WPS5191 Policy Research Working Paper 5191 Assessing the Impact of Infrastructure Quality on Firm Productivity in Africa Cross-Country Comparisons Based on Investment Climate Surveys from 1999 to 2005 Alvaro Escribano J. Luis Guasch Jorge Pena The World Bank Latin American and the Caribbean Region Sustainable Development Department January 2010 Policy Research Working Paper 5191 Abstract This paper provides a systematic, empirical assessment indices: per capita income, ease of doing business, firm of the impact of infrastructure quality on the total perceptions of growth bottlenecks, and the concept of factor productivity (TFP) of African manufacturing demeaned productivity (Olley and Pakes 1996). The firms. This measure is understood to include quality countries are divided into two blocks: high-income- in the provision of customs clearance, energy, water, growth and low-income-growth. Infrastructure quality sanitation, transportation, telecommunications, and has a low impact on TFP in countries of the first block information and communications technology (ICT). and a high (negative) impact in countries of the second. Microeconometric techniques to investment climate There is significant heterogeneity in the individual surveys (ICSs) of 26 African countries are carried out infrastructure elements affecting countries from both in different years during the period 2002­6, making blocks. Poor-quality electricity provision affects mainly country-specific evaluations of the impact of investment poor countries, whereas problems dealing with customs climate (IC) quality on aggregate TFP, average TFP, while importing or exporting affects mainly faster- and allocative efficiency. For each country the impact is growing countries. Losses from transport interruptions evaluated based on 10 different productivity measures. affect mainly slower-growing countries. Water outages Results are robust once controlled for observable affect mainly slower-growing countries. There is also fixed effects (red tape, corruption and crime, finance, some heterogeneity among countries in the infrastructure innovation and labor skills, etc.) obtained from the determinants of the allocative efficiency of African firms. ICSs. African countries are ranked according to several This paper--a product of the Sustainable Development Department, Latin American and the Caribbean Region--is part of a larger effort in the department to advocate and support the development and improvement of infrastructure services. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at jguasch@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 Assessing the Impact of Infrastructure Quality on Firm Productivity in Africa * Cross-Country Comparisons Based on Investment Climate Surveys from 1999 to 2005 by Alvaro Escribano, J. Luis Guasch and Jorge Pena§ Key words: Africa, Infrastructure, Total Factor Productivity, Investment Climate, Competitiveness. * 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 are indebted to Vivien Foster and her group of researchers at the World Bank for providing data sets and further insight into the African region. All errors in this report are the sole responsibility of the authors. Telefónica-UC3M chair in the economics of telecommunications, Department of Economics, Universidad Carlos III de Madrid; alvaroe@eco.uc3m.es The World Bank and the University of California, San Diego; jguasch@worldbank.org. § Department of Economics, Universidad Carlos III de Madrid; jpizquie@eco.uc3m.es. 1. The wide-ranging effects of infrastructure quality For Africa's awaited growth resurgence to occur, a broad range of factors--political, institutional, and economic--must be improved. The World Bank's landmark Africa Competitiveness Report (ACR) (2004 and 2007) focuses on problems that, in the words of Artadi and Sala-i-Martin (2003), constitute the most important growth tragedy of the twentieth century--a phenomenon that has received special attention in recent growth literature, such as that of Ndulu and O'Connell (2005). It is agreed that improving Africa's infrastructure is a crucial step toward penetrating international markets and meeting the goals of continuous growth and poverty reduction. Infrastructure quality has a pervasive influence on all areas of an economy. Low-quality infrastructure and limited transport and trade services increase logistical and transaction costs, rendering otherwise competitive products uncompetitive, as well as limiting rural production and people's access to markets--with adverse effects on economic activity and poverty reduction. A large number of empirical studies illustrate the impact of infrastructure on economic performance, including those of Calderón et al. (2003a and b), Calderón and Serven (2003), Canning (1998), Reinikka and Svensson (1999), Prud'homme (2004), Escribano and Guasch (2005), Escribano et al. (2005), and Guasch (2004). All suggest that Africa's infrastructure gap is an important growth bottleneck with a negative impact on productivity and the overall competitiveness of the region. Furthermore, several studies using the methodology of Escribano and Guasch (2005, 2008) and Escribano et al. (2008a and b and 2009) have found empirical evidence--in cases such as Brazil, Chile, Costa Rica, Mexico, Turkey, and several southeast Asian countries--that improvements in investment climate (IC) conditions in general, and in infrastructure quality in particular, may lead to important gains in productivity and in other economic performance measures: employment, real wages, exporting activities, and foreign direct investment (FDI) inflows. Disentangling the ways that infrastructure affects Africa's economic growth poses several difficulties because of the special characteristics of the African region. The comprehensive analysis found in Estache (2005) takes stock of the basic characteristics of infrastructure in Sub-Saharan Africa and the impact of 1990 reforms, pointing out that the impact of infrastructure in Africa may be different than in other regions. As Brunel (2004) signals, the colonial period has had a lasting effect on the use of space in the region, resulting in a productive structure that consists, in most cases, of coastal cities connected inland by railways designed to carry raw materials to main ports. This and other factors that are progressively modifying the continent's productive structure--such as continuous urbanization, the movement of economic activity from the agricultural to manufacturing and service sectors, and the increasing openness of African economies--has caused both a quantitative and qualitative mismatch between the current supply of infrastructure and ever-increasing demand. Factors such as inequality across income levels (affecting the affordability of infrastructure services), large and unoccupied areas, and regional variations in climate are increasingly becoming a concern for African policy makers managing infrastructure. 2 In addition to furthering the regional integration needed to support infrastructure investment, African governments made important contributions to infrastructure development in the decades following independence. The majority of African state monopolies were, however, characterized by inefficient bureaucracies. These became increasingly unable to satisfy customer demands, with increasing deficits. By the beginning of the 1980s, most African countries embarked on infrastructure sector reforms, with the aim of increasing private sector participation in provision. Despite attempts to introduce more competition and to attract private investors, Africa continues to trail the world in both the quantity and quality of its infrastructure, with bottlenecks particularly in the management of current stock. Figure 1.1 Geographical locations of the 26 countries considered in the investment climate assessment (ICA) Source: Authors' elaboration. 3 Figure 1.1 shows the geographical distribution of the countries considered in this study, both in North and Sub-Saharan Africa. The countries studied are divided into five main geographical areas, identified in some cases by the major multilateral organization of each region: (a) the North African region, or Maghreb, includes Morocco, Algeria, and Egypt; (b) the Economic Community of West African States (ECOWAS) includes Mauritania, Mali, Niger, Senegal, Burkina Faso, Benin, Cameroon, and Cape Verde; (c) the Horn of Africa region is composed of two countries, Eritrea and Ethiopia; (d) the East African Community (EAC) includes Tanzania, Kenya, Uganda, and Burundi; and (e) the South African Development Community (SADC), for which we have data for Malawi, Zambia, Namibia, Botswana, Swaziland, Lesotho, and Madagascar. South Africa and Mauritius are the last two individual countries included in the report. The 26 countries show enormous heterogeneity due to (a) geographical factors, such as whether a nation is landlocked (Cape Verde, Madagascar, and Mauritius), tropical (with landmass for the most part covered by rainforests), or dominated by deserts (such as the North African countries Mauritania and Namibia); (b) social or political factors, such as civil wars, armed conflicts, early democracies, dictatorships, and colonial heritage; and (c) economic factors, which this paper discusses for all countries, from the most affluent (Mauritius) to the poorest (Eritrea). Figure 1.2 clarifies the different evolutions of per capita income across the countries included in this analysis. Of the 26 African countries analyzed, Mauritius was, in 1950, the country with the highest per capita income (measured in terms of per capita gross domestic product, GDP), followed closely by South Africa, and, by a wider gap, Namibia and Algeria. But the per capita income situation in 2003 was somewhat different; Mauritius was still ranked first, followed by Swaziland, South Africa, and Botswana--and, by a wider gap, Algeria, Cape Verde, Egypt, and Morocco. Panel B of figure 1.2 shows the five-year growth rate of per capita income. Mauritius and Botswana are the countries that have experienced the highest, sustained per capita income growth during the recent years. Lesotho is the median country, splitting the cross-section into two blocks. The first block comprises countries with faster and steadier growth rates (Mauritius, Swaziland, South Africa, Botswana, Namibia, and Lesotho in the south; Algeria, Morocco, and Egypt in the north; and Cape Verde and Cameroon in central Africa). In the second are countries with lower and more irregular growth rates (Mauritania, Senegal, Benin, Mali, Niger, and Burkina Faso in the central west; Uganda, Kenya, Zambia, Tanzania, Malawi, Burundi, Madagascar, Ethiopia, and Eritrea from the central east), periods of positive expansion fluctuate with those of persistent reductions in per capita income. 4 Figure 1.2 The evolution of gross domestic product (GDP) per capita and ranking based on the ease of doing business in African countries A. Country by country evolution of GDP per capita, 1950 - 2003 B. Country by country five-year rate of growth of GDP per capita, 1950 - 2003 C. Ranking of countries on the ease of doing business, World Bank's Doing Business Report MUS SWZ ZAF BWA DZA NAM CPV EGY MAR CMR LSO MRT SEN BEN KEN MLI UGA BFA ZMB TZA NER MWI BDI MDG ETH ERI 2007 32 76 29 48 116 42 125 165 115 152 114 148 146 137 83 155 107 163 102 142 160 110 166 149 97 170 2006 32 67 28 44 123 39 125 165 117 147 116 146 152 139 80 166 103 171 92 150 170 106 160 148 96 168 Source: Source: Penn World Table, Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania, 2006 and World Bank's Doing Business Report 2007. 5 These per capita income rankings are correlated with the rankings obtained from the World Bank's 2007 Doing Business Report (DBR), presented in panel C of figure 1.2. In 2007 Mauritius, Swaziland, South Africa, Botswana, and Namibia rank 32nd, 76th, 29th, 48th, and 42nd in the world based on the ease-of-doing-business indicators. This index considers questions such as the number of days required to start a business and the ease of dealing with licenses, registering a property, trading across borders, employing workers, and so on. Other 2007 rankings include 83rd for Kenya, 97th for Ethiopia, 165th for Egypt, and 170th for Eritrea. To better understand the convergence or divergence of trends, we plotted the per capita income of each African country relative to the per capita income of the United States (see panel A of Figure 1.3). Convergence is observed only in Mauritius, Swaziland, and Botswana. For all other study countries, including South Africa, per capita income is diverging from the United States, while in a few (Egypt, Morocco, and Cape Verde) the ratio has remained stable. While persistently positive GDP growth allowed Mauritius's per capita income to reach 45 percent of the United States in 2003, this is clearly the exception (together with Swaziland and Botswana). For the rest of the countries, including South Africa, relative per capita income was much lower in 2003 than in 1960 (indicating divergence). In fact, the 2003 per capita income of several countries was no larger than 5 percent of the per capita income of the United States. As expected, labor productivity is the main factor explaining this divergence in per capita income in Africa (panel B of Figure 1.3), given that labor force participation has a steady influence (panel C of Figure 1.3). 5 Since TFP is usually a key factor explaining the evolution of labor productivity, in this paper we seek to use investment climate surveys (ICSs) to identify the main infrastructure-related TFP bottlenecks in Africa. 5 The per capita income of country J (YJ/PJ) is decomposed into labor productivity (YJ/LJ) and the employment-population rate (LJ/PJ) by following the expression: (YJ/PJ) = (YJ/LJ)*(LJ/PJ), where Y is GDP, L is total labor force, and P is total population. 6 Figure 1.3 Evolution of per capita income in Africa relative to the United States, 1960­2003 Per capita income of country J (YJ/PJ) is decomposed into the product of labor productivity (YJ/LJ) and the labor participation rate (LJ/PJ) by: (YJ/PJ)= (YJ/LJ)*(LJ/PJ). Therefore, per capita income relative to the United States becomes: [(YJ/PJ)/(YUS/PUS)]= [(YJ/LJ)/ (YUS/LUS)]*[(LJ/PJ)/ (LUS/PUS)]. Source: Penn World Table, Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania, 2006. Figure 1.4 shows the percentage of firms that perceive telecommunications, electricity, customs clearance, and transport as major obstacles to their economic performance. Only in Benin, Kenya, and Zambia do more than 50 percent of firms identify telecommunications as a severe obstacle. The quality 7 of electricity provision is a major problem for more than 50 percent of firms in more than half of the countries in our sample. In Burundi, Cameroon, Benin, Burkina Faso, and Cape Verde, the percentage of firms considering electricity as a severe or very severe obstacle exceeds 80 percent; on the other hand, only 20 percent of firms in Morocco, South Africa, Botswana, and Namibia consider electricity a severe obstacle. Customs clearance is considered an acute problem in Benin, Kenya, Madagascar, Senegal, and Algeria. Finally, transportation is considered a severe obstacle by more than 70 percent of firms in Burkina Faso and Benin. Figure 1.4 Percentage of firms that consider telecommunications, electricity, customs, and transport as severe or very severe constraints on economic performance (by country) A. Telecommunications B. Electricity Benin Burundi Kenya Cameroon Zambia Benin Niger Burkina Faso Ethiopia Cape Verde Lesotho Tanzania Madagascar Malawi Burkina Faso Kenya Algeria Madagascar Malawi Uganda Cameroon Ethiopia Tanzania Mauritania Swaziland Zambia Mali Eritrea Cape Verde Lesotho Eritrea Senegal Botswana Mali Burundi Swaziland Mauritania Niger Namibia Egypt Uganda Algeria Mauritius Mauritius South Africa Morocco Senegal South Africa Egypt Botswana Morocco Namibia 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Percentage of firms Percentage of firms C. Customs clearance D. Transport Benin Burkina Kenya Benin Madagas Cameroon Senegal Kenya Algeria Senegal Cameroon Zambia Tanzania Malawi Ethiopia Niger Zambia Burundi Uganda Madagascar Niger Tanzania Malawi Uganda Mauritius Cape Verde Mauritania Lesotho Mali Mali Egypt Mauritania South Namibia Lesotho Swaziland Cape Ethiopia Burundi Mauritius Burkina South Africa Morocco Botswana Namibia Eritrea Swaziland Morocco Botswana Egypt Eritrea Algeria 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Percentage of firms Percentage of firms Source: Authors' calculations based on IC data. Note: No data are available for perceptions of transport in Algeria. 8 Figure 1.5 offers another view of the state of infrastructure in Africa. The World Bank's ACR (2007) evaluates a wide range of factors related to economic activity, infrastructure among them. Once again there are clearly different performance levels across the two blocks of countries. While in Namibia, South Africa, Botswana, Egypt, and Morocco the quality of infrastructure exceeds the approval level; in the remaining countries this quality is rated low in most cases. The same holds for the disaggregated results, including the number of telephone lines and the quality of ports, air transport, and electricity supply. Figure 1.5 The state of infrastructure in Africa, at first glance A. Quality of overall infrastructure B. Quality of railroad infrastructure Namibia Namibia South Africa South Africa Mauritius Egypt Botswana Botswana Egypt Morocco Morocco Tanzania Algeria Burkina Tanzania Algeria Uganda Lesotho Mali Kenya Malawi Mauritania Kenya Malawi Madagasca Cameroon Lesotho Mauritius Ethiopia Mali Benin Uganda Burkina Madagasca Burundi Ethiopia Cameroon Benin Mauritania Burundi 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 Score (1-7) Score (1-7) C. Quality of port infrastructure D. Quality of air transport infrastructure Nam ibia South Africa Mauritius Mauritius South Africa Kenya Morocco Namibia Egypt Egypt Tanzania Morocco Kenya Ethiopia Algeria Botswana Botswana Tanzania Benin Algeria Mauritania Uganda Madagascar Madagascar Cam eroon Burkina Uganda Malawi Lesotho Mali Burundi Benin Burkina Mauritania Malawi Burundi Mali Lesotho Ethiopia Cameroon 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 Score (1-7) Score (1-7) E. Quality of electricity supply F. Telephone lines Mauritius Mauritius Nam ibia Egypt Morocco South Africa Egypt Algeria Botswana Botswana Algeria Nam ibia South Africa Morocco Ethiopia Lesotho Mali Mauritania Lesotho Benin Kenya Malawi Burkina Fasso Kenya Mauritania Ethiopia Benin Mali Malawi Burkina Cam eroon Cameroon Tanzania Tanzania Madagascar Madagascar Burundi Burundi Uganda Uganda 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Score (1-7) Total number of telephone lines per 100 inabitants Source: Africa Competitiveness Report (2007), World Bank, Washington, DC. Note: No data are available for Cape Verde, Eritrea, Niger, Senegal, Swaziland, or Zambia. 9 The difference between the two blocks becomes even clearer in figure 1.6, where the cross-plots between GDP per capita relative to the United States and firms' perceptions are presented. A preliminary analysis of the cross-plots suggests two points: first, that there is an intuitive and negative relation between income level and infrastructure constraints; and, second, that the diversion of the two blocks of countries remains intact, showing now the largest dispersion in the constraint perceptions of figure 1.6 for the lowest per capita income group. Figure 1.6 A simple illustration (cross-plots) of the relation between per capita GDP and infrastructure perceptions of severe or very severe obstacles to growth in Africa GDP per capita relative to United States A Telecom unications vs. GDP per capita . m B. Electricityvs. GDP per capita 100 100 90 90 80 80 70 70 %of firm considering electricity as a severe %of firm considering telecom as a severe 60 60 obstacle obstacle s 50 50 40 40 30 30 20 20 s s 10 10 0 0 0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50 GDP per capita (%of US) GDP per capita (%of US) s B. Custom clearance vs. GDP per capita D. Transport vs. GDP per capita 100 100 90 90 80 80 % of firms considering customs as a severe % of firms considering transport as a severe 70 70 60 60 obstacle obstacle 50 50 40 40 30 30 20 20 10 10 0 0 0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50 GDP per capita (% of US) GDP per capita (% of US) Source: Authors' calculations based on IC data. Note: No data are available on perceptions of transport in Algeria. The objective of this paper is to assess the impact of the quality of existing infrastructure on the TFP of African firms. This measure is understood to include quality in the provision of the following services: customs clearance, energy, water, sanitation, transportation, telecommunications, and 10 information and communications technology (ICT). We also want to identify infrastructure factors with statistically significant impacts on TFP, country by country. In the econometric evaluation we use 10 different measures of TFP and show that the results are robust--no matter what measure of TFP is used--if we follow the econometric methodology of Escribano and Guasch (2005, 2008), and Escribano et al. (2008). For the empirical analysis of infrastructure's constraints on TFP, we go down to the firm level since infrastructure is one of the key elements of a country-specific IC, and a significant component of country competitiveness. To provide reliable and robust estimates of the impact of infrastructure on economic performance is not a straightforward task. As we will see later on, we have to deal with the heterogeneity of the countries included in our sample, and the endogeneity of explanatory variables (inputs and IC variables) in several dimensions due to unobservable fixed effects, measurement errors, missing observations, and so on. To solve these problems, we take advantage of the useful and rich firm-level information provided by the ICSs undertaken by the World Bank in Africa from 2002 to 2006. These surveys capture firm-level information in a range of areas related to economic performance: infrastructure, financing, governance, corruption, crime, regulation, tax policy, labor relations, conflict resolution, supplies and marketing, quality, technology, and training, among others. These surveys offer information on the production function (PF) variables over one, two, or three years, depending on the African country. But for infrastructure and other IC and plant control (C) variables they only provide information for a single year. We will see how we can use this valuable information to evaluate how firms operate in Africa and to identify the main obstacles to productivity improvements. Section 2 of this report clarifies the link between this type of empirical work and existing literature on infrastructure and productivity. The properties and quality of the ICSs are analyzed in section 3. Why we classify the IC factors in broad categories or groups will also be discussed, together with the infrastructure variables (INFs) used. In section 4 we present the econometric methodology we use to estimate the impact of infrastructure and other IC variables and C characteristics on TFP. Once we have estimated the infrastructure and other IC elasticities and semi-elasticities on productivity, we evaluate the effects of infrastructure on aggregate productivity and on allocative efficiency, using the Olley and Pakes (O&P, 1996) decomposition. The main empirical results are described in the remaining sections. In particular, section 5 focuses on the relative importance of infrastructure in the IC of each country. Section 6 presents the empirical results country by country, and section 7 includes the main conclusions. Most of the tables and figures are included in the Appendix at the end of the paper. 2. How does infrastructure quality affect economic performance? Much literature discusses the different ways that infrastructure affects growth and other development outcomes at the macroeconomic level. For example, the World Bank's landmark World Development Report (1994) highlighted multiple links between infrastructure and development and emphasized how 11 policy can improve not only the quantity, but also the quality, of infrastructure services in developing countries. As Straub (2008) signals, macro level literature has too often sought to obtain the elasticity of infrastructure capital and compare it with the elasticity of private capital. Few papers go beyond measures of infrastructure spending and infrastructure stocks to consider the issue of infrastructure efficiency. Since the seminal paper of Aschauer (1989) found that infrastructure capital has a large impact on aggregate TFP, this finding has been replicated by a number of earlier studies: Munnell (1990a, 1990b, 1992) for the United States, Mitra et al. (2002) for that of India, and Easterly and Rebelo (1993) for cross-sectional country data. Loayza, Fajnzylber, and Calderón (2002) find that a telecommunications indicator is robustly related to growth in a large panel data set that includes both industrial and developing countries. For the case of Africa, studies exploring the relation between infrastructure and growth are 6 scarce. Traditionally, infrastructure services have been viewed as public goods in Africa, with their provision entrusted to government monopolies. The overall performance of government-owned providers of infrastructure in Africa has been very poor. This sector is characterized by high inefficiency, a lack of technological dynamism, and very poor service provision. In addition, the provision of infrastructure-related services in most African countries is characterized by high prices and long waits between the time of application for services and actual connection. Many African economies are also endowed with adverse natural and geographical attributes, such as lack of access to sea ports and tropical climates. The Economic Commission for Africa (ECA) Report (2005) and Sachs et al. (2004) have explored the African need for new investments in infrastructure, but without a properly systematic cross-country analysis. Estache et al. (2005) makes one of the first attempts to conduct a more systematic, quantitative assessment of the importance of Sub-Saharan Africa's infrastructure. The main finding of this paper is that electricity, water, roads, and telecommunications are crucial factors in promoting growth, with colonial and postcolonial histories also being important factors explaining some of the differences among countries. On the other hand, Esfahani and Ramirez (2003) estimate that Sub-Saharan Africa's poor growth performance is, in part, related to under-investments in electricity and telecommunications infrastructure. Estache (2005) estimates that if Africa had enjoyed Korea's quantity and quality of infrastructure, it would have raised its annual growth per capita by about 1 percent. Hulten (1996) finds that differences in the effective use of infrastructure resources explain one- quarter of the growth differential between Africa and East Asia, and more than 40 percent of the growth differential between low- and high-growth countries. Empirical explorations of infrastructure's effect on growth and productivity, however, have been characterized by ambiguous results with little robustness. The possible endogeneity of infrastructure measures has been advanced as a reason for contradictory findings of the impact of public 6 Estache (2005) points out the two main reasons for ignoring the role of infrastructure as one of the most important drivers of economic growth in Africa: (a) econometric focus on human capital, and (b) low quality of available data. 12 capital on long-run economic development indicators. Literature has signaled that endogeneity in this context might come from three sources: (a) measurement errors stemming from the use of public capital figures as proxies for infrastructure; (b) omitted variables, which may arise when there is a third variable, unobserved, that affects the infrastructure and growth measure; and (c) the fact that under the simultaneous determination of infrastructure and productivity or output, the bias and inconsistency of standard estimators would follow where infrastructure provision itself positively responds to productivity gains. 7 Possible reasons for such feedback would arise with increased reliance on the private sector for the provision of infrastructure services, or with successful lobbying by industry interest groups that experience either positive productivity gains or constraints on performance due to infrastructure provision. Various panel data and country studies have tried to address these issues. Röller and Waverman (2001) explicitly model and estimate the impact of telecommunications under simultaneity. In a cross- country panel estimation, Calderón and Serven (2003, 2005) employ generalized method of moments (GMM) panel estimation methods to control for the possibility of endogeneity, reporting significant improvements in results. Dessus and Herrera (1999) allow for simultaneity in a panel data set for 28 countries. Country-specific time series also confirm the presence of simultaneity between output and infrastructure measures--see Frutos et al. (1998) for Spain, and Fedderke et al. (2005) for South Africa. Also for South Africa, Fedderke and Bogetic (2006)--controlling for the potential endogeneity of infrastructure in estimation--robustly eliminate nearly all evidence of possible overinvestment in infrastructure. Indeed, controlling for the possible endogeneity of infrastructure measures renders the impact of infrastructure capital positive. Romp and Haan (2005) indicate that when simultaneity is taken into consideration, the elasticity estimates found in earlier studies considerably decrease. Another possibility behind the ambiguous results obtained from empirical studies of public capital impacts on output might simply be that aggregate measures of infrastructure hide the productivity impact of infrastructure at a more disaggregated level. A second batch of studies, focusing mainly on micro data, reveals the existence of the possible indirect impact of infrastructure on economic growth and economic performance beyond the effect of the simple accumulation of capital. Thus, for instance, Shioji (2001) finds that the positive impact of infrastructure arises in panel data on U.S. and Japanese industries once public capital is properly disaggregated. Agénor and Moreno- Dodson (2006) point out that improvement in the stock of infrastructure can reduce the adjustment costs of private capital by (a) lowering the logistical cost of the investment in private capital, and (b) allowing the replacement of unproductive private investments such as electricity generators or boreholes and wells with more productive investments in machinery and equipment. This assumption has been tested in the context of investment climate assessments (ICAs) with firm-level information. Reinikka and Svensson (1999) show that improvements in the infrastructure stock in Uganda make 7 Notice that we avoid using the terms causality or reverse causality, since there is no control group to compare against and the temporal dimension is not large enough to consider Granger-causality concepts. Therefore, we use the terms simultaneity and identification, which are more appropriate for ICSs. 13 infrastructure services more reliable, reducing the necessity of investing in less productive substitutes (such as generators) in order to avoid potential service interruptions, and thus freeing funding of private productive investments. Relationships at a more disaggregated level tend to be obscured by aggregated data, and are unobservable with country-level data. Another channel of infrastructure impact is via improvements in labor productivity through (a) improved transport between home and work, and (b) more efficient work processes. Another way that better infrastructure might increase labor productivity is through improvements in health and education, making existing human capital more efficient, and promoting successive investments in human capital (Galiani et al., 2005). The effect of infrastructure on firms' international integration has also been tested. Recent literature affirms that improvements in transportation services and infrastructure can lead to improvements in export performance. Thus, for instance, Francois and Manchin (2006) explore the role that infrastructure plays (among other factors such as policy reforms, institutional development, colonial history, development assistance, and general north-south differences) in the different trade performances observed in the so-called globalizer countries such as India and China, as well as other developing countries (many located in Africa and with a very different story to tell regarding the integration of the global economy). Limão and Venables (2001) show that infrastructure is quantitatively important in determining transport costs, concluding that poor infrastructure accounts for much of the different transport costs observed in coastal and landlocked countries. Bougheas et al. (1999), in the context of gravity models, find evidence in the European economy of a positive relationship between the level of infrastructure and the volume of trade. Wilson et al. (2004) consider ports, customs, regulations, and e-businesses as proxies of trade-facilitation efforts, finding that the scope and benefit of unilateral trade-facilitation reforms are very large, and that the gains fall disproportionately to exports. In a world where governments compete to attract more FDI inflows through a variety of investment and tax incentives and other policy preferences, the availability of good-quality physical infrastructure could also increase the inflow of FDI by subsidizing the cost of total investment by foreign investors and thus raising the rate of return. The favorable role of physical infrastructure in influencing patterns of FDI inflows has been corroborated by recent studies, such as those of Loree and Guisinger (1995), Mody and Srinivasan (1996), and Kumar (2001), among others. Multinational enterprises may consider the quality of available infrastructure especially important while deciding to relocate export-platform production undertaken for efficiency considerations. In other words, the quality of physical infrastructure could be an important consideration for multinationals in their location choices, for FDI in general, and for efficiency-seeking production in particular. As has been pointed out, the main concern of this paper is to offer a robust assessment of the various channels through which infrastructure quality may impact TFP. Thus, instead of the quantity of macro variables, we use, as an explanatory variable, the quality of existing infrastructure stock. Instead of aggregate infrastructure measures usually included in macro models, such as kilometers of paved roads or total number of telephone lines, we incorporate measures that allow us to identify direct 14 relationships between infrastructure and economic performance at a more disaggregated level. Additionally, by going down to the firm level, we avoid the endogeneity problems of the macro level variables. Nevertheless, micro level data have specific endogeneity problems, and several variables cannot be considered to be exogenously determined; for instance, public investment decisions are likely to be affected by expected returns on investment, and firms faced with different quality and availability of infrastructure services would choose different technologies. The solutions proposed in this methodology allow us to obtain a robust assessment of the impact of infrastructure quality on TFP. 3. Country-level data and their treatment in the study Produced by the World Bank, ICSs of private enterprises explore the difficulties that firms located in developing countries encounter in starting and running businesses. More precisely, the surveys capture firms' experiences in a range of areas related to economic performance: financing, governance, corruption, crime, regulation, tax policy, labor relations, conflict resolution, infrastructure, supplies and marketing, quality, technology, training, and so on. For that purpose, we classify IC factors in five categories to evaluate the impact of each group on economic performance. In the first group-- infrastructure--we include all related variables such as customs clearance, power and water supply, telecommunications (including phone connection and information technology, IT), and transportation. In the second group--red tape, corruption, and crime--we include IC factors relating to tax rates, conflict resolution, crime, bureaucracy, informalities, corruption, and regulations. The next group comprises financial and corporate governance and includes factors related to management, investments, informalities in sales and purchases, access to finance, and accountability (or auditing). The last group of IC variables includes quality, innovation, and labor skills, as well as quality certifications, technology usage, product and process innovation, research and development (R&D), quality of labor, training, and managers' experience and education. The last group--other C variables--are not properly a group of IC factors, but a group of other firms' control characteristics. We classify in this group all the factors that may have an important impact on economic performance but are not considered IC factors: exports and imports, age, FDI, number of competitors, firm size, and so on. Table A.2 (see the Appendix) includes the whole list of IC and C variables, as well as a description of how each is measured. Likewise, not all surveys provide the same information on ICSs, although there is a common group of variables in each group that is available for all the countries; although the regressions among them are slightly heterogeneous, we can use this common group as a benchmark for comparison purposes. The ICSs provide information on TPF variables, output (sales), employment, intermediate materials, capital stock, and labor costs. Table A.1 includes information on these variables and indications of how they were measured. The ICSs do not provide information on prices at the firm level, so the production function (PF) variables were deflated by using the World Bank's country- specific consumer price index (CPI), base 2000. The information on the net book value of the capital stock (NBVC) is not available for Algeria, Kenya, Mali, Senegal, and Uganda; in these cases the 15 NBVC is substituted by the replacement cost of machinery and equipment, which, in the surveys, is only available for a single year. We thus recursively estimate the missing values of the NBVC from the information on the replacement cost of and the net investment in machinery and equipment by using the permanent inventory method, according to which the capital stock at moment t is given by Kit=K it-1(1- )+ Iit. By inverting this formula we can obtain the value of the capital at moment t-1 as Kit-1=(K it - Iit)/(1-) where Kit is approximated with the replacement cost of machinery and equipment, Iit is the net investment in machinery and equipment, and is the depreciation rate of the machinery and equipment. 8 In this paper we focus on the manufacturing sector, and while classifying the establishments by their international standard of industrial classification (ISIC) code, we end up with establishments from the following eight sectors: (a) food and beverages; (b) textiles and apparels; (c) chemicals, rubber, and plastics; (d) paper, printing, and publishing; (e) machinery and equipment/metallic products; (f) wood and furniture; (g) nonmetallic products; and (h) other manufacturing. Classification of countries by geographical area For the classification of countries by groups used in the regression analysis, we take into account the following facts: (a) the surveys provide different information on PF variables and on IC and C variables; (b) the surveys were carried out in different years during the period 2002­6; (c) the quality of the data varies across surveys; and (d) not all the surveys provide panel data information (recall data) for the PF variables. Thus, we end up with two types of country databases. For those countries with a large enough number of observations available for regression analysis (see column 6 of table 3.1) and with panel data information for the PF variables (for more than one year), we carry out the analysis country by country. For the countries in which surveys were collected in 2006 (which only offer one year of information for PF variables) and the number of firms surveyed was lower than in the previous surveys, we follow the estimation strategy of pooling the information according to the similarity of geographical and economic factors--thus gaining efficiency in the parametrical estimation of the IC parameters (with more observations in the regressions) at the cost of having common IC parameters for some countries. We end up with two pools of 2006 countries: (a) ECOWAS countries, such as Mauritania, Cameroon, Niger, and Burkina Faso; and (b) SADC countries, such as Botswana, Namibia, and Swaziland. Finally, since Eritrea has only 179 observations available, we consider this country as a special region of Ethiopia and carry out a joint analysis of the two, constituting the third pool of countries considered in the analysis. 8 The depreciation rate used is 15 percent, a standard percentage commonly applied in other works. Other percentages were also used in order to check robustness. Alternatively, to check whether the results were robust for other ways of constructing the NBVC, we used the next formula Kit-1=K it(1-Iit), where Kit is approximated by the replacement cost and Iit=(Iit-Iit-1)/Iit- 1 is the rate of growth of the net investment in machinery and equipment. In both cases the main results were maintained. 16 Table A.3 Summary of the investment climate assessment (ICA) surveys, sorted by geographical area Years of production Final number of Year of Total number of function (PF) observations available for survey observations1 variables regression analysis2 Algeria 2002 2000­1 952 706 North Africa Egypt 2004 2001­3 2,931 2,629 Morocco 2003 2000­2 2,550 2,422 Senegal 2003 2000­2 783 535 Benin 2004 2001­3 591 475 Mali 2003 2000­2 462 309 Economic Community of Cape Verde3 2006 2005 47 47 West African States (ECOWAS) Mauritania* 2006 2005 80 80 Burkina Faso* 2006 2005 51 51 Niger* 2005 2004 64 48 Cameroon* 2006 2005 119 118 Ethiopia** 2002 1999­2001 1,281 1,142 Horn of Africa Eritrea** 2002 2000­1 237 179 Kenya 2003 2000­2 852 577 East African Community Uganda 2003 2001­2 900 635 (EAC) Tanzania 2003 2000­2 828 561 Burundi3 2006 2005 102 101 Malawi 2005 2004­5 320 288 Madagascar 2005 2002­4 870 623 Southern African Zambia 2002 1999­2001 564 417 Development Community Lesotho3 2003 2000­2 225 79 (SADC) Botswana*** 2006 2005 114 112 Namibia*** 2006 2005 106 104 Swaziland*** 2006 2005 70 69 Mauritius 2005 2002­4 636 417 South Africa 2003 2001­2 1,737 1,492 Source: Authors´ calculations; ICA data. Note: 1 Total number of observations is equal to the total number of firms surveyed multiplied by the total number of years. 2 The observations available for regression analysis are the total number of observations minus the observations with any PF variable missing and/ or outlier after the cleaning process. 3 Countries for which no regression analysis was conducted. * Countries pooled for regression analysis: Mauritania, Burkina Faso, Niger, and Cameroon. ** Countries pooled for regression analysis: Ethiopia and Eritrea. *** Countries pooled for regression analysis: Botswana, Namibia, and Swaziland. Table A.3 offers an initial overview of the data we use in the analysis. We have data for 26 countries from five different geographical regions. Cape Verde, Lesotho, and Burundi are special cases. The PF information for Lesotho is rather poor and it is impossible to make reliable statistical inferences 17 with only 79 observations. We did not group Lesotho with the pool of SADC countries because the survey of this country is from 2003 and the information on the IC and C variables is quite different. Burundi presents similar problems--the information on the PF for this country is for a single year (2005), and the number of observations is only 101. Although Burundi belongs to the EAC--along with Kenya, Uganda, and Tanzania--we did not pool Burundi with any of these countries because the information on the PF and the IC comes from different years and with different information on the IC and C variables. Cape Verde is another country with information for a single year (2006) and with only 47 observations available for regression analysis. Because of its obvious difference from the rest of the ECOWAS countries--different per capita income and its condition as an insular state, as well as other geographical considerations--we did not include Cape Verde in this pool. As a result, no regression analyses were conducted for Cape Verde, Lesotho, and Burundi. By running the regressions country by country we can use as many infrastructure and other IC and C variables as are available. This allows us to gain heterogeneity estimating the impact of infrastructure on productivity. In addition, we can use more variables as proxies for firm-level, unobservable fixed effects, and we do not have to constrain ourselves to the subset of IC variables common to all the countries. Cleaning the data The IC databases are, in some respects, troublesome. From table B.1 (see the Appendix) it is clear that out of the total number of establishments surveyed there are a considerable number of observations with at least one PF variable missing, and/or with outlier observations in the PF variables. 9 This problem becomes more acute for some countries--such as Algeria, Senegal, Eritrea, Tanzania, and Mauritius--where more than half the observations are missing for the regression analysis (see the upper panel of table B.1), which results in the sample representativity being lost. To reduce the effects of this sample selection bias, we apply a preliminary data-cleaning process that allows us to retrieve a considerable number of establishments for the analysis; it is based on a robust simple version of the EM-algorithm of Dempster et al. (1977) (for more details see Little and Rubin, 1987; Escribano et al., 2008). First, we exclude those plants with missing values in all the PF variables--sales, materials, capital stock, and labor cost. We convert outlier observations of PF variables into missing observations, then proceed as follows: (a) we replace the missing values by the corresponding (cells) industry-region- size median of the variables keeping from 15 to 20 observations in each cell; (b) if we do not have enough observations in some cells we replace them with the corresponding industry-size medians; (c) if we still do not have enough observations in those cells we replace them with the region-size medians; and (d) if still necessary, in the last step we compute the medians only by size and/or by industry to replace those missing values. Table B.1 shows that the number of available observations in all the countries considerably increases with the application of this data-cleaning process. Tables B.2.1 and B.2.2 (see the Appendix) show the distribution of the observations (by country and year, and by 9 By outliers we mean those observations with ratios of materials to sales and/or labor cost to sales greater than 1. 18 country and industry, respectively) before and after the cleaning process. From these tables it is clear that this process does not alter much of the original representativity of the ICSs. The importance of infrastructure among IC variables As has been previously pointed out, we classify the IC factors in several categories to evaluate the impact of each group on economic performance. The infrastructure group of variables (INFs) is intended to be part of the country-specific IC. Within the infrastructure group we consider the next list of IC variables: customs clearance, energy, water, telecommunications, ICT, and transportation. Table B.4 describes the main INFs used in the empirical analysis. The variables listed in table B.4 are common to almost all the countries considered, and are therefore intended to be a benchmark for comparison purposes; however, there are other country- specific variables not listed. For a description of the complete set of variables, along with the countries for which they are available and the response rate of the variables, see table B.3 in the Appendix. Within each infrastructure subgroup we consider different factors. Thus, in the customs clearance group the factor considered is the time required to clear customs for exports and imports, and the time to get an import license. In the energy group we consider variables that describe the quality of power provision (number and average duration of power outages, and subsequent losses), the use of a generator as a substitute for the public provision of power, the price of energy either from the public grid or from private generators, and the average time waiting to be hooked up for electricity supply. Similarly, for the group of water we consider provision quality, price, the use of alternative supplies of water (such as private wells or boreholes), and the time to get water supply. In the telecommunications and ICT group, the variable considered is the quality of the phone provision and the time to obtain a phone connection, as well as the use of ICT technologies (such as Internet or e-mail) in communications with clients and suppliers. The transport group mainly incorporates a description of the quality of transportation services and dummy variables for the use of own-transport services (roads, transportation for workers, and so on). From the econometric point of view we use three types of variables: (a) variables in logs, whose coefficients are interpreted as elasticities; (b) variables in percentages, whose coefficients can be interpreted as semi-elasticities; and (c) dummy variables, for which coefficients from the regressions are interpreted as semi-elasticities. Finally, some of the variables in the same group are likely to be correlated since they provide similar information; for instance, the number and average duration of power outages and subsequent losses. In order to avoid multicolinariety problems we do not simultaneously use all variables in the regressions, but in the final model specification we test for possibly omitted variables. 19 Table B.4 Classification of the main infrastructure variables (INFs) Name of the variable Description of the variable Customs Days to clear customs to import Average number of days to clear customs when importing (logs) clearance Days to clear customs to export Average number of days to clear customs when exporting directly (logs) Wait for an import license Number of days waiting for an import license (logs) Energy/ Dummy for own power infrastructure Dummy taking value 1 if the firm provides its own power infrastructure, excluding generators Electricity Dummy for own generator Dummy variable taking value 1 if the firm has its own power generator Electricity from own generator Percentage of the electricity used by the plant provided by the own generator Cost of electricity from generator Estimated annual cost of generator fuel as percentage of annual sales Cost of electricity from public grid Average cost per kilowatt-hour (Kw/H) when using power from the public grid (logs) Dummy for equipment damaged by Dummy taking value 1 if any machine or equipment was damaged by power fluctuations / Value of the power fluctuations / Equipment losses of machinery and equipment damaged by power fluctuations as a percentage of the net book damaged by power fluctuations value of machinery and equipment (NBVC) Power outages / Average duration of Total number of (logs) / Average duration of (logs) / Percentage of sales loss due to power outages power outages / Sales lost due to same suffered by the plant in the last fiscal year (LFY) (conditional on the plant reports having power outages) Power fluctuations / Average duration of Total number of (logs) / Average duration of (logs) power fluctuations suffered in hours (conditional on power fluctuations the plant reports having power fluctuations) Wait for electric supply Number of days waiting to obtain an electricity supply (logs) Water Water outages / Average duration of Total number of (logs) / Average duration of (logs) / Percentage of sales lost due to water outages water outages /Losses due to same suffered by the plant in LFY (conditional on the plant reports having water outages) Dummy for own well or water Dummy taking value 1 if the plant has its own or shared borehole or well or builds its own water infrastructure infrastructure Water from own well or water Percentage of firm's water supply from its own or shared well infrastructure Cost of water from own well Total annual cost of self-provided water as a percentage of total annual sales Cost of water from public system Unit cost of using water from the public water system (logs) Wait for a water supply Number of days waiting for a water supply (logs) Telecom. Phone outages / Average duration of Total number of (logs) / Average duration of (logs) / Percentage of sales lost due to phone outages and ICT phone outages / Losses due to same suffered by the plant in LFY (conditional on the plant reports having phone outages) Wait for phone connection Number of days waiting to obtain a phone connection (logs) Dummy for e-mail Dummy variable taking value 1 if the plant mainly uses e-mail to communicate with clients and suppliers Dummy for web page Dummy variable taking value 1 if the plant uses its own Web page to communicate with clients and suppliers Transport Transport failures / Average duration of Total number (logs) of / Average duration of (logs )/ Percentage of sales lost due to transport failures transport failures / Sales lost due to suffered by the plant in LFY (conditional on the plant reporting on transport failures) same Dummy for own roads Dummy taking value 1 if the firm provides its own roads Dummy for own transportation for Dummy taking value 1 if the firm provides its own transportation for workers workers Dummy for contract with transportation Dummy taking value 1 if the firm arranges transport services for the delivery of finished products or raw company materials by directly contracting with the transportation company Dummy for own transportation Dummy taking value 1 if the firm arranges transport services for the delivery of finished products or raw materials with its own transportation Products with own transport Percentage of products delivered with firm's own transport Transport delay Percentage of times that transport services are late in picking up sales for domestic (or international) markets at the plant for delivery Shipment losses Percentage of the consignment value of the products shipped for domestic (or international) transportation lost while in transit because of theft, breakage, or spoilage Sales lost due to delivery delays Percentage of domestic (or international) sales lost due to delivery delays from suppliers in LFY Low quality supplies Percentage of domestic inputs/supplies that are of lower than agreed-upon quality Source: ICS data. 20 4. Evaluating the impact of infrastructure on total factor productivity (TFP) Escribano and Guasch (2005, 2008) relate infrastructure and other IC and C variables with firm-level productivity (TFP) according to the following observable fixed-effects system of equations: log Yit = L log Lit + M log M it + K log K it + log TFPit (4.1a) ai + D D log TFPj ,it = i + P + wit (4.1b) ai = INF INFi + IC ICi + i (4.1c) where, Y is firms' output (sales), L is employment, M denotes intermediate materials, K is the capital stock, INF is a time-fixed vector of observable infrastructure variables, IC is a time-fixed effect vector of other investment climate and other control variables, and D is a vector of industry and year dummies. The usually unobserved time fixed effects ( ai ) of the TFP equation (4.1b) is here proxy by the set of observed time-fixed components INF and IC variables of (4.1c) and a remaining unobserved random effect ( i ). The two random error terms of the system, i and wit , are assumed to be conditionally uncorrelated with the explanatory L, M, K, INF and IC variables 10 of equation (4.2): log Yit L log Lit + M log M it + K log K it + INF INFi + IC ICi + D Di + P + uit . = (4.2) Therefore, the regression equation (4.2) is representing the conditional expectation plus a composite RE error term equal to uit = i + wit . Providing reliable and robust estimates of the impact of infrastructure on productivity is not a straightforward task. First, because the functional form of the PF is not observed and there is no available single salient TFP measure. Second, there is an identification issue separating TFP from PF. When any PF inputs are influenced by unobserved common causes affecting productivity--such as a firm's fixed effects--there is a simultaneous equation problem in equation (4.1a). Third, we could expect that several IC variables have at least some degree of endogeneity, questioning therefore the conditional lack of correlation of (4.2). In what follows of this section, we briefly review the solutions to these questions suggested in Escribano and Guasch (2005, 2008) and Escribano et al. (2008). Estimating infrastructure's impact on productivity TFP or multifactor productivity refers to the effects of any variable different from the inputs--labor (L), intermediate materials (M), and capital (K)--affecting the production (Y) process. Since there is no 10 Under this formulation (and other standard conditions) the OLS estimator of the productivity equation (4.2) with robust standard errors is consistent, although a more efficient estimator (GLS) is given by the random effects (RE) estimator that takes into consideration the particular covariance structure of the error term, i + wit , which introduces a certain type of heteroskedasticity in the regression errors of (4.2). 21 single salient measure of productivity (or logTFPi), any empirical evaluation of the productivity impact of INFs might critically depend on the particular productivity measure used. Escribano and Guasch (2005, 2008) suggested--following the literature on sensitivity analysis of Magnus and Vasnew (2006)--to look for empirical results (elasticities) that are robust across several productivity measures. This is also the approach we follow in this paper. In particular, we want the elasticities of INFs on productivity (TFP) to be robust (with equal signs and similar magnitudes) for the 10 different productivity measures used. Alternative productivity measures come from considering: · Different functional forms of the PFs (Cobb-Douglas and Translog) · Different sets of assumptions (technology and market conditions) to get consistent estimators based on Solow's residuals, ordinary least squares (OLS), or random effects (RE), and so on · Different levels of aggregation in measuring input-output elasticities (at the industry level or at the aggregate country level) Box 4.1 Summary of productivity measures and estimated investment climate (IC) elasticities 1. Solow´s Residual Two-step 1.1 Restricted coefficient 1.1.a OLS 1.1.b RE 2 (Pit) measures estimation 1.2 Unrestricted coefficient 1.2.a OLS 1.2.b RE 4 (IC) elasticities 2. Cobb-Douglas Single-step 2.1 Restricted coefficient 2.1.a OLS 2.1.b RE 4 (Pit) measures estimation 2.2 Unrestricted coefficient 2.2.a OLS 2.2.b RE 4 (IC) elasticities 3. Translog Single-step 3.1 Restricted coefficient 3.1.a OLS 3.1.b RE 4 (Pit) measures estimation 3.2 Unrestricted coefficient 3.2.a OLS 3.2.b RE 4 (IC) elasticities 10 (Pit) measures Total 12 (IC) elasticities Source: Author's estimations. Note: Restricted coefficient = equal input-output elasticities in all industries. Unrestricted coefficient = different input-output elasticities by industry. Box 4.1 summarizes the productivity measures used for the IC evaluation. The two-step estimation starts from the nonparametric approach based on cost shares from Hall (1990) to obtain Solow's residuals in logs under two different assumptions: 11 (a) the cost shares are constant for all plants located in the same country (restricted Solow residual), and (b) the cost shares vary among industries in the same country (unrestricted by industry Solow residual). Once we have estimated the Solow residuals (logTFPi) in the first step, in the second step we can estimate equation (4.3) by OLS with robust standard errors for the countries that have a single year of data (2006) on PF variables. For the remaining 14 countries and for the blocks of countries described in section 3, we can also estimate (4.3) by RE to obtain the corresponding IC elasticities and semi-elasticities, INF INFi + log TFPit = IC Ci + D Di + P + i + wit (4.3a) 11 The advantage of the Solow residuals is that they require neither the inputs (L, M, K) to be exogenous nor the input-output elasticities to be constant or homogeneous (Escribano and Guasch, 2005 and 2008). The drawback is that they require having constant returns to scale (CRS) and, at least, competitive input markets. 22 where INF is the observable fixed effects vector of infrastructure variables, and IC is the observable fixed effects vector of other IC and control variables listed in table A.2 in the Appendix. In all the panel data regressions, we always control for several sector-industry dummies (Dj, j = 1, 2, . . ., qD), and in the cases having more than one year of observations we also include a set of time (Dt, t = 1, 2, . . ., qT) dummy variables and always a constant term (P). For cross-country comparisons based on TFP we use the concept of demeaned TFP, 12 which gets rid of the constant term as well as the constant effects by industry and by year, concentrating therefore on the part of TFP that is influenced by INF, IC, and the other plant-level control variables, = Demeaned log TFPit INF INFi + IC ICi . (4.3b) In the single-step estimation approach, we consider the parametric estimation by OLS and RE of the extended PF (4.2). To address the well-known problem of the endogeneity of inputs, we follow the approach proposed by Escribano and Guasch (2005, 2008). That is, we proxy the usually unobserved firm-specific fixed effects (which are the main cause of inputs' endogeneity) by a long list of observed firm-specific fixed effects coming from the ICSs. Controlling for the largest set of IC variables and plant characteristics, we can get--under standard regularity conditions--consistent and unbiased least squares estimators of the parameters of the PF and the INF and IC elasticities. Furthermore, we use two different functional forms of the PF--Cobb-Douglas and Translog--under two different assumptions on the input-output elasticities: equal input-output elasticities in all industries (restricted case) and different input-output elasticities by industries (unrestricted case). Notice that even if we are only interested in assessing the impact of infrastructure on TFP, we do not limit the scope of the control analysis to only this subset of IC variables. We include (and therefore control for) all the IC factors because of the crucial role IC variables play as proxies for the unobserved fixed effects; this is the key feature of this methodology in order to provide robust empirical results. If we tried to estimate only the impact of infrastructure, without controlling for the other blocks of IC variables, we might get different signs on certain coefficients because of the omitted variables problem (Escribano and Guasch, 2008). Another econometric problem we have to face when estimating the parameters of the INF and IC variables--either from the two-step or single-step procedure--is the possible endogeneity of some of these variables. That is, many INF or IC variables are likely to be determined simultaneously along with any TFP measure. With these productivity equations, the traditional instrumental variable (IV) approach is difficult to implement, given that we only have information for one year, and therefore we cannot use the natural instruments for inputs, such as those provided by their own lags. As an alternative correction for the endogeneity of the INF and IC variables, we use the region-industry-size 12 Notice that the demeaned TFP concept of equation (4.3b) corresponds to the observable part of the fixed effects equation (4.1c). 23 average of firm-plant-level INF and IC variables instead of the crude variables, 13 which is a common solution in panel data studies at the firm level 14 (see Veeramani and Goldar, 2004, for other use of industry-region averages with IC variables). Using industry-region-size averages also mitigates the effect of having certain missing individual INF and IC observations at the plant level, which--as mentioned in section 3--represent one of the most important difficulties of using ICSs. As an alternative, we also follow a second strategy to deal with the missing values of some INF and IC variables. In order to keep as many observations in the regressions as possible to avoid losing efficiency, when the response rate of the variables is large enough, we decided to replace those missing observations with the corresponding industry-region-size average. 15 Thus, we gain observations, efficiency, and representativity at the cost of introducing measurement errors into some variables. 16 The econometric methodology applied for the selection of the variables (INF and IC) goes from the general to the specific. The otherwise omitted variables that we encounter--starting from a too- simple model--generate biased and inconsistent parameter estimates. We start the selection of variables with a wide set compounded by up to 90 variables (depending on the country). We avoid simultaneously using time variables that provide the same information and are likely to be correlated, mitigating the problem of multicollinearity that could otherwise arise. We then start removing the less significant variables from the regressions one by one, until we obtain the final set of variables, all significant in at least one of the regressions and with parameters varying within a reasonable range of values. Once we have selected a preliminary model we test for omitted INF and IC variables. The robust coefficients of the INF and IC variables in productivity, along with their level of significance, are available upon from request. The parameters estimated in the two step procedure with restricted input-output elasticites can be found in figures 6.1 to 6.23. Infrastructure assessment based on O&P decompositions According to the O&P (1996) decomposition, aggregate productivity for a given country, industry, or region may be decomposed into two terms: (a) average productivity, and (b) a covariance term measuring whether the economy is able to efficiently reallocate resources from less productive establishments to more productive ones. Once we have estimated a robust set of parameters for the IC 13 Because of the low number of available regions in most of the countries, we had to use the industry-region-size variables instead of the region-industry averages. For the creation of cells a minimum number of firms are imposed--there must be at least 15 to 20 firms in each industry-region-size cell to create the average, otherwise we apply the region-industry averages. If the problem persists, we apply the industry-size or the region-size average. 14 This two-step estimation approach is a simplified version of an instrumental variable (IV) estimator (two-stage least squares, 2SLS). 15 Notice that this replacement strategy has a straightforward weighted least squares interpretation since we are giving a greater weight to those observations with more variance (Escribano et al., 2008). 16 The measurement error introduces a downward bias in the parameters that depends on the ratio between the variances of the variables and the measurement error. Since those explanatory variables are constant within regions, sizes, and industries we expect their variances will be small. 24 factors with statistically significant impacts on firms' productivity, we exploit the exact relation, proposed by Escribano et al. (2008a), between the terms of the O&P decomposition, and the IC factors affecting productivity. The IC infrastructure variables affect both the average productivity of African establishments (or firms) as well as their allocative efficiency component. It is well known that competitive markets efficiently allocate resources under certain conditions. But in a world of imperfect information a turbulent IC introduces distortions into markets, and, as a result, affects the efficiency of the economy as a whole. The allocative efficiency term of the O&P decomposition should therefore reflect those imperfections. In the second part of this analysis--taking advantage of the robustness of the INF, IC, and C elasticities estimated--our aim is to concentrate on the TFP measure that comes from the restricted Solow's residuals in order to evaluate the infrastructure effects on average productivity and on allocative efficiency based on the O&P decomposition of aggregate productivity in levels, = TFPq TFP q + N q cov( sq,it , TFPq ,it ) . ^ Y (4.4a) Furthermore, we want to exploit the log-linear properties of the following mixed 17 O&P decomposition for each of the African countries considered in order to obtain closed form O&P decompositions in terms of IC and C variables: = log TFPq log TFP q + N q cov( sq,it , log TFPq ,it ) . ^ Y (4.4b) Aggregate log productivity of country q, say (logTFPq), is equal to the sum of the sample average log productivity of the establishments of country q, and the covariance between the share of sales (sYq) and log productivity of that country (allocative efficiency of country q). The index q could also indicate a particular industry, region, size, and so on. The useful additive property of equation (4.2) in logarithms allows us to obtain an exact closed form solution of the decomposition of aggregate log productivity according to equation (4.5). Following Escribano et al. (2008), we can express aggregate log productivity as a weighted sum of the average values of the IC, dummy D variables, the intercept, ö and the productivity average residuals ( u ) from (4.2), and the sum of the covariances between the share of sales and investment climate variables IC, dummies D, and the productivity residuals ( u ):ö = INF INF q IC IC q + D Dq + p + uq ,it + N q INF cov( sq,it , INFq ,i ) log TFPq ^ ^ ^ ^ ^ ^ ^ Y (4.5) + N q IC cov( sq,it , ICq ,i ) + N q Ds cov( sq,it , Di ) + N q cov( sq,it , uq ,it ) ^ ^ Y ^ ^ Y ^ Y ^ where the set of estimated parameters used comes from the two-step TFP estimation, having the restricted Solow's residual as a dependent variable in (4.2). From equation (4.5) each INF and IC variable may affect the aggregate log productivity through both its average and covariance (with respect to the share of sales). This complements the information provided by the marginal effects (INF and IC elasticities). Suppose that an INF variable with a low 17 It is called a mixed Olley and Pakes (O&P) decomposition because in the original O&P decomposition both TFP and the share of sales were in levels, while TFP in (4.4b) is in logs (log P). 25 impact--in terms of marginal effects (elasticities)--affects most of the firms in a given country; the impact of such an IC variable in terms of average productivity could be very high. It is therefore very important for policy analysis to combine the empirical evidence from the estimated IC elasticities on productivity with their corresponding INF impact on the two components of O&P decompositions: average productivity and allocative efficiency. A variable with a negative marginal effect on average productivity (or logTFP) may have either a positive or a negative effect on allocative efficiency. If the covariance of that IC variable and the market share is positive, then the greater proportion of sales in the hands of establishments with high levels of that variable, the larger the negative impact on aggregate productivity will be, therefore decreasing the allocative efficiency. In contrast, a negative covariance means that those establishments with the highest levels of the IC variable have the lowest market shares, and therefore the negative effect of the IC variable on average productivity is somehow compensated through the effect on the reallocation of resources among firms. By operating in (4.5) Escribano et al. (2008) obtained the next expression, which allows us to obtain a direct decomposition of the impacts of each INF and IC variable on aggregate productivity (logTFPq): 100 =100 [ IC INF q + IC ICq + D Di + p + ui + N q INF cov( sq,i , INFq ,i ) ^ ^ ^ ^ ^ ^ ^ Y log TFPq (4.6) + N q IC cov( s , ICq ,i ) + N q D cov( s , Di ) + N q cov( s , uq ,i )]. ^ ^ Y q,i ^ ^ Y q,i ^ Y ^ q,i There are several advantages of using equation (4.6). First, we can compare net contributions by isolating the impact of INF and other IC variables from the impact of industry dummies, the intercept, and the residuals. Second, we can express what portion of aggregate productivity is explained by INF, IC, and C variables (demeaned log TFP), and what proportion is due to the constant term, industry dummies, and so on. To make cross-country comparisons based on IC impacts on TFP and to avoid the problem of comparing apples and oranges, it is desirable to create an index (demeaned TFP). After subtracting the mean (that is, the constant term, time effects, industry effects, and country-specific effects) from aggregate productivity we can concentrate on the contributions of IC variables to the demeaned TFP. Similarly, we can construct the demeaned counterparts of expressions (4.5) and (4.6) and compute the percentage contribution of each INF variable or block of IC variables--as in equations (4.7) and (4.8), respectively--obtaining the following demeaned mixed O&P decomposition: = INF INF q IC ICq + N q INF cov( sq,it , INFq ,i ) + N q ´IC cov( sq,it , ICq ,i ) Demean log TFPq ^ ^ ^ ^ Y ^ ^ Y (4.7) 100 100 = [ ´IC INF q + ´IC ICq + N q INF cov( sq,i , INFq ,i ) + N q IC cov( sq,i , ICq ,i )]. ^ ^ ^ ^ Y ^ ^ Y (4.8) Demean log TFPq So far, we have exploited the linear properties of the logarithm form of the mixed O&P decomposition of TFP. But the original O&P decomposition is based on TFP and the share of sales (in 26 levels), and is therefore also capturing nonlinear relations between market shares and IC variables coming from (4.3a). To know to what extent these nonlinear terms are affecting this relation, we perform simulation experiments 18 on INF, IC, and C variables, and evaluate the consistency of the results with the ones obtained from the previous mixed O&P decomposition--see (4.4b). The IC simulations are done variable by variable (one at a time) keeping the rest of the variables constant; that is, we propose a scenario in which the level of one of the IC variables improves by 20 percent in all establishments (20 percent less power outages, 20 percent less shipment losses, etc). We compute the corresponding rate of change of aggregate productivity, average productivity, and allocative efficiency caused by such a 20 percent improvement. We repeat the same experiment for the rest of the IC and C variables, and, for comparative purposes, we also evaluate the relative group of IC variables. 5. The contribution of infrastructure to the investment climate (IC) of Africa In section 4 we described the econometric methodology used in section 6 to assess the impact of infrastructure on productivity. We suggested three key elements of empirical evaluations of infrastructure and other IC and C impacts on productivity: the marginal productivity effects, the percentage contributions of infrastructure to aggregate log productivity (mean and efficient components), and the simulations of infrastructure improvements on aggregate productivity (in levels). In this section we focus on presenting the results of infrastructure contributions to aggregate productivity from simulation experiments. In addition to the results of the econometric analysis, we consider African firms' perceptions of the main obstacles to economic performance. In the first subsection the objective is to assess how African firms perceive infrastructure quality. This is followed in the next subsections by the results of the econometric analysis, focusing on the infrastructure's effect on productivity after controlling for other IC factors. Finally, to complement the robustness of the results we check the consistency of the conclusions obtained from both the IC contributions to average log TFP and from the TFP simulation experiments. Do African firms perceive infrastructure as an obstacle to growth? In ICSs, firms are asked to rate a number of IC factors as obstacles to economic performance. The survey options offered are no obstacle, minor obstacle, moderate obstacle, major obstacle, or very severe obstacle on a broad range of IC aspects: infrastructure, red tape, corruption and crime, finance, and labor skills. Figure 5.1 shows the degree to which each group of IC factors is perceived by firms as an obstacle to economic development. These perceptions are sorted in descending order by their perceived contribution to the total, after being normalized to 100. For example, in panel A of figure 5.1, we observe that in Cape Verde 25 percent of firms believe infrastructure to be a major or very severe obstacle to economic growth; 40 percent find red tape, corruption, and crime as a major or very severe 18 We are indebted to Ariel Pakes for this suggestion. 27 obstacle; 23 percent finance; and 10 percent a lack of labor skills. The countries in which infrastructure is perceived as an especially great obstacle to growth are--in descending order--Cape Verde, Burundi, Burkina Faso, Mauritania, Cameroon, Ethiopia, and Niger. Countries where a relatively low number of firms perceive infrastructure as a major constraint are Egypt, Mauritius, Morocco, South Africa, and Botswana. Figure 5.1 Rankings of firms' perceptions of severe and very severe obstacles to growth Source: Authors' calculations from IC data. Note: * = Number of questions on perceptions by blocks of IC variables: Infrastructure, 4 questions; red tape, corruption, and crime, 9 questions; finance and corporate governance, 2 questions; labor skills, 2 questions. 28 The factor group with the largest number of sub-factors is red tape, corruption, and crime. So it is not surprising that this has the largest percentage of all. In order to isolate this effect, we normalize after computing the mean of each IC factor type. For example, in the infrastructure group there are four IC sub-factors (telecommunications, customs, electricity, and transportation), whereas the red tape group contains eleven sub-factors. But the importance of infrastructure is very similar across countries, although obviously the relative contribution of the red-tape group is more balanced, gaining relatively more relevance to the other IC groups. In the next subsection the results of the econometric analysis--estimating the relative impact of infrastructure on average productivity using simulations--are compared with firms' perceptions of obstacles to growth. The question of interest is: are the econometric results consistent with firm perceptions? Impact of infrastructure on productivity The impact of infrastructure factors on productivity is evaluated here in terms of their effect on the O&P decomposition. Infrastructure's contribution to the aggregate productivity of each country's manufacturing sector is decomposed into its contribution to (a) average productivity and (b) allocative efficiency (the ability of markets to reallocate resources from less productive to more productive establishments). Figure 5.2 presents the two alternative country-by-country O&P decompositions given by equations (4.4a) and (4.4b), sorted by aggregate productivity in descending order. The productivity measure used to calculate the O&P decomposition is the restricted Solow residual obtained from the two-step estimation approach (see section 4). We present two sets of results with O&P decompositions. Panel A of figure 5.2 shows the O&P decomposition with the restricted Solow residual in levels, and panel B shows the mixed O&P decomposition with the restricted Solow residual, in logs, weighted by the share of sales. This is important because the results of the simulations are associated with the O&P decomposition in levels, and the results from the percentage contributions to the average use the convenient additive property of the TFP equation in logs. Notice that both panels A and B preserve the rankings of average productivity, but this is not necessary true for aggregate productivity. The reason is clear: aggregate productivity is simply the sum of average productivity and allocative efficiency, and this second term depends somewhat on whether we use TFP in levels or in log form. 19 From panel A of figure 5.2 we observe a positive reallocation of output. That is, output is moving from less productive establishments to more productive ones, since in all the countries the allocative efficiency is positive, with the greatest effects found in Benin, Burkina Faso, Cameroon, Niger, and Eritrea. 19 For a deeper discussion of this issue see Escribano et al. (2008). 29 Figure 5.2 Olley and Pakes (O&P) decompositions of total factor productivity (TFP) A. Olley and Pakes decomposition of TFP B. Mixed Olley and Pakes decomposition of TFP Source: Authors' calculations from IC data. Notes: The Olley and Pakes (O&P) decomposition of TFP in levels is obtained from equation 4.4a of section 4. The mixed O&P decomposition is obtained from equation 4.4b. Sales in levels are used to compute the share of sales in both O&P decompositions. 30 Figure 5.3 Demeaned O&P decompositions of TFP A. Demean Olley and Pakes decomposition of TFP B. Mixed Demean Olley and Pakes decomposition of TFP Source: Authors' calculations from IC data. Notes: The demeaned Olley and Pakes (O&P) decomposition of TFP in levels is given by equation 4.7. It is derived from equation 4.4a, using as the productivity measure the demeaned counterpart of the restricted Solow residual (see equation 4.3b) in levels. The demeaned mixed O&P decomposition comes from equation 4.4b, with the demeaned log-TFP of equation 4.3b in logs. Sales in levels are used to compute the share of sales in both O&P decompositions. 31 Notice that we avoid direct comparisons of TFP across countries but we suggested in section 4 to compare demeaned productivity decompositions (see figure 5.2, panels A and B, ). Remember that the demeaned productivity (or demeaned TFP) at the firm level is simply firm-level productivity minus the constant term of the productivity equation and the industry and year dummies (see equation 4.7).20 We assume that all the productivity differences resulting from units of measurement, different deflators, and so on are contained in the constant, industry, and time-fixed effects, and therefore what is left in the productivity measure are only the TFP effects of the infrastructure, IC, and other C variables. We obtain the O&P decomposition using the demeaned productivity either in levels or in logs. This demeaned TFP set of cross-country comparisons is presented in figure 5.3 . Panel A shows the decomposition of the demeaned productivity in levels; it is interpreted as the productivity that stems from IC conditions after controlling for all the other elements. The results are not at all surprising since they are basically consistent with those provided by the per capita income and by the DBR (2007). Rankings based on demeaned productivity are topped by South Africa and Mauritius, closely followed by Botswana, Algeria, Egypt, Namibia, and Swaziland. The lowest-ranked countries are those with the most antiproductive IC, in other words, those whose IC conditions pose difficulties to economic development. These countries are Tanzania, Malawi, Uganda, Benin, Mauritania, and Zambia. Symmetrically, as for the regular O&P decompositions, the contribution of the IC to aggregate demeaned productivity is decomposed into its contributions to average demeaned productivity and the allocative demeaned efficiency term (see equation 4.7). Notice that, in Africa, the allocative efficiency component is always lower than the effect of average productivity. Nevertheless, in Madagascar, Botswana, Mauritius, and other countries, the IC has a considerable effect on the efficient reallocation of resources among establishments. Alternatively, this demeaned productivity may be interpreted as a sum of pro- and anti- productive infrastructure, as well as other IC and C factors. Examples of pro-productive infrastructure factors are the use of e-mail and websites. Negative or anti-productive infrastructure factors include the number of power outages, the average duration of water outages, and so on. As a consequence, productivity will decrease as the importance of anti-productive factors becomes larger and larger; this picture becomes even clearer in panel B of figure 5.4 . The demeaned O&P decomposition of TFP in logs (see panel B) shows how aggregate productivity may be negative (in Tanzania, Benin, Malawi, and so on) when the negative TFP aspects of IC dominate over the positive (pro-productive IC factors weigh more than the negative ones), as in the case of South Africa, Mauritius, Egypt, Botswana, and so on. In sum, in African countries the IC has important effects on the aggregate productivity of the manufacturing industry, and this net effect may be positive or negative depending on which IC aspect matters more--the pro-productive or the anti-productive. The aim now is to know to what extent such decreases or increases in productivity are due to infrastructure or other IC groups. 20 Obviously, the year dummies are only subtracted from the productivity measure of the countries with panel data. 32 Figure 5.4 Demeaned productivity by groups of IC variables: simulations and average contributions A. Percentage productivity gains from a 20% improvement in the investment climate conditions B. Productivity gains and losses from the average investment climate conditions Source: Authors' calculations from IC data. Note: The simulations are done variable by variable. The total percentage productivity gain from each group of variables (infrastructure; red tape, etc.) is computed as the sum of the individual productivity gains caused by the improvement in the IC variables of that group (one by one). Therefore, the final productivity gain should be interpreted in ceteris paribus terms: how much does productivity increase when the corresponding variable improves by 20 percent, holding everything else constant? The productivity gains and losses from the average investment climate come from the decomposition of the demeaned Olley & Pakes decomposition in logs by groups of variables (4.8). The productivity gain or loss from the infrastructure group for each country is computed as the sum of the percentage contributions to average log-TFP caused by the average individual infrastructure variables. 33 The same holds for the rest of the groups of IC and C variables. Figure 5.4, panel A, provides the decomposition of demeaned productivity in levels using simulations of improvements to IC variables. When the IC factor improves by 20 percent it could mean, for example, that 20 percent more firms are using e-mail, or that there is a 20 percent reduction in power outages, and so on--which implies decreases in the negative IC factors and increases in the positive ones. The total effect of improving each IC by 20 percent, maintaining the rest of IC factors constant, implies that aggregate productivity could increase in South Africa by 55 percent, in Mauritius by 30 percent, and so on. From panel A of figure 5.4 it is clear that there are some economies that are more likely to be affected by the IC. These are therefore more sensitive to changes in IC conditions. This is the case in Kenya and Benin, where the aggregate productivity could increase by 70 percent and 85 percent, respectively. At the other extreme are Egypt, Morocco, and Eritrea. Lastly, improvements to aggregate productivity come in almost all countries via improvements to average productivity, and, to a lesser extent, allocative efficiency, with the exception of Algeria, Kenya, and Benin. The role of infrastructure in the composition of aggregate productivity is considerable in all the countries, but is the greatest in Uganda, Benin, Malawi, Cameroon, and Zambia. This suggests that these countries are the most sensitive to changes in infrastructure quality. Panel B of figure 5.4 shows a more static interpretation using the O&P decomposition in logs by group of variable. In particular, panel B offers information on the actual and current situation of the IC and its effect on aggregate productivity; in other words, gains and losses generated by the average IC conditions (O&P decomposition of TFP in logs decomposed by groups of INF, IC, and C variables). For example, in South Africa, aggregate demeaned TFP is 0.83 (see panel B of figure 5.3). Out of this level of productivity -0.9 is explained by the overall contribution of the infrastructure factors; 0.95 by red tape, corruption, and crime; and the rest by the remaining IC and C variables. Notice that in panel B the contributions of the different groups are not in absolute value, so the positive effect of the pro- productive factors compensate for the negative effect of the anti-productive ones. Even taking this into account, the overall contributions of the infrastructure group are negative in all the countries, implying that the pro-productive infrastructure IC factors never compensate for the negative IC effects, with the exception of Kenya (where it is slightly positive) and Madagascar, Ethiopia, and Algeria (where the contribution of the infrastructure group is close to zero and almost negligible). As expected, the largest and most negative infrastructure effect is found in Benin, followed by Malawi, Uganda, Mauritania, Cameroon, and Zambia. 34 Figure 5.5 Simulation of infrastructure absolute effects on productivity (20 percent improvement) A. Aggregate productivity 10 0 % 80% 60% 40% 20% 0% MWI ERI UGA MLI SEN NAM CMR ETH TZA NER SWZ MRT ZMB BEN MUS BWA DZA MAR EGY BFA MDG ZAF KEN In f r as tr uc tu re s Re d tap e , c or ru ption a nd c rime Fina n c e a nd c or po ra te go v er na nc e Qua lity , inn ov a tio n an d lab o r s kils Othe r c o ntr ol v a riab le s B. Average productivity 100% 80% 60% 40% 20% 0% ERI MWI SEN UGA MLI DZ A BEN ETH TZA NER CMR MRT BFA SWZ KEN BWA NA M MUS EGY ZA F MA R ZMB MDG Inf ras truc tures Red tape, c orr uption and c rime Financ e and c or porate gov ernanc e Quality , innov ation and labor s kils Other c ontrol v ar iables C. Allocative efficiency 100% 80% 60% 40% 20% 0% MWI SEN UGA NA M MLI TZ A DZA ERI ETH BEN CMR KEN NER SWZ MRT EGY Z MB MA R MDG MUS BFA ZA F BWA Inf ras tr uc tur es Red tape, c or ruption and c rime Financ e and c orpor ate gov er nanc e Quality , innov ation and labor s kils Other c ontrol v ariables Source: Authors' calculations from IC data. Note: The percentage contribution of the infrastructure group is computed as the sum of the absolute values of the percentage contributions of the individual infrastructure variables, divided by the cumulative sum in absolute terms of the percentage contributions of all the IC and C variables, including infrastructure. The holds for the rest of the IC blocks of variables. Continuing with the same idea in figure 5.5, we are interested in obtaining the weight of the infrastructure group relative to the IC as a whole. Thus, by normalizing to 100 the contribution of the 35 IC to aggregate productivity, average productivity, and allocative efficiency, we find via simulations that the relative 20 percent improvement of infrastructure in Malawi reaches 58 percent, in Eritrea 50 percent, and in Uganda 45 percent (as panel A of figure 5.5 shows). The same holds for average productivity (panel B) where the rankings do not change, and for allocative efficiency (panel C), where, once again, Malawi, Senegal, and Uganda show the largest contributions of the infrastructure group. Figure 5.6 Infrastructure absolute effects on productivity: Mixed demeaned O&P decomposition A. Aggregate productivity 100% 80% 60% 40% 20% 0% MWI UGA ZMB ETH BEN SEN DZA MLI CMR ERI TZA NER MAR MRT MDG EGY ZAF BFA MUS NAM KEN SWZ BWA Inf ras truc tur es Red tape, c orr uption and c r ime Financ e and c orpor ate gov ernanc e Quality , innov ation and labor s kils Other c ontrol v ariables B. Average productivity 100% 80% 60% 40% 20% 0% MWI BEN SEN UGA ETH ZMB DZ A ERI MLI CMR MRT BFA NER TZ A MA R MDG KEN ZAF MUS EGY SWZ BWA NA M Inf ras truc tures Red tape, c orr uption and c rime Financ e and c or porate gov ernanc e Quality , innov ation and labor s kils Other c ontrol v ar iables C. Allocative efficiency 100% 80% 60% 40% 20% 0% MWI SEN NA M DZ A UGA TZ A MDG BFA MLI ZMB CMR EGY ERI ETH MRT KEN ZAF MUS MA R SWZ BEN NER BWA Inf ras tr uc tur es Red tape, c or ruption and c rime Financ e and c orpor ate gov er nanc e Quality , innov ation and labor s kils Other c ontrol v ariables Source: Authors' calculations from IC data. Note: The percentage contribution of the infrastructure group is computed as the sum of the absolute values of the percentage contributions of the individual infrastructure variables, divided by the cumulative sum in absolute terms of the percentage contributions of all the IC and C variables, including infrastructure. The holds for the rest of the IC blocks of variables. A similar picture is provided by figure 5.6, where, instead of simulations, we consider the relative contributions by groups of variables to average demeaned log productivity and to the demeaned 36 efficiency term--see equation (4.8). In this case we sum up the different contributions of the INF, IC, and C factors of equation (4.8), but in absolute value so that the negative effects do not compensate for the positive ones and vice versa, and we compute the relative contribution of each group within the IC group as a whole. Therefore, the relative contribution of the infrastructure group is the sum in absolute value of all individual infrastructure variables divided by the total absolute contribution of all INF and IC variables--multiplied by 100. The largest relative effects of infrastructure on aggregate log productivity are found in Malawi (60 percent), Uganda (50 percent), Benin (50 percent), Zambia (47 percent), and Ethiopia (46 percent). The lowest contributions are in Kenya, Swaziland, and Botswana. A similar ranking is provided by panel B, where the effects on the average log productivity are isolated from those from the allocative efficiency, as seen in panel C. Once again, Malawi, Benin, Senegal, Uganda, and Ethiopia lead a ranking closed by Mauritius, Egypt, Swaziland, Botswana, and Namibia. Panel C offers the results for allocative efficiency. In Malawi, Senegal, Namibia, and Algeria, the effects of infrastructure on the efficient reallocation of results among firms appear to be very significant, reaching the relative contributions of 54 percent, 48 percent, and 46 percent, respectively. 21 Cross-country comparisons Table C.1 summarizes the empirical results discussed in previous sections. The first column shows ranking of African countries based on per capita income, the second based on the DBR (2007), the third column based on quality of overall infrastructure (1 minimum, 7 maximum) given in the ACR (2007), the fourth column the demeaned aggregate productivity, and the fifth column shows the ranking of firms' perceptions of the quality of infrastructure (from 23rd being the poorest quality to 1st being the best quality in our sample). Columns 6 and 7 show the percentage of absolute contributions of infrastructure to average log productivity and to allocative efficiency, with TFP in logs, while columns 8 and 9 show the percentage absolute contributions of infrastructure to average productivity in levels and to allocative efficiency via simulations. 21 It is useful to clarify that the differences between the rankings of the contributions to the aggregate productivity via simulations of panel A of figure 5.5 and the rankings of the contributions to the aggregate log productivity of panel A of figure 5.6 come mainly from the role of the allocative efficiency. Notice that in figure 5.6 the allocative efficiency term based on log TFP does not have the same scale as the efficiency term when TFP is in levels. As a result, since aggregate productivity is simply the sum of the average productivity and the efficiency term in levels, the role of the efficiency term with TFP in levels will increase with respect to its counterpart in logs and therefore, could alter the rankings of countries based on the two alternative measures of aggregate productivity (weighted productivity). 37 Table C.1 Summary of cross-country comparisons based on alternative rankings of economic performance Ranking DBR 2007 Ranking Demeaned Firms' Perc. Contributions of infrastructure to Perc. contributions of infrastructure to based on (rank) ACR'07 (rank Aggregate perceptions: log-productivity productivity via simulations per capita within Productivity infras. As an GDP sample) (rank) obstacle % abs. Average log- Allocative efficiency Average Allocative efficiency contribution productivity in logs productivity (rank) % abs. contribution % abs. contribution % abs. contribution % abs. contribution (rank) (rank) (rank) (rank) MUS (1) 32 (2) 2.0 (2) 4.2 (1) 13.9 (2) 26.6 (19) 17.1 (18) 21.8 (6) 12.4 (4) SWZ (2) 76 (5) 1.4 (7) n.a 22.4 (10) 25.6 (21) 14.3 (20) 27.4 (10) 17.6 (10) ZAF (3) 29 (1) 2.3 (1) 4 (3) 16.2 (5) 28.6 (18) 19.7 (17) 17.4 (4) 11.0 (2) BWA (4) 48 (4) 1.7 (3) 3.4 (6) 15.6 (4) 17.5 (22) 7.41 (23) 23.2 (8) 8.8 (1) DZA (5) 116 (12) 1.5 (4) 2.9 (7) 18.3 (7) 48.6 (7) 31.1 (4) 34.9 (18) 26.4 (17) NAM (6) 42 (3) 1.5 (6) 4.2 (2) 18.3 (6) 16.5 (23) 32.9 (3) 22.7 (7) 36.7 (20) EGY (7) 165 (22) 1.5 (5) 3.7 (4) 14.0 (3) 26.0 (20) 23.8 (12) 19.9 (5) 16.1 (8) MAR (8) 115 (11) 1.1 (9) 3.6 (5) 9.9 (1) 31.3 (15) 16.6 (19) 16.2 (3) 14.8 (6) CMR (9) 152 (18) 0.8 (16) 1.9 (18) 27.5 (23) 41.6 (10) 25.4 (11) 31.2 (13) 23.2 (13) MRT (10) 148 (16) 0.6 (19) 2.1 (15) 25.3 (17) 35.4 (11) 21.1 (15) 28.3 (12) 16.2 (9) SEN (11) 146 (15) 0.9 (12) n.a 22.7 (11) 58.5 (3) 40.9 (2) 52.1 (21) 42.2 (22) BEN (12) 137 (13) 0.6 (20) 2.1 (11) 25.6 (18) 59.9 (2) 12.4 (21) 33.3 (17) 23.3 (14) KEN (13) 83 (6) 1.0 (11) 2.8 (8) 25.6 (19) 30.3 (17) 19.9 (16) 26.1 (9) 23.2 (12) MLI (14) 155 (19) 0.9 (14) 2.1 (14) 21.6 (9) 42.7 (9) 26.8 (9) 42.5 (19) 33.5 (19) UGA (15) 107 (9) 0.6 (21) 2 (17) 23.3 (12) 58.4 (4) 29.8 (5) 45.4 (20) 42.0 (21) BFA (16) 163 (21) 0.8 (15) 2.1 (12) 26.9 (22) 35.3 (12) 27.0 (8) 27.6 (11) 12.0 (3) ZMB (17) 102 (8) 0.7 (18) n.a 24.0 (14) 50.6 (6) 26.8 (10) 15.4 (2) 15.1 (7) TZA (18) 142 (14) 0.2 (23) 2.7 (9) 24.3 (15) 34.1 (14) 28.3 (6) 32.3 (15) 29.1 (18) NER (19) 160 (20) 0.8 (17) n.a 26.2 (20) 34.7 (13) 11.1 (22) 31.6 (14) 22.1 (11) MWI (20) 110 (10) 0.4 (22) 2.1 (13) 24.5 (16) 65.9 (1) 45.8 (1) 53.7 (22) 55.2 (23) MDG (21) 149 (17) 1.4 (8) 2 (16) 23.5 (13) 30.6 (16) 27.9 (7) 11.1 (1) 14.3 (5) ETH (22) 97 (7) 1.0 (10) 2.3 (10) 26.7 (21) 52.6 (5) 21.9 (14) 33.2 (16) 25.0 (15) ERI (23) 170 (23) 0.9 (13) n.a 20.7 (8) 46.1 (8) 22.5 (13) 54.7 (23) 25.3 (16) Source: Authors' calculations using ICA data, DBR (2007), ACR (2007), and Penn World Table. Note: n.a = not available. The rankings presented in the first five columns are very consistent. In particular, the ranking based on demeaned aggregate productivity (column 4) shows a clear positive correlation to per capita GDP and with the ranking based on the DBR rankings. From the results of the rankings obtained from the first five columns we find two groups of African countries, as was suggested by looking at the growth rates of per capita GDP (see figure 1.2, panel B). That is: (a) countries in the north and south of Africa are relatively more successful, and (b) countries from the central-east and central-west regions of Africa are relatively less successful. The last four columns show two alternative measures of the percentage absolute contribution of infrastructure to productivity, along with the ranking in parentheses. In particular, column 6 shows a negative correlation between the ranking based on the contribution of IC to average log TFP and per capita GDP and also with the ranking based on the DBR, indicating that low infrastructure quality is one of the key growth bottlenecks in Africa. The results show a great homogeneity among the rankings in the first four columns and the results from the econometric analysis. Thus, for instance, Mauritius is ranked 1st in terms of per capita income and quality of overall infrastructure in the ACR (2007), 2nd according to the DBR (2007) and firms' perceptions, 19th (out of 23) according to the impact of the INF variables on the average log productivity, and 18th (out of 23) with respect to the allocative efficiency in logs. In these cases the correlation with firm growth is negative, signaling again that infrastructure quality is an important growth bottleneck in Africa. 38 Egypt and Morocco are interesting cases. Both countries show a relatively high quality of infrastructure according to the ACR and perception rankings, and the results of the econometric analysis confirm this. Egypt is 4th and Morocco is 5th in the rankings based on demeaned aggregate productivity, and both have one of the lowest contributions of infrastructure to TFP. Countries with the poorest infrastructure quality, according to ACR (2007) and firms' perceptions, are Cameroon, Burkina Faso, Niger, and Ethiopia. Once again this is consistent with the econometric analysis done for these countries, showing a great negative influence of infrastructure on productivity. Cameroon and Burkina Faso are ranked among the countries with the highest contribution of infrastructure to average productivity. The influence of infrastructure on Ethiopia's manufacturing sector productivity is also very high, with a relative contribution equal to 52.6 percent of the total IC effect. The following set of figures provides some additional evidence on the relation between measures of countries' economic performance and TFP based on our econometric analysis. The previous conclusions become more apparent by looking at the cross-plots. Figure 5.7 shows a clear positive correlation between GDP per capita and demeaned aggregate productivity, with a correlation coefficient equal to 0.81. Notice that this positive relationship has a decreasing dispersion as per capita income grows; that is, for those African countries with a per capita GDP lower than 10 percent of that of the United States, demeaned aggregate productivity presents a more heterogeneous behavior. The conclusion obtained from figures 5.8 and 5.9 are similar. Those countries that are high in the ranking based on ease of doing business (DBR) also have a large demeaned aggregate productivity, with a correlation coefficient of 0.77. The same is true for the positive relationship between the quality of overall infrastructure and the ACR (2007)--the more productive the manufacturing firm is, the higher the contribution of overall infrastructure quality to TFP (correlation coefficient equal to 0.76). A question of interest is whether those countries with the lowest demeaned aggregate productivity levels are also those with the greatest impact of infrastructure on firm's perceptions, on average productivity, and on allocative efficiency. Figures 5.10 and 5.12 provide clear answers to these questions. Figure 5.10 shows the negative correlation between the mixed demeaned aggregate productivity and firm's perceptions of growth bottlenecks, with a correlation coefficient of 0.76. The absolute contribution of infrastructure to both average log productivity and average productivity via simulations decreases as the demeaned aggregate productivity increases. This relation is stronger in the case of the absolute percentage contribution to the average log productivity since the corresponding coefficients of correlations are -0.60 (figure 5.11) and -0.49 (figure 5.12). 39 Figure 5.7. Cross-plot between Figure 5.8. Cross-plot between Figure 5.9. Cross-plot between demeaned aggregate productivity demeaned aggregate productivity demeaned aggregate productivity and GDP per capita (% of US) and ranking on the ease of doing and quality of overall business* infrastructure from ACR 2007* 3.0 2.5 2.5 R² = 0.6 Demeaned Aggregate Productivity R² = 0.66 Demeaned Aggregate Productivity 2.5 Demeaned Aggregate Productivity 2.0 2.0 R2 = 0.58 2.0 1.5 1.5 1.5 1.0 1.0 1.0 0.5 0.5 0.5 0.0 0.0 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0 1 2 3 4 1.5 2.0 2.5 3.0 3.5 4.0 4.5 GDP per capita (relative to U.S) Ranking Ease of Doing Business 2007 Quality of overall Infrastructure ACR 2007 *Rank is computed as: (total number of firms in DBR-Rank)/ total number of firms in DBR Source: Authors´ calculations with IC data, Doing Business Report (2007) and Penn World Table. Figure 5.10. Cross-plot between Figure 5.11. Cross-plot between Figure 5.12. Cross-plot between demeaned aggregrate productivity demeaned aggregate productivity demeaned aggregate productivity and firms´ perceptions on and percentage absolute and percentage absolute infrastructure as an obstacle contribution of infrastructure to contribution of infrastructure to average log-productivity average productivity via simulations 2.5 2.5 2.5 Demeaned Aggregate Productivity Demeaned Aggregate Productivity R² = 0.47 Demeaned Aggregate Productivity 2.0 2.0 2.0 R2 = 0.25 1.5 1.5 1.5 1.0 1.0 1.0 R2 = 0.36 0.5 0.5 0.5 0.0 0.0 0.0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 8 13 18 23 28 33 Percentage Absolute Contribution of Infrastructure to Average Log- Percentage Absolute Contribution of Infrastructure to Average Firms' perceptions of infrastructure as an abstacle Productivity Productivity via simulations Source: Authors´ calculations with IC data. Figures 5.13 and 5.14 show the linear correlations between demeaned aggregate productivity and percentage absolute contribution to allocative efficiency TFP in logs and allocative efficiency via simulations, correspondingly. There is a negative relation in both figures. But the linear correlation is smaller in the case of the average log productivity (correlation coefficient equal to -0.31) than in the case of allocative efficiency with TFP in levels (correlation coefficient equal to -0.49). 40 Finally, figures 5.15 and 5.16 show the strong linear positive relation between the infrastructure contributions to the two components of the O&P decomposition based on TFP in logs and TFP in levels. Their corresponding coefficients of correlation are 0.69 and 0.77, respectively. Figure 5.13. Cross-plot between Figure 5.14. Cross-plot between demeaned aggregate productivity and demeaned aggregate productivity and percentage absolute contributions of percentage absolute contributions of infrastructure to allocative efficiency infrastructure to allocative efficiency via (TFP in logs) simulations 2.5 2.5 Demeaned Aggregate Productivity Demeaned Aggregate Productivity 2.0 2.0 1.5 1.5 1.0 1.0 R 2 = 0.1 2 R = 0.24 0.5 0.5 0.0 0.0 0 10 20 30 40 50 60 0 10 20 30 40 50 Percentage Absolute Contribution of Infrastructure to Allocativ e Percentage Absolute Contribution of Infrastructure to Allocativ e Efficiency v ia simulations Efficiency (TFP in logs) Figure 5.15. Cross-plot between Figure 5.16. Cross-plot between percentage absolute contribution to percentage absolute contribution to average log­productivity and allocative efficiency (with TFP in logs) contributions via simulations and contributions via simulations 70 60 Percentage Absolute Contribution of Infrastruct 60 50 Percentage Absolute Contribution of Infrastruct Average log-Productivity 50 Allocative Efficiency (TFP in logs) 40 40 30 30 20 20 2 R = 0.47 R 2 = 0.59 10 10 0 0 0 10 20 30 40 50 60 Percentage Absolute Contribution of Infrastructure to Av erage 0 10 20 30 40 50 Percentage Absolute Contribution of Infrastructure to Allocative Productiv ity v ia simulations Efficiency via simulations Source: Authors' calculations with IC data. All effects of infrastructure are not supposed to be negative in all cases. There are positive factors intended to stimulate productivity and economic activity, such as the use of ICT or a firm's own electricity generator. A question of interest is to what extent the impacts listed in table 5.1 are due to positive factors that enhance economic performance or negative factors that constrain economic activity. Since the absolute percentage contributions are constructed based on absolute values, at this point we still cannot say anything about the direction of the effect of infrastructure on economic performance. But it is easy to analyze the effect of the individual INF factors, and this is one of the aims of the following section. 41 6. Country-by-country results In the preceding section we evaluated the relative weight of infrastructure among IC variables. In this section the objective is to present a summary of the main results, country by country, focusing on the impact of the individual infrastructure factors or variables. We measure the strength of infrastructure's impact on TFP through three different procedures: (a) elasticities or semi-elasticities, (b) simulations, and (c) evaluation of the IC regressions impact on the sample mean of the variables. These three sets of results provide complementary information. The elasticities and semi- elasticities measure the impact of a change in an independent variable (infrastructure factors or other IC and C variables) on the dependent variable (productivity). But elasticities and semi-elasticities are not directly comparable. On the other hand, simulations measure how the dependent variable changes from scenario A, in which the infrastructure and other IC and C factors are as observed by the survey, to scenario B, in which one of the infrastructure factors improves by, say, 20 percent. From this we can make the following assertions: If the number of power outages suffered by firms in country X is reduced by 20 percent, then the average productivity (or the allocative efficiency) could increase by Y percent, holding everything else constant. Finally, the evaluation at the sample means of the regression variables, as opposed to the simulations, is a static exact decomposition of the terms of the mixed O&P decomposition. We can evaluate the contributions of all the INF, IC, and C factors to the sample mean of average log productivity, identifying the relative importance of each infrastructure variable (for example, losses due to the number of water outages or transport failures) in net terms or in absolute terms. The next subsection focuses on the results of each of the 23 countries. The results are presented in a series of country-specific figures 6.1­6.23. The figures also report the results for the productivity equation. The first panel of each figure shows the elasticities and semi-elasticities; the second and third panels focus on the relative contribution of the INF variables to the average log productivity and to the allocative efficiency in logs; and the fourth and fifth panels present the results of the simulations (that is, how much the average productivity and the allocative efficiency would increase if we improve the INF variables). Note that the results on the elasticities and semi-elasticities are not comparable since they use different measurement scales. For purposes of comparison we should rely on the simulations and on the results of the contributions to the average. In the interest of space, we focus only on the major results for each country. 42 Figure 6.1 Impact of infrastructure on productivity in Algeria A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.4 0.319 0.3 0.2 0.166 0.1 -0.001 0 -0.019 -0.012 -0.008 -0.1 -0.058 -0.2 -0.3 -0.288 -0.4 1 2 3 4 5 6 7 8 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Cost of exports Dummy for own 2 generator 3 Power outages Losses due to water 4 outages 5 Dummy for own well Cost of water from 6 public system 7 Dummy for e-mail 8 Low quality supplies Figure 6.2 Impact of infrastructure on productivity in Benin A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.3 0.257 0.2 0.1 0.003 0 -0.011 -0.018 -0.012 -0.1 -0.071 -0.2 -0.3 -0.4 -0.383 -0.435 -0.5 1 2 3 4 5 6 7 8 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Days to clear customs to export 2 Sales lost due to power outages 3 Sales lost due to water outages 4 Wait for phone connection 5 Water from own well or water infrastructure 6 Dummy for own transportation 7 Sales lost due to delivery delays, domestic 8 Sales lost due to transport delays, international 43 Figure 6.3 Impact of infrastructure on productivity in Botswana Figure 6.4 Impact of infrastructure on productivity in Burkina Faso 44 Figure 6.5 Impact of infrastructure on productivity in Cameroon A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0 -0.02 -0.015 -0.04 -0.037 -0.06 -0.08 -0.069 -0.081 -0.1 -0.12 -0.14 -0.16 -0.18 -0.174 -0.2 1 2 3 4 5 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Days to clear customs to import 2 Average duration of power outages 3 Average duration of water outages 4 Shipment losses 5 Wait for phone connection Figure 6.6 Impact of infrastructure on productivity in Egypt A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.015 0.011 0.01 0.005 0.001 0 -0.005 -0.01 -0.007 -0.015 -0.02 -0.021 -0.022 -0.025 -0.03 -0.035 -0.032 1 2 3 4 5 6 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Electricity from own generator 2 Average duration of power outages 3 Average duration of water outages 4 Dummy for own transportation 5 Shipment losses, exports 6 Days of inventory of main supply 45 Figure 6.7 Impact of infrastructure on productivity in Eritrea A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.02 0.006 0 0 0 -0.02 -0.008 -0.04 -0.038 -0.06 -0.08 -0.1 -0.12 -0.14 -0.138 -0.16 1 2 3 4 5 6 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Days to clear customs to import 2 Cost of electricity from public grid 3 Electricity from own generator 4 Shipment losses 5 Low quality supplies 6 Days of inventory of main supply Figure 6.8 Impact of infrastructure on productivity in Ethiopia A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.1 0.048 0.05 0.006 0 -0.008 -0.004 -0.05 -0.038 -0.1 -0.15 -0.138 1 2 3 4 5 6 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Days to clear customs to import 2 Cost of electricity from public grid 3 Electricity from own generator 4 Shipment losses 5 Low quality supplies 6 Days of inventory of main supply 46 Figure 6.9 Impact of infrastructure on productivity in Kenya A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.25 0.226 0.2 0.142 0.15 0.096 0.1 0.072 0.067 0.053 0.05 0 -0.019 -0.013 -0.02 -0.05 -0.019 -0.053 -0.07 -0.1 1 2 3 4 5 6 7 8 9 10 11 12 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 1 Dummy for own generator 20% improvement in infrs. vars. 20% improvement in infrs. vars. 2 Cost of electricity from public grid 3 Water outages 4 Dummy for own well or water infrastructure 5 Sales lost due to power outages 6 Dummy for own transportation for workers 7 Dummy for contract with transportation company 8 Transport delay, outcoming domestic merch. 9 Transport delay, incoming export merch. 10 Shipment losses, domestic 11 Dummy for webpage 12 Days of inventory of main Figure 6.10 Impact of infrastructure on productivity in Madagascar A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.2 0.151 0.15 0.127 0.1 0.05 0.0004 0 -0.017 -0.023 -0.021 -0.05 -0.049 -0.1 1 2 3 4 5 6 7 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Electricity from own generator 2 Average duration of power outages 3 Cost of water from own well 4 Phone outages 5 Dummy for own transportation for workers 6 Dummy for contract with transportation company 7 Sales lost due to delivery delays, domestic 47 Figure 6.11 Impact of infrastructure on productivity in Malawi A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.1 0.06 0.05 -0.002 0.002 0.002 -0.001 -0.003 0 -0.05 -0.031 -0.011 -0.042 -0.051 -0.057 -0.1 -0.15 -0.2 -0.25 -0.3 -0.317 -0.35 1 2 3 4 5 6 7 8 9 10 11 12 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 1 Days to clear customs to 20% improvement in infrs. vars. 20% improvement in infrs. vars. import 2 Power outages 3 Electricity from own generator 4 Cost of electricity from own generator 5 Average duration of water outages 6 Water from own well or water infrastructure 7 Phone outages 8 Products with own transport 9 Shipment losses, domestic 10 Dummy for e-mail 11 Low quality supplies 12 Sales lost due to delivery delays, domestic Figure 6.12 Impact of infrastructure on productivity in Mali A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 2 1.453 1.5 1 0.5 0.338 -0.021 0.008 -0.002 -0.006 0 -0.128 -0.5 -1 -0.969 -1.5 1 2 3 4 5 6 7 8 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Dummy for equipment damaged by power fluctuations 2 Losses due to phone outages 3 Electricity from own generator 4 Low quality supplies 5 Cost of water 6 Dummy for own roads 7 Dummy for own waste disposal 8 Sales lost due to delivery delays, domestic 48 Figure 6.13 Impact of infrastructure on productivity in Mauritania Figure 6.14 Impact of infrastructure on productivity in Mauritius A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0 -0.02 -0.015 -0.04 -0.037 -0.06 -0.08 -0.069 -0.081 -0.1 -0.12 -0.14 -0.16 -0.18 -0.174 -0.2 1 2 3 4 5 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Days to clear customs to import 2 Average duration of power outages 3 Average duration of water outages 4 Shipment losses 5 Wait for phone connection 49 Figure 6.15 Impact of infrastructure on productivity in Morocco A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.06 0.049 0.04 0.02 0.002 0 -0.003 -0.002 -0.02 -0.04 -0.035 -0.06 -0.051 -0.08 -0.086 -0.1 1 2 3 4 5 6 7 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Days to clear customs to import 2 Sales lost due to power outages 3 Electricity from own generator 4 Shipment losses 5 Dummy for webpage 6 Wait for phone connection 7 Low quality supplies Figure 6.16 Impact of infrastructure on productivity in Namibia A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.04 0.018 0.02 -0.001 0 -0.02 -0.023 -0.04 -0.06 -0.08 -0.1 -0.104 -0.12 1 2 3 4 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Shipment losses in customs to export 2 Power outages 3 Electricity from own generator 4 Average duration of transport by road 50 Figure 6.17 Impact of infrastructure on productivity in Niger A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0 -0.02 -0.015 -0.04 -0.037 -0.06 -0.08 -0.069 -0.081 -0.1 -0.12 -0.14 -0.16 -0.18 -0.174 -0.2 1 2 3 4 5 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Days to clear customs to import 2 Average duration of power outages 3 Average duration of water outages 4 Shipment losses 5 Wait for phone connection Figure 6.18 Impact of infrastructure on productivity in Senegal A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.15 0.118 0.118 0.103 0.1 0.05 0 -0.05 -0.032 -0.037 -0.034 -0.051 -0.1 -0.15 -0.171 -0.2 1 2 3 4 5 6 7 8 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Power outages 2 Losses due to phone outages 3 Dummy for own generator 4 Low quality supplies 5 Dummy for equipment damaged by power fluctuations 6 Cost of water 7 Dummy for own roads 8 Dummy for contract with transportation company 51 Figure 6.19 Impact of infrastructure on productivity in South Africa A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0 -0.009 -0.02 -0.02 -0.04 -0.06 -0.08 -0.1 -0.1 -0.12 -0.116 -0.124 -0.14 -0.144 -0.16 1 2 3 4 5 6 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Days to clear customs to import 2 Sales lost due to power outages 3 Water outages 4 Average duration of transport failures 5 Wait for electric supply 6 Sales lost due to delivery delays Figure 6.20 Impact of infrastructure on productivity in Swaziland A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.04 0.018 0.02 -0.001 0 -0.02 -0.023 -0.04 -0.06 -0.08 -0.1 -0.104 -0.12 1 2 3 4 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Shipment losses in customs to export 2 Power outages 3 Electricity from own generator 4 Average duration of transport by road 52 Figure 6.21 Impact of infrastructure on productivity in Tanzania A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.5 0.411 0.4 0.3 0.2 0.125 0.1 0.002 0.001 0 -0.007 -0.017 -0.018 -0.1 -0.059 -0.138 -0.2 1 2 3 4 5 6 7 8 9 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Electricity from own generator 2 Losses due to water outages 3 Water from own well or water infrastructure 4 Losses due to phone outages 5 Transport outages 6 Dummy for own roads 7 Dummy for webpage 8 Wait for a water supply 9 Low quality supplies Figure 6.22 Impact of infrastructure on productivity in Uganda A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.3 0.226 0.2 0.1 0.003 0.004 0 -0.01 -0.006 -0.1 -0.055 -0.049 -0.074 -0.108 -0.2 -0.203 -0.226 -0.3 -0.4 -0.405 -0.5 1 2 3 4 5 6 7 8 9 10 11 12 Elasticity Semi-elasticity 1 Average number of days to clear an incoming container D. Percentage average productivity gain from E. Percentage allocative efficiency gain from through port 2 Electricity from own 20% improvement in infrs. vars. 20% improvement in infrs. vars. generator 3 Cost of electricity from generator 4 Average duration of power fluctuations 5 Sales lost due to power outages 6 Dummy for own well or water infrastructure 7 Cost of water from own well 8 Phone outages 9 Transport delay, incoming domestic merch. 10 Transport delays in international sales 11 Illegal payments to obtain public utilities 12 Days of inventory of finished goods 53 Figure 6.23 Impact of infrastructure on productivity in Zambia A. Infrastructure elasticities and semi- B. Percentage absolute contribution of C. Percentage absolute contribution of elasticities with respect to productivity infrastructure variables to average log- infrastructure variables to allocative productivity efficiency in logs 0.02 0.011 0 -0.02 -0.012 -0.011 -0.022 -0.04 -0.039 -0.06 -0.08 -0.1 -0.12 -0.123 -0.14 -0.141 -0.16 1 2 3 4 5 6 7 Elasticity Semi-elasticity D. Percentage average productivity gain from E. Percentage allocative efficiency gain from 20% improvement in infrs. vars. 20% improvement in infrs. vars. 1 Electricity from own generator 2 Cost of electricity from public grid 3 Average duration of power outages 4 Average duration of phone outages 5 Transport failures 6 Low quality supplies 7 Shipment losses 54 6.1 Infrastructure impacts on TFP by country High-growth countries in southern Africa Mauritius (MUS). Mauritius is the top-ranked in terms of per capita GDP and demeaned aggregate productivity, and the second according to the DBR (2007) and firms' perceptions of the quality of infrastructure (Table C.1). The relative contribution of infrastructure to average log productivity is 26.6 percent--one of the lowest among the African countries considered (see figure 5.5, panel B, and figure 6.14, panel B). The most important constraint on productivity comes from the number of days to clear customs for exports (17 percent on average) (see figure 6.14); 5.2 percent is due to the use of IC technologies (positive factor); and low-quality provision of electricity and water accounts for only 2 percent of the average log productivity. Swaziland (SWZ). Swaziland is ranked second in terms of per capita GDP (Table C.1). Productivity of firms is negatively affected by shipment losses in customs, the number of power outages, and the average duration of transport by road (see figure 6.20). These results are common to Namibia and Botswana since the countries are pooled together for estimation. The use of generators has a positive sign, meaning that it stimulates productivity. Country-specific results for Swaziland show that the largest contribution to average log productivity comes from problems in customs during the export process (10 percent), and from the number of power outages (9 percent). South Africa (ZAF). South Africa is ranked third based on per capita GDP and demeaned aggregate productivity (Table C.1). Productivity is negatively affected by the days to clear customs to import, the sales lost due to power outages, the number of water outages, the time waiting for an electricity supply, and sales lost due to delivery delays. Therefore, the low quality of the customs services, electricity services, and water affects productivity performance at the firm level in South Africa. The contribution to average log TFP of electricity provision is 6.9 percent, and the contribution of water provision is 5.7 percent. Time wasted in customs while importing accounts for 9.4 percent of the average log productivity. Lastly, problems in transport services represent 5.7 percent of average log productivity (see figure 6.19). Botswana (BWA). Botswana is ranked fourth based on per capita income (DBR, 2007) and on perceptions and sixth in terms of demeaned aggregate productivity. The productivity of firms located in Botswana is affected by shipment losses in customs while exporting (negative), power outages (negative), the percentage of electricity that comes from firm's own generators (positive), and the average duration of transport by road (negative) (see panel A of figure 6.3 and panel B of figure 5.5). These marginal effects are common to Namibia and Swaziland since the countries are pooled together for estimation. Country-specific results show that the largest contributor to both average log productivity and to allocative efficiency is shipment loss in customs while exporting (panel B). Simulations show that the largest productivity improvement comes from reduction in power outages (panel D of figure 6.3). That is, according to our simulations, if the number of power outages suffered by firms in Botswana were reduced by 20 percent, average productivity could increase by 2.1 percent. 55 Namibia (NAM). Namibia is ranked sixth in terms of per capita GDP and ACR (2007) (Table C.1). Productivity of firms is negatively affected by shipment losses in customs while exporting, the number of power outages, and the average duration of transport by road (see figure 6.16). These results are common to Botswana and Swaziland since the countries are pooled together for estimation. Country-specific results show that the impact of infrastructure on the productivity of manufacturing firms in Namibia mainly comes from problems in customs while exporting--this factor represents 9 percent of the average log productivity. Problems from electricity provision (power outages) and from use of alternative power infrastructure (such as a generator) represent 2.2 percent and 3 percent of the average log productivity, respectively, in absolute terms (figure 6.16). High-growth countries in North Africa Algeria (DZA). Algeria is ranked fifth in terms of per capita GDP, and seventh in terms of firm perceptions and demeaned aggregate productivity. The results on the productivity impact of infrastructure (see figure 6.1) show the total effect on absolute value is as large as 48.6 percent of average log TFP. The quality of infrastructure variables affecting TFP are: cost of exports, having an own generator, number of power outages, losses due to water outages, having an own well, the cost of water from the public system, having e-mail, and low-quality supplies. The largest and most positive effect comes from having e-mail, which could represent 14.5 percent of average log TFP. Egypt (EGY). Egypt is ranked seventh based on per capita GDP, sixth in terms of ACR (2007), fourth in terms of demeaned aggregate TFP, and third in firms' perceptions of infrastructure quality. From the econometric analysis, the contribution of infrastructure to the average log productivity in Egypt is only 26 percent. The main infrastructure factors affecting firms' productivity are the average duration of water and power outages (both with negative effects), the percentage of firms with their own generator (positive effect), the dummy for own transportation (positive), shipment losses in exports (negative), and days of inventory of the main supply (negative) (figure 6.6). Morocco (MAR). The perceptions of the managers of the Moroccan firms suggest that infrastructure is not a major concern when compared to other IC constraints; it is ranked first in Table C.1. But according to figure 6.15, the contribution of infrastructure to average log productivity is 31.3 percent, with the largest impacts coming from (a) the average time to clear customs to import, and (b) the time wasted to obtain a phone connection. A 20 percent reduction of average customs delays for imports could increase average productivity by 1.6 percent and allocative efficiency by 0.4 percent. Notice, that the ranking based on the econometric analysis (demeaned aggregate productivity) is consistent with the ranking based on per capita GDP (see Table C.1), which establishes that Morocco is in eighth position, not first. Low-growth countries in central-west Africa Cameroon (CMR). Cameroon is rank ninth in term of per capita GDP, which is somehow surprising if we compare it with the results of the rest of the rankings based on DBR (2007), ACR (2007), firms' perceptions, and demeaned aggregate productivity. We found that the productivity of manufacturing 56 firms in Cameroon is reduced by the following factors: number of days required to clear customs for imports, average duration of power and of water outages, shipment losses, and time waiting for a phone connection (panel A of figure 6.5). These factors are common to Mauritania, Burkina Faso, and Niger since the countries are pooled together for estimation purposes. Country-specific results show that the largest contributions to average log productivity come from the number of days waiting to clear customs, duration of power outages, and from the time waiting for a phone connection (panel B of 6.5). Infrastructure represents 41.6 percent of average log TFP. Mauritania (MRT). Mauritania ranks 10th in terms of per capita GDP, 13th in term of the total absolute contribution of infrastructure to average log productivity, 17th in terms of firm perceptions, and 15th in terms of demeaned aggregate productivity. These results are common to Burkina Faso, Cameroon, Niger, Ethiopia, and Eritrea since those countries are pooled together for estimation purposes. Delays in customs while importing represent 22 percent of average log productivity and 12 percent of allocative efficiency. In terms of simulation, a 20 percent improvement in this variable could cause a 13.9 percent increase in average productivity and a 7.8 percent one in allocative efficiency. Low-quality provision of electricity and water and its indirect costs also reduce average productivity in Mauritania (see figure 6.13). Senegal (SEN). Senegal reveals a high infrastructure impact on the TFP of manufacturing firms. The percentage contribution of infrastructure to the average log productivity of this country is 58.5 percent; the indirect costs stemming from the low-quality provision of electricity represent 9.3 percent of this (see figure 6.18). The use of own-power infrastructure partially alleviates the negative impact of the low quality of electricity provision. The relative importance of problems in transport services (such as low-quality supplies) is very high; this variable represents 23.4 percent of average log productivity and 14 percent of allocative efficiency. A 20 percent reduction in the percentage of low- quality supplies received may cause a 3.2 percent increase in average productivity. Notice that the simulation of a 20 percent improvement in the percentage of low-quality supplies received causes a decrease in allocative efficiency of -0.4 percent. The reason for this phenomenon is clear: the allocative efficiency is simply the covariance between productivity at the firm level and share of sales. Therefore, a negative rate of change of the allocative efficiency indicates that the firms receiving a larger share of low-quality supplies are those with the largest market shares. Benin (BEN). Benin is ranked 12th in per capita GDP, 11th in terms of demeaned aggregate productivity, and 13th in terms of DBR (2007). The time waiting for phone connections and to clear customs in order to export are the two factors that most negatively contribute to average log productivity (see panel B of figure 6.2). An independent 20 percent improvement in these two variables could increase average productivity by 3.8 percent and 4.3 percent, respectively (panel D of figure 6.2). The same holds for the allocative efficiency term (panels C and E). Mali (MLI). Mali is ranked 14th in per capita GDP terms, in demeaned aggregate productivity, and in the ACR (2007). The total contribution of infrastructure to average log productivity in Mali is 42.7 percent. The low quality of electricity, water, and phone provision accounts for almost 32 percent 57 of the average log productivity. The use of firms' own roads is a factor that increases productivity (figure 6.12). Burkina Faso (BFA). Burkina Faso ranks 16th and 12th in terms of per capita GDP and demeaned aggregate TFP, respectively. The main infrastructure problems are clearing customs while importing, the average duration of power and water outages, shipment losses, and time waiting to obtain a phone connection. These results are common to Mauritania, Cameroon, and Niger since these countries are pooled for estimation. In particular, for Burkina Faso, all of these factors reduce productivity at the firm level (see figure 6.4) and can contribute to 35 percent of average log TFP. Niger (NER). Niger is one of the poorest countries in our sample: it ranks 19th based on per capita income, and managers' perceptions show a great concern regarding quality of current infrastructure (20th in the rank). The absolute contribution of infrastructure to average log productivity is 34.7 percent, with 20.7 percent due to problems clearing customs while importing, 9.4 percent due to the average time wasted in obtaining a phone connection, and 4.5 percent due to the cumulated negative effect of the low-quality provision of electricity and water and the poor transport system (see figure 6.17). Low-growth countries in central-east Africa Kenya (KEN). Kenya ranked 13th in per capita GDP, 6th in terms of DBR (2007), and 8th in terms of demeaned aggregate productivity. The results of the productivity equation in Kenya show multiple interrelationships between productivity at the firm level, on one hand, and infrastructure, on the other. Factors such as the use of generators, the cost of electricity from the public grid, water outages, sales lost due to power outages, and so on, reduce productivity at the firm level (see figure 6.9). Although the contribution to the average log productivity of any of these factors is over 6 percent, infrastructure as a whole accounts for 30 percent of average log productivity, and for 19.9 percent of allocative efficiency (which illustrates the major influence infrastructure has on Kenyan firms' productivity). Uganda (UGA). Uganda is ranked 15th in per capita GDP and 17th in demeaned aggregate productivity. Uganda's average log productivity is strongly influenced by infrastructure conditions, representing 58.4 percent. The two main factors affecting average productivity are the time to clear customs and the provision of electricity (figure 6.22, panel D). Zambia (ZMB). Zambia ranks 17th in terms of per capita GDP, 14th in terms of firm perceptions, and 18th in terms of the ACR (2007) .The contribution of the cost of electricity from the public grid to average log productivity in Zambia is 32.5 percent. The contribution of the average duration of power outages to average log TFP is 9.1 percent (see figure 6.23). The total contribution of infrastructure to average TFP is 50.6 percent. Tanzania (TZA). Tanzania is 18th based on per capita GDP, 14th in terms of DBR (2007), and 15th in terms of firm perceptions. The absolute contribution of infrastructure to average log productivity in Tanzania is 34.1 percent. Out of this, 14.8 percent is due to time wasted waiting for water supply and 5.5 percent is due to the number of transport outages (figure 6.21). 58 Malawi (MWI). Malawi is ranked 20th in per capita GDP terms and 22nd in terms of ACR (2007). The econometric evidence shows that the aggregate productivity of Malawian's manufacturing firms are dramatically affected by infrastructure quality (83 percent of average log TFP). Delays in clearing customs while importing account for 25 percent of the average log productivity. A 20 percent improvement in this variable could increase the average productivity by 6 percent and the allocative efficiency by 1.5 percent (see panels D and E of figure 6.11). Symmetrically, productivity decreases as the number of power outages increases; the percentage contribution of this variable to average log productivity is 9.2 percent and to the allocative efficiency is 3.5 percent. The cost of electricity from firms' own generators is another antiproductive factor. Water provision also impacts the productivity of Malawian firms (figure 6.11). Madagascar (MDG). Madagascar is ranked 21st in per capita GDP terms, and 16th and 17th in terms of demeaned aggregate productivity and DBR (2007), respectively. How infrastructure may impact firm-level productivity is clear from the results obtained in figure 6.10. The factors related to electricity supply are intimately linked to productivity. Water costs and the number of phone outages also reduce productivity at the firm level. The total contribution of infrastructure to average log productivity in Madagascar is 31 percent and to allocative efficiency is 28 percent. Ethiopia (ETH). Ethiopia is ranked 22nd in terms of per capita GDP and 21st in terms of firms' perceptions. The productivity of Ethiopian manufacturing firms is negatively affected by the days to clear customs to import, the cost of electricity from the public grid, shipment losses while in transit, and the percentage of supplies that are of lower than agreed-upon quality. Positive effects on productivity come from the percentage of electricity from firms' own generators and from the days of inventory of the main supply. These results are common to Eritrea since the countries are pooled together for estimation. The largest contribution to the average log productivity is by the days of inventory of main supply, days to clear customs to import, the cost of electricity from the public grid, and the electricity from firms' own generators (figure 6.8). In total, infrastructure represents 32 percent of average log TFP, which seems too low. This might indicate that pooling with Eritrea for estimation of IC elasticities might not be a good idea in this case. Eritrea (ERI). Eritrea is the last country in the ranking, in 23rd position in terms of per capita GDP. The total number of days waiting to clear customs to import, the cost of electricity from the public grid, shipment losses, and low-quality supplies are the factors with negative effects on firms' productivity (see figure 6.7). Remember that these results are common to Ethiopia since the countries are pooled together for estimation. The cost of energy from the public grid accounts for a 22.9 percent of average log productivity, almost 50 percent more than the second factor (days to clear customs to export) in order of importance. In total, infrastructure represents 48 percent of average log TFP. 6.2 Summary of the main empirical results The aim of this subsection is to summarize the main empirical results obtained, country by country, including the absolute percentage contribution of the infrastructure group of variables to the sample 59 means of productivity in logs. The customs clearance subgroup includes those variables related to the ease or difficulty of clearing customs when exporting or importing. Within provision of electricity we have grouped all the variables related to low-quality provision of electricity (number of power outages, power fluctuations, cost of electricity, and so on). Use of power infrastructure is intended to enclose all the variables related to the use of alternative sources of energy, such as generators. Similarly, the subgroups provision of water and provision of phone connections includes all the variables related to the quality of the provision of these utilities, whereas use of water infrastructure includes the use of firms' own wells or boreholes to replace the public provision of water. Obviously, use of ICT takes into account the use of ICT in firms' commercial operations. Lastly, transport services contains all variables relating to the quality of transportation services, such as shipment losses in transit, transport delays, delivery delays, and so on. Own transport infrastructure includes the use of own transportation for products or workers. Figure 6.24 shows the prominent influence of low-quality electricity provision on average log productivity in the different countries considered in this report. The percentage absolute contribution of this group of variables to average log productivity ranges from 34.1 percent in Zambia to 0.3 percent in Morocco, being a negative effect in all cases. Only in Tanzania was the low quality of electricity provision not statistically significant, probably due to the significant and very influential effect of water provision in this country. The low quality of electricity and the continuous outages are partially alleviated by the use of own-power infrastructure, as the positive effect of the group use of power infrastructure shows. Another group of variables with a statistically high impact on average log productivity is customs clearance. The contributions of this group are negative and very large in most countries, indicating a clear and pervasive constraining effect of the time wasted in customs when importing or exporting. Figure 6.24 Infrastructure's impact on average log productivity by key factors (I) A. Percentage absolute contribution of infrastructure to average log-productivity 100.0 90.0 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 MWI BEN SEN UGA ETH ZMB DZA ERI MLI CMR MRT BFA NER TZA MAR MDG KEN ZAF MUS EGY SWZ BWA NAM Infrastructures Other IC 60 B. Percentage absolute contribution of infrastructure to average log-productivity by key factors 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% MWI BEN SEN UGA ETH ZMB DZA ERI MLI CMR MRT BFA NER TZA MAR MDG KEN ZAF MUS EGY SWZ BWA NAM Customs clearance Provision of electricity Use of generator or power infrastructures Provision of water Use of own water infrastructure Provision of phone Use of ICT Transport services Own transport infrastructures Other Source: Author's calculations with IC data. Note: Customs clearance includes: days to clear customs to export and import; shipment losses in customs; inspections in customs; wait for an import license. Provision of electricity includes: power outages; avg. duration of power outages; losses due to power outages, wait for an electricity supply; power fluctuations; avg. duration of power fluctuations; cost of electricity from the public grid; cost of electricity from private system. Use of power infrastructures includes: dummy for own generator; electricity from own generator; dummy for own power infrastructures (excl. generators). Provision of water includes: water outages; avg. duration of water outages; losses due to water outages, wait for a water supply; cost of water from the public grid; cost of water from private system. Use of water infrastructures includes: dummy for own water infrastructures; water from own well. Provision of phone includes: phone outages; avg. duration of phone outages; losses due to phone outages, wait for a phone connection. Use of ICT includes: dummy for e- mail; dummy for webpage. Transport services includes: sales lost due to transport delays; sales lost due to delivery delays; shipment losses; low quality supplies; transport delays. Own transport infrastructures include: dummy for own roads; dummy for own transportation for workers; products with own transport. Other: inventories, illegal payments to obtain public utilities. Regarding provision of water, the relative importance of this group of variables is lower when compared to the provision of electricity; nevertheless, there are some cases (such as Tanzania, Mali, and Kenya) where the contribution of this group of variables is very high, even compared to the provision of electricity. As with the provision of electricity, the use of alternative water infrastructure such as boreholes or wells has a positive impact on plants' productivity--an effect that only appears to be significant when there is a negative effect of water provision, suggesting the existence of a replacement effect between the public provision of water and alternative supplies of water. The poor quality of phone provision is negatively related to productivity in 14 countries; nevertheless, the quantitative contribution of this group of variables is, in general, lower than the impact of the electricity provision. Benin is an exception. The contribution to the average log productivity of telephone provision in this country is 40.8 percent--more than 20 times the contribution of electricity provision, which is only 2 percent. The use of IC technologies is positively related to productivity, but the use of these technologies was only significant in the productivity regressions of six countries: Malawi, Algeria, Tanzania, Kenya, Morocco, and Mauritius, with the largest impacts in Algeria (14.5 percent) and Mauritius (5.2 percent). Problems with product transport are negatively related to productivity in all the cases, with the exception of Botswana, Swaziland, and Namibia, for which no variables for this group were significant in productivity regressions. The largest impacts of this subgroup of variables were seen in Senegal, 61 Tanzania, Madagascar, South Africa, and Zambia; nevertheless, the contributions of transport services to average log productivity are relatively lower than the impact of the provision of electricity or the customs clearance subgroups. On the other hand, the use of own-transport infrastructure stimulates productivity growth; in all the cases in which any variable belonging to this subgroup was significant in the productivity regressions, it appeared with a positive sign. But the positive effects on productivity of these factors were concentrated in only seven countries: Malawi, Benin, Senegal, Eritrea, Tanzania, Kenya, and Egypt. Figure 6.25 provides similar information, but, in this case, we have grouped the different infrastructure factors in only five groups: customs clearance, electricity, water, telecommunications and ICT, and transportation. From this figure the high influence of electricity factors become even clearer. When we include the provision of electricity and use of generators or power infrastructure into a single group, the resulting block of electricity factors can explain more than half of average log productivity in Uganda, Ethiopia, Zambia, Eritrea, Swaziland, and Botswana. The water group is relatively important in Tanzania, Kenya, and Egypt. The customs clearance group gains importance in those countries with a more patent export orientation, such as Mauritius, Botswana, and Namibia. Finally, as has been signaled, the transportation subgroup explains more than half of the whole infrastructure impact on average log productivity in Senegal and Madagascar. Figure 6.25 Infrastructure impact on average log productivity by key factors (II) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% MWI UGA BEN SEN ETH ZMB ERI MLI CMR DZA MRT BFA NER TZA KEN MAR MDG ZAF EGY SWZ MUS BWA NAM Customs clearance Electricity Water Telecoms and ICT Transportation Other Source: Author's calculations with IC data. Note: For a description of the variables contained in each group see footnote in Figure 6.24 Figure 6.26 reports the absolute percentage contribution of infrastructure by key factors via simulations. The results are fully consistent with the ones provided by figure 6.24. The relative weights of the electricity factors dominate in more than half the countries, and the water and telecommunications and ICT subgroups tend to play a secondary role in explaining average productivity when compared to electricity. The main difference with respect to figure 6.24 is the lower relative weight of the customs clearance group. Once again, it should be pointed out that the 62 information provided by the simulations complements the results from the evaluation of the sample average of log productivity. In this case, we are talking about a cumulative effect, all other things being equal, since we evaluate the change in the average productivity when one of the INFs changes. Figure 6.26 Infrastructure's impact on average productivity by key factors via simulations 100% 80% 60% 40% 20% 0% ERI MW I SEN UGA MLI DZA BEN ETH TZA NER CMR MRT BFA SW Z KEN BW A NAM MUS EGY ZAF MAR ZMB MDG Customs clearance Electricity Water T elecoms and ICT T ransportation Source: Author's calculations with IC data. Note: For a description of the variables contained in each group see footnote in Figure 6.24 The summary of results is complemented by the analysis of allocative efficiency. Figures 6.27 and 6.28 show that the impact of infrastructure on allocative efficiency is equally distributed among the different infrastructure factors. The impact of water is, in this case, larger than in the case of average productivity, while the impact of energy provision is considerably less--gaining relative importance with the use of own generators. The positive effect of the use of own generators on allocative efficiency indicates that those firms that accumulate a larger proportion of market sales are also the firms that use their own generators. The same holds for the use of IC technologies and the use of own-water infrastructure. The customs clearance group has important implications for the allocative efficiency of Namibia, Mauritania, and Botswana. Finally, the quality of transport services plays a secondary role in explaining the behavior of allocative efficiency in the different countries included in our sample. Figure 6.27 Infrastructure impact on allocative efficiency in logs by key factors (I) A. Percentage absolute contribution of infrastructure to allocative efficiency in logs 100.0 90.0 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 MWI BEN SEN UGA ETH ZMB DZA ERI MLI CMR MRT BFA NER TZA MAR MDG KEN ZAF MUS EGY SWZ BWA NAM Infrastructures Other IC 63 B. Percentage absolute contribution of infrastructure to allocative efficiency in logs by key factors 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% MWI BEN SEN UGA ETH ZMB DZA ERI MLI CMR MRT BFA NER TZA MAR MDG KEN ZAF MUS EGY SWZ BWA NAM Customs clearance Provision of electricity Use of generator or power infrastructures Provision of water Use of own water infrastructure Provision of phone Use of ICT T ransport services Own transport infrastructures Other Source: Author's calculations with IC data. Note: For a description of the variables contained in each group see footnote in Figure 6.24 Figure 6.28 organizes the different subgroups of infrastructure factors into five key groups. From this figure, the important contribution of the electricity subgroup becomes even clearer. Transportation explains more than 50 percent of the allocative efficiency in logs of Morocco. Once again, the relative importance of the water, telecommunications, and ICT subgroups is lower when compared to the contribution of electricity and customs clearance. Figure 6.28 Infrastructure's impact on allocative efficiency in logs by key factors (II) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% MWI UGA BEN SEN ETH ZMB ERI MLI CMR DZA MRT BFA NER TZA KEN MAR MDG ZAF EGY SWZ MUS BWA NAM Customs clearance Electricity Water T elecoms and ICT T ransportation Other Source: Author's calculations with IC data. Note: For a description of the variables contained in each group see footnote in Figure 6.24 64 7. Conclusions For Africa's awaited growth resurgence to occur, a broad range of factors--political, institutional, and economic--must be improved. The World Bank's landmark Africa Competitiveness Reports (2004 and 2007) focus on problems that inhibit economic growth. It is agreed that improving Africa's infrastructure is a crucial step toward penetrating international markets and meeting the goals of continuous growth and poverty reduction. Infrastructure quality has a pervasive influence on all areas of an economy. Low-quality infrastructure and limited transport and trade services increase logistical and transaction costs, rendering otherwise competitive products uncompetitive, as well as limiting rural production and people's access to markets--with adverse effects on economic activity and poverty. A large number of empirical studies illustrate the impact of infrastructure on economic performance. All suggest that Africa's infrastructure gap is an important growth bottleneck, with a negative impact on productivity and the overall competitiveness of the region. Using the methodology of Escribano and Guasch (2005, 2008) and Escribano et al. (2008), several studies have found empirical evidence--in cases such as Brazil, Chile, Costa Rica, Mexico, Turkey, and several Southeast Asian countries--that improvements in investment climate conditions in general, and in infrastructure quality in particular, may lead to important gains in productivity and in other economic performance measures: employment, real wages, exporting activities, and inflows of foreign direct investment (FDI). Approach and methods This paper provided a systematic, empirical assessment of the impact of infrastructure quality on the TFP of African manufacturing firms. We applied micro-econometric techniques to investment climate surveys of 26 African countries to gauge the impact of infrastructure quality on TFP. For each country we estimated, by regression techniques, the impact of infrastructure quality based on 10 different productivity measures and showed that the results were robust once we controlled for other observable fixed effects (red tape, corruption and crime, finance, innovation and labor skills, and so on) obtained from the investment climate surveys (see Escribano and Guasch, 2005, 2008). We pooled data from the investment climate surveys only for the few African countries for which we did not have sufficient observations for estimation purposes. Otherwise, we performed a country-by-country estimation to reveal firm and industry information by country. After pooling the data from several countries, the econometric results were then suitably disaggregated following the method of Olley and Pakes (1996) (Escribano et al., 2008), which allowed us to make country-specific evaluations of the impact of investment-climate quality on aggregate TFP, average TFP, and allocative efficiency. We ranked the African countries in the study according to several aggregate indices: per capita income, ease of doing business, firm perceptions of growth bottlenecks, and the recent concept of 65 demeaned productivity (demeaned TFP), which overcame the problem of comparing apples and oranges when doing TFP cross-country comparisons (Escribano et al., 2008). We found the concept of demeaned productivity very useful because it is highly correlated with per capita income, ease of doing business indices, firm's perceptions of growth bottlenecks, and the results of the Africa Competitiveness Reports. Furthermore, the information obtained from the investment-climate determinants of demeaned TFP provided a much deeper insight into the firm-level investment-climate infrastructure elements that are constraining productivity growth in African countries. We distinguished two clear blocks of countries in Africa. The first block comprised countries with faster, steadier growth rates. These are mainly in the south, including Mauritius, Swaziland, South Africa, Botswana, Namibia, and Lesotho. The block also included Algeria, Morocco, and Egypt from the north, and Cape Verde and Cameroon from Central Africa. In southern Africa Botswana, Namibia, and Swaziland, were pooled for estimation purposes. In the second block were Mauritania, Senegal, Benin, Mali, Niger, and Burkina Faso in the central-west; and Uganda, Kenya, Zambia, Tanzania, Malawi, Burundi, Madagascar, Ethiopia, and Eritrea in the central-east. These countries have experienced lower and more irregular growth rates, with periods of both positive increase and persistent decrease in per capita income. Pooled for estimation purposes were the West African states (Burkina Faso, Cameroon, Mauritania, and Niger) and Eritrea and Ethiopia. Of the 26 African countries analyzed, Mauritius was, in 1950, the country with the highest per capita income (measured in terms of per capita gross domestic product, GDP), followed closely by South Africa, and, by a larger gap, Namibia and Algeria. But the per capita income levels in 2003 were somewhat different; Mauritius was still ranked first, followed by Swaziland, South Africa, and Botswana--and, by a wider gap--Algeria, Cape Verde, Egypt, and Morocco. Mauritius and Botswana experienced the highest sustained per capita income growth during recent years. Lesotho's rate is the median, splitting the study into two blocks. To better understand the convergence or divergence of trends, we plotted the per capita income of each African country relative to the per capita income of the United States. Convergence was observed only in Mauritius, Swaziland, and Botswana. For all other study countries, including South Africa, per capita income was found to be diverging from the United States, while, in a few (Egypt, Morocco, and Cape Verde) the ratio was stable. While persistently positive GDP growth allowed Mauritius's per capita income to reach 45 percent of the United States' in 2003, this was clearly the exception (together with Swaziland and Botswana). For the rest of the countries, including South Africa, relative per capita income was much lower in 2003 than in 1960 (indicating a divergence). In fact, the 2003 per capita income of several countries was no larger than 5 percent of the per capita income of the United States. As expected, labor productivity was the main factor explaining this divergence, given that labor force participation has a steady influence. Since total factor productivity (TFP) is usually a key factor explaining the evolution of labor productivity, in this paper we used investment climate surveys to identify the main infrastructure-related TFP bottlenecks in Africa. 66 The per capita income rankings were correlated with the rankings obtained from the World Bank's 2007 Doing Business report. In 2007 Mauritius, Swaziland, South Africa, Botswana, and Namibia ranked 32nd, 76th, 29th, 48th, and 42nd in the world based on the ease-of-doing-business indicators. This index considers questions such as the number of days required to start a business and the ease of dealing with licenses, registering a property, trading across borders, employing workers, and so on. Other 2007 rankings include 83rd for Kenya, 97th for Ethiopia, 165th for Egypt, and 170th for Eritrea. We showed the percentage of firms that perceived telecommunications, electricity, customs clearance, and transport as major obstacles to their economic performance. Only in Benin, Kenya, and Zambia did more than 50 percent of firms identify telecommunications as a severe obstacle. Meanwhile, the quality of electricity provision is a major problem for more than 50 percent of firms in more than half of the countries in our sample. In Burundi, Cameroon, Benin, Burkina Faso, and Cape Verde, the percentage of firms considering electricity as a severe or very severe obstacle exceeded 80 percent; on the other hand, only 20 percent of firms in Morocco, South Africa, Botswana, and Namibia considered electricity a severe obstacle. Customs clearance was considered an acute problem in Benin, Kenya, Madagascar, Senegal, and Algeria. Finally, transportation was considered a severe obstacle by more than 70 percent of firms in Burkina Faso and Benin. The World Bank's 2007 Africa Competitiveness Report evaluated a wide range of factors related to economic activity, with infrastructure among them. Once again there were clearly different performance levels across the two blocks of countries. While in Namibia, South Africa, Botswana, Egypt, and Morocco, the quality of infrastructure exceeded the approval level, in the remaining countries this quality was rated low in most cases. The same held for the disaggregated results, including the number of telephone lines and the quality of ports, air transport, and electricity supply. The difference between the two blocks becomes even more apparent when looking at the cross- plots between GDP per capita relative to the United States and firms' perceptions. A preliminary analysis of the cross-plots suggests two points: first, that there is an intuitive and negative relation between income level and infrastructure constraints; and, second, that the division of the two blocks of countries remains intact, showing now the largest dispersion in the constraint perceptions for the lowest per capita income group. Findings Among the countries of the high-income-growth block, infrastructure has a low impact on TFP (see panel B of figures 5.5 and 5.6 and panel A of figure 6.24). Red, tape, corruption, and crime dominate over infrastructure in countries such as Mauritius, Egypt, and South Africa (figure 5.5, panel B). Infrastructure quality has a high impact on TFP in the countries of the low-income-growth block (see panel B of figures 5.5 and 5.6 and panel A of figure 6.24), but the impact is very negative (see panel B of figures 5.3 and 5.4), identifying important bottlenecks for TFP growth. 67 We found much heterogeneity among individual infrastructure factors affecting countries in both the high- and low-growth blocks (see figure 6.26). Among related factors that most influence the average productivity TFP of African firms are: · Poor-quality electricity provision, which affects mainly poor countries, such as Eritrea, Ethiopia, Mali, Senegal, Uganda, Zambia, and Kenya. It also affects countries that are growing faster, in relative terms, such as Botswana, Namibia, and Swaziland. · Problems dealing with customs during importing or exporting affects mainly fast-growing countries, such as Mauritius, Morocco, and Swaziland. But low quality of customs also affects slow-growing countries, such as Niger, Mauritania, Cameroon, Malawi, Burkina Faso, and others. · Losses from transport interruptions affect mainly slower-growing countries, such as Madagascar, Kenya, Tanzania, and Senegal. · Water outages affect mainly slower-growing countries, such as Tanzania, Kenya, Burkina Faso, Mauritania, Niger, and Mali. But it also affects some of the faster-growing countries, such as Egypt. Of the infrastructure determinants that most influence the allocative efficiency of African firms there is also some heterogeneity across countries. · Poor-quality electricity provision affects the allocative efficiency of mainly poor countries, such as Zambia, Mali, Uganda, Eritrea, and Kenya. · Problems dealing with customs while importing or exporting affects mainly slow-growing countries, such as Mauritania, Niger, and Cameroon. But it also affects the allocation efficiency of countries that are growing fast, such as Morocco, Namibia, and Mauritius. · Transport services affects the allocative efficiency of mainly slower-growing countries, such as Madagascar, Senegal, and Tanzania. · Water provision affects the allocative efficiency of mainly slower-growing countries, such as Tanzania, Kenya, and Mali. 68 References Agénor, P. R., and B. Moreno-Dodson. 2006. "Public Infrastructure and Growth: New Channels and Policy Implications." Policy Research Working Paper 4064, World Bank, Washington DC. Artadi, E. V., and X. Sala-i-Martin. 2003. "The Economic Tragedy of the XXth Century: Growth in Africa." National Bureau of Economic Research (NBER) Working Paper 9865, Cambridge, MA. Aschauer, D. A. 1989. "Is Public Expenditure Productive?" Journal of Monetary Economics 23(2): 177­ 200. Bougheas, S., P. Demetriades, and E. Morgenroth. 1999. "Infrastructure, Transport Costs and Trade." Journal of International Economics 47: 169­89. Brunel, S. 2004. "L'Afrique--Un Continent en Réserve de Développement." Editions Bréal, Rosny- sous-Bois. Calderon, C., and L. Serven. 2003. "The Growth Cost of Latin America's Infrastructure Gap." In William Easterly and Luis Serven (eds.), Adjustment Undetermined? Public Deficits and Growth in Latin America, 1980­2000. Princetown, N.J.: Princetown University Press. ------. 2005. "The Effects of Infrastructure Development on Growth and Income Distribution." Mimeo, World Bank, Washington, DC. Calderon, C., W. Easterly, and L. Serven. 2003a. "Infrastructure Compression and Public Sector Solvency in Latin America." In William Easterly and Luis Serven (eds.), Adjustment Undetermined? Public Deficits and Growth in Latin America, 1980­2000. Princetown, N.J.: Princetown University Press. ------. 2003b. "Latin America's Infrastructure in the Era of Macroeconomic Crisis." In William Easterly and Luis Serven (eds.), Adjustment Undetermined? Public Deficits and Growth in Latin America, 1980­2000. Princetown, N.J.: Princetown University Press. Canning, D. 1998. "A Database of World Infrastructure Stocks, 1950­95." Policy Research Working Paper 1929, World Bank, Washington, DC. Center for International Comparisons of Production, Income and Prices. "Penn World Table." http://pwt.econ.upenn.edu/cic_main.html. Dempster, A. P., N. M. Laird, and D. B. Rubin. 1977. "Maximum Likelihood Estimation From Incomplete Data Via the EM Algorithm." Journal of Royal Statistical Society B39: 1­38. Dessus, S., and R. Herrera. 1999. "Public Capital and Growth: An Econometric Study on Panel Data." Revue Economique 50(1): 113­26. Easterly, W., and S. Rebelo. 1993. "Fiscal Policy and Economic Growth." Journal of Monetary Economics 32: 417­58. Economic Commission for Africa (ECA). 2005. "Assessing Regional Integration in Africa." ECA Policy Research Paper. Escribano, A., and J. L. Guasch. 2008. "Robust Methodology for Investment Climate Assessment on Productivity: Application to Investment Climate Surveys From Central America." Working Paper # 08-19 (11), Universidad Carlos III de Madrid. ------. 2005. "Assessing the Impact of Investment Climate on Productivity Using Firm Level Data: Methodology and the Cases of Guatemala, Honduras and Nicaragua." Policy Research Working Paper 3621, World Bank, Washington, DC. Escribano, A. and J. Pena. 2008. "Empirical Econometric Evaluation of Alternative Methods of Dealing with Missing values in Investment Climate Surveys." Unpublished manuscript, Universidad Carlos III de Madrid. Escribano, A., J. L. Guasch, M. de Orte and J. Pena. (2008a). "Investment Climate Assessment Based on Demean Olley and Pakes Decompositions: Methodology and Applications to Turkey's Investment Climate Survey." Working Paper # 08-20 (12), Universidad Carlos III de Madrid. Escribano, A., J. L. Guasch, M. de Orte and J. Pena. (2008b). "Investment Climate and Firm's Performance: Econometric and Applications to Turkey's Investment Climate Survey." Working Paper # 08-20 (13), Universidad Carlos III de Madrid. Escribano, A., J. L. Guasch, M. de Orte and J. Pena. (2009). "Investment Climate Assessment in Indonesia, Malaysia, Philippines and Thailand: Results from Pooling Firm Level Data". Forthcoming in Singapore Economic Review, Special Issue on the Econometric Analysis of Panel Data, 2009. Escribano, A., N. Peltier, L. Garrido, and H. Sing. 2005. "The Impact of Infrastructure on Competitiveness in Latin America: A Firm Level Analysis Based on Investment Climate Assessments." Paper presented at the Joint Conference of the Inter-American Development Bank (IADB) and the World Bank (WB), June 6­7. Esfahani, H. S., and M. T. Ramirez. 2003. "Institutions, Infrastructure, and Economic Growth." Journal of Development Economics, 70: 443­77. Estache, A. 2005. "What Do We Know About Sub-Saharan Africa's Infrastructure and the Impact of Its 1990 Reforms?" Mimeo, World Bank, Washington, DC. Estache, A., B. Speciale, and D. Veredas. 2005. "How Much Does Infrastructure Matter to Growth in Sub-Saharan Africa?" Mimeo, World Bank, Washington, DC. Fedderke, J. W., P. Perkins, and J. M. Luiz. 2005. "Infrastructural Investment in Long-Run Economic Growth: South Africa 1875­2001." World Development, forthcoming. Fedderke, J. W., and Z. Bogetic. 2006. "Infrastructure and Growth in South Africa: Direct and Indirect Productivity Impacts of 19 Infrastructure Measures." Working Paper 3989, World Bank, Washington, DC. Francois, J., and M. Manchin. 2006. "Institutional Quality, Infrastructure, and the Propensity to Export." Unpublished manuscript, The World Bank, Washington D.C. Frutos, R. F., M. Gracia-Diez, and T. Perez-Amaral. 1998. "Public Capital Stock and Economic Growth: An Analysis of the Spanish Economy." Applied Economics 30(8): 985­94. Galiani, S., P. Gertler, and E. Schargrodsky. 2005. "Water for Life: The Impact of Privatization of Water Services on Child Mortality." Journal of Political Economy 113: 83­120. Guasch, J. L. 2004. "Granting and Renegotiating Infrastructures Concessions, Doing It Right." World Bank Institute (WBI) Development Studies, Report 28816, World Bank, Washington, DC. 70 Hall, R. 1990. "Invariance Properties of Solow's Productivity Residual." In P. Diamond (ed.), Growth, Productivity, Unemployment. Essays to Celebrate Bob Solow's Birthday. Cambridge, MA: Massachusetts Institute of Technology (MIT) Press. Hulten, C. R. 1996. "Infrastructure Capital and Economic Growth: How Well You Use It May Be More Important than How Much You Have." NBER Working Paper 5847, Cambridge, MA. Kumar, N. 2001. "Infrastructure Availability, Foreign Direct Investment Inflows and Their Export- orientation: A Cross-Country Exploration." Research and Information System for Developing Countries, New Delhi, India. Limao, N., and A. Venables. 2001. "Infrastructure, Geographical Disadvantage, Transport Costs and Trade." World Bank Economic Review 15: 451­79. Little, R. J. A., and D. B. Rubin. 1987. "Statistical Analysis With Missing Data." Wiley Series in Probability and Mathematical Statistics. John Wiley and Sons (eds). Loayza, N., P. Fajnzylber, and C. Calderón. 2002. "Economic Growth in Latin America and the Caribbean." Mimeo, World Bank, Washington, DC. Loree, D. W., and S. E. Guisinger. 1995. "Policy and Non-Policy Determinants of U.S. Equity Foreign Direct Investment." Journal of International Business Studies 26(2), Second Quarter 1995. Magnus, J. R., and A. L. Vasnev. 2007. "Local Sensitivity and Diagnostic Tests." Econometrics Journal (10): 166­92. Mitra, A., A. Varoudakis, and M. Véganzonès-Varoudkis. 2002. "Productivity and Technical Efficiency in Indian States' Manufacturing: The Role of Infrastructure." Economic Development and Cultural Change 2002: 395­426. Mody, A., and K. Srinivasan. 1996. "Japanese and United States Firms as Foreign Investors: Do They March to the Same Tune?" Mimeo, World Bank, Washington, DC. Munnell, A. H. 1990a. "Why Has Productivity Growth Declined? Productivity and Public Investment." New England Economic Review January/February 1990: 3­22. ------. 1990b. "How Does Public Infrastructure Affect Regional Economic Performance?" New England Economic Review September/October 1990: 11­32. ------. 1992. "Infrastructure Investment and Economic Growth." Journal of Economic Perspectives 6(4): 189­98. Ndulu, B. J., and A. O'Connell. 2005. "Sub-Saharan Africa: Growth Econometrics and Country Experiences." Paper prepared for the African Economic Research Consortium (AERC)/Harvard workshop on, "Explaining African Economic Growth," Weatherland Center, Harvard University. Olley, G. S., and A. Pakes. 1996. "The Dynamics of Productivity in the Telecommunications Equipment Industry." Econometrica 64 (6): 1263­97. Prud'Homme, R. 2004. "Infrastructure and Development." Paper presented at the Annual Bank Conference on Development Economics, World Bank, Washington DC. Reinikka, R., and J. Svensson. 1999. "How Inadequate Provision of Public Infrastructure and Services Affect Private Investment?" Policy Research Working Paper 2262, World Bank, Washington, DC. 71 Roller, L. H., and L. Waverman. 2001. "Telecommunications Infrastructure and Economic Development: A Simultaneous Approach." American Economic Review, American Economic Association, 91(4): 909­23. Romp, W., and J., de Haan. 2005. "Public Capital and Economic Growth: a Critical Survey." European Investment Bank Papers 10(1). Sachs, J., J. McArthur, G. Schmidt-Traub, M. Kruk, C. Bahadur, M. Faye, and G. McCord. 2004. "Ending Africa's Poverty Trap." Brookings Papers on Economic Activity, 117­240. Shioji, E. 2001. "Public Capital and Economic Growth: A Convergence Approach." Journal of Economic Growth 6(3): 205­27. Solow, R. 1957. "Technical Change and the Aggregate Production Function." Review of Economics and Statistics 39: 312­20. Straub, S. 2008. "Infrastructure and Growth in Developing Countries: Recent Advances and Research Challenges." Policy Research Working Paper Series 4460, World Bank, Washington, DC. Veeramani, C., and G. Goldar. 2004. "Investment Climate and Total Factor Productivity in Manufacturing: Analysis of Indian States." Working paper #127, Indian Council for Research on International Economic Relations, New Delhi, India. Wilson, J. S., C. L. Mann, and T. Otsuki. 2004. "Assessing the Potential Benefit of Trade Facilitation: A Global Perspective." Working Paper WPS 3224, World Bank, Washington, DC. World Bank. 1994. World Development Report 1994: Infrastructure for Development. New York: Oxford University Press. ------. 2004. Africa's Competitiveness Report 2004. Washington, DC: World Bank. ------. 2006. Doing Business in 2007: How to Reform. Washington, DC: World Bank. ------. 2007. Africa's Competitiveness Report 2007. Washington, DC: World Bank. 72 Appendix: Tables Table A.1 General information on firms and industries and on production function (PF) variables (a) Food and beverages; (b) textiles and apparels; (c) chemicals, rubber, and plastics; (d) paper, Industrial printing, and publishing; (e) machinery and equipment/metallic products; (f) wood and furniture; (g) classification nonmetallic products; and (h) other manufacturing. Small firms: less than 20 employees; medium firms in between 20 and 99 employees; large firms Size classification more than 99 employees. 1) Algeria: Region A, Region B, Region C, Region D 2) Benin: South (coastal), rest of country (rainforest) 3) Botswana: Francistown, Gaborone 4) Burkina Faso: Ouagadougou, rest of country 5) Burundi: Bujumbura 6) Cameroon: Bafoussam, Douala, Yaounde 7) Cape Verde: Mindelo, Praia 8) Eritrea: Eritrea 9) Ethiopia: Addis Ababa, Awasa, Bahir Dar, Dire Adwa, Mekele, Nazareth, Gondar, Adigrat, Harar, Adwa, rest of country 10) Kenya: Nairobi, rest of country General information at 11) Madagascar: Antananarivo, rest of country firm level 12) Malawi: Blantyre, Lilongwe, rest of country Country/Region 13) Mali: Bamako, rest of country classification 14) Mauritania: Noauadhibou, Nouakchott 15) Mauritius: Port Louis, Beau Bassin, Vacoas Phoenix, Curepipe, Quatre Bornes, other 16) Morocco: Settat, Nador, Casablanca, Rabat, Fes, Tanger 17) Namibia: Walvis Bay, Windhoek 18) Niger: Maradi, Niamey 19) Senegal: Dakar, rest of country 20) South Africa: Gauteng, Kwazulu, Natal, Western Cape, Eastern Cape 21) Swaziland: Matsapha, Manzini, Mbabane 22) Tanzania: Dar es Salaam, Kilimanjaro, Tanga/Arusha, Lake Victoria, South, Zanzibar; 23) Uganda: Central, North East, South West 24) Lesotho: Maseru, rest of country 25) Egypt: Cairo, Suez Channel, Qualyubia, Menoufiya, Alexandria, Nile Delta, Sharkiya, Lower Egypt 26) Zambia: Lusaka, Ndola, Kitwe, rest of country Used as the measure of output for the PF estimation. Sales are defined as total annual sales. The Sales series are deflated by using the consumer price index (CPI), base 2000. Employment Total number of permanent (full-time) and temporal (full-time) workers. Total hours worked Total number of employees multiplied by the average hours worked per year. per year PF variables Total costs of intermediate and raw materials used in production (excluding fuel). The series are Materials (productivity) deflated by using the CPI deflator, base 2000. Net book value of machinery and equipment (NBVC); for those countries which the net book value is Capital stock not available it is replaced by the replacement cost of machinery and equipment. The series are deflated by using the CPI deflator, base 2000. The user cost of capital is defined in terms of the opportunity cost of using capital; it is defined as the User cost of capital 15 percent of the value of the capital stock. Labor cost Total expenditures on personnel, deflated by using the CPI deflator, base 2000. Source: IC data. Note: All figures are in U.S. dollars. 73 Table A.2.1 Definition of investment climate (IC) variables; infrastructure Name of the variable Description of the variable Days to clear customs to import Average number of days to clear customs when importing. Longest number of days to clear Longest number of days to clear customs when importing. customs to import Days to clear customs to export Average number of days to clear customs when exporting directly. Longest number of days to clear Longest number of days to clear customs when exporting directly. customs to export Total cost to clear customs for a typical consignment as a percentage of the consignment value (including Cost to clear customs to export payments to clearing agents, storage fees, container handling fees, and gifts or informal payments to customs officials). Inspections in customs Percentage of establishment's exports that were physically inspected during last financial year (LFY). Percentage of the consignment value of the products shipped to be exported that was lost while in transit because Shipment losses in customs to export of breakage or spoilage. Dummy taking value 1 if the plants enjoy a export facility such as customs duty drawback, duty suspension on Dummy for profit from export facilities imported inputs, profit tax exemption, and so on. Cost of exports Percent of the value of export earnings was transport costs. Dummy for public mechanism to cover Dummy taking value 1 if the firm has a public mechanism to cover risk of nonpayment of exported products. risks in exports Dummy for outside clearing agent for Dummy variable taking value 1 if the firms uses an agent to facilitate customs clearance for imports. imports Average number of days to clear an Average time of clearing an outgoing container through a port clear (including preshipment inspection). outgoing container through port Cost to clear an outgoing container Average cost of clearing an outgoing container through a port clear (including preshipment inspection). through port Average number of days to clear an Average time of clearing an incoming container through a port clear (including preshipment inspection). incoming container through port Cost to clear an incoming container Average cost of clearing an incoming container through a port clear (including preshipment inspection). through port Dummy for own power infrastructure Dummy taking value 1 if the firm provides its own power infrastructure, excluding generators. Dummy for own generator Dummy variable taking value 1 if the firm has its own power generator. Electricity from own generator Percentage of the electricity used by the plant provided by a own generator. Cost of electricity from generator Estimated annual cost of generator fuel as percentage of annual sales. Cost of electricity from public grid Average cost per kilowatt-hour (KwH) when using power from the public grid. Dummy for equipment damaged by Dummy taking value 1 if any machine or equipment was damaged by power fluctuations. power fluctuations Equipment damaged by power Value of the losses of machinery and equipment damaged by power fluctuations as a percentage of the net book fluctuations value of machinery and equipment (NBVC). Power outages Total number of power outages suffered by the plant in LFY. Average duration of power outages Average duration of power outages suffered in hours, conditional on the pant reports having power outages. Power fluctuations Total number of power fluctuations suffered by the plant in LFY. Average duration of power fluctuations suffered in hours, conditional on the plant reports having power Average duration of power fluctuations fluctuations. Losses due to power outages as a percentage of total annual sales, conditional on the plant reports having power Sales lost due to power outages outages. Water outages Total number of water outages suffered by the plant in LFY. Average duration of water outages Average duration of water outages suffered in hours, conditional on the plant reports having water outages. Losses due to water outages as a percentage of total annual sales, conditional on the plant reports having power Sales lost due to water outages outages. Dummy for own well or water Dummy taking value 1 if the plant has its own or shared borehole or well, or builds its own water infrastructure. infrastructure Water from own well or water Percentage of firm's water supply from its own or shared well. 74 infrastructure Cost of water from own well Total annual cost of self-provided water as a percentage of total annual sales. Cost of water from public system Unit cost of using water from the public water system. Phone outages Total number of phone outages suffered by the plant in LFY. Average duration of phone outages Average duration of phone outages suffered in hours, conditional on the plant reports having water outages. Losses due to phone outages as a percentage of total annual sales, conditional on the plant reports having power Losses due to phone outages outages. Transport failures Total number of transport failures suffered by the plant in LFY. Average duration of transport failures Average duration of transport failures suffered in hours, conditional on the plant reports having water outages. Losses due to transport failures as a percentage of total annual sales, conditional on the plant reports having Sales lost due to transport failures power outages. Source: IC data. 75 Table A.2.1 Definition of IC variables; infrastructure (cont.) Name of the variable Description of the variable Average duration of transport Time in hours that it takes to ship the inputs transported by road from the point of origin to the establishment. Public postal service interruptions Total number of public postal service interruptions suffered by the plant in LFY. Average duration of public postal service Average duration of public postal service interruptions suffered in hours, conditional on the plant reports having interruptions water outages. Sales lost due to public postal service Losses due to public postal service interruptions as a percentage of total annual sales, conditional on the plant interruptions reports having power outages. Dummy for own roads Dummy taking value 1 if the firm provides its own roads. Dummy for own transportation for Dummy taking value 1 if the firm provides its transportation for workers. workers Dummy for own waste disposal Dummy taking value 1 if the firm provides its own waste disposal. Dummy for contract with transportation Dummy taking value 1 if the firm arranges transport services for the delivery of finished products, or raw company materials by direct contract with transportation company. Dummy taking value 1 if the firm arranges transport services for the delivery of finished products, or raw Dummy for own transportation materials with its own transportation. Products with own transport Percentage of products delivered with firm's own transport. Transport delay, outgoing domestic Percentage of times that transport services are late in picking up sales for domestic markets at the plant for merchandise delivery. Transport delay, outgoing export Percentage of times that transport services are late in picking up sales for exports at the plant for delivery. merchandise Transport delay, incoming domestic Percentage of times that transport services are late in dropping off supplies from domestic sources at the plant merchandise for delivery. Transport delay, incoming export Percentage of times that transport services are late in dropping off direct imports at the plant for delivery. merchandise Percentage of the consignment value of the products shipped for domestic transportation lost while in transit Shipment losses, domestic because of theft, breakage, or spoilage. Percentage of the consignment value of the products shipped for international transportation lost while in transit Shipment losses, exports because of theft, breakage, or spoilage. Dummy for e-mail Dummy variable taking value 1 if the plant mainly uses e-mail to communicate with clients and suppliers. Dummy for Web page Dummy variable taking value 1 if the plant uses its own Web page to communicate with clients and suppliers. Wait for phone connection Number of days waiting to obtain a phone connection. Dummy for gifts to obtain a phone Gifts expected or requested to obtain a phone supply. connection Wait for electric supply Number of days waiting to obtain an electricity supply. Dummy for gifts to obtain a electric Gifts expected or requested to obtain an electrical connection. supply Wait for a water supply Number of days waiting for a water supply. Dummy for gifts to obtain a water supply Gifts expected or requested to obtain a water supply. Wait for an import license Number of days waiting for an import license. Dummy for gifts to obtain an import Gifts expected or requested to obtain an import license. license Low quality supplies Percentage of domestic inputs/supplies that are of lower than agreed-upon quality. Sales lost due to delivery delays, Percentage of domestic sales lost due to delivery delays from suppliers in LFY. domestic Sales lost due to delivery delays, imports Percentage of exports lost due to delivery delays from suppliers in LFY. Transport delays in domestic sales Percentage of domestic sales lost due to delays in transportation services in LFY. Transport delays in international sales Percentage of exports lost due to delays in transportation services in LFY. Amount (as a percentage of total annual sales) spent by a typical establishment in "unofficial payments" for Illegal payments to obtain public utilities public utilities (that is, power, water and sewage, and telephone). Days of inventory of main supply Average number of days (measured in production days) that the main input is available on stock. Days of inventory of finished goods Average number of days (measured in production days) that the main output is available on stock. 76 Source: IC data. Table A.2.2 Definition of IC variables; red tape, corruption, and crime Name of the variable Description of the variable Manager's time spent in bureaucratic issues In typical week percentage of manager's time spent dealing with bureaucratic issues. Payments to deal with bureaucratic issues Total payments as a percentage of total annual sales to "speed up" bureaucratic issues. Amount (as a percentage of total annual sales) spent by a typical establishment in "unofficial payments" for Illegal payments to obtain licenses licenses from government institutions, for example, a city council. Amount (as a percentage of total annual sales) spent by a typical establishment in "unofficial payments" to tax Illegal payments to tax administration administration. Wait for a construction permit Days waiting to obtain a construction permit. Dummy for gifts to obtain a construction Gifts expected or requested to obtain a construction permit. permit Wait for an operating license Days waiting to obtain a main operating license. Gifts to obtain an operating license Gifts expected or requested to obtain a main operating license. Sales declared to taxes Percentage of total annual sales that a typical firm operating in plant's sector reports for tax purposes. Workforce declared to taxes Percentage of total workforce that a typical firm operating in plant's sector reports for tax purposes. Days in inspections Total number of inspections from regulatory agencies received by the plant in LFY. Dummy for gifts in inspections Dummy taking value 1 if any informal gift or payment were requested during inspections from regulatory agencies. Dummy for lawyer/consultant to help deal Dummy taking value 1 if the plant uses/used a lawyer and/or consultant to help obtaining all the permissions and with permissions licenses needed to operate/enter the market. Payments to obtain a contract with the Payments to obtain a contract with the government as a percentage of contract value. government Dummy for law-influencing firm Dummy taking value 1 if the firm seeks to influence local or national laws. Overdue payments to private customers Percentage of total sales to private enterprises that involved overdue payments in LFY. Overdue payments to state-owned Percentage of total sales to government agencies or SOEs that involved overdue payments in LFY. enterprises (SOEs) Weeks to resolve a case of overdue Percentage of overdue payments that required the action of a court to be solved. payment Overdue payments in courts Percentage of total sales to private enterprises that involved overdue payments that were resolved in courts in LFY. Weeks to resolve an overdue payment in Weeks that it takes to resolve a typical case of overdue payment in courts courts Security expenses Security expenses as a percentage of annual total sales. Dummy for security expenses Dummy taking value 1 if the plant has security expenses. Illegal payments in protection Cost in illegal payments to avoid violence, for example to criminal organizations. Dummy for payments in protection Dummy taking value 1 if the plant has cost in illegal payments to avoid violence. Cost to avoid pilferage from workers Cost in illegal payments to reduce pilferage by workers. Dummy for cost to avoid pilferage from Dummy taking value 1 if the plant has costs to reduce pilferage by workers. workers Crime losses Crime losses as a percentage of annual total sales in LFY. Dummy for crime losses Dummy taking value 1 if the plant has experienced losses due to criminal attempts in LFY. Crimes reported to police Percentage of criminal attempts reported to the police. Crimes solved by police Percentage of criminal attempts solved by the police. Days of production lost due to civil unrest Total number of production days lost due to civil unrest during LFY. Days of production lost due to absenteeism Total number of production days lost due to employees absenteeism during LFY. Dummy for tax exemption Dummy variable that takes value 1 if the labor regulation has affected plant's employment decisions. Dummy for lawsuit in the last 3 years Dummy taking value 1 if the plant had any lawsuit during the last 3 years 77 Dummy for "gifts" for credit Dummy if the firm had to offer a gift or an informal payment to get a credit. Dummy for interventionist labor regulation Dummy taking value 1 if plant's decisions on hiring and/or firing workers have been influenced by labor regulations. Total days spent with licenses Total number of days that were spent dealing with licenses LFY. Dummy for accountant to accomplish taxes Dummy if the firm uses an accountant or consultant to accomplish taxes. Dummy for gifts to tax inspectors Dummy if the firm had to offer a gift or an informal payment to tax inspectors. Gifts to tax inspectors Amount (as a percentage of total annual sales) paid to tax inspectors in gifts and/or irregular payments. Dummy for labor conflicts Dummy taking value 1 if the firm had any conflict with employees during LFY. Average time to hire a skilled worker Average days that it takes to hire a skilled production worker. Dummy for conflicts with suppliers Dummy taking value 1 if the firm had any conflict with suppliers during LFY. Dummy for conflicts with clients Dummy taking value 1 if the firm had any conflict with clients during LFY. Cost of entry Cost of entry to the market in terms of licenses and permissions needed. Dummy for consultant to help deal with Dummy taking value 1 if the firm uses consultants and/or lawyers to help deal with licenses and permissions. permissions Source: IC data. 78 Table A.2.3 Definition of IC variables; finance and corporate governance Name of the variable Description of the variable Dummy for trade chamber Dummy taking value 1 if the firm belongs to a trade chamber or association. Dummy for credit line Dummy that takes value 1 if the firm has access to a credit line or overdraft facility. Credit unused Percentage of the overdraft that is not being used currently. Dummy for loan Dummy that takes value 1 if the firm has access to a loan line. Dummy for loan with collateral Dummy that takes value 1 if the firm has access to a loan line with collateral (conditional on having a loan line). Value of the collateral Value of the collateral as a percentage of the loan value (conditional on having a loan with collateral). Interest rate of the loan The interest rate applied to the last loan. Dummy for short-term loan Duration of the loan in years. Borrowings in foreign currency Percentage of firm's borrows denominated in a foreign currency. Dummy for external auditory Dummy that takes value 1 if the firm has its annual statements externally audited. Owner of the lands Percentage of the lands in which the plant operates owned by the firm. Owner of the buildings Percentage of the buildings in which the plant operates owned by the firm. Dummy for owner of the buildings Dummy taking value 1 if the almost all the buildings in which the plant operates are owned by the firm. Dummy for owner of the buildings and Dummy taking value 1 if the almost all the lands in which the plant operates are owned by the firm. lands Largest shareholder Percentage of firm's capital owned by the largest shareholder. Working capital financed by internal Percentage of firm's working capital financed by internal funds. funds Working capital financed by commercial Percentage of firm's working capital financed by funds from private domestic banks. banks Working capital financed by foreign Percentage of firm's working capital financed by funds from foreign banks. commercial banks Working capital financed by leasing Percentage of firm's working capital financed by leasing. Working capital financed by state Percentage of firm's working capital financed by funds from state services (for example, Brazilian Development services Bank, BNDES; Mexican labor and income generation program, PROGER; and so on). Working capital financed by supplier or Percentage of firm's working capital financed by trade credit (supplier or customer credit). customer credit Working capital financed by credit cards Percentage of firm's working capital financed by credit card. Working capital financed by equity Percentage of firm's working capital financed by equity, sale of stock. Working capital financed by family/friends Percentage of firm's working capital financed by funds from family or friends. Working capital financed by informal Percentage of firm's working capital financed by funds from informal sources (for example, money lender). sources Working capital financed by other funds Percentage of firm's working capital financed by other funds. New investments financed by internal Percentage of new investments in new lands, buildings, or machinery financed by internal funds. funds New investments financed by commercial Percentage of new investments in new lands, buildings, or machinery financed by funds from private domestic banks banks. New investments financed by foreign Percentage of new investments in new lands, buildings, or machinery financed by funds from foreign banks. commercial banks New investments financed by leasing Percentage of new investments in new lands, buildings, or machinery financed by leasing. New investments financed by state Percentage of new investments in new lands, buildings, or machinery financed by funds from state services (for services example, BNDES, PROGER, and so on). New investments financed by supplier or Percentage of new investments in new lands, buildings, or machinery financed by trade credit (supplier or customer credit customer credit). New investments financed by credit cards Percentage of new investments in new lands, buildings, or machinery financed by credit card. New investments financed by equity Percentage of new investments in new lands, buildings, or machinery financed by equity, sale of stock. New investments financed by Percentage of new investments in new lands, buildings, or machinery financed by funds from family or friends. family/friends 79 New investments financed by informal Percentage of new investments in new lands, buildings, or machinery financed by funds from informal sources (for sources example money lender). New investments financed by other funds Percentage of new investments in new lands, buildings, or machinery financed by other funds. Share of net profits reinvested Share of net profits reinvested in the firm in the LFY. Sales bought on credit Percentage of establishment's inputs that were purchased on credit in LFY. Dummy for inputs bought on credit Days that it takes for the establishment to pay off the supply credit. Inputs bought on credit Percentage of establishment's total sales that were bought on credit during LFY. Source: IC data. 80 Table A.2.3 Definition of IC variables; finance and corporate governance (cont.) Name of the variable Description of the variable Time to pay off the credit for inputs Average days that it takes to pay off the credits. Inputs bought on credit with delayed Share of inputs bought on credit. payment Wait to clear a check Total number of days needed on average to clear a check from the establishment's financial institution. Charges to clear a check Average fee charged for a check. Total number of days needed on average to clear a domestic currency wire from the establishment's financial Wait to clear a domestic currency wire institution. Charges to clear a domestic currency Average fee charged for a domestic currency wire. wire Total number of days needed on average to clear a foreign currency wire from the establishment's financial Wait to clear a foreign currency wire institution. Charges to clear a foreign currency wire Average fee charged for a foreign currency wire. Wait to clear a letter of credit Total number of days needed on average to clear a letter of credit from the establishment's financial institution. Charge to clear a letter of credit Average fee charged for a letter of credit. Delay of payments of domestic clients Total number of days needed on average to clear a payment from a domestic customer. Charges to get payments from domestic Average fee charged to clear a payment of a domestic customer. clients Delay of payments of foreign clients Total number of days needed on average to clear a payment from a foreign customer. Charges to get payments from foreign Average fee charged to clear a payment of a foreign customer. clients Dummy for current or saving account Dummy taking value 1 if the firm has a checking or saving account. Dummy for foreign current or saving Dummy taking value 1 if the firm has a foreign checking or saving account. account Dummy for accountant Dummy taking value 1 if the firm uses an accountant to finish annual statements. Source: IC data. 81 Table A.2.4 Definition of IC variables; quality, innovation, and labor skills Name of the variable Description of the variable Dummy for foreign technology Dummy taking value 1 if the plant uses technology licensed from a foreign-owned company. Dummy for International Organization for Dummy taking value 1 if the firm has any kind of quality certification. Standardization (ISO) quality certification Sales with warranty Percentage of sales bought with warranty. Dummy for new product Dummy taking value 1 if the firm developed a major new product line during LFY. Dummy for product improvement Dummy taking value 1 if the firm improved an existing product line during LFY. Dummy for discontinued product line Dummy taking value 1 if the firm discontinued at least one product line during LFY. Dummy for equipment improvement Dummy taking value 1 if the firm improved the equipment during LFY. Dummy for R&D Dummy taking value 1 if the firm had expenses in R&D during LFY. R&D expenditures R&D expenditures as a percentage of annual total sales. Workers engaged in design/R&D Percentage of workers in staff engaged in R&D and design tasks. Dummy for subcontracted R&D Dummy taking value 1 if the firm subcontracted R&D activities during LFY. Royalties expenditures Total expenses in royalties as a percentage of total annual sales. Dummy taking value 1 if the firm introduced a new technology that substantially changed the way that the main Dummy for new technology product is produced. Dummy for joint venture Dummy taking value 1 if the firm agreed a new joint venture with foreign partner during LFY. Dummy for new license agreement Dummy taking value 1 if the firm obtained a new license agreement during LFY. Dummy taking value 1 if the firm outsourced a major production activity that was previously conducted in-house Dummy for outsourcing during LFY. Dummy taking value 1 if the firm brought in-house a major production activity that was previously outsourced during Dummy for in-house production LFY. Dummy for new plant Dummy taking value 1 if the firm opened a new plant during LFY. Dummy for closed plant Dummy taking value 1 if the firm closed an existing plant during LFY. Staff--management Percentage of management in staff. Staff--professional workers Percentage of professional production workers in staff. Staff--skilled workers Percentage of skilled production workers in staff. Staff--unskilled workers Percentage of unskilled production workers in staff. Staff--nonproduction workers Percentage of nonproduction workers in staff. Staff--foreign nationals Percentage of foreign national workers in staff. Average education of staff Average number of years of education of staff. Average tenure of staff Average number of years of experience of staff. Average age of staff Average age of staff. Dummy for training Dummy taking value 1 if the firm provides formal (either internal or external) training to its employees. Training to skilled workers Percentage of skilled workers receiving formal (either internal or external) training. Training to unskilled workers Percentage of unskilled workers receiving formal (either internal or external) training. Training to production workers Percentage of production workers receiving formal (either internal or external) training. 82 Training to nonproduction Percentage of nonproduction workers receiving formal (either internal or external) training. Weeks of training for skilled workers Weeks of training received by skilled workers. Weeks of training for unskilled workers Weeks of training received by unskilled workers. Workforce with computer Percentage of workforce using a computer at job. University staff Percentage of staff with at least 1 year of university education. Dummy for university staff Percentage of staff that regularly uses computer at job. Manager education Dummy taking value 1 if the manager of the establishment has a bachelor degree or higher education level. Manager's experience Years of experience of the manager in the same industry before joining the establishment. Source: IC data. 83 Table A.2.5 Definition of variables; other control variables Name of the variable Description of the variable Age Age of the firm. Dummy for incorporated company Dummy that takes value 1 if the firm is an incorporated company. Dummy for limited company Dummy that takes value 1 if the firm is a limited company. Dummy for SOE Dummy variable that takes value 1 if the plant is a SOE. Dummy for foreign direct investment Dummy that takes value 1 if any part of firm's capital is foreign. (FDI) Dummy for holdings Dummy variable that takes value 1 if the firm has holdings or operations in other countries. Share of the local market Percentage of local market that is made up by the sales of the establishment. Share of the national market Percentage of national market that is made up by the sales of the establishment. Dummy for direct exports Dummy taking value 1 if the firm exports more than 10% of the total annual sales. Share of exports Share of exports over total annual sales. Exporting experience Number of years of exporting experience. Dummy for direct imports Dummy taking value 1 if the firm imports more than 10% of the total purchases of intermediate materials. Share of imports Share of imported inputs over total purchases of intermediate materials. Number of competitors Total number of competitors in the local market of the establishment's main product line. Capacity utilization Percentage of capacity utilized. Trade union Percentage of workforce unionized Dummy for privatized firm Dummy variable that takes value 1 if the firm was previously state-owned. Dummy for industrial zone Dummy variable that takes value 1 if the firm is located in an industrial zone. Days of production lost due to strikes Total number of production days lost due to strikes. Dummy for small firm Dummy taking value 1 if the firm has less than 20 employees. Dummy for medium firm Dummy taking value 1 if the firm has in between 20 and 100 employees. Dummy for large firm Dummy taking value 1 if the firm has more than 100 employees. Workers infected by HIV Percentage of workers infected by HIV/AIDS. Dummy variable that takes value 1 if the HIV/AIDS epidemic has negatively affected the firm through absenteeism Dummy for negative impact of HIV of workers or high staff turnover. Cost in HIV-prevention programs Medical expenses for staff (HIV/AIDS related) as percentage of total sales. Source: IC data. 84 Table B.1 Total number of observations available for the PF variables before and after cleaning missing values and outliers Percentage over total number of observations in parentheses Northern Africa Western Africa--Economic Community of West African States (ECOWAS) Horn of Africa DZA EGY MAR SEN BEN MLI MRT BFA CPV NER CMR ETH ERI Total number of observations 952 2,931 2,550 783 591 462 80 51 47 64 119 1281 237 a) Before cleaning Missing observations 605 (63.5) 1,543 (52.6) 95 (3.73) 513 (65.5) 199 (33.6) 211 (45.6) 1 (1.25) 1 (1.96) 0 49 (76.5) 2 (1.68) 150 (11.7) 171 (72.1) of which: firms with one PF variable missing 419 (44.0) 1,009 (34.4) 29 (1.14) 189 (24.1) 146 (24.7) 39 (8.44) 0 1 (1.96) 0 11 (17.1) 1 (0.84) 33 (2.58) 88 (37.1) firms with two PF variables missing 0 34 (1.16) 1 (0.04) 88 (11.2) 18 (3.05) 25 (5.41) 1 (1.25) 0 0 2 (3.13) 0 (0.00) 9 (0.70) 2 (0.84) firms with three PF variables missing 0 319 (10.8) 2 (0.08) 57 (7.28) 8 (1.35) 18 (3.90) 0 0 0 25 (39.0) 0 (0.00) 7 (0.55) 30 (12.6) firms with four PF variables missing 186 (19.5) 181 (6.18) 63 (2.47) 179 (22.8) 27 (4.57) 129 (27.9) 0 0 0 11 (17.1) 1 (0.84) 101 (7.88) 51 (21.5) Outliers 62 (6.51) 131 (4.47) 103 (4.04) 29 (3.70) 42 (7.11) 10 (2.16) 0 0 0 1 (1.56) 0 83 (6.48) 10 (4.22) of which: outliers in materials 24 (2.52) 78 (2.66) 69 (2.71) 23 (2.94) 31 (5.25) 5 (1.08) 0 0 0 0 0 83 (6.48) 4 (1.69) outliers in labor cost 21 (2.21) 33 (1.13) 18 (0.71) 3 (0.38) 4 (0.68) 3 (0.65) 0 0 0 1 (1.56) 0 0 4 (1.69) outliers in both materials and labor cost 17 (1.79) 20 (0.68) 16 (0.63) 3 (0.38) 7 (1.18) 2 (0.43) 0 0 0 0 0 0 2 (0.84) Available observations after replacing 316 (33.1) 1,317 (44.9) 2,352 (92.2) 253 (32.3) 364 (61.5) 242 (52.3) 79 (98.7) 50 (98.0) 47 (100.) 14 (21.8) 117 (98.3) 1,048 (81.8) 61 (25.7) (outliers and missing excluded) b) After cleaning Missing observations 198 (20.8) 225 (7.68) 71 (2.78) 179 (22.8) 42 (7.11) 129 (27.9) 0 0 0 11 (17.1) 1 (0.84) 101 (7.88) 51 (21.5) of which: firms with one PF variable missing 12 (1.26) 9 (0.31) 8 (0.31) 0 9 (1.52) 0 0 0 0 0 (0.00) 0 (0.00) 0 (0.00) 0 (0.00) firms with two PF variables missing 0 0 0 0 2 (0.34) 0 0 0 0 0 (0.00) 0 (0.00) 0 (0.00) 0 (0.00) firms with three PF variables missing 0 34 (1.16) 0 0 1 (0.17) 0 0 0 0 0 (0.00) 0 (0.00) 0 (0.00) 0 (0.00) firms with four PF variables missing 186 (19.5) 182 (6.21) 63 (2.47) 179 (22.8) 30 (5.08) 129 (27.9) 0 0 0 11 (17.1) 1 (0.84) 101 (7.88) 51 (21.5) Outliers 60 (6.30) 82 (2.80) 65 (2.55) 69 (8.81) 77 (13.0) 24 (5.19) 0 0 0 5 (7.81) 0 38 (2.97) 7 (2.95) of which: outliers in materials 16 (1.68) 46 (1.57) 35 (1.37) 48 (6.13) 58 (9.81) 22 (4.76) 0 0 0 4 (6.25) 0 38 (2.97) 2 (0.84) outliers in labor cost 18 (1.89) 10 (0.34) 14 (0.55) 12 (1.53) 8 (1.35) 0 (0.00) 0 0 0 1 (1.56) 0 0 (0.00) 3 (1.27) outliers in both materials and labor cost 26 (2.73) 26 (0.89) 16 (0.63) 9 (1.15) 11 (1.86) 2 (0.43) 0 0 0 0 (0.00) 0 0 (0.00) 2 (0.84) Available observations after replacing 706 (74.1) 2,629 (89.7) 2,422 (94.9) 535 (68.3) 475 (80.3) 309 (66.8) 80 (100) 51 (100.) 47 (100) 48 (75.0) 118 (99.1) 1,142 (89.1) 179 (75.5) (outliers and missing excluded) Source: IC data. Note: The PF variables are: sales, materials, capital stock, and labor cost; the total number of hours worked per year are not included here. For the countries with panel data, the total number of observations is equal to the total number of firms surveyed, multiplied by the total number of years. For the countries with cross-sectional data the total number of observations is equal to the total number of firms surveyed. Outliers are defined as those observations with the ratio of materials to sales and/or labor cost to sales greater than 1. By useful observations we mean those observations available to run regression and to make statistical inference. Missing observations and/or outliers in sales, materials, or labor cost are therefore not initially considered useful available observations. 85 Table B.1 (cont.) Total number of observations available for the PF variables before and after cleaning missing values and outliers Percentage over total number of observations in parentheses) Eastern Africa--East African Southern Africa--Southern African Development Community (SADC incl. Burundi) Community (EAC excl. Burundi) MUS ZAF KEN UGA TZA MWI MDG ZMB BDI BWA LSO NAM SWZ Total number of observations 852 900 828 320 870 564 102 114 225 106 70 636 1737 a) Before cleaning Missing observations 426 (50.0) 652 (72.4) 457 (55.1) 106 (33.1) 456 (52.4) 153 (27.1) 0 (0.00) 4 (3.51) 187 (83.1) 5 (4.72) 3 (4.28) 340 (53.4) 487 (28.0) of which: firms with one PF variable missing 112 (13.1) 288 (32.0) 189 (22.8) 76 (23.7) 184 (21.1) 26 (4.61) 0 3 (2.63) 38 (16.8) 5 (4.72) 2 (2.85) 117 (18.4) 241 (13.8) firms with two PF variables missing 48 (5.63) 40 (4.44) 75 (9.06) 8 (2.50) 62 (7.13) 0 0 0 7 (3.11) 0 1 (1.42) 37 (5.82) 37 (2.13) firms with three PF variables missing 62 (7.28) 95 (10.5) 32 (3.86) 0 (0.00) 30 (3.45) 6 (1.06) 0 0 12 (5.33) 0 0 13 (2.04) 11 (0.63) firms with four PF variables missing 204 (23.9) 229 (25.4) 161 (19.4) 22 (6.88) 180 (20.6) 121 (21.4) 0 1 (0.88) 130 (57.7) 0 0 173 (27.2) 198 (11.4) Outliers 53 (6.22) 41 (4.56) 55 (6.64) 10 (3.13) 40 (4.60) 20 (3.55) 2 (1.96) 1 (0.88) 6 (2.67) 1 (0.94) 0 28 (4.40) 34 (1.96) of which: outliers in materials 46 (5.40) 19 (2.11) 25 (3.02) 9 (2.81) 20 (2.30) 18 (3.19) 2 (1.96) 1 (0.88) 0 (0.00) 1 (0.94) 0 9 (1.42) 12 (0.69) outliers in labor cost 4 (0.47) 16 (1.78) 19 (2.29) 1 (0.31) 17 (1.95) 2 (0.35) 0 (0.00) 0 (0.00) 5 (2.22) 0 0 14 (2.20) 14 (0.81) outliers in both materials and labor cost 3 (0.35) 6 (0.67) 11 (1.33) 0 (0.00) 3 (0.34) 0 (0.00) 0 (0.00) 0 (0.00) 1 (0.44) 0 0 5 (0.79) 8 (0.46) Available observations after replacing 377 (44.2) 232 (25.7) 325 (39.2) 208 (65.0) 383 (44.0) 391 (69.3) 109 (106.) 100 (87.7) 37 (16.4) 100 (94.3) 67 (95.7) 271 (42.6) 1,229 (70.7) (outliers and missing excluded) b) After cleaning Missing observations 205 (24.0) 234 (26.0) 164 (19.8) 22 (6.88) 181 (20.8) 122 (21.6) 0 (0.00) 1 (0.88) 131 (58.2) 0 (0.00) 1 (1.42) 174 (27.3) 199 (11.4) of which: firms with one PF variable missing 0 5 (0.56) 3 (0.36) 0 1 (0.11) 0 0 0 1 (0.44) 0 0 1 (0.16) 1 (0.06) firms with two PF variables missing 0 0 0 0 0 0 0 0 0 0 0 0 0 firms with three PF variables missing 1 (0.12) 0 0 0 0 0 0 0 0 0 0 0 0 firms with four PF variables missing 204 (23.9) 229 (25.4) 161 (19.4) 22 (6.88) 180 (20.6) 122 (21.6) 0 (0.00) 1 (0.88) 130 (57.7) 0 (0.00) 1 (1.42) 173 (27.2) 198 (11.4) Outliers 70 (8.22) 35 (3.89) 106 (12.8) 10 (3.13) 66 (7.59) 25 (4.43) 1 (0.98) 1 (0.88) 16 (7.11) 2 (1.89) 0 46 (7.23) 47 (2.71) of which: outliers in materials 64 (7.51) 13 (1.44) 74 (8.94) 4 (1.25) 35 (4.02) 25 (4.43) 0 (0.00) 1 (0.88) 2 (0.89) 1 (0.94) 0 28 (4.40) 18 (1.04) outliers in labor cost 2 (0.23) 14 (1.56) 12 (1.45) 4 (1.25) 22 (2.53) 0 1 (0.98) 0 6 (2.67) 1 (0.94) 0 11 (1.73) 13 (0.75) outliers in both materials and labor cost 4 (0.47) 8 (0.89) 20 (2.42) 2 (0.63) 9 (1.03) 0 0 0 8 (3.56) 0 (0.00) 0 7 (1.10) 16 (0.92) Available observations after replacing 577 (67.7) 635 (70.5) 561 (67.7) 288 (90.0) 623 (71.6) 417 (73.9) 101 (99.0) 112 (98.2) 79 (35.1) 104 (98.1) 69 (98.5) 417 (65.5) 1,492 (85.9) (outliers and missing excluded) Source: IC data. Note: As for previous table. 86 Table B.2.1 Representativity of PF variables before and after cleaning missing values and outliers, by country and year 1999 2000 2001 2002 2003 2004 2005 Total Perc. #Obs #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs available Original sample 952 952 1,904 Algeria Without replacing 552 42.0 562 41.0 1,114 With replacing 700 26.5 712 25.2 1,412 Original sample 197 197 197 591 Benin Without replacing 112 43.1 123 37.6 129 34.5 364 With replacing 143 27.4 164 16.8 168 14.7 475 Original sample 114 114 Botswana Without replacing 109 4.4 109 With replacing 113 0.9 113 Original sample 51 51 Burkina Faso Without replacing 50 2.0 50 With replacing 51 0.0 51 Original sample 102 102 Burundi Without replacing 100 2.0 100 With replacing 101 1.0 101 Original sample 119 119 Cameroon Without replacing 117 1.7 117 With replacing 118 0.8 118 Original sample 47 47 Cape Verde Without replacing 47 0.0 47 With replacing 47 0.0 47 Original sample 977 977 977 2,931 Egypt Without replacing 631 35.4 686 0 100 1,317 With replacing 795 18.6 902 932 4.6 2,629 Original sample 79 79 79 237 Eritrea Without replacing 0 100 38 51.9 23 70.9 61 With replacing 50 36.7 62 21.5 67 15.2 179 Original sample 427 427 427 1,281 Ethiopia Without replacing 316 26.0 344 19.4 388 9.1 1,048 With replacing 351 17.8 377 11.7 414 3.0 1,142 Original sample 284 284 284 852 Kenya Without replacing 110 61.3 119 58.1 131 53.9 360 With replacing 185 34.9 185 34.9 215 24.3 585 Original sample 75 75 75 225 Lesotho Without replacing 9 88.0 12 84.0 16 78.7 37 With replacing 20 73.3 26 65.3 33 56.0 79 Source: Authors' elaboration with IC data. Note: Original sample includes all establishments surveyed. Without replacing includes establishments without missing values and/or outliers in PF variables. With replacing includes establishments without missing values and/or outliers in the PF variables. 87 Table B.2.1 (cont.) Representativity of PF variables before and after cleaning missing values and outliers, by country and year 1999 2000 2001 2002 2003 2004 2005 Total Perc. #Obs #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs available Original sample 290 290 290 870 Madagascar Without replacing 113 61.0 134 53.8 136 53.1 383 With replacing 183 36.9 212 26.9 228 21.4 623 Original sample 160 160 320 Malawi Without replacing 93 41.9 115 28.1 208 With replacing 136 15.0 152 5.0 288 Original sample 154 154 154 462 Mali Without replacing 62 59.7 78 49.4 102 33.8 242 With replacing 74 51.9 93 39.6 142 7.8 309 Original sample 80 80 Mauritania Without replacing 79 1.3 79 With replacing 80 0.0 80 Original sample 212 212 212 636 Mauritius Without replacing 77 63.7 97 54.2 97 54.2 271 With replacing 122 42.5 142 33.0 153 27.8 417 Original sample 850 850 850 2,550 Morocco Without replacing 754 11.3 794 6.6 804 5.4 2,352 With replacing 780 8.2 813 4.4 829 2.5 2,422 Original sample 106 106 Namibia Without replacing 100 5.7 100 With replacing 104 1.9 104 Original sample 64 64 Niger Without replacing 14 78.1 14 With replacing 48 25.0 48 Original sample 261 261 261 783 Senegal Without replacing 59 77.4 84 67.8 110 57.9 253 With replacing 135 48.3 183 29.9 217 16.9 535 Original sample 579 579 579 1,737 South Africa Without replacing 373 35.6 406 29.9 450 22.3 1,229 With replacing 457 21.1 498 14.0 537 7.3 1,492 Original sample 70 70 Swaziland Without replacing 67 4.3 67 With replacing 69 1.4 69 Original sample 276 276 276 828 Tanzania Without replacing 113 59.1 124 55.1 88 68.1 325 With replacing 193 30.1 205 25.7 163 40.9 561 Original sample 300 300 300 900 Uganda Without replacing 102 66.0 112 62.7 154 48.7 368 With replacing 169 43.7 249 17.0 277 7.7 695 Original sample 188 188 188 0 564 Zambia Without replacing 114 39.4 127 32.4 150 20.2 0 391 With replacing 126 33.0 136 27.7 155 17.6 0 417 Source: Author's elaboration with IC data. Note: As for first part of table. 88 Table B.2.2 Representativity of PF variables before and after cleaning missing values and outliers, by country and industry Mach and Chemicals, rubber, Paper, edition, and Nonmetallic Food and beverages Textiles and apparels equipment/metallic Wood and furniture Other manufacturing and plastics publishing products products Perc. Country #Obs #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. available Original sample 204 372 404 308 440 144 32 Algeria Without replacing 114 44.1 200 46.2 280 30.7 162 47.4 256 41.8 98 31.9 4 87.5 With replacing 174 14.7 258 30.6 332 17.8 204 33.8 320 27.3 114 20.8 10 68.8 Original sample 120 36 135 66 189 45 Benin Without replacing 71 40.8 29 19.4 75 44.4 37 43.9 125 33.9 27 40.0 With replacing 98 18.3 34 5.6 110 18.5 48 27.3 147 22.2 39 13.3 Original sample 12 27 0 75 Botswana Without replacing 12 0.0 26 3.7 0 71 5.3 With replacing 12 0.0 27 0.0 0 74 1.3 Original sample 14 12 25 Burkina Faso Without replacing 13 7.1 12 0.0 25 0.0 With replacing 14 0.0 12 0.0 25 0.0 Original sample 19 24 59 Burundi Without replacing 18 5.3 24 0.0 58 1.7 With replacing 19 0.0 24 0.0 58 1.7 Original sample 31 17 19 11 18 23 Cameroon Without replacing 31 0.0 17 0.0 18 5.3 11 0.0 18 0.0 22 4.3 With replacing 31 0.0 17 0.0 18 5.3 11 0.0 18 0.0 23 0.0 Original sample 12 16 19 Cape Verde Without replacing 12 0.0 16 0.0 19 0.0 With replacing 12 0.0 16 0.0 19 0.0 Original sample 468 915 453 672 174 249 Egypt Without replacing 225 51.9 393 57.0 219 51.7 303 54.9 67 61.5 110 55.8 With replacing 416 11.1 815 10.9 414 8.6 602 10.4 152 12.6 230 7.6 Original sample 54 51 18 114 Eritrea Without replacing 14 74.1 11 78.4 8 55.6 28 75.4 With replacing 38 29.6 39 23.5 15 16.7 87 23.7 Original sample 285 279 618 99 Ethiopia Without replacing 233 18.2 207 25.8 531 14.1 77 22.2 With replacing 258 9.5 240 14.0 557 9.9 87 12.1 Original sample 249 141 144 147 171 Kenya Without replacing 99 60.2 69 51.1 62 56.9 57 61.2 73 57.3 With replacing 172 30.9 95 32.6 97 32.6 91 38.1 130 24.0 Original sample 54 102 69 Lesotho Without replacing 17 68.5 8 92.2 12 82.6 With replacing 31 42.6 24 76.5 24 65.2 Source: Author's elaboration with IC data. Note: As for previous table. 89 Table B.2.2 Representativity of PF variables before and after cleaning missing values and outliers, by country and industry (cont.) Mach and Textiles and Chemicals, rubber, Paper, edition, and Nonmetallic Food and beverages equipment/metallic Wood and furniture Other manufacturing apparels and plastics publishing products products Perc. Country #Obs #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. #Obs Perc. available Original sample 150 267 108 93 60 192 Madagascar Without replacing 77 48.7 106 60.3 57 47.2 51 45.2 24 60.0 68 64.6 With replacing 110 26.7 175 34.5 76 29.6 76 18.3 44 26.7 142 26.0 Original sample 112 70 42 48 48 Malawi Without replacing 72 35.7 52 25.7 30 28.6 27 43.8 27 43.8 With replacing 101 9.8 66 5.7 42 0.0 40 16.7 39 18.8 Original sample 153 30 69 33 66 54 57 Mali Without replacing 82 46.4 8 73.3 29 58.0 19 42.4 47 28.8 23 57.4 34 40.4 With replacing 97 36.6 14 53.3 47 31.9 27 18.2 56 15.2 30 44.4 38 33.3 Original sample 27 12 13 28 Mauritania Without replacing 26 3.7 12 0.0 13 0.0 28 0.0 With replacing 27 0.0 12 0.0 13 0.0 28 0.0 Original sample 117 219 72 54 93 33 18 30 Mauritius Without replacing 53 54.7 97 55.7 32 55.6 32 40.7 29 68.8 20 39.4 8 55.6 0 100.0 With replacing 86 26.5 139 36.5 47 34.7 50 7.4 63 32.3 23 30.3 9 50.0 0 100.0 Original sample 216 1,722 414 147 51 Morocco Without replacing 196 9.3 1,584 8.0 383 7.5 140 4.8 49 3.9 With replacing 205 5.1 1,635 5.1 390 5.8 142 3.4 50 2.0 Original sample 18 5 83 Namibia Without replacing 18 0.0 5 0.0 77 7.2 With replacing 18 0.0 5 0.0 81 2.4 Original sample 18 14 32 Niger Without replacing 6 66.7 0 100.0 8 75.0 With replacing 12 33.3 12 14.3 24 25.0 Original sample 279 69 147 108 75 48 57 Senegal Without replacing 78 72.0 20 71.0 55 62.6 48 55.6 19 74.7 15 68.8 18 68.4 With replacing 186 33.3 46 33.3 106 27.9 73 32.4 49 34.7 29 39.6 45 21.1 Original sample 189 180 285 159 561 147 66 150 South Africa Without replacing 131 30.7 107 40.6 187 34.4 120 24.5 435 22.5 102 30.6 43 34.8 104 30.7 With replacing 162 14.3 144 20.0 241 15.4 137 13.8 498 11.2 131 10.9 50 24.2 129 14.0 Original sample 14 20 36 Swaziland Without replacing 12 14.3 19 5.0 36 0.0 With replacing 13 7.1 20 0.0 36 0.0 Original sample 243 93 102 75 87 195 33 Tanzania Without replacing 108 55.6 29 68.8 42 58.8 33 56.0 26 70.1 68 65.1 19 42.4 With replacing 168 30.9 58 37.6 69 32.4 55 26.7 65 25.3 117 40.0 27 18.2 Original sample 366 45 75 69 63 162 120 Uganda Without replacing 148 59.6 22 51.1 17 77.3 19 72.5 33 47.6 74 54.3 55 54.2 With replacing 292 20.2 37 17.8 58 22.7 44 36.2 53 15.9 120 25.9 91 24.2 Original sample 273 69 63 75 84 Zambia Without replacing 188 31.1 54 21.7 44 30.2 52 30.7 53 36.9 With replacing 201 26.4 58 15.9 50 20.6 54 28.0 54 35.7 Source: Authors' elaboration with IC data. Note: As for first part of table. 90 Table B.3.1 Response rate of infrastructure IC variables in the final sample Horn of Eastern Africa-- Northern Africa Western Africa--ECOWAS Southern Africa--SADC (incl. Burundi) Africa EAC MUS ZAF DZA EGY MAR SEN BEN MLI MRT BFA CPV NER CMR ETH ERI KEN UGA TZA MWI MDG ZMB BDI BWA LSO NAM SWZ Days to clear customs to import 53.6 23.7 70.5 58.7 32.1 41.1 47.5 41.2 51.1 60.0 52.9 23.7 73.7 61.1 23.3 66.3 64.5 37.6 74.1 76.0 67.0 48.6 65.2 69.8 Longest number of days to clear customs 52.3 23.3 70.4 57.3 32.1 41.7 47.5 41.2 51.1 60.0 52.9 23.4 72.6 35.6 22.4 64.5 37.6 73.2 74.7 66.0 48.6 63.8 69.1 to import Days to clear customs to export 4.0 18.7 58.4 21.3 17.0 15.9 12.5 23.5 4.3 16.3 26.9 6.3 12.8 50.0 41.9 67.0 31.7 1.0 16.1 42.7 23.6 37.1 55.4 59.1 Longest number of days to clear customs 4.0 18.3 58.4 21.3 17.0 15.9 12.5 23.5 4.3 16.3 26.9 6.3 12.3 18.0 40.8 23.3 31.7 1.0 16.1 40.0 23.6 37.1 55.4 58.4 to export Cost to clear customs to export 99.4 12.5 19.6 2.1 7.5 26.9 1.0 13.4 18.9 37.1 Inspections in customs 62.7 13.8 25.5 4.3 80.0 39.5 2.0 14.3 20.8 35.7 Shipment losses in customs to export 13.8 25.5 4.3 16.3 37.8 2.0 17.0 24.5 37.1 Dummy for profit from export facilities 47.2 21.2 16.3 25.5 4.3 17.5 39.5 5.2 12.8 57.7 18.9 26.1 23.3 28.9 65.9 Cost of exports 99.5 91 Dummy for public mechanism to cover 47.0 21.8 20.1 43.6 24.7 risks in exports Dummy for outside clearing agent for 65.0 34.9 20.7 100.0 39.2 51.1 61.3 53.8 59.1 26.6 98.0 43.7 66.4 2.0 17.0 24.5 37.1 68.6 imports Average number of days to clear an 46.9 28.9 12.3 29.4 outgoing container through port Cost to clear an outgoing container through 22.7 10.6 port Average number of days to clear an 55.1 22.5 42.5 incoming container through port Cost to clear an incoming container 34.6 20.0 through port Dummy for own power infrastructure (excl. 99.2 99.0 98.0 100.0 generator) Dummy for own generator 98.1 99.9 99.6 99.2 100.0 95.5 100.0 98.0 100.0 80.0 100.0 99.7 100.0 97.4 100.0 96.6 100.0 100.0 100.0 100.0 100.0 92.0 98.1 98.6 Electricity from own generator 98.9 99.8 99.1 89.7 84.5 97.4 100.0 96.1 100.0 75.0 95.8 12.1 97.2 93.5 99.5 0.0 93.4 99.0 98.8 100.0 98.2 82.7 96.2 97.1 Cost of electricity from generator 37.3 24.4 39.8 93.5 89.4 60.1 32.9 76.7 18.8 95.0 92 Cost of electricity from public grid 73.3 93.2 2.8 22.9 89.7 88.0 58.7 Dummy for equipment damaged by power 97.9 95.6 97.4 95.4 95.3 98.7 fluctuations Equipment damaged by power fluctuations 83.2 93.9 91.7 93.7 70.1 Power outages 99.2 96.6 32.5 80.0 87.0 86.7 100.0 98.0 100.0 75.0 95.8 97.1 96.1 89.3 85.7 76.6 90.6 82.2 100.0 97.0 99.1 74.7 96.2 98.6 98.1 65.1 Average duration of power outages 86.6 32.9 86.5 84.0 85.8 100.0 96.1 95.7 71.3 86.6 92.4 96.1 88.5 81.9 75.5 92.4 98.2 100.0 97.0 100.0 84.0 96.2 97.1 97.1 64.2 Power fluctuations 64.4 88.8 74.5 97.9 62.5 90.8 81.4 84.7 84.7 95.9 93.1 90.2 89.6 88.6 95.4 Average duration of power fluctuations 68.4 73.3 77.3 100.0 Sales lost due to power outages 99.2 77.0 33.3 83.4 85.8 94.1 88.8 62.3 91.2 98.8 73.3 51.1 Water outages 99.2 54.6 6.8 86.7 79.6 94.2 68.8 98.0 95.7 70.0 97.5 91.3 96.1 87.2 77.2 33.5 89.2 93.4 100.0 50.5 97.3 70.7 50.0 95.7 95.0 29.9 Average duration of water outages 49.6 7.6 86.5 79.2 93.5 68.8 98.0 91.5 68.8 97.5 81.4 76.1 32.6 90.6 94.9 99.5 50.5 97.3 73.3 50.0 95.7 95.7 29.5 Sales lost due to water outages 99.4 50.7 7.1 86.7 65.1 93.5 22.2 85.8 96.8 65.3 94.2 20.2 Dummy for own well or water infrastructure 94.1 5.1 100.0 100.0 99.7 100.0 94.9 97.8 97.5 100.0 100.0 50.5 97.3 50.0 95.7 93 Water from own well or water infrastructure 51.8 100.0 96.6 99.8 99.7 68.8 84.3 95.7 73.8 95.8 90.7 100.0 95.2 71.9 100.0 90.7 Cost of water from own well 71.2 68.5 76.7 43.3 48.8 52.4 67.9 Cost of water from public system 68.7 52.0 Phone outages 98.9 7.4 94.3 81.3 87.7 87.2 96.1 92.1 58.4 17.5 87.2 100.0 60.0 97.4 36.4 Average duration of phone outages 8.1 93.9 77.1 87.7 92.1 57.5 15.7 87.5 100.0 62.7 96.6 36.1 Losses due to phone outages 4.0 94.5 86.1 10.7 40.0 70.3 25.9 Transport failures 7.3 29.7 33.0 13.2 89.9 100.0 62.7 30.1 Average duration of transport failures 7.6 29.5 33.0 11.3 90.6 99.3 62.7 29.7 Sales lost due to transport failures 8.1 29.1 36.2 8.4 83.7 64.0 21.8 Average duration of transport 100.0 88.4 91.5 98.6 94 Public postal service interruptions 57.3 21.6 Average duration of public postal service 50.7 21.2 interruptions Sales lost due to public postal service 2.7 15.2 interruptions Source: Authors' elaboration with IC data. 95 Table B.3.1 Response rate of infrastructure IC variables in the final sample (cont.) Horn of Eastern Africa-- Northern Africa Western Africa--ECOWAS Southern Africa--SADC (incl. Burundi) Africa EAC MUS ZAF DZA EGY MAR SEN BEN MLI MRT BFA CPV NER CMR ETH ERI KEN UGA TZA MWI MDG ZMB BDI BWA LSO NAM SWZ Dummy for own roads 99.2 100.0 100.0 96.8 100.0 98.0 100.0 98.8 Dummy for own transportation for 99.2 100.0 100.0 96.8 100.0 98.0 100.0 98.8 workers Dummy for own waste disposal 99.2 100.0 99.7 94.3 100.0 98.0 100.0 98.8 Dummy for contract with transportation 96.6 96.8 99.0 92.3 81.3 78.2 100.0 98.8 company Dummy for own transportation 99.9 100.0 96.6 96.8 99.0 100.0 98.0 97.9 80.0 100.0 92.3 81.3 100.0 100.0 100.0 98.1 98.6 98.6 Products with own transport 99.2 100.0 98.0 95.7 77.5 100.0 95.8 100.0 100.0 98.1 98.6 72.9 Transport delay, outgoing domestic 93.1 53.4 90.3 60.9 Transport delay, outgoing export 77.9 43.7 88.1 53.1 Transport delay, incoming domestic 94.4 78.8 89.3 58.6 96 Transport delay, incoming international 86.4 44.7 87.0 50.7 Shipment losses, domestic 99.4 99.0 50.9 83.0 13.8 100.0 97.9 75.0 99.2 98.3 100.0 83.6 21.9 91.1 100.0 98.7 100.0 2.0 17.0 86.7 24.5 37.1 95.9 99.7 Shipment losses, exports 28.8 54.2 67.2 9.9 100.0 65.5 91.8 Dummy for e-mail 89.3 99.9 72.5 98.9 100.0 98.4 100.0 98.0 100.0 80.0 100.0 99.9 100.0 97.4 100.0 96.2 100.0 100.0 99.3 100.0 100.0 98.7 98.1 98.6 99.5 100.0 Dummy for Web page 85.7 99.8 97.3 99.2 100.0 99.4 100.0 100.0 100.0 80.0 100.0 99.9 100.0 96.6 100.0 94.8 100.0 100.0 99.3 100.0 100.0 98.7 98.1 98.6 96.9 100.0 Wait for phone connection 31.7 13.1 97.5 55.6 53.8 73.5 30.0 21.6 19.1 23.8 16.0 34.3 33.5 61.3 42.8 23.1 48.3 23.8 6.9 24.1 36.0 38.7 20.0 43.9 39.9 Dummy for gifts to obtain a phone 0.0 12.6 56.8 56.1 74.1 30.0 23.5 100.0 23.8 16.8 66.0 43.1 30.8 49.3 26.0 13.9 25.9 42.7 38.7 24.3 42.7 39.9 connection Wait for electric supply 4.2 9.1 94.6 42.7 58.4 69.9 18.8 13.7 8.5 17.5 10.1 27.2 29.6 48.8 52.6 24.0 29.2 14.3 15.8 11.6 25.3 23.6 11.4 27.1 33.7 Dummy for gifts to obtain a electric 9.2 44.0 58.4 70.6 18.8 15.7 100.0 16.3 10.1 50.2 54.2 30.2 28.1 14.0 15.8 12.5 37.3 23.6 11.4 27.8 33.7 supply Wait for a water supply 4.8 94.2 25.5 47.9 65.4 13.8 11.8 6.4 12.5 5.9 17.7 11.5 6.3 0.0 9.8 20.0 14.2 7.1 11.8 30.1 Dummy for gifts to obtain a water supply 4.6 29.1 48.9 65.7 15.0 11.8 100.0 11.3 6.7 25.8 11.8 5.8 0.0 9.8 32.0 14.2 7.1 12.0 30.1 97 Wait for an import license 9.4 26.9 27.3 49.5 10.0 21.6 19.1 37.5 33.6 46.6 17.8 14.0 22.9 3.2 15.6 19.8 14.3 25.3 21.7 15.7 8.4 26.9 Dummy for gifts to obtain an import 9.1 32.2 29.4 54.7 11.3 21.6 100.0 38.8 31.1 47.4 18.7 25.2 22.6 5.8 19.8 16.1 33.3 20.8 18.6 8.4 27.1 license Low quality supplies 99.4 100.0 98.6 89.7 96.4 97.4 99.1 96.6 94.6 96.8 95.5 99.2 99.5 86.7 97.6 99.7 Sales lost due to delivery delays, 99.4 98.7 75.4 21.8 92.6 99.1 92.9 94.3 85.9 27.8 86.7 99.3 88.0 92.1 99.3 domestic Sales lost due to delivery delays, imports 42.7 93.7 40.1 76.3 30.2 2.1 48.5 Transport delays in domestic sales 63.6 75.2 86.8 92.9 90.7 94.0 86.0 86.7 Transport delays in international sales 62.4 43.0 93.7 41.1 75.1 30.2 48.8 Illegal payments to obtain public utilities 62.7 86.8 62.3 31.7 Days of inventory of main supply 98.4 89.1 98.4 100.0 94.1 97.9 72.5 98.3 85.3 78.1 91.5 81.9 94.1 99.3 100.0 99.1 86.7 97.2 94.3 95.2 99.0 Days of inventory of finished goods 85.1 98.4 99.6 39.7 96.6 85.7 82.8 90.7 96.4 0.0 Source: Authors' elaboration with IC data. 98 Table B.3.2 Response rate of red tape, corruption, and crime IC variables in the final sample Horn of Eastern Africa-- Northern Africa Western Africa--ECOWAS Southern Africa--SADC (incl. Burundi) Africa EAC MUS ZAF DZA EGY MAR SEN BEN MLI MRT BFA CPV NER CMR ETH ERI KEN UGA TZA MWI MDG ZMB BDI BWA LSO NAM SWZ Manager's time spent in bureaucratic issues 99.4 99.6 83.8 92.9 94.2 100.0 90.2 100.0 76.3 99.2 97.8 98.3 93.1 97.6 96.6 97.9 95.8 100.0 100.0 98.2 92.0 98.1 97.1 98.1 99.1 Payments to deal with bureaucratic issues 20.6 97.6 78.3 88.2 93.9 75.0 45.1 89.4 43.8 68.1 78.5 55.3 90.0 53.5 96.6 77.0 92.1 84.8 72.0 74.5 91.4 80.8 93.3 Illegal payments to obtain licenses 63.2 62.1 31.2 Illegal payments to tax administrators 60.4 Days spent with regulation agencies 98.7 23.5 Cost dealing with regulation agencies 76.2 11.7 Wait for a construction permit 99.4 89.4 20.4 30.0 54.4 12.5 7.8 19.1 11.3 4.2 24.3 24.1 9.7 21.5 1.0 6.3 18.7 17.0 4.3 16.1 Dummy for gifts to obtain a construction permit 21.1 32.4 57.9 12.5 7.8 100.0 11.3 5.9 27.5 25.2 2.0 22.6 1.0 8.0 32.0 18.9 7.1 16.4 Wait for an operating license 24.0 96.8 20.6 38.7 56.6 5.0 9.8 12.8 25.0 83.2 81.6 98.6 67.3 34.0 9.9 67.0 37.3 36.8 30.0 25.9 99 Gifts to obtain a operating license 24.3 24.8 40.3 63.4 5.0 9.8 100.0 25.0 78.2 79.4 97.0 100.0 36.5 9.9 68.8 53.3 35.8 31.4 26.1 Sales declared to taxes 33.2 98.4 98.1 78.1 90.1 96.4 95.0 94.1 87.2 76.3 96.6 87.9 69.9 90.9 83.7 98.9 77.5 100.0 98.2 64.0 95.3 98.6 86.3 94.3 Workforce declared to taxes 99.4 98.5 91.6 96.3 94.1 91.5 71.3 95.8 81.3 100.0 100.0 100.0 94.3 94.3 86.1 Days in inspections 99.9 99.6 91.8 100.0 90.9 87.5 98.0 100.0 73.8 95.8 99.6 98.3 96.0 99.7 95.3 99.3 75.3 100.0 84.2 49.1 82.7 24.5 70.0 95.9 97.8 Dummy for gifts in inspections 100.0 88.2 89.1 66.7 87.5 82.4 74.5 57.5 61.3 60.5 40.6 90.5 92.7 35.2 96.6 84.2 49.1 12.0 27.4 75.7 6.5 79.1 Dummy for lawyer/consultant to help deal with 38.7 62.8 permissions Payments to obtain a contract with the 99.4 98.6 47.6 89.9 89.0 73.8 68.6 93.6 40.0 49.6 49.6 53.5 64.8 94.8 98.7 91.1 87.5 41.3 72.6 92.9 88.7 74.1 government Dummy for law-influencing firm 99.8 100.0 100.0 96.2 99.8 95.9 0.0 100.0 93.3 99.8 Overdue payments to private customers 99.4 91.9 98.9 91.0 96.6 96.8 100.0 100.0 100.0 80.0 100.0 90.7 98.3 43.8 96.9 96.1 99.3 100.0 100.0 68.0 98.1 98.6 95.4 98.6 Overdue payments to SOEs 99.4 44.8 51.0 26.9 31.8 0.0 43.4 100.0 100.0 50.7 98.1 98.6 69.4 100 Weeks to resolve a case of overdue payment 49.9 57.6 70.9 77.9 85.4 82.2 49.8 47.8 87.6 79.1 50.7 85.1 93.4 Overdue payments in courts 99.4 97.1 60.6 92.6 64.1 96.8 100.0 100.0 100.0 80.0 100.0 29.8 18.3 10.0 96.1 47.0 41.3 88.0 93.8 Weeks to resolve an overdue payment in 11.8 14.4 17.1 19.7 51.8 26.7 9.6 10.4 18.4 4.3 23.3 14.7 14.4 36.4 courts Security expenses 68.7 97.2 88.2 81.9 96.1 38.8 82.4 87.2 26.3 82.4 97.4 48.6 93.7 96.4 93.0 94.8 97.3 99.8 39.6 30.4 64.0 28.3 42.9 77.0 99.5 Dummy for security expenses 68.1 97.2 88.2 81.9 96.1 100.0 100.0 97.9 80.0 100.0 97.4 48.6 93.7 96.4 93.0 94.8 97.3 99.8 100.0 100.0 64.0 98.1 98.6 77.0 99.5 Illegal payments in protection 85.6 65.9 83.2 93.2 90.1 91.7 86.6 96.2 98.7 100.0 72.2 99.5 Dummy for payments in protection 85.6 65.9 83.2 93.2 90.1 91.7 86.6 94.8 98.7 100.0 70.3 99.5 Cost to avoid pilferage from workers 90.7 Dummy for cost to avoid pilferage from 90.7 workers Crime losses 99.4 96.4 89.9 97.7 94.8 98.8 100.0 100.0 76.3 97.5 95.2 92.1 99.7 36.5 87.2 97.0 99.8 99.0 98.2 44.0 97.2 98.6 95.7 98.6 101 Dummy for crime losses 99.4 96.4 89.9 97.7 94.8 100.0 100.0 100.0 80.0 100.0 95.2 92.1 99.7 36.5 73.6 97.0 99.8 100.0 100.0 58.7 98.1 98.6 79.9 98.6 Crimes reported to police 6.5 44.2 17.6 75.4 11.6 27.5 33.9 64.6 70.8 22.5 82.7 61.3 15.6 81.9 Crimes solved by police 6.0 18.1 13.2 40.5 8.9 27.7 28.0 34.9 66.0 22.0 65.9 38.7 16.1 69.4 Days of production lost due to civil unrest 99.4 96.4 87.4 94.2 98.4 1.3 99.0 99.5 100.0 58.7 95.0 92.4 Days of production lost due to absenteeism 97.6 93.7 88.0 96.1 98.0 97.6 98.1 98.8 68.0 87.3 88.6 Dummy for tax exemption 97.5 97.2 98.1 95.4 Dummy for lawsuit in the last 3 years 91.8 99.8 Dummy for "gifts" for credit 96.2 Dummy for interventionist labor regulation 100.0 70.8 7.5 96.1 97.9 80.0 96.6 100.0 99.0 0.0 7.1 3.8 12.9 Total days spent with licenses 86.8 102 Dummy for accountant to accomplish taxes 100.0 Dummy for gifts to tax inspectors 100.0 Gifts to tax inspectors 100.0 Dummy for labor conflicts 92.5 Average time to hire a skilled worker 91.5 Dummy for conflicts with suppliers 98.9 Dummy for conflicts with clients 99.6 Cost of entry 20.8 10.6 Dummy for consultant to help deal with 20.9 permissions Source: Authors' elaboration with IC data. 103 Table B.3.3 Response rate of finance and corporate governance IC variables in the final sample Horn of Eastern Africa-- Northern Africa Western Africa--ECOWAS Southern Africa--SADC (incl. Burundi) Africa EAC MUS ZAF DZA EGY MAR SEN BEN MLI MRT BFA CPV NER CMR ETH ERI KEN UGA TZA MWI MDG ZMB BDI BWA LSO NAM SWZ Dummy for trade chamber 99.2 99.9 99.1 99.4 100.0 100.0 100.0 100.0 100.0 80.0 100.0 100.0 82.2 100.0 98.4 99.7 99.5 100.0 93.3 99.5 99.7 Dummy for credit line 97.9 99.6 100.0 99.4 96.8 99.7 98.9 98.9 95.3 100.0 94.3 100.0 100.0 100.0 100.0 100.0 89.3 98.1 98.6 98.1 100.0 Credit unused 97.9 99.9 100.0 91.2 21.2 95.5 100.0 100.0 100.0 80.0 100.0 94.3 64.6 97.6 99.3 82.7 83.0 74.6 Dummy for loan 100.0 100.0 100.0 96.6 96.2 98.7 100.0 100.0 100.0 80.0 100.0 99.5 98.9 91.3 100.0 44.0 100.0 100.0 100.0 100.0 100.0 100.0 98.1 98.6 97.6 100.0 Dummy for loan with collateral 100.0 97.7 53.2 99.8 27.1 100.0 17.5 13.7 31.9 11.3 16.8 94.7 45.3 44.7 21.6 44.0 99.3 100.0 100.0 100.0 100.0 14.7 98.1 98.6 97.4 88.9 Value of the collateral 87.0 99.7 44.8 79.6 20.8 90.0 100.0 98.0 100.0 75.0 98.3 62.2 38.0 34.4 18.6 41.7 89.9 97.8 100.0 12.9 21.4 27.4 14.3 87.3 89.3 Interest rate of the loan 94.7 44.5 93.3 26.3 95.8 100.0 100.0 100.0 80.0 100.0 40.1 20.0 44.0 97.2 97.9 98.8 100.0 97.3 13.3 97.2 98.6 94.7 66.0 Dummy for short-term loan 21.0 52.4 62.9 75.6 67.0 5.2 18.6 97.6 81.5 100.0 100.0 100.0 98.1 98.6 98.3 100.0 Borrows in foreign currency 99.4 99.9 93.9 40.1 93.2 100.0 100.0 100.0 80.0 100.0 77.7 95.4 85.0 96.5 22.0 70.3 100.0 100.0 18.7 98.1 98.6 95.5 Dummy for external auditory 99.1 99.8 97.7 100.0 100.0 98.0 100.0 77.5 95.8 99.4 98.9 98.0 100.0 99.8 100.0 100.0 100.0 88.0 98.8 98.6 104 Owner of the lands 97.1 90.3 80.6 94.5 96.9 88.2 100.0 100.0 100.0 100.0 92.0 98.1 98.6 91.6 99.8 Owner of the buildings 98.7 100.0 91.3 95.5 99.4 95.7 100.0 100.0 90.7 95.2 99.5 Dummy for owner of the buildings 99.7 Dummy for owner of the buildings and 98.0 100.0 92.2 95.7 80.0 94.1 99.2 lands Largest shareholder 99.4 100.0 96.6 93.7 95.0 96.1 100.0 100.0 97.9 80.0 99.2 16.1 16.8 95.7 95.1 95.7 97.2 95.8 99.3 100.0 100.0 98.7 94.3 95.7 96.6 98.6 Working capital financed by internal 89.9 100.0 99.8 95.8 96.6 93.9 100.0 100.0 97.9 80.0 99.2 98.5 98.9 93.1 99.8 94.3 98.6 99.5 99.0 100.0 99.1 80.0 98.1 98.6 93.3 99.0 funds Working capital financed by commercial 89.9 100.0 99.8 95.8 96.6 93.9 98.5 98.9 93.1 99.8 94.3 98.6 99.5 99.0 100.0 99.1 80.0 98.1 98.6 93.3 99.0 banks Working capital fin. by foreign 100.0 99.8 95.8 96.6 93.9 98.5 93.1 99.8 94.3 98.6 99.5 99.0 80.0 93.3 99.0 commercial banks Working capital financed by leasing 100.0 99.8 95.8 96.6 93.9 98.5 93.1 99.8 94.3 98.6 99.5 99.0 80.0 93.3 99.0 Working capital financed by state 100.0 99.8 95.8 96.6 93.9 100.0 100.0 97.9 80.0 99.2 98.5 98.9 93.1 99.8 94.3 98.6 99.0 80.0 99.0 services Working capital fin. by supplier or 89.9 100.0 99.8 95.8 96.6 93.9 98.5 98.9 93.1 99.8 94.3 98.6 99.5 99.0 100.0 99.1 80.0 98.1 98.6 93.3 99.0 customer credit 105 Working capital financed by credit cards 100.0 99.8 95.8 96.6 93.9 93.1 99.8 94.3 98.6 99.5 99.0 80.0 93.3 99.0 Working capital financed by equity 100.0 99.8 95.8 96.6 93.9 100.0 100.0 97.9 80.0 99.2 98.5 98.9 93.1 99.8 94.3 98.6 99.5 99.0 80.0 93.3 99.0 Working capital financed by 89.9 100.0 99.8 95.8 96.6 93.9 100.0 100.0 97.9 80.0 99.2 98.5 98.9 93.1 99.8 94.3 98.6 99.5 99.0 100.0 99.1 80.0 98.1 98.6 93.3 99.0 family/friends Working capital financed by informal 100.0 99.8 95.8 96.6 93.9 98.5 98.9 93.1 99.8 94.3 98.6 99.5 99.0 100.0 99.1 80.0 98.1 98.6 93.3 99.0 sources Working capital financed by other funds 73.3 99.8 95.8 96.6 93.9 35.0 96.1 93.6 77.5 80.7 98.5 93.1 99.8 94.3 98.6 70.9 99.0 80.0 93.3 99.0 New investments financed by internal 62.8 73.3 90.5 90.3 85.3 77.7 35.0 96.1 93.6 77.5 80.7 42.6 77.1 70.6 66.6 60.8 68.8 70.9 71.2 51.5 48.2 56.0 56.6 62.9 79.9 89.6 funds New investments financed bcommercial 62.8 73.3 90.5 90.3 85.3 77.7 42.6 77.1 70.6 66.6 60.8 68.8 70.9 71.2 51.5 48.2 56.0 56.6 62.9 79.9 89.6 banks New investments fin. by foreign 73.3 90.5 90.3 85.3 77.7 42.6 70.6 66.6 60.8 68.8 70.9 71.2 56.0 79.9 89.6 commercial banks New investments financed by leasing 73.3 90.5 90.3 85.3 77.7 42.6 70.6 66.6 60.8 68.8 70.9 71.2 56.0 79.9 89.6 New investments financed by state 73.3 90.5 90.3 85.3 77.7 35.0 96.1 93.6 77.5 80.7 42.6 70.6 66.6 60.8 68.8 71.2 56.0 89.6 services New investments fin. by supplier or 62.8 73.3 90.5 90.3 85.3 77.7 42.6 70.6 66.6 60.8 68.8 70.9 71.2 51.5 48.2 56.0 56.6 62.9 79.9 89.6 customer credit 106 New investments financed by credit 73.3 90.5 90.3 85.3 77.7 70.6 66.6 60.8 68.8 70.9 71.2 56.0 79.9 89.6 cards New investments financed by equity 73.3 90.5 90.3 85.3 77.7 35.0 96.1 93.6 77.5 80.7 42.6 77.1 70.6 66.6 60.8 68.8 70.9 71.2 56.0 79.9 89.6 New investments financed by 62.8 73.3 90.5 90.3 85.3 77.7 35.0 96.1 93.6 77.5 80.7 42.6 77.1 70.6 66.6 60.8 68.8 70.9 71.2 51.5 47.3 56.0 56.6 62.9 79.9 89.6 family/friends New investments financed by informal 73.3 90.5 90.3 85.3 77.7 42.6 77.1 70.6 66.6 60.8 68.8 70.9 71.2 51.5 48.2 56.0 57.5 62.9 79.9 89.6 sources New investments financed by other funds 75.5 90.5 90.3 85.3 77.7 42.6 70.6 66.6 60.8 68.8 70.9 71.2 56.0 79.9 89.6 Share of net profits reinvested 99.8 96.9 83.0 69.1 92.9 100.0 100.0 100.0 80.0 100.0 90.1 79.1 90.5 76.7 78.0 68.1 82.7 95.0 98.3 Sales bought on credit 97.9 95.1 100.0 94.5 99.5 97.3 100.0 98.4 100.0 100.0 100.0 84.0 98.1 98.6 Dummy for inputs bought on credit 99.4 99.8 99.4 99.7 100.0 100.0 100.0 78.8 99.2 95.3 99.8 98.4 99.5 98.6 Inputs bought on credit 96.2 57.4 97.7 85.6 99.8 97.9 99.3 99.0 100.0 98.2 98.1 97.1 Time to pay off the credit for inputs 87.4 62.0 61.7 99.0 107 Inputs bought on credit with delayed 68.8 payment Wait to clear a check 99.6 88.0 77.2 97.2 81.2 95.3 99.0 61.3 96.9 90.8 Charges to clear a check 52.0 86.0 37.9 74.6 14.7 11.4 Wait to clear a domestic currency wire 99.2 87.2 43.2 21.8 52.1 88.6 76.0 32.0 80.6 86.4 Charges to clear a domestic currency 28.1 19.0 36.0 58.0 13.3 9.7 wire Wait to clear a foreign currency wire 96.1 63.9 11.8 10.1 34.5 61.3 82.3 48.0 83.0 68.9 Charges to clear a foreign currency wire 6.0 10.1 6.3 64.7 6.7 10.9 Wait to clear a letter of credit 25.7 47.3 Charge to clear a letter of credit 18.9 8.1 Delay of payments of domestic clients 91.8 96.4 83.6 95.3 108 Charges to get payments from domestic 70.3 84.8 48.2 71.0 clients Delay of payments of foreign clients 51.2 31.7 48.6 22.2 Charges to get payments from foreign 36.6 20.4 100.0 98.0 97.9 80.0 100.0 28.5 16.7 clients Dummy for current or saving account 99.6 93.7 99.2 96.1 99.8 100.0 100.0 98.1 98.6 Dummy for foreign current or saving 99.2 100.0 account Dummy for accountant 99.8 95.0 99.4 100.0 Source: Authors' elaboration with IC data. 109 Table B.3.4 Response rate of quality, innovation, and labor skills IC variables in the final sample Horn of Eastern Africa-- Northern Africa Western Africa--ECOWAS Southern Africa--SADC (incl. Burundi) Africa EAC MUS ZAF DZA EGY MAR SEN BEN MLI MRT BFA CPV NER CMR ETH ERI KEN UGA TZA MWI MDG ZMB BDI BWA LSO NAM SWZ Dummy for foreign technology 99.0 99.3 97.3 100.0 99.0 100.0 100.0 100.0 53.8 99.2 91.3 98.4 100.0 100.0 98.6 100.0 100.0 94.7 98.1 98.6 99.5 100.0 Dummy for ISO quality certification 100.0 99.1 99.6 99.4 100.0 98.8 98.0 100.0 56.3 100.0 100.0 97.2 98.0 97.9 98.6 100.0 100.0 100.0 100.0 93.3 98.1 98.6 99.5 99.3 Sales with warranty 99.9 Dummy for new product 99.4 100.0 100.0 100.0 100.0 100.0 100.0 55.0 99.2 98.5 94.7 98.4 100.0 100.0 100.0 100.0 100.0 89.3 97.2 98.6 98.8 99.7 Dummy for product improvement 99.8 100.0 99.4 100.0 100.0 100.0 80.0 99.2 94.7 98.4 100.0 100.0 100.0 100.0 100.0 90.7 97.2 98.6 99.5 99.7 Dummy for discontinued product line 99.7 Dummy for equipment improvement 100.0 Dummy for R&D 99.8 87.0 98.1 89.3 100.0 100.0 100.0 80.0 100.0 98.7 78.8 89.9 99.1 100.0 R&D expenditures 100.0 69.5 81.9 80.9 100.0 100.0 97.9 80.0 96.6 95.4 40.2 74.3 91.1 110 Workers engaged in design/R&D 97.3 Dummy for subcontracted R&D 97.5 100.0 Royalties expenditures 98.4 Dummy for new technology 37.4 100.0 99.8 100.0 96.4 100.0 98.4 100.0 100.0 90.7 99.5 99.7 Dummy for joint venture 99.9 99.8 98.4 100.0 86.7 99.7 Dummy for new license agreement 100.0 99.4 98.4 100.0 100.0 88.0 99.7 Dummy for outsourcing 100.0 99.5 98.4 100.0 100.0 100.0 100.0 77.3 98.1 98.6 99.7 Dummy for in-house production 99.5 97.9 100.0 100.0 76.0 99.5 Dummy for new plant 99.7 Dummy for closed plant 99.7 Staff--management 100.0 97.1 98.7 99.4 88.1 100.0 91.3 97.6 93.0 97.6 97.1 100.0 78.7 100.0 111 Staff--professional workers 68.7 99.5 100.0 95.6 98.9 98.4 88.2 98.3 91.5 97.6 62.3 98.3 99.3 57.3 82.0 100.0 Staff--skilled workers 68.7 99.5 100.0 95.4 98.7 98.7 100.0 100.0 100.0 72.5 97.5 87.9 100.0 91.5 97.6 84.3 97.6 97.1 100.0 100.0 99.1 65.3 97.2 98.6 80.3 100.0 Staff--unskilled workers 68.7 99.5 100.0 95.4 98.7 98.7 100.0 100.0 100.0 72.5 97.5 87.6 98.3 90.9 98.1 66.7 96.9 96.6 100.0 100.0 99.1 64.0 97.2 98.6 77.5 100.0 Staff--nonproduction workers 68.7 99.5 100.0 95.4 98.9 98.4 87.0 100.0 91.5 98.1 70.3 96.9 96.6 100.0 57.3 76.7 100.0 Staff--foreign nationals 96.4 0.0 85.2 98.3 95.8 99.3 78.7 99.3 Average education of staff 95.2 Average tenure of staff 94.3 Average age of staff 95.4 Dummy for training 97.1 99.8 99.5 99.4 92.6 100.0 100.0 100.0 100.0 80.0 100.0 96.8 93.3 93.3 100.0 85.9 99.3 100.0 100.0 100.0 100.0 92.0 98.1 98.6 96.6 100.0 Training to skilled workers 89.1 99.2 98.4 25.3 55.7 23.9 100.0 100.0 100.0 78.8 100.0 98.7 8.9 87.7 91.5 93.0 97.2 98.2 100.0 89.3 81.8 97.6 Training to unskilled workers 89.1 98.9 97.5 23.0 23.1 17.2 100.0 100.0 100.0 78.8 100.0 98.6 14.5 76.9 79.2 89.4 97.9 99.2 99.0 84.0 74.3 95.5 Training to production workers 99.2 96.7 100.0 100.0 98.1 98.6 112 Training to nonproduction 98.7 96.7 100.0 100.0 96.2 98.6 Weeks of training for skilled workers 16.0 22.1 19.1 100.0 80.4 70.2 72.5 79.8 75.9 91.0 69.8 88.2 97.8 94.2 86.7 87.5 96.0 Weeks of training for unskilled workers 4.4 8.6 3.2 100.0 80.4 70.2 70.0 79.8 56.7 79.2 55.6 69.1 98.1 78.9 82.7 83.0 91.2 Workforce with computer 99.4 98.7 98.1 100.0 99.4 83.4 99.5 97.5 98.6 84.2 84.8 University staff 97.1 100.0 80.8 90.8 79.3 96.8 91.6 79.1 97.0 89.4 74.3 96.9 77.3 99.8 Dummy for university staff 100.0 96.1 91.5 62.5 99.2 100.0 99.1 96.2 97.1 Manager's education 99.9 99.2 96.4 32.1 99.4 100.0 100.0 100.0 80.0 100.0 99.1 100.0 98.0 99.1 99.3 26.3 99.3 100.0 100.0 92.0 98.1 98.6 100.0 99.8 Manager's experience 99.8 98.8 83.0 98.1 98.4 100.0 98.0 100.0 76.3 99.2 99.7 100.0 66.4 79.7 80.1 96.5 98.8 100.0 100.0 77.3 98.1 98.6 74.6 99.8 Source: Authors' elaboration with IC data. 113 Table B.3.5 Response rate of other control C variables in the final sample Horn of Eastern Africa-- Northern Africa Western Africa--ECOWAS Southern Africa--SADC (incl. Burundi) Africa EAC MUS ZAF DZA EGY MAR SEN BEN MLI MRT BFA CPV NER CMR ETH ERI KEN UGA TZA MWI MDG ZMB BDI BWA LSO NAM SWZ Age 99.4 99.7 100.0 99.2 99.6 99.7 100.0 100.0 100.0 33.8 100.0 100.0 100.0 97.4 99.4 98.9 97.9 99.7 99.3 100.0 100.0 100.0 98.1 95.7 98.8 99.8 Dummy for incorporated company 98.5 99.4 99.9 99.8 100.0 99.7 100.0 100.0 100.0 80.0 100.0 100.0 100.0 98.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 98.1 98.6 96.4 100.0 Dummy for limited company 99.2 99.4 99.9 99.8 100.0 99.7 100.0 100.0 100.0 80.0 100.0 100.0 98.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 98.1 98.6 100.0 100.0 Dummy for SOE 98.7 100.0 100.0 99.2 100.0 99.7 100.0 100.0 100.0 80.0 100.0 100.0 100.0 98.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 94.7 98.1 98.6 100.0 100.0 Dummy for FDI 98.7 100.0 100.0 99.2 100.0 99.7 100.0 100.0 100.0 80.0 100.0 98.6 100.0 98.0 100.0 97.9 100.0 100.0 100.0 100.0 100.0 94.7 98.1 98.6 100.0 100.0 Dummy for holdings 99.8 100.0 99.7 100.0 94.1 100.0 20.0 100.0 100.0 100.0 98.7 98.8 100.0 Share of the local market 41.0 64.1 53.5 82.4 83.8 93.8 62.7 83.0 61.3 92.4 78.1 96.6 96.0 98.2 57.3 90.6 81.4 92.9 Share of the national market 36.6 63.2 52.4 83.0 79.3 90.0 66.7 72.3 60.0 84.0 94.5 62.2 86.9 82.3 49.4 75.8 98.0 96.4 40.0 88.7 81.4 57.3 93.3 Dummy for direct exports 97.5 99.7 99.9 96.4 95.8 87.4 100.0 100.0 100.0 80.0 100.0 98.9 100.0 65.0 99.5 98.9 99.3 99.7 100.0 100.0 100.0 86.7 98.1 98.6 97.6 99.5 Share of exports 97.5 99.7 99.9 96.4 95.8 87.4 100.0 100.0 100.0 80.0 100.0 98.9 100.0 94.5 99.5 98.9 99.3 99.7 100.0 100.0 100.0 86.7 98.1 98.6 97.6 Exporting experience 97.7 97.4 99.9 94.9 92.4 91.9 100.0 100.0 100.0 72.5 100.0 98.0 92.1 17.8 27.2 99.3 31.9 98.8 100.0 99.1 89.3 97.2 94.3 97.6 Dummy for direct imports 97.9 100.0 90.1 97.9 92.9 68.8 100.0 100.0 80.0 100.0 98.9 100.0 92.9 92.6 93.6 99.0 97.6 92.6 78.2 100.0 100.0 98.1 98.6 93.8 97.4 Share of imports 99.9 100.0 90.1 97.9 92.9 100.0 100.0 100.0 80.0 100.0 97.9 100.0 94.1 93.7 94.8 99.3 99.0 100.0 100.0 100.0 93.3 98.1 98.6 93.8 114 Number of competitors 94.3 83.2 95.0 65.4 94.5 98.6 96.1 99.3 72.4 63.5 86.7 72.7 99.5 Capacity utilization 95.2 99.7 99.2 82.7 97.7 98.4 100.0 96.1 91.5 53.8 99.2 87.0 100.0 90.1 94.8 99.5 97.6 95.2 99.8 100.0 98.2 89.3 95.3 94.3 97.1 98.1 Trade union 99.4 95.9 97.3 95.6 97.5 98.7 98.8 100.0 100.0 76.3 100.0 98.6 93.9 91.3 99.4 91.6 100.0 94.7 86.3 100.0 100.0 64.0 97.2 95.7 97.1 99.1 Dummy for privatized firm 97.7 98.1 97.3 98.5 98.7 95.6 100.0 89.5 93.7 97.6 96.8 100.0 98.7 96.9 100.0 Dummy for industrial zone 99.9 100.0 99.2 99.2 99.7 100.0 96.1 100.0 80.0 98.3 97.6 100.0 98.0 100.0 98.6 99.3 100.0 100.0 98.1 98.6 99.5 0.0 Days of production lost due to strikes 99.4 97.4 96.4 87.4 95.5 98.8 89.7 98.9 1.6 65.3 99.5 100.0 65.3 95.0 92.6 Workers infected by HIV 68.2 65.1 57.0 56.3 98.3 70.0 80.8 35.4 Dummy for negative impact of HIV 64.4 76.7 90.6 100.0 98.0 97.9 80.0 99.2 99.9 84.0 94.3 94.8 91.0 64.8 98.0 100.0 62.7 93.4 98.6 88.2 100.0 Cost in HIV-prevention programs 67.0 75.8 75.1 99.9 19.0 66.3 48.7 87.5 78.5 90.9 44.7 Source: Authors' elaboration with IC data. 115