WPS5177 Policy Research Working Paper 5177 Infrastructure and Economic Growth in Egypt Norman V. Loayza Rei Odawara The World Bank Middle East and North Africa Region Social and Economic Development Group (Egypt) & Development Research Group Macroeconomics and Growth Team January 2010 Policy Research Working Paper 5177 Abstract In the past half a century, Egypt has experienced this paper suggests that an increase in infrastructure remarkable progress in the provision of infrastructure in expenditures from 5 to 6 percent of gross domestic all areas, including transportation, telecommunication, product would raise the annual per capita growth rate of power generation, and water and sanitation. Judging gross domestic product by about 0.5 percentage points from an international perspective, Egypt has achieved in a decade's time and 1 percentage point by the third an infrastructure status that closely corresponds to what decade. If the increase in infrastructure investment did could be expected given its national income level. The not imply a heavier government burden (for instance, present infrastructure status is the result of decades of by cutting down on inefficient expenditures), the purposeful investment. In the past 15 years, however, a corresponding increase in growth of per capita gross worrisome trend has emerged: Infrastructure investment domestic product would be substantially larger, in has suffered a substantial decline, which may be at odds fact twice as large by the end of the first decade. This with the country's goals of raising economic growth. highlights the importance of considering renewed Improving infrastructure in Egypt would require a infrastructure investment in the larger context of public combination of larger infrastructure expenditures and sector reform. more efficient investment. The analysis provided in This paper--a joint product of the Middle East and North Africa Region, Social and Economic Development Group (Egypt); and the Macroeconomics and Growth Team, Development Research Group--is part of a larger effort is part of a larger effort to understand the relationship between infrastructure and economic growth.. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at nloayza@worldbank. org or rodawara@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Infrastructure and Economic Growth in Egypt* Norman V. Loayza Rei Odawara World Bank George Washington U. and World Bank * This study is part of the Public Expenditure Review process, sponsored by the World Bank in conjunction with the Egyptian Ministry of Finance and funded partly with the Dutch Government's Trust Fund established for this purpose. We also acknowledge the financial support of the Japanese Consultant Trust Fund, which funded Rei Odawara's participation in the project. For insightful advice and support throughout the preparation of the study, we are especially grateful to Santiago Herrera (Lead Country Economist for Egypt). We also thank the useful comments and suggestions provided by Farrukh Iqbal, Xavier Devictor, Michel Bellier, Sherine H. El-Shawarby, Luis Servén, César Calderón, Auguste Kouame, Alex Kremer, Ziad Nakat, Paul Noumba Um, Mustapha Rouis, Andrew Stone, Marijn Verhoeven, and other colleagues and seminar participants at the World Bank. We gratefully recognize the Egyptian Center for Economic Studies (ECES) and its director Dr. Hanaa Kheir-El-Din for hosting the dissemination of the study in Cairo; and Dr. M. Fathy Sakr (Principal Advisor to the Minster of Economic Development), Eng. Fawzeya Abou Neema (First Under-Secretary, Ministry of Electricity), and other participants of the ECES seminar for their valuable feedback and suggestions. Tomoko Wada provided excellent research assistance. The views expressed in this paper are those of the authors and do not necessarily reflect those of the World Bank, its Board of Directors, or the countries they represent. Infrastructure and Economic Growth in Egypt I. Introduction Over the last five decades, infrastructure in Egypt has experienced a remarkable improvement. This has undoubtedly supported the relatively strong economic growth performance of the country, as well as contributed to the progress in social and economic well-being of its citizens. Despite this progress, in the last years there has been a slowdown or even a decline in some areas of infrastructure, particularly power generation and transportation. Associated with this decline, capital expenditures in Egypt have been reduced in the last decade, raising concerns that the country may have reached an unsustainably low level of infrastructure investment. This paper analyzes the situation, trends, and effects of infrastructure in Egypt. It does so by placing the Egyptian experience in an international context. The paper examines the major sectors of infrastructure, including electricity generation, transportation, telecommunication, and water and sanitation. It assesses how infrastructure measures in Egypt currently compare with the rest of the world, particularly countries at similar level of economic development. It also reviews the historical trends in these infrastructure measures and projects their likely improvement in the future. Then, the paper describes the trends in infrastructure expenditures in Egypt, comparing to the extent possible their differing patterns across types of infrastructure and for different times in the last five decades. To serve as benchmark, the paper also presents the trends in infrastructure expenditures in a few other countries, paying special attention to the increasing role in private investment in certain infrastructure sectors. The paper links the progress in infrastructure with an increase in the rate of economic growth in the country. This is a central task of the paper. It consists of first estimating how infrastructure investment expenditures have led to infrastructure improvements and this, in turn, to higher economic growth. Estimating the connection between expenditures and growth cannot be done in a single step for lack of sufficient data. Thus, it is done in two steps. First, using panel (cross-country and time-series) data, the paper estimates the link between the level of infrastructure and economic growth. 2 Using panel data allows considering various complexities in assessing the impact on economic growth, chief among them controlling for the effect of other growth determinants. In the second step, the paper evaluates how expenditures in infrastructure translate into improvements in the level of infrastructure. This is a limited and direct exercise for which only Egypt-specific data are used. Focusing on Egypt is both a necessity (given that comparable data do not exist for a sufficiently large group of countries) and an advantage (given that the expenditure-improvement connection may vary significantly across countries). The paper takes great care to make the two chains in the estimation process consistent with each other. For instance, the choice of infrastructure measures used in the growth analysis is driven by the existing data on infrastructure expenditures in Egypt. Since the latter are historically presented for only two categories, power generation and transport/telecommunication, the paper constructs indices of the level of infrastructure aggregated at exactly those categories. Then, using the estimates just described, the paper generates some projections for the likely impact of further increases in infrastructure expenditures on the rate of economic growth of Egypt. It considers a couple of scenarios, including a moderate and a strong increase in expenditures. In assessing the growth impact of higher infrastructure investment, the paper considers the importance of evaluating the fiscal burden that these expenditure increases may entail. The rest of the paper is structured as follows. Section II provides a review of the literature on, first, the connection between infrastructure and economic growth across the world, and, second, related issues that are especially important to Egypt. Section III analyzes the situation of infrastructure in Egypt, introducing indicators that quantify and place the Egyptian situation in an international context. Section IV presents some new results on the relationship between infrastructure measures and economic growth. Section V reviews the trends in infrastructure expenditures in the country, analyzes how they are related to improvements in infrastructure measures, and finally estimates the economic growth impact of further infrastructure improvements in Egypt. Section VI presents a summary of the paper and offers some concluding remarks.1 1 The appendices provide information on sources of data used in the paper, present additional cross-country comparisons, describe details on econometric methodologies, and present additional regression analysis. 3 II. Literature Review A. The Impact of Infrastructure on Growth The impact of infrastructure on long-run economic growth has been studied extensively. The basic theoretical framework of the impact of public capital on economic growth was developed first by Arrow and Kurz (1970). Based on this framework, the endogenous growth literature shows that an increase in the stock of public capital can raise the steady state growth rate of output per capita, with permanent growth effects (Barro 1990, 1991, and Barro and Sala-I-Martin, 1992). Other studies focus on the differential impact of capital and current components of public spending on growth (Devarajan et al., 1996), showing a positive effect from capital expenditures and often negative effects from current or consumption expenditures. The body of empirical literature on infrastructure and its link to economic performance has adopted various estimation methodologies on a variety of data (panel and time series data) and measures of infrastructure.2 A majority of the literature finds a positive impact on the relationship between infrastructure and output, growth, or productivity. However, the results largely depend on the measures of infrastructure employed in the analysis. The empirical literature uses various measures of infrastructure such as physical units of infrastructure, stocks of public capital, and infrastructure spending flows.3 Straub (2008) claims that the positive effect of infrastructure on growth is often obtained when physical indicators of infrastructure are used. The results are not so clear when infrastructure spending flows are used as proxies for infrastructure.4 This might be due to the fact that political and institutional factors (i.e. inefficient government) (not the level of infrastructure investment) often affect the level of infrastructure stocks 2 Empirical studies in regards to the impacts of infrastructure on growth and productivity include: Aschauer (1989), Easterly and Rebelo (1993), Canning and Fay (1993), Canning (1999), Sanchez-Robles (1998), Demitriades and Mamuneas (2000), Roller and Waverman (2001), Esfahani and Ramirez (2003), Calderon (2008), Calderon and Serven (2004, 2008). 3 Some studies use the indices of infrastructure as proxy for infrastructure. Sanchez-Robles (1998) constructs an index of infrastructure stock by using transportation facilities, electricity generating supplies, and communications. Calderon (2008) and Calderon and Serven (2004, 2008) build synthetic indices that captures the stock of the different types of infrastructure assets and the quality of service in different infrastructure sectors. 4 Straub (2008) surveys both theoretical and empirical papers linking infrastructure and growth. 4 and the quality of services in different infrastructure sectors, particularly in developing countries. Calderon and Serven (2008) and Calderon (2008) analyze the impact of infrastructure on economic performance of African countries. Using panel data for a large sample of countries for the period 1960-2005, they employ growth regressions estimated through a Generalized Method of Moments estimator and evaluate the impact of several types of infrastructure assets, as well as measures of quality of their services. Their findings suggest that both infrastructure stock and quality are positively and significantly related to real GDP per capita growth. In addition, the latter study evaluates the impact of a higher infrastructure development in African countries over the last 15 years (comparing 2001-05 to 1991-1995). At the country level, Egypt has attained the largest contribution of infrastructure development to growth (1.51%) among Northern African countries, with a rate higher than the average of the Africa region (0.99%). Finally, infrastructure also affects economic performance through an indirect channel related to income distribution. Higher access to infrastructure services often helps reduce income inequality by lowering logistics costs or raising the value of human capital or land (Estache, Foster and Wodon, 2002, Estache (2003), Calderon and Chong, 2004, Calderon and Serven, 2004a, 2008, Galiani et al., 2005). B. The Impact of Infrastructure in Egypt The share of public investment to GDP in the Middle East and North Africa (MENA) region exceeds other regions in the developing world. In particular, historically Egypt has had a high share of public investment in infrastructure even among MENA countries. Over the last few decades, however, public infrastructure investment in Egypt has been falling, and the decline in public investment has not been compensated by a rise in private investment.5 Reflecting the specific situation of Egypt, the impact of infrastructure in the country has been discussed from the following perspectives in the literature. 1) 5 IFC (2003) reports that private participation in infrastructure investment in the MENA region declined in the 2000s compared to the 1990s and in fact its cumulative investment for 1990-2001 is smaller than other regions, even smaller than Sub-Saharan Africa. The World Bank (2003) concludes that the MENA region especially suffers from an unfavorable investment environment that prevents private participation in the last decade. 5 infrastructure as one of the determinants and binding constraints of growth performance, 2) the importance of infrastructure in order to improve the business climate and encourage private participation in the economy, and 3) the effect of infrastructure on private investment. The first strand of the literature attempts to identify the determinants and constraints of economic performance in Egypt over time. Using diagnostic approach developed by Hausmann, et al. (2005) and growth regressions, Dobronogov and Iqbal (2005)6 and Enders (2007) find that inadequate infrastructure is not among most urgent binding constraints in Egypt, but inefficient financial intermediations and high public debt are critical growth constraints.7 Kamaly (2007) analyzes the sources of growth in Egypt for the last three decades (1973-2002)8. Using a new consistent estimate for capital stock and growth accounting technique, he claims that capital stock seems to be the most important source of growth, and the downward trend in real output growth since the 1980s could be attributed to the slowdown in capital growth, including infrastructure. Nabli and Vefganzounes-Varoudakis (2007) investigate the linkage between economic reforms, human capital, infrastructure, and economic growth in the MENA region. 9 Employing growth regressions that include different composite indicators of infrastructure10 on panel data consisting of 44 countries from 1970 (or 1980) to 1999, they find that the contribution of infrastructure on growth is substantial. At the country level, comparing the period for 1980-89 to 1990-99, the contribution of infrastructure to growth in Egypt fell from 1.0 to -0.9, while that of the average of MENA countries fell from 1.4 to 1.0. The drop in the contribution from infrastructure in Egypt was due to the decline in their measure of road networks experienced in the 1990s. 6 They conducted growth regressions that determine Egypt's GDP per capita growth for 1986-2003 on key variables, but they did not include infrastructure as one of explanatory variables. 7 Egypt has a dense road network, including the new Cairo-Alexandria highway, major ports in Suez and Alexandria, and a new airport in Cairo. Electricity is cheap and highly subsidized, as is natural gas (Enders, 2007). 8 Kheir-El-Din and Moursi (2003) also examined the growth experience in Egypt by using the data from 1960-1998. 9 They generate the aggregate indicators for economic reforms, human capital, and physical infrastructure using principal component analysis. 10 The physical infrastructure indicator is based on the density of the road network (in km per km2) and the number of phone lines per 1,000 people (both are in logs). 6 As for the second strand of the literature, the World Bank report (2008) emphasizes the importance of securing long-term fiscal sustainability in its basic infrastructure sectors while sustaining the quality of service delivery in them. Moreover, Ragab (2005) argues that better performance of infrastructure and more efficient regulatory framework are critical to improve the business climate and promote private domestic and foreign investment in Egypt. The third strand of the literature has analyzed the effects of public investment on private capital formation to identify whether public infrastructure investment complements or crowds out private investment in Egypt. The majority of previous studies on this topic find a positive impact of public infrastructure investment on private investment. Shafik (1992) claims that public investment tends to crowd in private investment through infrastructure investment in Egypt. Dhumale (2000) finds a positive effect of public infrastructure investment on private investment in the non oil-exporting countries (including Egypt) within the MENA region, while a crowding out effect in oil- exporting countries. In a recent paper, Agenor et al. (2005) investigate the impact of public infrastructure on private investment in three countries in the MENA region (Egypt, Jordan, and Tunisia). They use a vector auto regression (VAR) model that accounts for both the flows and stocks of public infrastructure and controls for simultaneous interactions between these variables and private credit, output, and the real exchange rate.11 The impulse response analysis indicates that public infrastructure has both flow and stock effects on private investment in Egypt. III. The State of Infrastructure in Egypt: International Context A. Cross-Country Comparison Using Current Data We start by presenting cross-country data of various infrastructure indicators for different sectors in order to compare the performance of Egypt with the rest of the world. 11 They propose two aggregate quality indicators of infrastructure: an "ICOR-based" and "excess demand" measures. They combine these two quality indicators in order to derive the composite index by using the principal component analysis technique. 7 We collect a pooled data set of cross-country observations for 150 countries12 using the latest available data of each indicator. As for the measures of infrastructure assets, we select different indicators in stock and quality of services from four infrastructure sectors: transport, telecommunications, electricity, and water and sanitation. All the indicators used in this section are the following. (a) Transport: total road length in km, normalized by square root of the county's 1,000 workers multiplied by its mean arable land13 (in logs), paved roads (the ratio to total road length), quality of roads, quality of railroads, quality of port facilities, and quality of air transport. (b) Telecommunications: main phone lines per 1,000 workers (in logs), cell phone lines per 1,000 workers (in logs), telephone faults per 100 main lines, and waiting list for main line installation as ratio of main lines14. (c) Electricity: electricity generating capacity (EGC), megawatts per 1,000 workers (in logs), power loss (% of total output), access to electricity (% of electrification rate), and quality of electricity supply. (d) Water and sanitation: access to improved water source (% of population with access), and access to sanitation facilities (% of population with access). Figure 1 provides correlations between various indicators for different infrastructure sectors by using per capita real income level (average of 1995-2007) in PPP terms and indicates the expected level of different infrastructure indicators at a given level of economic development across countries. Panels (a) through (c) display that Egypt is located above or on the predicted regression line, except for total road length and 12 In this exercise, we exclude countries with less than 1 millions population. The pooled data set is unbalanced. 13 The country's arable land varies over time. Thereby, we use mean arable land for the period 1971-2005 for each country. 14 This indicator serves as a proxy of unmet demand for main line installation. 8 access to sanitation facilities, which are located just below the line15. The results suggest that Egypt has attained (or exceeded for some cases) the level of infrastructure performance, expected to achieve at a given level of development in comparison with the rest of the world. In Appendix 2, we provide two additional sets of figures to examine whether these results remain the same when we compare Egypt with a group of fast growing countries16 (in Figure 1-A), and when we use per capita real income growth (in Figure 1- B) instead of per capita real income level. The former confirms that infrastructure performance in Egypt has achieved what is expected (or more than expected in some cases) at a given level of development even compared with a group of fast growing countries. 15 As for telephone faults, waiting list, and power loss, the lower value means high quality in services. Thereby, Egypt being below the predicted regression line indicates that the performance of these indicators in Egypt is better than the expected level at a given level of economic development. 16 There is a criteria used to select the sub-group of countries. Using real GDP per capita data for 1983- 2005, we calculate average growth rates for each country. The sub-group consists of countries that satisfy greater than median real growth rates, which are close to the average real growth rates of Sweden (0.019). 9 Figure 1. Full-sample, correlations between infrastructure indicators vs per capita GDP, PPP (constant 2005 int'l $) (a) Transport lo of total ro len (sqrt o 1,00 w rkers x a le land) OMN AZE 4 100 rab JOR THA MUS DNK AUT CZE ITAHKG ARE ISR DEU SGP SVN FRA GBRCHE IRL UKR BGR TWN BGD GEO GRC SVN KGZ ARM LTU EST BEL UZB BLR SVK KOR MDA Paved roads (% of total roads) JPN NLD KWT 3 LVA IRL NOR KAZ HRV SWESGP 80 JAM AUT CHE CHN EGY MYS LVA CRI BWA NLD DOM BEL NOR NZLFRA PRT JPN 0 o LTU AUS USA TJK JAM IRN HUN PNG CZE ITA FIN GRC DZA BRA MYS GBR DNK MKD TUN POL 2 GIN YEM ZAF POL GAB PRT ESP PAK FIN USA NZL SVK 60 BOL ROM IDN RUS SWE CHL ISR GMB BLR TUR HRV NPL MAR LBY VNM PHL ECU PAN COL f BIH ROM TTO ZAR CAF CIV IND LSOMRT LKA EGY PER BGR SLE MDG GHA ZMB KGZNIC AGOJOR DOM DZA ARGMEX IND ESP TCD SYRSLV 1 GNB IDN IRN MUS HUN BEN 40 RWA KEN PAK LKA AUS BDI HND ALB ad gth TZA KHM CHN UKR KAZ RUS GTM MAR AGO MLI TUN URY SDN GTM URY BWA MEX TUR CAN NGA TGOBFA VEN PAN SEN SEN NGA ARG SAU UGA ARE OMN NPL TGOBFA NER 0 HTI CRI ZAF CHLEST 20 ERI PHL HND HTI THA ETHRWA GMB MOZ MLI LSO SYRSLV CMR GHA ZMB KEN CMRYEM ECU COL ETH NER GNB GIN CIVVNM MDG BEN MRT NIC COG NAM PER BDI PRY GAB SLE TZA BOL BRA LBR KHM -1 UGA PNGMNG BGD ZAR CAF TCD 0 g 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 8 8 CHE DEU SGP FRACHE JPN FRA DEU DNK HKG HKG JPN USA 6 AUT 6 BEL TWN NLD ARE NLD MYS GBR CAN BEL DNK FIN SGP SWE TWN SWE Quality of railroads FIN PRT ESP CHL KOR AUS KOR CAN Quality of roads ISR KWT MYS AUTUSA SLV THA JOR TUNZAF NAM LTU ESP GBR HRV IND SVK AUS NOR SVNGRC NZL NOR TUN MUS ITA PRT CZE 4 LTU LVA 4 CHN PAN URY RUS ISR MEX SVKCZE CHN AZE UKR NAM THA EST SVN POL DOM GTM JAM TUR BWA EST IRL PAK ZAF HUN NZL PAK MAR AZEEGY LVA HUN EGY BGR KAZ BWA GRC IND ARM MKD ARGPOL TTO MAR ITA BGD TZA GMB HND GEO HRV IRL KHM TZA TJK ROM GEO LKA DZA ECU MDA CHL MWI GA U PHL PER COL MNG LKA IDN PAN MLI BEN VNM NIC VEN MEX BDI ETH MDG ZMB NGA NPL LSO BFA KEN BRA BGR UKR KAZ RUS BGD VNM MKD TUR DZA ARG 2 TJK IDN CRI ROM KGZ BFA KENNGA JOR ARM ARE 2 BOL ALB MRT KWT MOZ KGZ MRT MNG PRY CMR AGO BIH MWI MOZ MLI KHM CMR PHLBOL PER BRA BIH MUS MDA MDG ZMB ETH UGA BEN LSO AGO URY TCD BDI GMBTCD GTM COL CRI JAM VEN SLV ALB DOM TTO NPL NIC HND PRY ECU 0 0 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 10 Figure 1 (continued). Full-sample, correlations between infrastructure indicators vs per capita GDP, PPP (constant 2005 int'l $) (a) Transport (continued) 8 7 SGP DEU HKG NLD FRA ARE GBR JPN DNK SGP FIN USA NLD HKG DEU CHE 6 BEL DNK MYS AUS CAN BEL NOR FIN JPN TWNSWE 6 FRA ARE SLV ZAF PRT ESP ISR PANMYS SWE NZL AUT Quality of port facilities USA CHL IRL Quality of air transport CAN THA KOR NZLGBR NOR TWN JAM DOM GRC JAM EST KORESP CHE MUS AUS AZE PAN LVAEST NAM CHL ISR IND CZE KWT 5 TUN JOR TUN THA MUS LVA PRT GRC AUT KEN NAMCOL HND ZAF URY SVN EGY TUR MEX KWT IRL ZMB PAK BRA JOR TTO LTUHUN SVN 4 GMB LTU ETH MAR GTM PAK MAR CRI RUS ITA LKA DOM CHN AZE SLV GTM TZA KHM IND GEO EGYUKR ARGPOL MEX RUS SVK BGR HRV CZE IDN MDA PHLLKA ECUKAZ HRV 4 KEN HND TUR DZAROM TTO HUN ITA GMBKHM VNMNIC VEN NGA COL ECUBRA BWA TZA ARM ARGPOL BWASVK VNM PHL VEN KAZ CHN ROM MOZ BGDBEN MRT AGOPRY DZA MDG NIC IDN CMR PER CRI MOZ NGA GEO BGR ALB UGA NPL PER 2 BDI MDG AGO URY BFALSO UGA ZMB MKD KGZ ALB MWI 3 TJKKGZ MNG BIH BOL UKRMKD PRY ETH MLI NPL TCD MDA BOL MWI BFA TJK MNG ARM MLI BGDBEN MRT BIH BDI LSO CMR TCD 0 2 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 11 Figure 1 (continued). Full-sample, correlations between infrastructure indicators vs per capita GDP, PPP (constant 2005 int'l $) (b)Telecommunications 8 8 ISR ITA log of cell phone lines (lines per 1,000 workers) HKG log of total mainlines (lines per 1,000 workers) LTUHUN SWE ESTCZE ESPNLD NOR BEL AUT GBR SWE DEU CHE FRA GRC USA JAM SVN FIN IRL PRT GRC SGP DNK DEU FRA AUS DNK CAN AUS BGR CHLSVK KOR JPN KWT GBR ISR ITAHKG BEL IRL KORESPNLD NOR HRV LVA TUR POL NZL CHEARE BGR HRVHUN SVN FIN SGP NZL AUT JPN MKD MYS SAU USA CRI BLR MUS EST PRT TUR POL CZE TUNZAF RUS ROM MUS TTO UKRMKD ARG SVK SAU IRN URY LVA RUS CHL TTO LTU PHL JOR MAR ALB THAPANMEX GTM BIH BRA BWA ARG GAB OMN CAN MDA ARM EGY BIH BRA MEX ARE 6 CHN COL SYRSLV ROM LBN JOR JAM KAZ MYS KWT SLV PRY DOM VEN UKR ECU LBN GEO GTM TUNZAF ECU PAN VEN OMN CRI VNM AZE DOM MRT MDA CHN COL BLR BOLNAM 6 KGZ ALB DZA NAM THA PER BWA NIC GEO SWZ MNG AZE DZAURY EGY PERKAZ BOL YEM HND SWZ MNG LKA MAR GMB SEN IDN SYR LKA COG SDN NIC PHL PRY IND LSOCIVCMR HND PAK IDN GHA NGA YEM GAB KEN ARM IRN 4 SEN LSO TGO KHM NPL CIV BEN KGZVNM IND AGO HTI LAO GHA MRT TZA ZMB SDNPAK UGAHTI LBY TGO ZAR MLI LAO PNG LBR ERI BEN ZMB NGA KEN GNB MOZ BFA TJK 4 MWI MLI ETH MMRBFA CMR AGO UZB BGD MDGBGD COG MWI RWA GIN TCD TZA CAF 2 BDI MOZMDG NER CAFUGAGIN RWA BDI NER TCD ERI NPL PNG 2 0 ETH ZAR MMR 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 100 MMR SDN HND 1 ERI aiting list (the ratio of waiting list to m lines) IND NPL Telephone faults (per 100 m lines) ain TJK KEN 80 .8 ain ZMB LSO GHA 60 .6 ETH UZB CIV NER NPL ETH GNB TGO YEM SWZ SWZ JAM LSO ALB .4 BDI 40 MOZ TCD ALB KEN DZA ERI ARM MWI LKA CAFMDG GMB ZAF MMR BEN AZE KAZ BGD MNG NAMUKR GAB MUS .2 BFA JAM UKR 20 COL MOZ BFA MDA LAO RUS TZA GHA MRT MNG LKA SYR MAR TUNBLR MYS CHL CAF ZMB KGZ GEO ARM ROM IRN BWA BLR NGA GEO IDN BIH THA GAB RWA SLVECU PAN TUR ROM EST SVN COL KAZ EGY TUN POL TZA CIV PAK GIN TJK SEN IND BGR POL JOR HRV LVASVK PRT GRC USA LTUHUN UZB AZE SLV PRY MUS SAU MKD AUS CZE ESP IRL DNK MDG BOL CRI LVA HUN SVNGRC KWT NAM ECUBRA CHLESTOMN JOR PER ZAF MYS SVKCZENZLGBR SGP PAN LTU ARG TUR 0 W BDI TGO BEN SDN NIC PHL PRY SEN MDA THACRI RUS GBR BEL KWT AUT PNG MAR URY HRV SWE DNK NLD FRA FIN JPNCHE TWNAUT NOR KOR ITAHKG ARE CAN DEU AUS HND DZA VEN BGR BRA MEX TWNHKG ARE OMN JPN SGP SAU GIN KOR 0 EGY 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 12 Figure 1 (continued). Full-sample, correlations between infrastructure indicators vs per capita GDP, PPP (constant 2005 int'l $) (c) Electricity 80 NOR 2 SWE KWT CAN FIN USA COG log of EGC (megawatts per 1,000 workers) DNK AUS ISRESP CHE SAU FRA EST OMN JPN BGR LTU CZENZL BEL SGP AUT PRY ARM UKR ITAHKG RUS SVK SVN DEU IRL BIH KAZ LVA HUN TTOKOR GRCGBR ARE NLD TJK ZAF MYS GEO MKD LBNLBY ROM HRV PRT Power loss (% of total output) KGZ BLR ARGPOL IRN CHL VEN 60 MDA AZE DOM JAM URY ALB TUR MUS UZB BRA MEX PAN 0 JOR SYR CRI MNG EGY TUN THA GAB HTI ECU COL CHN GTM PER HND NAM MAR SLV DZA ZMB PHL PAK LKA SWZ LAO NIC IND BOL MOZ VNM IDN ECU 40 LBR PNG BWA YEM GHA CIV MDA -2 MRTCMR ALB ZAR LSO NGA AGO TGO NGA MMR KEN SDN COG DOM URY BGD GIN HTI SEN MLI KGZ MWI NPL SLE TZA IND VEN LBY GMBBEN PAK NIC GNB ERI TZA YEM SYR HND BFA 20 MDG MKD ETH UGA CAF MMR NPL CMR NAMCOL LVA -4 NER KEN PAN GAB BDI CIV SDN TJK GHA SEN LKA MAR UKR KAZ MEX ARM GEO AGO DZA IRN BIH BRA HRV OMN TUR ARG LBN BEN IDN JOR PHLBOL SLV AZE BGR EGY JAMBLR RUS EST TUN CRI HUN NZL HKG KWT TGO RWA TCD MOZ ETH BGD VNM UZB PER ROM POL BWA PRT ESP IRL NOR GRC CHN THA CHL SAU GBR SWE LTU CZE ITA CAN ARE ZAF MYS TTOSVN DEU USA CHE AUS KHM ZMB PRY GTM JPN SGP FRA TWNAUT SVK KOR FINBEL DNK NLD ZAR ISR 0 -6 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 150 8 Access to Electricity (% of population) DNK JPN DEU FRA CHE NLD HKG GBR ARE FIN SGP BEL NOR AUT ISR SWE KWT CAN KOR AUS USA SVKCZE ESP IRL Quality of electricity supply 6 TWN SVN PRT MYS JOR TUN MUS CHLEST 100 ZMB THACRI URY JOR TUN CRI MYS CHN LBN KWT SGP ITA EGY THAURY CHL TTO TWN VEN DZA IRN LBY BRA SAU ISR LTUHUN NZL ARG MUS OMN MAR SLV EGY NAM HRV PAN LVA BRA BWA GRC SYRDOMECU ARE PER JAM PAN PRY COL GTM COL VNM MAR JAM KAZ MKD POL TTO PHL SLV DZA VEN PHLARM BIH ZAF MEX TUR 4 BOL CHN AZE BGR ARG PER ZAF LKA ROMRUS NIC ETH UKR MNG LKA BOL PAK VNM IDN MDA HND HND MOZ MLI KGZ LSO BFA KEN IND PRY ECU IND PAK IDN MNG MRT NIC GEO 50 GHA CIV MWI BEN CMR NGACMR GAB GMBKHM NPL TZA AGO BWA 2 HTI YEM MDG NPL BGD SEN NAM BDI TJKNGA SDN UGABGD ALB DOM ERI BEN KHM COG KOR TCD ZMB ETH TGO MDG KEN AGO MMR LSO TZA UGA ZAR MWI MOZ BFA 0 0 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 13 Figure 1 (continued). Full sample, correlations between infrastructure indicators vs per capita GDP, PPP (constant 2005 int'l $) (d) Water and Sanitation 120 Access to Improved Water Source (% of population) Access to Sanitation Facilities (% of population) 100 URY HRVHUN BGR PRT GRC USA FIN SGP JPN DEU SVKCZE ESPNLD AUS DNK AUT CAN CHE SWE UZB KAZ ALB THACRI LBY ARE BIH DZA MUSMYS EST CHL 100 TJKKGZ ARM UKR BLR ARG GEO SYR TTO MKD LBN HRVHUN ISRESPNLD ARE URY LVAEST PRT FIN SGP MUS GEOJOR BIH BLR MYS SVKCZE GRC USA BGR EGY ARM ALB THACRI RUS GBR FRA AUS DNK AUT CAN CHE DEU NOR JPN SWE MKD TUR RUS TUR GTM UKR KAZ MEX DOM ECU BWA ARG CHL TTO LKA SLV JOR TUN GTM JAMECU TUN MMR VNM PHL COL NAM JAM ZAF PAN BRA MDA PHLAZE DOM MEX NPL PAK MDA KGZIND SYR COL BRA LVA UZB CHN ROM GAB GMB BOL HND MAR SLV DZA PER MAR PER PAN ROM LKA PRY 80 MMRGHA CIV NIC IDN BGD HND EGY LSO SEN AZE PRY VNM CHN MWI BFA MNG MWI ZAF BDI COG PAK GIN SDNCMR TJK ZMB GMB CMR IDN SWZ 50 CAF RWA YEM LBR UGA BENKHM MNG LAO NIC AGO BWA 60 ERI TGOMLI MRT LAO SWZ MLI YEM PNG GNB ZMB HTIKEN KEN BOL BDI TZA SLE BGDLSOSDN NAM GAB AGO GNB MOZ TZA MDG TCDNGA LBR ZAR CAFUGA BEN NGA ZAR SEN IND NPL KHM ETH MOZ NER MRT CIV 40 PNG RWA GIN HTI COG BFA ETH MDG SLE TGO GHA NER TCD ERI 20 0 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 14 Table 1. Definitions and Sources of Infrastructure Quantity and Quality Indicators Variable Definition Year Source roads Length of total roads (km, sqrt of 1,000 workers x mean arable land for 1971-2005) 2004 International Road Federation (IRF) paved roads Paved roads (the ratio of paved roads to total road length) 2004 International Road Federation (IRF) ml The number of main phone lines (per 1,000 workers) 2004 Int'l Telecommunications Union (ITU) cell The number of cell phone lines (per 1,000 workers) 2004 Int'l Telecommunications Union (ITU) telf Telephone faults (the number of reported telephone faults for the year Avg. of 2001-06 Int'l Telecommunications Union (ITU) per 100 main phone lines) wl Waiting list for main line installation (the ratio of waiting list to main lines) Avg. of 2000-04 Int'l Telecommunications Union (ITU) egc Electricity generating capacity (megawatts, per 1,000 workers) 2004 Statistical Yearbook, United Nations. US- Energy Information Administration pl Power loss (% of total output) 2004 WDI, The World Bank. q_roads Quality of roads 2006 Global Competitiveness Report q_railroads Quality of railroads 2006 Global Competitiveness Report q_ports Quality of port facilities and inland waterways 2006 Global Competitiveness Report q_air Quality of air transport 2006 Global Competitiveness Report q_elec Quality of electricity supply 2006 Global Competitiveness Report elec_accesss Access to electricity: Electrification rate (% ) 2006 World Energy Outlook water Access to water: Improved water sources (% of population with access) 2006 WDI, The World Bank. sanitation Access to sanitation: Improved sanitation facilities (% of population with access) 2006 WDI, The World Bank. 15 Table 2 displays pairwise correlations of the components of infrastructure by sector (used in Figure 1) and the correlations between the representative components from each sector. As shown in the top panel of Table 2, all the components within sector are significantly mutually correlated at either the 5 or 10 percent level of significance. The bottom panel also indicates that the representative components are correlated across sectors at the 5 percent significance level. Among all the indicators listed above, our main focus is on the following five indicators: total road length in km per square root of the country's 1,000 workers multiplied by its mean arable land, paved roads as ratio of total road length, the number of main phone lines per 1,000 workers, electricity generating capacity in megawatts per 1,000 workers, and power loss (% to total output). These five indicators are used to construct sectoral infrastructure indices for transport, telecommunications, and electricity, which are explained in detail in Part C in this section. The definitions and sources of the entire set of infrastructure indicators used in Figure 1 as well as Figures 1-A and 1-B in Appendix 2 are shown in Table 1. B. Cross-Regional Comparison Using Trend Data In this section, we assess the time trends in the main infrastructure indicators over the last few decades, comparing Egypt with developing countries as well as the group of both developing and developed countries (called "World"). We select five main indicators of infrastructure quantity and quality from the three core infrastructure sectors, transport, telecommunications, and electricity. A quality indicator for telecommunications, telephone faults per 100 main lines, is excluded as the time dimension of the indicator is very limited and is only available for the last few years17. 17 The cross-country comparison of telephone faults by using the most recent data is shown in Panel (b) of Figure 1. 16 Table 2. Pairwise correlation of infrastructure measures 1.) Components by sector (a) Transport roads (in logs) paved roads q_roads q_railroads q_ports q_air roads (in logs) 1 paved roads 0.2701** 1 q_roads 0.5106** 0.5382** 1 q_railroads 0.5787** 0.5787** 0.7769** 1 q_ports 0.5487** 0.4610** 0.8900** 0.7579** 1 q_air 0.5506** 0.4737** 0.8565** 0.6957** 0.8690** 1 (b) Telecommunications ml (in logs) cell (in logs) telf wl ml (in logs) 1 cell (in logs) 0.8223** 1 telf -0.4902** -0.5916** 1 wl -0.3950** -0.4665** 0.1866* 1 (c) Electricity egc (in logs) pl q_elec elec_access egc (in logs) 1 pl -0.4230** 1 q_elec 0.7331** -0.6391** 1 elec_access 0.8295** -0.2005* 0.6069** 1 (d) Water & Sanitation water sanitation water 1 sanitation 0.8112** 1 Notes: ** denotes the significance level at 5 percent, and * at 10 percent. 2.) The representative component from each sector roads (in logs) ml (in logs) egc (in logs) water roads (in logs) 1 ml (in logs) 0.5727** 1 egc (in logs) 0.6374** 0.8727** 1 water 0.4902** 0.8644** 0.7785** 1 Notes: ** denotes the significance level at 5 percent, and * at 10 percent. 17 Panels (a) through (e) of Figure 2 display the evolution of the main infrastructure indicators 18 : the length of total roads per square root of the country's 1,000 workers multiplied by its mean arable land (in km), paved road (the ratio to total road length), the number of main phone lines per 1,000 workers, and EGC per 1,000 workers (in megawatts), and power loss (percentage of total output), respectively. In each case, the group median for each decade is shown. Transport: Panel (a) presents road networks as a measure of transport stock and the ratio of paved roads to total road length as a proxy of transport quality, respectively. We normalize the measure of transport stock, dividing it by the square root of the country's 1,000 labor force times its mean arable land. Although Egypt has lagged behind the other two groups in terms of total road length since the 1970s through the 1990s, the growth in total road length drastically picked up in the 1990s onwards, while that in the other two groups has stagnated. As Panel (b) shows, Egypt has far exceeded the typical country in the two comparator groups regarding the ratio of paved roads to total road length.19 Telecommunications: Egypt experienced rapid growth in its quantity indicator of telecommunications over time. As for the number of main lines in Panel (c), the gap between Egypt and World has significantly narrowed in the latest period. 18 The balanced data is used for all five indicators. 19 We also normalized total road length by 1,000 workers and obtained very similar results to those in Panel (a). 18 Figure 2. Transition of main infrastructure indicators over time (medians) a.) Total road le ngth b.) Pave d roads (the share to total roads) (s qrt of 1,000 worke rs x me an arable land) 0.9 5.0 0.8 0.7 4.0 0.6 3.0 0.5 0.4 2.0 0.3 0.2 1.0 0.1 0.0 0.0 Egypt Developing countries World (60) Egypt Developing countries World (60) (40) (40) c.) Main line s pe r 1,000 worke rs d.) EGC pe r 1,000 worke rs (me gawatts ) 350 1.2 300 250 1.0 200 0.8 150 0.6 100 0.4 50 0.2 0 0.0 Egypt Developing countries World (60) Egypt Developing countries World (60) (40) (40) e.) Power loss (percentage of total output) 0 -5 -10 -15 -20 Egypt Developing countries World (60) (40) 1971-80 1981-90 1991-2000 2001-05 Note: The balanced data are used. 19 Electricity: Panel (d) shows the trends in EGC per 1,000 workers (in megawatts). Egypt exceeds developing countries in this measure, but has fallen far behind World. The gap between Egypt and World has even widened in the recent period as growth in EGC has stagnated in Egypt since the 1990s. The quality indicator of electricity, power loss (in percentage of total output), displayed in panel (e), improved in the 1990s, but it reversed to the 1980s level in the most recent period. 20 Although a decline in electricity quality seems to be a worldwide trend, the electricity sector in Egypt shows some signs of weakening in both indicators of quality and quantity. C. Infrastructure Indices by Sector Some empirical literature that studies the impact of infrastructure on economic performance uses a single infrastructure sector (i.e. the number of main lines for telecommunications) as a proxy for infrastructure.21 Others build aggregate indices of infrastructure quantity and quality that capture the stock of different types of infrastructure assets and the quality of services in different infrastructure sectors separately (i.e. quantity and quality indices for telecommunications).22 In this paper, however, instead of focusing on a single sector or building synthetic indices capturing either quantity or quality aspect of infrastructure assets, we construct indices by major infrastructure sectors that simultaneously capture both quantity and quality features of infrastructure. In order to build the sectoral indices, we combine the information of quantity and quality indicators for the following three sectors, respectively: transport, telecommunications, and electricity.23 20 The measure of power loss is transformed in a way such that an increase in value indicates an improvement in quality. 21 Easterly (2001) and Loayza, et al (2005). 22 In a series of works by Calderon (2008) and Calderon and Serven (2004, 2008), they constructed synthetic indices of infrastructure quantity and quality, which consist of quantity and quality indicators from the three infrastructure sectors: roads, telecommunications, and power. 23 In order to construct the indices, we used time series data for infrastructure stock and quality indicators in Egypt. When we closely checked the original data, we encountered some issues. That is, some indicators had missing observations, and others fluctuated in an unreasonable way over time. In order to solve these problems, we first interpolated the missing observations for the length of roads and the paved road ratio to 20 The construction of a single infrastructure index per sector is based on the assumption that both quantity and quality aspects of infrastructure assets are closely related to and even depend on each other. Consider for example, the electricity sector: higher power loss (a proxy for quality) may be caused by having too few power plants in the country and thereby not holding enough electricity generating capacity (a proxy for quantity) to satisfy the electrical power demand. For the transport index, we select the length of total roads in km as a measure of quantity and the ratio of paved roads to total roads as a proxy for quality. As for the communications index, we use the number of main lines for the physical measure of communications. As mentioned earlier, the time series coverage of the quality indicator of telecommunications (telephone faults per 100 main lines) is very limited, and thereby only the physical measure of telecommunications is used to construct the index. Lastly, the electricity index consists of electricity generating capacity (EGC) in megawatts and power loss (percentage of total output) for quantity and quality indicators, respectively. All stock measures are normalized by 1,000 workers and then transformed in logs. The exception is total road length, which is normalized by square root of the country's 1,000 workers multiplied by its surface area. Table 3. Variance by Sector Using Principal Component Analysis Sector Variance Transport 0.7231 Telecommunications 1.000 Electricity 0.7330 Transport & Telecommunications 0.9018 In order to standardize the components for each sector, we use the principal component analysis (PCA) technique, which allows us to obtain a series of uncorrelated and normalized linear combinations of the components for each sector. We standardize a total road. Then, we smoothed out some indicators (power loss and paved road ratio), so that the time path in each component became relatively smooth. 21 pair of quantity and quality indicators for each sector (only a quantity indicator for telecommunications), and then obtain the new index. The principal component for each sector explains over 70 percent of the variance of the underlying individual indicators, as shown in Table 3. Once the three indices are constructed, the transport and telecommunications indices are combined and transformed into a single new index by using principal component analysis again. These infrastructure indices: the electricity index, the transportation index, telecommunications index, and a combined transport and telecommunications index are used in order to investigate both the effect of infrastructure on economic growth and the relationship between the performance of infrastructure assets and infrastructure expenditures in each sector (to be presented in Sections IV and V). We now observe the transition of the combined transport and telecommunications and electricity indices with respect to Egypt's own history for 1971-2005, given in the top left panel of Figure 3. Panels (a) through (c) illustrate the time paths of all five indicators used to construct these indices. As shown in the top left panel, starting with a negative value, the transport and telecommunications index was clearly on an upward trend after 1981. After turning to a positive value in the early 1990s, the index has continued to rise. In fact, the index rose more than 1.8 point for over three decades. Also starting with a negative value, the electricity index has been falling during the 1970s and through the mid-1980s (except for a sharp increase in 1982). After having a positive turn in the mid-1980s, it has been on a rise along with the transport and telecommunications index for a decade. Reaching its peak in the late 1990s, the electricity index declined until 2002 but has shown some recovery in recent years. As for the components of transport and telecommunications index, Panels (a) displays a clear upward trend in the length of total roads after the late 1980s, and the ratio of paved roads to total road length after the early 1980s. For the later, a rise in the indicator became slow down in the late 1990s. The number of main lines shown in Panel (b) also illustrates the upward trend, which became more prominent in the early 1980s, and the indictor continued to rise through recent years. Thus, all three components 22 contribute to a continuous improvement in the transport and telecommunications index over the last three decades. Finally, Panel (c) reveals the quite volatile time path of the electricity index which stems from both components, EGC and power loss. Their trends are clearly more volatile compared to the other three components used to construct the transport and telecommunications index. Having stagnated during the 1970s through the mid-1980s, EGC finally took off in 1983 and has maintained the level for a decade. After a sharp increase in 1997, it has continued to fall. As for the other component, power loss rose during the 1970s, declined through the mid-1980s, and then recovered till 1997. It then fell until recently, where some signs of improvement are visible. 23 Figure 3. Infrastructure Indices by Sector in Egypt (1971-2005) Infrastructure Indices by Sector (a) Components of Transport Index 1.0 0.0 0.9 0.8 -0.2 0.8 0.6 -0.4 0.7 0.4 -0.6 0.6 0.2 0.0 -0.8 0.5 -0.2 -1.0 0.4 -0.4 -1.2 0.3 -0.6 -1.4 0.2 -0.8 -1.6 0.1 -1.0 -1.8 0 -1.2 1971 1974 1977 1980 1983 1986 1989 1992 1995 1998 2001 2004 1971 1974 1977 1980 1983 1986 1989 1992 1995 1998 2001 2004 Transport & Telecommunications Electricity Roads, sqrt of 1000 workers x surface area (in logs) Paved roads (the ratio to total road length) (b) Components of Telecommunications Index (c) Components of Electricity Index 6.0 0.0 18.0 -0.2 16.0 5.5 14.0 -0.4 5.0 12.0 -0.6 10.0 4.5 -0.8 8.0 4.0 6.0 -1.0 4.0 3.5 -1.2 2.0 3.0 -1.4 0.0 1971 1974 1977 1980 1983 1986 1989 1992 1995 1998 2001 2004 1971 1974 1977 1980 1983 1986 1989 1992 1995 1998 2001 2004 Main lines per 1,000 workers (in logs) EGC per 1,000 workers (in logs) Power loss (% of output) 24 IV. Infrastructure and Growth: Regression Analysis The purpose of this section is to study the effect of infrastructure on the rate of per capita GDP growth. The goal is to obtain a sensible measure for this effect that can be used in the quantitative projections for Egypt. To perform our estimations, we use pooled cross-country and time-series data covering 78 countries over the period 1961- 2005. The data is organized in non-overlapping five-year periods, with each country having at most 9 observations. The panel is unbalanced, with some countries having more observations than others. We build on the panel-data growth regression literature that uses a GMM procedure to address endogeneity and control for unobserved country- specific factors. This was introduced by Arellano and Bond (1991 and 1998) and applied, for example, in Levine, Loayza, Beck (2000) and Dollar and Kraay (2004). 24 The econometric methodology is explained in detail in Appendix 3. A. Data and Regression Specification Our point of departure is a standard growth regression equation designed for estimation using (cross-country, time-series) panel data: y i ,t y i ,t 1 0 y i ,t 1 1 ' CVi ,t 2 I i ,t t i i ,t (4.1) Where the subscripts i and t represent country and time period, respectively; y is the log of output per capita, CV is a set of control variables, and I represents infrastructure; t and i denote unobserved time- and country-specific effects, respectively; and is the error term. The dependent variable (yi,t-yi,t-1) is the average rate of real output growth, that is, the log difference of GDP per capita normalized by the length of the period. The regression equation is dynamic in the sense that it includes the level of output per capita (yi,t-1) at the start of the corresponding period in the set of 24 The estimation results presented in the text follow the standard procedure. To check robustness, we applied the standard-error correction proposed by Windmeijer (2005) in exercises not presented here. The qualitative results were, however, the same. In particular, the coefficients related to the infrastructure indices remained statistically significant. 25 explanatory variables. Unless stated otherwise, all data are obtained from the World Bank's World Development Indicators. The set of explanatory variables can be divided into four groups. The first consists of the initial level of per capita GDP and is included to capture "transitional convergence," that is, the tendency of economies to grow slower as they become richer and converge to their steady state. The second set of variables accounts for the role of external conditions related to international prices and global economic conditions. To capture changes in international prices, we use the rate of change of the terms of trade; and to account for global conditions, we use a period specific dummy variable. The third group focuses on macroeconomic stability in both aggregate domestic prices and output. It includes the (log of 1 plus) the CPI inflation rate and a measure of "crisis" volatility based on negative deviations of trend beyond a certain threshold (see Hnatkovska and Loayza, 2004, for details). The fourth group consists of variables that measure structural conditions in areas such as educational investment, financial depth, trade openness, government burden, and infrastructure, the variables of particular interest for this study. Educational investment is measured as (the log of) the gross rate of enrollment in secondary school. Financial depth is proxied by the ratio of private domestic credit by private financial institutions to GDP. The outward orientation of the economy is proxied by (the log of) the volume of trade (exports and imports) to GDP. Government burden is measured as (the log of) the ratio of general government expenditure to GDP. And infrastructure is measured by the indices of electric power generation, transportation, and telecommunications, presented in detail in the previous section. Except for the variables measuring external conditions, all variables are potentially jointly endogenous with economic growth (that is, caused by previous and current innovations in per capita GDP growth, the dependent variable). B. Estimation Results We now present and discuss the estimation results. We present different variations dealing with how the measures of infrastructure are included in the growth regression. Table 4 presents the estimation results when we include the infrastructures 26 indices one at a time. Table 5 presents the results when we include the indices simultaneously in the regression. To establish the validity of our results in the context of the growth literature, let's start by analyzing the results corresponding to the standard growth determinants and the regression specification tests. In brief, the results are consistent with the previous empirical literature. Initial GDP per capita carries a significantly negative coefficient, commonly interpreted as evidence of "conditional convergence"; that is, holding constant (or conditioning for) structural and stabilization conditions, poorer countries tend to grow faster and, thus, converge towards richer ones. External shocks are also important growth determinants. Specifically, favorable terms-of-trade shocks affect positively economic growth. Representing global conditions, the period shifts (not shown in the tables to save space) indicate that the international trend in economic growth experienced a declining drift over 1960-2005, resulting in less favorable external conditions in the last three decades than in the previous ones. Suggesting a beneficial impact on economic growth, the proxies of educational investment, depth of financial intermediation, and trade openness have positive and statistically significant coefficients. Government expenditures, price inflation, and crisis volatility, on the other hand, carry negative and statistically significant coefficients, indicating the harmful consequence of government burden and macroeconomic instability. Finally, regarding the specification tests, the Hansen tests indicate that the null hypothesis of correct specification cannot be rejected, lending support to our estimation results. This is the case for all exercises presented below, and we only mention it here in order to avoid redundancy. Let's now focus the discussion on the growth effects of infrastructure. As mentioned above, in Table 4 we include the infrastructure indices one at a time. They are, respectively, the electricity index, the transportation index, the telecommunications index, and a combined transportation and telecommunications index. The latter is relevant for our purposes because the historical data for Egypt aggregates investment in transportation and telecommunications. All of them carry positive and significant coefficients. This is a strong indication that infrastructure in general is an important determinant of economic growth. It also suggests that each of the aspects of infrastructure considered in the 27 analysis ­electricity, transportation, and telecommunications--is relevant for economic growth. However, this cannot be stated conclusively given that each index can capture the significance of the rest when they are introduced one at a time in the regression. For this reason, we turn to Table 5. In Table 5, the infrastructure indices enter simultaneously in the regression. In the first column, we introduce the electricity, transportation and telecommunications indices together. In the second column, we replace the latter two by the combined transportation and telecommunication index. Interestingly, all infrastructure indices carry positive and statistically significant coefficients. This indicates that each aspect of infrastructure considered here is a relevant determinant of economic growth. The size of the estimated coefficients is also informative. The coefficients presented in Table 5 are much smaller than those in Table 4, where the indices were introduced one at a time. This confirms the conjecture mentioned above that each index represents not only its own specific area but, to some extent, overall infrastructure. In the next section, we consider in detail the quantitative importance of the estimated coefficients for the Egyptian case. Here we only consider two brief exercises. For both, we use the estimated coefficients presented in Table 5, column 1, where the three basic indices and included simultaneously in the regression. The first exercise is to consider the growth effect of changing each of the indices by 1 standard deviation of its world sample distribution. The estimated effects of this improvement are 0.89 percentage points of per capita GDP growth for electricity, 1.24 for transportation, and 1.26 for telecommunications. The second exercise is to measure the growth effect of changing each of the indices from the 25th to the 75th percentile of its world sample distribution. This is a much larger improvement and, therefore, the estimated growth effects are correspondingly stronger. They are 1.23 percentage points of per capita GDP growth for electricity, 2.05 for transportation, and 2.08 for telecommunications. 28 Table 4. Economic Growth and Infrastructure ­ Individual Effects Sample: 78 countries, 1961-2005 (5-year period observations) Estimation Method: System GMM Dependent Variable: GDP per capita Growth [1] [2] [3] [4] Infrastructure Variables: 1 Electricity Index 1.539 *** [6.436] 2 Transportation Index 2.45 *** [5.631] 3 Telecommunication Index 1.476 *** [6.687] 4 Transportation & Telecommunication Index 2.81 *** [7.171] Control Variables: Initial GDP per capita -1.592 *** -2.072 *** -1.512 *** -2.688 *** in logs [-5.175] [-5.900] [-7.133] [-7.576] Education 0.949 ** 1.008 *** 0.239 0.367 secondary school enrollment rate, in logs [2.424] [2.973] [0.813] [1.186] Financial Depth 0.403 ** 0.719 *** 1.206 *** 1.075 *** private credit/GDP, in logs [2.114] [4.226] [7.165] [5.925] Crisis Volatility -1.876 *** -1.734 *** -1.937 *** -1.761 *** 5 std dev of GDP per capita growth [-15.070] [-15.400] [-20.300] [-16.120] Government Burden -0.919 * -0.224 -0.274 0.102 government expenditure/GDP, in logs [-1.957] [-0.429] [-0.611] [0.213] Inflation -0.227 -2.033 *** -3.036 *** -2.841 *** 1+Growth rate of CPI, in logs [-0.362] [-3.189] [-5.071] [-4.561] Trade Openness 4.221 *** 2.062 *** 1.287 ** 1.586 *** (exports+imports)/GDP, in logs [9.487] [4.358] [2.432] [3.504] Growth rate of Terms of Trade 0.038 *** 0.035 *** 0.046 *** 0.045 *** log differences of terms of trade index [3.294] [2.942] [4.167] [4.019] Constant 0.733 16.826 *** 21.379 *** 26.997 *** [0.208] [3.624] [5.036] [5.750] Observations 522 522 522 522 Number of Countries 78 78 78 78 Number of Instruments 58 58 58 58 Arellano-Bond test for AR(1) in first differences 0.000 0.000 0.000 0.000 Arellano-Bond test for AR(2) in first differences 0.064 0.0517 0.134 0.072 Hansen test of overidentifying restrictions 0.182 0.357 0.471 0.435 Numbers in brackets are the corresponding t-statistics. * significant at 10%; ** significant at 5%; *** significant at 1% Period fixed effects were included (coefficients not reported). 1 First principal component of two indicators: power loss (as % of electricity output) & electricity generating capacity (in MW per 1000 workers, in logs). 2 First principal component of two indicators: share of paved roads in the overall road network & length of roads (in km, sqrt of per 1000 workers*surface area, in logs). 3 First principal component of two indicators: main telephone lines per 1000 workers, in logs & main telephone lines per 1000 workers, in logs. 4 First principal component of two indicators: Transportation index & Telecommunication index. 5 Crisis volatility is the portion of the standard deviation of GDP per capita growth that corresponds to downward deviations below the world-wide 1-std-dev threshold. 29 Table 5. Economic Growth and Infrastructure ­Joint Effects Sample: 78 countries, 1961-2005 (5-year period observations) Estimation Method: System GMM Dependent Variable: GDP per capita Growth [1] [2] Infrastructure Variables: Electricity Index1 0.749 *** 0.975 *** [5.353] [5.292] Transportation Index2 1.093 *** [3.102] Telecommunication Index3 1.097 *** [4.754] Transportation & Telecommunication Index4 2.135 *** [5.637] Control Variables Initial GDP per capita -2.452 *** -2.814 *** in logs [-8.264] [-8.092] Education 0.604 * 0.668 * secondary school enrollment rate, in logs [1.749] [1.925] Financial Depth 0.859 *** 0.849 *** private credit/GDP, in logs [5.486] [5.494] Crisis Volatility -1.679 *** -1.627 *** std dev of GDP per capita growth5 [-18.420] [-13.560] Government Burden -0.530 -0.413 government expenditure/GDP, in logs [-1.390] [-0.864] Inflation -2.918 *** -2.113 *** 1+Growth rate of CPI, in logs [-6.696] [-3.781] Trade Openness 1.881 *** 2.265 *** (exports+imports)/GDP, in logs [5.164] [5.657] Growth rate of Terms of Trade 0.060 *** 0.060 *** log differences of terms of trade index [6.785] [6.424] Constant 26.779 *** 23.764 *** [7.257] [5.336] Observations 522 522 Number of Countries 78 78 Number of Instruments 70 64 Arellano-Bond test for AR(1) in first differences 0.000 0.000 Arellano-Bond test for AR(2) in first differences 0.170 0.107 Hansen test of overidentifying restrictions 0.164 0.340 Numbers in brackets are the corresponding t-statistics. * significant at 10%; ** significant at 5%; *** significant at 1% Period fixed effects were included (coefficients not reported). 1 First principal component of two indicators: power loss (as % of electricity output) & electricity generating capacity (in MW per 1000 workers, in logs). 2 First principal component of two indicators: share of paved roads in the overall road network & length of roads (in km, sqrt of per 1000 workers*surface area, in logs). 3 First principal component of two indicators: main telephone lines per 1000 workers, in logs & main telephone lines per 1000 workers, in logs. 4 First principal component of two indicators: Transportation index & Telecommunication index. 5 Crisis volatility is the portion of the standard deviation of GDP per capita growth that corresponds to downward deviations below the world-wide 1-std-dev threshold. 30 V. Infrastructure Investment and Growth in Egypt A. Trends in Infrastructure Investment We now turn to reviewing the long-term trends in infrastructure investment in Egypt. The investment data are disaggregated by sector of origin: public and private, and by destination of industry. Infrastructure investment includes both capital expenditures (the construction of new infrastructure) and current expenditures (operations and maintenance spending). There are two kinds of time series data available for Egypt's infrastructure investment. The first data range from 1960 through 2007 with the disaggregation of two infrastructure sectors: transportation (including Suez Canal) and communications, and electricity. The second data cover the more recent period: 2003-2007, with a higher degree of disaggregation, that is, five infrastructure sectors: transportation, communications, electricity, water, and Suez Canal. Figure 4 offers a comprehensive view of infrastructure investment in Egypt relative to GDP from 1960 through 2007. Panel (a) of Figure 4 illustrates the time path of total infrastructure investment that consists of investment in two infrastructure sectors: transportation and communications and electricity, with a disaggregation of public and private investment. Total investment rose until the late 1980s, and then it declined until the mid 2000s, when it stabilized. Total investment in recent years has returned to its level in the early 1960s at roughly 5 percentage of GDP. It is clear that public investment has been a dominant force for more than four decades in Egypt. In contrast, after having stagnated for more than two decades since the 1960s, private investment finally took off in the mid-1980s. The magnitude of private investment has been growing and is roughly two-thirds of the amount of public investment in recent years. Though considerable, rising private investment in the last two decades has not fully offset the decline in public investment. From Panels (b) and (c) of Figure 4, there is an apparent downward trend in total investment which originated from a decline in public investment in both the transportation/communications and electricity sectors. Public investment in the former has been declining since the early 1980s, while that in the later since the late 1980s. 31 Figure 4. Infrastructure Investment in Egypt (1960-2007) (Percentage of GDP) (a) Total Investment 14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 (b) Transportation (incl. SC) & Communications 10.0 8.0 6.0 4.0 2.0 0.0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 (c) Electricity 6.0 5.0 4.0 3.0 2.0 1.0 0.0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 Public Private 32 Conversely, private investment in transportation and communications has maintained its level after its sharp increase in the mid-1980s and has been on the rise again in recent years. Private participation in the electricity sector is still quite limited. Figure 5 depicts the time path of more recent investment for 2003-2007 by destination of five infrastructure sectors and by sector of origin. In particular, investment data in the three core infrastructure sectors (transportation, communications, and electricity) are presented separately, allowing a more precise assessment of their recent trends.25 As shown in Panels (a) through (e) of Figure 5, the shares of investments in both water and Suez Canal are quite small compared to three core sectors. Investment in the water sector increased drastically in 2007, while that in Suez Canal has fallen since 2004. In additions, the figure reveals that public sector continues to play a major role in investment in electricity, water, and Suez Canal.26 Repeating the findings in Figure 4, public investment in transportation, communications, and electricity has been falling, while a rise in private investment has become more obvious in transportation and communications in recent years. In particular, the private sector has become a dominant player in investment in the communications sector since 2005. B. Cross-Country Comparison of Infrastructure Investment In this section, we compare the level of infrastructure investment in Egypt to that of other developing countries. Panels (a)­(c) of Figure 6 plot infrastructure investment (total, public, and private) as ratio to GDP for 2000-05 against GDP per capita in 2000 across countries.27 Total and public infrastructure investment fall as the income level of the country increases, while the opposite happens for private investment. 25 In Appendix 4, we present disaggregated investment data for transportation (including investment in Suez Canal) and communications as ratio to GDP for 1983-2007 (see Figure 4-A). These data has been made available only recently. 26 There is no private investment for water and Suez Canal. As for electricity, no private investment has been carried out after 2004. 27 Infrastructure investment data are obtained from Calderon, Odawara, and Serven (2008). Total investment consists of investment in three major infrastructure sectors: transport, telecommunications, and electricity. 33 Figure 5. Infrastructure Investment in Egypt (2003-2007) (Percentage of GDP) (a)Transportation 3.0 2.5 2.0 1.5 1.0 0.5 0.0 2003 2004 2005 2006 2007 (b) Communications 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 2003 2004 2005 2006 2007 (c) Electricity 2.0 1.5 1.0 0.5 0.0 2003 2004 2005 2006 2007 Public Private 34 Figure 5 (continued). Infrastructure Investment in Egypt (2003-2007) (Percentage of GDP) (d) Water 0.6 0.5 0.4 0.3 0.2 0.1 0.0 2003 2004 2005 2006 2007 (e) Suez Canal 0.10 0.08 0.06 0.04 0.02 0.00 2003 2004 2005 2006 2007 Public Private 35 This may indicate that poorer countries, especially through their public sectors, tend to make larger infrastructure investments until they build a certain level of infrastructure assets (a number of electricity plants and a road network). As they become richer, they tend to spend less in new infrastructure, focusing on maintaining existing assets. Panels (a) shows that Egypt is located on the predicted regression line for total investment. In contrast, for public and private investment Egypt shows interesting discrepancies with respect to the international norm (Panels (b) and (c)). While public investment in infrastructure in Egypt is much larger than what would be expected according to the country's income level, its private investment is considerable smaller than the standard set by other developing countries. Finally, we turn to reviewing the trends of infrastructure investment across countries. Figures 7.1 and 7.2 provide an overview of the level of infrastructure investment in six countries, divided into two groups: India (IND), Pakistan (PAK), and Indonesia (IDN) in Figure 7.1, and Egypt (EGY), Turkey (TRK), and South Africa (SA) in Figure 7.2. Panels (a) through (c) of Figures 7.1 and 7.2 offer the trends of total, public, and private investment as a share of GDP for over the last two decades, respectively. The top panels of Figures 7.1 and 7.2 illustrate that total investment has been slightly falling over time. This tendency is more obvious in the mid 1990s or the early 2000s, when even countries which investment had remained roughly constant over time (i.e. India, Pakistan, and Turkey) experienced a downward trend. The middle panel shows that the path of public investment has been declining in all six countries. In contrast, private investment, shown in the bottom panel, presents a clear upward trend (except for Indonesia). Private investment in infrastructure has increased so much than it has equaled and, in some cases, even surpassed public investment. Thus, public and private investments display contrasting patterns, which indicates that those countries have experienced a composition shift between public and private investment. The share of private investment to total investment has been rising, while that of public investment has been falling. However, rising private investment has not fully offset the fall in public investment in most cases. 36 Figure 6. Infrastructure Investment (average of 2000-05, % of GDP) (a) Total Investment 10 f D) T ta In s c re In e tm n (%o G P VNM 8 JOR vs et BOL 6 TRK CHL o l fra tru tu IND COL EGY 4 PHL THA MNG PAK KHM IDN SA BRA 2 PER ARG MEX 7 8 9 10 Log per capita GDP at year 2000 (b) Public Investment VNM 8 f D) P b In s c reIn e tm n (%o G P 6 vs et JOR HUN ALB 4 u lic fra tru tu UKR SVK BOL EGY COL MNG BLR TRK LTU KHM ARM SVN IND IDN 2 PAK THA CHL MAR PHL SA LVA GEO DZA BRA PER MEX ARG 0 7 8 9 10 Log per capita GDP at year 2000 (c) Private Investment 4 f D) P a In s c reIn e tm n (%o G P CHL BOL TRK 3 vs et IND JOR PHL 2 THA riv te fra tru tu COL PAK VNM EGY BRA ARG SA 1 PER MEX KHM MNG IDN 0 7 8 9 10 Log per capita GDP at year 2000 Data source: Calderón, Odawara, and Servén (2008). 37 Figure 7-1. Infrastructure Investment across Countries (percentage of GDP) (a) Total Investment 8 7 6 5 4 3 2 1 0 1981 1984 1987 1990 1993 1996 1999 2002 2005 (b) Public Investment 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 1981 1984 1987 1990 1993 1996 1999 2002 2005 (c) Private Investment 3.0 2.5 2.0 1.5 1.0 0.5 0.0 1981 1984 1987 1990 1993 1996 1999 2002 2005 IND PAK IDN Data source: Calderón, Odawara, and Servén (2008). 38 Figure 7-2. Infrastructure Investment across Countries (percentage of GDP) (a) Total Investment 14 12 10 8 6 4 2 0 1980 1985 1990 1995 2000 2005 (b) Public Investment 12 10 8 6 4 2 0 1980 1985 1990 1995 2000 2005 (c) Private Investment 6 5 4 3 2 1 0 1980 1985 1990 1995 2000 2005 EGY TRK SA Data source Calderón, Odawara, and Servén (2008). 39 C. Infrastructure Investment Expenditures and Infrastructure Improvement The final objective of the paper is to link infrastructure investment expenditures with economic growth in Egypt. To accomplish this objective, we need the estimation of the growth effect of improvements in the infrastructure indices, obtained in section IV above. In addition, we need information concerning the connection between infrastructure investment expenditures and improvement in the infrastructure indices for Egypt. This is the goal of this subsection. For this purpose, we use historical Egyptian data to answer the question: how much improvement in an infrastructure index is obtained for given expenditure in the corresponding area of infrastructure? Table 6. Electricity Expenditure and Improvement Electricity Estimation Method: Quantile regression Dependent variable: Change in Electricity Infrastructure Index [1] [2] [3] [4] Ratio of expenditure to labor force 0.006 *** (expenditure on electricity per 100,000 workers) [5.00] Ratio of expenditure to labor force 0.051 *** (expenditure on electricity per 100,000 workers, in logs) [5.04] Ratio of expenditure to GDP 0.005 *** (expenditure on electricity / 1,000 GDP) [6.84] Ratio of expenditure to GDP 0.079 *** (expenditure on electricity / 1,000 GDP, in logs) [8.02] Constant -0.056 *** -0.094 *** -0.084 *** -0.206 *** [3.57] [3.89] [5.32] [7.39] Observations 34 34 34 34 R-squared 0.37 0.32 0.41 0.37 Notes: The dependent variable is smoothed by using the Hodrik Prescott filter. All the expenditure variables are the moving average of expenditures in the last three years. Numbers in brackets are the corresponding t-statistics. * significant at 10%; ** significant at 5%; *** significant at 1% We run a series of regressions for which the dependent variable is the change in a given infrastructure index and the explanatory variable is the corresponding investment 40 expenditure in average for the previous three years. Investment expenditure is normalized in different ways. We consider, in turn, the ratio of expenditure to the labor force and the ratio of expenditures to GDP; and in addition, we consider the logarithm of each of these two ratios. As mentioned above, publicly available Egyptian historical data has not (until recently) disaggregated between the transportation and telecommunication sectors;28 therefore, we use the infrastructure index that combines those two sectors. For electricity, on the other hand, there is sector specific expenditure data; and therefore, we use its own infrastructure index. The results are presented in Table 6 for electricity and Table 7 for transport and telecommunications.29 Table 7. Transportation and Telecommunication Expenditures and Improvement Transportation and Telecommunication Estimation Method: Quantile regression Dependent variable: Change in Transportation & Telecommunication Infrastructure Index [1] [2] [3] [4] Ratio of expenditure to labor force 0.002 *** (expenditure on transportation & telecommunication per 100,000 workers) [14.22] Ratio of expenditure to labor force 0.038 *** (expenditure on transportation & telecommunication per 100,000 workers, in logs) [9.46] Ratio of expenditure to GDP 0.002 *** (expenditure on transportation & telecommunication / 1,000 GDP) [5.08] Ratio of expenditure to GDP 0.061 *** (expenditure on transportation & telecommunication / 1,000 GDP, in logs) [3.96] Constant -0.016 *** -0.076 *** -0.036 ** -0.18 *** [3.84] [6.16] [2.08] [3.16] Observations 45 45 45 45 R-squared 0.47 0.44 0.19 0.22 Notes: All the expenditure variables are the moving average of expenditures in the last three years. Numbers in brackets are the corresponding t-statistics. * significant at 10%; ** significant at 5%; *** significant at 1% 28 As stated in footnote 24, disaggregated investment data for transportation and telecommunication has recently become available for the period 1983-2007. It is not used here, however, because of its limited time coverage and because it has not undergone the necessary quality controls. 29 These regressions are estimated using the Quantile (Median) estimator, rather than the simpler and more common Ordinary Least Squares procedure. Although the results from both methods are similar to each other, in this case we prefer the Quantile estimator given its lower sensitivity to outlier observations. 41 The results are qualitatively similar for both electricity and transportation/telecommunications. The coefficients on the four measures of investment expenditure are positive and statistically significant, implying that there is a relevant connection between investments and improvement in infrastructure. However, the small size of the coefficients, as well as the low R2s, reveals that any economically important improvement in the infrastructure indices would take a relatively long time if investment expenditures are maintained at moderate levels. A rapid improvement in the indices would occur only under quite sizable investment. Since most public expenditure planning is done relative to GDP, in what follows we only use the estimated coefficient in Column 3 of each table. In addition, we conduct a more complex regression analysis of infrastructure improvement by linking it not only to investment expenditures but also to the current stock of infrastructure. This extension may be necessary to take into account the infrastructure "product cycle," that is, the different investment needs at different stages of infrastructure development. In most infrastructure cases, larger investment is needed when infrastructure is incipient, as when it depends on expensive plants to initiate production. This is known in the economics literature as "non-convex investment," characterized by large fixed costs and low marginal costs. In other cases, initial infrastructure improvements may be obtained relatively easily; but as development occurs, further progress is increasingly costly. We conduct this extended empirical analysis by estimating a regression of the change in infrastructure index on past investment expenditure, the current infrastructure index, and the interaction between the two. The results are presented in Appendix 5. For electricity, the three explanatory variables are statistically significant, carrying the following pattern of signs. The respective coefficient on investment expenditure is positive, on current infrastructure index is negative, and on the interaction is positive. The positive coefficients on both investment expenditure and its interaction with the current infrastructure level indicate that electricity-related infrastructure behaves as "non- convex" investment, that is, electricity improvements are more costly in the early stages of this type of infrastructure, when large plants and generators have to be constructed. As electricity infrastructure develops, there is a larger expansion in infrastructure for given 42 amount of expenditures. For transport and telecommunications, the pattern of signs is the same, although only the investment expenditure and, in one case, its interaction with current infrastructure level are statistically significant. Given that both of these variables carry positive coefficients, the results indicate that the transport and telecommunications sector is also characterized by necessary large outlays at incipient levels of progress, with lower costs afterwards.30 Notwithstanding their insightfulness, the results on the complex specification of the infrastructure regression are more tentative than those of the simple specification presented above because of data constraints and robustness concerns. Since the simple specification implies constant costs to infrastructure improvement (rather than declining ones, as in the complex specification), using the simple specification results entails a more conservative approach to growth projections. We use the simple specification results in order to produce more conservative projections and because we are more confident in the quality of its estimates. D. Growth Projections We are now ready to produce the main projections of the paper; that is, the projections on growth improvement given an increase in public infrastructure expenditures. For this purpose, we need three pieces of information. The first relates to the determinants of economic growth. From this analysis, the most important element to consider is the effect of the infrastructure indices on economic growth. We also need to consider that a rise in infrastructure entails an increase in investment expenditures, which may involve a corresponding increase in the government burden (through distortionary taxation, public debt overhang, or bureaucratic red tape). Lastly, we must take into account that improving growth becomes more difficult as per capita GDP gets larger (the convergence effect). These effects are measured through the estimated coefficients on, respectively, the two infrastructure indices, the proxy of government burden, and the initial level of per capita GDP, which we take from Table 5, Col. 2. 30 The negative coefficient on current infrastructure stock (attenuated by the positive coefficient on the interaction) can be interpreted as an indication of capital depreciation. 43 The second piece of information is the effect of infrastructure expenditure on improvements in the infrastructure indices in Egypt. These parameters are obtained from the analysis on electricity and transportation/telecommunication improvements; specifically, we use the estimated coefficients reported in Tables 6 and 7, Col. 3. The third piece of information is the projected new level of infrastructure expenditure. In recent years, total infrastructure expenditures in Egypt are about 5% of GDP, allocated around 1.3% to electricity and 3.7% to transportation and telecommunication. For projected expenditures, we consider three scenarios. In the first, we allow total infrastructure expenditures to increase from 5% to 6% of GDP permanently. In the second one, the increase is substantially larger, from 5% to 8% of GDP permanently. In the third scenario, we consider a large initial increase that gradually diminishes over time. Specifically, infrastructure expenditures rise from 5% to 7% of GDP for the first twenty years, then decline to 6.5% of GDP for the following 20 years, and thereafter converge to 6% of GDP forever. In all cases, the increase is allocated proportionally to expenditures for electricity and transportation/telecommunication. Figure 8: Projected Growth Improvement from Higher Infrastructure Investment Growth Improvement with respect to current infrastructure spending of 5%GDP 4.00 3.50 3.00 2.50 %2.00 1.50 1.00 0.50 0.00 0-5 5-10 10-15 15-20 20-25 25-30 30-35 35-40 40-45 45-50 Five-year periods Scenario 1: 6% GDP Scenario 2: 8% GDP Scenario 3: 7/6.5/6% GDP 44 Figure 8 presents the growth projections under the three scenarios for the next 50 years. They represent the change in growth with respect to the path corresponding to the current infrastructure investment of 5% of GDP. The growth improvement is not constant over time because, first, as the infrastructure stock builds up, it renders higher growth; and, second, as per capita GDP rises, it becomes more difficult to grow further. In addition, the government burden of increased expenditures reduces the growth improvement. These three forces play against each other dynamically to render the projected growth path. The effect of higher infrastructure investment dominates over time, and that's why the growth improvement path rises gradually. In the first years of larger infrastructure investment, the growth returns are rather small. However, towards the end of the first decade, the per capita GDP growth improvement is already considerable, reaching almost 0.5 pp, 1.5 pp, and 1 pp for three scenarios, respectively. By the third decade, the growth improvement continues to rise, amounting to 1 pp, 3 pp, and 1.75 pp, correspondingly for the three scenarios. Note that, by the design of the third scenario, its growth improvement tends to flatten by the fifth decade. As time goes on, the infrastructure stock continues to increase, promoting larger growth; however, due to decreasing returns (manifested through conditional convergence in the growth regression), these growth improvements are increasingly harder to obtain. The combination of these two forces makes the growth improvement converge to a steady value. Specifically, in the long run, the growth gains from the three scenarios are, respectively, 1.6 pp, 4.7 pp, and 1.6 pp. The first and third scenarios render the same long run increase because in both cases infrastructure expenditure eventually stays at 6% of GDP. The projections just presented assume, realistically, that the increase in infrastructure investment may involve a heavier government burden (through higher government expenditures). However, if the rise in infrastructure investment occurs in the context of public expenditure reform, entailing a redistribution of funds from less to more efficient uses and a co-participation of the private sector in the investment process, then the growth improvement could become substantial even in the first years of the program. Figure 9 compares the results for the first two scenarios (i.e., a permanent increase of infrastructure investment to 6% and 8% of GDP, respectively) between low and high 45 government burden. High government burden corresponds to the projections analyzed above (and presented in Figure 8). Low government burden corresponds to the case when the increase in infrastructure investment does not imply an increase in government expenditures (setting the corresponding coefficient to zero). The results are remarkable. Having a low government burden is crucial to increase the growth improvement from higher infrastructure investment particularly in the early stages of the program. Even when the increase in infrastructure investment is modest (from 5% to 6% of GDP), the per capita GDP growth improvement can reach 1 pp by the end of the first decade provided the government burden is low (this rise is twice as high as under high government burden). If government burden is a big concern, it may be better to increase infrastructure investment by a modest amount, particularly in the first years. An implication from this analysis is that the merits of increasing infrastructure investment have to be weighed against the costs of an expansion in government size. One way to deal with this challenge is to increase the quality of infrastructure investment by giving the private sector a more active role in the provision of infrastructure. Figure 9: Projected Growth Improvement under Different Fiscal Burden Growth Improvement comparing low and high government burden to increase infrastructure expenditure 4.50 4.00 3.50 3.00 2.50 % 2.00 1.50 1.00 0.50 0.00 0-5 5-10 10-15 15-20 20-25 25-30 30-35 35-40 40-45 45-50 Five-year periods 6% GDP, Low Burden 6% GDP, High Burden 8% GDP, Low Burden 8% GDP, High Burden 46 The simulations can be used to inform three critical values used in public policy analysis. The first is the measure of the fiscal multiplier. This is the impact on per capita GDP of a given increase in public expenditures. The simulations show that the fiscal multiplier that applies to infrastructure expenditures gradually rises over time. A permanent increase in public infrastructure expenditures of 1 percentage point of GDP leads to a positive but small change in per capita GDP in the first few years of implementation. The low short-term impact is explained by the fact that infrastructure takes time to build, while the negative impact of government burden is rather immediate. As the stock of infrastructure grows, so does the positive effect on per capita GDP. Thus, the fiscal multiplier reaches 1 percentage point of per capita GDP about three decades after implementation, gradually converging to a long-run fiscal multiplier of 1.6 percentage points. The second is the analysis of financial fiscal sustainability. Specifically the issue is whether the improvement in infrastructure can render a sufficient increase in government revenues to fully finance infrastructure investment. In layman's terms: does public infrastructure pay for itself? According to our simulations, if the proceeds derived from the new infrastructure projects are limited to general government revenues, public infrastructure investment will not pay for itself. In Egypt, government revenues amount to 25-30% of GDP. Applying this range of rates, the rise in government revenues given by the increase in GDP will only cover a fraction of the increase in infrastructure expenditure. This fraction would be rising given the increasing impact of infrastructure investment on GDP growth. Thus, assuming a government revenue rate of 30% of GDP, the fraction of self-finance infrastructure investment would be about 35% during the first five years, 50% by the end of the second decade, and 75% in the long run. (Correspondingly, the revenue rates needed for infrastructure investment to break even would be 85%, 60%, and 40% for the respective time horizons.) The third, and most important, is the analysis from a social perspective. This considers the value of the increase in per capita GDP (or income) for society as a whole. The calculation should take into account the time value of money and, therefore, discount the future stream of income with an appropriate rate. Assuming a discount rate of 5% over the per capita GDP growth rate, an increase in infrastructure expenditure of 1 47 percentage point of GDP leads to a net present value gain of 6 percentage points of per capita GDP for the first 25 years of implementation and 10.5 percentage points of per capita GDP for the first 50 years.31 This is a sizable yet reasonable improvement, and it can be further increased if the program's financing and implementation are less burdensome to the economy. Finally, we must note that this improvement is only a fraction of the total effect of infrastructure on social welfare. A complete evaluation would also take into account the direct benefit of infrastructure on the health, comfort, and happiness of its recipients. VI. Summary and Conclusion In the last half century, Egypt has experienced remarkable progress in the provision of infrastructure, which is the result of decades of purposeful and costly investment. In the past 15 years, however, a worrisome trend has emerged: Infrastructure investment has suffered a substantial decline. To be sure, some of this decrease is to be expected as infrastructure projects mature and demand less costly outlays. Nevertheless, the investment decline should be of concern for two reasons. The first is that the rate of progress in the measures of infrastructure, particularly road networks and power generation, has slowed down. A developing country like Egypt cannot afford such stagnation. The second reason is that, although acceptable at Egypt's current income level, infrastructure in the country has much room to improve to be consistent with its goals of future economic growth. The study first reviews the current state of several infrastructure indicators in Egypt compared with the rest of the world. For this purpose, a dataset for 150 countries was collected containing different indicators of quantity and quality of services for four major sectors of infrastructure: transport, telecommunications, electricity, and water and sanitation. The overall results suggest that Egypt has attained a level of infrastructure 31 The first-order condition for consumer welfare maximization implies, r = + g, where r is the real interest rate, is the subjective rate of time preference, is the inverse of the coefficient of intertemporal substitution, and g represents the per capita GDP growth rate. It is customarily assumed that the coefficient of intertemporal substitution is equal to 1 and the subjective rate of time preference is 0.05. This would entail a real interest rate equal to 0.05 over the growth rate, which is our working assumption here. Thus, if the per capita GDP growth rate were around 5%, the real interest used as discount rate in the NPV calculations would be 10%. 48 performance consistent with what is expected given its level of economic development. In particular, Egypt performs as well as or better than other countries of similar per capita GDP regarding the following infrastructure indicators. For transportation: road network length, paved roads, quality of railroads, and quality of air transport. For telecommunication: main telephone lines, telephone faults, and waiting list of mainline installation. For power generation and water: access to electricity, quality of electricity supply, access to improved water source, and access to sanitation facilities. Then, the study examines the trends in infrastructure spending in Egypt from 1960 to 2007, disaggregated by sector of origin (public and private) and by two main sectors, transportation (including Suez Canal) and communications, and electricity. Relative to GDP, total infrastructure expenditure in Egypt fell after reaching its peak in the late 1980s, mainly due to a decline in public spending. Private investment took off in the mid-1980s, and its magnitude has been growing since then. In recent years, the rise in private investment is clear in transportation and telecommunications, with the private sector becoming a principal player in telecommunications since 2005. Rising private investment in the last two decades, however, has not entirely offset the decline in public spending. In comparison with other developing countries, the share of public in total infrastructure spending in Egypt is larger than what would be expected according to the country's income level. In contrast, by the same standard its private infrastructure spending is considerably smaller. The trends of infrastructure expenditure across countries also reveal declining public spending and a clear upward trend in private investment. Thus, not only Egypt but countries such as India, Indonesia, Pakistan, South Africa, and Turkey have experienced a composition shift from public to private infrastructure spending. In some cases, such as India and Turkey, private investment has increased sufficiently to offset the decline in public spending in infrastructure. A large body of literature has found that improving infrastructure has a statistically significant and economically substantial positive effect on economic growth. Using data for 78 countries for the period 1960-2005, this study confirms the result on the beneficial growth impact of infrastructure in telecommunication, transport, and power generation. This impact is found to be larger if infrastructure development does not 49 involve an increase in government burden on the economy. Moreover, using Egypt- specific data, the study finds a positive and significant link between infrastructure expenditures and infrastructure development. Based on these results, this study concludes that improving infrastructure in Egypt will have a beneficial effect on economic growth and that, in turn, improving infrastructure will require a combination of larger infrastructure expenditures and more efficient investment. The analysis provided in this study suggests that a permanent increase in infrastructure expenditures has a gradually rising effect on per capita GDP growth. As the infrastructure stock builds up, it renders higher growth; on the other hand, as per capita GDP rises, it becomes more difficult to grow (due to diminishing capital returns). In addition, the government burden of increased expenditures may reduce any growth improvement. These three forces play against each other dynamically to render the projected growth path of a permanent rise in infrastructure expenditure. The effect of higher infrastructure investment dominates over time, and, thus, economic growth rises gradually to converge to a positive value. The quantitative estimates obtained in this study allow some tentative growth projections. If infrastructure expenditure in Egypt rose permanently from its current level of 5% of GDP to 6%, the growth returns would be rather small in the first years but would gradually rise. Towards the end of the first decade, the gain in per capita GDP growth would reach almost 0.5 percentage points, by the third decade it would amount to 1 percentage point, and eventually it would converge to 1.6 percentage points. If the increase in infrastructure expenditures were more pronounced, from 5% to 8%, the rise in per capita GDP growth would amount to 1.5, 3, 4.7 percentage points in the corresponding time horizons. These quantitative estimates can be used to elucidate questions of fiscal sustainability and social welfare. First, on financial fiscal sustainability, the issue is whether the improvement in infrastructure can render a sufficient increase in government revenues to fully finance infrastructure investment. According to the simulations presented in the study, if the proceeds derived from the new infrastructure projects were limited to general government revenues, public infrastructure investment would not pay for itself. Considering the rates of government revenues prevailing in Egypt (25-30% of 50 GDP), the rise in government revenues given by the increase in GDP would only cover a fraction of the increase in infrastructure expenditure. This fraction would be rising given the increasing impact of infrastructure investment on GDP growth. Thus, the fraction of self-finance infrastructure expenditure would be about 35% during the first five years, 50% by the end of the second decade, and 75% in the long run. Second, on social welfare, the estimates can be used to quantify the gains from an expansion in infrastructure in terms of per capita GDP. The calculation should take into account the time value of money and, therefore, discount the future stream of income with an appropriate rate. Assuming a discount rate of 5% over the per capita GDP growth rate, an increase in infrastructure expenditure of 1 percentage point of GDP would lead to a net present value gain of 6 percentage points of per capita GDP for the first 25 years of implementation and 10.5 percentage points of per capita GDP for the first 50 years. These projections assume, realistically, that the increase in infrastructure expenditure may involve a heavier government burden (through higher government expenditures). However, if the rise in infrastructure expenditure occurs in the context of public expenditure reform, entailing a redistribution of funds from less to more efficient uses and a co-participation of the private sector in the investment process, then the growth improvement could become substantial even in the first years of the program. If infrastructure expenditure rose permanently from 5% to 6% of GDP without entailing more government burden, the per capita GDP growth improvement could reach 1 percentage point by the end of the first decade. If government burden is a big concern, it may be better to increase infrastructure investment by a modest amount, particularly in the first years. One implication from this analysis is that the merits of increasing infrastructure investment have to be weighed against the costs of an expansion in government size. A second implication is that renewed infrastructure investment should be considered in the larger context of public sector reform. Rationalizing public expenditures can release resources to be used in the generation of infrastructure. Moreover, an improvement in the quality of infrastructure investments can result in faster progress at lower costs. In this regard, Egypt would do well in considering the experience of other countries where 51 stronger participation of the private sector in the provision of infrastructure has led to significant productivity gains. Finally, a limitation of this study should be acknowledged. It has focused on the economic growth impact and corresponding income improvement obtained from higher infrastructure investment. This is, however, only one of the positive effects of infrastructure on social welfare. A complete evaluation would also take into account the direct benefit of infrastructure on people's health, accessibility, comfort, and, ultimately, happiness. 52 References Agenor, Pierre-Richard, Nabli, Mustapha K., and Tarik M. Yousef (2005). "Public Infrastructure and Private Investment in the Middle East and North Africa." World Bank Policy Research Working Paper 3661. The World Bank. Alonso-Borrego, C. and M. Arellano (1999). "Symmetrically Normalized Instrumental- Variable Estimation Using Panel Data." Journal of Business and Economic Statistics 17(1): 36-49. Arellano, M., and S. Bond (1991). "Some Tests of Specification for Panel Data: Montecarlo Evidence and an Application to Employment Equations." Review of Economic Studies 58(2): 277-97. _____ (1998). "Dynamic Panel Data Estimation Using DPD98 for Gauss: A Guide for Users." Mimeo. CEMFI. ftp://ftp.cemfi.es/pdf/papers/ma/dpd98.pdf Arellano, M., and O. Bover (1995). "Another Look at the Instrumental-Variable Estimation of Error-Components Models." Journal of Econometrics 68(1): 29-52. Aschauer, David A. (1989). "Is Government Spending Productive?" Journal of Monetary Economics 23:177-200. Arrow, Kenneth, J. and Mordekai Kurz (1970). "Public Investment, The Rate of Return and Optimal Fiscal Policy." Johns Hopkins. Barro, Robert J. (1990). "Government Spending in a Simple Model of Endogenous Growth." The Journal of Political Economy 98 (5): 103-125. ______, (1991). "Endogenous Growth in a Cross Section of Countries." Quarterly Journal of Economics 106: 404-444. Barro, Robert J. and Xavier Sala-I-Martin (1992). "Public Finance in Models of Economic Growth." Review of Economic Studies 59: 645-661. Blundell, R. and S. Bond (1998). "Initial Conditions and Moment Restrictions in Dynamic Panel Data Models." Journal of Econometrics 87(1): 115-143. Calderon, Cesar (2008) "Infrastructure and Growth in Africa," Unpublished working paper, The World Bank. Calderon, Cesar and Alberto Chong (2004). "Volume and Quality of Infrastructure and the Distribution of Income: An Empirical Investigation." Review of Income and Wealth 50: 87-105. 53 Calderon, Cesar and Luis Serven (2004). "The Effects of Infrastructure Development on Growth and Income." World Bank Policy Research Working Paper No. 3400. The World Bank. _____ (2004). "The Trends in Infrastructure in Latin America." World Bank Policy Research Working Paper 3401. The World Bank. _____ (2008). "Infrastructure and Economic Development in Sub-Saharan Africa." World Bank Policy Research Working Paper No. 4712. The World Bank. Calderon, Cesar, Odawara, Rei, and Luis Serven (2008). "Infrastructure Investment in Developing Countries: A Quarter-Century Retrospective." Mimeo. The World Bank. Canning, David (1999). "The Contribution of Infrastructure to Aggregate Output." World Bank Policy Research Working Paper 2246. The World Bank. Canning, David and Marianne Fay (1993). "The Effect of Transportation Networks on Economic Growth." Columbia University Department of Economics Discussion Paper Series. Columbia University. Demitoriades, Panicos O. and Theofanis P. Mamuneas (2000). "Intertemporal Output and Employment Effects of Public Infrastructure Capital: Evidence from 12 OECD Economies." Economic Journal, Royal Economic Society 110(465): 687-712. Devarajan, Shantayanan, Swaroop, Vinaya, and Heng-Fu Zou (1996). "The Composition of Public Expenditure and Economic Growth." Journal of Monetary Economics 37: 313- 344. Dhumale, Rahul (2000). "Public Investment in the Middle East and North Africa: Towards Fiscal Efficiency," Development Policy Review Vol.18, pp.307-324. Dobronogov, Anton, and Farrukh Iqbal (2005). "Economic Growth in Egypt: Constraints and Determinants." MENA Working Paper Series No. 42. The World Bank. Dollar, D. and A. Kraay (2004). "Trade, Growth, and Poverty." Economic Journal 114(493): F22-49. Easterly, William (2001). "The Lost Decades: Explaining Developing Countries' Stagnation in spite of Reform 1980-98." Journal of Economic Growth 6 (2): 135-157. Easterly, William and Sergio Rebelo (1993). "Fiscal Policy and Economic Growth: An Empirical Investigation." Journal of Monetary Economics 32 (3): 417-458. Enders, Klaus (2007). "Egypt-Searching for Binding Constraints on Growth." IMF Working Paper WP/07/57. International Monetary Fund. 54 Esfahani, Hadi S. and Maria T. Ramirez (2003). "Institutions, Infrastructure and Economic Growth." Journal of Development Economics 70 (2):.443-477. Estache, Antonio (2003). "On Latin America's Infrastructure Privatization and its Distributional Effects." Mimeo. The World Bank. Estache, Antonio, Foster, Vivian, and Quentin Wodon (2002). "Accounting for Poverty in Infrastructure Reform: Leaning from Latin America's Experience." WBI Development Studies. The World Bank. Everhart, Stephane S., and Mariusz A. Sumlinski (2001). "The Impact on Private Investment of Corruption and the Quality of Public Investment," Discussion Paper No. 44, International Finance Corporation. Galiani, Sebastian, Gertler, Paul, and Ernesto Schargrodsky (2005). "Water for Life: The Impact of the Privatization of Water Services on Child Mortality." Journal of Political Economy 113(1): 83-120. Griliches, Z. and J. Hausman (1986). "Errors in Variables in Panel Data." Journal of Econometrics 31(1): 93-118. Hausmann, Ricard, Rodrik, Dani, and Andres Velasco (2005). "Growth Diagnostics." Working paper. Harvard University. Hnatkovska V. and N. Loayza, 2004, "Volatility and Growth," in Managing Volatility and Crises: A Practitioner's Guide Overview, The World Bank. Hoffmaister, Alexander W., Jorge E. Roldós, and Peter Wickham, 1998. "Macroeconomic Fluctuations in Sub-Saharan Africa." IMF Staff Papers, 45(1). International Monetary Fund Holtz-Eakin, D. W. Newey, and H. S. Rosen (1988). "Estimating Vector Autoregressions with Panel Data." Econometrica 56(6): 1371-95 IMF (2003). "Fund Assistance for Countries Facing Exogenous Shocks", Policy Development and Review Department, International Monetary Fund. International Finance Corporation (2003). "Private Participation in Infrastructure: Trends in Developing Countries in 1990-2001." International Finance Corporation. Kamaly, Ahmed (2007). "Economic Growth Before and After Reform: The Case of Egypt." International Journal of Applied Econometrics and Quantitative Studies 1-4. Kheir El Din, Hanaa and Tarek A. Moursi (2003). "Sources of Economic Growth and Technical Progress in Egypt: An Aggregate Perspective." Economic Research Forum, Egypt. 55 Levine, R., N. Loayza, and T. Beck (2000). "Financial Intermediation and Growth: Causality and Causes." Journal of Monetary Economics 46(1): 31-77. Loayza, Norman, Fajnzylber, Pablo, and Cesar Calderon (2005). "Economic Growth in Latin America and the Caribbean: Stylized Facts, Explanations and Forecasts." The World Bank. Mankiw, N. Gregory, Romer, David, and David N. Weil. "A Contribution to the Empirics of Economic Growth." The Quarterly Journal of Economics 107 (2): 407-437. Nabli, Mustapha Kamel, and Marie-Ange Veganzones-Varoudakis (2007). "Reform Complementarities and Economic Growth in the Middle East and North Africa." Journal of International Development 19: 17-54. Ragab, Adla (2005). "Infrastructure and Investment." Working Paper Series #002. Social Research Center, The American University in Cairo. Roller Lars-Hendrik and Leonard Waverman (2001). "Telecommunications Infrastructure and Economic Development: A Simultaneous Approach" American Economic Review 91: 909-923. Sanchez-Robles, Blanca (1998). "Infrastructure Investment and Growth: Some Empirical Evidence." Contemporary Economic Policy 16: 98-108. Straub, Stephane (2008). "Infrastructure and Growth in Developing Countries: Recent Advances and Research Challenges." World Bank Policy Research Working Paper No. 4460. The World Bank. Windmeijer, Frank (2005). "A finite sample correction for the variance of linear efficient two-step GMM estimators." Journal of Econometrics 126 (1): 25­51. The World Bank (2003). "Trade, Investment, and Development in the Middle East and North Africa." The World Bank. _____ (2008). "Public-Private Partnership and Infrastructure Subsidy and Tariff Reform." Unpublished technical note. The World Bank. World Development Indicators (2007), www.worldbank.org. 56 APPENDIX 1: Data sources - Investment in infrastructure data for Egypt is obtained from Ministry of Economic Development, Egypt. (http://www.mop.gov.eg/English/Economic%20Indicators.html). The unit is in local currency unit, millions, at current prices. - Investment in infrastructure data used for Figure 6 and 7 are obtained from César Calderón, Rei Odawara and Luis Servén (2008). - Real per capita GDP, PPP adjusted 2005 international dollars from WDI database, The World Bank. - Labor force from WDI database, The World Bank. - Arable land from WDI database, The World Bank. - Surface area from WDI database, The World Bank. 57 APPENDIX 2: Additional Cross-Country Comparisons Figure 1-A. Sub-sample: countries with real per capita growth>0.019 Correlations between infrastructure indicators and per capita GDP, PPP (constant 2005 int'l $) (a) Transport ra le n ) lo o to l ro d le g (s rt o 1 0 w rk rs x a b la d 10 0 OMN MUS DNK AUT HKG IRL SVN GBR SGP THA 4 BGD TWN ARM KOR NLD SVN 0 EST CHNGY E EGY MYS P ve ro d (%of to l ro s) 8 JPNIRL NOR 3 DOM PRT JPN NOR SWE SGP ta ad g f ta a n th q f ,0 0 o e AUT CRI BWA NLD AUS USA TUN FIN PAK POL FIN USA GBR 0 MYS DNK IDN SWE 2 POL 6 PRT ESP NPL CHL TUR BIH VNM IND ESP LSO IND LKA EGY DOM a d as TCD 0 AUS 1 MUS LKA 4 PAK IDN BWA CAN KHM CHN TUN TUR OMN NPL CRI 0 0 CHL EST 2 MOZ LSO THA VNM KHM -1 0 BGD TCD 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 8 8 SGP JPN DNK HKG HKG JPN USA 6 6 AUT NLD TWN NLD MYS GBR CAN DNK SWE SWE SGP FIN uality of railroads FIN PRT ESP TWN CHL KOR AUS KOR uality of roads CAN USA THA MYS ESPAUT TUN GBR IND AUS SVN NOR TUN NOR MUS PRT 4 4 CHN DOM TUR BWAEST IRL CHN EST POL PAK PAK THA SVN IND EGY ARM POL EGY BWA BGD Q KHM IRL Q LKA CHL VNM LKA IDN NPL LSO BGD VNM TUR IDN CRI 2 2 ARM MOZ BIH MOZ KHM BIH MUS TCD LSO TCD CRI NPL DOM 0 0 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 58 Figure 1-A (continued). Sub-sample: countries with real per capita growth>0.019 Correlations between infrastructure indicators and per capita GDP, PPP (constant 2005 int'l $) (b) Telecommunications 8 8 orkers) rs) HKG SWE SWE GBR SGP EST PRT ESPAUT NOR SVN FIN IRL DNK NLD orke DNK CAN AUS KOR JPN GBR USA AUS KOR ESPHKG NOR CHL M POL JPNIRL AUT NLD SVN FIN SGP TUR YS USA TUR POL CRI PRT TUN log of cell phone lines (lines per 1,000 w MUS lo of total m in s (lin pe 1 00 w MUS EST BWA OMN CAN CHL THA BIH EGY BIH ARM DOM 6 CHN MYS TUN es r ,0 CHN CRI 6 VNM DOM OMN THA BWA EGY IDN LKA LKA LSO IND IDN PAK ARM KHM 4 LSO VNM IND NPL LAOPAK LAO MOZ a line 4 MMR BGD BGD TCD 2 MOZ TCD NPL 2 g 0 MMR 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 100 MMR 1 Waiting list (the ratio of waiting list to main lines) NPL IND Telephone faults (per 100 m lines) 80 .8 ain LSO 60 .6 NPL LSO 40 .4 MOZ TCD ARM LKA MMR BGD MUS 20 .2 LKA TUN MYS CHL MOZ IDN BIH LAO ARM TUR EST SVN BWA POL USA THA PRT ESP AUS DNK PAK EGY TUN THA CRI GBRIRL AUT IND MUS POL OMN TWN HKG KOR JPN SGP CRI MYS EST OMN FIN SGP TUR 0 CHL 0 EGY SVN GBR KOR JPN TWNDNK NOR SWE AUS HKG AUT NLD CAN 6 7 8 9 10 11 log of real per capita GDP 6 7 8 9 10 11 log of real per capita GDP 59 Figure 1-A (continued). Sub-sample: countries with real per capita growth>0.019 Correlations between infrastructure indicators and per capita GDP, PPP (constant 2005 int'l $) (c) Electricity NOR DOM 2 30 SWE CAN FIN USA C egawatts per 1,000 workers) DNK AUS EST OMN JPN AUT ESPHKG SGP ARM SVN IRL BIH KOR GBR NLD IND MYS PRT Power loss (% of total output) POL 25 CHL DOM TUR PAK MUS 0 CRI EGY TUN THA CHN 20 PAK LKA IND LAO NPL MMR MOZ IDN VNM BWA -2 LKA LSO ARM BIH 15 TUR OMN MMRBGD NPL IDN log of EG (m HKG EGY TUN CRI BWAEST 10 VNM -4 MOZ BGD POL PRT ESP THA CHL GBRIRL NOR TCD SWE CHN CAN AUS USA KHM SVN SGP 5 MYS AUT JPN DNK NLD TWN -6 KOR FIN 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 8 100 CHN THA TWN SGP EGY TUN CRI MYS CHL Access to Electricity (% of population) JPN DNK MUS OMN NLD HKG GBR SGP FIN DOM AUT SWE NOR CAN USA KOR ESP IRL VNM Quality of electricity supply AUS 6 MYS PRTTWN SVN TUN MUSCHL EST THA CRI EGY BWA LKA POL ARM BIH TUR 4 CHN IND PAK IDN 50 LKA PAK VNM IDN MOZ LSO IND BWA KHM NPL BGD NPL 2 BGD KOR DOM KHM TCD MMR LSO MOZ 0 0 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 60 Figure 1-A (continued). Sub-sample: countries with real per capita growth>0.019 Correlations between infrastructure indicators and per capita GDP, PPP (constant 2005 int'l $) (d) Water and Sanitation 12 0 ater S rce (%of p ulation) acilities (%of population) 100 FIN USA JPN SGP SWE AUS DNK NLD CAN PRT ESPAUT THACRI BIH MUSMYS EST CHL op ARM TUR LKA TUN 0 80 MMR 10 MUS EST PRT ESPAUT NOR JPN SGP FIN USA GBR AUS SWE DNK NLD CAN THACRI MYS EGY BIH ARM TUR DOM DOM BWA CHL TUN VNM NPL PAK IND EGY CHN VNM CHN ou 60 LKA PAK 80 BGD MMR IDN LSO IDN anitation F LAO ccess to Im ved W BWA 40 BGDLSO KHM pro MOZ 60 ccess to S LAO NPL KHM IND 20 TCD TCD A MOZ 40 A 0 6 7 8 9 10 11 6 7 8 9 10 11 log of real per capita GDP log of real per capita GDP 61 Figure 1-B. Full-sample, correlations between infrastructure indicators and average per capita GDP growth (1995-2007), PPP (a) Transport .15 .15 BIH Average real per capita GDP growth Average real per capita GDP growth .1 .1 ARM AZE CHN CHN LVA GEO LVAEST EST AGO AGO LTU LBR TTO LTU KHM BLR VNM KHMVNM BLR KAZ KAZ ALB IRL .05 IRL .05 RWA IND RWA IND SVK SVK HRV POL MOZ POL HRV DOM BWA MNG BWA SDN DOM KOR RUS LKA HUN SVN LKA HUN RUS TWNSVN TUN MUS BGR ROM GRCFIN SGP CHL ROM CUB TUN FIN BGR PRI MUS MDA GRC SGP BGD BGD ETH TCDEGY CHL UGA UKR IRN PER TUR MYSCZE TCD TZA PERETH UGA TUR PAN IRN EGY UZB MYS KGZ UKR CZE THA BFA KGZ PAN MLITZA IDN GHA PHL ESP GBR AUS CRINOR NIC SWE MLI CRI BFA NICGHAPHL OMN CAN AUS ESP MAR SWE HKG THA MAR JOR PAK DZA NGA HND LSO NLD OMN NAMLSO YEMHND NGA IDN MKDTJK NOR NLD USA DZA PRT PAK GBR JOR AUT DNK USA AUT BEL PRT YEM SEN URY BEN ZMB ECU ISRZAF NZL ZAF ARG URY NZL BEL DNK ISR FRA CMR NPL GTM SLVARG COL BOLBRA MEX SYR MRT GIN ITA FRA JPN BRA CMR BOL GIN SYR BENCOLSLV MRTECU ZMB SLE KEN SEN MEX GTM LBY NPL JPN KWT DEU ITA CHE ARE KEN SLE GMB CHE COG GMB MDG SAU JAM ARE TGO MDG JAM TGO 0 PRY 0 NER CAF CAF GAB CIV NER CIV GAB HTI HTI PNG PNG ERI BDI GNB BDI WBG GNB ZWE ZWE -.05 -.05 -1 0 1 2 3 4 0 20 40 60 80 100 log of total road length (sqrt of 1,000 workers x arable land) paved roads (% of total roads) .15 .15 thlegend BIH BIH Average real per capita GDP growth .1 .1 verage real per capita G P grow ARM AZE ARM AZE CHN CHN GEO LVA GEO LVA EST EST D AGO AGO TTO LTU TTO LTU VNM KHM KHM VNM ALB KAZ ALB IRL KAZ .05 .05 IRL IND IND MOZ POL SVK BWA HRV MOZ HRVBWA POL SVK MNG RUS LKA DOM SVN KOR TWN DOM MNG LKA SVN RUS KORTWN HUN HUN MDA ROM PER BGD TUR MUS TUN CHL BGR TCD KGZETHUGA TZA UKR CZE GRC PAN FIN SGP MUS ETHPER BGDMDA UGA ROM BGR TURCHL TCD BFA PAN GRCUKR TUN CZE FIN SGP BFA EGY MYS EGY MYS CRINIC TJK MLI HND PAK NOR THAESPCAN HKG SWE NICCRI KGZ IDN TZAMAR THA MLI NOR ESP SWE HKG LSO PHL IDNNGADZA MKD MAR NAMAUS GBR DNK JOR PRT NLD URY NZL ZAF USA AUT FRA BEL PHLDZA HND URY JOR LSONGAMKD TJK NAM PRT GBRCAN DNK PAK ZAF AUS USA AUT NLD BEL CMR NPLCOL ECU ARGGTMMEX ISR SLV CMR ARG ECU BOL NZL ISR FRA BOL BRA MRT ZMB MWI BEN GMB ITA SLV KWT DEU JPN CHE GTM MWI MEX NPL ZMB COL BRA BEN MRT KWT ITA DEU JPN CHE KEN ARE GMB KEN ARE MDG JAM JAM MDG 0 0 PRY PRY BDI BDI A ZWE ZWE -.05 -.05 0 2 4 6 8 0 2 4 6 8 Quality of roads Quality of railroads 62 Figure 1-B (continued). Full-sample, correlations between infrastructure indicators and average per capita GDP growth (1995-2007), PPP (a) Transport (continued) .15 .15 BIH BIH th Average real per capita GDP growth Average real per capita G P grow .1 .1 ARM AZE ARM AZE D CHN CHN GEO LVA GEO LVA AGO EST EST AGO LTU LTU VNMTTO KAZ KHM KHM VNM KAZ TTO ALB .05 IRL ALB .05 IND IRL MOZ SVK POL HRV SVK MOZ BWA IND BWA POLHRV MNG RUS DOM LKA SVN KORTWN MNG RUS SVN LKA DOM KOR ETH UGA BGD MDA HUN ROM BGR CZE TUN MUS GRCCHL FIN SGP BGD ROM BGR HUN TUN MUSCHLTWN MDA ETH TUR CZE GRC FIN SGP TCD BFAPER TUREGY UKR PAN TCD UKR UGA BFAKGZPER PAN KGZ MLI LSO CRI PHL TJK NIC TZA ESP MYS THA AUSGBR JOR AUTNAM NOR SWE HKG MLI TJK TZA NIC PHL CRI EGY ESP MYS THA SWE NORAUS GBRHKG IDN NGA MKDCMR ECU MEXMAR URY PRT DZA PAK HND ZAF ISR USA DNKNLD CAN NZL FRA BEL LSO MKD NGA ARG IDN MAR DZA HND JOR PAK NAM AUT CAN DNK NZL BEL USAFRA DEU PRT NLD NPL ZMB MRT COL ARG BOL MWI BEN SLV KWT BRA ITA GTM JPN DEU CMR MRT MWI URY BEN BOL NPL ECU GTM MEX KWT ZMBCOL BRA ZAF ISR SLV KEN GMB CHE ARE GMB ITA KEN CHE JPNARE MDG JAM MDG JAM 0 PRY 0 PRY BDI BDI ZWE ZWE -.05 -.05 0 2 4 6 8 2 3 4 5 6 7 Quality of port facilities Quality of air transport 63 Figure 1-B (continued). Full-sample, correlations between infrastructure indicators and average per capita GDP growth (1995-2007), PPP (b) Telecommunications .15 .15 BIH BIH Average real per capita GDP growth Average real per capita GDP growth .1 .1 AZE ARM ARM AZE CHN CHN GEO LVA GEO LVA AGO EST AGO EST MMR LTU MMR LTU TTO BLR LBR KHM BLR TTO VNM VNM ALB KAZ IRL KAZ ALB IRL .05 .05 RWA IND RWA MOZ SVK POL HRV MOZIND BWA SVK POL HRV LAO SDN BWA MNG DOM KOR RUSSVN SDN LAO MNG DOMRUS KOR LKA LKA SVN HUN CUB ROM HUN TUN CHLPRI BGR MUS GRC FIN CUB TUN PRI ROMBGR MUS CHL FIN TCD UGA BGD ETH IRNSGP MDATUR PER PANEGYCZE ETH BGD TCD UZB IRN MDA SGP CZE GRC TZA BFA UKR MYS CRI HKG BFA UGA EGY UKRPAN TUR HKG TZA GHA PER CRI THA MYS SWE MLI GHA LSO IDNNIC KGZ THA PHLHND OMN NORSWE ESP AUS GBR TJKMLI KGZ NIC MAR AUS OMN NLD PHL ESP GBR NOR NGA PAKMAR DZA JOR URYNLD NAM YEM ZAF LBN MKD BEL CAN USA PAK LSO DZA NGA IDN HND NAM JOR CANUSA DNK CMR MWI NPLSEN ECU MEX PRT DNK BOL GTM BRA JPN DEU COL AUT ARG NZLFRA ISR NPL MWI LBY YEM CMR URY LBN MEX NZLAUT ARG MKDDEU SEN MRT SLV ZAFKWT ISR ECU PRT BEL FRA GIN ZMB MRT BEN SLV KWT SYR ITA CHE GIN ZMB BEN BOL GTM SYR COL BRA JPNCHE ITA COG KEN ARE KEN GMB ARE MDG TGO JAMSAU TKM MDG TGO COG SAU JAM 0 PRY 0 PRY NER CAF CIV GAB SWZ NER CAF CIV SWZ GAB HTI HTI ERI PNG ERI PNG BDI WBG BDI WBG GNB ZWE ZWE -.05 -.05 0 2 4 6 8 0 2 4 6 8 log of total main lines (lines per 1,000 workers) log of cell phone lines (lines per 1,000 workers) .15 .15 BIH th Average real per capita GDP growth Average real per capita G P grow .1 .1 AZE ARM AZE ARM D LVA GEO EST LVA EST GEO LTU MMR LTU MMR BLR KAZ KAZBLR ALB ALB .05 IRL .05 SVKRWA IND SVK IND POL MOZ HRV POL BWA MOZ SDN HRV MNG KORHUN SVNLKA RUS KOR LAO MNG SDN TWN TUN TWN TUN RUS SVN HUN LKA SGP CUB ROM BGR GRC CHL MDA PAN CZE MUS TUR MYS BFA ETH UZB SGP BGR ROM FIN MUS IRNMDA CHL CZE GRCUZB TUR BGD ETH EGY UKR TCD PAN PER UKR OMNJOR IDNTZA HKG AUS CRI THA NIC PHLESP GBR GHA HKG EGY BFA MYS SWETZA KGZ CRI OMN THA AUS NOR GBR GHA LSO MAR LSO TJK JOR MAR NLD TJK CAN PAK SENDNK ECU COL NAMZAF DZA USA HNDPRT NGA AUT MEXMKD SLV BEL AUT URY NZL NAM DNK COL ZAF ARG FRASLV DZA YEM HND BRA KWT JPNBEN GIN MRT SYR NPL ZMB DEUSEN ECU BOL BRA KWT JPN GIN ITA ZMB MWI BEN GMB NPL ARE KEN CHE ARE KEN SAUTGO JAM MDG TKM SAU MDG TKM JAMTGO 0 0 PRY SWZ NER PRY SWZ GAB CAF CIV CAF CIV GAB ERI PNG ERI BDI WBG WBG BDI GNB ZWE ZWE -.05 -.05 0 20 40 60 80 100 0 .2 .4 .6 .8 1 Telephone faults (per 100 main lines) Waiting list (the ratio of waiting list to main lines) 64 Figure 1-B (continued). Full-sample, correlations between infrastructure indicators and average per capita GDP growth (1995-2007), PPP (c) Electricity .15 .15 BIH BIH Average real per capita GDP growth Average real per capita GDP growth .1 .1 AZE ARM ARM AZE CHN CHN GEO LVA GEO AGO EST ESTAGO LVA MMR LTU MMR KHM LBR VNM BLR TTO LTU TTO VNM BLR KAZ ALB KAZ IRL IRL ALB .05 .05 RWA IND IND MOZ BWALAO POL SVK HRV SVK POL MOZ HRV BWA SDN SDN MNG DOM KOR RUS KOR RUS LKA DOM LKA HUNSVN TWNSVN HUN TUN CUB MUS ROM CZE PRI BGR FIN CHLTUNCUB ROM BGR IRN TCD ETH UGA TZA BGD UZBIRN UKRSGP FIN CHL PANMDA PER EGY TUR GRC SGP ETH TUR BGD CZEPER UKR GRC UZB MDA BFA MYS ESP KGZ HKG MYSESPHKG PAN SWEEGYOMN TZA KGZ MLI PHL THACRI GHA IDN NIC LSO YEM PAKMAR NAM DZA HND JOR MKD GBR AUSSWE TJK OMN USA NOR NLD AUT CAN AUS CRI GHA THA JOR NLDNOR PHL NAMYEM CANGBR IDNTJK MAR DZA NIC NGA URYPRTDEU DNK ZAF BEL NZL HND ZAF BOLLBN CMR PAKLBY NGA USA AUT DNK PRT NZL BEL SEN MKD URY SEN MWI NPL BENSLE CMR MRT SLV ECU MEX LBY ISR ZMB COL BRA ARG ITA JPN KWT BOLGTM SYR LBN FRA ISR GTMDEU KWT ARGCOL SYR FRA ZMB MEX BEN NPL SLV BRA ECU GMBGINKEN COG CHE ARE SAU JPN ITA CHE ARE SAU TKM KEN COG TGO MDG TKM JAM JAM TGO 0 0 NER SWZ GAB PRY PRY CAF CIV GAB CIV HTI HTI ERI PNG WBG BDI GNB ZWE ZWE -.05 -.05 -6 -4 -2 0 2 0 20 40 60 80 log of EGC (megawatts per 1,000 workers) Power loss (% of total output) .15 .15 BIH th Average real per capita G growth Average real per capita G P grow .1 .1 ARM AZE DP D CHN CHN GEO LVA EST AGO AGO MMR TTO LTU TTO KHM VNM KAZ KHM VNM ALB .05 .05 IRL IND MOZ SVK MOZ IND POL BWAHRV BWA DOM MNG RUS KOR KOR SDN MNG DOM TWN LKA HUN SVN LKA ROM MDA BGR TUN TWN CUB TUN TCD BGD TZA UGA ETH TUR PERGRC MUS CZE SGP PAN CHL FIN UGA ETH BGD PER PAN MUSCHL IRNSGP BFA UKR KGZ EGY THA CRI MYS HKG SWE BFATZA EGY MYS CRI THA NICMLI ESP CANNLD GHAIDN NIC TJK NGA LSO PAK PHL IDN HND DZA MKD ZAF NAM JOR AUS NOR MAR GBR URY PRT USA FRA AUT BELDNK LSO NAM YEM NGA PAK HND PHLMAR OMN JOR DZA NPL CMRECU MWIMRT BEN COL NZL ARG GTM SLVZMB MEX BOL BRA ISR DEU KWTCHEJPN MWI ZMB BEN SEN NPL CMR BOL ZAF SLV COL URY ISR ECU ARG SYR BRA LBYLBN KWT GMB KEN ITA ARE KEN COG ARESAU MDG JAM MDG TGO JAM 0 PRY 0 PRY CIV GAB HTI ERI BDI ZWE ZWE -.05 -.05 0 2 4 6 8 0 20 40 60 80 100 Quality of electricity supply Access to Electricity (% of population) 65 Figure 1-B (continued). Full-sample, correlations between infrastructure indicators and average per capita GDP growth (1995-2007), PPP (d) Water and Sanitation .15 .15 BIH BIH Average real per capita GDP growth Average real per capita GDP growth .1 .1 ARM AZE ARM AZE CHN CHN GEO LVA GEO LVA EST AGO EST AGO MMR MMR LBR TTO KHM LBR TTO BLR KHM VNMKAZ BLR VNM KAZ ALB .05 ALB .05 RWA IND RWA IND SVK MOZ SVK HRV MOZ BWA HRV LAO SDN BWA DOM SDN LAO MNG DOM LKA RUS MNG LKA RUSHUN ROM MDA TUN HUN ROMCUBTUN BGR ETH BGD MUSBGR CUB FIN SGP CHL CZE ETH TCD BGD MDA CHL MUS UZB PAN TUR FIN SGP CZE GRC TCD UGA PAN PER TUR UKRGRC UZB TZA MLI UGA BFA PER KGZ UKR EGY MYS GHA BFA TZA NIC MLI IDN EGY MYS SWE KGZ ESP CRIAUS THA TJK NIC GHA IDN HND PAK LSO MARDZA PHL CRI THA NAM JOR SWE ESP AUS NOR GBR CAN NLD LSO NGANAM YEMCMR PAK ZAF HND MAR PHL JOR TJK PRT MKDDZA CANNLD USA AUT DNK URY NGA USA ZMB YEMCMR MWI MRT BEN GIN SEN NPL MEXPRT ZAF SLV SYR COL MKD AUT ARGDNK ECU URY BOL BRA GTMISRLBN FRA DEU SLE SEN NPL GIN MRT BEN BOL ZMB MWI BRA GTM ARG LBY MEXSLV COL ECU SYR ARE DEU JPN CHE SLE KEN GMB JPN CHE ARE KEN GMB MDG TGO COG JAM MDG COG TGO JAM 0 NER SWZ PRY 0 NER SWZ CAF PRY CIV CAF GAB CIV GAB HTI HTI ERI PNG PNG ERI WBG GNB BDI WBG GNB BDI ZWE ZWE -.05 -.05 20 40 60 80 100 0 20 40 60 80 100 Access to Improved Water Source(% of population) Access to Sanitation Facilities(% of population) 66 APPENDIX 3: Econometric Methodology Although our focus is on the estimation of the growth effects of public infrastructure, we must make sure that the full growth regression is correctly specified and estimated. Thus, we need to ensure that all relevant variables are included, that their potential endogeneity is controlled for, and that we account for unobserved effects. The growth regression presented above poses some challenges for estimation. The first is the presence of unobserved period- and country-specific effects. While the inclusion of period-specific dummy variables can account for the time effects, the common methods of dealing with country-specific effects (that is, within-group or difference estimators) are inappropriate given the dynamic nature of the regression. The second challenge is that most explanatory variables, including the public infrastructure indices, are likely to be jointly endogenous with economic growth, so we need to control for the biases resulting from simultaneous or reverse causation. The following outlines the econometric methodology we use to control for country-specific effects and joint endogeneity in a dynamic model of panel data. We use the generalized method of moments (GMM) estimators developed for dynamic models of panel data that were introduced by Holtz-Eakin, Newey, and Rosen (1988), Arellano and Bond (1991), and Arellano and Bover (1995). These estimators are based, first, on differencing regressions or instruments to control for unobserved effects and, second, on using previous observations of explanatory and lagged-dependent variables as instruments (which are called internal instruments). After accounting for time-specific effects, we can rewrite equation 4.1: yi ,t yi ,t 1 ' X i ,t i i ,t (4.2) To eliminate the country-specific effect, we take first differences of equation 4.2: y i ,t y i ,t 1 y i ,t 1 y i , t 2 ' X i , t X i , t 1 i , t i , t 1 (4.3) 67 The use of instruments is required to deal with the likely endogeneity of the explanatory variables and the problem that, by construction, the new error term, i,t ­ i,t­1, is correlated with the lagged dependent variable, yi,t­1 ­ yi,t­2. The instruments take advantage of the panel nature of the data set in that they consist of previous observations of the explanatory and lagged-dependent variables. Conceptually, this assumes that shocks to economic growth (that is, the regression error term) be unpredictable given past values of the explanatory variables. The method does allow, however, for current and future values of the explanatory variables to be affected by growth shocks. It is this type of endogeneity that the method is devised to handle. Under the assumptions that the error term, , is not serially correlated and that the explanatory variables are weakly exogenous (that is, the explanatory variables are assumed to be uncorrelated with future realizations of the error term), our application of the GMM dynamic panel estimator uses the following moment conditions: E y i , t s i , t i , t 1 0 for s 2; t 3, ..., T (4.4) E X i ,t s i ,t i , t 1 0 for s 2; t 3, ..., T (4.5) for s 2 and t = 3,..., T. Although in theory the number of potential moment conditions is large and growing with the number of time periods, T, when the sample size in the cross-sectional dimension is limited, it is recommended to use a restricted set of moment conditions in order to avoid overfitting bias (we return to this issue below). In our case, we work with the first five acceptable lags as instruments.32 As mentioned above, the indicator of natural disasters and the measure of external shocks (i.e. growth rate of terms of trade) are treated as exogenous variables. The GMM estimator based on the conditions in 4.4 and 4.5 is known as the difference estimator. Notwithstanding its advantages with respect to simpler panel data 32 Specifically, regarding the difference regression corresponding to the periods t and t-1, we use the following instruments: for the variables measured as period averages --financial depth, government spending, trade openness, inflation, and crisis volatility-- the instrument corresponds to the average of period t-2; for the variables measured as initial values --per capita output and secondary school enrollment- - the instrument corresponds to the observation at the start of period t-1. 68 estimators, the difference estimator has important statistical shortcomings. Blundell and Bond (1998) and Alonso-Borrego and Arellano (1999) show that when the explanatory variables are persistent over time, lagged levels of these variables are weak instruments for the regression equation in differences. Instrument weakness influences the asymptotic and small-sample performance of the difference estimator toward inefficient and biased coefficient estimates, respectively.33 To reduce the potential biases and imprecision associated with the difference estimator, we use an estimator that combines the regression equation in differences and the regression equation in levels into one system (developed in Arellano and Bover, 1995, and Blundell and Bond, 1998). For the equation in differences, the instruments are those presented above. For the equation in levels (equation 4.2), the instruments are given by the lagged differences of the explanatory variables.34 These are appropriate instruments under the assumption that the correlation between the explanatory variables and the country-specific effect is the same for all time periods. That is, E[ yi ,t p i ] E[ yi ,t q i ] and (4.6) E[ X i ,t p i ] E[ X i ,t q i ] for all p and q Using this stationarity property and the assumption of exogeneity of future growth shocks, the moment conditions for the second part of the system (the regression in levels) are given by: E[ yi ,t 1 yi ,t 2 i i ,t ] 0 (4.7) E[ X i ,t 1 X i ,t 2 i i ,t ] 0 (4.8) 33 An additional problem with the simple difference estimator involves measurement error: differencing may exacerbate the bias stemming from errors in variables by decreasing the signal-to-noise ratio (see Griliches and Hausman, 1986). 34 The timing of the instruments is analogous to that used for the difference regression: for the variables measured as period averages, the instruments correspond to the difference between t-1 and t-2; and for the variables measured at the start of the period, the instruments correspond to the difference between t and t-1. 69 We thus use the moment conditions presented in equations 4.4, 4.5, 4.7, and 4.8 and employ a GMM procedure to generate consistent and efficient estimates of the parameters of interest and their asymptotic variance-covariance (Arellano and Bond 1991; Arellano and Bover 1995). These are given by the following formulas: ^ ( X ' Z 1 Z ' X ) 1 X ' Z 1 Z ' y ^ ^ (4.9) ^ ^ AVAR( ) ( X ' Z 1 Z ' X ) 1 (4.10) where is the vector of parameters of interest (, ); y is the dependent variable stacked first in differences and then in levels; X is the explanatory-variable matrix including the lagged dependent variable (yt­1, X) stacked first in differences and then in levels; Z is the ^ matrix of instruments derived from the moment conditions; and is a consistent estimate of the variance-covariance matrix of the moment conditions.35 Note that we use only a limited set of moment conditions. In theory the potential set of instruments spans all sufficiently lagged observations and, thus, grows with the number of time periods, T. However, when the sample size in the cross-sectional dimension is limited, it is recommended to use a smaller set of moment conditions in order to avoid over-fitting bias (see Arellano and Bond 1998; for a detailed discussion of over-fitting bias in the context of panel-data GMM estimation, see Roodman 2007). This is our case, and therefore we use two steps to limit the moment conditions. First, as described in detail above, we use as instruments only five appropriate lags of each endogenous explanatory variable. Second, we use a common variance-covariance of moment conditions across periods. This results from substituting the assumption that the average (across periods) of moment conditions for a particular instrument be equal to zero for the assumption, conventional but more restrictive, that each of the period moment conditions be equal to zero.36 At the cost of the reduced efficiency, our two 35 Arellano and Bond (1991) suggest the following two-step procedure to obtain consistent and efficient GMM estimates. First, assume that the residuals, i,t, are independent and homoskedastic both across countries and over time; this assumption corresponds to a specific weighting matrix that is used to produce first-step coefficient estimates. Second, construct a consistent estimate of the variance-covariance matrix of the moment conditions with the residuals obtained in the first step, and then use this matrix to re-estimate the parameters of interest (that is, second-step estimates). 36 This uses the "collapse" option of xtabond2 for STATA. 70 steps decrease over-fitting bias in the presence of small samples by accommodating cases when the unrestricted variance-covariance is too large for estimation and inversion given both a large number of explanatory variables and the presence of several time-series periods. The consistency of the GMM estimators depends on whether lagged values of the explanatory variables are valid instruments in the growth regression. We address this issue by considering a specification test. This is the Hansen test of overidentifying restrictions, which tests the validity of the instruments by analyzing the sample analog of the moment conditions used in the estimation process. Failure to reject the null hypothesis gives support to the model. 71 APPENDIX 4: Infrastructure Investment ­ Recently Published Disaggregation Figure 4-A. Infrastructure Investment in Egypt (1983-2007) (Percentage of GDP) (a) Total Investment 14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0 1983 1987 1991 1995 1999 2003 2007 (b) Transportation (incl. SC) 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 1983 1987 1991 1995 1999 2003 2007 (c) Communications 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 1983 1987 1991 1995 1999 2003 2007 (d) Electricity 6.0 5.0 4.0 3.0 2.0 1.0 0.0 1983 1987 1991 1995 1999 2003 2007 Public Private 72 APPENDIX 5: Additional Regression Analysis on Infrastructure Expenditure and Improvement Table 6-A. Electricity Expenditure and Improvement Electricity Estimation Method: Quantile regression Dependent variable: Change in Electricity Infrastructure Index [1] [2] [3] [4] Ratio of expenditure to labor force 0.010 *** (expenditure on electricity per 100,000 workers) [6.02] Ratio of expenditure to labor force 0.112 *** (expenditure on electricity per 100,000 workers, in logs) [4.13] Ratio of expenditure to GDP 0.007 *** (expenditure on electricity / 1,000 GDP) [6.38] Ratio of expenditure to GDP 0.129 *** (expenditure on electricity / 1,000 GDP, in logs) [6.24] Initial value of infrastructure index -0.232 ** -0.542 ** -0.226 ** -0.687 *** [2.43] [2.06] [2.22] [2.76] Initial value of infrastructure index * Expenditure 0.009 0.184 * 0.010 * 0.219 ** [1.43] [1.82] [1.99] [2.61] Constant -0.135 *** -0.281 *** -0.145 *** -0.372 *** [5.05] [3.87] [5.83] [6.11] Observations 34 34 34 34 R-squared 0.46 0.35 0.47 0.42 Notes: The dependent variable and initial value of index are smoothed by using the Hodrik Prescott filter. All the expenditure variables are the moving average of expenditures in the last three years. Numbers in brackets are the corresponding t-statistics. * significant at 10%; ** significant at 5%; *** significant at 1% 73 Table 7-B. Transportation and Telecommunication Expenditures and Improvement Transportation and Telecommunication Estimation Method: Quantile regression Dependent variable: Change in Transportation & Telecommunication Infrastructure Index [1] [2] [3] [4] Ratio of expenditure to labor force 0.002 *** (expenditure on transportation & telecommunication per 100,000 work [3.04] Ratio of expenditure to labor force 0.053 ** (expenditure on transportation & telecommunication per 100,000 workers, in logs) [2.65] Ratio of expenditure to GDP 0.001 *** (expenditure on transportation & telecommunication / 1,000 GDP) [2.71] Ratio of expenditure to GDP 0.052 *** (expenditure on transportation & telecommunication / 1,000 GDP, in logs) [3.05] Initial value of infrastructure index 0.008 -0.082 -0.007 -0.089 [0.40] [1.19] [0.30] [1.29] Initial value of infrastructure index * Expenditure 0 0.029 0.001 0.033 * [0.26] [1.36] [1.62] [1.71] Constant -0.005 -0.132 * -0.017 -0.149 ** [0.24] [1.92] [0.74] [2.37] Observations 45 45 45 45 R-squared 0.50 0.47 0.46 0.45 Notes: All the expenditure variables are the moving average of expenditures in the last three years. Numbers in brackets are the corresponding t-statistics. * significant at 10%; ** significant at 5%; *** significant at 1% 74