Report No. 33709-KZ Kazakhstan Reforming Intergovernmental Fiscal Relations Lessons From International Experience January 13, 2006 Poverty Reduction and Economic Management Unit Europe and Central Asia Region Document of the World Bank FOREWORD This report is based on the findings o f missions to Kazakhstan in August and December 2004. It was prepared by William Dillinger (Task Manager) with the assistance o f Ilyas Sarsenov and Stepan Titov o f the World Bank. The team wishes to acknowledge the valuable advice and information provided by Minister Kelimbetov and other officials o f the Ministry o f Economy and Budget Planning, the Department o f Economic Policy o f the President's Administration, staff o f the oblasts o f South Kazakhstan and Karaganda, and the Public Policy Research Center. This volume i s a product o f the staff o fthe International Bank for Reconstruction and Development/ The World Bank. The findings, interpretations, and conclusions expressed inthis paper do not necessarily reflect the views o fthe Executive Directors o f The World Bank or the governments they represent. The World Bank does not guarantee the accuracy o f the data included inthis work. The boundaries, colors, denominations, and other information shown on any map inthis work do not imply any judgment on the part o f The World Bank concerning the legal status o f any territory or the endorsement or acceptance o f such boundaries. TABLE OF CONTENTS Executive Summary......................................................................................................... i Background..................................................................................................................... Introduction..................................................................................................................... 1 Assignment of Functions ............................................................................................ 2i Assignment of Revenues............................................................................................. 6 Trends inFiscal Performance ..................................................................................... 7 Issues............................................................................................................................. 10 FundingDiscretionary Expenditures ........................................................................ 11 Implications for Kazakhstan ..................................................................................... 16 Education .................................................................................................................. 17 Health........................................................................................................................ 20 Conclusions ................................................................................................................... 22 EXECUTIVE SUMMARY Kazakhstan's present system for allocating revenues to subnational governments is largely based on historical trends. These, in turn, have their origins in the Soviet-era practice o f assigning jurisdictions fixed shares o f the revenues collected within their boundaries. This has given rise to substantial disparities in per capita revenues among oblasts. Such disparities are not inherently undesirable. Costs vary across jurisdictions. Different levels of spending are required to achieve equivalent outcomes. But the size o f the current disparities may leave poorer jurisdictions unable to provide adequate levels o f basic education and health. This has important implications for Kazakhstan's efforts to reduce poverty. This report looks at ways to address this problem. It draws, in part, on the practices o f large Western federations. It also draws on the experiences o f the recent EU accession countries--all o f which once had systems similar to Kazakhstan's. The report makes two major recommendations. First, it recommends that the Government provide separate, earmarked funding for functions with critical distributional implications. This would include primary education and primary health. These services would be financed on a capitation basis (Le., an equal amount per pupil and registered patient, respectively.) The remaining subnational functions would be financed from an annually-determined share o f the budget. Inprinciple, these funds could be distributed on a per capita basis. To ease the adjustment to the new system, part o f the funding could continue to be distributed on the basis o f origin. INTRODUCTION 1. Kazakhstan is in the process o f transforming the way it allocates public sector resources. The present system has its roots inthe Soviet model, where the state allocated resources through annual negotiations over physical targets and resources within a single planning framework. The country is moving toward a system inwhich separate functions are assigned to separate tiers o f government and some effort i s made to provide a stable, transparent source o f revenue to each tier. While some progress has been made on both fronts, much remains to be done. In particular, the Government has yet to derive a transparent, equitable basis for allocating resources among tiers o f government. 2. The Government requested Bank advice on this issue as part o f the ongoing Joint Economic Research Program (JEW). In particular, the Government sought the experience o f relevant comparator countries. As agreed during discussions o f the concept paper, these include the large scale federal systems o f the West (including Canada, Germany and Australia). They also include a selected group o f unitary countries from the former Socialist bloc: the eight Eastern European countries that recently joined the European Union. All eight began the 1990's with a Soviet-era intergovernmental structure, much like Kazakhstan's. Their efforts to conform to Western European models contain useful lessons for Kazakhstan. BACKGROUND 3. Kazakhstan i s a unitary state. The President i s directly elected for a 7-year term, (most recently in January 1999). The President serves as commander-in-chief, sets foreign policy, can initiate legislation, and appoints the Prime Minister, subject to Parliamentary approval. The legislative branch is comprised o f a lower house (the Majlis) and upper house (the Senate). Sixty-seven o f the 77 members o f the Majlis are elected from single mandate districts. The other ten members are elected by party-list vote on the basis o f proportional representation. The Senate has 39 members. Two senators are selected by each o f the elected assemblies o f the 16 first-tier subnational units (14 oblasts, plus Astana and Almaty). The President appoints the remaining seven senators. Although the legislature has the authority to initiate legislation, most is proposed by the Government. 4. Subnational administration i s composed o f two tiers o f government. At the provincial level, there are 14oblasts and two cities with oblast status, Almaty and Astana. Oblasts are further subdividedinto 168 districts (rayons) and 39 towns with rayon status, yielding a total o f 207 third-tier units o f government. Both tiers o f subnational administration government are headed by chairmen (akims). The President appoints the akims o f the oblasts and Astana and Almaty. Akims o f other administrative units may be appointed or elected to office, at the discretion o f the President. Both tiers also have legislative bodies (maslikhats) whose members are elected by universal adult suffrage for four year terms. The powers o f subnational maslikhats are severely circumscribed. Under the Constitution, the akims are an integral part o f the central government executive branch and function as representatives o f the President. The role o f maslikhats i s largely limitedto reviewingand approving actions initiated bythe akims. I Assignmentof Functions 5. From independence until 1998, the division o f functions between tiers o f government largely reflected the practices o f the previous Soviet era. Expenditure assignments were not formalized and were often subject to change. In 1998, new legislation explicitly distinguished the respective functions o f the central government from those of subnational governments (a term which included both oblasts and rayons). Beginning in 2004, there have been attempts to distinguish the respective functional responsibilities o f the two tiers o f subnational governments, although some overlaps and ambiguities remain. (See Table 2.) 6. Defense, justice andpublic security. The current Budget Law (as amended through July 8, 2005) assigns the Republican budget exclusive responsibility for (inter alia) defense- and operation of the judiciary. and penal system. Although all three levels o f government are responsible for `ensuring public as %total order and public security' and for military public sector as %total spending on subnational recruitment, the Republic level dominates this function spending category o f expenditure. As shown in Table 1, admin 26% 4% subnational governments account for only 17 defense 17% 2Yo percent o f public expenditure on defense and 21 police 21% 4% percent o f public expenditure on police. 7. Education. Education dominates the I social asst I 15% 1 7% I expenditures o f subnational governments. As utilities 98% 12% shown in Table 1, subnational governments account for nearly 90 percent o f total public total 43% 100% expenditure on education. By the same token, education spending consumes about one-third o f subnational budgets. 8. Under the current budget code, the Republic's role in education is limited to higher education and the development o f textbooks and teaching materials. The role o f oblasts i s similarly limited. The current budget code confines oblasts' responsibilities to vocational education. This is reiterated in sectoral legislation', which limits the responsibilities o f oblasts to `elementary and secondary vocational education, special education, general education under special educational programs (e.g. for the handicapped), and general education for talented children inspecialized schools. Rayons 2 therefore bear the legislative brunt o f compulsory education. Under the budget code, they are exclusively responsible for the organization and provision o f compulsory education (grades 1-11). Sectoral legislation, similarly, requires them to `secure provision o f citizens with compulsory secondary general education (Le., grades 1-11.) ' Law.. ~ * Local .on Education (No 389-1) as amended through December 20,2004 Government Law nevertheless assigns both oblasts and rayons overlapping responsibility for `ensuring implementation o f civil right for free (secondary) education. 2 9. Health Care. Subnational governments also play a preeminent role in health care. Under the budget code, the Republic's role i s largely limited to the provision o f specialized medical assistance and medical rehabilitation at `state-run public health institutions o f Republic significance.' Oblasts bear the brunt of health care responsibilities. The budget code assigns them responsibility for ensuring `healthy sanitary/epidemiological conditions for the population and for providing `other health services with the exception o f those provided by the Republican budget'. This i s further elaborated inthe Local Government Law4, which requires oblasts to ensure that `citizens' rights to guaranteed free medical care are implemented according to national standards'. Neither the budget code nor the local government law assigns any health care responsibilities to the rayon level. 10. Social protection. The Budget Code assigns the majority o f social protection responsibilities to the Republic level. These include old-age pensions, state social allowances, and childbirth allowances. This i s reiterated in sectoral legislation, which specifically assigns these functions to the Ministry o f Labor and Social Protection. Oblasts are assigned responsibility for other vulnerable groups, including orphans and the disabled. Under the budget code, rayons are responsible for the administration o f state targeted social assistance and social assistance to `some categories o f indigent citizens at the discretion o f the local akimat.' Sectoral legislation further defines the latter to include housing assistance, social help provided at home and `social adaptation of homeless'. As noted in Table 2, the local government law i s less explicit on the division o f functions between oblasts and rayons. Oblasts are required to `develop and implement programs facilitating employment and poverty reduction, coordinate social assistance to socially vulnerable groups'. Rayons must `address issues o f social assistance, protection o f maternity and childhood, provide targeted assistance to socially vulnerable strata and coordinate charity assistance.' 11. Other functions. Taken together, social services-education, health, and social assistance-account for about sixty percent o f total subnational expenditure. (In2004, the proportion was 58 percent, down from 60 percent in 2003). Expenditures on housing, public utilities, and road transportations account for much o f the rest. The budget code provides no direct role for the central government inthe provision o f housing and related utilities. The oblasts' role i s limited to the provision o f gas. Rayons therefore bear the brunt o f expenditures in these sectors. Under the budget code, their responsibilities include construction o f public housing, construction and maintenance o f water supply facilities, water treatment and draining systems, sewerage, heating and electricity networks (when owned by the municipalities); solid waste management, street lighting. (Note that in Almaty and Astana, the roles o f rayons are performed by oblast level administrations.). Expenditures on housing and public utilities account for twelve percent of subnational expenditures, and nearly all (98 percent) o f public sector spending in this sector. Responsibility for road construction and maintenance is divided among the three 'Unlike many former-Soviet-bloc countries, Kazakhstan has no extra-budgetary health insurance system. In1998, the Compulsory Medical InsuranceFund(CMIF) was extinguishedand its functions integrated into the respective budgets o f the central and subnational governments. Law.. . No 148-11o f January 2001, as amended through April 15,2005 3 levels of government according to the national system of road classification. Roads of international and Republic significance and assigned to the Republic level, and roads of oblast andrayon significance are assigned to their respective tiers. 12. It is not clear that the legally-mandated divisionof functions between levels of government i s precisely observed inpractice-particularly as regards to the respective responsibilities of oblasts and rayons. A recent report by the Center for Research and Development states that the "demarcation of functions between oblast andrayon levels in each oblast i s established by oblast akims at their own discretion. For example, health services and education incertain oblasts are financed by oblast budgets, while inother oblasts they are financed by rayon and city budgets." The report finds, for example, that inSouthKazakhstanoblast, the oblast budget finances medical assistance to AIDS infected patients, campaigns against epidemics, and specialist medical assistance to cancer patients andthose with drug abuse issues. Rayons finance rural medical attention and obstetrical center^.^ Box 1T h e Macroeconomic Context of Intergovernmental Fiscal Reform Kazakhstan's current macroeconomic situation has several implications for intergovernmental fiscal reforms. Overall, it provides an supportive environment. GDP grew by over 9% in 2004. Rapid growth is expected to continue. As a result, the system for distributingrevenues can altered without imposing substantial cuts on any onejurisdiction. Butgrowth has beenfueled mainlyby the oil sector, which attracts $4-5 billion inFDI annually. This i s widening economic disparities among oblasts. Rapid growth i s concentrated inthe oil-rich regions and the two largest cities where the bulkofprivate and public resources are spent. As these economic disparities increase, the urgency o f redistributive transfers will grow as well. Increasesin oil and gas production and highworld prices are also contributing to rapid increases in central government revenues. As a result, the country has run fiscal surpluses (averaging 2% o f GDP) since 2001. Oil revenues are inherently volatile. These surpluses have therefore been set aside ina separate oil fund managed by the central government. As noted in the World Bank' recent country economic memorandum, these reserves will have to be managed prudently. Any sudden increase in the total volume o f public spending will require further adjustments inthe division o frevenues between tiers o f government. 5Center for Research and Development ,2004, TheDistribution of Governmental Budget Expenditures Between Central,Regional and Local Administrative Levels in the Republic of Kazakhstan, Almaty 4 .-E c 5- 0 4 .-0 c 0 > Assignment of Revenues 13. Tuxes. Subnational governments derive the majority o f their revenues from centrally-administered taxes collected within their boundaries. Prior to 2001, four major tax bases were shared with the oblasts: the personal income tax (PIT), the corporate income tax (CIT), certain excise taxes and (until 1997) the VAT. The proportion o f revenues to be shared varied from oblast to oblast and from year to year, on a negotiated basis. 14. A 2001 reformintroduced identical sharingrates for eachtax for all oblasts. Local governments were assigned 100 percent o f the PIT and 100 percent o f the payroll tax (which had formerly supported various extra-budgetary social funds), and 100percent o f the excise tax on gambling, along with 50 percent o f the CIT, 50 percent o f the excise tax on alcoholic beverages, and 15 percent o f excise tax on crude oil, including condensed gas. In addition, local governments were assigned 100 percent o f the land tax, the property tax and the vehicle tax and the authority to impose fees for use o f local roads, the registration o f firms, other administrative fees and penalties. A subsequent reform (in 2002) eliminated the corporate income tax and the excise on crude oil from the pool o f taxes subject to sharing. As shown in Figure 1, the majority o f subnational tax revenues are now derived from the personal income tax and the payroll tax, with smaller contributions from excise taxes. In 2004, payroll taxes accounted for 45 percent o f total tax revenues; the PIT accounted for 27 percent. Figure 1 15. Current Transfers. Returning T r e n d s i n C o r n p o s i t i o n o f S u b n a t i o n a l R e v e n u e s tax revenues strictly on the basis o f origin [ a s \ G D P I would result in wide disparities in per I , capita revenues among oblasts. To reduce I these disparities, the Government uses a I I system o f gap filling subventions and withdrawals. This is based on system o f revenue and expenditure projections made I during the budget process. Oblasts whose projected expenditures exceed their projected revenues are allocated additional funds (subventions) to make up the part o f the difference. Oblasts whose revenues are projected to exceed their 2 C % C O $ 2 C B 1 3 % 6 C B h 3 1 C O i 2 3 % expenditures are subject to withdrawals; Le.,. a reduction in the percentage o f the centrally administered taxes they are allowed to retain. 6 16. Intheory, the calculation ofrevenues is to be based on the `tax potential' of each oblast6 Estimates o f expenditures are to be based on service norms. The current Budget Code states that projected local expenditures shall be calculated "accounting for their distribution for current budget programs and development budget programs, budget provision/ security, and established norms..." Article 71 o f the budget code states that `in-kindnorms shallbemandatoryinbudgetplanning and execution'. 17. But in practice, both estimates are largely derived from historical trends. According to interviews conducted for this report, estimates o f revenues and expenditure needs are largely determined though annual budget negotiations between oblast akims and the Ministry o f Economy and Budget Planning (MOEBP). Each oblast akim submits a budget proposal slightly higher than the previous year's, and this i s whittled down by the MOEBP based on resource availability and presidential priorities. History i s not the only factor. Projections o f aggregate economic growth guide the calculation o f revenues. Projections o f expenditures reflect changes inthe major expenditure drivers-particularly public sector wage increases-and the impact o f specific Presidential initiatives (such as the construction o f new schools). 18. This approach is, ineffect, codified in the most recent Government resolution on the methodology for calculating general official transfers.' This specifies (in section 3.5) that the projection o f expenditures will be `based on the level o f expenditures inthe year preceding the planning period', adjusted according to the consumer price index (and further adjusted in light o f resource availability), along with the impact of any laws or presidential decrees providing for increases or decreases in expenditures'. In the current (2005-2007) exercise, these include US$ 111.1million for guaranteed free medical aide, US$ 20.2 million for teachers' salaries, US$ 21.3 million for the wage bills o f rural districts and US$17 million for the maintenance o f newly commissioned education and health care institutions. 19. Capital transfers The Government also makes a variety o f other transfers to subnational governments. The largest these, the targeted investment transfer, finances regional and local capital works. In 2003 and 2004, a large volume o f such transfers were allocated to Astana. (These accounted for 28% and 45% o f the city's receipts in those years, respectively.) With this exception, capital transfers do not constitute a major source o f subnational receipts. In 2003, they comprised 3.4% o f total subnational net receipts (when Astana is excluded). In2004, the figure was six percent. Trends in FiscalPerformance 20. As shown inFigure 1, total subnational revenues have remained roughly constant, as a percent o f GDP, over the last six years. Total revenues (including net transfers) were equivalent to 10.8 percent o f GDP in 1999 and again in 2004, despite the substantial changes in tax assignment which took effect in 2002. The elimination o f the corporate income tax from the pool o f taxes subject to sharing was largely compensated by an increase in net transfers from the Government budget. While withdrawals exceeded 'Budget Code, as amendedthrough July 8,2005 Resolutionof the GovernmentNo. 916Onthe Methodology for Computation of General Official Transfers 7 subventions duringthe period 1999-2001, subventions exceeded withdrawals in2002 and have continued to increase. In 2004, net transfers (subventions less withdrawals) totaled KZT 142. billion, or roughly a quarter o f aggregate revenues. Figure 2 I 21. Despite the system o f I m p a c t o f E q u a l i z a t i o n o n P e r C a p i t a R e v e n u e s 2 0 0 4 subventions and withdrawals, A i t a n a c ~ t y substantial variations in revenues M a n g i r t a ~i 0 b l a s t A t y r a u ~ b l a l t exist among oblasts. Figure 2 K y i y i o r d a o b i s i t Aim a t y city compares the per capita tax W e s t K a z a k h r t a n o b i l i t revenues o f each oblast in 2004 A l m o l a o b l a r t A k l o b e o b l i i t prior to withdrawals, with its net D D P a v I o d I t 0 b l a s t c N o i t h K a z a k h s t a n e b l a r t per capita revenues after K a i a g a n d a 0 b l a s t withdrawals and subventions. As E s i t K a z a k h r t a n 0 b i a s 1 K o i t s n s , 0 b l a s t shown, the principal impact o f the 2 h a m b y l o b l a s t S o u t h K a i a k h r t a n 0 b l 2 6 1 transfer system is to reduce the A l m ~ I ~ o b I a S t revenues o f the two oil-producing 0 0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 I 2 0 0 1 4 0 0 oblasts (Atyrau and Mangistau) and ~u n c t i e v e n u e m g i a t r t a r i r v e n v e _ _ ~ the city o f Almaty to the benefit o f the remaining thirteen jurisdictions. Astana benefits disproportionately from the transfer transfer system increases its revenues, to a level roughly 3.5 times the national average. system. While its pre-transfer tax revenues are the third highest in the country, the8 With this exception, the transfer system in general reduces the average variation in per capita resources. In2004, it reduced the standard deviation intax revenue among oblasts from KZT 25,000 to KZT 8,540 or 24 percent o f the mean (excluding Astana). This leveling effect has persisted over several years, particularly since the elimination o f the corporate income tax from the revenue sharing pool. As shown in Figure 3, the transfer system has roughly halved the standard deviation per capita revenues among oblasts in each o f the past five years. Nevertheless, substantial variations in per capita remain. As shown in Figure 2, per capita revenues in 2004 ranged from KZT 130,000 in Astana to KZT 24,000 inthe Almaty oblast. Even excluding Astana, the variations are significant. Per capita revenues inthe richest jurisdiction (excluding Astana) are 2.2 times that o f the poorest. 22. Under the terms o f current Government policy, this pattern o f regional disparities will persist. One o fthe innovations o fthe most recent decree on transfers was to fix the level o f subventions/withdrawals for each oblast innominal terms for three year periods. As shown in Figure 3, the legislation provides for an abrupt decline in subventions in 2005, followed by a gradual rise thereafter. lo).The distributional impact o f the proposed transfers i s shown in Figure 4, which compares actual net revenues in 2004 with the proposed distribution o f net revenues in 2005. As shown, the proposed distribution of * This largely reflects the highvolume o ftargeted investment grants allocated to Astana in2004. Ifthese are excluded, Astana's receipts are only twice the national average. L a w on the amounts o f generalpurpose transfers between the republic and oblast budgets, budget o f cities under state jurisdiction and the capital for 2005-2007. (August 31, 2004) loThis largely reflects a substantial cut intransfers to Astana. Under the proposal, Astana would become a net donor to the transfer system, rather than a net recipient. 8 withdrawals and subventions would further widen the disparities in per capita net revenues among the oblasts, roughly doubling the standard deviation (from KZT 25,000 in2004 to KZT 51,000 in2005)." Figure 3 _ _ ~ ~ _ _ ~ 23. Variations in revenues are ~ ~ .- T r e n d s i n S u b v e n t i o n s a n d W i t h d r a w a l s reflected in substantial variations in the level of spending on specific 2 5 0 sectors. Table 3, below, shows the ~~ ~ -+subventlolns Correlation between per capita --+--w I t h d r a w'as revenues and per capita expenditures -_ on each o f the major functions. (The ~ calculations exclude Astana.) As \QP \+' \."? 2." . 3 $ 6 \%$\$ \." \\+ shown, the correlations are all positive, implying that the more revenue an oblast has, the more it spends across the board. But the relationship between revenues and expenditures is complex. In2004, the correlationbetweenper capita revenues and per capita spendingon education was fairly strong. ( ~ 7 3 ) .Least square regression analysis suggests that for every 1 tenge increase in revenues, spending on education increases by KZT 0.23. But the relationship i s highly influenced by certain outliers. Spending on education inthe two oil-producing oblasts is considerably higher than the national average. Removing Atyrau and Mangistau from the analysis would reduce the correlation to .47. In Almaty, per capita spending on education i s far below the national average. Removing Almaty from the analysis would increase the correlation coefficient to .90. In the health sector, the impact o f outliers i s less pronounced. While health spending inMangistau is considerably higherthan the national average, spending inAtyrau is not. Removing the two oil oblasts from the analysis would have little impact on the relationship (r=.89). Figure 4 24. As shown in Table 3, the P e r C a p i t a N * t R . Y ~ I I U . I C u r r a n t a n d P r o p o s e d relationship between revenues and A r t a n a c i t y sectoral spending appears to be highly M a n g 1618 u I b i a s t A t y r a u 0 b l a 6 t unstable over time. The correlation K y z y l a r d a o b i a i l between revenues and education A i m a l y c i t y W e s t K a t a k h r t a n a b l a ~ t spending has declined sharply over the A k m o l a a b l a r t last three years, as has the correlation A k t o b e o b l a s t with housing and transport. The P a v l o d a i e b l a r l N o r t h K a z a k h s t a n a b l a s t correlation with health care spending, K a i a g a n d a o b l a r t on the other hand, has increased. E a r l K a z a k h r t a n o b 1 1 6 1 K a s ta n a I 0 b i a s t Further research would be required to Z h a m b y I o b i a r t precisely define the relationship S o u t h K a z a k h s t a n o b i a s 1 A l m i l y o b l a s t between revenues and the sectoral 5 0 0 0 0 1 0 0 0 0 1 5 0 0 0 2 0 0 0 0 pattern o f spending. What is unambiguous i s that, despite the "Estimatesfor2005arebasedonactualgrossrevenuecollections in 2004, adjusted for withdrawals and subventions as proposed inthe plan. 9 budgetary norms, spendingon each sector varies considerably among the jurisdictions. ISSUES 1itself a cause for concern. Even in a unitary state such as Kazakhstan, where the priorities o f the central government are intended to be observed throughout the hierarchy o f subnational jurisdictions, there i s reason to expect variations inthe level and composition o f subnational spending. Costs vary across jurisdictions. Different levels o f spending are required to achieve equivalent outcomes. Priorities also vary across jurisdictions. Housing and public utilities may be a priority in the rapidly growing republican cities, whereas education and health may be more important inthe stagnant ruraljurisdictions. 26. The problem lies instead with the method used to determine the net resources available to eachjurisdiction. Because it i s largely based on history, it does not provide a rational basis for implementing the spending priorities o f the Government. Instead, spendingpatterns largely reflect the historical distribution o f taxable activity. Intheory, one solution would be to precisely calculate the level o f spending required to meet national outcome standards for each function in eachjurisdiction. Oblast akims would be required to allocate their budgets accordingly. Withdrawals and subventions would be calculated on the basis'o f the difference betweenthese costs and the projected revenue o f eachjurisdiction. But this level o fmicromanagement i s administratively infeasible. While outcomes in some sectors are relatively easy to define-student performance o f standardized tests, levels o f infant mortality-the level o f spending required to achieve them is difficult to determine. In other sectors, desirable outcomes are themselves difficult to define. It i s not clear what sort o f outcomes the Government would wish to pursue under the heading o f `culture and sports', or `administration' nor is it clear that there i s an overriding national interest in ensuring that such outcomes are achieved uniform1y. 27. The alternative is a system o f subnational finance that ensures that overall government objectives are pursued, while leaving micro-allocation decisions to the oblast akims and their subordinates at the rayon level. In the Kazakh case, this might imply a relatively precise specification o f the funding of education and health-where there i s a clear national interest in uniform outcomes. For other, more discretionary functions, it would imply a financing formula that i s not tied to specific sectors but instead reflects the Government's interest in ensuring a minimal level o f discretionary resources in each 10 jurisdiction. This approach--a combination o f general revenue equalization with sector specific-financing for education and health-+, in fact, used in the majority o f countries reviewed for this report. FundingDiscretionaryExpenditures 28. In mechanical terms, all transfer systems confront two basic design questions: how to determine the total amount to be transferred and how to distribute that amount among individual jurisdictions. Some countries address these questions simultaneously: by fixing a normative level o f spending to be achieved in alljurisdictions, for example, and then providing a sufficient level o f central government subsidies to achieve this result. Other countries address them independently, fixing the amount to be transferred on the basis o f the yield o f a particular tax, for example, and distributing the proceeds according proportionately. Countries also vary widely in terms o f the aims o f their transfer systems. Some attempt to equalize per capita revenues. Others attempt to take differingcost variables into account. 29. Germany. O f the three large federal revenue sharing systems, Germany has the most straightforward. The German system is essentially aimed at reducing disparities in per capita revenues among the states. The principal sources o f state revenues in Germany are the personal income tax (PIT) and the VAT. While both taxes are administered by the states, they are imposed at nationally-uniform rates and are subject to federally-mandated sharing arrangements. 30. Revenues from the personal income tax are divided in three ways-- 42.5 percent to the federal government, 42.5 percent to the states, and 15 percent to the municipalities- -andare distributed on the basis o f origin. Revenues from the VAT are divided between the federal government and the states. (While the proportions are subject to change, the federal government currently retains approximately half o f the revenues.) The remainder i s distributed to the states. Seventy five percent o f the states' share i s distributed on the basis o f population. The remaining 25 percent o f the states' share is distributed to states whose per capita revenues from the PIT (as well as minor business taxes) fall below the national average. 31. The combined effect o f distributing 75 percent o f the states' share o f the VAT on the basis o fpopulation and usingthe remainder to compensate states whose PIT revenues fall below the national average i s a substantial flattening o f inter-state revenue disparities. In2001, the percapita revenues ofthe richest state were Approximately six times those of the poorest. After redistribution, the ratio dropped to approximately 1.5: 1. But because the system does not attempt to equalize total revenues, substantial disparities remain amongjurisdictions. 32. Canada. The Canadian system o f intergovernmental transfers is more complex, largely because provinces are permittedto set the rates o f their major taxes. The principal sources ofprovincial revenue are the PIT, the payroll tax, the corporate income tax (CIT) and a VAT, termed the tax on goods and services. All these taxes (along with 29 other minor taxes) are administered by the provincial governments at rates determined by each province. Provinces retain 100 percent o f the tax revenues collected intheir jurisdictions. 11 33. The aim o f the Canadian equalization system i s to reduce disparities, not in tax revenues, but intax bases. (Because each province can set its own tax rate, any attempt to equalize tax revenues would prompt the provinces to drop their tax rates to zero.) To accomplish this, the federal government first calculates a per capita provincial revenue standard applicable to the nation as a whole. This i s determined by: (1) calculating the average per capita tax base o f the five middle income provinces-Quebec, Ontario, Manitoba, Saskatchewan and British Columbia and then (2) applying the average o f all provincial tax rates to it. For each o f the poorer provinces, the Government then calculates the revenues that would result from applying the average tax rate to the actual tax base o f the province. The amount o f the transfer is then based on the difference betweenthe two figures. 34. Canada's equalization transfers are funded from general federal government revenues. Until recently, the amount o f funding was subject to annual budget negotiations. As a result, the level o f funding for equalization varied frmi year to year. (Allocations in the current FY 2004/2005 budget year were sufficient to bring all provinces up to 95 percent o f the five province standard--C$6217). According to a policy paper adopted in 2004, now intends to fix the Government aggregate amount o f the equalization transfer on the basis o f a formula. In 2005-06, funding levels will be set at C$10.9 billion and are to be increased by 3.5 percent per year thereafter.12 35. Australia. The Australian system i s yet more complicated, as it attempts to equalize not merely tax revenues but service outcomes. The Federal government monopolizes all but one o f the principal tax bases in Australia: the PIT, the CIT, excise taxes and (as o f 2000) the newly created VAT. The only major revenue source allocated to the states is the payroll tax. Both rich and poor states are therefore dependent upon transfers from the federal government. 36. The aim o f the Australian transfer system i s ostensibly to bridge the gap between the own-source revenues o f each state and its expenditure requirements. This is determined by calculating, first, the relative revenue raising capacity and expenditure needs o f eachjurisdiction. The revenue capacity o f each state i s determined largely on the basis o f the yields o f the payroll tax. (Because the payroll tax is imposed at nationally uniform rates, the complex revenue-base equalization procedure required inCanada is not needed.) Expenditure needs are calculated on the basis o f separate sets of indicators for each o f 41 expenditure categories. The relative costs o f general medical services, for example, are based on population adjusted for age and sex composition, indices o f health statistics (e.g. infant mortality) and social composition (i.e., percent aboriginal population). Each o f these service indicators i s assigned a weight, which is then used to aggregate the relative expenditure needs of each state. The indicator o f expenditure needs, less the indicator o f revenue raising capacity, determines the relative expenditure gap o f eachjurisdiction. l2Government o f Canada, Ministry o f Finance `Equalization Program' www.tin.~c.ca/FEDPROV/eq~e.html. Note that the equalization transfer is not the largest form o fFederal transfer to provinces. The largest i s Canada Health and Social Transfer, which is earmarked for health care and social assistance, including early childhood development. (The allocation for the CHTiCST was C$ 21.8 in200314, as opposed to C$ 8.7 billionfor equalization transfers.) 12 37. The amount o f funding for equalization i s determined through a separate exercise. Inessence, it is based on the yield ofthe VAT. Total VAT yields (net of administration costs) multiplied by each state's share o f the total expenditure gap equals the amount o f the transfer to eachjurisdiction. As a result, the Australian system makes no pretence of achieving any specific level o fper capita revenues in each state or any standardized level o f spending. The complex system for calculating revenue capacity and expenditureneeds is merely a mechanism for distributingan exogenously determinedlevel o f resources: the annual yield o f the VAT. Table 4 Revenue Sharing Systems inSelected Industrialized directly from its profits. If Federations Country Basis of Distributable Distribution Objective the enterprise were -Pnnl --_ subordinate to a higher tier Germany 25 percent ofVAT equalize per capita revenues o f government, the Canada annual budget equalize per capita tax bases municipality would receive revenues and estimated expenditures government would receive revenues from shares o f the turnover tax (on consumer goods, foods and certain extractive and light industries) and from the personal income tax. As a result, a municipality's revenues were partly dictated by the number and scale o f enterprises within its territory, and by their ownership. But not entirely. Local budgets were also subject to a complex subventions and withdrawals. Budgets could be supplemented from higher levels o f government and by contributions from large enterprises within their boundaries. Thus the level o f resources available depended upon the ability o f the local party secretary to remain on good terms with the party bosses at higher levels and with the managers o f state farms or factories. 39. All eight o fthe EUaccession countries beganto dismantle the Soviet era structure inthe early 1990s. Ineconomic terms, the dismantlingo fthe Soviet-era structure took the form o f privatization. This had two important implications for local governments. First, it eliminated an important provider o f public services. Second, it drastically altered the revenue base, as profits from local-government-owned enterprises were no longer a source o f revenue. In political terms, this took the form o f an end to the Communist Party's monopoly on political power andthe rise o f competing parties at both the national local level. This, in turn, increased pressure for local fiscal autonomy while raising the risk that local governments would use that autonomy to thwart central government fiscal and equity objectives 40. hresponse, all the EU8 countries have radically reorganized the structure of intergovernmental fiscal relations. While many retain the system o f subventions and l3The Czech Republic, Estonia, Latvia, Lithuania, Hungary, Poland, Slovakia, and Slovenia. 13 withdrawals, the scope for bargaining has been considerably reduced. At the same time, the allocation o f transfers has become considerably more transparent and more effective inachieving specific equityobjectives. 41. The primary source o f subnational revenue in all the EU8 countries is the PIT.I4 Inall the EU8 countries, the PIT is administered by the central government at nationally uniform rates. Fixed shares o f the revenues are then assigned to local government. The percentage o f the PIT that is shared with municipalities varies among the EU8, ranging from 30 percent inthe Czech Republic and Poland to 73 percent in Latvia and 94 percent inSlovakia. 42. Distributingthe PIT solely on the basis o f origin would result in extremely wide variations in revenues among local jurisdictions. As shown in Table 5, all the EU8 countries employ some form o f formula to reduce these variations. Three techniques are used. The first is to distribute the PIT on the basis o f origin and provide additional funding to poorerjurisdictions from general central government revenues. InEstonia, for example, the local share o f the PIT is distributed solely on the basis o f origin. The central government then provides additional funding to poorer jurisdictions until they reach a fixed percentage o f the national per capita average. In Slovenia, similarly, local governments receive 35 percent of the PIT, distributed on the basis o f origin. The Government then uses a complex formula to supplement the revenues o f poorer jurisdictions. The aim o f the Slovenian system is to ensure an equal level o f `relevant' spending' in all jurisdictions. In principle, relevant spending i s defined as the amount needed to comply with local government tasks as provided in legislation. Inpractice, it i s nationally uniform per capita figure fixed annually by the Parliament. To calculate the amount o f subsidyto a givenmunicipality, this figure i s first multiplied by the population o f the municipality and four adjustment factors (favoring municipalities with lengthy road networks, large territories, and disproportionately large numbers o f children and people over 65). The expected tax revenues o f the jurisdiction (including revenue from the PIT and various local taxes) are then subtracted from this figure. Municipalities whose relevant expenditures exceed their projected revenues receive a subsidy covering the difference. Municipalities whose expenditure falls below estimated revenues are allowed to keep the surplus. l4The only exception i s the Czech Republic, where the PIT, the VAT and the corporate income tax are all subject to sharing. As discussed later inthe report, subnational governments also receive significant funding earmarkedtransfers for education and (indirectly) health inmost o f the EU8 countries. 14 43. The second technique is to distributethe PIT entirely on an origin base and then require richer jurisdictions to share part o f their PIT revenues with poorer ones. This so- called fraternal equalization method i s used in Lithuania and Latvia. In Lithuania municipalities with per capita revenues over 165 percent o f national average must transfer all o f the surplus to an equalization fund. The fund is then used to bringpoorer jurisdictions to 65 percent o f the national average. Any funds that remain once this target i s achieved are transferred to municipalities on the basis o f need indicators, the most important o fwhich i s school enrollment. 44. Latvia uses a more complicated technique. The process begins with a calculation of expenditure needs. Aggregate expenditure needs are first calculated on the basis of total municipal expenditures in the preceding year, adjusted for inflation. (The figure i s further refinedduringnegotiations with the Uniono fLocal Governments.) The projection o f aggregate spending i s then divided between republic cities and other local governments on a 45/55 basis. Within each o f these two groups, the relative expenditure needs o f each municipality are then calculated on the basis o f six criteria: population, number of children ages 0-6; number o f children ages 7-18, population above retirement age, number o f children in orphanages and number o f elderly in retirement homes. If a given municipality's projected revenues exceed its projected expenditure needs by more than 10 percent, the municipality is subject to a revenue cap. Forty-five percent o f the surplus i s taken away and allocated to an equalization fund. Ifprojected tax revenues are less than 90 percent o f estimated expenditures needs, the gap is fully met, usingproceeds from the equalization fund. Any shortfall between contributions into the equalization fund and payments out o fit are to becovered bythe Government. 45. The thirdtechnique is to distribute the PIT on a basis other than origin. The Czech Republic, for example, distributes the municipal share o f the PIT (as well as the VAT and CIT) on a per capita basis, weighted to favor larger jurisdictions. (A separate, more complicated, formula i s used to distribute the PIT to regional government.) In Slovakia, similarly, the municipal share o fthe PIT i s distributed on a per capita basis. 46. Poland uses a combination o f the first and second methods. Municipalities are entitled to 27.6 percent o f the PIT. (Intermediate levels o f subnational government 15 receive another 2.5 percent.) Municipalities whose per capita revenues exceed 150 percent o f the average are required to contribute to an equalization hnd. These contributions, plus additional funding from the central budget equal to at least one percent o f state budgetreceipts, are usedto bringpoorer jurisdictions up to 85 percent of the national average. 47. All o f these approachesproduce a largely similar result: the disparities that would result from distributing the PIT on the basis o f origin are substantially flattened. Using general government revenues to top up the PIT receipts o f poorerjurisdictions has much the same effect as requiring richer municipalities to transfer some o f their PIT revenues to poorer ones, or distributing the PIT on a per capita basis. (The degree o f flattening depends, o f course, on the specific parameters o f the formula.) Implicationsfor Kazakhstan 48. The characteristic that distinguishes all these systems from the current system in Kazakhstan is the use o f formulas to determine level o f resources at subnational level. Rather than relying on annual negotiations and historic trends, the large Western federal systems and the EU8 countries allocate resources on the basis o f objective criteria. These clearly vary in terms o f objectives and complexity. The system for determining the total amount to be transferred varies from the simple--a fixed percentage o f a designated tax base--to the highly complex, where the amount i s determined on the basis o f equalization targets. The criteria used for distribution also range from the simple-an equal amount per capita-to the highly complex formulas aimed at factoring invarious determinants of costs. Some are more relevant to Kazakhstan than others. The complex system used to equalize tax bases in Canada, for example, would not be required in Kazakhstan, as oblasts have no control over tax rates. 49. Determining the amount to be distributed. Inprinciple, international experience would suggest that the amount to be distributed should be based on a fixed share o f a designated tax base. Other things being equal, transparency and predictability are desirable characteristics in a system o f intergovernmental fiscal relations. Designating a fixed source o frevenue is one means o f achieving this. 50. As noted earlier, this is the practice in Germany, Australia, and most of the EU8 Box 2 PreviousAttemptsat FormulaBasedCriteria Formula based distribution criteria have alreadybeen consideredinKazakhstan. Resolution 529 of 1999 would have grouped oblasts into three categories based on natural conditions (temperature, level of desertification, salinization) and other indicators of need (incidence of contagiousdiseases andlengthofroads) andthenforced convergence inper capita expenditures within each group over several years. Revenues would have been based on 1999 revenues. Aggregate subnationalexpenditureswould haveremainedfixed inrealterms through 2001. This proposal was never implemented. The World Bank's previous public expenditure review* reported that `the reality is that Resolution 529/99 is not being implemented anymore, and because the level oftransfers still varies from year to year ina non-transparentway, uncertainty of financial flows to andfrom localbudgetsstill exists * World Bank KazakhstanPublicExpenditureReview2000 16 countries. But it i s not clear that this would be the best approach inKazakhstan. There are two reasons. First, given Kazakhstan's extremely high growth rates, the yield o f any particular tax i s itself unpredictable. Tying subnational revenues to a particular tax base would therefore not ensure stability. Second, the political and institutional conditions that give rise to this approach in the Western countries do not exist, to the same extent, in Kazakhstan. Western countries rely on designated tax bases in order to limit intergovernmental conflicts over resources. Once an agreement on the revenue base is reached, annual conflicts over central government support can be avoided. In Kazakhstan's more centralized political system, this is not so important a consideration. 51. Instead, there i s a good case for allowing the Government to set the overall level for subnational spending as part o f the budget process. Stability could be achieved by fixing the level on a multi-year basis, either by allowing for adjustments on the basis o f population growth and inflation or by tying the overall level to increases in aggregate public sector revenues. 52. With respect to distribution criteria, transparency would argue for a limited number o f readily measurable variables. Provided that the Government institutes separate, more precise mechanisms for financing key social sectors (as recommended earlier), there would appear to be little justification for fine tuning the distribution of funding for administration, culture, transport, housing or other discretionary expenditures. One attractive approach would be to simply equalize revenues per capita. Such a radical change may not be feasible inthe short term. But the level o f equalization need not be 100 percent. In line with the Canadian, German and Latvian examples, equalization could be aimed at simply reducing the variance among oblasts, ensuring for example, that no jurisdiction falls below a fixed percentage o f the national average level ofper capita revenues. Education 53. With respect to key social sectors, however, there is a strong case for ensuring a . standard minimum level o f spending. In theory, the existing budget process already incorporates a mechanism for ensuring adequate, uniform financing for compulsory education (grades 1-11). Budget norms specify standard classroom sizes (which should permit numbers of pupils to be converted into numbers o f classrooms); teaching loads (which should permit numbers o f classrooms to be converted into hours o f teacher time) and salaries for various levels o f teachers (which should permit hours o f teacher time to be converted into a wage bill). Similar norms exist for the non-wage components of compulsory education. In principle, this should yield a uniform level o f per education spending, per pupil, in every jurisdiction. (Salary scales for teachers are uniform across the co~ntry.'~)Figures on education spending, noted earlier, suggest that this is not the case. Per capita spending on education in Atyrau oblast, for example, is nearly twice the level inKostanai oblast. 54. There are a variety o f solutions to this problem. One would be to implement the norms and provide earmarked funding to cover the resulting costs. Given the scale o f 'jIBRD2005 Kazakhstan: Reforming the Public Sector Wage SystemPolicyPaper 17 Kazakhstan and the variety o f environment and socio-economic conditions within it, this approach i s likely to be both practically and administratively infeasible. Monitoring and enforcing standard classroom sizes, teacher loads and teacher remuneration (including ad hoc salary supplements) may be beyond the capability o f the central administration. An alternative solution would be to shift to a capitation-based financing system, with allocations to each jurisdiction based on enrollment. This would permit the oblast akims to allocate funding to individual rayons-and rayon akims to allocate funds to individual schools--in light o f local conditions, while ensuring an adequate overall level o f funding for this function. 55. Five o f the EU8 countries-Poland, Hungary, Lithuania, the Czech Republic, and Slovakia-now employ some form o f capitation-based financing for primary education. None o f them employs a single, nationally uniform amount, however. In all five countries, capitation rates are adjusted to reflect ostensible differences in the costs providing education. In Poland, for example, rural schools receive a 33 percent supplement over the basic per capita amount. Town with populations under 5,000 receive an 18 percent supplement. InLithuania, the capitation amount is determined according to a complex formula that differentiates among both school characteristics and grade levels. (For students in grades 1-4, rural elementary schools receive roughly twice as much funding per student as those inmunicipal capitals andbig cities. Schools with fewer than 150 children receive 50 percent more than those in the 150-299 student bracket and nearly twice as much as those with 300 or more students.) In the Czech Republic, capitation figures distinguish among different levels and forms o f education as well as among regions. Regional variations are intended to reflect variations in labor costs, and therefore favor, rather than discriminate against, Prague. Box3 EfficiencyBenefitsof Capitation-BasedEducationFinancing One o f the major attractions o f capitation-based financing in Europe i s that it decentralizes the political onus of closing schools in jurisdictions where enrollment i s declining. This i s a common problem in Eastern Europe, where the population o f school age children i s declining. Keeping under-enrolled schools open i s costly. Shifting to capitation-based financing cuts the funds available to schools with declining enrollments, forcing local governments to reduce teaching loads and in some cases, close schools. The same demographic conditions appear to exist in Kazakhstan. A recent report by the Asian Development Bank* predicts that Kazakhstan's school age population will decrease by 32 percent by 2015. It also found that teacher: pupil ratios are already low. As a result, the ADB advocated a shift to capitation-based financing, at least for general secondary education (grades 1-11). Under the ADB plan, the system would initially be used to allocate funding among oblasts. Itwould later be used to allocate funds directly to individual schools. 18 56. Capitation-based education financing has already received some consideration by the Government o f Kazakhstan. The State Program for Education (Sept 2000) and the Strategic Plan o f Development to 2010 both call for the introduction o f capitation-based financing throughout the education system. The more recent National Program of Education Development, however, makes no reference to changes in the system for distributing education funding. It merely calls for an increase inspending at the republic and local level. 57. European experience suggests that two ancillary reforms are required for the successful implementation of capitation-based education financing. First, local governments must have some degree o f control over the determinants o f education costs. Where falling enrollment triggers a drop in funding, local governments must have the legal authority to reduce expenditures by a corresponding amount. The lack of expenditure autonomy has hampered capitation-based financing in several o f the EU8 countries. InPoland, for example, school directors are, in theory, free to make their own decisions about staffing levels. But regulations constrain dismissals. Although a teacher may be dismissed when a school i s totally or partially liquidated, local government must provide six month's severance pay and re-employ the teacher at the first opportunity. Similar constraints on downsizing exist inLithuania and Hungary. 58. Second, the central government must be willing to increase the level o f capitation financing to reflect centrally-mandated increases in costs. In Europe, as in Kazakhstan, local government typically have little control over wage rates. The principal determinant o f costs-wage levels-is largely determined by the central Government. In Lithuania, for example, the national civil service laws sets out a structure o fpay scales for municipal employees, based on grade, years o f employment and size-class o f municipality (and in the case o f teachers, number of classes taught). InPoland, similarly, teachers' salaries are determinedon the basis o f the Teacher's Charter and annual ministerial regulations on the remuneration of teachers. Legislation in some o f the EU8 countries requires the Government to increase capitation grants in line with annual salary adjustments. The legislation enabling Slovakia' recent shift to capitation-based financing, for example, requires the capitation amount to be based on annual salary negotiations.l6 59. Neither o f these problems presents an insuperable obstacle to the introduction of capitation-based school financing in Kazakhstan. Subnational governments already have considerable legal discretion to dismiss redundant teaching staff. (Teachers do not have civil service protection.) The Government, for its part, is legally obligated to provide additional funds to cover the costs of any new expenditures it imposes on subnational governments. As a result, subnational governments have the means to adjust to declining enrollments and an assurance that any centrally-mandated increases in costs will be covered by increases intransfers. l6EuropeanCommission (EurydiceEducationDataBase). 2003. TheEducation System in Slovakia 2001/2002. 19 Health 60. Inprinciple, a similar approach could be usedinfinancing some aspects ofhealth care. One o f the major sources o f inefficiency in the Soviet model o f health care was its excessive use o f secondary and tertiary facilities. Primary physicians were trained in a narrow range o f fields, requiring referrals to more expensive secondary and tertiary facilities. Performance incentives were also weak. Compensation to medical staff was not linked to the volume o f services they provided. To varying degrees the EU8 countries have addressed the excessive use o f secondary and tertiary by introducing the so-called family medicine model o f primary care. Under this approach, patients register with an individual doctor of their choice who becomes the primary point o f contact in the heath care system. These doctors receive training ina broad range o f primary health care fields, limiting the need for referrals. To encourage physicians to register patients, the Government health systempays providers a standard rate for each patient on their roster. To encourage physicians to actually serve these clients, they often provide additional funding on a fee for service basis. 61. To an extent, the Kazakh Government appears to be considering the same approach. The proposed National Program o f Health Sector Reform'7 would consolidate all subnational spending on health at the oblast level. Primary health care would be separated from hospital care and would be financed on a per capita basis. The Program 'would not, however, incorporate performance incentives. According to the Program document, the per capita norm for primary health care facilities (w)ould be determined `taking into account the required volume o f health care at this level, sex and age o f members covered, and geographical area o f the territory serviced'. It would not, however, provide additional funding to providers on the basis o f the volume o f services they perform 62. The recent experience o f the EU8 countries suggest that this may be the wisest approach, at least in the short term. Performance incentives have proven incompatible with cost containment due to the tendency o fproviders to respond to payment-for-service forms o f compensation by increasing the volume o f services they ostensibly provide. Slovakia, for example, introduced a capitation-based for primary care system in 1994. To reward performance, it replaced it with a mixed system (60 percent capitation, 40 percent payment-for-service) in the following year. The resulting explosion in the number o f claims forced the Government to revert to an exclusively capitation-based system in 1998. It has now reintroduced a payment-for service element but limited it to specific preventive procedures. 63. The Czech Republic arrived at a similar system by a different route. The Czech Republic initially attempted to control primary care costs by fixing an overall spending ceiling for primary care and allocated it among individual physicians on a fee for service basis. Physicians responded by inflating the number o f treatments they claimed. Because the spending ceiling was fixed, more services meant smaller reimbursements per treatment. This prompted physicians to further inflate the number o f treatments they l7National Program o f Health sector Reform and Development inthe Republic o f Kazakhstan for 2005, 2010; para 5.1.1 20 claimed. The Czech Republic eventually brought an end to this vicious circle by introducing a strict capitation-based system in 1997. Under the current system, primary physicians receive a fixed amount per registered patient, differentiated by age group. (Zero-to four-year olds, for example, are worth 4.2 times as much as patients inthe 20-24 year old range.) Hungary, similarly, now reimburses primary care solely on a capitation basis, using a points system that distinguishes among patients in five different age groups. 64. Shifting to a purely capitation-based system for financing primary care does have one clear risk. It gives primary providers an incentive to immediately refer patients to higherlevel health care facilities, rather than treating patients on site. Europeancountries have addressed this problem in several ways. Some continue to reimburse providers on a fee-for-service basis for priority items, such as immunizations and screenings. Some set up administrative procedures to restrict referrals. Others rely on individual health care institutions to limit referrals. In Lithuania, for example, the HIF reimburses health facilities on the basis o f the number o fpatients registered with it. Each facility, inturn, is allowed to reimburseindividual physicians on a fee-for-service or capitation basis." 65. Rationing secondary and tertiary care has been even more difficult. For secondary and tertiary care, EU8 countries have attempted to shift the basis for financing from one based on inputs (salaries, equipment, supplies, etc.) to one based on outputs (treatments provided) Again, this approach i s endorsed in the Kazakh national health Program. Hospital care would be provided through contracts with individual health care institutions, and would be financed on the basis o f treatment-specific reimbursements (diagnosis related groups--DRG's). The treatments eligible for reimbursement would be definedingeneral package o fhealthcare benefits (GPFHC). 66. Inprinciple, this would have the same salutary effect as the proposed reforms in primary health care. Fiscal flows would become more transparent. Funds would be allocated more efficiently. By providing funding on the basis o f outputs rather than inputs, it would encourage the efficient use o f staff and facilities. And by forcing health providers to compete for patients, it would encourage client-responsiveness-at least in theory. 67. Experience in the EU8 suggests that this approach should be implemented with caution. All the EU8 countries have experimented some form o f output based compensation, generally following the model used in Germany-the so called `German points system'. Under the German points system, each treatment eligible for reimbursement is assigned a number o f points, based on its complexity and cost. A total budget (in monetary terms) for eligible health care is then agreed between individual health insurance funds and regional physicians associations. Individual physicians report the number o f points they have accumulated to the physicians' association, which calculates the value o f a point by dividing the agreed budget by the total number o f points reported. Individual physicians are reimbursed according to the number o f points they have billed and the value o f an individual point. In effect, this system encourages the 18 Salaries, however, remain the predominant form o f compensation. 21 physicians' association to ration the supply o f health services, as more points for one physician means lower reimbursements for all. 68. Hungary experimented with this approach during the 1990's. Between 1993 and 1997, separate points systems were introduced for outpatient services, acute inpatient services and chronic inpatient services. The amount o f funding for each category o f service was fixed in the yearly budget o f the Hungarian health insurance fund. Points were added up nationally and the monetary value o f one point was calculated by dividing the predetermined budget by the total number o f points. But the Hungarian health insurance fund proved less adept at restraining performance inflation than the German physicians' associations. Claims increased and the value o f a point dropped. In2000, the health insurance fund abandoned this approach. For outpatient care, the value o f a point is now fixed in advance. To control the supply o f services, fiscal penalties are imposed on individual providers. As o f 2004, providers receive 100 percent reimbursements only for the level o f services provided in the previous year-or more precisely for 98 percent o f that level. The excess, up to five percent, is reimbursed at 60 percent; between five and ten percent, 30 percent; and above that level, at only ten percent. Individual providers therefore have a strong incentive to turn away patients once their quota has been filled. 69. The Czech Republic has largely abandoned the points system altogether. The Czech Republic experimented with points system 1993-97. A total o f 4,500 distinct procedures were reimbursable, with points based on the amount o f time required to carry out a given procedure. Hospitals also billed points for each day spent in the hospital. Under the Czech system o f regional sickness funds, each fund calculated the value o f a point separately, by dividingits revenues from contributions by the total number o fpoints billed to it. But, as in Hungary, the volume o f claims exploded. In 1997, the Ministry o f Health published a new list o f procedures with new point numbers. This met with criticism both from providers, who believed that the new point numbers were insufficient to cover the costs o f services, and from insurers who argued that their revenues from contributions would be insufficient to pay the providers even the stipulated amounts. The points system was therefore largely abandoned. Sickness funds now reimburse individual hospitals on the basis o f aggregate budgets, which are largely based on expenditure inthe preceding year. The points system is merely used to determine whether each hospital has delivered an equivalent volume o f services. 70. Poland now rations the supply o f secondary and tertiary care thorough individual contracts with health care facilities. The number o f cases to be treated (under separate diagnostic-related groups) is agreed in advance and specified in the annual contracts between the sickness funds and individual health care providers. Similar methods are used inmost Western European countries (excluding Germany). Such an approach bears consideration inKazakhstan. CONCLUSIONS 71. The current system for financing subnational governments in Kazakhstan i s in need o f reform. Because it i s largely based on history, it does not provide a rational basis for implementing the spending priorities o f the Government. Experience in the large 22 Western federal countries and in the more recent, Eastern European, members of the EuropeanUnion illustrate how the system might be reformed. 72. First, it suggests that the current system of historically-based subventions and withdrawals be replaced by one that explicitly reflects Government objectives in the form a transparent, readily measurable, formula. One such objective i s clearly the reduction of interregional resource disparities. International experience suggests a variety of means of doing so. One promising approach would have the Government fixing a minimum per- capita floor for subnational spending for all oblasts. Oblasts exceeding the floor would be required to contribute a percentage of the surplus to the Government-although not necessarily 100%. Oblasts falling below the floor would receive a subsidy to bring them up to the minimum.Stability could be achieved by fixing the minimumlevel on a multi- year basis, either in real per capita terms or as a percentage of aggregate public sector revenues. 73. Separate, sector specific formulas may be required to achieve the Government's equity and efficiency objectives in the social sectors. With respect to primary education, there i s a strong case for providing financing on a per-pupil basis. This would not only ensure an adequate level of spending for this function, but also provide an incentive for subnational governments to close under-enrolled schools and shift resources to more crowded facilities. Ancillary reforms are required to achieve this result, however. Subnational governments must have some degree of control over the determinants of costs, particularly including the authority to dismiss excess staff. And the Government must be willing to increase the level of capitation-financing to reflect any centrally- mandated increases in salaries. 74. There i s a similar case for capitation based financing for primary health care. This would again ensure an adequate level of funding for this function and encourage treatment at the primary-as opposed to secondary and tertiary level Some countries have attempted to include performance incentives into their system o f primary health financing. Kazakhstan should not follow suit. Until Kazakhstan has the means to closely monitor the validity of claims, performance incentives are likely to encourage fraud and drive up costs. Box 4 FinancingCapitalInvestment As noted earlier, capital grants constitute a relatively small proportion of subnational receipts. Access to credit i s also limited-as it should be. This suggests that most subnational capital expenditures are financed from current account savings, i.e., the surplus of taxes and general subventions over recurrent expenditures. Such surpluses financed about two-thirds of subnational capital expenditures in 2002 and 2003. In 2004, the figure fell to 43%, due to the growth of targeted capital transfers to Astana. Financing capital investment involves a distinct set of issues and i s therefore not addressed in this report. It would be a worthwhile subject of future joint economic research. 23 75. Reforms in the financing of secondary and tertiary health care should be similarly modest. While performance-based financing (inthis case on the basis of DRGs)would, in principle, reward efficiency, in practice such systems have been difficult to administer. Kazakhstan would do well to follow the example of most European countries, and ration secondary and tertiary health on the basis of service contracts with individual institutions. 76. There i s one larger lesson to be drawn from international experience-particularly the experience of the recent EU accession countries: reforms along the lines of those recommended here can be implemented in a relative short period of time. Fifteen years ago, most of EU8 countries hadintergovernmental structure very similar to Kazakhstan's. All of them have now largely completed the transition to Western style-system. Adaptations would clearly be required to fit the country's particular political and economic circumstances. Detailed design issues would also have to be addressed. These include the level of equalization to be attempted in the funding of discretionary expenditures and the specific formulas to be used for capitation-based financing in the social sectors. Such issues could be addressedinfollow-on work by the World Bank. 24