WPS4912 P olicy R eseaRch W oRking P aPeR 4912 Making Sense of Africa's Infrastructure Endowment A Benchmarking Approach Tito Yepes Justin Pierce Vivien Foster The World Bank Africa Region African Sustainable Development Front Office April 2009 Policy ReseaRch WoRking PaPeR 4912 Abstract The paper's objective is to explain factors underlying sector, with generating capacity and household access to Africa's weak infrastructure endowment and to identify electricity at half the levels observed in South Asia. While suitable infrastructure goals for the region based on it is often assumed that Africa's infrastructure deficit is benchmarking against international peers. The authors largely a reflection of its relatively low income levels, use a dataset covering the stocks of key infrastructure-- the authors find that African countries have much more including information and communication technology limited infrastructure than income peers in other parts (ICT), power, roads, and water--across 155 developing of the developing world. Countries that face the most countries over the period 1960 to 2005. The paper challenging environment, with low population density, also examines subregional differences within Africa. weak governance, and history of conflict, have the They make use of regression techniques to control poorest infrastructure endowments. for a comprehensive set of economic, demographic, At the outset of the data series, Africa was doing geographic, and historic conditioning factors, as well as significantly better than other developing regions for road adjusting for potential endogeneities. density, generation capacity, and fixed-line telephones, Results show that Africa lags behind all other regions but Africa's relative position has deteriorated over time. of the developing world in its infrastructure endowment, The most dramatic loss of ground has come in electrical except in ICT. By far the largest gaps arise in the power generating capacity, which has stagnated since 1980. This paper--a product of the African Sustainable Development Front Office, Africa Region--is part of a larger effort in the region to gauge the status of public expenditure, investment needs, financing sources, and sector performance in the main infrastructure sectors for 24 African focus countries, including energy, information and communication technologies, irrigation, transport, and water and sanitation. Policy Research Working Papers are also posted on the Web at http://econ. worldbank.org. The author may be contacted at jirving@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 Making Sense of Africa's Infrastructure Endowment: A Benchmarking Approach Tito Yepes, Justin Pierce, and Vivien Foster Acknowledgments This study is part of the Africa Infrastructure Country Diagnostic (AICD), a project designed to expand the world's knowledge of physical infrastructure in Africa. Financing for AICD is provided by a multi-donor trust fund to which the main contributors are the Department for International Development (United Kingdom), the Public Private Infrastructure Advisory Facility, Agence Française de Développement, and the European Commission. Inquiries concerning the availability of datasets should be directed to vfoster@worldbank.org. The authors thank Cesar Calderón for generously providing some of the data used in our analysis of the interplay of infrastructure and growth. Contents 1 Methodological framework 1 2 The impact of environmental factors on infrastructure endowments 5 Variations across regions 5 Variations across types of countries 10 3 The impact of historical trends on infrastructure endowments 17 Variations over time 17 Panel data models 23 Benchmarking 26 4 Conclusions 28 References 29 Annex A Alternative regression models 31 Annex B Country-specific deviations from regression models 32 Annex C Regional infrastructure endowment 36 About AICD 38 The authors Tito Yepes is an economist in the World Bank's Latin America and Caribbean Region. Justin Pierce (justinrpierce@gmail.com) is a doctoral student in economics at Georgetown University. Vivien Foster (vfoster@worldbank.org) is a lead economist in the World Bank's Africa Region and the staff director of the Africa Infrastructure Country Diagnostic. iii T he international community has recently committed itself to scaling up development assistance to Africa, in part to address the continent's major infrastructure deficit. But given its income and other constraints, what level of infrastructure should Africa aim for? A related issue is how well African institutions currently perform in sustaining and expanding infrastructure stocks based on the resources that they have at their disposal. Both questions are amenable to analysis through cross-country benchmarking of infrastructure stocks. The objective of this paper is to shed light on Africa's current infrastructure endowment and clarify its future infrastructure goals. We pursue that objective by benchmarking the region's stock of key infrastructure--including information and communication technology (ICT), power, roads, and water--against a dataset comprising 155 countries over the period 1960 to 2005. The paper also sheds light on subregional differences, by comparing infrastructure stocks between southern, central, eastern, and western Africa. In addition to pinpointing the factors that have contributed to low infrastructure development on the continent, the benchmarking models serve to identify suitable infrastructure targets that take into account the environmental difficulties that countries face. The paper is divided into three main sections. Section 1 lays down the methodological framework used in the paper and relates it to the academic literature. Section 2 undertakes a cross-sectional analysis that identifies the extent to which differences in infrastructure stocks across countries can be explained in terms of differences in the geographical, demographic, and economic environment that they face. Section 3 presents a panel data analysis that incorporates the evolution of infrastructure stocks over time and thus clarifies the role of historical factors in determining today's infrastructure endowments. 1 Methodological framework Benchmarking is a technique widely used in the management and regulatory fields to compare the performance of firms against relevant peer groups. The benchmarking technique employed in this paper was first proposed by Shleifer (1985), as a means of regulating so-called "natural monopolies." Shleifer argued that regulators should select socially optimal prices for markets served by natural monopolies based on cost data from "similar" firms. Our approach also resembles that used by Battese and Coelli (1993) to determine levels of "technical inefficiency" among firms in a particular industry. Battese and Coelli estimated industry-level production functions based on firm-level data and defined the distance between a firm's actual and predicted levels of production as its "technical inefficiency." More recently, the World Bank's annual Doing Business report has illustrated the application of benchmarking techniques at the country level. By creating a global database of investment climate indicators and conducting cross- country comparisons, Doing Business has prompted a global debate about how to reduce red tape and motivated policy makers to improve their competitiveness. A similar approach has been applied to the infrastructure sectors in recent World Bank work at the country level (World Bank, 2004, 2005a, 2005b). Research in this area has benefited greatly from the availability of an international panel of data on infrastructure stocks provided in Estache and Goicoechea (2005). MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Here we perform a country-level benchmarking exercise for infrastructure stocks in Sub- Saharan Africa, comparing them with a global peer group of developing countries. The paper extends and deepens the work of Bogetic (2006) for the countries of the Southern African Development Community (SADC), in particular by using regression techniques to control for a wide range of differences in the operating environment faced by particular countries. In this way, the benchmark against which each country is compared is individually adjusted to control for a comprehensive set of economic, demographic, geographic, and historic conditioning factors. The analysis is based on a panel dataset of all developing countries for the years 1960 to 2004. Developing countries are defined to be those in the low-income, lower-middle-income, and upper-middle-income categories as defined by the World Bank. Data availability varies over the period according to the specific infrastructure variable under consideration. Thus, road data are available from 1960, ICT data from 1970, and electricity generation capacity data from 1980. In the case of water and sanitation, only two time periods are available (1990 and 2002), while for access to electricity no consistent time series could be found. The data are drawn from a variety of global sources, including the World Bank's World Development Indicators, the International Telecommunications Union, and the World Economic Forum's Global Competitiveness Report, among others. Our study contributes to the international benchmarking literature in three ways. In terms of scope, it is the first study to benchmark the entire region of Sub-Saharan Africa, with an emphasis on comparing regions of the developing world as well as the diverse subregions of Africa. It is also the first study to base its predicted infrastructure levels on a panel dataset and to control for potential endogeneity of regressors. The benchmarking exercise compares each country's actual infrastructure endowments to an expected, or predicted, value based on its socioeconomic structure. The predicted values are derived from an econometric model that explains variation in infrastructure levels among developing countries based on a set of economic, demographic, and structural variables. This exercise produces a measure of infrastructure endowment that controls for differences in all of the socioeconomic variables included in the model. Thus, for example, it would be incorrect to explain a lower-than-predicted infrastructure endowment for a particular country in terms of the country's low income, because the model includes income as an independent variable. Hence the result already controls for differences in income. Note that here the concept of an expected, or predicted, infrastructure level does not refer to any concept of demand, since actual levels of infrastructure may also be driven by supply factors. Moreover, the expected value should not be treated as an ideal; it simply expresses the average endowment of countries with comparable characteristics. Each country's infrastructure endowment is measured by the deviation between its actual endowment and the endowment predicted by the model (equation 1). A positive deviation indicates that the country outperforms the benchmark provided by the econometric model (i.e., the average for the relevant peer group) and vice versa. The larger the absolute size of the deviation, the greater the extent of the corresponding over- or underachievement. 2 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH (1) deviationi (actuali predictedi ) / predictedi Separate econometric models are estimated for nine different infrastructure variables--ICT (fixed and mobile telephone lines, Internet connections), power (generating capacity and access to electricity), roads (total and paved), water, and sanitation. All variables are normalized to facilitate comparisons across countries. Roads are measured in terms of their density across the country's surface area. However, to allow for the fact that some countries include large areas of uninhabitable wilderness, the total land area and total arable area are used as alternative means of normalization. Generation capacity is normalized per million inhabitants; ICT variables are normalized per thousand inhabitants. Electricity, water, and sanitation are expressed as percentage household access rates. Access to water and sanitation correspond to the definitions of improved water and sanitation specified in the Millennium Development Goals (United Nations 2007). Our data reflect only the quantity of infrastructure; they say nothing of the quality and hence economic value of those stocks. For example, two countries may have the same paved road density, but one network may be well maintained and the other nearly impassable. Unfortunately, there is no global dataset available that documents the quality of infrastructure stocks, although some research does indicate a close correspondence between quality and quantity of infrastructure (Calderon and Serven 2004). [ The first step is to estimate a simple, cross-sectional, ordinary least squares (OLS) model, based on the most recent year of data available for each of the nine forms of infrastructure . This is done following the specification given in equation (2), where y is an infrastructure stock, X is a vector of independent variables (including economic, demographic, and environmental variables as discussed below), and is an error term. (2) yi ' X i i Our approach extends the work of Canning (1998) to include a much wider set of explanatory variables. In his seminal paper, Canning found that a significant portion of cross-country variation in infrastructure endowments could be explained by economic and demographic variables, such as per capita income, population density, urbanization rate, and growth of urban population. To Canning's set of regressors we add several demographic, public sector, and structural variables. Ethnic fractionalization is also included as a regressor, since competition between ethnic groups may affect infrastructure building programs. Similarly, a governance term based on Transparency International's Corruption Perceptions Index accounts for the impact of wasteful or corrupt government management of infrastructure projects. A per capita measure of foreign aid designated for infrastructure controls for significant variations in aid activity between countries. Lastly, structural variables such as the share of manufacturing, agriculture, and exports in GDP are incorporated, since the structure of the economy may affect demand for specific infrastructure services. Our choice of explanatory variables draws upon an extensive exploratory data analysis presented in the next section of the paper. 3 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Several important limitations of this approach call for the use of panel data models. The simple OLS cross-section is of some interest in that it replicates the results of earlier literature (Canning 1998) and serves to isolate the effect of specific environmental variables that can be identified as relevant from the exploratory data analysis. Nevertheless, it is subject to important methodological limitations. First, the environmental variables only imperfectly control for the myriad differences that arise in different country situations, which can be fully reflected only by means of country-specific factors. Second, the long lag times in the development of infrastructure stocks mean that historic trends play an important role in explaining a country's current infrastructure endowment. Third, the potential reverse causality from per capita income to infrastructure stocks raises the possibility of endogeneity. All of these issues can be addressed using panel data models that analyze a repeated cross-section for the same countries across a number of years. Our second step is therefore to estimate an OLS panel model that controls for country differences (fixed effects) that affect the ease or difficulty of providing infrastructure services. This is represented by equation (3), where yit is defined as the infrastructure level for country i at time t, Xit is a matrix of socioeconomic and structural explanatory variables, and i is a time- invariant country-specific fixed effect. (3) yit ' X it i it Because the OLS fixed-effects model does not correct for the potential endogeneity of the explanatory variables, our third step is to control for potential endogeneity of relevant regressors using instrumental variables. It seems likely that per capita gross domestic product (GDP) will be endogenous in the specification given in (3). In fact, there is already a large literature based on the concept that causation lies in the opposite direction--that is, that expansion of infrastructure services increases income and income growth. This direction of causality has been examined extensively in Easterly and Serven (2003) and Calderon and Serven (2003, 2004). We draw on this literature to choose appropriate instruments for per capita GDP. Specifically, we employ some of the standard growth regressors from Calderon and Serven (2004), including trade openness (trade as share of GDP), inflation, political risk index, government involvement in economy (government consumption as share of GDP), domestic credit available to the private sector (as share of GDP), and the terms of trade index. Adding these regressors gives us the following equation: (4) yit ' X it t i it Xit contains the same set of regressors included in the cross-sectional OLS study. Per capita GDP is instrumented using the standard growth variables described above. Lastly, it is important to note that any time-invariant variables from the OLS regressions are dropped once country fixed effects are included, which accounts for the smaller sets of coefficients reported for panel data models later in the paper. 1 1 The regressors used in (4) were also considered in a dynamic panel specification, using the technique developed by Blundell and Bond (1998). These results are reported in Annex A.1 for comparison purposes, 4 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH 2 The impact of environmental factors on infrastructure endowments Here we examine cross-country variations in today's infrastructure stocks in the context of the very different operating environments that countries face. We begin with a simple benchmarking of infrastructure stocks by regions, subregions, income groups, and other subcategories. We then use a cross-sectional OLS model to isolate the impact of individual factors on infrastructure stocks in a multivariate framework. All averages presented are simple unweighted averages across countries, as opposed to population-weighted averages for the different country groupings. As a result, regional summary statistics may appear to differ from commonly reported regional averages, which typically are based on a population weighting of countries. Variations across regions Africa lags behind all other regions of the developing world in its infrastructure endowment, except in ICT (table 2.1). This finding holds across a wide range of indicators including the density of roads and paved roads, per capita capacity to generate electricity, and household access to electricity, water, and sanitation. By far the largest gaps arise in the power sector, with generating capacity and household access to electricity in Africa at around half the levels observed in South Asia, and about a third of the levels observed in East Asia. The conclusion on paved road density differs depending on whether one is considering total land area (in which case Africa comes in last) or only arable area (in which case Africa comes in ahead of South Asia and East Asia). In ICT, Africa significantly outperforms South Asia in density of mobile telephones and Internet connections and comes close in terms of fixed-line density. but they should be treated with caution because the slow adjustment process for infrastructure stocks may overstate the significance of the lagged dependent variable. 5 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 2.1 Infrastructure endowments by world region America and Central Asia Middle East Europe and and Pacific South Asia Caribbean and North East Asia Saharan Africa Africa Latin Sub- Sector and measure Transport Density of paved road network 49 149 59 335 418 482 2 (km/1,000 km , 2001) Density of paved road network 1,087 675 588 1,208 4,826 6,890 2 (km/1,000 arable km , 2001) Density of total road network 152 306 237 576 740 599 2 (km/1,000 km , 2001) Density of total road network 2,558 1,400 5,385 2,160 8,850 30,319 2 (km/1,000 arable km , 2001) Information and communication technology Density of fixed-line telephones 33 39 90 261 197 100 (subscribers per 1,000 people, 2004) Density of mobile telephones 101 86 208 489 350 224 (subscribers per 1,000 people, 2004) Density of Internet connections 2.8 1.7 6.6 16.4 14.1 10.1 (subscribers per 100 people, 2004) Energy Electrical generating capacity (MW per 1 70 154 231 970 464 496 million people, 2003) Access to electricity 18 44 57 -- 79 88 (% of households with access, 2004) Water and sanitation Water 63 72 75 87 90 85 (% of households with access, 2002) Sanitation 35 48 60 78 77 77 (% of households with access, 2002) Sources: For transport, Easterly & Serven (2003); for ICT, International Telecommunications Union; for energy, Energy Information Agency, U.S. Department of Energy); for water and sanitation, World Development Indicators. Africa's deficit remains even when the countries of the region are compared with others in the same income bracket (table 2.2). It is often assumed that Africa's infrastructure deficit is largely a reflection of its relatively low income levels. But the comparison with other developing countries in the same income bracket shows that income does not tell the whole story. Africa's low-income countries (LICs) lag substantially behind those in other regions, while the same is true for lower-middle-income countries (LMCs) and upper-middle-income countries (UMC). The divergence is particularly striking for power and paved roads. Electrical generating capacity and access to electricity in Africa are less than a third of the levels found in other UMCs around the world. 6 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 2.2 Infrastructure endowments by income group, Sub-Saharan Africa vs. other world regions Country group Low-income Lower-middle-income Upper-middle-income Africa Other Ratio Africa Other Ratio Africa Other Ratio (C=B/ (F=E/ (I=H/ Sector and measure (A) (B) A) (D) (E) D) (G) (H) G) Transport Density of paved road network 2 31 134 4.3 94 141 1.5 238 781 3.3 (km/1,000 km , 2001) Density of paved road network 2 290 728 2.5 1,176 1,919 1.6 11,086 7,415 0.7 (km/1,000 arable km , 2001) Density of total road network 2 137 211 1.5 215 343 1.6 293 1,171 4.0 (km/1,000 km , 2001) Density of total road network 2 1,535 2,194 1.4 4,233 10,624 2.5 14,179 13,375 0.9 (km/1,000 arable km , 2001) Information and communication technology Density of fixed-line telephones 10 78 7.7 106 131 1.2 120 274 2.3 (subscribers per 1,000 people, 2004) Density of mobile telephones 55 86 1.6 201 298 1.5 422 554 1.3 (subscribers per 1,000 people, 2004) Density of Internet connections 2.0 3.2 1.6 5.1 8.0 1.6 10.3 26.2 2.5 (subscribers per 100 people, 2004) Energy Electrical generating capacity (MW per 1 37 326 8.8 256 434 1.7 246 861 3.5 million people, 2003) Access to electricity 16 41 2.6 35 80 2.3 28 95 3.4 (% of households with access, 2004) Water and sanitation Water 60 72 1.2 75 86 1.2 90 93 1.0 (% of households with access, 2002) Sanitation 34 51 1.5 48 74 1.5 39 90 2.3 (% of households with access, 2002) Sources: As for table 2.1. Nevertheless, the African average masks significant variations within the region (table 2.3). To compare subregional endowments, we used four country groupings. They are SADC (all the SADC countries except Tanzania); the East African Community (EAC) of Kenya, Tanzania, and Uganda; the Economic Community of West African States (all the ECOWAS countries); and Central Africa, a default category comprised of Burundi, Cameroon, the Central African Republic, Chad, Comoros, Congo, Rep., Equatorial Guinea, Eritrea, Ethiopia, Gabon, Madagascar, Mauritania, Mozambique, Rwanda, Sao Tome and Principe, Somalia and Sudan. Comparing infrastructure endowments across these subregions reveals that the SADC countries have a substantial advantage over the others. That advantage is most pronounced in the case of paved roads, ICT, and power, where SADC is ahead of the other regions by several multiples. Generating capacity (per capita) in SADC, for example, is more than five times that reported in other parts of Africa, although, strikingly, household access to electricity is relatively similar. At the other end of the spectrum, EAC has the lowest infrastructure endowment on most measures. Western and central Africa are similar in their results. The least divergence in stocks 7 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH across subregions is seen in water and sanitation, with all regions reporting water access in the 60­70 percent range and sanitation access in the 30­45 percent range. Table 2.3 Infrastructure endowments by African subregion African subregion ECOWAS EAC SADC Central Transport Density of paved road network 2 38 8 92 41 (km/1,000 km , 2001) Density of paved road network 2 301 93 3,636 416 (km/1,000 arable km , 2001) Density of total road network 2 144 105 214 132 (km/1,000 km , 2001) Density of total road network 2 1,279 1,286 6,164 1,790 (km/1,000 arable km , 2001) Information and communication technology Density of fixed-line telephones 28 6 74 13 (subscribers per 1,000 people, 2004) Density of mobile telephones 72 54 180 74 (subscribers per 1,000 people, 2004) Density of Internet connections 2.4 2.1 5.5 1.7 (subscribers per 100 people, 2004) Energy Electrical generating capacity (MW per 1 million 31 24 175 44 people, 2003) Access to electricity 18 7 21 18 (% of households with access, 2004) Water and sanitation Water 63 64 71 58 (% of households with access, 2002) Sanitation 35 45 43 28 (% of households with access, 2002) Sources: As for table 2.1. East Asia is the closest comparator for the SADC countries; South Asia for the other African subregions. Figure 2.1 brings together the regional and subregional analysis by plotting infrastructure stocks against income for all of the regions and subregions we have considered so far. In income terms, the SADC countries are most comparable to East Asia and the Pacific. Nevertheless, the SADC countries have less infrastructure than the countries of East Asia and the Pacific by all measures, particularly access to sanitation and electricity. The only exception is paved road density, where SADC has some edge over EAP. By contrast, all other African subregions are, in income terms, similar to each other and to South Asia. On the whole, their endowments do not diverge substantially from that of the whole of Sub-Saharan Africa, although the Central African countries lag behind in terms of water and sanitation. 8 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Figure 2.1 Infrastructure endowments plotted against national income, by world region a. Percentage of households with improved water b. Percentage of households with improved sanitation 80 LCR 90 ECA MNA LCR ECA % HHolds With Improved Sanitation % HHolds With Improved Water MNA 70 80 60 EAP EAP 50 SAS SADC SAS 70 EAC SADC 40 EAC SSA ECOWAS ECOWAS SSA 60 30 CENTRAL CENTRAL 0 1000 2000 3000 4000 0 1000 2000 3000 4000 Income Per-Capita Income Per-Capita c. Percentage of households with electricity d. MW installed generating capacity per 1 million people 1000 MNA ECA 80 LCR 800 % HHolds With Electricity MW per 1 Mil. People 60 EAP 600 SAS MNA LCR 40 400 SADC EAP 20 200 SSA CENTRAL ECOWAS SADC SAS EAC SSA CENTRAL ECOWAS EAC 0 0 0 1000 2000 3000 4000 0 1000 2000 3000 4000 Income Per-Capita Income Per-Capita e. Total roads (km/km2) f. Paved roads (km/km2) 800 500 MNA LCR LCR 400 600 Paved Road (km/km^2) ECA MNA Tot. Road (km/km^2) ECA 300 400 200 SAS EAP SAS SADC 200 100 SSA ECOWAS CENTRAL SADC EAC SSA EAP CENTRAL ECOWAS EAC 0 0 0 1000 2000 3000 4000 0 1000 2000 3000 4000 Income Per-Capita Income Per-Capita 9 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Figure 2.1 Infrastructure endowments plotted against national income, by world region, continued g. Mobile telephone subscribers per 1,000 inhabitants h. Landline telephone subscribers per 1,000 inhabitants 500 ECA Mainline Subscribers Per 1000 Inhabitants ECA 250 Mobile Subscribers Per 1000 Inhabitants 400 200 LCR LCR 300 150 MNA EAP 200 100 SADC MNA EAP SADC 100 SSA 50 SAS CENTRAL ECOWAS SAS EAC SSA ECOWAS CENTRAL EAC 0 0 0 1000 2000 3000 4000 0 1000 2000 3000 4000 Income Per-Capita Income Per-Capita Source: As for table 2.1. Variations across types of countries Relative to other developing regions, Sub-Saharan Africa faces a difficult environment for the development of infrastructure, and the region's infrastructure shortfall may be traceable to those environmental disadvantages (table 2.4). Building infrastructure tends to be more difficult in countries characterized by low population density and low levels of urbanization, weak governance, high incidence of conflict, and geographical isolation--all of which distinguish Africa from other developing regions. Africa has the lowest population density, the lowest governance index, and the second-lowest urbanization rate of all developing regions. The incidence of conflict in Africa is similar to that in other regions in percentage terms, but the absolute number of countries in conflict is higher in Africa than in any other region. Table 2.4 Difficulty of infrastructure environment across regions and Caribbean East Asia and Latin America Sub-Saharan Central Asia Middle East Europe and South Asia and North Pacific Africa Africa Regional characteristic Population density (pop./km2) 67 436 121 71 133 150 Urbanization rate (%) 37 23 42 57 63 69 Governance index (0=most corrupt; 10=least corrupt) 3.7 4.3 4.2 3.9 3.8 4.5 Landlocked countries (%) 30 38 9 29 6 0 Countries in conflict (%) 23 25 14 25 12 25 Income per capita (US$) 999 842 1,801 3,628 3,649 4,193 Sources: As for table 2.1. To explore the potential impact of environmental variables, we compare the infrastructure endowment of countries within the region according to the difficulty of their environment. As 10 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH expected, countries facing a more challenging environment perform systematically worse on infrastructure (table 2.5). The differences in endowment are largest for landlocked countries, countries in conflict, and countries with low population density. Countries with low urbanization perform worse on service coverage, but better on road density. Countries with poor governance also perform worse than their comparators, but the differences are less pronounced. Overall, paved road density is the infrastructure variable that shows the largest variation across types of countries, while water and sanitation show the least variation. Striking differences also emerge when countries are grouped according to their oil-exporting status and language (table 2.6). Oil-exporting countries score systematically and substantially worse than oil-importing countries, suggesting that oil revenues are not being channeled into infrastructure investments. With respect to language groupings, francophone countries have the lowest infrastructure stocks overall. Once again, the largest differences are to be found in the area of paved road density, and the smallest in water and sanitation. 11 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 2.5 Infrastructure endowments by demographic, political, and geographic variables Population Conflict Urbanization Governance Coastline density status Sector and measure Low High Low High Poor Good Yes No No Yes Transport Density of paved road network 2 11 131 51 21 45 54 12 62 16 65 (km/1,000 km , 2001) Density of paved road network 2 273 2,829 1,144 308 354 1,819 212 1,378 218 1,492 (km/1,000 arable km , 2001) Density of total road network 2 79 315 155 112 178 126 135 158 148 154 (km/1,000 km , 2001) Density of total road network 2 1,878 4,065 2,585 2,178 1,532 3,631 1,927 2,762 1,174 3,183 (km/1,000 arable km , 2001) Information and communication technology Density of fixed-line telephones 21 56 33 28 10 51 15 36 14 50 (subscribers per 1,000 people, 2004) Density of mobile telephones 92 133 82 321 68 128 44 115 49 131 (subscribers per 1,000 people, 2004) Density of Internet connections 2.2 4.1 2.7 3.8 2.6 3.0 1.0 3.3 1.3 3.5 (subscribers per 100 people, 2004) Energy Electrical generating capacity (MW per 79 49 46 410 47 92 39 79 40 83 1 million people, 2003) Access to electricity (% of households with access, 2004) 18 17 16 49 15 21 10 21 9 22 Water and sanitation Water 59 72 62 77 63 64 52 67 61 64 (% of households with access, 2002) Sanitation 34 39 34 48 35 35 27 38 32 36 (% of households with access, 2002) Sources: As for table 2.1. While the preceding analysis is suggestive, it does not take into account the high correlations that exist among these different variables, and with income. Indeed, correlation coefficients between these different indicators of "difficulty" range in absolute value between 0.3 and 0.6. In order to isolate the effect of individual factors we must perform a multivariate regression analysis that looks at all of the effects simultaneously. As described in equation (1) above, we perform simple OLS cross-sectional analysis for each category of infrastructure. Many control variables are considered--among them income per capita; demographic measures (population density, urbanization, urban growth rates, and ethnic fractionalization); measures of the quantity and quality of public spending on infrastructure (proxied by governance and infrastructure aid per capita); measures capturing the geographical and cultural heritage of the country (including language group, location, and natural resources); and indicators of the structure of the economy (such as the share of exports, agriculture, and manufacturing in GDP). In addition, an Africa dummy is used to see if any specific disadvantages are associated with the continent as a whole. 12 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 2.6 Infrastructure endowments by historical variables Not oil Oil-exporter Lusophone Anglophone Francophone exporter Transport Density of paved road network 14 57 95 84 31 2 (km/1,000 km , 2001) Density of paved road network 246 1,273 1,315 2,705 188 2 (km/1,000 arable km , 2001) Density of total road network 70 173 152 240 136 2 (km/1,000 km , 2001) Density of total road network 1,909 2,720 2,295 4,861 1,425 2 (km/1,000 arable km , 2001) Information and communication technology Density of fixed-line telephones 16 38 148 48 4 (subscribers per 1,000 people, 2004) Density of mobile telephones 118 97 86 139 46 (subscribers per 1,000 people, 2004) Density of Internet connections 1.7 3.1 3.7 4.8 1.5 (subscribers per 100 people, 2004) Energy Electrical generating capacity (MW per 1 million 66 71 49 145 22 people, 2003) Access to electricity 26 16 15 23 15 (% of households with access, 2004) Water and sanitation Water 59 64 59 72 60 (% of households with access, 2002) Sanitation 34 35 35 46 29 (% of households with access, 2002) Sources: As for table 2.1. The OLS regression analysis reveals that levels of infrastructure stocks are primarily related to per capita income and demographic variables, and less so to other variables (table 2.7). Income is statistically significant in almost all cases, except for household access to water and electricity. Demographic variables also seem to be important. Higher levels of urbanization are associated with significantly higher rates of household access to water, sanitation, electricity. By contrast, higher rates of urban growth significantly hold back rates of access to services, suggesting difficulties in keeping up with the rate of urban expansion. Moreover, countries with higher population density seem to have significantly less road area, although this difference disappears when arable land is used to measure road density. None of the other geographical, cultural, or structural variables prove to be statistically significant in explaining infrastructure stocks. This suggests that in the earlier analysis, the seeming effect of geographical and cultural variables was likely being confounded with the effect of income and demographic variables with which they are associated. Finally, it is noteworthy that the coefficients for infrastructure aid per capita and governance are rarely significant and do not always have the expected signs. An alternative log-log specification of the model, designed to test the sensitivity of the estimates, yields similar results, while revealing relatively low income elasticities and somewhat 13 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH higher urbanization elasticities (table 2.8). The main differences from the first OLS model are these: In the log-log model, income is no longer significant for road density, and some of the structural variables (exports, agriculture, manufacturing) become more significant. Overall, the fit of the model is also somewhat improved. The log-log specification also allows the coefficients on continuous variables to be interpreted as elasticities. The results suggest that the income elasticity of infrastructure stocks is generally well below unity, with the highest values (around 0.5 to 0.7) found for ICT and generating capacity and the lowest (around 0.1 to 0.2) for access to water, sanitation, and electricity, and for roads. The elasticity of urbanization is somewhat higher, particularly for access to electricity and ICT services. In summary, Africa presents a major infrastructure deficit relative to other developing regions that appears to reflect low urbanization as much as low income. The region lags behind all others in the developing world in almost all areas of infrastructure. The strongest endowments are found in the ICT sector, where Africa is somewhat ahead of South Asia. By far the lowest endowment is in the power sector. Rates of access to electricity in the middle-income countries of Africa, for example, are a fraction of those found in middle-income countries in other regions. The regional average masks significant geographic variations in infrastructure endowments from country to country. The SADC countries are substantially ahead of the others, while the EAC countries are significantly behind. Africa's deficit remains even when countries are compared with others in the same income bracket. The reason appears to be that--compared with other regions of the world--Africa is a difficult environment in which to develop infrastructure. The proof is that infrastructure is weakest in those countries that face the most challenging environment. However, some environmental factors are more important than others. When all environmental variables are considered simultaneously, demographic variables (notably urbanization) appear to have the most substantial effect on infrastructure endowments. 14 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 2.7 Effect of country variables on infrastructure endowments, part 1 Cross-sectional OLS analysis, data in levels paved roads paved roads connections Generating telephones telephones Density of Density of Density of Density of Access to Access to Access to Fixed-line sanitation electricity capacity (arable) (arable) Internet Mobile roads roads water Variable Income per 0.0001 0.002 0.0001 0.001 0.001 0.003 0.001 0.0001 0.003 0.056 0.021 capita (5.73)** (3.07)** (4.20)** (3.15)** ­0.66 (2.76)** ­0.84 (4.87)** (5.25)** (5.75)** (5.37)** ­0.515 ­8.845 ­0.288 ­2.697 26.93 40.957 63.355 0.345 3.878 105.594 97.429 Urbanization ­1.66 ­1.16 ­1.06 ­0.85 (2.66)** (3.31)** (4.81)** ­1.55 ­0.59 ­0.82 ­1.89 Population 0.001 ­0.007 0.002 0.003 0.031 0.007 0.019 ­0.0001 0.01 ­0.004 0.074 density (4.29)** ­0.87 (5.53)** ­0.84 (3.04)** ­0.55 ­1.4 ­0.56 ­1.49 ­0.03 ­1.36 ­0.052 0.451 ­0.061 ­0.052 ­2.081 ­4.273 ­3.097 ­0.095 ­0.485 ­16.292 ­15.805 Urban growth ­1.92 ­0.67 (2.55)* ­0.19 (2.23)* (3.66)** (2.11)* (4.62)** ­0.82 ­1.43 (3.34)** Fractionali- 0.036 ­1.447 ­0.048 ­2.14 0.567 4.632 14.682 0.297 ­0.583 71.167 ­9.624 zation ­0.23 ­0.38 ­0.36 ­1.37 ­0.11 ­0.72 (2.23)* (2.56)* ­0.18 ­1.08 ­0.36 0.003 ­0.197 0.005 0.198 0.507 ­0.4 ­0.717 ­0.011 0.283 ­6.042 0.024 Governance ­0.15 ­0.39 ­0.29 ­0.96 ­0.74 ­0.46 ­0.73 ­0.74 ­0.66 ­0.74 ­0.01 Infrastructure 0.002 0.003 ­0.002 ­0.032 0.035 0.263 0.231 0.0004 ­0.059 0.635 0.55 aid (per capita) ­0.89 ­0.07 ­1.34 ­1.81 ­0.62 (3.76)** (2.15)* ­0.33 ­1.38 ­0.72 ­1.69 ­0.119 ­1.954 0.034 1.226 ­7.198 ­19.246 ­35.293 ­0.198 ­2.237 ­73.992 ­29.09 Africa ­1.03 ­0.69 ­0.34 ­1.05 ­1.95 (4.27)** (7.13)** (2.37)* ­0.89 ­1.39 ­1.42 0.141 3.394 ­0.001 1.001 ­3.101 7.351 ­5.68 ­0.103 0.088 8.696 ­1.339 Anglophone ­1.1 ­1.08 ­0.01 ­0.78 ­0.73 ­1.41 ­1.04 ­1.08 ­0.03 ­0.15 ­0.06 0.043 ­0.113 0.061 0.01 ­6.511 ­0.918 1.79 ­0.034 1.266 28.44 ­2.927 Francophone ­0.36 ­0.04 ­0.59 ­0.01 ­1.68 ­0.2 ­0.39 ­0.39 ­0.5 ­0.55 ­0.14 0.012 ­0.864 0.041 ­0.397 ­1.434 ­2.953 ­4.667 0.075 ­0.131 20.189 ­0.019 Landlocked ­0.12 ­0.36 ­0.49 ­0.4 ­0.44 ­0.75 ­1 ­1.06 ­0.06 ­0.47 0 ­0.109 0.698 0.037 0.213 ­3.356 ­6.323 5.269 0.001 ­3.03 ­53.872 ­15.892 Oil-exporting ­1.25 ­0.33 ­0.48 ­0.24 ­1.15 ­1.79 ­1.35 ­0.01 ­1.65 ­1.48 ­1.08 ­ 0.09 0.104 0.003 ­0.173 ­6.223 5.548 ­2.748 ­2.37 4.995 ­2.097 Conflict 0.0001 ­0.85 ­0.04 ­0.03 ­0.16 ­1.84 ­1.36 ­0.58 0.0766 ­1.08 ­0.11 ­0.12 Exports ­0.0003 0.07 0.0004 0.038 ­0.099 ­0.068 ­0.131 0.001 0.114 1.497 ­0.029 (% GDP) ­0.15 ­1.29 ­0.2 ­1.75 ­1.38 ­0.71 ­1.44 ­0.63 (2.45)* ­1.62 ­0.08 Agriculture 0.002 ­0.056 0.003 ­0.004 ­0.164 0.192 ­0.196 0.006 0.052 ­1.379 0.578 (% GDP) ­0.52 ­0.49 ­0.71 ­0.09 ­1.05 ­1.02 ­1 ­1.63 ­0.52 ­0.64 ­0.73 Manufacturing ­0.008 ­0.275 ­0.006 ­0.062 ­0.294 0.113 0.572 0.002 0.138 2.601 0.304 (% GDP) ­1.21 ­1.79 ­1.04 ­0.96 ­1.47 ­0.46 ­1.98 ­0.56 ­1.05 ­0.97 ­0.29 0.343 7.331 0.126 0.385 81.37 45.274 38.376 0.092 ­4.259 44.459 41.711 Constant ­1.13 ­0.99 ­0.48 ­0.13 (8.18)** (3.51)** (2.93)** ­0.42 ­0.66 ­0.35 ­0.84 Observations 105 104 104 103 101 99 82 110 108 93 107 R­squared 0.62 0.28 0.55 0.35 0.61 0.74 0.87 0.69 0.63 0.74 0.74 Note: * = significant at 5%; ** = significant at 10%. T-statistics are reported below coefficients. Parentheses denote that T- statistics have been reported in absolute value. 15 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 2.8 Effect of country variables on infrastructure endowments, part 2 Cross-sectional OLS analysis, data in logs (except dummy variables, which are in levels) paved roads paved roads connections Generating telephones telephones Density of Density of Density of Density of Access to Access to Access to Fixed-line sanitation electricity capacity (arable) (arable) Internet Mobile roads roads water Variable Income per 0.112 0.262 0.082 0.222 0.135 0.041 ­0.061 0.451 0.693 0.492 0.748 capita (2.87)** (2.93)** ­0.57 ­1.65 ­0.75 ­0.2 ­0.27 (2.30)* (5.19)** (3.05)** (5.13)** 0.188 0.461 0.991 0.341 0.426 ­0.126 ­0.023 0.942 0.885 0.859 0.54 Urbanization (3.19)** (3.40)** (4.93)** ­1.62 ­1.5 ­0.38 ­0.06 (3.05)** (4.07)** (3.53)** (2.29)* Population 0.071 0.081 0.09 0.729 1.111 ­0.036 0.341 0.193 0.062 ­0.048 0.121 density (4.80)** (2.38)* ­1.72 (14.12)** (16.12)** ­0.44 (3.95)** (2.57)* ­1.18 ­0.75 (2.11)* ­0.07 ­0.149 ­0.181 ­0.208 ­0.598 0.201 ­0.205 ­0.243 ­0.138 ­0.488 ­0.346 Urban growth (2.26)* (2.04)* ­1.46 (2.23)* (4.76)** ­1.39 ­1.31 ­1.72 ­1.53 (4.42)** (3.28)** Fractionali- ­0.011 ­0.018 0.047 ­0.088 ­0.013 ­0.146 ­0.073 ­0.019 ­0.022 0.08 ­0.08 zation ­0.69 ­0.51 ­0.88 ­1.61 ­0.18 ­1.72 ­0.81 ­0.24 ­0.4 ­1.28 ­1.37 0.035 0.034 0.086 ­0.308 ­0.122 ­0.259 ­0.05 ­0.191 ­0.339 0.016 0.035 Governance ­0.81 ­0.33 ­0.55 (2.09)* ­0.62 ­1.13 ­0.2 ­0.88 (2.43)* ­0.09 ­0.22 Infrastructure ­0.006 0.023 0.033 0.008 ­0.006 0.025 0.009 0.002 0.067 ­0.016 0.019 aid (per capita) ­0.67 ­1.05 ­0.99 ­0.27 ­0.14 ­0.54 ­0.17 ­0.04 (2.24)* ­0.45 ­0.57 ­0.05 ­0.251 ­0.871 ­0.097 ­0.029 ­0.249 ­0.169 ­0.042 ­0.003 ­0.998 ­0.451 Africa ­1.01 (2.19)* (5.14)** ­0.55 ­0.12 ­0.9 ­0.57 ­0.16 ­0.02 (4.93)** (2.26)* ­0.024 0.265 ­0.135 0.526 ­0.214 0.565 ­0.144 0.039 0.269 0.246 ­0.364 Anglophone ­0.44 (2.13)* ­0.73 (2.81)** ­0.85 ­1.94 ­0.46 ­0.14 ­1.4 ­1.08 ­1.67 ­0.153 ­0.23 ­0.075 ­0.11 ­0.322 ­0.347 ­0.53 ­0.331 ­0.057 ­0.621 ­0.825 Francophone (3.01)** ­1.96 ­0.45 ­0.61 ­1.33 ­1.24 ­1.76 ­1.26 ­0.31 (2.95)** (4.19)** 0.018 0.059 ­0.221 0.314 ­0.107 0.016 ­0.474 ­0.04 0.142 0.107 ­0.076 Landlocked ­0.36 ­0.52 ­1.18 ­1.92 ­0.48 ­0.06 ­1.67 ­0.16 ­0.89 ­0.54 ­0.42 ­0.017 ­0.033 0.195 ­0.453 ­0.001 ­0.305 0.156 ­0.294 ­0.014 0.005 0.002 Oil exporting ­0.41 ­0.34 ­1.34 (3.06)** 0 ­1.32 ­0.63 ­1.37 ­0.1 ­0.03 ­0.01 ­0.091 0.011 ­0.109 0.088 ­0.347 0.002 ­0.425 ­0.272 0.171 ­0.169 ­0.286 Conflict ­1.98 ­0.11 ­0.66 ­0.54 ­1.58 ­0.01 ­1.55 ­1.15 ­1.08 ­0.89 ­1.62 Exports ­0.078 0.079 ­0.015 0.365 0.479 0.585 0.666 0.444 0.474 0.285 0.034 (% GDP) (2.18)* ­0.92 ­0.12 (2.86)** (2.82)** (2.92)** (3.12)** (2.42)* (3.88)** ­1.93 ­0.24 Agriculture 0.081 0.296 0.07 ­0.031 ­0.124 ­0.759 ­0.864 0.114 0.231 0.265 0.344 (% GDP) ­1.88 (2.97)** ­0.47 ­0.21 ­0.63 (3.33)** (3.52)** ­0.53 ­1.57 ­1.45 (2.13)* Manufacturing ­0.064 0.04 0.142 ­0.148 ­0.193 ­0.367 ­0.428 0.014 0.089 0.035 ­0.033 (% GDP) ­1.79 ­0.47 ­1.12 ­1.17 ­1.13 ­1.86 (2.01)* ­0.07 ­0.67 ­0.24 ­0.23 3.706 1.111 3.526 ­6.44 ­8.298 1.309 ­0.262 ­3.128 ­1.791 ­5.408 ­2.015 Constant (9.19)** ­1.16 (2.41)* (4.59)** (4.39)** ­0.6 ­0.11 ­1.56 ­1.35 (3.19)** ­1.34 Observations 94 92 82 95 95 94 94 98 84 98 97 R­squared 0.71 0.69 0.81 0.85 0.87 0.55 0.69 0.7 0.87 0.86 0.87 Note: * = significant at 5%; ** = significant at 10%. T-statistics are reported below coefficients. Parentheses denote that T- statistics have been reported in absolute value. 16 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH 3 The impact of historical trends on infrastructure endowments Variations over time Africa's present-day infrastructure is strongly influenced by the endowment that the countries inherited at independence. Because infrastructure is costly to build, stocks usually change slowly over time. Hence, differences across developing regions will reflect differences in their history. The current dataset contains time series data going back as far as 1960 for roads, 1970 for fixed telephone lines, 1980 for generating capacity, and 1990 for water and sanitation. No consistent time series data are available for access to electricity. The data points from the 1960s and 1970s are of particular interest for Africa, since they describe the situation at around the time of independence. Overall, the data show some evidence of worldwide convergence in infrastructure levels (table 3.1). They also reveal the divergent starting points and differing rates of infrastructure growth in world regions. Convergence requires, of course, that regions with a low starting point grow faster than regions with a high starting point. We computed correlation coefficients between starting levels and growth rates. These coefficients are uniformly negative and always smaller than ­0.28, suggesting that convergence is underway. The strongest evidence of convergence is found for water and sanitation (with correlation coefficients between ­0.6 and ­0.7) and roads (with correlation coefficients between ­0.45 and ­0.55). The lowest is for generating capacity (with a correlation coefficient of ­0.28). Bucking the trend toward worldwide convergence, the infrastructure gap between Africa and other developing regions is larger today than it was some decades ago. Slow convergence has occurred for some forms of infrastructure. At the beginning of the data series for paved roads and access to water and sanitation, Africa had the lowest endowment of any developing region. The region managed to achieve relatively high growth rates in each category, but they were not great enough to compensate for the region's starting position. On the other hand, at the outset of the data series, Africa was doing significantly better than other developing regions with regard to overall road density, generation capacity, and fixed-line telephones. But growth rates in these categories have been slower than in other regions, so that by 2000 Africa had lost substantial ground relative to the other regions. Africa's failure to converge with the rest of the developing world is most clearly illustrated by the comparison between Africa and South Asia. At the outset of the period under study, South Asia was ahead of Africa in all forms of infrastructure except for fixed-line telephones and electrical generating capacity (figure 3.1). Their positions have since been reversed. The record on generating capacity is particularly stark. In 1980, Africa had almost three times as much generating capacity (per million people) as South Asia. Since then, capacity in South Asia has expanded at an average annual rate of 9 percent, while in Africa it has stagnated. Consequently, 17 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH by 1990 South Asia had overtaken Africa; by 2000 it had almost twice the generating capacity (per million people) of Africa. Indeed, Africa had the slowest rate of growth in generating capacity of any region in the developing world. The story for telephone lines is similar, if less dramatic. In 1970, Africa had twice the teledensity of South Asia. However, with faster average annual growth rates in South Asia (9 percent against 6 percent in Africa), the two regions had converged by 2000. Africa has now fallen behind South Asia with respect to fixed-line telephones, but in mobile telephony and Internet connections Africa has been growing more rapidly than South Asia and currently maintains a lead. Indexing infrastructure stocks in each region and plotting the evolution of the indexes over time sheds light on the sequencing and relative growth path of different aspects of infrastructure. In all regions, fixed-line telephones have been by far the fastest-growing component of infrastructure since the 1980s (albeit from low starting points), but the growth varies substantially across regions from around tenfold in Africa, Latin America, and the Middle East, to more than forty-fold in East Asia (figure 3.2). After telephone lines, paved roads generally have expanded most quickly, particularly in East Asia and Eastern Europe. Generating capacity (per million people) has grown only very slowly across all regions, except in South Asia in the late 1980s, when a significant expansion took place. Within Africa, the SADC countries started with a larger infrastructure endowment than the other subregions and extended it more rapidly (table 3.2). At independence there already were substantial variations in infrastructure endowment across the continent--particularly with respect to paved roads, generating capacity, and telephone lines. By 1980, the SADC countries had more than three times the generating capacity of other subregions. By 1970, they had five times the teledensity. In the case of roads, ECOWAS was in a much stronger position than the other subregions in the 1960s, but was overtaken by SADC after 1980. In water and sanitation, the differences between subregions have been relatively small. 18 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 3.1 Trends in infrastructure endowments, by world region, 1960­2000 Average a annual Region 1960 1970 1980 1990 2000 b growth (percent) SSA 6 8 21 26 39 5.0 SAS 60 76 115 113 146 2.3 Density of paved road network EAP 12 21 37 39 54 4.0 2 (km/1,000 km ) ECA 54 239 312 362 351 4.9 LAC 132 185 59 274 361 2.6 MNA 14 19 26 39 45 3.1 SSA 71 120 544 652 1,037 7.1 SAS 477 434 585 711 670 0.9 Density of paved road network EAP 239 392 563 453 556 2.2 2 (km/1,000 arable km ) ECA 110 498 687 800 1,197 6.3 LAC 472 808 527 1,973 3,958 5.6 MNA 196 320 3,283 4,594 5,729 9.0 SSA 73 80 106 119 143 1.7 SAS 104 145 204 204 303 2.8 Density of total road network EAP 55 75 144 157 175 3.0 2 (km/1,000 km ) ECA 298 576 567 630 620 1.9 LAC 267 198 241 516 583 2.0 MNA 45 50 77 92 104 2.2 SSA 1,525 1,577 2,106 2,343 2,556 1.3 SAS 757 699 888 1,084 1,408 1.6 Density of total road network EAP 3,064 2,671 3,244 3,063 3,252 0.2 2 (km/1,000 arable km ) ECA 651 1,188 1,261 1,543 2,045 3.0 LAC 1,962 2,138 2,743 4,929 7,021 3.3 MNA 558 760 33,263 28,850 30,704 10.8 SSA -- 4 7 11 22 5.6 SAS -- 2 2 5 21 8.9 Density of fixed-line telephones EAP -- 5 10 21 58 8.9 (subscribers per 1,000 people) ECA -- 32 76 134 229 6.8 LAC -- 24 44 75 155 6.4 MNA -- 9 18 41 84 7.7 SSA -- -- 71 82 73 0.1 SAS -- -- 26 127 137 8.6 Electrical generating capacity EAP -- -- 123 178 229 3.2 (MW per 1 million people) ECA -- -- 587 786 1,022 2.8 LAC -- -- 309 409 495 2.4 MNA -- -- 207 421 413 3.5 SSA -- -- -- 51 63 1.0 SAS -- -- -- 72 70 ­0.2 Water EAP -- -- -- 73 73 ­0.03 (% of households with access) ECA -- -- -- 93 89 ­0.4 LAC -- -- -- 81 89 0.7 MNA -- -- -- 83 83 ­0.02 19 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Sanitation SSA -- -- -- 30 35 0.4 (% of households with access) SAS -- -- -- 31 47 1.3 EAP -- -- -- 57 53 ­0.4 ECA -- -- -- 86 80 ­0.5 LAC -- -- -- 64 76 1.0 MNA -- -- -- 69 75 0.5 Sources: As for table 2.1. Note: SSA = Sub-Saharan Africa, SAS = South Asia, EAP = East Asia and Pacific, ECA = Europe and Central Asia, LAC = Latin America and Caribbean, MNA = Middle East and North Africa. a. Water and sanitation data are from 2002. b. Calculated over entire period available for each sector. -- = data not available. 20 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 3.2 Trends in infrastructure endowments by African subregion, 1960­2000 Average a annual 1960 1970 1980 1990 2000 b growth (percent) Central Africa 1 2 5 8 9 6 Density of paved road network EAC 3 6 12 13 8 3 2 (km/1,000 km ) ECOWAS 11 14 21 24 38 3 SADC 4 9 46 53 92 8 Central Africa 19 75 156 161 159 6 Density of paved road network EAC 46 106 135 148 95 2 2 (km/1,000 arable km ) ECOWAS 122 161 213 237 305 2 SADC 64 123 1,715 2,015 3,649 11 Central Africa 57 60 82 94 101 1 Density of total road network EAC 56 77 82 94 105 2 2 (km/1,000 km ) ECOWAS 84 93 109 121 144 1 SADC 81 90 140 154 211 2 Central Africa 2,101 1,720 1,776 1,677 1,587 ­1 Density of total road network EAC 1,053 1,201 1,073 1,124 1,297 1 2 (km/1,000 arable km ) ECOWAS 1,137 1,259 1,391 1,319 1,300 0.3 SADC 1,588 2,058 4,002 5,045 6,174 4 Central Africa -- 1 2 4 7 5 Density of fixed-line telephones EAC -- 2 3 4 6 4 (subscribers per 1,000 people) ECOWAS -- 2 3 5 17 8 SADC -- 10 18 32 56 6 Central Africa -- -- 49 57 48 ­0.1 Electrical generating capacity EAC -- -- 21 20 23 0.4 (MW per 1 million people) ECOWAS -- -- 38 34 31 ­1 SADC -- -- 177 204 185 0.2 Central Africa -- -- -- 46 56 1 Water EAC -- -- -- 42 64 2 (% of households with access) ECOWAS -- -- -- 51 63 1 SADC -- -- -- 60 73 1 Central Africa -- -- -- 22 27 0.5 Sanitation EAC -- -- -- 44 45 0.1 (% of households with access) ECOWAS -- -- -- 28 35 1 SADC -- -- -- 37 42 0.4 Sources: As for table 2.1. a. Water and sanitation data are from 2002. b. Calculated over entire period available for each sector. -- = data not available. 21 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Figure 3.1 Index of infrastructure growth by sector and by region Tot. Rd. Density - Indexes Relative to SSA Paved Rd. Density - Indexes Relative to SSA 1,600 25,000 1,400 20,000 1,200 SSA SSA 1,000 EAP 15,000 EAP 800 ECA ECA 600 LCR 10,000 LCR 400 MNA MNA 5,000 200 SAS SAS 0 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 1960 1965 1970 1975 1980 1985 1990 1995 2000 Mainline Density - Indexes Relative to SSA Mobile Density - Indexes Relative to SSA 7,000 350,000 6,000 300,000 5,000 250,000 SSA SSA 200,000 EAP 4,000 EAP ECA ECA 150,000 3,000 LCR LCR 100,000 2,000 MNA MNA 50,000 SAS 1,000 SAS 0 0 90 92 94 96 98 00 02 04 1975 1980 1985 1990 1995 2000 19 19 19 19 19 20 20 20 Internet Density - Indexes Relative to SSA Generation Capacity - Indexes Relative to SSA 16,000 1,600 14,000 1,400 12,000 SSA 1,200 SSA 10,000 EAP 1,000 EAP 8,000 ECA 800 ECA 6,000 LCR 600 LCR 4,000 MNA 400 MNA 2,000 SAS SAS 0 200 0 93 94 95 96 97 98 99 00 01 02 03 04 19 19 19 19 19 19 19 20 20 20 20 20 1980 1985 1990 1995 2000 Water Access - Indexes Relative to SSA Sanitation Access - Indexes Relative to SSA 200 350 180 1990 1990 160 2002 300 2002 140 250 120 200 100 80 150 60 100 40 50 20 0 0 SSA EAP ECA LCR MNA SAS SSA EAP ECA LCR MNA SSA Sources: As for table 2.1. 22 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Figure 3.2 Indexes of infrastructure growth by region, 1960­2003 Index of Infrastructure Growth - SSA Index of Infrastructure Growth - SAS 2,000 2,000 1,800 Paved Road Paved Road 1,800 1,600 Tot. Road 1,600 Tot. Road 1,400 Gen. Cap. 1,400 Gen. Cap. 1,200 Mainline 1,200 Mainline 1,000 1,000 800 600 800 400 600 200 400 0 200 1960 1965 1970 1975 1980 1985 1990 1995 2000 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 Index of Infrastructure Growth - EAP Index of Infrastructure Growth - ECA 2,000 2,000 1,800 Paved Road 1,800 Paved Road 1,600 Tot. Road 1,600 Tot. Road 1,400 Gen. Cap. 1,400 Gen. Cap. 1,200 Mainline 1,000 1,200 Mainline 800 1,000 600 800 400 600 200 400 0 200 1960 1965 1970 1975 1980 1985 1990 1995 2000 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 Index of Infrastructure Growth - LCR Index of Infrastructure Growth - MNA 2,000 2,000 1,800 Paved Road 1,800 Paved Road 1,600 Tot. Road 1,600 Tot. Road 1,400 Gen. Cap. 1,400 Gen. Cap. 1,200 Mainline 1,200 1,000 Mainline 1,000 800 600 800 400 600 200 400 0 200 1960 1965 1970 1975 1980 1985 1990 1995 2000 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 Sources: As for table 2.1. Panel data models The preceding exploratory data analysis illustrated the importance of historical perspective in understanding countries' present-day infrastructure position. We use panel data models to integrate that perspective into our formal analysis. As described in the first section of this paper, two different panel model specifications are estimated: an OLS fixed effects model and an instrumental variables specification. Unfortunately, it is not possible to apply either of the panel data models to access to electricity, for which only a single cross-section is available. (However, we do apply a cross-sectional instrumental variables specification to the electricity access data so as to compare them to the OLS estimates. That comparison is reported in table 3.4.) The fixed effects model confirms the importance of income, demography, and economic structure in driving stocks of household services and broader economic infrastructure (table 3.3). 23 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH The income and demographic variables that proved statistically significant in the cross-sectional OLS results were also statistically significant in the fixed effects panel results. To a greater extent than before, however, variables capturing the economic structure of the economy (export orientation, and shares of agriculture and manufacturing compared with services) become statistically significant in explaining infrastructure stocks, particularly in the case of roads and ICT. It is important to remember, however, that these results do not yet account for dynamics in infrastructure provision, nor have they been corrected for endogeneity. Table 3.3 Results of OLS fixed effects panel data model Data in logs Access to water roads (arable) paved roads paved roads connections Generating telephones telephones Density of Density of Density of Density of Access to Access to Fixed-line sanitation electricity capacity (arable) Internet Mobile roads Variable 0.051 0.079 0.196 0.227 ­0.022 0.126 0.371 0.247 1.768 2.059 0.633 Income (3.25)** (4.49)** (10.60)** (11.15)** ­0.47 ­1.33 (2.52)* (11.87)** (7.73)** (11.77)** (27.68)** 0.244 0.131 0.477 0.365 0.664 0.141 1.096 0.504 2.316 3.158 0.986 Urbanization (3.52)** ­1.69 (5.90)** (4.10)** (3.30)** ­0.36 (5.11)** (5.50)** (2.88)** (4.72)** (10.98)** Population 0.366 0.094 0.587 0.312 0.103 1.01 0.119 0.062 21.884 16.909 1.524 density (6.43)** ­1.48 (8.76)** (4.24)** ­0.64 (3.12)** (2.18)* ­0.84 (26.44)** (25.86)** (20.14)** Infrastructure aid ­0.005 ­0.011 ­0.011 ­0.017 ­0.004 0.019 0.006 0.003 ­0.019 0.08 ­0.003 (per capita) ­1.87 (3.90)** (3.76)** (5.08)** ­0.54 ­1.25 ­0.17 ­0.96 ­0.8 (3.98)** ­0.85 Exports 0.001 0 0.001 0.001 0.001 0 0.04 0.004 0.073 0.034 0.006 (% GDP) ­1.13 ­0.35 (2.20)* ­1.82 ­0.35 ­0.01 ­0.29 (5.98)** (11.28)** (6.85)** (8.60)** Agriculture ­0.004 ­0.003 0.001 0.002 ­0.001 0.009 0.188 ­0.005 ­0.112 ­0.154 ­0.01 (% GDP) (3.40)** (2.58)** ­0.95 ­1.38 ­0.49 ­1.43 ­1.17 (3.29)** (7.11)** (11.30)** (5.69)** Manufacturing 0.001 0.002 ­0.001 0.001 ­0.007 0.01 0.398 0.004 ­0.151 ­0.205 ­0.002 (% GDP) ­0.85 ­1.43 ­0.37 ­0.27 ­1.56 ­1.13 (2.85)** (2.23)* (7.55)** (13.39)** ­0.99 ­3.401 ­0.185 ­6.602 ­3.424 4.764 ­1.021 ­0.22 ­10.59 ­92.962 ­70.033 ­5.856 Constant (11.67)** ­0.57 (19.38)** (9.13)** (5.72)** ­0.61 ­0.15 (27.63)** (23.72)** (23.53)** (14.93)** Observations 1919 1910 1875 1866 167 165 85 2223 1100 1230 2323 Plant fixed effects 0.21 0.07 0.39 0.24 0.67 0.52 0.68 0.22 0.68 0.76 0.73 Sources: As for table 2.1. The results from the instrumental variables estimation indicate endogeneity bias in the OLS specification. If--as suggested by Calderón and Servén (2004) and others that expansion of infrastructure is associated with higher levels of income and growth, then we would expect the OLS coefficient on per capita GDP to be upwardly biased. Indeed, the results of our instrumental variables estimation show that the magnitude of the coefficient on per capita GDP decreased substantially once we controlled for endogeneity using the standard regressors from the growth literature, as discussed above. The traditional Hausman test of the exogeneity of our instruments indicates that they are appropriate in most infrastructure sectors. The exception is in the estimates for access to water and sanitation, where the limited number of observations (from just two years of data) may have affected the results. But it is clear that addressing the endogeneity of per capita GDP is critical to obtaining consistent coefficient estimates. 24 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 3.4 Fixed effects with instrumental variables (data in logs) Access to water roads (arable) paved roads paved roads connections Generating telephones telephones Density of Density of Density of Density of Access to Access to Fixed-line sanitation electricity capacity (arable) Internet Mobile roads Variable ­0.065 0.0002 0.00002 0.0002 0.001 0.003 0.001 0.0000001 0.004 0.103 0.03 Income ­1.15 ­1.43 (3.35)** (2.91)** ­0.23 ­1.11 ­0.3 (8.70)** (8.12)** (10.97)** (12.32)** 0.042 6.246 ­0.041 3.011 40.797 ­12.423 57.96 0.0002 ­7.384 ­30.302 31.93 Urbanization ­0.46 (5.32)** ­0.57 (4.96)** ­1.02 ­0.32 (2.12)* (2.09)* ­0.67 ­0.18 ­1.16 Population ­0.145 0.00001 0.001 0.0002 ­0.002 0.093 0.01 ­0.0000002 ­0.076 ­1.312 ­0.128 density ­0.8 ­0.01 (4.75)** ­0.23 ­0.04 (2.04)* ­0.51 ­1.35 (3.28)** (4.23)** (2.97)** Infrastructure 0.0003 ­0.0005 0.00005 ­0.0004 ­0.256 ­0.018 ­0.047 0.0000003 ­0.007 0.222 0.009 aid (per capita) ­0.09 ­0.22 ­0.37 ­0.34 (2.05)* ­0.14 ­0.16 ­1.74 ­0.54 ­1.16 ­0.18 Exports ­0.001 0.001 0.0002 0.001 0.016 0.037 ­0.03 2.92E­06 0.228 3.58 0.699 (% GDP) ­0.64 ­0.3 ­0.67 ­0.69 ­0.12 ­0.25 ­0.22 (8.16)** (8.43)** (8.05)** (7.15)** Agriculture ­0.006 ­0.008 0.001 0.001 ­0.213 0.097 ­0.791 0.000002 0.224 5.128 ­0.011 (% GDP) (2.38)* ­0.82 ­1.47 ­0.26 ­1.03 ­0.49 (2.20)* (2.80)** (2.83)** (3.67)** ­0.05 Manufacturing 0.009 0.025 0.0004 0.011 ­0.27 0.206 2.369 0.000002 0.204 1.001 ­0.384 (% GDP) (4.38)** (2.47)* ­0.65 (2.08)* ­0.89 ­0.69 (5.60)** (2.85)** (2.58)** ­0.75 ­1.63 ­1.186 ­0.382 ­0.006 ­1.046 63.256 46.968 6.852 ­0.0001 ­9.948 ­243.622 ­25.85 Constant ­1.19 ­0.53 ­0.13 (2.85)** (3.07)** (2.09)* ­0.39 (2.43)* ­1.37 (2.06)* ­1.59 Observations 1,247 1,299 1,308 1,296 119 116 78 1,474 765 892 1,525 Plant fixed Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes effects Year fixed Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes effects Hausman p­ 0.0523 0.0405 0.0001 0.0000 0.7338 0.7772 N/A 0.0000 0.0000 0.0000 0.0000 value Sources: As for table 2.1. When we model our data using the instrumental variables estimation, per capita GDP, urbanization, population density, and certain structural variables continue to explain the most country-to-country variance in infrastructure stocks (table 3.4), although the magnitude of their coefficients changes from the basic fixed effects specification. For the most part, per capita GDP remains highly significant in explaining national variations in infrastructure, although the magnitude of the coefficients decreased from the simpler model. Urbanization becomes slightly less important, with significance in only four sectors, compared to eight under the first fixed effects specification. Population density, too, loses significance in several sectors. The significance of structural variables varies by sector, but export share of GDP is highly significant with respect to generating capacity and ICT. 25 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Benchmarking The main objective of the regression models is to predict expected levels of infrastructure stocks for the purposes of benchmarking. As described in the first section of the paper, benchmarking is done by comparing the infrastructure stock in each country with the value predicted by the regression model and calculating the deviation. The value predicted by the model takes into account the effect of each country's history and environment in determining the level of infrastructure stocks the country might be expected to have. A negative deviation indicates that the country has performed below the benchmark, while a positive one indicates the opposite. The deviations are averaged across countries to produce unweighted average deviations at the regional and subregional level. For the purposes of sensitivity analysis, deviations are calculated for each of the two model specifications described above. Following Bogetic (2006), a deviation of less than 10% in absolute value is considered not to constitute a major divergence from the benchmark. Hence, attention focuses on larger deviations of more than plus or minus 10% that are taken to be indicative of a substantial degree of over or underachievement. Table 3.5 Mean deviations from predicted values, OLS Access to water roads (arable) paved roads paved roads connections Generating telephones telephones Density of Density of Density of Density of Access to Access to Fixed-line sanitation electricity capacity (arable) Internet Mobile roads Region Subregions of Africa Central ­6 ­6 ­12 ­13 2 4 ­7 ­18 ­62 ­39 ­10 EAC ­12 ­10 ­55 ­52 5 ­13 ­26 ­11 ­117 ­57 ­72 ECOWAS ­7 ­14 ­13 ­20 ­0.1 ­1 ­10 ­9 ­143 ­30 ­4 SADC 9 12 13 16 0.2 ­3 ­11 ­21 9 7 ­17 World regions SSA ­4 ­5 ­6 ­8 1 ­1 ­11 ­17 ­78 ­25 ­15 SAS 24 25 13 14 ­0.4 16 30 29 ­36 35 35 EAP ­7 ­6 ­4 ­4 ­2 ­1 9 ­2 ­33 ­20 11 ECA 6 9 ­6 ­3 0.02 ­0.5 N/A ­7 75 95 ­18 LAC ­7 ­7 ­0.3 0.4 ­0.1 ­1 3 16 ­11 30 9 MNA 0.1 3 1 4 ­3 ­3 5 10 4 ­2 ­2 Note: SSA = Sub-Saharan Africa, SAS = South Asia, EAP = East Asia and Pacific, ECA = Europe and Central Asia, LAC = Latin America and Caribbean, MNA = Middle East and North Africa. 26 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Table 3.6 Mean deviations from predicted values, instrumental variables Access to water roads (arable) paved roads paved roads connections Generating telephones telephones Density of Density of Density of Density of Access to Access to Fixed-line sanitation electricity capacity (arable) Internet Mobile roads Region Subregions of Africa Central 57 ­57 ­63 ­93 ­11 ­48 ­58 ­81 ­110 ­79 ­44 EAC 11 106 ­87 20 ­7 ­20 ­39 ­77 ­48 ­54 ­60 ECOWAS 21 ­4 ­50 ­59 ­11 ­37 ­24 ­81 ­12 ­49 ­64 SADC 6 84 ­18 85 ­1 ­20 ­69 ­38 ­76 ­89 ­38 World regions SSA 44 0.3 ­61 ­7 ­10 ­34 ­41 ­66 ­50 ­72 ­64 SAS ­31 ­53 ­11 215 20 ­42 ­8 ­51 ­108 ­118 40 EAP 32 31 ­47 ­42 ­8 ­1 24 ­29 ­29 ­24 ­34 ECA ­25 ­43 173 57 12 38 81 167 ­10 27 105 LAC 14 ­34 ­62 ­62 1 34 ­1 16 ­14 ­26 26 MNA 29 9 ­14 ­45 5 36 42 ­11 ­11 ­10 29 Note: SSA = Sub-Saharan Africa, SAS = South Asia, EAP = East Asia and Pacific, ECA = Europe and Central Asia, LAC = Latin America and Caribbean, MNA = Middle East and North Africa. Relative to the levels predicted by the instrumental variables model, Africa performs well in total road density (paved and unpaved). In the remaining sectors, it performs significantly below the levels predicted based on the region's economic, demographic, and structural characteristics. In fact, it is the worst-performing world region (that is, the region with the largest negative deviation as shown in table 3.6) in access to electricity, electrical generating capacity, and density of fixed-line telephones. Africa also exhibits large negative deviations in paved road density, sanitation access, mobile telephone density, and Internet density. Once historic and environmental factors are taken into account and steps are taken to control for the endogeneity of per capita GDP, the relative standing of Africa's subregions is not as clear cut as it first appeared. The initial comparison of infrastructure stock indicators by subregions (see table 1.3) revealed a very clear ranking: SADC was the top performer on all dimensions and EAC the worst performer on nearly all dimensions, with Central Africa and ECOWAS somewhere in between. The picture is more nuanced once the regions are compared in terms of the magnitude of their deviation from benchmarks. Except in road density (paved and unpaved densities and arable paved density), every African subregion underperformed relative to its expected value in every infrastructure sector. This underachievement is most severe in Central Africa, which exhibits the largest average negative deviation in Africa in water access, sanitation access, electricity generation capacity, and Internet density. While SADC--like the other African subregions--performs worse than predicted by the model in nearly every sector, its deviation from expectations is less great than that of other subregions, with the exception of electricity access and mobile telephone density, where it ranks last. ECOWAS and ECA generally fall somewhere between SADC and Central Africa, although ECOWAS exhibits the largest negative deviation in density of fixed-line telephones and ECA ranks last in paved road density. 27 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH History has a large influence. Deficiencies present at the beginning of our time series have become more pronounced over time. The historical data show that at the outset of the series, Africa was behind all other regions with respect to paved road density, water, and sanitation, but was doing significantly better than other regions in terms of road density, generating capacity, and telephone lines. Over time, Africa has lost ground, so that today it is the lowest-performing region in all areas except ICT. The most dramatic loss of ground has come in electrical generating capacity, which has largely stagnated since 1980. Comparing subregions within Africa reveals that most SADC countries started with a larger infrastructure endowment and have extended it more rapidly than other subregions on the continent. Given the importance of historical trends, panel models provide the appropriate analytical framework for a benchmarking exercise. The OLS fixed effects specification underscores the importance of income and demographic variables in driving infrastructure endowments. The importance of these variables is confirmed by the instrumental variables model, which controls for potential endogeneity of per capita GDP. When Africa's actual infrastructure levels are compared to those predicted by the instrumental variables model, it is clear that Africa underperforms relative to its benchmark in nearly every sector. In fact, Africa tends to have the largest magnitude of underachievement relative to other regions in many sectors, with the most severe underachievement found in Central Africa. 4 Conclusions We have tried to shed light on the status of infrastructure in Sub-Saharan Africa by benchmarking the region against other developing countries and against the level of development predicted by our models based on the region's income, demography, and economic structure. Our analysis has shown that with but one exception Africa has the worst infrastructure endowment of any developing region today, particularly with respect to electrical generating capacity. (The exception is the modern ICT sector.) African countries even perform poorly compared with their peers in the same income group in other parts of the world. Part of the explanation seems to lie in difficult environmental factors that complicate the development of infrastructure services, in particular low rates of urbanization. The region's very low level of infrastructure at the time of independence a half centry ago is another explanation for today's results. Today, the SADC countries are far ahead of the rest of Sub-Saharan Africa--their infrastructure situation is more similar to that of East Asia than to other African subregions. EAC has the weakest infrastructure situation today; its position, like that of Central Africa and ECOWAS, most closely resembles that of South Asia. When controlling for income, demographic, and structural variables, Central Africa misses its benchmark level of predicted infrastructure achievement by the largest margin of any African subregion, while SADC comes closest, despite generally falling below the predicted values. 28 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH References Arellano, Manuel and Bond, Steven (1991), "Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Questions," Review of Economic Studies 76 (1991), pp. 277­297. Battese, G.E. and Coelli, T.J., "A Stochastic Frontier Production Function Incorporating a Model for Technical Inefficiency Effects," Working Papers in Econometrics and Applied Statistics, University of New England, 1993. Blundell, Richard, and Steven Bond. 1998. "Initial Conditions and Moment Restrictions in Panel Data Models." Journal of Econometrics 87, pp. 115­143. Bogetic, Zeljko, and Johannes W. Fedderke. 2006. "International Benchmarking of Infrastructure Performance in the Southern African Customs Union Countries." Policy Research Working Paper 3987, World Bank, Washington, DC. Calderón, Cesar, and Luis Servén. 2003. "Infrastructure Compression and Public Sector Solvency in Latin America." In William Easterly and Luis Serven, eds., The Limits of Stabilization: Infrastructure, Public Deficits and Growth in Latin America (119­138). Palo Alto, CA: Stanford University Press. ------. 2004. "The Effects of Infrastructure Development of Growth and Income Distribution." Central Bank of Chile Working Papers 270. Santiago de Chile. Canning, David. 1998. "A Database of World Stocks of Infrastructure, 1950­1995," The World Bank Economic Review, No. 3, Vol. 12, pp. 529­47. Easterly, William and Luis Serven, eds. (2003), The Limits of Stabilization: Infrastructure, Public Deficits and Growth in Latin America, Stanford University Press and the World Bank, 119­38. Estache, Antonio and Goicoechea, Ana. 2005. "A `Research' Database on Infrastructure Economic Performance", World Bank Policy Research Working Paper No. 3643, available at SSRN: http://ssrn.com/abstract=757364. Shleifer, Andrei, "A Theory of Yardstick Competition," The RAND Journal of Economics, 16 (1985) pp. 319­327. United Nations. 2007. "UN Millennium Development Goals." http://www.un.org/millenniumgoals/. World Bank. Annual since 2000. Doing Business. Private Sector Development Vice Presidency, World Bank, Washington DC. World Bank (2004), Colombia: Recent Economic Developments in Infrastructure, Finance, Private Sector and Infrastructure Department, Latin America and Caribbean Region, World Bank, Washington DC. 29 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH World Bank (2005a), El Salvador: Recent Economic Developments in Infrastructure, Finance, Private Sector and Infrastructure Department, Latin America and Caribbean Region, World Bank, Washington DC. World Bank (2005b), Costa Rica: Country Economic Memorandum, Poverty Reduction and Economic Management Department, Latin America and Caribbean Region, World Bank, Washington DC. 30 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Annex A Alternative regression models Annex table A.1 Blundell and Bond Dynamic Panel Estimator roads (arable) paved roads paved roads connections Generating telephones telephones Density of Density of Density of Density of Fixed-line capacity Internet (arable Mobile roads Variable 1.006 1.004 1.034 1.004 1.167 1.186 1.255 1.098 Lagged dependent variable (119.55)** (54.94)** (52.45)** (102.82)** (17.17)** (23.45)** (43.17)** (48.39)** 0 0 0 0 0 0 0.007 ­0.002 Income ­0.59 ­1.23 ­0.13 ­1.19 ­0.9 ­1.57 ­1.65 ­0.81 ­0.004 ­0.33 ­0.005 ­0.203 ­0.0001 ­0.785 ­21.369 ­7.481 Urbanization ­0.54 ­1.33 ­0.66 ­1.3 (2.01)* ­1.16 ­1.25 (2.42)* 0 0 0 0 0 0.001 ­0.011 0 Population density ­0.05 ­0.27 ­0.81 ­1.42 ­0.69 ­1.41 ­1.28 ­0.12 0.005 ­0.015 0.001 0.008 0 ­0.063 3.153 ­1.167 Fractionalization ­1.56 ­0.32 ­0.2 ­0.28 ­0.5 ­0.28 ­0.74 ­1.01 0 0.005 0 ­0.009 0 ­0.004 ­1.476 ­0.058 Governance ­0.72 ­0.58 ­0.38 ­1.62 ­0.4 ­0.13 (2.53)* ­0.29 0 ­0.003 0 0 0 ­0.001 ­0.043 0.003 Infrastructure aid (per capita) ­0.47 ­0.75 ­0.24 ­0.39 ­0.55 ­0.33 ­0.49 ­0.33 Exports 0 0.001 0 0 0 0.007 0.028 ­0.032 (% GDP) ­0.28 ­0.5 ­0.37 ­0.03 ­1.81 ­1.82 ­0.55 ­1.3 Agriculture 0 0.003 0 0.002 0 0.005 0.083 ­0.077 (% GDP) ­0.85 ­1.23 ­0.46 ­1.21 ­1.03 ­0.52 ­0.4 ­1.08 Manufacturing 0.0004 0.003 0 0.003 0 0.005 ­0.004 ­0.069 (% GDP) (2.37)* ­0.27 ­1.36 ­1.67 ­1.59 ­0.41 ­0.03 ­0.8 ­0.008 ­0.075 0 ­0.044 0 ­0.196 8.932 7.611 Constant ­1.03 ­0.45 ­0.04 ­0.66 ­1.67 ­0.38 ­1.15 ­1.8 Observations 1889 1880 1848 1839 2186 995 1128 2308 Hansen J-test (p­value) 0.299 0.322 0.15 0.203 0.055 0.064 0.062 0.003 AB test for AR(1) (p­value) 0.008 0.288 0.129 0.004 0.000 0.013 0.377 0.023 AB test for AR(2) (p­value) 0.163 0.302 0.42 0.356 0.268 0.357 0.412 0.576 31 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Annex B Country-specific deviations from regression models Annex table B.1 Country deviations, OLS Percent Access to water roads (arable) paved roads paved roads connections Generating telephones telephones Density of Density of Density of Density of Access to Access to Fixed-line sanitation electricity capacity (arable) Internet Mobile roads Country Burundi 42 48 1 7 0 0 -- 24 93 ­58 31 Cameroon ­65 ­55 ­1 9 1 22 5 ­3 ­9 117 ­16 Central African ­4 2 14 21 0 0 -- ­15 78 150 26 Republic Chad ­2 ­1 ­50 ­50 20 ­7 ­55 ­40 46 12 3 Comoros ­8 ­2 ­8 ­1 0 0 ­3 ­32 ­12 0 32 Congo, Rep. 0 ­4 ­5 ­10 0 0 ­3 ­55 ­17 ­45 ­127 Equatorial Guinea 0 2 -- -- -- -- -- ­2 -- -- ­6 Eritrea ­5 ­11 ­6 ­12 0 0 39 -- 177 0 13 Central Africa Ethiopia 0 0 0 0 -- -- -- 0 -- -- 0 Gabon ­7 ­4 5 8 0 0 ­16 ­15 ­107 ­9 ­56 Madagascar ­3 ­1 ­19 ­17 ­4 29 ­20 ­33 ­177 8 ­59 Mauritania ­12 ­19 ­18 ­24 3 6 16 ­23 ­210 ­117 66 Mayotte -- -- -- -- -- -- -- -- -- -- -- Mozambique ­12 ­13 ­16 ­17 0 0 ­33 ­45 ­206 ­115 ­78 Rwanda ­18 ­30 14 1 2 0 ­34 ­27 ­405 ­426 ­92 Sao Tome and ­4 ­15 ­6 ­17 0 0 10 ­31 ­57 4 5 Principe Somalia 0 1 ­83 ­82 -- -- -- 9 -- -- 16 Sudan ­2 ­7 ­13 ­18 0 0 15 ­6 ­62 ­70 70 Average ­6 ­6 ­12 ­13 2 4 ­7 ­18 ­62 ­39 ­10 Kenya ­12 ­12 ­70 ­69 1 ­8 ­37 ­13 ­67 ­53 ­105 Tanzania ­1 ­2 ­24 ­26 15 ­20 ­17 ­4 ­57 ­25 ­53 EAC Uganda ­23 ­15 ­70 ­62 0 ­12 ­25 ­15 ­226 ­93 ­59 Average ­12 ­10 ­55 ­52 5 ­13 ­26 ­11 ­117 ­57 ­72 Benin ­42 ­65 ­9 ­33 ­3 36 ­9 61 ­370 ­26 ­45 Burkina Faso 8 ­5 ­2 ­14 3 ­18 9 ­8 ­165 ­13 6 Cape Verde ­28 ­22 30 36 0 0 ­23 ­53 ­351 ­250 7 Cote d'Ivoire ­11 ­22 ­8 ­18 1 4 ­2 ­62 ­117 ­127 ­14 ECOWAS Gambia, The 23 3 11 ­8 0 0 ­34 9 ­36 82 44 Ghana 8 ­19 28 ­1 12 7 18 ­34 8 ­2 18 Guinea ­6 ­9 ­7 ­10 2 ­28 ­38 9 11 81 1 Guinea­Bissau ­11 ­2 ­5 4 0 0 ­26 5 ­83 0 ­43 Liberia 12 20 14 24 0 0 -- ­11 0 0 ­39 Mali ­13 ­55 ­41 ­84 3 3 ­11 40 ­361 ­226 ­4 32 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Niger ­69 ­70 ­157 ­157 ­5 6 3 ­5 ­76 103 ­47 Nigeria 29 41 24 36 ­1 ­18 17 ­35 ­132 ­19 ­25 Senegal ­15 ­10 ­16 ­12 ­5 8 ­5 ­37 ­159 ­114 33 Sierra Leone 19 25 ­57 ­51 -- -- -- ­3 107 0 ­1 Togo ­14 ­21 ­2 ­8 ­8 ­14 ­29 ­16 ­415 62 41 Average ­7 ­14 ­13 ­20 0 ­1 ­10 ­9 ­143 ­30 ­4 Angola ­44 ­37 16 23 9 ­10 ­7 ­42 ­245 ­217 ­71 Botswana ­19 ­8 22 34 ­7 ­4 ­6 ­82 ­56 ­54 ­13 Congo, Dem. Rep. 5 15 27 36 ­4 5 ­ ­1 329 231 ­104 Lesotho 4 6 24 26 0 ­7 ­43 ­16 222 184 46 Malawi 53 44 12 4 10 8 ­20 6 ­249 ­117 ­9 Mauritius ­2 1 ­5 ­2 ­2 ­6 1 35 ­3 ­14 45 SADC Namibia 13 11 35 34 8 ­7 8 -- ­6 ­34 ­41 Seychelles 4 6 10 12 0 -- ­24 ­34 ­2 127 ­4 South Africa 39 45 11 16 ­5 ­14 ­3 ­12 ­110 ­18 ­34 Swaziland ­14 ­12 11 11 0 0 ­19 ­42 ­36 ­51 ­17 Zambia 70 77 1 8 ­5 4 ­16 ­29 ­3 ­54 ­29 Zimbabwe ­2 ­5 ­3 ­5 0 0 4 ­11 270 101 31 Average 9 12 13 16 0 ­3 ­11 ­21 9 7 ­17 33 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Annex table B.2 Country deviations, instrumental variables Percent Access to water roads (arable) paved roads paved roads connections Generating telephones telephones Density of Density of Density of Density of Access to Access to Fixed-line sanitation electricity capacity (arable) Internet Mobile roads Country Burundi ­72 ­472 ­82 ­117 44 ­32 -- ­80 ­100 ­100 ­130 Cameroon 32 ­76 ­84 ­89 ­16 ­9 17 ­76 ­88 ­64 ­90 Central Afr. Rep. 84 ­22 ­98 ­89 ­17 ­46 ­44 ­94 ­99 ­99 ­93 Chad 123 6 ­99 ­104 ­46 ­85 ­80 ­98 ­96 ­97 ­96 Comoros ­60 ­8 70 -- 33 ­69 -- ­82 ­108 ­102 69 Congo, Rep. 94 92 ­63 ­38 ­50 ­80 ­89 ­90 ­94 ­67 ­96 Equatorial Guinea 13 36 -- -- -- -- -- ­80 -- -- ­72 Eritrea 85 16 ­80 ­139 ­17 ­83 ­84 ­100 ­220 ­61 ­35 Central Africa Ethiopia 64 ­348 ­95 ­107 -- -- -- ­91 -- -- ­138 Gabon 106 ­49 ­94 ­87 ­10 ­28 ­92 ­48 ­88 ­42 ­86 Madagascar 31 39 ­76 -- ­34 ­37 ­63 ­88 ­93 ­88 ­90 Mauritania 237 ­50 ­92 ­80 ­30 ­9 ­19 ­79 ­89 14 ­76 Mayotte -- -- -- -- -- -- -- -- -- -- -- Mozambique 89 ­51 ­78 58 ­40 ­47 ­84 ­10 ­85 ­71 ­90 Rwanda ­66 ­408 ­82 ­114 31 ­50 ­33 ­139 ­102 ­106 ­110 Sao T&P ­48 139 129 -- 3 ­61 -- ­53 ­229 ­136 12 Somalia 62 289 ­94 ­218 -- -- -- ­93 -- -- 403 Sudan 189 ­95 ­95 ­83 ­4 ­33 ­65 ­85 ­49 ­86 ­24 Average 57 ­57 ­63 ­93 ­11 ­48 ­58 ­81 ­110 ­79 ­44 Kenya 12 ­8 ­76 36 ­15 ­12 ­59 ­75 68 ­33 ­74 Tanzania 18 73 ­93 128 6 ­14 18 ­83 ­83 ­76 ­88 EAC Uganda 4 251 ­92 ­106 ­11 ­32 ­74 ­72 ­129 -- ­19 Average 11 106 ­87 20 ­7 ­19 ­39 ­77 ­48 ­54 ­60 Benin 40 ­86 ­79 ­85 ­9 ­41 ­5 ­88 130 ­60 ­68 Burkina Faso 59 ­15 ­87 ­111 ­19 ­78 ­61 ­86 ­82 ­63 ­50 Cape Verde ­37 ­25 159 86 11 ­25 -- ­92 ­300 74 81 Cote d'Ivoire ­2 ­33 ­73 ­72 11 ­28 ­34 ­78 ­81 ­59 ­79 Gambia, The ­68 78 ­14 -- 13 ­12 ­46 ­86 227 54 ­40 ECOWAS Ghana ­21 ­29 ­26 73 10 0 73 ­67 ­49 ­46 ­61 Guinea 8 252 ­48 ­488 ­28 ­75 ­64 ­65 314 ­84 ­92 Guinea­Bissau 6 ­1 ­80 159 ­2 ­38 1 ­94 ­84 ­100 ­82 Liberia 20 28 ­88 ­62 ­18 ­48 -- ­42 ­99 ­100 ­95 Mali 144 ­68 ­96 ­151 ­33 ­10 -- ­84 ­94 ­85 ­85 Niger 173 ­48 ­98 ­105 ­30 ­77 76 ­90 ­98 ­96 ­94 Nigeria ­28 ­66 ­36 ­49 ­23 ­38 3 ­77 ­151 ­23 ­87 34 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Senegal 36 ­76 ­53 ­69 ­10 0 ­35 ­87 296 9 ­58 Sierra Leone ­15 107 ­85 65 -- -- ­153 ­84 ­134 ­95 ­87 Togo ­3 ­79 ­46 ­12 ­28 ­42 ­38 ­95 25 ­67 ­69 Average 21 ­4 ­50 ­59 ­11 ­37 ­24 ­81 ­12 ­49 ­64 Angola 83 6 ­39 333 ­37 ­40 ­82 ­80 ­91 ­80 ­92 Botswana 153 ­9 ­84 55 8 ­21 ­79 ­84 ­85 ­42 ­59 Congo, Dem. 37 163 ­97 ­200 ­30 ­45 ­136 ­71 ­99 ­95 ­99 Rep. Lesotho ­14 24 ­41 381 11 ­36 ­88 ­80 ­78 ­56 ­60 Malawi ­31 401 ­65 ­129 9 ­25 ­66 ­73 ­84 ­79 ­32 Mauritius ­99 ­34 122 64 41 ­14 ­45 17 ­194 ­384 152 SADC Namibia 69 359 ­79 334 12 ­43 ­53 ­100 ­81 ­67 ­49 Seychelles ­91 -- 165 -- ­7 -- -- ­62 ­34 ­31 ­14 South Africa ­33 ­34 ­11 ­60 1 21 ­23 115 ­57 ­4 ­24 Swaziland ­6 ­8 ­45 56 ­19 ­20 ­79 ­73 ­87 ­78 ­59 Zambia 23 ­9 ­60 ­48 ­26 ­9 ­53 26 ­41 ­77 ­80 Zimbabwe ­25 66 18 147 20 5 ­60 12 18 ­75 ­37 Average 6 84 ­18 85 ­1 ­20 ­69 ­38 ­76 ­89 ­38 Source: As for table 2.1. Note: Deviations larger than 500% or less than ­500% have been dropped. 35 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Annex C Regional infrastructure endowment Infrastructure category Transport Information and communication technology Energy Water Sanitation Density of fixed- Density Density Electrical Access Density of paved Density of total line of mobile of Internet generating to road network road network phones phones connections capacity electricity % km per 1,000 km2, 2001 house- subscribers MW per holds per 1 million with Region/ 1,000 people, subscribers per people, access, % households country total arable total arable 2004 100 people, 2004 2003 2004 with access, 2002 ECOWAS 38 301 144 1,279 28 72 2.4 31 18 63 35 Benin 12 55 60 273 9 -- 1.4 18 22 68 32 Burkina Faso 7 48 46 298 6 31 0.4 10 13 51 12 Cape Verde 213 2,047 273 2,625 148 133 5.3 21 20 80 42 Côte d'Ivoire 15 156 156 1,603 -- 86 1.8 55 39 84 40 Gambia 85 338 496 1,982 -- 118 3.4 21 5 82 53 Ghana 58 326 194 1,088 14 78 1.7 63 54 79 58 Guinea 20 565 124 3,427 -- -- 5.8 35 5 51 13 Guinea-Bissau 13 117 122 1,138 -- -- 2.0 13 8 59 34 Liberia 6 149 95 2,410 -- -- -- 98 5 62 26 Mali 1 39 12 319 6 30 0.5 24 8 48 45 Niger 1 18 8 225 2 11 0.2 9 8 46 12 Nigeria 65 208 210 672 8 71 1.4 43 40 60 38 Senegal 22 170 74 579 -- 90 4.7 23 30 72 52 Sierra Leone 13 187 158 2,263 -- -- 0.2 22 5 57 39 Togo 42 91 132 287 -- -- 4.4 8 9 51 34 EAC 8 93 105 1,286 6 54 2.1 24 7 64 45 Kenya 13 165 110 1,364 9 76 4.6 36 8 62 48 Tanzania 4 87 93 2,060 -- 44 0.9 24 11 73 46 Uganda 8 29 113 436 3 42 0.7 12 4 56 41 SADC 92 3,636 214 6,164 74 180 5.5 175 21 71 43 Angola 15 638 41 1,712 -- 61 1.2 50 12 50 30 Botswana 10 1,479 18 2,690 77 319 3.3 75 22 95 41 Congo, DR 2 55 67 2,262 -- -- -- 48 7 46 29 Lesotho 36 329 196 1,796 21 88 2.4 45 5 76 37 Malawi 31 132 240 1,024 7 18 0.4 27 5 67 46 Mauritius 956 1,939 980 1,989 287 413 14.6 548 100 100 99 Namibia 10 983 76 7,619 64 142 3.7 -- 34 80 30 Seychelles 700 31,532 829 37,337 253 589 24.7 359 30 87 -- South Africa 60 498 297 2,445 -- 428 7.9 883 66 87 87 Swaziland -- -- -- -- -- 101 3.3 118 20 52 52 36 MAKING SENSE OF AFRICA'S INFRASTRUCTURE ENDOWMENT: A BENCHMARKING APPROACH Zambia 10 141 121 1,716 -- 26 2.1 172 12 55 55 Zimbabwe 48 575 252 3,025 25 31 6.3 150 20 83 83 Central Africa 41 416 132 1,790 13 74 1.7 44 18 58 28 Burundi 37 97 520 1,369 -- -- 0.4 6 5 79 36 Cameroon 9 70 72 564 -- 96 1.0 56 49 63 48 Cen. Afr. Rep. 1 34 38 1,233 3 15 0.2 10 5 75 27 Chad 0.2 7 26 910 1 13 0.7 3 3 34 8 Comoros 302 841 395 1,099 -- -- 1.0 17 20 94 23 Congo, Rep. 4 670 37 6,905 4 99 0.9 32 21 46 9 Eq. Guinea -- -- 103 2,218 -- 113 1.0 20 20 44 53 Eritrea 7 150 34 689 9 5 1.2 -- 17 57 9 Ethiopia 4 33 29 268 -- 3 0.2 10 13 22 6 Gabon 3 248 32 2,510 28 359 3.0 305 31 87 36 Madagascar 10 194 89 1,755 -- 18 0.5 17 8 45 33 Mauritania 2 177 7 1,569 -- 175 0.5 42 50 56 42 Mozambique 7 139 38 744 -- 36 0.7 125 7 42 27 Rwanda 38 94 456 1,125 3 16 0.4 5 5 73 41 San Tome and 227 3,633 333 5,333 -- -- 12.2 64 20 79 24 Principe Somalia 4 245 35 2,075 25 63 1.7 8 3 29 25 Sudan 2 25 5 70 29 30 3.3 23 30 69 34 -- = data not available. 37 About AICD This study is part of the Africa Infrastructure Country Diagnostic (AICD), a project designed to expand the world's knowledge of physical infrastructure in Africa. AICD will provide a baseline against which future improvements in infrastructure services can be measured, making it possible to monitor the results achieved from donor support. It should also provide a more solid empirical foundation for prioritizing investments and designing policy reforms in the infrastructure sectors in Africa. AICD will produce a series of reports (such as this one) that provide an overview of the status of public expenditure, investment needs, and sector performance in each of the main infrastructure sectors, including energy, information and communication technologies, irrigation, transport, and water and sanitation. The World Bank will publish a summary of AICD's findings in November 2009. The underlying data will be made available to the public through an interactive Web site allowing users to download customized data reports and perform simple simulation exercises. The first phase of AICD focuses on 24 countries that together account for 85 percent of the gross domestic product, population, and infrastructure aid flows of Sub-Saharan Africa. The countries are: Benin, Burkina Faso, Cape Verde, Cameroon, Chad, Democratic Republic of Congo, Côte d'Ivoire, Ethiopia, Ghana, Kenya, Madagascar, Malawi, Mali, Mozambique, Namibia, Niger, Nigeria, Rwanda, Senegal, South Africa, Sudan, Tanzania, Uganda, and Zambia. Under a second phase of the project, coverage will be expanded to include additional countries. AICD is being implemented by the World Bank on behalf of a steering committee that represents the African Union, the New Partnership for Africa's Development (NEPAD), Africa's regional economic communities, the African Development Bank, and major infrastructure donors. AICD grew from an idea presented at the inaugural meeting of the Infrastructure Consortium for Africa, held in London in October 2005. Financing for AICD is provided by a multi-donor trust fund to which the main contributors are the Department for International Development (United Kingdom), the Public Private Infrastructure Advisory Facility, Agence Française de Développement, and the European Commission. A group of distinguished peer reviewers from policy making and academic circles in Africa and beyond reviews all of the major outputs of the study, with a view to assuring the technical quality of the work. This and other papers analyzing key infrastructure topics, as well as the underlying data sources described above, will be available for download from www.infrastructureafrica.org. Freestanding summaries are available in English and French. Inquiries concerning the availability of datasets should be directed to vfoster@worldbank.org.