WPS7414 Policy Research Working Paper 7414 Infrastructure Investment Demands in Emerging Markets and Developing Economies Fernanda Ruiz-Nuñez Zichao Wei Public-Private Partnerships (PPPs) Group September 2015 Policy Research Working Paper 7414 Abstract The authors have assembled 1960–2012 infrastructure billion per year, which implies that emerging markets and stock data from 145 countries to estimate the demand for developing economies should almost double their current infrastructure services in emerging markets and developing spending. The paper also estimates that half of the spend- economies. This paper identifies that the required resource ing should be allocated to maintenance of existing assets. flows to satisfy new demand while maintaining service for Acknowledging the challenges to compare infrastructure existing infrastructure amounts to $836 billion or 6.1 per- investment estimates across different methodologies, the cent of current gross domestic product per year over the authors recognize this result as a lower bound estimate and period 2014–20. The annual infrastructure investment gap compare the results with others available in the literature. for emerging markets and developing economies is $452 This paper is a product of the Public-Private Partnerships (PPPs) Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at fruiznunez@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Infrastructure Investment Demands in Emerging Markets and Developing Economies Fernanda Ruiz-Nuñez and Zichao Wei1 JEL Classification: H54, O13, O18, R42, E22 Keywords: Infrastructure, Infrastructure Investment Demand, Investment Requirement, Government Spending, Infrastructure Gap.                                                                   1 Fernanda Ruiz Nuñez is a Senior Economist at the Public-Private Partnerships (PPPs) Group at the World Bank; and Zichao Wei is a Financial Analyst (STC) at World Bank’s PPPs Group. The authors are grateful to Francesca Guiso for assisting with data analysis, and Tito Yepes who has provided database used in Yepes (2008)’s paper. The author would like to thank Marianne Fay, Tito Yepes and Atsushi Iimi for their valuable comments and discussions. The findings, interpretations and conclusions expressed herein do not necessarily reflect the views of the Board of the Executive Directors of the World Bank or the governments they represent. The authors’ email addresses are: fruiznunez@worldbank.org, and zwei@worldbank.org.   I. Introduction The links between infrastructure and development are well established, including the impact of infrastructure on poverty alleviation, equality, growth and specific development outcomes such as job creation, market access, health and education. 2 These relationships are complex and dynamic; even with respect to growth and job creation, infrastructure’s impacts are felt through multiple channels.3 The demand for infrastructure is rising with the accelerating pace of globalization and urbanization. Every month in the developing world more than five million people migrate to urban areas. This demand trend is compounded by the growing need for low CO2 and climate- resilient investments to combat the challenges of climate change.4 The developing world suffers from an undersupply of infrastructure, constraining economic growth rates, leaving the world’s most vulnerable communities without access to basic services, and hampering attempts to achieve broad-based poverty reduction. Currently, it is estimated that some $700 billion to $800 billion is invested worldwide ($550 billion in emerging markets and developing economies (EMDE)) in infrastructure every year, although data are scarce.5 In a nutshell, more than 70% of the current infrastructure funds come from the public sector. The second largest source of financing comes from the private sector, roughly 20% of the total funds. The remaining is covered by ODA.6 A range of estimates on infrastructure financing requirements applying different assumptions and methodologies for achieving various global goals have been estimated over the years. Despite the challenges comparing results across different methodologies, all the latest estimates of the global infrastructure investment requirements indicate that there is a significant gap. This paper reviews the recent global estimates and summarizes the main findings. Using a top-down approach, this paper estimates the global estimates on infrastructure investment requirements to satisfy consumer and producer’s demand for infrastructure services, assuming specific economic and demographic growth rates. This paper uses Yepes’ (2008) model addressing some of the limitations of the original model developed by Fay (2000), and gathers the most recent data on infrastructure stock and macro- economic variables up to 2012.7 It is important to address that the purpose of long- term projections is not to forecast the future. Rather, it is to explore how the future infrastructure investment might evolve given a set of economic, social, technical and                                                                 2 Straub (2008); and Serven and Calderon (2004b, 2008 and 2010) 3 Agénor and Moreno-Dodson (2006) for an overview. Estache, et al (2013) and Schwartz, et al (2009) for a treatment of infrastructure impacts on jobs and growth. 4 Fay and Toman (2010) and Bhattacharya, et al (2013) 5 These figures are author’s own estimates based on a number of sources. See table 5. 6   Adopted as a concept by the Development Assistance Committee (DAC) of the Organization for Economic Cooperation and Development in 1969, official development assistance (ODA) is a classification of resource flows from 22 DAC donor countries.  7   It also explicitly list the countries used for the estimation for which data is available and extrapolates results for all Emerging Markets and Developing Economies (EMDE). Moreover, one of the main contributions of this study is to simplify future updates of investment needs by providing a model and database simple to update when most recent data is available.  2      political assumptions. The estimates from this paper should be considered as complement to the existing estimates in the literature by providing lower bound figures and filling the gap for certain sectors or regions. II. Definitions and Previous Literature A range of estimates on infrastructure financing requirements applying different assumptions and methodologies for achieving various goals have been estimated over the years. Terms such as “needs”, “demand”, and “gap” have been used in the literature interchangeably generating some confusion when comparing results. In order to quantify “needs”, normative targets should be established usually assuming an optimal level of infrastructure. Different targets will lead to different estimated needs. “Demand,” as used in this paper, refers to the infrastructure investment requirements to satisfy the consumer and producer demand, based on predicted GDP growth and other growth trends. The terms “gap” is used as the difference between the estimated requirements and the actual level of investments. There are various methodologies used in the literature to estimate infrastructure investment requirements; they can be broadly classified in two categories: “top-down” and “bottom-up” (See Annex I). As the data of macroeconomic variables are relatively more comprehensive compared with the data of individual infrastructure projects, majority of the studies in fact adopt the “top-down” approach to conduct the empirical analysis. The most recent estimates of annual infrastructure investment requirements for Emerging Markets and Developing Economies (EMDE) are Fay et al (2011) which estimate the annual needs of US$0.9-1.1 trillion with an annual gap of US$100-300 billion by extrapolating estimates from Yepes (2008) using price inflation rate. A recent G24 policy paper, Bhattacharya et al. (2013), calculates investment annual needs of US$2 trillion and investment gap of US$1 trillion by extrapolating Fay et al (2011) using economic growth rate and price inflation rate. Both studies are extrapolations of previous studies and they do not account for actual changes in infrastructure stock since 2006.8 McKinsey (2013)’s uses three approaches to estimate infrastructure investment requirements. First, they use the historical spending approach, which assumes that future needs will follow past spending as percentage of GDP. Using this assumption, they project spending to 2030. The main limitation of this approach is that the data are mainly limited to OECD countries and extrapolated for the rest of the world. Secondly, they use the stock of infrastructure approach that estimates its value by using a perpetual inventory model. Unfortunately, they only have data for 12 countries9 and                                                                 8 Moreover, they fail to mention the list of countries for which data was not available.  9 The model was populated with data from the following countries: Brazil, Canada, China, Germany, India, Italy, Japan, Poland, South Africa, Spain, United Kingdom and United States  3      they use this information to estimate the average value of infrastructure stock (calculated at 70% of GDP). This average value is used to extrapolate results for the rest of the countries until 2030. Thirdly, they use the independent estimates of future needs approach that aggregate the estimates from three different sources: (i) Global Water Intelligence for the water sector; IEA (2011) for the power sector and OECD (2012) for the remaining infrastructure sectors. Based on these three approaches, they estimate that global infrastructure investment requirements for 2013-2030 range from US$57 trillion to US$67 trillion, or about US$3.2 trillion to US$3.7 trillion annually on average. The OECD (2006) report concluded that global infrastructure investment requirements across the transport (road, rail), telecom, electricity and water sectors would amount to around US$53 trillion over 2010-30, or about US$ 2.7 trillion annually average. Annual investment requirements for these sectors amount to some 2.5% of world GDP, which would rise to 3.5% of GDP; if electricity generation and other energy-related infrastructure investments in oil, gas and coal are included. The methodology used built on existing estimates, and extrapolates those to 2030. For the transport sector, they apply the estimated regional level elasticity (the ratio of paved road capital stock in USD billion to GDP) from Fay and Yepes (2003) assuming a fixed convergence rate. Asset replacement is calculated by assuming the economic life provision of 30 years without controlling by the current quality level. For the water sector, they extrapolate the proportion of GDP allocated to water services, varying from 0.3% to 2% based on different levels of income. For telecoms, they identify the underlying trends and drivers, and extrapolate the percentage of GDP invested in telecoms from Fay and Yepes (2003. Lastly, for the energy sector, they adopt estimates from IEA’s World Energy Outlook, which uses a scenario approach to analyze the possible evolution of energy markets to 2030. There are other estimations in the literature, however, they are based on Fay and Yepes (2003), McKinsey (2013), and OECD (2006). For example, Kennedy and Corfee- Morlot (2013), based on OECD work, are in the high part of the range, with annual investment requirements estimated at US$ 6.6-7.2 trillion when buildings and vehicles included at low carbon (2°C) scenario, and US$ 2.7-3.2 trillion when those are excluded. Using McKinsey’s (2013) estimates of infrastructure investment requirements, KPMG (2014) and Standard & Poor's Ratings Services present the annual investment requirements of $3.2 trillion by 2030. The World Economic Forum’s (2013) estimation of annual infrastructure spending need for 2020 is calculated by taking the Fay et al (2010) estimate of US$ 1.25-1.5 trillion annually in 2013 and assuming a 4% annual growth rate from 2013-20, and an additional US$ 200-300 billion annual requirement to make the infrastructure sustainable (by providing for climate change mitigation and adaptation). There have been previous efforts to calculate regional estimates, but some of them are outdated and do not account for changes in infrastructure stock or have significant 4      drawbacks. Gill and Kharas (2007) estimate infrastructure investment requirements for East Asia and the Pacific using data up to 2005. Data for Latin America and the Caribbean were collected from primary sources for 1990–2005 and then extrapolated in posterior work as presented in Yepes (2008). The most recent figures are from South Asia, where Andres et al. (2013) have extrapolated previous estimations at the country level for South Asia.10 The estimations from Sub-Saharan Africa come from Foster and Briceño-Garmendia (2010) that made a significant effort to collect data on infrastructure stock and present a bottom-up approach. Some of the most recent sectoral estimates come from the International Energy Agency (IEA). IEA indicates that the annual investments in energy supply of $1.6 trillion need to rise steadily over the coming decades towards $2 trillion. Under IEA’s expectation, future electricity usage underpins changes in the installed capacity of power generation technologies and in transmission and distribution networks, which assume global electricity demand growth rate of 2.2% per year, projected to future demand. Annual spending on energy efficiency, measured against a 2012 baseline, needs to rise from $130 billion today to more than $550 billion by 2035. OECD presents estimates at sectoral level for transport, telecoms, and water and sanitation. Comparing the different estimates presented some significant challenges (see Annex II). Recognizing the limitation of comparing results from different methodologies, the table below presents the most recent estimates on infrastructure investment requirements/demand in the literature.                                                                 10  Estimates by Chatterton and Puerto (2006) follow Fay (2000) methodology to estimate investment needs in South Asia but estimates are up to 2010.  5      Table 1. Estimations on annual infrastructure investment requirements (US$ billion) Global Estimates (USD billion) $3,200-$3,700 McKinsey, 2013 Global $2,680-$3,200 Kennedy&Corfee-Morlot, 2013 $2,650 OECD, 2006 $836 Ruiz-Nuñez&Wei, 2015 $2,000 Bhattacharya, 2013 $1,250-$1,500 WEF, 2013 Developing $900-$1,100 Fay, 2011 $1,113 Yepes, 2008 $465 Fay&Yepes, 2003 Regional Estimates (USD billion) $498 Bhattacharyay, 2010 EAP $165 Gill and Kharas, 2007 $34 Bhattacharyay, 2010 $260-$434 Kohli&Basil, 2011 LAC $186 Perrotti& Sanchez, 2011 $57 Fay, 2000 MENA $78.50 World Bank. 2009 $137-$206 Andrés et al, 2013 SAR $234 Bhattacharyay, 2010 $62 Chatterton&Puerto, 2006 SSA $93 Foster&Briceño-Garmendia,2010 Sectoral Estimates (USD billion) Electricity $2,000 IEA,2012 Transport $805 OECD, 2011 Telecoms $600 OECD, 2007 W&W+WW $1,320 OECD, 2007   Source: Authors’ summary III. Methodology This paper uses Fay’s (2000) model to estimate infrastructure investment demand and tries to address some of the limitations in the methodology following Yepes (2008). See Annex III with details of the methodology used. The model estimates the infrastructure investment requirements needed to satisfy consumer and producer demand given an expected growth rate. The model follows a top-down approach and accounts for global, regional and sectoral estimates. As described in the previous section, most of the methodologies in the literature have significant drawbacks, and Yepes (2008) is not innocuous to those. This methodology does not account for regional or sector specificities and supply constraints. 11 Importantly, the model does not estimate the level of infrastructure that would maximize growth or socio-economic targets, that is to say the model assumes no optimality. The relationship between income level and infrastructure service demand is                                                                 11  Data limitations does not allow to address those constraints.  6      established on the basis of past observed behavior in a sample of countries and extrapolated for the future using predicted income growth. Using Yepes (2008) as starting point, we constructed an annual panel of infrastructure stocks, macro variables, and demographics for 145 countries for 1960–2012. Due to missing data the numbers of periods observed for each country differ and therefore, the database is an unbalanced panel. (See Annex IV for data sources.) The infrastructure variables gathered are:  Telephones (mobile and landlines) subscribers per 1,000 persons;  Kms of paved roads per Km2 of land  Kms of unpaved roads per Km2 of land  Kms of rail per 1,000 persons  Total TEU handled  KW of installed electricity generation capacity per capita;  Percentage of households with access to electricity  Percentage of households with access to water and sanitation  Percentage of households with access to sanitation  Percentage of households with access to wastewater treatment After obtaining the annual database, we followed Yepes (2008) by taking five-year averages of all variables for each country, and obtaining a panel that contains the five- year average of annual flows (for flow variables) or the average stock held over the five-year period (for stock variables). The main motivation for this transformation comes from the lack of annual data on infrastructure stock and the fact that infrastructure accumulation is a slow process that happens over long periods of time. Therefore, we are not able to capture the short-term dynamics of the investment process. Instead, Yepes’s (2008) model aims to reveal the relationship between income flows and infrastructure stocks at low frequencies. Therefore the model is estimated on a panel with five-year periodicity. The benchmark GDP growth scenarios are created by extrapolating the 5-year-long regional growth rate to predict future GDP. This approach provides a region-specific scenario that is tailored to forecast. Then, the model applies the average growth rate of 5 years (2010–2015) GDP series, including the forecasted value of year 2014 and 2015, to predict the GDP trend for 2014–2020, which results in an implied annual growth rate for regions GDP projection. Table 2. Regional GDP growth rate used to forecast GDP 2014-2020 Region EAP ECA MENA LAC SAR SSA GDP Growth rate 7.79% 3.95% 1.64% 3.32% 6.32% 4.63% Source: Authors’ estimation based on Global Economic Prospects, WDI 7      To obtain the demand function for infrastructure, the model includes some additional regressors: (i) agricultural share of value added, (ii) manufacturing share of value added, (iii) urbanization rate, and (iv) population density. For each country, we created scenarios for these variables by obtaining their growth rate for the 1990–2012 period, and then applying that growth rate to obtain a forecast for the 2014–20 period. Projections of GDP composition are not available (share in agriculture and in manufacturing) so we keep them at 2010 values. For urbanization and population, we use the 2014–20 projections provided by the United Nations. Since infrastructure stocks tend to change reasonably slowly over time and have a long life span, we included a lagged value of the dependent variable in the regression in order to increase the explanatory power. We take Fay and Yepes (2003) estimate equation below to estimate the infrastructure demand: j j Ii,t  0  1Ii,t1  2 yi,t  3 Ai,t  4Mi,t  5Di  6Dt  i,t where all variables are in natural logs to linearize the model, Iji,t is demand for infrastructure stock of type j in country i at time t; Iji,t-1 is the lagged value of the infrastructure stock; y is income per capita; A is share of agriculture value added in GDP; M is the share of manufacturing value added in GDP; Di is a country fixed effect; Dt is a time dummy; and  is the error term.12 In order to calculate maintenance cost, the report provides estimates based on a top- down approach of using data on infrastructure and parameters for maintenance costs. The maintenance cost is calculated by applying a depreciation rate to the predicted total value of the capital stock. The unit costs used to value infrastructure were obtained by applying GDP deflator inflation rate to previously estimated unit cost values presenting prices at 2011 level. Unit costs are assumed constant over time. Since more recent data were not available, these numbers were validated by World Bank experts. For transport, energy and water sectors, a GDP deflator inflation rate is applied to the lasted available price in the literature and they were later validated by World Bank experts. In terms of telecoms sector, the costs have come down significantly in the last decade, mainly due to competition, technological change, market entry by Chinese companies and market volume. However, according to experts in the field, the projected price level is expected to remain constant in the coming decade.13                                                                 12 Manufacturing rather than industry was used here because industry includes mining, which has very different implications on the demand for infrastructure.    13  This is due to: (i) decline in shipments, due to market saturation, (ii) technological change is now pushing in favor of higher performance networks (eg DSL, fibre, 4G) which will tend to swap out performance for price, (iii) the shake up to the market 8      Table 3. Unit Costs (USD), by Sector 2011 Price Level Sector Cost per unit Unit Electricity generation 2,700 kW Paved roads 500,000 kilometer Unpaved roads 51,000 kilometer Rails 1,200,000 kilometer Rural water and sanitation 150 person Urban sanitation 150 person Urban water 80 person Main telephone lines* 200-300 line Mobile lines* 90-130 line Access to electricity 250 person Ports 360 TEU Wastewater treatment 150 person Source: Authors’ estimation validated with World Bank experts. *Main telephone lines unit price ranges from US$200-US$300 per line per connections in areas greater than 3km of an existing exchange. The lower bound is used for India (US$200) and the upper bound for the rest of the countries. *Mobile lines unit price is related to the urbanization rate. US$100 for new 3G connection in urban areas and US$ 160 for new 3G connections in rural areas. We take the average of the rural and the urban areas, as the projected urbanization rate is nearly 50% in developing countries. Therefore, US$130 unit price is used all countries except India where the numbers would be closer to US$60 and US$120, respectively with an average of US$90 per connection. IV. Main Results The results of the estimated regression are included in Annex V. This study indicates that the developing world will require annual investments about $819 billion (6 percent of GDP in developing countries) through 2020 to satisfy consumer and producer demand for infrastructure services. Of this amount, about 49 percent is required for capital expansion of infrastructure and 51 percent for maintenance of both current and future infrastructure. In addition, high income countries will require about $ 285 billion (0.8 percent of GDP in developed countries) of annual infrastructure investments during the 2014- 2020 period. The table below describes the results for the expected annual investment requirements, including capital expenditure and maintenance cost, by income groups (low income, upper                                                                 caused by the entry of Chinese manufacturers like Huawei and ZTE won’t happen again for a while.  9      middle-income and lower middle income), regions, and infrastructure sector (i.e transport, telecom, electricity, and water and sanitation). As expected, the investment requirements as a share of GDP are the highest in low-income countries (14.1%). This is followed by lower-middle income (3.4%), and upper-middle income (2.6%) countries. The sectors accounting for the greatest share of infrastructure investment requirements is electricity (39%). Transport is the second largest (31%), followed by telecommunications (23%). The Water and sanitation sector only account for 7% of the share. Table 4: Expected Annual Investment Requirements, 2014–20 (at 2011 prices)   Capital Maintenance Total   US$Mn %GDP US$Mn %GDP US$Mn %GDP By income group             Low Income 176,217 7.1% 172,089 6.9% 348,306 14.1% Lower Middle Income 35,821 1.1% 72,063 2.3% 107,884 3.4% Upper Middle Income 160,251 2.0% 203,047 2.6% 363,298 2.6%         Developing countries by region        SAR 159,074 7.8% 144,427 7.1% 303,501 14.9% SSA 28,946 3.2% 27,108 3.0% 56,054 6.2% EAP 115,897 2.0% 96,244 1.7% 212,140 3.7% MENA 14,826 1.1% 32,244 2.5% 47,070 3.7% LAC 65,828 1.7% 74,450 1.9% 140,277 3.6% ECA 24,178 0.8% 36,267 1.1% 60,445 1.9%         Developing countries by sector        Electricity 152,577 1.1% 167,754 1.2% 320,331 2.4% Transport 76,201 0.6% 178,818 1.3% 255,019 1.9% Telecoms 123,351 0.9% 63,755 0.5% 187,105 1.4% W&S+WW 20,160 0.1% 36,873 0.3% 57,033 0.4%                     All developing countries 819,488 6.0%     High Income countries 285,045 0.8%     World 1,104,532 2.2%   Source: Authors’ estimation. Note: GDP percentage is calculated based on the projected GDP2015 by income level and by region. For sectors, we used average projected GDP2015 for developing countries. Income level, regions and sectors are defined according to the World Bank’s World Development Indicators Report. 10      Figure 1 shows the distribution of investment requirements by region. South Asia has the largest annual infrastructure investment requirements of $304 billion (14.9 percent of GDP). The investment requirements in ECA, MENA and SSA range from $47-$60 billion annually. Figure 1. Total Annual Infrastructure investment requirements, by Region  350 16.0% US $ Billion  300 14.0%  250 12.0% 10.0%  200 8.0%  150 6.0%  100 4.0%  50 2.0%  ‐ 0.0% ECA LAC MENA EAP SSA SAR Capital Maintenace GDP Percentage Source: Authors’ estimation Figure 2 shows each sector’s total investment requirements, giving an idea of the variation in priorities and requirements between two different investment types (capital expenditure and maintenance cost). Except for the telecommunication sector, the other three sectors have greater percentage in maintenance cost than capital expenditure. Figure 2. Total Annual Infrastructure investment requirements, by Sector Source: Authors’ estimation 11      V. Infrastructure Investment Gap The infrastructure investment gap is estimated as the difference between the estimated investment requirements and actual spending on infrastructure. The lack of data on infrastructure spending has been long recognized as a serious impediment to deriving reliable gaps estimates (Fay et al., 2011). There have been many efforts to calculate figures on infrastructure spending but they have had significant drawbacks (See Annex VI with the description of those attempts). The table below describes total spending in infrastructure by applying GDP deflator for inflation to the latest and most reliable regional data. Table 5: Total Spending in Infrastructure by region in 2011 USD billion. Spending % of Region in 2011 USD Source of figure as reported in Fay and Toman (2010) GDP billion Yepes (2008a) with Original source from Gill and EAP 6.8 356 Kharas (2007) Estimates of India’s public infrastructure spending from SAR 4.2 67 the country’s public expenditure plans and extrapolated to the rest of the region. Estimates from original data collection efforts. Data was LAC 1.2 40.3 collected for 1990–2005 as presented in Yepes (2008b). SSA 3.2 27.4 Foster and Briceño-Garmendia, 2010. No data. Investments share is assumed to be similar to ECA 1.2 34.6 that of Latin America. No data. Assumed to be the weighted sample average. Based on Agénor et al. (2005), Public capital MENA 4.2 46 expenditure in infrastructure (as % of GDP) is estimated as 2 percent. TOTAL 4.3 571.3 Source: Authors’ estimation Note: The developing countries that are included in our dataset are listed by regions: EAP ( Indonesia, Lao PDR, Fiji, Malaysia, Mongolia, China, Papua New Guinea, Cambodia, Vietnam, Philippines, and Thailand); SAR ( Pakistan, Sri Lanka, India, Nepal, and Bangladesh); LAC (Argentina, Uruguay, Costa Rica, Guatemala, Guyana, Trinidad and Tobago, Jamaica, El Salvador, Venezuela, RB, Colombia, Peru, Bolivia, Nicaragua, Honduras, Brazil, Suriname, Ecuador, Haiti, Panama, Paraguay, Mexico, Dominican Republic, Belize, and Chile); SSA (Chad, Niger, Angola, Mauritania, Kenya, Swaziland, Gambia, The, Tanzania, Comoros, Burundi, Ghana, Botswana, Cape Verde, Equatorial Guinea, Liberia, Nigeria, Central African Republic, South Africa, Mali, Zambia, Mozambique, Cameroon, Lesotho, Ethiopia, Congo, Rep., Zimbabwe, Malawi, Uganda, Senegal, Cote d'Ivoire, Sudan, Benin, Madagascar, Namibia, Djibouti, Guinea-Bissau, Guinea, Rwanda, Gabon, Togo, Mauritius, Sierra Leone and Burkina Faso); ECA (Lithuania, Belarus, Russian Federation, Armenia, Estonia, Turkey, Albania, Croatia, Kyrgyz Republic, Tajikistan, Bulgaria, Poland, Azerbaijan, Georgia, Kazakhstan, Slovak Republic, Czech Republic, Romania, Uzbekistan, Latvia, Turkmenistan, Hungary, and Ukraine); and MENA (Tunisia, Iran, Islamic Rep., Oman, Saudi Arabia, Egypt, Arab Rep., Jordan, Morocco, Syria, Yemen, and Algeria) 12      The figure below shows the annual infrastructure investment requirements against actual investments disaggregated over 2014-2020 for the countries included in the sample. Figure 3. Infrastructure investment requirements Vs. Actual for the subset of EMDE countries that were included in the estimation 900 Amount, in 2011 US$(billion) 800 700 600 500 400 300 200 100 0 EAP SAR LAC SSA ECA MENA Total Actual 356 67 40 27 35 46 571 Needs 212 304 140 56 60 47 819 Source: Authors’ calculation for the countries included in the estimation (See Note in Table 2). In order to present figures for all EMDE countries,14 we extrapolated the results of investment requirements and spending as percentage of GDP for the countries not included in our estimation due to data limitations.15 We apply the estimated regional investment requirements (as percentage of GDP) to the countries not included in the estimation. Figure 4 illustrates infrastructure investment requirements for EMDE countries and disaggregates the results by region. The infrastructure investment requirements for EMDE countries is $836 billion per year over 2014-2020. The estimated actual spending is $580 billion per year. This estimated gap only includes the following sectors: roads (paved and unpaved), ports, railways, water, sanitation, waste water management, electricity (generation and distribution), telephone and mobile lines. As presented in the figure below, South Asia has the highest investment gap. The opposite occurs in East Asia and the Pacific as China drives up the average. In that region, investments actually exceed requirements to keep up with economic and                                                                 14 The definition of EMDE country list comes from two parts: Emerging market countries come from IMF World Economic Outlook Update: http://www.imf.org/external/pubs/ft/weo/2012/update/02/index.htm, and Developing economies list come from WBG: http://data.worldbank.org/about/country-and-lending-groups. 15 List of countries not included in the regression due to limited data: EAP : American Samoa, Kiribati, Korea, Dem. Rep., Marshall Islands, Micronesia, Fed. Sts, Myanmar, Palau, Samoa, Solomon Islands, Timor-Leste, Timor-Leste, Tuvalu, Tonga, Vanuatu; SAR: Afghanistan, Bhutan, Maldives; LAC: Belize, Bolivia, Grenada, St. Lucia, St. Vincent and the Grenadines SSA: Botswana, Cabo Verde, Côte d'Ivoire, Eritrea, São Tomé and Principe, Seychelles, Somalia ECA: Kosovo, Macedonia, FYR, Moldova, Montenegro, Serbia MENA: Djibouti, Lebanon, Libya, Tunisia, West Bank and Gaza  13      demographic growth. However, if we exclude China from the EAP estimation, the rest of EAP countries do have a positive investment gap. Figure 4. EMED Infrastructure investment requirements Vs. Actual 900 800 Amount, in 2011 US$(billion) 700 600 500 400 300 200 100 0 EAP EAP* SAR LAC SSA ECA MENA Total Actual 356 35 68 41 28 35 52 580 Needs 212 87 309 141 58 62 53 836 Source: Authors’ estimation Note: EAP*: Exclude China’s data from EAP region. VI. Main Conclusions This study presents estimations of investment requirements over 2014-2020 accounting for the most recent changes in stock of infrastructure. The annual infrastructure investment requirements for EMDE countries is $836 billion or 6.1% of current GDP per year over 2014-2020. The table below shows the investment gap for EMDE excluding China.16 The annual infrastructure investment gap for EMDE countries is $452 billion over 2014-2020, which implies that EMDE countries should almost double their current spending.                                                                 16 China is overinvesting in infrastructure, and including it in the calculation of the infrastructure investment gap will underestimate the gap for the rest of EMDE countries since investment is done at country level.  14      Table 6: EMDE* Infrastructure investment requirements, Actual Spending and Investment Gap in annual US$ billion over 2014-2020    EAP*  SAR  LAC  SSA  ECA  MENA  Total  Requirements  87  309  141  58  62  53  711  Actual  35  68  41  28  35  52  259  Gap  52  241  100  30  27  1  452  Source: Authors’ estimation Note: (*) EAP* includes EMDE in EAP excluding China. The breakdown between capital expenditure and maintenance cost shows that roughly half of the estimated amount should be spent on maintenance. When results are disaggregated by region, SAR has the biggest gap followed by LAC and EAP (without China). It is important to note that the results are less robust when data are further disaggregated as the quality of data varies significantly across regions. Our estimates are lower than the most recent extrapolation presented in the literature. While our methodology has some drawbacks, other available estimates have used extrapolation techniques not accounting for infrastructure stock changes in developing countries and are based on a very small sample of countries mainly in the developed world. The estimation of this study should be used to complement the existing ones and it should be considered as a lower bound of the actual global requirements since, due to data limitations, it does not include the cost of upgrading and rehabilitating existing infrastructure in bad condition (except for roads where the regional percentage of road in bad conditions is available). Moreover, this estimation does not include sectors such as airports, urban transport, energy transmission, and solid waste management that could account for a significant cost. Expenditures related to GHG mitigation and adaptation to climate change could add some $170-$220 billion per year to the cost of developing country infrastructure.17 The table below compares the estimations of this paper with the latest estimates in the literature. While the methodologies are not strictly comparable, it offers a sensitivity analysis around estimates that are highly constrained by data availability.                                                                 17 From Fay and Toman (2010): This includes some $30-40 billion for adaptation in infrastructure (World Bank 2009a) and some $140-175 billion associated with mitigation in the energy sector (World Bank 2009b). 15      Table 7: Global Estimates Comparison Global Estimates (USD billion) $836 Ruiz-Nuñez&Wei, 2015 $2,000 Bhattacharya, 2013 $1,250-$1,500 WEF, 2013 Developing $900-$1,100 Fay, 2011 $1,113 Yepes, 2008 $465 Fay&Yepes, 2003 $3,200-$3,700 McKinsey, 2013 Global $2,680-$3,200 Kennedy&Corfee-Morlot, 2013 $2,650 OECD, 2006 Source: Authors’ summary   16      REFERENCES Agénor, Pierre-Richard, M. K. Nabli, and T. Yousef, 2005, “Public Infrastructure and Private Investment in the Middle East and North Africa”. World Bank Policy Research Working Paper 3661, July 2005 Agénor, Pierre-Richard and Blanca Moreno-Dodson, 2006. ‘Public Infrastructure and Growth: New Channels and Policy Implications’, World Bank Policy Research Working Paper 4064. Andrés, L., D. Biller, and M. Herrera Dappe, 2013. Reducing Poverty by Closing South Asia's Infrastructure Gap. World Bank, Washington, DC. Extracted from http://www.worldbank.org/content/dam/Worldbank/document/SAR/Reducing%20Pov erty%20by%20Closing%20South%20Asia's%20Infrastructure%20Gap_Web.pdf on June 15, 2013 Bhattacharyay, B. 2010. Estimating Demand for Infrastructure in Energy, Transport, Telecommunications, Water and Sanitation in Asia and the Pacific: 2010-2020. ADBI Working Paper 248. Tokyo: Asian Development Bank Institute. Available: http://www.adbi.org/working- paper/2010/09/09/4062.infrastructure.demand.asia.pacific/ Bhattacharya A and M Romani, 2013, Meeting the Financing Challenge for Sustainable Infrastructure’,Paper prepared for the Global Green Growth Summit 2013 ‘The Future of Green Growth: Finance, Innovation and Policy’, Incheon City, 10– 11 June. Calderón, C. and Servén, L. , 2004a, “Trends in infrastructure in Latin America, 1980-2001”; Central Bank of Chile, Working Paper No. 269. Calderón C., Servén L., 2004b, ‘The Effects of Infrastructure Development on Growth and Income Distribution’, World Bank Policy Research Working Paper No 3400. Calderón C., Servén L., 2008, ‘Infrastructure and Economic Development in Sub- Saharan Africa’, World Bank Policy Research Working Paper No 4712. Calderón C., Servén L., 2010, ‘Infrastructure in Latin America’, World Bank Policy Research Working Paper No 5317. Chatterton I. and Puerto O. , 2006, ‘Estimation of infrastructure investment requirements in the South Asia region’, World Bank Estache A., E. Ianchovichina, R. Foucart, G. Garsous and T. Yepes., 2013. ‘Job creation through infrastructure investment in the Middle East and North Africa’, World Development, Volume 45, Pages 209-222. Fay, Marianne. 2000. "Financing the Future: Infrastructure Needs in Latin America, 2000-05." World Bank Working Paper No. 2545. The World Bank, Washington DC. http://econ.worldbank.ora/resource.php Fay, Marianne & Yepes, Tito, 2003. "Investing in infrastructure: what is needed from 2000 to 2010?,"Policy Research Working Paper Series 3102, The World Bank. 17      Fay, M., A. Iimi, and B. Perrissin-Fabert, 2010. Financing Greener and Climate Resilient Infrastructure in Developing Countries: Challenges and Opportunities., EIB Papers 15 (2): 35–59. Fay, M., M. Toman, D. Benitez and S. Csordas (2011), Infrastructure and Sustainable Development. In: Fardoust, Shahrokh,Yongbeom Kim, and Claudia Paz Sepúlveda (eds.) Postcrisis Growth and Development: A Development Agenda for the G-20. http://siteresources.worldbank.org/DEC/Resources/PCGD_Consolidated.pdf Foster Vivien and Cecilia Briceño-Garmendia, eds., 2010. ‘Africa’s Infrastructure: A Time for Transformation’; International Bank for Reconstruction and Development and World Bank, 2010. Gill Indermit and Homi Kharas., 2007. ‘An East Asian Renaissance: Ideas for Economic Growth; The World Bank. Ingram, Gregory K. and Marianne Fay. 1994. "Valuing Infrastructure Stocks and Gains from Improved Performance." Background Paper for the 1994 World Development Report. The World Bank, Washington, DC. International Energy Agency (2014). World Energy Investment Outlook, June 2014 http://www.worldenergyoutlook.org/investment/ Kennedy, Christopher and J. Corfee-Morlot, 2013. ‘Past performance and future needs for low carbon climate resilient infrastructure– An investment perspective’ Energy Policy 59 (2013) 773–783 Kohli H., and P Basil (2011). Requirements for Infrastructure Investment in Latin America Under Alternate Growth Scenarios 2011–2040. Global Journal of Emerging Market Economies. McKinsey Global Institute, 2013. Infrastructure Productivity: How to save $1 trillion a year, McKinsey Infrastructure Practice, January. OECD., 2006. ‘Infrastructure to 2030: Telecom, Land transport, Water and Electricity’. June 2006. OECD., 2007. ‘Infrastructure to 2030: Mapping policy for electricity water and transportation’ Volume 2 and ‘Infrastructure to 2030: Telecom, Land transport, Water and Electricity’. Volume 1 OECD., 2011. ‘Strategic Transport Infrastructure Needs to 2030’.  Perrotti, D. y Sánchez, R. (2011). La brecha de infraestructura en América Latina y el Caribe. Santiago: CEPAL. Schwartz J., L. Andres & G. Dragoiu, 2009. ‘Infrastructure Investment, Employment and the Expectations of Stimulus’, Journal of Infrastructure Development (12/2009) and Policy Research Working Paper, 5009, The World Bank, July 2009. 18      Straub S., 2008. ‘Infrastructure and Growth in Developing Countries: Recent Advances and Research Challenges’, World Bank Policy Research Working Paper No 4460. United Nations, 2012. “Financing for sustainable development: Review of global investment requirement estimates” Working paper prepared for the UNTT Working Group on Sustainable Development financing http://sustainabledevelopment.un.org/content/documents/2096Chapter%201- global%20investment%20requirement%20estimates.pdf UNTT Financing for sustainable development: Review of global investment requirement estimates World Bank. 2009a. ‘Economics of Adaptation to Climate Change’. The World Bank, Washington, DC. Word Bank. 2009b. ‘World Development Report 2010: Development and Climate Change’. The World Bank, Washington DC. World Bank, 2013a, ‘Long-term Investment Financing for Growth and Development: Umbrella Paper’, Presented at Meeting of the G20 Ministers of Finance and Central Bank Governors, February 2013, Moscow, Russia. World Economic Forum, 2013. ‘The green investment report. The ways and means to unlock private finance for green growth’. Geneva, Switzerland Yepes, Tito., 2008a. “Investment Needs for Infrastructure in Developing Countries 2008-15”, unpublished working paper of the World Bank Yepes , Tito., 2008b. “Trends in Investment in Infrastructure in Latin America”. Working Paper, World Bank,   19      ANNEX I Methodologies for estimating investment requirements18 The methodologies for estimating investment requirements can be classified as “top- down” and “bottom-up.” The “top-down” approach utilizes econometric analysis techniques to quantitatively estimate national infrastructure needs and, by extension, regional need. It involves creating an econometric model that could be used across countries to project physical capacity needs per sector during a determined period (most of the studies are based on Fay and Yepes (2003)). After the projections of the physical capacities were derived, standard unit costs based on international “best practice” norms were applied to estimate the investment requirements for new capacity. In addition, it projected the investments required to maintain or replace the existing stock is included. Within the “top-down” approach, there are two econometric technics are used to estimate requirements. (1) Kohli et al. (201X) developed a structural equation model with instrumental variables implemented using a simultaneous two-stage least-square regression frame work. A structural equation model treats certain variables as endogenous to the system a remaining variables as exogenous. (2) Computable General Equilibrium (CGE) models are commonly used tools for quantifying the costs and benefits of policy. One group of CGE model applications were practically implemented with the focus of modeling the joint production of the goals at the country level for MDG costing estimates (Reddy and Heuty, 2005, 2006). For instance, the World Bank’s MAMS model provides a general equilibrium framework for countries to simulate the effect of improvements in one MDG on progress in others (Bourguignon et al., 2008). The “bottom-up” approach reviews infrastructure investment demand at the project level, specifically for regional or cross-border projects. The “bottom-up” methodology is a conservative approach that identifies individual infrastructure projects and estimate costs of their implementation. The compiled project information is then used to obtain the total infrastructure service demand by region, by sub-regional program, and by sector.                                                                 18  Bhattacharyay (2010) and United Nations (2012).   20      Annex II Challenges to compare estimates19 The United Nations (2012)’s report points out to the main challenges to compare estimates across different methodologies: Data issues: Data in infrastructure are very limited not only for stock of capital but also for quality, making the analysis more challenging. Data are in many cases not standardized across countries, making international comparison very difficult. Varying scope: The definition of infrastructure and of the sectors varies and there is not a well-accepted definition. For example in the energy sector, some models focus on electricity production, others on broader definitions of the energy system. The types of technologies considered differ and therefore their costs. Choice of the baseline: Estimating additional investment requirements under any scenario requires the choice of a baseline. This choice has a critical importance on the “needs” or requirements that will be estimated. There is no a clear agreement on the BAU assumptions. Choice of goals and targets: Modelers producing estimates of costs or investment requirements usually do not choose exactly the goals or same targets in their scenarios. Inconsistency: Cost estimates obtained from different models generally cannot be added. Methods used differ in the sectors included (usually due to data availability), in the assumptions used, as well as, in the degree of detail to which specific policies are modeled. This can affect estimated needs several-fold. Non-additivity: The cost or implied investment requirement of reaching a set of goals is not equal to the sum of the costs of reaching the goals separately. Moreover, targets may not be compatible to each other.. Differing time scales: A last caveat to keep in mind when comparing estimates is the differing time scale of various goals and targets. Access to a particular target could be reached in a decade while changing the technology used could take decades.                                                                 19  United Nations (2012) 21      Annex III Methodology to estimate Investment Needs This paper uses Fay and Yepes (2003)’ model and tries to address some limitations in the methodology. The model presented in Fay and Yepes (2003) estimates future demand for infrastructure, where infrastructure services are demanded both as consumption goods by individuals and as inputs into the production process by firms. On the consumption side, the amount of services demanded is a function of income and prices: Ijc = f(Yj; qI ) (1) Demand for a particular type of infrastructure service I by individual j is a function of j’s income, Yj, and the price of infrastructure service I, qI. Aggregating over the population, national per capita demand of infrastructure service for consumption, Ic, will then be given as: Ic 1 Y  j I c j  F ( ; qI ) (2) P P P where Y/P is income per capita. On the production side, the demand of each individual firm for infrastructure service i will be based on a profit maximization decision, which yields the usual first order condition: Yi q (3)  I  I ip wi where Yi is output of good i by the firm, and wi is the price of that good. To go any further, a specific functional form for the production function needs to be adopted. Assuming a Cobb-Douglas production function, the paper can rewrite the first order condition as: qI K i L  1 i  Ii  (4) wi where K is physical capital (excluding infrastructure), L is labor or human capital, and I is the flow of infrastructure services consumed by the individual firm in the production of good i. Solving for Ii yields the derived demand for infrastructure services of firm i: 1  w  (1 ) I ip   i Ki L i  (5)  qI  Aggregating over all firms yields the following: 22      1  w  (1 ) I  I i    i K i L p p i  (6) i i  qI  The derived demand for any given infrastructure service Ip is the sum of the weighted demand of individual firms. However, the usefulness of equation 6 is limited because there is no firm- level data. A reasonable proxy for firms’ aggregate demand for infrastructure is given by aggregate output. However, it is unlikely that the elasticity of demand for a particular infrastructure service, , is the same across all sectors of the economy. Thus the weight attributable to the demand of a given firm depends on the sectoral composition of the economy. Also, as technology changes,  may change. Finally, the weighted average of the relative price wi/qI can be proxied by the real price of the infrastructure good—qI/w, where w is the price level. The reduced form of equation 6 is then given as:  w   Y , q , Y AG , Yind ; A  I p F   (7)  I  where Y is aggregate output, YAG and YIND are the share of GDP derived from agriculture and industry, and A is a term representing technology level. Combining equations 2 and 7, and expressing infrastructure demand in per capita terms, yields the following equation for overall production and consumption demand for infrastructure services: I Y q  F ( ; I ; Y AG ; Y IND ; A ) (8) P P w Estimating infrastructure demand empirically To estimate the flow of services from the endowment of physical infrastructure, data on stocks of infrastructure is used, assuming that services are proportional to the physical stock (though intensity of use may vary). Therefore, equation (8) can easily be understood as demand for physical stocks of infrastructure. Lacking measures of technological change or actual real prices of infrastructure services, time dummies and country fixed effects are used as a proxy. The country fixed effect allows each country to have a different intercept that, combined with the time dummy, allows to capture (albeit roughly) the price variable. The purpose of this paper is not to establish a causal relationship between infrastructure stocks and various economic variables. Instead, the objective is to use this regression for projection, and therefore we needed to obtain the best fit possible and the highest explanatory power. Thus, since infrastructure stocks tend to change reasonably slowly over time and have a long life span, we include lagged value of the dependent variable in the regression in order to increase explanatory power. Fay and Yepes (2003) estimate equation (8) as follows: 23      j j Ii,t  0  1Ii,t1  2 yi,t  3 Ai,t  4Mi,t  5Di  6Dt  i,t (9) where all variables are in natural logs to linearize the model, Iji,t is demand for infrastructure stock of type j in country i at time t; Iji,t-1 is the lagged value of the infrastructure stock; y is income per capita; A is share of agriculture value added in GDP; M is the share of manufacturing value added in GDP; Di is a country fixed effect; Dt is a time dummy; and  is the error term.20 Most infrastructure goods are provided through networks so that the price of the service is often reduced with higher population density. Urbanization, in particular, allows easier and cheaper access to electricity and telephone service. Average costs of water and sanitation actually tend to be higher in urban areas, but this is because the standard service offered there is typically much higher. However, access is always much higher in cities, partly because of the higher income of the population and partly because of public health considerations that make piped water and reasonably sophisticated sanitation services necessary. In the case of roads, kilometers per capita tend to decrease with higher population density. Therefore, equation (9) also includes urbanization and population density to capture the density effect and its impact on demand (both direct and through price). Using ordinary least squares regression with fixed effects, both the basic model described by equation (9) and an extended model that included density and urbanization was run on all five infrastructure variables. Country fixed effects proxy for differences in technology and price across nations. Their use also allows us to obtain consistent parameter estimates. Canning (1998) shows that per capita infrastructure levels are nonstationary, which implies that running the regressions in levels may produce misleading results unless the variables used in the regressions are cointegrated. Unfortunately, cointegration would not yield an easy system with which to make predictions, leaving us with two possible solutions. One is to run the regressions on first differences, which Canning shows to be stationary. This would reduce our sample size considerably since we only have up to eight time-series observations, and the series are often incomplete. The second possibility—which we use—is to include fixed effects. Kao (1997) shows that in this case the estimates of parameters are consistent even if the estimated relationship is not cointegrating. In this paper, the author use the updated Yepes(2008)’s refined methodology and take into cosideration of the infrastrucure qaunlity. For example in transport sector, the maintenance component is obtained by applying a depreciation rate to the predicted total value of the stock in kilometers. This is the same method used in Fay and Yepes (2003). Here, however, the depreciation comes from a stylized model of a maintenance and rehabilitation program. Though a maintenance program is a complex dynamic process associated with level of traffic, geography, and other factors, we rely on fixed initial parameters and an assumption of equal                                                                 20  Manufacturing rather than industry was used here because industry includes mining, which has very different  implications on the demand for infrastructure.    24      levels of traffic among regions to justify the same cycle of routine and periodic maintenance. Below, we argue that a reasonable model of maintenance and rehabilitation over a 10-year horizon can be collapsed into a time invariant annual maintenance and rehabilitation unit cost. The cost factor is obtained from the following equation: AMUC   % Kr  % Bad i , j / 2   RoutineAUC       PeriodicAU C  % Kp  i, j   j     j   (10)   Re habilitati onAUC    % Kh  % Bad i , j / 10      j   Where i indexes regions j indexes types of roads (paved or unpaved). At any time period between year 1 and year 10, each kilometer of road is in one of four states: routine, periodic, rehabilitation, or bad. The time zero fraction of type j roads in bad state is denoted by %Bad . A bad road is one that is 16 years old or older. At any point in time, a given kilometer of good roads might be under routine or periodic rehabilitation maintenance. Rehabilitation maintenance is applied to good roads that are 15 years old. The fraction of 15- year old, good roads, is denoted by % Kh . Good roads that are 5 years old undergo periodic maintenance. The fraction of good roads that are 5 years old is denoted by % Kp . All good roads have to undergo routine maintenance every year except at years 5 and 15, when they receive one of the other two types of maintenance. Bad roads do not receive routine or periodic maintenance, as this would be a waste of resources. To account for bad roads, we assume that there is a special rehabilitation program to reduce the percentage of bad roads to zero over a 10-year span. The program operates by performing rehabilitation maintenance on 10 percent of the time zero stock of roads that are bad. As bad roads undergo rehabilitation and enter the normal maintenance cycle for good roads, some of them will undergo routine and periodic maintenance. We simplify things by ignoring periodic maintenance of these roads and noting that, on average, half of the initial number of bad roads will undergo routine maintenance over the 10-year period. Furthermore, if we assume that the growth rate of kilometers prior to time zero was constant, then the fraction of roads in each state will be constant in the absence of the special rehabilitation program. This completes the necessary assumptions for equation 10. Levels of quality used in these estimates were collected for this study from different World Bank sources. 25      ANNEX IV Data Sources GDP in constant USD of year 2000, comes from the World Development Indicators (WDI) of the World Bank (http://www.worldbank.org/data/.) Agriculture share and manufacture share of value added come from the WDI database of The World Bank (http://www.worldbank.org/data/.) Total population and urban population, in percentage are from the United Nations Population Projections (http://www.un.org/popin/wdtrends.htm) Historical data on telephone lines, paved roads and rails come from Canning (1998) for 1960 to 1995. See data section for information on data combination. Data are available at: http://www.worldbank.org/html/dec/Publications/Workpapers/WPS1900series/wps1929/canni ng1.xls Containerization, as the total number of trafficking containers is obtained from the Containerization International Yearbook (1970-2006), published by Containerization International (http://www.ci-online.co.uk). Telephone lines, mobile phones (in subscribers per 1000 inhabitants), paved roads and rails (in thousands of kilometers), come from the WDI database of The World Bank (http://www.worldbank.org/data/.) Electricity generating capacity in millions of kilowatts per hour is from the US Energy Information Administration, (http://www.eia.doe.gov/neic/historic/hinternational.htm). Electrification rate, measured as the fraction of population with access to electricity, is obtained from Annex B of the World Energy Outlook 2006, published by the International Energy Agency (www.iea.org/w/bookshop/pricing.html). Access to improved water and sanitation in urban and rural areas is defined as fraction of total population with access to services. It comes from the Joint Monitoring Program (JMP) for water supply and sanitation of the World Health Organization and UNICEF (http://www.wssinfo.org/en/watquery.html). Waste Water treatment, measured as the fraction of the population connected to public waste water treatment plants in 2004, is obtained from the United Nations Statistical Division (http://unstats.un.org/unsd/environment/wastewater.htm 26      ANNEX V RESULTS FROM EMPIRICAL ESTIMATION All relevant variables in logarithms Standard errors in parentheses *** p<0.01, **p<0.05, *p<0.1 27      ANNEX VI Efforts to calculate spending on Infrastructure When calculating the actual spending on infrastructure, the authors faced significant challenges. The data on infrastructure spending is very limited and there is no common statistical standard. There have been many efforts to calculate these figures but they have significant drawbacks. In this part, we illustrate relevant research results, which we adopted for different regions, to address the dataset resources of infrastructure spending. We relied on the shares of GDP reported in Yepes 2008 for East Asia and the Pacific (6.8%; original source: Gill and Kharas 2007); for South Asia (4.2%) we used estimates of India’s public infrastructure spending from L. Andres (Pers. Com.); for LAC (1.2%) from original data collection efforts. These shares were applied to 2011 GDP figures from the World Development Indicators. Figures for Sub-Saharan Africa are from Foster and Briceño-Garmendia (2010). For ECA we assumed investments share was similar to that of Latin America; for MENA, where no data was available either, we used the weighted sample average (4.2%). No data are available on how much is spent on maintenance except in Africa where O&M represents 45% of total expenditure on infrastructure (Foster and Briceño-Garmendia 2010). 28