WPS7987 Policy Research Working Paper 7987 Rising Incomes and Inequality of Access to Infrastructure among Latin American Households Marianne Fay Stéphane Straub Sustainable Development Global Practice Group February 2017 Policy Research Working Paper 7987 Abstract This paper documents access to services and ownership of only. In addition, few country fundamentals appear to be infrastructure-related durables in the water, energy, telecom, significant in explaining this variability, pointing to policy and transport areas, based on harmonized household survey differences as an important determinant. The paper derives data covering 1.6 million households in 14 Latin Ameri- the income elasticity of infrastructure access for the full can countries during 1992 to 2012. The paper provides set of indicators, and uses these to estimate the time that a systematic disaggregation of access and ownership rates would be needed to close the remaining gap for households at different levels of income and over time, and econo- at different levels of the income distribution under a “busi- metrically derives the country infrastructure premium, a ness as usual” hypothesis. Under that scenario, universal measure of how much a household benefits from simply access appears to be decades away for many countries in the being located in a given country. The results show extensive region. The last part discusses the policy challenges, argu- inequality of access, within countries across the income ing that in a context in which public budgets face strong distribution, but also across countries for households at constraints and significant increases in private investment similar levels of income. For water and electricity, for are unlikely to be forthcoming, a large part of the solu- example, up to two-thirds of the variability in individual tion lies in refocused investment strategies, better demand percentile access to infrastructure services and consump- management, and improved public spending efficiency. tion of related assets can be explained by country residence This paper is a product of the Sustainable Development Global Practice 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 mfay@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 Rising Incomes and Inequality of Access to Infrastructure among Latin American Households Marianne Fay and Stéphane Straub1 JEL Codes: H54, O54, D12 Keywords: Infrastructure, Latin America and the Caribbean, Household Consumption                                                              1 Marianne Fay: World Bank. Email address: mfay@worldbank.org. Stéphane Straub: Toulouse School of Economics, University of Toulouse Capitole (IDEI and IAST) and World Bank. E-mail address: stephane.straub@tse-fr.eu. We thank Charles Fox and the World Bank LAC Stats team for help with data access, María-José Graña for outstanding research assistance, and Uwe Deichmann, Augusto de la Torre, and Stephane Hallegatte, and participants at the World Bank ‘Unbaked and Wired’ Infrastructure workshop for useful comments and suggestions. 1. Introduction In Latin America, as in large parts of the developing world, aggregate indicators show that access to infrastructure services is still largely rationed, with both households and firms suffering from shortfalls across all major types of infrastructure (Estache and Fay 2009). This situation is likely to have been exacerbated over the last two decades by the rapid growth in the share of households attaining middle-income status and the growth in demand for infrastructure services they generated (Ferreira et al. 2013, Lakner and Milanovic 2015). However, despite its potential importance in designing infrastructure policy, there is no systematic information on how access to infrastructure varies across the income distribution. This paper addresses this gap by documenting access to services and ownership of infrastructure-related durables in the water, energy, telecom, and transport areas, based on large scale harmonized household survey data from Latin America covering 14 countries over the 1992 to 2012 period. It provides a systematic disaggregation of access and ownership rates across the income distribution, allowing for an assessment of the situation of households at different levels of income and over time across the countries of the region. For a number of indicators, the data reveal wide disparities across households at similar levels of income in different countries. We then econometrically derive country infrastructure premia, which show how much a household at a specific level of the income distribution benefits from simply being located in a given country, and assess the progressivity or regressivity of infrastructure access patterns across the income distribution. We then show that these premia are only partly explained by per capita GDP and are not systematically aligned with other country overall characteristics such as inequality, density, or urbanization, which means they are very likely to be for a large part policy driven. Second, we compute the income elasticity of infrastructure access for our set of indicators, and use these to forecast the time that would be needed to close the remaining gap for households at different levels of the income distribution under a business as usual hypothesis. The general message is rather somber: without fundamental changes to the pace and efficiency of infrastructure delivery, universal access is in many cases still decades away. We also highlight an asset ownership “pecking order”, by which, as their income grows, poorer households tend to acquire first a fridge, then a washing machine, and finally a car.   2    The last part of the paper discusses the road ahead given the magnitude of the challenge. Based on the observed performance of countries in mobilizing investment in infrastructure, it argues that in a context in which public budgets face strong constraints and large increases in private investment are unlikely to be forthcoming, the solution may lie in other parts of the delivery equation. Specifically, we discuss three promising margins: refocused investment goals, demand management, and interventions to improve public spending efficiency. The paper is organized as follows. Section 2 reviews the related literature and shows how this paper contributes to it. Section 3 presents the data used in the analysis. Section 4 documents the evolution in infrastructure demand throughout the period in the different countries, and the implications for infrastructure delivery looking at the future. Section 5 discusses the policy challenges and concludes. 2. Literature Review For decades, the economic literature on infrastructure has mostly focused on aggregate data at the regional or more often the national level. Traditionally, infrastructure has either been measured through supply-side physical indicators, such as electricity generation capacity or kilometers of roads, or demand-side ones such as aggregate electricity or water connection rates. Given the fact that such data ignore both the spatial nature of and the distributional disparities in infrastructure access, the policy relevance of the conclusions have been limited (see Straub 2011 for a critical review). There were of course micro-econometric studies using household-level data to look at specific local outcomes, such as Gibson and Rozelle (2003), Galiani, Gertler, and Schargrodsky (2005), or Dinkelman (2011), to cite only a few. More recently, the growing use of spatial data has been instrumental in generating a renewal of the field, allowing for example for local (often municipal) level infrastructure usage indicators.2 However, there have been surprisingly few papers using micro-data to analyze systematically how infrastructure demand at the household level varies along the income distribution. As a result, little is known about the way demand for infrastructure varies with income at any point in time, and how it evolves over time as household income grows.                                                              2 See Straub (2008, 2015) for reviews.   3    A recent exception is Gertler, Shelef, Wolfram, and Fuchs (2016), which uses household survey data in several large developing countries to document an S-shaped relationship between the income distribution and both consumer durable ownership and energy use, in the cross-section and over time as income grows. They then make use of a cash-transfer program-generated exogenous shock to incomes in Mexico to establish that such relationship is indeed causal. In a related paper, Davis and Gertler (2015) analyze energy consumption using Mexican household data, looking at both the extensive margin (whether households adopt air conditioning or not) and the intensive margin (the dollar amount of electricity consumption). Lee, Miguel, and Wolfram (2016) use an experiment to derive the demand for electricity connections among poor rural households in Kenya, and find that consumer surplus is far less than total costs at all price levels, suggesting that residential electrification may reduce social welfare—at least in the absence of externalities. Our contribution is first to systematically document infrastructure access and asset ownership across a panel of 14 countries in Latin America over 20 years, in the four main infrastructure sectors. The SEDLAC database provides a harmonized sample of surveys, allowing in particular for the comparison of households at similar levels of income across countries of the region. While the household surveys used here are not following the same households over time, we are able to construct a pseudo-panel by grouping households by income quantiles and looking at the within-quantile evolution of infrastructure access over time. This then allows for the estimation of income elasticities of infrastructure access in a dynamic setting. This enables us to document extensive inequality of access, within countries across the income distribution, and across countries in Latin America for households at similar levels of income. In addition, the scope of the dataset allows us to test the relevance of country fundamentals vs. policies in explaining these differences, and to provide comparative estimates of the time needed to reach universal coverage. This paper focuses exclusively on the extensive margin, i.e., whether households have access to services or own durables, and provides descriptive evidence on infrastructure demand in a context in which household income has experienced strong and sustained growth. It therefore stops short of providing insights on the intensive margin as captured by the value of consumption expenditures on these services. Finally, it does so in the context of potential binding supply constraints on the infrastructure provision side, and credit constraints on the demand side. Addressing these issues is the object of ongoing work.   4    3. Data We use household surveys microdata from the Socio-Economic Database for Latin America and the Caribbean (SEDLAC), developed by The Center for Distributional, Labor and Social Studies (CEDLAS) of the University of La Plata, in partnership with World Bank Latin America and the Caribbean Poverty and Gender Group (LCSPP). Based on the availability of key variables related to infrastructure access and infrastructure- related assets, we build a five-year interval dataset for 14 countries between 1992 and 2012.3 The final dataset covers 1.6 million unique household observations and includes the surveys shown in table 1. Table 2 shows the coverage by country and variable. The main variables of interest are the per capita household income adjusted by purchasing power parity, in 2005 US dollars, and infrastructure. Income here includes labor and non-labor income as well as the imputed rent of owner-occupied housing. Infrastructure is measured through a set of dummies related to access to water, sewerage, electricity, fixed and mobile phones, and internet access. Another set of dummies indicates households´ ownership of infrastructure-related appliances and assets. We focus on the following variables: refrigerators, washing machine, television, computer, car, and motorcycle. The use of access / ownership dummies means that we are bound to concentrate on the extensive margin, i.e., on whether households are connected or use a service or not, but are unable to offer insights on the intensive margin due to the absence of specific expenditure data. Finally, we add a number of standard country-level variables such as GDP per capita, income inequality, population density, and urbanization, all from the World Bank Indicators database. To exploit the panel dimension of the data, we split it into income quantiles, alternatively percentiles or deciles, defined by reference to the regional income distribution.4 We define quantiles corresponding to the region-wide income distribution at each date. In this way, a household at say, the 11th percentile or the 6th decile of the distribution in 2012 is at similar                                                              3 We allow for +/-1 year when the survey for the exact year is not available. 4  Following SEDLAC practice when computing inequality measures, we drop observations reporting zero income (see CEDLAS and The World Bank, 2014).   5    level of income regardless of whether it is from Honduras or Chile for example. In what follows, we refer to this as the global quantiles (deciles / percentiles) distribution.5 This way of defining income class allows for easy comparisons across countries, and for the disentangling of country- vs. individual determinants of households’ situation. Based on this, the next section examines access to infrastructure across countries and at different levels of the income distribution. 4. The Demand for Infrastructure The last decade witnessed significant changes in the shape of the income distribution of developing and emerging countries (Lakner and Milanovic 2015). In Latin America specifically, Ferreira et al. (2013) estimated that the number of people that can be considered as middle class increased by 50% over the 2003 to 2009 period. They define the concept of middle class following a criteria of economic security, whereby a household is considered middle class if it has a probability of falling back into poverty no larger than 10% over a five- year interval. This implies a middle class income threshold of $10 a day, which is at the 68th percentile of the Latin American income distribution in 2009, while the moderate poverty line of $4 leads to 30.5% of the population falling below it. Looking at our data and relying on a consistent sample of 14 countries for which data are available from 2002 to 2012, Latin America was characterized by important income gains across the income distribution during the 2002-2012 period. Indeed, over that ten-year period, the annual growth rate of the incomes of the lowest three deciles was above 5%, that of vulnerable households at above 4%, and that of the top three deciles, which includes the middle class and rich households at almost 3%.6 4.1. Access Patterns in Latin America: The Situation as of 2012                                                              5 We often take the semantic shortcut of referring to households “in a given quantile” when talking of those below a given threshold (e.g., in quantile 1 for those below the first quantile threshold) or between two successive ones (e.g., in quantile 5 for those between the quantiles 4 and 5 thresholds). 6  Household surveys of the type used here typically do not include very rich individuals because of random sampling, non-response, or large under-reporting. This may lead to the figure for the growth of income of the top decile being lower than the actual one, but it is inconsequential for our purpose.   6    This rapid income growth led to a surge in the demand for home equipment and mobility, hence for infrastructure services such as water, electricity connections, and transport, and for related durables such as electric appliances and vehicles. Yet, as of 2012, access and ownership rates still vary widely, as can be seen in Figure 1, which graphs Latin American average access and ownership rates by deciles, as well as average rates in the worst- and best-performing countries of the sample respectively.7 For water, Latin American averages, which go from around 71% in decile 1 to 98% at the top of the distribution, mask wide differences, from an abysmal 19% of the poorest decile households in El Salvador having access to a source of water in the house or lot (a measure akin to basic access as defined in the MDGs) to 99% for households of similar income in Argentina. Beyond El Salvador which could be considered an outlier, three countries, Bolivia, the Dominican Republic, and Peru, have barely more than half of the bottom decile households with access to water. The situation is more critical for sanitation. In 2012, only 1% of households in the poorest decile lived in a dwelling connected to a public sewerage system in Paraguay. With the exception of Chile, where 82% of the poorest households are connected, no country has more than 46% connections, for a Latin American average of 35% in the poorest decile. The range narrows only slightly when moving up the income distribution. At the eighth decile, roughly the lower limit for middle class households according to Ferreira et al. (2013), average Latin American access rates are still only 72%, hovering below 80% for 9 out of 14 countries. Electricity is clearly the dimension in which the region performs best. As of 2012, regional averages start at 89% in decile 1, and reach 99% from decile 6 up. At the bottom of the distribution coverage rates start at 67%, and all countries in the sample are providing above 90% coverage in all deciles, except El Salvador in the first decile, Bolivia in the first two, and Honduras and Peru in the first three. But these rather encouraging electricity connections figures are not exactly mirrored in appliances ownership rates. Overall, less than 80% of households in the bottom three deciles                                                              7 Average Latin American rates are simple averages over the total 420,204 households’ sample. Household survey weights are used, but we do not weight by country population or otherwise. Note however that bigger countries have larger samples: the number of households varies from 4,889 for Paraguay to 111,612 for Brazil, and looking only at 2012, the correlations of country sample size with population and GDP are .67 and .64 respectively.   7    own refrigerators and less than 60 percent washing machines—with access considerably lower in some countries. Finally, despite the frequently heard claim that access to cars has increased widely, and despite increased credit availability in many countries (World Bank 2017), cars are still luxury goods and the average ownership rates remain modest—only exceeding 40% for the top two deciles and below 20% for a large majority of deciles and countries. This is partly compensated by larger ownership rates of motorcycles, particularly for intermediate income groups, from deciles 5 to 8. However, these averages mask important differences across countries, as motorcycle use is mostly concentrated in the Dominican Republic, Colombia, Paraguay, and Uruguay. The numbers above show that household income alone has limited explanatory power when it comes to infrastructure access. But how much of infrastructure access is explained by households’ country of residence, and how does this vary along the income distribution? We start with a simple decomposition, focusing on the situation in 2012.8 Consider that access to a specific type of infrastructure service A in country c is given by: DAic = α + θc + εic, (1) where DAic is the average access to infrastructure asset A, in country c, by households in income percentile i, θc are country fixed effects, α is a constant, and εic is the error term. The results are in Table 3. Each cell in Panel A reports the R-squared from the estimation of (1) using the percentiles of the infrastructure service access / assets indicated at the top of the column. Given the large number of indicators, we discuss the results referring to infrastructure “clusters”, concentrating on indicators for which coverage is large enough: - The water cluster, which includes access to water, sewerage, and a toilet connected to a sewerage system or to a septic tank; - The electricity cluster, including access to electricity and ownership of the following assets: refrigerator, washing machine, and television; - The ICT cluster, which includes landline and cell phones, internet access, and ownership of a computer;                                                              8 See Milanovic (2015) for an application to inequality. Note however that our approach differs in that this paper compares country-level quantiles, which correspond to different levels of income, whereas we use a common income distribution and look at differences in infrastructure access for households at the same level of income.   8    - The transport cluster, including ownership of cars and motorcycles. The results indicate that a large share of the variability in individual percentile access to infrastructure services and consumption of related assets can be explained by country residence only. For the water cluster, country dummies explain between 40 and 67% of the variability of individual percentile-level access across Latin America. The electricity cluster displays an equally strong explanatory power of locational aspects, with country dummies explaining 33% of the variability in access to electricity, and between 54% (for refrigerators), 64% for television, and 66% (for washing machines) of the variability in related assets ownership. The fact that country dummies explain more of the difference in durables than infrastructure access itself, despite the fact that access may be constrained, may be related to differences in access to credit or in cost, such as variation in taxation of durables across countries. The outcomes for the ICT cluster indicate that 16 and 40% of landlines and cell phones variability in access is explained by location respectively, while it is 25% for computers, and 18% for internet access. Finally, in the transport cluster country dummies explain only 10% of the variability in ownership of cars, but this jumps to 85% for motorbikes due to the strong concentration in a few countries noted above. Panel B presents the country premium, as given by the coefficients of the corresponding country dummy, relative to the poorest country in the sample, which in this case is Honduras. This indicates, for each indicator, the average premium related to living in each country rather than Honduras. In Figure 2, we summarize graphically how countries perform given for example their level of development, by plotting the average premium against their 2012 per capita GDP. Honduras is the benchmark, i.e., a premium equal to zero. For water, six countries are at or below the horizontal zero line, thus doing equal or worse than Honduras despite boasting considerably higher per capita GDP. The situation is particularly bad for El Salvador, and to a lesser extent Bolivia, Peru, the Dominican Republic, Ecuador, and Paraguay. The other group of dots above the horizontal axis corresponds to more virtuous   9    countries, although Colombia, Costa Rica, Uruguay, and Argentina appear to fare relatively better than Mexico and Chile.9 This tale of several groups of countries also characterizes the regional sanitation panorama, although the composition of these groups is slightly different, with Costa Rica, the Dominican Republic and Paraguay now the strong underperformers, Brazil, Mexico, and Uruguay in an intermediate and rather weak position, and Bolivia, Ecuador, Peru and Colombia performing well. Finally, for electricity, the region displays overall a much smoother concave picture, with the notable exception of Peru, whose poor performance makes it an outlier. In addition, the use of household survey micro data allows for an unprecedented window into the relative treatment of households at different levels of the income distribution within countries. The comparison of within-country premia for each decile tells us whether the current access situation is progressive, meaning that a given country does relatively better delivering infrastructure to the lower rather than the upper part of the distribution, or regressive (Figure 3). For water, there is again a clear dichotomy, between countries that display a rather regressive pattern (El Salvador, the Dominican Republic, Peru, Bolivia, and Paraguay) and those that display the opposite pattern, especially Colombia, Costa Rica, Chile, and Argentina. Note that to a large extent, these two groups overlap with the ones that emerged from the comparison of per capita GDP and average premium in the preceding figure. What this means is that for water, given the convergence to high access levels at the top of the income distribution, overall underperformance mostly stems from a failure to address service shortfalls for the lowest income quantiles. This situation contrasts with that of sewerage. Indeed, in this case, a few countries fare badly overall, as witnessed by their negative premium in figure 2 above and the fact that the entire set of decile-level premiums are negative in Figure 3, despite boasting a rather progressive pattern. This is the case of Paraguay, Costa Rica, and the Dominican Republic. The only country with a clearly regressive pattern is Peru, whose premium goes from -1% at the first decile, to +27% at the tenth decile.                                                              9 No official figure is available for Argentina’s per capita GDP around 2012. We set it at 20,000$.   10    Finally, a similar story prevails for electricity, where all countries appear to have progressive delivery patterns, except again for Peru and to a lesser extent Bolivia. The fact that location may imply such large differences in access for otherwise similar households speaks to the importance of country-level factors in determining infrastructure rollout. In particular, it begs the question of whether these locational premia can be attributed to some specific country-level fundamental characteristics. To address this, we estimate equation (2): DAic = α + Xicγ1 + εic, (2) where Xic includes country-level per capita GDP, inequality (Gini), population density, and urbanization. Because of the small number of clusters (14 at most), standard errors are computed using wild cluster bootstrapping (Cameron and Miller 2015). The results are in Table 4. The elasticity of infrastructure indicators with respect to countries’ per capita GDP is strong and significant for water (0.43***) and toilet (0.74***), but not for sewerage (-0.75NS). In the electricity cluster, it is strong for electricity access (0.24***), refrigerator (0.75***) and washing machine (1.43) although this last one is imprecisely estimated. In the ICT cluster, it is only significant for internet access (0.56***), but is actually negative and not significant for cell phones (-0.18NS) and computers (-0.09NS). Finally, the elasticity of vehicle ownership is positive for cars (0.23**) and motorcycles (0.45NS). Interestingly, other determinants do not appear to matter much. Inequality as measured through the Gini coefficient is almost never a significant determinant of infrastructure indicators (a higher Gini only marginally increases internet access). Higher density reduces water access, and increases computer ownership. Finally, a higher urbanization rate leads to lower toilet access, lower electricity access, but higher computer ownership. In addition, the comparison of the R-square across Tables 3 and 4 shows that for most infrastructure indicators, there is a large drop in explanatory power when using specific country characteristics rather than generic country dummies. While we must be cautious in interpreting these results, they suggest that the wide disparity in access to infrastructure is only partially the result of countries’ fundamentals. While richer countries have to some extent been more successful in providing infrastructure to their citizens, for example in water, electricity and internet, this has not been the case for other services such   11    as sanitation or phones. The fact that aspects such as the pattern of urbanization, population density, and inequality, i.e., generic aspects that move slowly over time and are unlikely to be much affected in the short term by economic policies, also appears to have little impact suggests that other policy dimensions not necessarily captured by macro indicators may be key to explaining the extension of services in the last few decades. Before discussing these, we turn to the analysis of the dynamic aspects of infrastructure provision. 4.2. The Dynamic View: The Elasticity of Infrastructure Access and Durables Ownership There is a variety of patterns in the evolution of access to infrastructure across types of services and countries in the last two decades. For example, in 1997 water access at similar levels of income differed widely across countries: approximately 80% of households in the first decile of the Latin American income distribution had access to water in Chile, but for households at similar level of income, the rate was around 45% in Brazil, 20% in El Salvador and 10% in Paraguay. Clearly, most of the disparity concentrated in the lowest part of the income distribution, as in all four countries the top decile enjoyed access rates of 90% or more, even in earlier periods. In addition, there were wide differences in how these access rates evolved over time. For example, between 1997 and 2012, Paraguay successfully increased access rates of all income deciles above 60%, and above 80% from the third decile up, while El Salvador displayed much less progress, with only households in the sixth decile and above enjoying access rates of 60% or more. On the other hand, when looking at sewerage access rates, fewer differences are apparent between households in countries with very different levels of per capita GDP, in line with the previous result that the elasticity of per capita GDP is negative and not significant. Surprisingly, more progress appears to have been made in countries such as Peru and Bolivia, especially for households in lower deciles, than in richer countries such as Costa Rica and Uruguay. Electricity offers a slightly more optimistic panorama. Across Latin America, access rates have converged to very high levels. Countries in which two decades ago access rates for lower deciles were as low as 20 to 40%, such as Peru and El Salvador, have increased them to 70% or more   12    across the whole income distribution, while in others like Mexico or Brazil, universal access is now almost the rule. While this has in some cases led to large progress in access to energy consuming devices such as refrigerators or washing machines, different patterns emerge. In Brazil, ownership of refrigerators did increase in lower income deciles following connection to the electric grid, but the process was much slower in Mexico. Similarly, comparing El Salvador and Peru, two countries with similar profiles of electrification, reveals quite different evolutions. Access to refrigerators stagnated and if anything progress was greater at the top of the income distribution in the latter, while the former witnessed large increase in ownership. Next, we use the complete panel, based on the global quantiles distribution, to analyze how the income elasticity of infrastructure access varies across services and countries, and within these, at different levels of the income distribution over the 1992-2012 period. Table 5 presents two sets of results using the household survey data. First, in Panel A, we report an average Latin American elasticity given by the following estimation: DAdct = α + log(income)dct γ1 + θct + εdct, (3) where DAdct is the average access / ownership rate for infrastructure asset A, in country c, by households in income decile d, at time t, log(income)dct is the average decile income, and θct are country-time fixed effects.10 This simple specification assumes that infrastructure demand indicators at the decile level are a function of average income in that decile, and of the price of infrastructure as proxied by country-time fixed effects. The underlying assumption is thus that this price is uniform within countries in a given period, but varies over time. This is much milder than proxying prices through country dummies in the country-level panel, in effect assuming that prices are invariant across time period. In addition, the country-time fixed effects absorb any other relevant country- level determinants, such as the ones discussed above. Second, in Panel B, we differentiate elasticities by country, through the following specification: DAdct = α + log(income)dct γ1 + (log(income)dct.θc) γ2c + θct + εdct. (4)                                                              10 Using the percentiles instead yields very similar results.   13    The elasticity of country c is then given by γ1 + γ2c, where γ2c is the coefficient of the corresponding country interaction. The resulting elasticities of access are in the expected range and strongly significant throughout. For example, the Latin American average water access elasticity is 0.08, and country-specific values vary between 0.02 for Argentina and Costa Rica, and 0.18 for Peru. The values for toilet ownership and sanitation access are higher on average, at 0.16 and 0.17 respectively, and range between 0.07 and 0.25. Similarly, the average Latin American electricity access elasticity is 0.06, and it ranges from 0.01 for Argentina, 0.02 for Costa Rica and Ecuador, to 0.16 for Peru. Interestingly, elasticities for appliances ownership are in general quite larger, with 0.15 averages for refrigerators and washing machines, and ranges between 0.08 and 0.24, indicating faster progress on the intensive margin of consumption. In the ICT cluster, elasticities are high across all indicators. Notably, these are higher for landline phones than cell phones. Finally, the elasticity of vehicle ownership is strong for cars, with a Latin American average of 0.16, ranging between 0.09 in Bolivia and 0.20 in Mexico, but very small for motorcycles, with an average of 0.01 only. Gertler et al. (2016), looking at refrigerators, suggested that the elasticities of asset ownership vary along the income distribution. The Latin American data display a rich pattern, with heterogeneity across assets. Figure 4 shows the rate increase for each decile between 2002 and 2012 for refrigerators, washing machines and cars, as well as the related income elasticity computed separately at each decile. While the lowest income groups have experienced the largest increase for refrigerators, a clear inverted U-shaped pattern emerges for washing machine, with the largest increase in the middle of the income distribution. Finally, car ownership displays the opposite pattern, with access increasing mostly for the richest households. Table 6 offers a summary comparison of the demand elasticity of infrastructure assets and access using different approaches: Latin American average, countries average, country-level elasticities computed at the 10th, 25th, 75th and 90th percentiles respectively, as well as cross- country averages from Fay and Yepes (2003) and some more recent estimations for Latin America from Perrotti and Sanchez (2011). The first immediate observation is that elasticities   14    estimated using micro data from household surveys are smaller than those produced by macro studies. This is especially true for elasticities of electricity access, previously computed in the 0.18 to 0.43 range, while our estimate is at most 0.06 on average and 0.16 for the highest country value (Peru). Similarly, macro estimates for cell phones ownership and internet access are lower than when estimated with household survey data. The closest values are for water and sewerage, although the macro estimates tend to match the highest household data country estimates. It is probable that this disparity is partly due to the ability of disaggregated household data to control adequately for the prices of infrastructure through country-time fixed effects, while macro data can only include time invariant country fixed effects. Consequently, the difference in estimates appears bigger when dealing with services for which the price matters more and has experienced large shifts over the last two or three decades, such as cell phones, internet access, and to a lesser extent electricity. Several factors are likely to be key in understanding the differences in how countries have been able to accommodate the demand pressure on infrastructure stemming from growing incomes. First, demand side issues are likely to matter, for example the extent that credit constraints have limited the ability of households to access infrastructure and related equipment. In addition, on the supply side, it is likely that constraints related to limited investment and institutional capacity are playing a role. Given these likely binding supply constraints, it is enlightening to simply ask what these elasticities mean in terms of the ability of policy makers throughout the region to meet very simple universal access goals, akin to the early MDGs criteria, such as for water “access to a source of water in the house or lot.” Making simple assumptions on future income growth, we can forecast for each country the number of years it would take to offer universal access to a given infrastructure service under a “business as usual” scenario. Consider very simple mechanical projections, assuming a high income growth scenario similar to the one of the 2002-2012 period and the average historical elasticities just discussed (Table 7).11 With the exception of Argentina, which is close to full access according to this criterion,                                                              11 Using average elasticities is a safe approximation given that Table 5 shows they display little variations across income quantiles. Given that the values in Table 5 give “semi-elasticities”, for any value ε, ε/100 is the unit change in access when income increases by 1%. The number of years to close the gap to full access is then given by N = (1-C)/(ε/100.g), where C is the current coverage (between 0 and 1), and g is the income growth rate.   15    the panorama is rather bleak for the lowest income deciles: at current elasticities, countries of the region will take between three and more than nine decades to provide water to the poorest 10% of households. In a number of countries, this extends to higher quantiles and even to the top of the distribution: it will still take between 40 years in Ecuador and Honduras, and 73 years in El Salvador to provide water to households in the fourth decile, i.e., above the poverty line, and from 39 years in the Dominican Republic, to 48 years in El Salvador, and 45 years in Bolivia to serve those in the middle class range, i.e., above the seventh decile. Of course, these projections would be even less optimistic in case of a growth slowdown. Just halving the income growth rate, to 3% annually for the poorest 10% and 2,3% for households just above the poverty rate, would lead to delays in closing the water gap in decile 1 of 28 years in Argentina, 61 years in Uruguay, 72 years in Brazil and Mexico, and 86 years in Peru. Note that the value of the elasticity plays here a crucial role. Consider the three countries with similar coverage rates at the bottom of the income distribution, Bolivia (53%), the Dominican Republic (54%), and Peru (54%). The different income elasticities, of 0.08, 0.12, and 0.18, translate into delays of respectively 43, 64, and 96 years to achieve universal coverage among the lowest decile. Similar exercises for sewerage reveal an even grimmer picture. Even the best performing countries in the region are not on course to achieve universal basic sanitation (defined as a dwelling connected to a public sewerage system) in less than three decades at the seventh, “middle class decile”. At the current rate, and with high income growth, the region’s best performer, Chile, would need 30 years to achieve universal sanitation for the poorest 10% of households and 31 years for middle class ones above the 70th percentile, while Peru, due to its much more regressive pattern, would attain universal connections in 54 years at the bottom deciles and in 13 years for middle class households in the top three deciles. Finally, for electricity, most countries are on track to achieve full connection for richer households (above the 70th percentile) in the next 13 years or less, but delays increase to 22 years for Ecuador and Colombia at the 40th percentile, and to more than 30 years for the poorest decile for half of the countries in the region. This elasticity exercise is in some sense a reduced form way to think of the “last-mile” infrastructure challenge. Obviously, the “business as usual” scenario does not constitute a   16    satisfactory option for governments in the region. The question becomes to identify the realistic components of a successful infrastructure strategy that cuts down on the time needed to achieve the necessary improvements. The next section addresses this issue. 5. The Policy Challenges For a start, let us think of the potential gains to be had from speeding up improvements in infrastructure access. Taking again the example of water, and raising the income elasticity to 0.2 (that is above the current regional best performer) would allow the 9 best performing countries to connect all decile 1 households in less than 21 years. An even higher elasticity of 0.5 would allow all countries, except El Salvador, to complete that task in 16 years or less. The elasticity used in our reduced form exercise is in fact the outcome of a model combining amount invested and efficiency of such investments on the supply side, with amount demanded as a result of price and quality options available on the demand side. However, the conventional wisdom has generally taken investment amounts as the main, if not the only, policy lever. Unfortunately, experience shows that overall investment levels are unlikely to adjust significantly in the short run. As discussed in World Bank (2017), governments in the region face tight fiscal constraints. In the last decade, the combination of the commodity price bonanza, very low international interest rates, real exchange rate appreciations, and rising consumer credit, allowed major commodity exporters to rely on terms-of-trade windfalls to expand domestic demand, in particular through fiscal spending. As the recent, sharp and likely durable deterioration of their terms of trade brought down dramatically the purchasing power of their incomes, these countries (which include virtually all the major countries in South America) are now forced to reduce domestic spending. The current level of total public investment is 3.4 percent. On average, about a third of this public investment goes to infrastructure. As for public-private partnerships (PPPs), the best figures available, from the World Bank PPI database, puts them at about 40 percent of Latin America’s infrastructure investments. However, an often-disregarded fact is that they depend heavily on government support. New figures that have been gathered starting in 2015 show that about a third of the recorded investments actually come from public sources, and about half are   17    backed by some type of government guarantees.12 In that way, constraints to public investment spill over to private investment in infrastructure. In summary, considering average levels of PPP investments of between 0.5 and 1 percent of GDP, and public infrastructure investments of 1 to 1.5 percent, Latin American average investment in infrastructure is unlikely to exceed 1.5-2.5 percent of GDP in the near future, very much in line with current levels.13 This leaves investment efficiency improvements, including refocusing spending “needs”, reducing costs, improving cost recovery, and managing demand as the only realistic avenues to speed up infrastructure delivery in the short and medium run. In terms of setting adequate targets or goals, consider the cases of water and electricity, highlighted in World Bank (2017). By one estimate (Hutton and Varughese 2016) achieving universal access to basic water, sanitation, and hygiene by 2030, in line with the early Millennium Development Goal (MDG), would require an investment of only 0.05 percent of GDP a year through 2030. On the other hand, achieving the much more ambitious Sustainable Development Goal (SDG) of universal access to safely managed water and sanitation services by 2030 would cost five times more, at 0.25 percent of GDP, an amount equivalent to current Latin American spending on water and sanitation. The potential for demand management, through better pricing schemes, incentives and behavioral nudges, is probably large. The experience of Brazil in 2001, when the drought- induced energy crisis led to the implementation of a large scale electricity saving program, is a case in point (Costa and Gerard 2015). Faced with strong incentives, including quotas and penalties for exceeding them, households responded sharply, reducing electricity consumption. More interestingly, the policy led to a permanent and stable reduction in average electricity use of 11% four years later, arguably through a change in consumption habits. The possibility of hysteresis, whereby temporary policies may have permanent effects by shifting agents to a different equilibrium, has clearly been underestimated when defining policy goals in the context of infrastructure investments, which lock in behaviors over a long period of time. Finally, boosting the efficiency of investment spending should be high on the agenda. It has been known at least since Pritchett (2000) that (cumulated) investment flows may translate into                                                              12 See https://ppi.worldbank.org/ 13 See http://infralatam.info/   18    very different effective capital stocks, and hence services, depending on the environment in which they take place. Among the reasons for this discrepancy are government inefficiency, corruption, or departures from efficiency for redistributive motives among others. Given this, improving public investment efficiency can have direct effects on infrastructure delivery. On the cost side, there is first an ex ante aspect, stressed in the public investment management (PIM) model (IMF 2015). A first step towards useful policy recommendations would be to clarify what the relevant investment efficiency model is, something that is likely to be very context dependent, and to collect relevant empirical evidence to test this model. Considering the two main stages of the investment process corresponding first to the planning stage, in which projects are selected and their characteristics (size, timing, technology) are decided, and second the execution stage, in which they are carried out, this means understanding how the different stages interact, depending on the sector, the level of institutional capacity, and the level of development. For example, contrary to what is claimed in IMF (2015), focusing on the efficiency of delivery in the poorest countries is likely to be a misguided strategy. Indeed, if the two stages are strongly complementary and planning capacity is low, such efficiency gains will be useless from a social welfare point of view if the wrong projects are selected to start with. In addition, other aspects are relevant here, such as competition policy, to the extent that competition in downstream sectors such as construction and related activities is likely to have an important impact on ex ante costs. Next, there is a post construction stage of the cost issue that relates to operational efficiency. This is a crucial aspect that has much more to do with the political economy of running and delivering the service. The focus in that case should be on utilities managers’ incentives, as regard in particular their relationship with politicians, soft-budget constraint issues, and consequences for example in terms of overstaffing and clientelism. Addressing this part of the cost structure separately may be more fruitful, and would entail in particular to look at the quality of the regulatory framework. Finally, there is much talk in policy circle about public investment efficiency improvements as catalysts for additional and sustained flows of private investment (see for example Leigland et al. 2016). However, there is again a lack of rigorous empirical evidence to evaluate the relevance of this potential channel.   19    Given the constraints discussed above, generating evidence to guide policy makers on the overall issue of how to improve investment efficiency appears to deserve a high rank on the list of applied research priorities. References Cameron, A. C. and D. L. Miller. (2015). “A Practitioner's Guide to Cluster-Robust Inference”, Journal of Human Resources, Vol.50, No. 2, pp.317-373. CEDLAS and The World Bank. (2014). “A Guide to SEDLAC Socio-economic database for Latin America and the Caribbean.” http://sedlac.econo.unlp.edu.ar/eng/methodology.php Costa, F., and F. Gerard. (2015). “Hysteresis and the Social Cost of Corrective Policies: Evidence From a Temporary Energy Saving Program”, mimeo. Davis, L., and P. Gertler. (2015). “Contribution of air conditioning adoption to future energy use under global warming”. Proceedings of the National Academy of Sciences USA 112(19): 5962–5967. Dinkelman, T. (2011). “The effects of rural electrification on employment: New evidence from South Africa”, American Economic Review, Vol. 101(7) pp. 3078-3108 Estache, A., and M. Fay. (2009). Current Debates on Infrastructure Policy. Comm. Growth Dev., World Bank Publication. Fay, M. and T. Yepes. (2003). “Investing in Infrastructure: What is Needed from 2000 to 2010?” World Bank Policy Research Working Paper No. 3102. Ferreira, F. H. G., Messina, J., Rigolini, J., López-Calva, L.-F., Lugo, M. A. and R. Vakis. (2013). Economic Mobility and the Rise of the Latin American Middle Class. Washington, DC: World Bank. 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: 83-120. Gertler, P., O. Shelef, C. Wolfram, and A. Fuchs. (2016). “The Demand for Energy-Using Assets among the World's Rising Middle Classes,” American Economic Review. Vol. 106, No. 6, pp. 1366-1401.   20    Gibson J. and S. Rozelle. (2003). “Poverty and Access to Roads in Papua New Guinea”, Economic Development and Cultural Change, Vol. 52 (1), pp. 159-185. Hutton, G., and M. Varughese. (2016). “The Costs of Meeting the 2030 Sustainable Development Goal Targets on Drinking Water, Sanitation, and Hygiene.” Water and Sanitation Program Technical Paper. Washington, D.C., World Bank. IMF. (2015). “Making Public Investment More Efficient.” International Monetary Fund, Washington DC. Lakner, C., and B. Milanovic. (2015). “Global Income Distribution: From the Fall of the Berlin Wall to the Great Recession”, World Bank Economic Review, Advance Access published September 26, 2015. Lee, K., Miguel, T., and C. Wolfram. (2016). “Experimental Evidence on the Demand for and Costs of Rural Electrification.” Unpublished working paper. Leigland, J., Trémolet, S. and J. Ikeda. (2016). “Achieving Universal Access to Water and Sanitation by 2030. The Role of Blended Finance.” World Bank Water Global Practice Discussion Paper. Milanovic, B. (2015). “Global inequality of opportunity: how much of our income is determined by where we live”, Review of Economics and Statistics. Pritchett, L. (2000). “The Tyranny of Concepts: CUDIE (Cumulated, Depreciated, Investment Effort) Is Not Capital.” Journal of Economic Growth, Vol. 5, Issue 4, pp 361–384. Straub, S. (2008). Infrastructure and growth in developing countries: Recent advances and research challenges. Policy Research working paper 4460. World Bank. Straub, S. (2011). Infrastructure and Development: A Critical Appraisal of the Macro-level Literature. Journal of Development Studies. 47, 683–708. Straub, S. (2015). “Policy Lessons from the Recent Literature on Transport Infrastructure Development,” in The Economics of Infrastructure Provisioning: The (Changing) Role of the State, in Picot, Florio, Grove, and Kranz eds., MIT Press. World Bank. (Forthcoming, 2017). Rethinking Infrastructure in Latin America and the Caribbean –Spending better to achieve more. Washington D.C.   21    Figure 1: Infrastructure Access across Deciles         22    Figure 2: Access Premium vs. per capita GDP   23      Figure 3: Countries’ Access Premium across the Income Distribution WATER 0.3 0.2 0.1 0 1 2 3 4 5 6 7 8 9 10 ‐0.1 ‐0.2 ‐0.3 ‐0.4 ‐0.5 ‐0.6 ‐0.7 ARG CHL URY MEX BRA CRI COL SLV PRY DOM ECU PER BOL   24    SANITATION 0.7 0.5 0.3 0.1 1 2 3 4 5 6 7 8 9 10 ‐0.1 ‐0.3 ‐0.5 ‐0.7 ARG CHL URY MEX BRA CRI COL SLV PRY DOM ECU PER BOL ELECTRICITY 0.3 0.25 0.2 0.15 0.1 0.05 0 1 2 3 4 5 6 7 8 9 10 ‐0.05 ‐0.1 ‐0.15 BOL BRA CHL COL CRI DOM ECU MEX PER PRY SLV URY   25    Figure 4: Durable ownership accross the income distribution  Change in households durables ownership 2002-2012, by deciles 25% 20% 15% 10% 5% 0% 1 2 3 4 5 6 7 8 9 10 Refrigerators Washmachines Cars Income elasticity of durable ownership, by deciles 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 1 2 3 4 5 6 7 8 9 10 Refrigerators Washmachines Cars   26    Table 1: Household surveys included in the database Time period 1 2 3 4 5 Argentina 1992 1998 2002 2007 2012 Bolivia 2002 2007 2012 Brazil 1997 2002 2007 2012 Chile 1992 1996 2003 2006 2011 Colombia 1996 2002 2008 2012 Costa Rica 1992 1997 2002 2007 2012 Dominican Republic 1997 2002 2007 2011 Ecuador 2003 2007 2012 Honduras 1992 1997 2003 2007 2011 Mexico 1992 1998 2002 2006 2011 Paraguay 1997 2002 2007 2011 Peru 1997 2002 2007 2012 El Salvador 1991 1998 2002 2007 2012 Uruguay 1992 1997 2002 2007 2012 Table 2: Missing indicators Variable 1992 (7 countries) 1997 (12 countries) 2002 (14 countries) 2007 (14 countries) 2012 (14 countries) Water ‐‐‐‐ DOM COL SLV  (all  obs=1) ‐‐‐‐ Toilet ‐‐‐‐ DOM COL ‐‐‐‐ ‐‐‐‐ Sewers ARG/CRI DOM COL ‐‐‐‐ ‐‐‐‐ Electricity ARG DOM COL/URY ARG ARG Refrigerator ARG/CHL/HND ARG/CHL/DOM/HND  ARG/COL/HND ARG ARG Washing machine ARG/CHL/HND ARG/CHL/DOM/HND  ARG/COL/ECU/HND ARG/HND ARG/CRI/HND Landline phone ARG/CHL/HND/URY ARG/CHL/DOM/HND/URY  ARG/BOL/COL/HND/URY ARG ARG/URY ARG/CHL/CRI/HND/MEX/SLV/ ARG/BRA/CHL/COL/CRI/DOM ARG/BOL/COL/CRI/DOM/ECU Cell phone ARG/COL/ECU ARG URY /HND/MEX/PRY/URY /HND/URY ARG/BRA/CHL/COL/DOM/HN Computer ARG/CHL/CRI/HND/URY ARG/BOL/COL/ECU/HND ARG ARG D/PRY/URY  ARG/CHL/CRI/HND/MEX/SLV/ ARG/BOL/COL/CRI/DOM/ECU Internet 12/12 ARG/BOL/DOM/ECU ARG/DOM URY /HND ARG/BOL/BRA/CHL/COL/DO Car ARG/CHL/HND ARG/BRA/CHL/COL/HND ARG/BRA/CHL ARG/BRA M/HND  ARG/BOL/BRA/CHL/COL/CRI/ ARG/BRA/CHL/COL/CRI/ECU/ Motorcycle ARG/CHL/CRI/HND/SLV/URY ARG/BRA/CHL/CRI/SLV ARG/BRA/CHL DOM/HND/SLV/URY HND/SLV/URY   27    Table 3: Cross-country variability in access to infrastructure – 2012 Panel A: Share of variability explained by country dummies only (R2) Washing Landline Cell pc GDP Water Toilet Sewers Electricity Refrigerator TV Computer Internet Car Motorcycle machine phone phone 0.669 0.395 0.657 0.334 0.542 0.662 0.643 0.157 0.399 0.246 0.184 0.099 0.852 Panel B: Country premium relative to Honduras - 2012 Argentina - 0.103 0.277 0.238 Bolivia 5 793 -0.0879 0.0405 0.122 0.0121 -0.176 -0.536 0.00895 -0.105 -0.0152 0.0540 -0.109 -0.0679 0.0536 Brazil 14 970 0.0467 0.108 0.104 0.0890 0.234 -0.191 0.134 -0.0198 -0.00150 0.154 0.142 Chile 20 266 0.0370 0.267 0.320 0.0853 0.173 0.0695 -0.390 -0.0719 -0.0597 0.149 0.0727 0.0144 Colombia 11 840 0.0386 0.304 0.370 0.0817 0.101 -0.141 0.105 0.00660 0.0396 0.113 0.0728 -0.106 0.161 Costa Rica 13 589 0.0894 0.321 -0.318 0.0848 0.192 0.113 0.103 0.00574 0.0854 0.148 0.0405 0.0494 Rep.Dom 10 322 -0.191 0.0151 -0.251 0.0670 0.00877 0.0142 0.0150 -0.139 -0.144 -0.0858 -0.0559 0.260 Ecuador 7 718 0.00373 0.293 0.151 0.0637 0.0582 -0.348 0.0357 0.0386 -0.0792 0.0386 -0.0169 -0.0257 0.0117 Mexico 16 136 0.0189 0.0291 0.0833 0.0834 0.0967 -0.0577 0.0773 0.0131 -0.165 0.0123 -0.0183 0.0227 -0.0102 Paraguay 10 851 -0.0410 0.135 -0.340 0.0813 0.132 0.0798 -0.141 -0.00581 -0.00253 -0.0009 0.0456 0.410 Peru 7 505 -0.119 0.0877 0.140 -0.0157 -0.280 -0.509 -0.0372 -0.138 -0.110 -0.00785 -0.0716 -0.156 0.0536 El Salvador 11 376 -0.409 0.211 0.0652 0.0390 -0.0509 -0.522 0.0194 -0.0730 0.00916 -0.0502 -0.0353 -0.0488 -0.0391 Uruguay 18 439 0.0875 0.236 0.0327 0.0870 0.223 -0.0450 0.121 -0.128 0.330 0.119 0.00391 0.289 Note: Honduras is taken as reference country for having the lowest pc GDP in 2012 ($4433). Argentinean pc GDP is unavailable for any year close to 2012.   28    Table 4: Cross-country variability in access to infrastructure explained by country determinants - 2012 Washing TV Landline Cell Water Toilet Sewers Electricity Refrigerator Computer Internet Car Motorcycle machine phone phone ln GDP pc 0.4300** 0.735*** -0.7500 0.243*** 0.745*** 1.4280 -0.0898 0.2650 -0.1790 -0.0937 0.3230*** 0.2280** 0.4470 Gini 0.0075 -0.0053 0.0090 0.0018 0.0148 0.0176 -0.0051 0.0051 0.0077* 0.0001 0.0097** 0.0018 0.0067 Density -0.0011*** -0.0002 0.0000 0.0000 0.0002 0.0000 -0.0001 0.0000 0.0002 -0.0004 0.0003* -0.0001 -0.0003 Urbanization -0.0031 -0.0078* 0.0185 -0.0018* -0.0031 -0.0104 -0.0016 -0.0036 0.0023 0.0074*** 0.0011 -0.0034 -0,0015 Observations 1,188 1,188 1,188 1,188 1,188 990 1,188 1,089 1,188 1,188 1,089 1,089 990 R2 0.52 0.18 0.17 0.20 0.29 0.41 0.04 0.03 0.15 0.19 0.14 0.02 0.20   29    Table 5: Household infrastructure income elasticity Washing Landline Cell Water Toilet Sewers Electricity Refrigerator Computer Internet Car Motorcycle machine phone phone Panel A LAC 0,085*** 0,157*** 0,168*** 0,060*** 0,151*** 0,153*** 0,197*** 0,099*** 0,134*** 0,122*** 0,156*** 0,012** Panel B Argentina 0,017*** 0,122*** 0,175*** 0,005*** - - - - - - - - Bolivia 0,081*** 0,128*** 0,136*** 0,103*** 0,178*** 0,075*** 0,141*** 0,102*** 0,148*** 0,092*** 0,094*** 0,021*** Brazil 0,094*** 0,161*** 0,174*** 0,03*** 0,099*** 0,212*** 0,224*** 0,129*** 0,181*** 0,164*** - - Chile 0,041*** 0,097*** 0,102*** 0,018** 0,082*** 0,132*** 0,165*** 0,091 0,154*** 0,135*** - Colombia 0,069*** 0,088*** 0,153*** 0,034*** 0,136*** 0,165*** 0,191*** 0,196*** 0,161*** 0,141*** 0,023*** Costa Rica 0,022*** 0,065*** 0,073*** 0,02*** 0,086*** 0,099 0,174*** 0,123 0,155*** 0,150*** 0,157*** 0,009*** Dom. Rep. 0,120*** 0,198*** 0,117*** 0,037* 0,133*** 0,124*** 0,207*** 0,105*** 0,134*** - 0,187*** -0,010 Ecuador 0,054*** 0,08*** 0,188*** 0,024*** 0,156*** 0,19*** 0,212*** 0,188*** 0,176*** 0,159*** 0,006 Honduras 0,073*** 0,195*** 0,182*** 0,137*** 0,177*** - 0,17*** 0,103** 0,146*** 0,098** 0,167*** 0,015*** Mexico 0,091*** 0,247*** 0,239*** 0,029*** 0,186*** 0,193*** 0,226*** 0,154*** 0,125*** 0,112*** 0,198*** 0,005*** Peru 0,177*** 0,188*** 0,253*** 0,164*** 0,241*** 0,182*** 0,233*** 0,114*** 0,096** 0,098** 0,099*** 0,038 Paraguay 0,169*** 0,22*** 0,092*** 0,059*** 0,164*** 0,182*** 0,17*** 0,113*** 0,139*** 0,088* 0,187*** 0,013 El Salvador 0,193*** 0,219*** 0,218*** 0,104*** 0,215*** 0,116*** 0,195*** 0,061*** 0,087** 0,088** 0,164*** Uruguay 0,026*** 0,151*** 0,211*** 0,025*** 0,095*** 0,164*** 0,259*** 0,01*** 0,086* 0,140*** 0,148*** -0,029**   30    Table 6: Comparative household infrastructure income elasticity HH surveys Fay & Perrotti & data LAC LAC Q 0,1 LAC Q 0,25 LAC Q 0,75 LAC Q 0,9 Yepes 2003 Sanchez 2011 Water 0,08 0,06 0,06 0,05 0,04 0,11 0,13 Toilet 0,16 0,14 0,15 0,14 0,12 Sewers 0,17 0,16 0,17 0,17 0,15 0,19 0,22 Electricity* 0,06 0,04 0.03 0,02 0,01 0,18 0,43 Refrigerator 0,15 0,13 0,14 0,13 0,12 Washing machine 0,15 0,14 0,16 0,15 0,14 Landline phone 0,20 0,19 0,20 0,20 0,18 0,39 0,12 Cell phone 0,10 0,09 0,08 0,09 0,09 0,64 1,00 Computer 0,13 0,11 0,12 0,14 0,14 Internet 0,12 0,09 0,10 0,13 0,12 0,37 Car 0,16 0,13 0,14 0,16 0,16 Motorcycle 0,01 0,00 0,01 0,01 0,02   31    Table 7: Years to Close Gap Water Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Argentina 14 7 3 2 3 3 2 3 3 3 Bolivia 96 58 55 43 48 36 45 45 56 90 Brazil 36 21 15 14 11 8 13 5 6 2 Chile 34 28 24 22 19 16 18 16 18 16 Colombia 56 41 29 24 20 20 18 17 24 30 Costa Rica 38 21 15 20 9 9 6 9 2 6 Dominican Rep , 64 57 59 53 50 39 38 38 31 26 Ecuador 71 52 46 41 40 31 27 26 21 24 Honduras 57 44 40 40 32 32 26 35 25 42 Mexico 36 28 36 19 17 12 16 12 11 10 Peru 43 38 30 24 21 19 17 13 14 11 Paraguay 36 26 19 23 16 16 19 9 8 23 El Salvador 85 81 73 65 64 52 48 41 42 37 Uruguay 31 21 17 18 17 12 14 8 5 3 Sewerage Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Argentina 58 64 62 61 60 53 52 51 48 37 Bolivia 97 81 75 76 70 77 73 75 80 122 Brazil 65 61 55 56 50 48 56 46 49 48 Chile 30 33 35 34 34 30 32 31 29 29 Colombia 71 62 49 43 32 26 25 25 17 18 Costa Rica 203 221 237 245 247 261 267 292 330 443 Dominican Rep , 122 134 142 145 153 150 156 162 186 260 Ecuador 56 54 47 43 41 33 30 26 20 29 Honduras 79 76 75 70 68 68 61 68 70 107 Mexico 52 49 47 41 38 29 27 30 22 22 Peru 54 50 37 30 25 19 17 13 13 10 Paraguay 180 202 216 224 237 236 254 269 301 406 El Salvador 53 40 40 35 36 24 29 27 27 25 Uruguay 57 61 63 60 59 56 54 52 44 42   32    Electricity Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Bolivia 51 26 18 12 10 10 9 12 14 12 Brazil 13 5 3 3 3 2 4 1 1 0 Chile 14 7 8 7 6 4 4 6 4 3 Colombia 52 26 25 21 15 15 14 9 5 4 Costa Rica 33 10 6 6 6 3 6 4 0 2 Dominican Rep , 21 13 13 14 16 13 4 2 4 2 Ecuador 53 39 25 22 19 14 17 12 6 15 Honduras 39 21 17 17 17 7 11 12 14 20 Mexico 16 9 8 7 4 0 3 4 5 0 Peru 33 24 15 8 6 4 5 3 3 2 Paraguay 15 11 7 8 3 5 1 0 1 0 El Salvador 28 17 13 9 9 4 5 3 1 1 Uruguay 9 11 8 8 10 5 6 4 3 1   33