. In this context, let me raise a few basic queries without having precise answers. Did not the very name of "zero pillar" originally disclose a preference for a residual conception of sheer poverty relief instead of a broader conception of welfare? Why not adopt the United Nations philosophy, as formulated from the time of the "Guiding Principles for Social Developmental Welfare Policies and Programmes in the Near Future" to the follow-up documents of the World Summit for Social Development (WSSD) 1995? Why not replace a concept of welfare as minority concern, program of poverty relief, and social control of those "living on welfare" by a more encompassing, comprehensive, and universal policy concern meant to "serve to raise the level of living of the widest possible sections of the population" (UN Guiding Principles; see European Centre 1993, pp. 212, 213 ff.)? Should guarantees regarding old-age pensions be restricted to protecting the marginal and vul- nerable instead of more universal minimal standards of human well-being and social inte- gration? Is protecting and compensating the most needy a sufficient complement to counting on self-help for all others, or may a strategy of enabling also require supporting those who generally take care of themselves--and assisting those who care for others? Are no other forms of minimum income guarantees and non-contributory social rights qua redistribution conceivable, desirable, or even preferable? How does which kind of coverage of unemployment and periods of illness, disability, military service, and family- related time off, such as maternity leave or care leave, supplement insufficient earnings- related NDC claims to a decent minimum income guarantee? As they add up to several years, and frequently more than a decade of absenteeism from work during working life (Marin 2000), they may better assist consumption smoothing and provide more targeted, social policy goal-oriented support beyond poverty relief and to equal opportunities than a social pension. Generous cash contributions for non-contributory periods (related to unemployment, sickness, disability, military service, family-related time-off such as maternity leave, care leave, etc.) may supplement insufficient earnings-related NDC accu- mulations much better than unconditional basic social pensions, achieving a decent mini- mum benefit level and at the same time providing better incentives. This, at least, is my major hypothesis. Why should social pensions be means-tested and restricted to "the vul- nerable elderly" and thus to the very margins of society, instead of being non-contributory universal grants available to specific categories of people, such as (working or lone) par- ents, caregivers, or the disabled? In short: why, after all, should the social underpinning of NDC cum voluntary funded pillar be conceptualized as if it still were what it is not any longer called: namely, a "zero" pillar? Did "zero pillar" not imply unintentionally but tellingly that getting something for nothing (or at least for less than one's contribution equivalent) may end up with getting next to nothing ("zero") for something (a tax-financed primary social policy goal "pillar," withering away with other "national preferences")? Why not conceive the social safety net below the mix of mandated NDC and a supplementary voluntary funded scheme as a "grounding," a basic pension guarantee--as much a core component as the crucial NDC-- and not as a "zero pillar," which may crack just when most needed? But I have to agree with Robert Holzmann's rejoinder in a personal communication to this critique that my "grandiose `pension guarantee' sounds fine but is not fully thought through with regard TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 283 incentives provided" and still "rather fluffy." Indeed, much more work needs to be done on the synchronization of minimum NDC pension, social pension, and the non-contribu- tory supplements to regular earnings-related NDC schemes for social policy goals on work-, family- and health-related interruptions of working life--open to everybody in need and not just to the poor. Disability Welfare: A Most Relevant Lacuna One crucial aspect of the pension problematique in general, and of massive pre-retirement as probably its single most important determinant in particular, that is completely ignored in the Holzmann model is disability pensions. (For recent publications, see OECD 2003; Prinz 2003; Marin and Prinz 2003; and Marin, Prinz, and Queisser 2004). This is even more surprising in view of two trends: first, the uncontrollable spending dynamics on disability policies during the last few decades; and second, its foreseeable future aggravation. With coverage and replacement rates of monthly pension incomes probably falling, as well as with retirement age rising under earnings-related schemes, one can assume that the propensity to exit early from the labor market via disability pension claims may increase. But can an "ideal" Pan-European pension pillar mix formula even be conceived without some solution to one of the major causes of early retirement, if massive early retirement itself is one of the major causes of unsustainable pension dynamics today? In addition to fiscal unsustainability, in several European economies with high non-employment rates, disability pensions play a major role in depressing labor force participation. Some facts and trends are puzzling indeed: How can invalidity pensions for the work- ing age population significantly increase with improved health and higher disability-free life expectancy, compression and postponement of morbidity? How is a steep rise in inca- pacity rates in working age possible along with a simultaneous reduction of chronic and occupational diseases, accidents, and work injuries--and with less exposure to infectious and contagious diseases (some of which have virtually disappeared) and to dust and to hazardous substances such as asbestos and other carcinogens? How can disability pen- sions in working age rise with a decline in disability of population groups of higher risk such as the elderly beyond working age 65? In Hungary, for instance, the majority of all new pensioners exit to retirement via inva- lidity (Marin, Stefanits, and Tarcali 2001, figure 2). In Austria, every second man retires during working age as disabled and the numbers climb to up to two out of three farmers or three out of four blue collar workers. For the age cohorts 55/56 years, invalidity pen- sions have increased by 555(!) percent in less than two decades. Despite many other avenues for early retirement, in the age group 60 to 64 years, 40 percent of males have an invalidity pension. For the OECD, the average is 23 percent. In the Netherlands, which has one of the world's most advanced medical and health care systems, almost 1 million per- sons of working age are on disability pension benefits; overall, invalidity recipiency rates have increased 86 percent between 1980 and 1997. In the United Kingdom, without any evidence of deterioration in health, government spending for sickness and disability has quadrupled over the past two decades, and 40 percent of working-age recipients of state benefits now claim sickness and disability compensation. Correspondingly, the general slowing down of the rate of welfare expenditure expansion (social spending roughly doubled between 1960 and 1980 and has increased around 20 per- cent since) has affected disability pensions less than any other social expenditure. Extension of programs, number of beneficiaries, and amount of expenditures for disability have steadily increased for about 35 years, even if one controls for the changing age structure of 284 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES societies. Periodic efforts at retrenchment (in the mid-1970s and 1990s) have succeeded in slowing down recipiency growth rates, but never the growth of beneficiaries as such; the stock of benefit recipients remained high, and the inflow rates much higher than outflow. As a consequence, even disability pension expenditures have begun to show reduced inflow rates: that is, continuing though slower expansion dynamics. But overall cost containment Figure 11.6. Variation in Public Expenditure for Disability Related Programs (percent- age of GDP) Non-EU (9): 2.07 OECD (20): 2.42 EU (11): 2.70 OECD (17): 2.72 Norway 5.58 Sweden 4.66 Netherlands 4.64 Poland 4.51 Denmark 3.78 Germany 3.08 Austria 2.92 Switzerland 2.40 Spain 2.28 Italy 1.83 Belgium 1.72 France 1.67 United Kingdom 1.54 Portugal 1.53 Turkey 1.46 Australia 1.44 United States 1.40 Canada 1.28 Korea, Rep.of 0.29 Mexico 0.29 0 1 2 3 4 5 percent Source: Marin and Prinz (2003, p. 36). Note: All disability-related programmes include broad disability benefits and employment-related programs for disabled people. OECD (17) excludes Republic of Korea, Mexico, and Turkey. The vertical lines in the graph indicate mean. TOWARD A COORDINATED PENSION SYSTEM IN EUROPE 285 will be a core challenge in the years to come--more than in any other social policy field. Expansion of disability pensions over the last decades has become uncontrollable (fig- ure 11.6). Social expenditures on disability now total several times the social costs of unemployment, even under adverse conditions of very high unemployment rates. In 19 out of 20 OECD countries investigated, disability costs were significantly higher than the costs for unemployment. On average, they were more than double the costs (2.17 times). In Norway, they are up to 11.9 times the unemployment costs. Public expenditures go up to 5.58 percent of GDP, and are 2.72 in OECD-17 and 2.70 in EU-11 countries. Still, high costs do not guarantee good targeting. A recent OECD report (2003) shows that a clear majority of severely disabled people, most in need of support, are not awarded a disability benefit, Figure 11.7. Disability Status of Disability Benefit Recipients not disabled severely disabled EU (10): 34.4 OECD (14): 31.0 EU (10): 34.4 OECD (14): 32.7 moderately disabled EU (10): 31.2 OECD (14): 36.2 Mexico 91.4 8.6 Sweden 48.9 20.0 31.0 United States 46.7 26.6 26.7 Italy 43.9 25.9 30.3 Belgium 43.4 31.8 24.8 United Kingdom 43.3 28.8 27.8 France 33.3 25.4 41.2 Netherlands 30.6 36.2 33.2 Switzerland 29.8 70.2 Portugal 28.6 27.5 44.0 Norway 28.4 30.4 41.2 Austria 27.7 38.3 34.0 Denmark 26.2 43.6 30.2 Spain 18.3 34.5 47.1 Australia 15.2 45.0 39.8 Korea, Rep. of 93.4 6.6 100 80 60 40 20 0 20 40 60 80 100 percent not moderately severely disabled disabled disabled Source: Marin and Prinz (2003, p. 57). Note: For Mexico and Switzerland, "severe" and "moderate" are one number. Mexico and Switzerland are excluded in the OECD average. The vertical lines in the graph indicate mean. 286 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES whereas more than 40 percent of disability pension recipients are self-declared non- disabled (figure 11.7).