26061 March 2003 Infrastructure for Poor People Public Policyfor Private Provision Edited by Penelope J. Brook Timothy C. Irwin E WORLD BANK J, X FRAIUVC5.ATet ADVISORY FACIIITY Infrastructure for Poor People Public Policyfor Private Provision Edited by Penelope J. Brook Timothy C. Irwin " THE WORLD BANK (D 2003 The Intemational Bantk for Reconotructiori andi Development / The World Bank 1818 H Street, NW \W.ashington. l)C 20433 Telephone 202-173-1000 Internet w%uw.worldbank org E-mail feedhack@a)worldbank.org All rights reserve(d 1 2 3 4 06 05 01 03 The findings, interpretations, dnd conclusions expressed herein are those of the author(s) and do not necessanly refle t the views of the Board of Executive Directors of the World Bank or the governments they represent. The World Bank doe" not guarantee the accuracy of the data included in this work. The boundanes, colors, denominations, and other information shown on any map in this work do not imply any judgrient on the part of the World Bank concerning the legal status of any territory or the endorsemnent or acceptance of such boundaries. R/rigth and Eandesiuns The material in this %tork is copyn,ghted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The World Bank encourages dissemination of its work and will normally grant permission promptly. For pemiission to photocopy or reprint any part of this work, please send a request with complete information to the Copynght Clearance Center, Inc., 222 Rosewood Drive, Dan- vers, MIA 01923, USA, telephone 978-750-8100, fax 978-750-4470, www.copyright com All other quenes on nghto. and licenses, including subsidiary rights,, should be addressed to the Office of the Publi,her, World Bank, 1818 H Street NW, Washington, DC 204.33, USA, fax 202-522-2422, e-mail pubrights@v.orldbank.org. LiE)rary of Congrenn Caialoging-in-1Dnb1ieaUon lData has been applied for. Contents Foreword v Acknowledgments vii Contributors Viti ]. Private Infrastructtire and the Poor: Lncreasing Access 1 Timoth1y Irwin and Penelope Brook 2. Universal Service: Empirical Evidence on the Provision of Tnfrastrtettire Services to Rural and Poor Urban Constumers 21 George R. G. Clarke and Scott J. Wallsten 3. Infrastructttre Coverage and the Poor: A Global Perspective 77 Kristin Komives, Dale Whtittington, and Xurn Wi 4. Measuring the Impact of Energy Interventions on the Poor-An Illustrationi from Guatemala 125 Vivienl Foster and Jean-Philippe Tre 5. Impact of Market Structure on Service Options for the Poor 179 Davtd Erlrilardi 6. Regulatintg Infrastructure for the Poor: Perspectives on Regulatory System Design 209 Warrick SmitIL 7. Regulation of thc Quality of Infrastructure Services in Developing Counitries 233 Bill Baker and Sophie Trenolet 8. Lifeline or Means-Testing? Electric Utibty Subsidlies in Honditras 277 Querntin Wodon, Molhamed Iitsan Ajwad, and Corinnze Staens Foreword A major challenge facing infrastructure policymakers in developing coun- tries today is wvorking out how best to help poor people improve their access to good, reliable infrastructure services. Currently poor people often lack access to safe water and sanitation, modern sources of energy, and elec- tronic means of communication. As a result, they incur high costs obtain- ing low-quality substitutes. For example, they travel long distances to col- lect water thlat is not safe to drink; they burn wood and kerosene for cooking and lighting instead of using electricity, even thoughi electricity, when available, is generally cheaper and less damaging to their health; and they communicate without the benefit of telephones or the Internet. Privatization can help address the critical problem of access if it leads to lower costs, better pricing, or smarter investment decisions, and thereby reduces the risk that profitable opportunities to serve poor people are left unrealized. However, privatization is no panacea. Ensuring that it does lead to improvements requires policymakers to pay attention to the details of market design and regulation. To begin witlh, these arrangements must transfer real risks from the government to the private sector; othervise, pri- vate owners will have little incentive to improve the performance of their businesses. In addition, competition should be permitted wherever possible. Small- scale providers, which are often better at serving poor people than are large utilities, should be allowed to flourish alongside the utilities and to com- pete witlh them when opportunities arise. Regulation should be designed to protect poor people from excessively high pricing, but it should not pre- vent investors from recouping their investment costs, nor should it restrict competition andl services by imposing unwarranted barriers to the entry of new firms, whether in the form of inappropriate quality standards or of out- right monopoly franchises. Last, if access is to improve-under either private or public ownership- governments must have credible policies about who will pay for infrastruc- ture services. The choices are limited to customers and taxpayers, and much of the problem of access has its origin in governments' unwillingness v FOREWORD to impose the costs on either of these parties. Governments have kept tar- iffs below cost to alleviate concerns about affordability, but their macro- economic policy goals have stopped them from raising taxes to fund the shortfall. Although privatization does not create this problem, it brings it to the fore, because private firms insist on knowing who will pay for their services. The most promising, though politically demanding, solution is usually to require customers to pay. Some specific policy goals, such as increasing access by poor people, may, however, justify government subsi- dies. In the past, subsidies usually failed to reach thie poor, in part because they were not linked to results. When subsidies are called for, donors and developing country governments would do well to consider reducing this risk by making the subsidies output based rather than input based, for example, by tying subsidies to increases in access among poor people. This book seeks to cast lighlt on the problem of access to infrastructure services by poor people and to help policymakers design private infra- structure reforms so that they maximize efficient and sustainable improve- ments in access. Michael Klein Director, Private Sector Advisory Services The World Bank vi Acknowledgments Earlier versions of chapters 3 through 7 were initially prepared for the conference on Infrastructure for Development: Private Solutions and the Poor, sponsored by the Public-Private Infrastructure Advisory Facility, the United Kingdom's Department for Infrastructure Development, and the World Bank and held in London during May 31-June 2, 2000. The Public-Private Infrastructure Advisory Facility and the World Bank also provided funding for the preparation of this volume. The book owes its existence to the support and guidance of Michael Klein. Alice Faintich of The Word Doctor copyedited the text. Monika Kosior provided invaluable assistance with the production of the book. vii Contributors Mohamed Ihsan Ajwad works for the World Bank's Interniational Trade Department (majwad@worldbank.org). Bill Baker works for National Economic Research Associates in London (bill.baker@nera.com). Penelope Brook works for the Wforld Bank's South Asia Vice Presidency (pbrook@worldbank.org). George Clarke works for the World Bank's Development Research Group (gclarke@worldbank.org). David Ehrhardt works for Castalia, a consulting company in Melbourne, Australia (david.ehrhardt@castalia.fr). Vivien Foster works for the World Bank's Latin America and Caribbean Vice Presidency (vfoster@worldbank.org). Timothy Invin works for the World Bank's Private Sector Advisory Services Department (tinvin@worldbank.org). Kristin Komives teaches at the Institute of Social Studies in The Hague, The Netherlands (komives@iss.nl). Corinne Siaenis works for tlie Worldl Bank's Latin America and Caribbean Vice Presidency (csiaens@worldbank.org). Wamck Smith works for the Wforld Bank's Private Sector Advisory Services Department (wsmith3@woildbank.org). Jean-Philippe Tre works for the World Bank Institute (tre@worldbank.org). Sophie Tr6molet works for Environmental Resources Maniagement in London (sophie.tremolet@enn.com). Scott Wallsten works for the World Bank's Development Research Group (swallsten@worldbank.org). Dale Whittington teaches at the University of North Carolina at Chapel Hill (dale_whittington@unc edu). Quentin Wodon works for the World Bank's Latin America and Caribbean Vice Presidency (qwodon@worldbank.org). Xun Wu teaches at tlie National University of Singapore (nippwuxun@nus.edu.sg). viii 1 Private Infrastructure and the Poor: Increasing Access Timothy Irwin and Penelope Brook TIMOTHY IRWIN AND PENELOPE BROOK 1 .1 [troduction In the last two decades many governments have allowed private compa- nies to provide infrastructure services previously provided only by state- owned businesses. In some cases they have privatized state-owned busi- ness. In others they have allowed private firms to invest in and operate those businesses under long-term concession or lease contracts. In still others they have allowed private firms to compete alongside former gov- ernment monopolists. Although private participation in such infrastruc- ture industries as electricity, water, and telecommunications has often led to improvements-such as lower costs and better services-it has also been controversial. Among other things, privatization entails a more com- mercial approach to the provision of the services, which can lead to the withdrawal of subsidies, an increase in prices, the more frequent discon- nection of nonpaying customers, and a reluctance to connect new cus- tomers unless they will be profitable. As a result observers have raised concerns about its effects on the poor, irrespective of its possible benefits in other respects. The chapters in this book examine the data on infrastructure and the poor in developing countries and consider how policies centered on private pro- vision can address their needs. Many of the chapters focus on the extent to which the poor have access to infrastructure services of reasonable quality, for example, to water that is safe to drink, to a reliable source of electricity, and to a nearby telephone. Access to such services is, of course, not the only infrastructure issue that matters to the poor; the poor who already have access to modem services care, for instance, about the price and reliabil- ity of those services. However, in most developing countries access is the key issue. In these countries most of the poor have no access to standard infrastructure services provided by utilities. Instead they often pay high prices for lower-quality substitutes: they might buy water by the bucket from a private vendor and use candles instead of electricity for lighting. They would rarely make a telephone call. The lack of ready access to good basic infrastructure services can directly reduce the well-being of the poor. For instance, unsafe water and sanitation not only cause disease, but also This chapter benefited from comments provided by Cecilia Bricefio-Garmendia, George Clarke, David Ehrhardt, Philip Gray, Michael Klein, Kristin Komives, Sophie Sirtaine, Scott Wallsten, and Chiaki Yamamoto. 2 PRIVATE INFRASTRUCTURE AND THE POOR INCREASING ACCESS hamper their chances of rising out of poverty, while the hours women ancl girls spend fetching water or gathering biomass fuels in many developing countries severely restrict the time they have for education and other pro- ductive activities. Therefore, in this introduction we review the problem of access, beginning with some of the facts about access, then considering why it is low and what governments can do to increase it. 1.2 Some Facts about Access According to one commonly cited source, a quarter of the population of the world's poorest countries lacks access to safe water and less than half has safe sanitation facilities (table 1.1). Only a tiny proportion has a fixed or mobile telephone (figure 1.1). WiLtlin developing countries, access by the poor to infrastructure serv- ices is lower than average, as two chapters in this volume clarify. George Clarke and Scott Wallsten look at access to electricity telephones, piped water, and flush toilets in selected developing and transition countries. Not surprisingly, they find that poorer households-identified as those whose head has no formal education-generally have much lower rates of access The poor have especially low rates of access in Africa (figure 1.2). Kristin Komives, Dale Whittington, and Xun Wu obtain similar results using a different dataset. They find that the urban poor in all parts of the world often have access to electricity, and that the poor in Eastern Europe and Cen- tral Asia have relatively high levels of access to all services. Apart from these bright lights, however, tle picture is dim. FIigure 1.3 illustrates one of their TABLE 1.1. Access to an Improved Water Source and Improved Sanitation Facilities by Country Income Group, 2000 (percentage of the population) Country income Water Sanitation Low income 76 45 Lower-middle income 80 52 Upper-middle income 87 81 High income 100 100 Source World Bank (2002), supplemented by authors' estimates for high-income countries 3 TIMOTHY IRWIN AND PENELOPE BROOK FIGURE 1.1. Telephone Mainlines and Mobile Telephones by Country Income Group, 2000 Access per 1,000 people 1,200 - 1,000 - D Telephone mainlines D Mobile phones 800 - 600 - 400- 200- 0 - 1 Incm Low Lower-middle Upper-middle High inmcome income income Note Data for telephone mainlines are weighted averages of the countries in the income groups, while data for mobile telephones are medians Source World Bank (2002), using data from the International Telecommunication Union results, showing how rates of access to various services tend to vary with income. As a point of reference, note that roughly half the individuals (not households) in the world live on about $60 a month or less. 1.3 Why Is Access Low? Part of the explanation for the poor's low rate of access to infrastructure services is simply that the services are relatively expensive. Access can thus be expected to increase as incomes rise, and reforms that increase incomes can be expected to increase access. Indeed, perhaps the main contribution infrastructure reforms can make to reducing poverty is to raise incomes by increasing economic efficiency. Yet there are reasons for think- ing that differences in income are not the only factor explaining differences in access and that changes in infrastructure policy can increase access to formal infrastructure services by the poor irrespective of their effect on 4 C~~~~~~~~~~C -o U~~~~~~~~~~~~~~C a) -n C~~~~~~~~~~~~~~ E -n - -= C~~~~~~~~~~~~~~~~~~~~~~~~C 4- 0 VIa) 14 CU' >- C * a) cu~~~~~~~~~~~~' Cor ~0 r cD Or' a, as ag 2 ° -CV~~~~~~~~~~~~~~~~~~~~~~~- CC~~~~~~~~~~~~~~~~~~C- 5.0 :3~~~~~~~~~~~~~~~~~~~~ TIMOTHY IRWIN AND PENELOPE BROOK FIGURE 1.3. Trend Rates of Access to Infrastructure Services by Household Income, Nonrandom Country Sample Percentage of households with access 100 - 90 - 80 - Electrcity 70 - --- In-house tap .. ------ Sewer 60 - - - - - - Telephone 50 - 40 - 30 - 20 - - ........ 10 -. - - - 0 - I I I I I I I I I I I I I lI I I l I I 150 300 450 600 750 900 1,050 1,200 Monthly household income in 1998 US$ Source Chapter by Komives, Whittington, and Wu in this volume incomes. First, we know that rates of access do not correlate perfectly with incomes. Figure 1.4, for example, suggests that differences in income explain some, but far from all, of the variation by country in access to safe water and the number of telephones per capita. Second, we know that the poor often pay high prices for services from informal or nonstandard infrastructure providers. They buy water from tanker trucks at prices higher than those charged per liter by water utili- ties and use wood and kerosene for heating and cooking even though the cost per kilowatt hour of these fuels is higher than that of electricity (Vivien Foster and Jean-Philippe Tre in this volume; Lyonnaise des Eaux 1998). They also spend a lot to cope with the low quality of the service their util- 6 * 0 - *- 0 ~~ ~~~~~~~~~~~~~~~~~. 0~~~~~~~~~~~~~~~~~ 's - CD * * 0;X: C) 0~~00 to> 0 I 040 0 0 .0 0 * 00 E * B o 0 CD. 'We o 0 - * _o _o s c ~ ~ e 3 .*_ - LA o.> 0 v e 3 0 t. .* Gl~~~~~~~~~~~~~~~~~~~~~c ,a E 0~~*0 0 o *E 0 * _- o ' oX S *0 *S S *o .0I 0 - * o0- a) *~~ -°~ 0 a w -0~~~~~~~~~~~~~~~~~~ V 0. _ O U, ' co o. o o o o o o a) Ia, ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~C) ,: r~~o -L oo*CLt c@ oup o~~~~~~~~~~~~~~~~~~o 0~~~~~~~~~~0 2 z _ _ Os * 0 o 0 a 0 C a,~~~~~a OS *~~~~~~~~~~~~~O CL~~~~~~~~~~a -v 0 _ o - -~r= C a) 0 0 I 0r S I - a,0~~~~~~~~~~~~~~~~~~ a) - TIMOTHY IRWIN AND PENELOPE BROOK ity provides, for example, by purchasing storage tanks to compensate for intermittent water supply (Water and Sanitation Program 1999). When the poor are asked, they often say they would be willing to pay more for serv- ices than nearby utilities charge (see, for example, Foster, G6mez-l,obo, and Halpern 2000; Water and Sanitation Program 1999). Typically the poor buy in small quantities, so the high per unit prices they pay or say they are willing to pay do not imply that they would always be prepared to pay all the costs of service from the utility, including connection charges, but they do suggest that low incomes are not always what impedes access. Third, the recent policies of most governments appear to have had the unintended effect of inhibiting access. From the 1950s to the 1990s most governments adopted a strategy of providing infrastructure services through state-owned businesses and prohibiting competition from other companies. As part of the strategy they often kept prices low for some ("cross-subsi- dized") customers or services, while raising prices for others. The hope was that exclusive government ownership of the businesses wouldl facilitate coherently planned investments in extending service and protect consumers from monopoly pricing, while cross-subsidies would make the services affordable to the poor. However, the combination of state ownership and monopoly tended to weaken the government-owned businesses' incentives and ability to operate and invest efficiently (for surveys of the evidence, see Gray 2001; Shirley and Walsh 2000: World Bank 1995). Political pressures also kept the prices many government-owned businesses charged below costs (World Bank 1994), starving the businesses of the cash thiey needed to fund investment internally and the creditworthiness they needed to support external borrowing. In addition, as table 1.1 and figures 1.1-1.1 suggest, cross-subsidies tended to help the middle-class more than the poor, because the poor often lacked access (Clark and Wallsten in this volume; World Bank 1994; Walker and others 2000). 1.4 Policies to Increase Access The problems of the government monopoly approach suggest that the adop- tion of a new model involving private participation and competition will lead to better infrastructure services generally, and to faster growth of access in particular. While relatively little evidence on the effects of recent reforms on access exists, some of the early research is positive. Much of it 8 PRIVATE INFRASTRUCTURE AND THE POOR INCREASING ACCESS TABLE 1.2. Effects of Private Participation in Water and Sewerage on Consumer Welfare and Coverage Category Mexico City Abidjan Conakry Buenos Aires Nature of private participation Service contracts Lease Lease Concession Consumer gains per capita (1996 USS) n a n a 8 147 Coverage. water (%) Before 95 72 38 70 After 97 82 47 81 Coverage sewerage (%) Before 86 35 (9) 58 After 91 (35) 9 62 Growth in new connections (annual average, %) Before na 40 -01 21 After 51 6 7 8 5 2 8 n a Not available Note The bracketed numbers for sewerage In Abidjan and Conakry indicate that coverage is not believed to have changed much The four cities are ordered according to a rough view of the extent of responsibility and risk transferred to the private company, with service contracts representing the smallest transfer Note that the reform in Abidjan involved the deepening of private participation rather than irs introduction Before the reform, according to Shirley and Menard (2002 5), the con- tract 'had some characteristics similar to a management contract" Source Shirley and Menard (2002) comes from telecommunications, where private participation and competi- tion have been associated with strong growth in access (see, for example, Gebreab 2002; Ramamurti 1996; Ros 1999). In other sectors, in whlich reforms have been less wivcespread and reliable data are harder to find, less systematic evidence exists. In water, however, one of the most careful col- lections of case studies finds that private participation is associated with increases in access (table 1.2). In electricity too, some evicdence indicates that private provision helps increase access. Estache, G6mez-Lobo, andl Leipziger (2000) reported evi- dence that privatization contributed to an increase in electricity connec- tions in Chile in the 1990s. In this volume Clarke and Wallsten find that electricity coverage among those with no formal education fell slightly in Colombia after privatization, but rose in Bolivia, Brazil, and Peru. Plane (I 999) notecl that access increased after the privatization of Cbte d'ivoire's electricity company in 1990. 9 TIMOTHY IRWIN AND PENELOPE BROOK While the limited available evidence suggests that private provision has probably increased access, the extent of any positive effects will depend crucially on several government policies. We highlight here five types of policy discussed in this volume that matter for access, namely: o Allowing entry and permitting competition o Clarifying property rights and facilitating contract enforcement o Allowing prices to cover costs, now and in the future o Setting appropriate quality standards o Targeting subsidies to the poor. Because the poor are often willing to pay the costs of service, much of the problem of increasing access involves facilitating a deal between willing buyers and willing sellers. Accordingly, the first four types of policy we dis- cuss have this aim. The fifth goes beyond the removal of obstacles in the way of a deal and aims to increase the poor's purchasing power. Allowing Entry and Permitting Competition Perhaps the simplest step the government can take to facilitate improvements in access is to ensure that infrastructure providers who could serve the poor are legally permitted to do so. In some industries, such as long distance and mobile telecommunications, allowing entry can lead to head-to-head competi- tion between relatively large formal sector providers. The benefits of such com- petition in terms of lower prices and better quality have been documented in many studies, which Gray (2001) and Winston (1993) survey. The main bene- fit of liberal entry policies in developing countries is often to accelerate the growth of access to services, as vividly illustrated by the case of telecommu- nications services in Uganda, a country that pursued the strategy of govern- ment monopoly until 1995 and then began to allow competition (figure 1.5). Liberal entry policies can also be helpful in industries such as piped water supply and electricity distribution, where firms tend to have local monopolies. Here the main benefit of allowing entry is to speed the exten- sion of services to areas currently unserved by the utility as firms compete to be the dominant supplier in each location.' In many developing countries legal prohibitions on entry into utility indus- tries are not strictly enforced, and small firms often operate illegally along- side the dominant utility. In these cases some of the benefits of liberal entry policies and competition are realized, even though entry is illegal. Nonethe- less, legalizing the operation of these firms can still be helpful, because the 10 PRIVATE INFRASTRUCTURE AND THE POOR INCREASING ACCESS FIGURE 1.5. Fixed and Mobile Subscribers, Uganda, 1960-2001 Subscribers per 1,000 inhabitants 18 - 16 - 14 - Fixed per 1,000 ........ Mobile per 1,000 12 - 10 - UTL January 2001 8- 6- 4- ......... MTN October 1998 2- CeDe May 199 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 Note The arrows show when mobile operators entered the market CelTel and MTN are private com- panies that started providing mobile services soon after they were permitted to do so UTL is the for- mer government-owned monopoly, which was partially privatized in February 2000 and launched its mobile service in January 2001 Source International Telecommunication Union (2002) threat of closure and the imposition of other penalties by the authorities may cause the illegal firms to raise prices, lower quality, and invest less. Because these providers typically serve the poor, several authors in this volume stress the importance of permitting them to operate alongside large utilities (see the chapters by David Ehrhardt, Warrick Smith, and Bill Baker and Sophie Tre- molet, as well as Brook and W Smith 2001; Solo 1998). Clarifying Property Rights and Facilitating Contract Enforcement Although the policy-focused chapters in this volume concentrate on poli- cies relating specifically to the infrastructure sector, some economywide 11 TIMOTHY IRWIN AND PENELOPE BROOK policies may be just as important to the expansion of access. In addition to policies that help by raising incomes, those that facilitate the making and enforcing of contracts between buyers and sellers should increase access. As Ehrhardt points out in this volume, clarifying and protecting the poor's property rights by ensuring that they have legal title to their property can help them contract with the utility and borrow to cover the cost of connec- tion (see also Water and Sanitation Program 2002). Likewise, utilities in developing countries often suffer high rates of theft through unauthorized connections and nonpayment of bills, in part because the countries' legal systems do not adequately protect the utilities' property rights and allow them to enforce their contracts. The direct beneficiaries of weak property rights and contract enforcement are sometimes poor-Foster and Tre in this volume describe a case in which the implicit subsidy represented by theft was found to be better targeted than a parallel legal subsidy-but theft reduces utilities' incentives to invest in expanding the network, espe- cially in poor areas, thereby reducing the chances of connections for poor customers who are willing to pay. 2 Allowing Prices to Cover Costs, Now and in the Future Although infrastructure companies are often monopolies and usually have market power, the most important challenge in infrastructure policy facing developing countries is, paradoxically, not preventing companies from charging more than costs, but rather convincing them they will be allowed to recover their costs in environments where politicians have long held prices below cost. Several factors combine to encourage governments to keep prices for elec- tricity, water, and other infrastructure services below cost. First, many people purchase the services, and those who do tend to care deeply about them. Because the services are provided by companies with strong market power or a monopoly, people also tend to doubt they are getting a good deal. In the case of water, some argue that the service is so important it should be free. The combination of these factors causes customers to pressure the government to reduce prices. Moreover, much of the investment required to provide the serv- ices is sunk; that is, once made the investment cannot be reversed or moved to a new market. Thus a government can intervene to push prices below the levels that allow investors to recoup the costs of their investment without 12 PRIVATE INFRASTRUCTURE AND THE POOR INCREASING ACCESS causing them to cease providlng services, so long as it allows prices to cover operating costs. The temptation to use regulation to push prices below cost- covering levels is thus great. Yet unless the government can persuade investors that it will resist future pressures to push prices below costs, those investors will not invest to provide new services to the poor. Smith's chapter in this volume discusses some ways that governments, especially those with weak administrative capacity, can address these problems, with a particular view to helping the poor. He discusses such options as the following: * Intervening sparingly and carefully to reduce the risk of regulatory overreaching * Making use of competition to reduce the politicization of the markets and limit the need for price regulation * Increasing the independence of regulatory agencies from day-to-day polit- ical influence to help them resist pressure to push prices below costs Setting Appropriate Quality Standards Poor infrastructure services can threaten health and safety, and the regula- tion of their quality is an important policy concern. Regulators in rich countries tend to set quality standards at high levels, reflecting a tradeoff between quality and affordability that is, more or less, appropriate in these countries. Historically, they have also relied extensively on the regulation of inputs, controlling the means by which quality standards are achieved, rather than defining acceptable outputs and allowing experimentation and innovation in ways of achieving these outputs. As Baker andl Tremolet argue in this volume, developing countries are tempted simply to adopt rich countries' approaches to quality regulation without modification, which may often be a mistake. Quality matters in developing as well as industrial countries and to the poor as well as to the rich, but improvements in quality have a cost, so regulating for rich-country standards in poor countries can make services prohibitively expensive- unless the standards are ignored. Baker and Tremolet argue that governments should regulate quality only when regulation effectively addresses a market failure and the standard does not discourage the provision of service to the poor. In addition, focusing regulation on outputs allows providers to innovate and to offer less expensive ways of delivering service of a given quality, for example, a reliable supply of potable water at a reasonable pressure. 13 TIMOTHY IRWIN AND PENELOPE BROOK Targeting Subsidies to the Poor Even if the government has removed all the impediments to increasing access discussed here, the poor may still find that formal infrastructure services are too expensive to buy or, if they can afford them, that they are extremely expen- sive relative to their incomes. Thus the question arises whether, and if so how, the government should subsidize their use of the services. Before advocating a subsidy for infrastricture services, policy analysts need to compare its likely effectiveness with subsidies for other goods and services and with a subsidy provided in cash. Even the best electricity sub- sidy, for instance, may help the poor less effectively than a good subsidy of the same amount spent on education or health. In this volume Quentin Wodon, Mohamed Ihsan Ajwad, and Corinne Siaens examine a budget- funded scheme for subsidizing electricity consumption in Honduras and, using techniques applicable to other countries and sectors, reveal target- ing problems that are common in infrastructure subsidies. They estimate that over 80 percent of the electricity subsidy goes to clients who are not poor, partly because the government subsidizes even quite high levels of consumption, and partly because many low-consumption households are not poor. Moreover, even considering only those households that are clients of the main electricity utility (and thus ignoring ehe predominantly poor nonclients), they find that subsidies based on electricity consumption do not target the poor as accurately as would subsidies based on other crite- ria, such as the size of the potential recipient's house (figure 1.6). The effectiveness of infrastructure subsidies in directly targeting poverty will vary from industry to industry, and in some industries and countries, especially where coverage is already high, infrastructure subsidies may help the poor more effectively than others types of subsidies. If the govern- ment concludes that infrastructure subsidies are part of a cost-effective program for alleviating poverty, the question then becomes how to design them to maximize their impact. As noted earlier, traditional cross-subsidies provided through government- owned monopolies have usually failed to reach the poor. In addition, even if they are well targeted, they tend to break down in the presence of competi- tion between suppliers, because the subsidizing customers can switch to a new supplier who does not overcharge them (see Clarke and Wallsten in this volume). However, subsidized provision-and even provision that is free to the user as advocated by some for water services-does not entail public 14 PRIVATE INFRASTRUCTURE AND THE POOR INCREASING ACCESS FIGURE 1.6. Accuracy of Criteria for Targeting the Poor among Electricity Consumers, Honduras Size of house Quality of house Demographics Education _ Electricity consumption Employment status Geographic location Access to water and sanitation 05 06 07 08 09 10 Accuracy of targeting Note The graph shows the accuracy, defined as the area under the relative- (or receiver-) operating- characteristic curve, of various indicators in targeting poverty among ciients of the dominant utility in Honduras Random allocation of subsidies would have accuracy of 0 5 and perfect targeting of 1 0 Source Wodon, Ajwad, and Siaens in this volume ownership or monopoly, and policy analysts have recently devoted attention to subsidies funded from general taxation or from special industry-specific levies that can work in competitive markets wvsth private providers. The most effective such subsidies will probably be output rather than input based; that is, Lhey will be linked not to the use of certain inputs, such as the construction of a dam, but to the achievement of certain out- puts, such as the provision of electricity or water (see Brook and S. Smith 2001 for a discussion of output-based aid and several examples). In some circumstances subsidies may be efficiently provided through a dominant main utility, but especially when the utility does not serve many of the poor, a better option may be to design the subsidies so that different firms can compete to provide the subsidized services. Often governments have a choice between subsidizing consumption and subsidizing access. When access rates are low, subsidizing access by new customers is likely to be 15 TIMOTHY IRWIN AND PENELOPE BROOIK more effective than subsidizing consumption (Estache, Foster, and Wodon 2002). One scheme that allows competition and promotes access is to auc- tion the obligation to provide services to new areas, with the winner being the firm that demands the lowest subsidy (see, for example, Cannock 2001). Finally, infrastructure subsidies may often be designed not directly to help the poor, but to reduce opposition to reforms. For example, they might be used to allow prices to be raised to cost-covering levels gradually rather than immediately (see, for example, Brook and Locussol 2001). If the reforms themselves help the poor, the subsidies may indirectly help them by reducing primarily middle-class opposition. 1.5 Conclusion Many of the poor in developing countries, especially in rural areas, lack ready access to good, reliable infrastructure services, and they pay the price in terms of high costs and poor quality, as well as in opportunities for escaping poverty. Improving access will likely require considerable inno- vation and entrepreneurial effort to find ways of better tailoring services to the needs of low-income households and reducing costs to improve afford- ability. The private sector, both for-profit and nonprofit, has a central role to play in this process. As discussed in this introduction and elaborated on in the chapters that follow, however, the policies that governments put in place to govern private provision and to subsidize access can make or break the efforts of both public and private service providers. Infrastruc- ture reforms have the potential to improve access to services by the poor, but policymakers need to pay attention to the details of the reforms to ensure that the poor do indeed benefit. Notes 1. Circumstances can be described in which legal moniopolies increase efficiency in industries characterized by natural nionopoly (see Train 1991, for example), but in our view the weight of the evidence favors free entry in practice. 2. The enforcement of contracts and laws will not iniprove access if the contracts or laws are themselves inimical to such improvements. Clauses in concession con- tracts that grant utilities regional monopolies aiicl regulations that impose unrea- sonably high quality standards are examples. 16 PRIVATE INFRASTRUCTURE AND THE POOR INCREASING ACCESS References The word "processed" describes informally reproduced works that may not be commonly available through libraries. Brook, Penelope J., and Alain Locussol. 2001. "Easing Tariff Increases: Financing the 'Transition to Cost-Covering Water Tariffs in Guinea." In Penelope J. Brook and Suzanne M. Smith, eds., Contra ctingfor Public Services: Output-Based Aid and Its Applicationts. Washington, D.C.: World Bank. Brook, Penelope J., and Suzanne M. Smith, eds. 2001. Conttractitngfor Public Services: Output-Based Aid and Its Applications. Washington, D.C.: World Bank. Brook, Penelope J., and Warrick Smith. 2001. "Improving Access to Infrastructure Services by the Poor: Institutional and Policy Responses." Background paper for the Private Sector Development Strategy of the World Bank. Washington, D.C. Cannock, Geoffrey. 2001. "Expanding Rural Telephony: Output-Based Contracts for Pay Phones in Peru." In Penelope J. Brook and Suzanne M. Smith, eds., Contractingfor Public Services: Output-Based Aid and Its Applications. Washington, D.C.: Wforld Bank. Estache, Antonio, Vivien Foster, and Quentin Wodon. 2002. Accounttinigfor Poverty in Infrastructure Reform: Learninigfrom l atit Armenca s Experience. Development Studlies. Washington, D.C: World Bank Institute. Estache, Antonio, Andres G6mez-Lobo, and Danny Leipziger. 2000. "Utilities 'Privatization' and the Poor's Needs in Latin America: Have We Learned Enough to Get It Right?" Paper prepared for the confer- ence on Infrastructure Development and Private Solutions and the Poor, May 31-June 2, London. Foster, Vivien A., Andres G6mez-Lobo, and Jonathan Halpern. 2000. "Designing Direct Subsidies for Water and Sanitation Services. Panama: A Case Study." Policy Research Working Paper no. 2344. Wlorld Bank, Washington, D.C. 17 TIMOTHY IRWIN AND PENELOPE BROOK Gebreab, Frew Amare. 2002. "Getting Connected: Competition and Diffu- sion in African Mobile Telecommunications Markets." Policy Research Working Paper no. 2863. World Bank, Washington, D.C. Gray, Philip. 2001. "Private Participation in Infrastructure: A Review of the Evidence." World Bank, Washington, D.C. Processed. Intemnational Telecommunication Union. 2002. Thte World Telecommunica- tion Development Report: Reinventing Telecoms. Geneva. Lyonnaise des Eaux. 1998. Alternative Solutions for Water Supply and Sanitation in Areas with Limited Financial Resources. Nanterre, France. Plane, Patrick. 1999. "Privatization, Technical Efficiency, and Welfare Consequences: The Case of C6te d'Ivoire Electricity Company (CIE)." World Development 27(2): 343-60. Ramamurti, Ravi. 1996. "The New Frontier of Privatization." In Ravi Ramamurti, ed., Privatizing Monopolies: Lessons from the Telecommuni- cattions and Transport Sectors in Latin America. Baltimore, Maryland: The Johns Hopkins University Press. Ros, Augustin J. 1999. "Does Ownership or Competition Matter? The Effects of Telecommunications Reform on Network Expansion and Effi- ciency." Journal of Regulatory Economics 14(1). Shirley, Mary M., and Claude Menard. 2002. "Cities Awash: A Synthesis of the Country Cases." In Mary M. Shirley, ed., Thirstingfor Efficiency: the Economics and Politics of Urban Water System Reform. Amsterdam: Pergamon Press. Shirley, Mary M., and Patrick Walsh. 2000. "Public Versus Private Own- ership: The Current State of the Debate." World Bank, Washington, D.C. Processed. Solo, Tova Maria. 1998. "Competition in Water and Sanitation: The Role of Small-Scale Entrepreneurs." Public Policy for the Private Sector Note no. 165. World Bank, Washington, D.C. 18 PRIVATE INFRASTRUCTURE AND THE POOR INCREASING ACCESS Train, Kenneth E. 1991. Optimal Regulation: The Economic Tlheory of Natural Monopoly. Cambridge, Massachusetts: MIT Press. Water and Sanitation Program. 1999. "Willing to Pay but Unwilling to Charge: Do "Willingness-to-Pay" Studies Make a Difference?" Field Note. United Nations Development Programme-World Bank, New Delhi. . 2002. New Designsfor Water and Sanitation Transactions: Mak- ing the Private Sector Workfor the Poor. Washington, D.C.: Water and Sanitation Program and Public-Private Infrastructure Advisory Facility. Walker, Ian, Fidel Ordoiiez, Pedro Serrano, and Jonathan Halpern. 2000. "Pricing, Subsidies, and the Poor: Demand for Improved Water Services in Central America." World Bank, Washlington, D.C. Processed. Winston, Clifford. 1993. "Economic Deregulation: Days of Reckoning for Microeconomists." Joutrnzal of Economic Literature 31(3): 1263-89. World Bank. 1994. IWorld Development Report 1994. Newv York: Oxford University Press. . 1995. Bureaucrats in Business: 77Te Economics and Politics of Government Ownership. World Bank Policy Research Report. Oxford, U.K.: Oxford University Press. 2002. World Development Itdicators. Washington, D.C. 19 2 Universal Service: Empirical Evidence on the Provision of Infrastructure Services to Rural and Poor Urban Consumers George R. G. Clarke and Scott J. Wallsten 21 GEORGE R. G CLARKE AND SCOTT J WALLSTEN 2.1 Introduction Although much of the discussion about regulatory reform and privatization of infrastructure has focused on efficiency, distributional issues have strongly influenced public policy toward infrastructure in both industrial and developing economies. Most countries specify universal access to cer- tain infrastructure utilities, including telecommunications, electricity, and piped water and sewerage, as a public policy goal. Specific laws and objec- tives differ by country and by industry, but the general goal is to ensure access for all people at affordable prices. Most universal access laws and regulations have a geographic component meant to promote service in rural areas and a targeted component meant to help the poor afford service. At least in theory, countries have traditionally financed these obligations through cross-subsidies: low-cost and high-income consumers paid prices above cost to subsidize high-cost and low-income consumers, who paid prices below cost. Some observers have wormed that even if privatization and competition in infrastructure utilities increase efficiency and improve average con- sumer coverage, such reforms could hurt the poor in at least two ways. First, new market structures, including competition, make cross-subsidies difficult to maintain and raise the possibility that private firms will "cream skim." that is, serve the most profitable customers and ignore the unprof- itable ones, namely, poor and rural consumers. Second, reforms often necessitate tariff rebalancing, that is, increasing prices to cover costs. Even if such rebalancing is necessary to ensure viable service over time, higher prices could make service increasingly unaffordable for the poor. This chapter reviews the evidence on universal access in developing countries. We first discuss the rationale for universal access laws and review the different ways subsidies can be financed and allocated, along with the implications of these various methods. We then evaluate the his- torical effectiveness of monopoly enterprises in providing service to the poor and how privatization has affected coverage. Overall we find little evidence that subsidies have actually been used to meet universal service goals under monopoly provision: outside Eastern We tlhdnk Frew Gehicalt an(d Rosario Kaue4hiro for research d1 sistance We are grateful to Mlary Shirley for conmments on an earlier draft. 22 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE Europe, infrastructure connections to rural areas and the poor are distress- ingly low. Moreover, many mechanisms ostensibly intended to help the poor end up helping only the wealthy. Subsidizedl service prices, for exam- ple, tend to benefit the wealthy, because they are mor-e likely to be con- nected to the neLwork and consume the service, whlile poor households without direct connections receive nothing. Meanwvhile the empirical evidence on the effect of reforms on the poor is limited. Nonetheless, case studies and data gleaned from household sur- veys reveal some important trends. To begin with, ther-e is no evidence that reforms tend to hurt poor or rural consumers, at least in terms of access to service. Even when service prices increase, the share of poor and rural res- idents with connections does not generally decrease. In many cases cover- age even increases, possibly because actual connection fees decrease once service is no longer rationed. Case studies reveal that allowing entry and competition in infrastructure can dramatically improve service to the poor. Competition results in a range of price andl quality options, making service possible to regioins and income levels that a monopoly provider would never have considered. However, it is also clear that laws and regulations must be carefully con- sidered to ensure access to the poor. In particular, riles that appear to help the poor at first glance may, in leality, only help mainlain monopoly profits. For example, both public and private firms often claim that they need some form of monopoly rights to fund universal service obligations. Often these claims are self-serving, meant simply to block competition and lbolster prof- its, not to help the poor. Competition, however, does break dlown the ability to cross-subsidize service. Countries must therefore turn to other metliods to help those who simply cannot afford service priced at cost whien society believes that everyone must be connected. If funding through general tax revenues is too dlifficult or too costly, countries may considler, for example, universal service funds to which all firms contribute and from which they may all draw when providing service to the poor or to high-cost areas. 2.2 Why Do Countries Promote Universal Service? Universal service policies are typically justified through a combination of three factors (Cremer and others 1998a,b). First, externalities related to the 23 GEORGE R G CLARKE AND SCOTT J WALLSTEN consumption of infrastructure services might make subsidizing prices for low-income consumers economically efficient. Second, infrastructure serv- ices might be merit goods. Finally, political factors or regional development goals may induce a government to transfer resources to rural or low-income constituents. Efficiency, equity, and politics all explain the common prac- tice of subsidizing infrastructure services to low-income and rural con- sumers, although the strength of these arguments varies considerably across services and countries. Externalities The most common justification for subsidizing infrastructure is that posi- tive externalities are associated with consuming some services. Positive externalities imply that the total benefits of supplying the service exceed the benefits to the individual who receives the service. In other words, soci- ety benefits more than the individual recipient does. If the total marginal benefit of service exceeds the private marginal benefit, then individuals will consume less than the optimal amount because the total marginal ben- efit exceeds the private cost. The externalities argument for subsidies is stronger in some infrastructure sectors than in others, witth sewerage and piped water the most likely candidates, telecommunications less so, and electricity and gas only weakly so, if at all. The case for positive externalities necessitating subsidies is probably strongest for sewage removal and treatment. Improperly treated sewage can pollute the environment and spread disease to people other than those who produce the waste, especially if households dispose of their waste in pub- lic facilities, for instance, public waterways, parks, or roads. Furthermore, if society values the individual health benefits associated with improved sewerage more than the individuals do-either because individuals under- estimate the costs that communicable diseases can impose on the rest of soci- ety or because they are unable to fully assess the health risks associated with poor sewerage disposal-there might be additional scope for subsidies. Access to piped water may also have health benefits.' If society as a whole benefits more from an individual's improved health than the individ- ual does, the individual may undervalue piped water.2 Consequently, if piped water is a normal good (that is, poor users tend to consume less thlan wealthier users) and the health benefits associated with piped water are 24 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE greatest for low levels of consumption, for example, water for drinking and cleaning, focusing subsidies on low-income households by subsidizing basic access, for instance, through public standpipes or single household taps either indoors or in courtyards, would probably be reasonable. How- ever, if, as Esrey (1996) suggested, piped water improves health outcomes only when combined with improved access to high-quality sewerage such as flush toilets or water-seal latrines, then subsidies for basic access might have only modest health externalities. In telecommunications services, network externalities mean that the benefits a new conisumer accrues from connecting (the private benefits) are less than the total benefits to society, because everyone on the network benefits when an additional person connects. Because the private benefits from subscribing are less than the total benefits, individuals may not face a strong enough incentive to subscribe, thereby requiring subsidies to induce subscription. However, ignoring for now the crucial question of whether subsidies actually promote access successfully, the argument that network externalities justify subsidies does not necessarily hold up under closer inspection (see Cremer 1 998b for a more complete discussion of this issue). First, even if the benefits to the new subscriber are less than the total benefits, the private benefit may still exceed the cost. Second, because the firm's services become more valuable when more people are connected, the firm can capture some of the benefits from network exter- nalities. Consequently, even though network externalities are external to the individual, they are not necessarily external to the firms providing the service, potentially removing the need for subsidies. In other words, net- work externalities by themselves do( not necessarily imply telephone under- subscription and a need for subsidies. Merit Goods Even if infrastructure services generatedl no externalities, some services might be merit goods, that is, goods and services that society believes everyone shoulcd have. A policy dlecision that certain goods and services are more important than others for people to consume may come from a belief that society functions better when everyone has access to a minimum set of services or a concern that indlividluals are unable to accurately assess the private benefits of consuming these services. For example, some argue 25 GEORGE R G CLARKE AND SCOTT J WALLSTEN that people might not fully appreciate the benefits of consuming clean water if they are unaware of the costs associated with consuming polluted water or unable to fully assess the risks associated with doing so (Shirley and Menard 2002). If society is more concerned about the consumption of merit goods than it is about the overall level of utility attained by poor individuals, subsi- dies for these goods might be preferable to direct monetary transfers, because people may choose to spend cash transfers on something other than the service society intended. That some infrastructure services are merit goods and that society must ensure their provision is easily justified: people in cold climates will die without heat, for example. However, the justification is less clear in other sectors, and why universality is legally mandated in some sectors but not others is not at all clear. For example, why (1o so many countries have laws mandating universal access to telecommunications but not, say, to health care? Universal service laws in telecommunications, it turns out, do not have their roots in the desire to ensure telephone access to all people. Instead they originated with a desire by the Bell company in the United States in the early 20th century to stifle competition. Universal service in this context did not mean that everyone should have a telephone, but that everyone who did should have a Bell telephone (Mueller 1997). In other words, in telecommunicatLions universal service was meant to preserve monopoly profits, not to ensure service to everyone. As a result, laws to pro- mote universal service in telecommunications have tended to benefit monopolists instead of consumers. Understanding the origins of these laws in the various sectors can lead to a clearer understanding of where univer- sal service laws are justified and, where they are justified, the best ways to achieve their goals. Nonetheless, economics has nothing to say about what should be a merit good, and a good deal of evidence suggests that many countries consider some goods and services provided by network industries, most notably heat and clean water, to be merit goods. Estache, Foster, and Wodon (2001) note that pressure groups often demand access to utility services, suggesting that those demands might have more political resonance than demands for cash transfers. These groups appear to perceive that society values the con- sumption of infrastructure services more than it values the alternative uses of income by low-income households. In addition, vulnerable groups are 26 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE often guaranteed access to utility services, suggesting that society wants to ensure some level of access independent of income. Federal health legis- lation from the ] 930s prevents water companies in Mexico, for example, from completely disconnecting residential customers who fail to pay (Hag- garty, Brook, anci Zuliaga 2001). In the telecommunications sector many countries have explicit policies of promoting universal access to telecom- munications services, although these policies often remain unenforced. For example, Madagascar has a policy of providing a telephone in each village; Zambia has a policy of provilding telephone booths in public places, such as schools and health clinics; and Kenya has a policy of providing a tele- phone within walking distance of all residences (International Telecommu- nications Union 1998, table 4.5). The International Telecommunications Union lists universal access policies in 22 developing and transition economies, although in practice, many of the countries listed had not put any legal obligations upon the operator to actually provide access in line with the statedl policies (International Telecommunications Union 1998). Legal requirements and explicit policies designed to guarantee or promote access to specific services suggest that societies see utility services, espe- cially water, as more important than other goods or services. Polittcs and Regional Developmentt Strategies A government might wish to subsidize poor or rural consumers for political reasons or as part of a development strategy. For example, it may wish to subsidize service to the urban poor or to rural consumers because these groups have disproportionate political influence or to transfer resources to their supporters. 'To the extent that this promotes either efficiency or equity for the reasons outlined earlier, these transfers might appear relatively attractive. However, if subsidies are driven primarily by political motives, they can end up hurting rural and low-income urban households, which may have little political influence. In this situation wealthy households might benefit from tariffs set below cost, while poor households, which often remain unconnectedl, get nothing. Studies from several developing countries have found that subsidies often benefit middle-class and rich households, while having little impact on low-income groups. Politics often affect the distribution of subsicdies even when the subsidies were originally intendled to promote eqjuity. Once subsidies are introduced, 27 GEORGE R G CLARKE AND SCOTT J WALLSTEN they are often expanded to cover increasingly large portions of the population. For example, Boland and Whittington (2000) noted that most water supply utilities subsidize much higher levels of water consumption than is necessary to meet basic needs. Even though a five-person household would need to con- sume only between 4 and 5 cubic meters per month to meet international stan- dards for basic water use, 15 of the 17 water utilities in Asia for which Boland and Whittington had data subsidized more than this level of consumption, and 5 utilities subsidized more than 20 cubic meters per month. In other words, the biggest beneficiaries of the subsidies were large consumers, who are more likely to be wealthy. Furthermore, the authors noted that users reach the high- est tariffs only at extremely high rates of consumption, for example, about 80 times basic needs for a family with five members in La Paz, Bolivia. 2.3 FiPnancing Universal Serpvice Objecgives Justifying universal service is usually easier than implementing it. When society believes that everyone is entitled to a minimum set of services, sub- sidies may be necessary. This section reviews methods of funding univer- sal service obligations. These methods include cross-subsidies, the most common approach when a single firm provides service, and newer methods more consistent with competition and liberalization, such as universal service funds and auctions. Cross-Subsidies Under monopoly provision, whether by the state or by privately owned com- panies, cross-subsidies have been the primary way of funding universal service obligations. Cross-subsidies imply that some users are charged prices above cost to subsidize other users who are charged prices below cost. Cross-subsidies have several problems. First, they are inherently inefficient: by separating price from cost they distort consumption and investment decisions. Second, they are typically not transparent, making it difficult to determine who receives subsidies and who funds them. Third, they do nothing to encourage service to high-cost regions or to the poor, because the existence of monopoly profits from one group does not induce the firm to provide service to another group (Brook and Smith 2000). When- 28 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE ever any class of users is charged prices below cost, suppliers will have lit- tle incentive-or little ability if the enterprise has cash flow problems-to serve these users. Consequently, while subsidies will increase users' incen- tives to obtain service, they decrease the suppliers' incentives to serve them. Fourth, to make matters worse, most cross-subsidy programs were not care- fully clesigned to meet expansion goals in the first place (Chisari, Estache, and Laffont 1999); thus their tendency to be ineffective is not surprising. Even when subsidies can, in principle, promote efficiency, using cross- subsidies to pay for them may not be efficient. If positive externalities are present at all consumption levels, cross-subsidizing service to low-income consumers by charging high-income consumers prices above cost might be inefficient. Although this policy would encourage low-income consumers to consume more of that service-assuming that the infrastructure services are ordinary goods-the higher prices would discourage high-income consumers from consuming the same service. Dependling upon the relative price elas- ticities, cross-subsidies might therefore either increase or decrease total con- sumption of that service, with a similar impact on efficiency. Despite the problems of cross-subsidies, in many developing countries they might be more efficient than other methods of raising funds. While lump sum transfers from general tax revenues are, in theory, the most effi- cient means of subsidizing the people society wishes to help, the tax ancl transfer systems in developing and transition economies are themselves often distortionary and inefficient. For example, if countries rely heavily on tariffs or export taxes, redistribution by cross-subsidizing infrastructure prices might not be less efficient than redistributing income through the tax and transfer system (Cremer and others 1998a,b). Despite the possibility that cross-subsidies might be relatively efficient in certain circumstances, to our knowledge no general equilibrium studies of developing countries have compared tthe relative inefficiency of the two methods of redistribution; thus assessing these claims is difficult. Further- more, despite this uncertainty some policies, such as restricting competition in telecommunications, a service used by relatively few low-income house- holds in most developing countries, in order to allow a monopoly provider to cross-subsidize consumption, are likely to be highly distortionary, even in relative terms. Nonetheless, we will return to this issue later when discussing universal service funds, a method of funding service to the poor in the pres- ence of competition. 29 GEORGE R G CLARKE AND SCOTT J. WALLSTEN An Elusive Goal: Nondistortionary; Inexpensive, and Competitively Neutral Financing Mechanisms The disadvantages of cross-subsidies do not change the fact that, even where competition is likely to reduce prices, some sort of intervention is probably necessary to achieve full coverage of low-income households, especially in high-cost areas. For example, demand in some villages might be too low to support the cost of installing a payphone without subsidiza- tion. If low-income households are unwilling to pay the full cost of infra- structure services, subsidies might be required to ensure full coverage of vulnerable groups. Given that the empirical evidence that suggests that some low-income households in some countries appear to have limited willingness to pay for some infrastructure services, and that competition will make internal (firm-level) cross-subsidies difficult to maintain, gov- ernments will need to find new ways of financing service to low-income and rural households if universal service is to be a goal of public policy. Financing mechanisms should strive to be nondistortionary, inexpensive, and competitively neutral, that is, they should not distort consumption and investment decisions, should keep the cost of raising the funds low, and should not benefit one firm at the expense of others. Mechanisms to support universal service are competitive-neutral when one or several firms do not benefit or suffer relative to others in the indus- try. Nonneutrality would arise if one firm were obliged to provide universal service and raise funds for it while others were exempt from these require- ments. For example, in some cases incumbent firms may end up with an advantage over potential competitors if they receive subsidies or are allowed to maintain monopolies over certain services (for instance, inter- national service in the telecommunications sector) or in certain regions to meet universality conditions. In other cases the incumbent could be disad- vantaged if it is required to serve high-cost areas or low-income people while new entrants can choose to serve the most profitable customers with- out having to serve other less attractive customers. Unfortunately these policy goals are not entirely consistent with each other and present a difficult policy problem. Simultaneously minimizing distortion and costs and ensuring absolute competitive neutrality is impos- sible, and the point of universal access is not these three objectives, but to maximize the number of people connected to the network. Regulators and 30 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE policymakers should thus think of the problem as maximizing access sub- ject to sorne maximum acceptable level of consumption and investment distortions and expense. Two methods are typically proposed as ways to finance universal service obligations: through the country's general taxation system or through a uni- versal service fund. Economists typically suggest that subsidies should be financed through the general tax andl transfer system, and several develop- ing countries have developed programs along these lines. For example, Argentina provides tariff subsidies to pensioners by providing a dlirect pay- ment of US$13.50 per month to each person who receives the minimum pension in order lo pay for gas, electricity, and water (Chisari and Estache 1999). Although in theory this is the most efficient way to provide subsi- dies, practical problems are associated with it Taxation and redistribution systems in developing countries tend to be notoriously inefficient andl inef- fective; therefore raising, andl distributing money through these systems are likely to be expensive relative to other niethods (Chisari, Estache, and Laf- font 1999). Consequently, when this is the case, funding service to low- income households and high-cost areas through the firms that provide serv- ice may make more sense. While taxing some services and firms to SLIpport others is another form of cross-subsidly, competition requires mechanisms that support multiple firms. Universal service funds are just such mechanisms. All firms can be required to contribute to a universal service fund, and all firms that provide service slhould, be eligible to receive funds from it. In Lheory, a universal service fund provides a wider tax base and potentially reduces the potential for creanm skimming (Cremer andl others 1998b). Universal service fundls also make the financing mechanism more transparent, less costly, and more competitively neutral than cross-subsidies (Intven and 'retrault 2000). A practical probliem with these mechanisms is that someone must still deter- mine who, is. eligible to receive subsidies and how large they should be. How Large Should Subsidies Be? One approach for determining the appropriate level of subsidies is to auc- tion them, that is, firms can bid competitively for subsidies. In a fair bid- ding process, with multiple biddlers, firms will end up with the smallest sub- sidy necessary for them to provide service. Auctions can be especially 31 GEORGE R G CLARKE AND SCOTT J WALLSTEN useful in rural areas with little or no existing service. Chile and Peru were among the first to implement this method, giving licenses to those opera- tors that agreed to serve areas for the smallest subsidy (Cannock 2001). In Chile the average winning subsidy from 1995 to 1999 was about half the maximum subsidy offered, while in Peru it was only about one-quarter the subsidy offered (Intven and Tetrault 2000). These experiences have two important implications. First, the idea of auctioning subsidies is not merely a theoretical pipe dream; it has been successfully implemented in devel- oping countries. Second, it reveals that the subsidies necessary to serve remote locations are probably far lower than monopolists had previously claimed. As long as regulators lack information on the true costs of provid- ing service to remote areas-and regulators will always have less informa- tion than the firms-auctions can be an effective way to determine the true costs of providing service. 2.4 Who RTeceives Subsidies? Once funds have been raised, an important issue is how to distribute them most effectively to those in need. Another related question is how to iden- tify who should benefit from the subsidies. As discussed previously, uni- versality laws typically have two components: subsidies to high-cost, typi- cally rural, areas and subsidies to the poor. Each objective presents its own challenges and each mechanism to achieve the goal has its own unintended consequences. Rural Areas In industrial countries subsidies have tended to focus more on high-cost, mainly rural, areas, with less emphasis on supporting the poor. Crandall and Waverman (2000) estimated that in 1998 the United States spent close to US$2 billion subsidizing telecommunications in rural areas and only about US$400 million subsidizing telecommunications for the poor. Subsi- dies for high-cost urban and rural areas are also common in developing countries; for example, geographic price averaging (mandating uniform prices across the country) is common in telecommunications and postal services almost everywhere in the world. 32 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE Although these subsidies are sometimes explicit, for instance, cirecL sub- sidies for providers in rural areas, they are often implicit, with a single company providing the service and charging a uniform price across differ- enl regions regardless of cost. For example, in C6te d'Ivoire a single, pri- vate monopoly provides water services to more than 400 towns, charging a uniform tariff. The main justification for this arrangement is that, in theory, it allows the company to subsiclize service in high-cost small towns through the profits it earns in Abicljan, the largest city in C6te d'Ivoire (M6nard andl Clarke 2002a)." Similar arrangements are relatively common in the water supply and sanitation sectors elsewhere in Africa. The Water Utilities Part- nership (2000) reported that out of 48 African countries for which informa- tion was available, a single national company provided water supply serv- ices in 26 countries and a single company provided sanitation services in 25 countries.4 As discussed earlier, one problem with subsidizing service in high-cost areas by keeping prices below cost is that while low prices will generally increase demand in these areas, they will simultaneously reduce providers' ability and incentive to serve those regions (note that this will be true for both cash-strapped state-owned utilities and profit maximizing, regulated, privale utilities). Even worse, potential competitors have no incentive to serve high-cost areas if they are forced to charge low prices to everyone who happens to live there regardless of their willingness and ability to pay. The result of a policy of geographic price averaging can easily be no serv- ice or only limited service. Examples from developing countries where cross-subsidies have had this effect are plentiful. For example, Wellenius (2000) notedl Lthat in the 1980s nearly 400,000 Brazilian farmers and rural cooperatives were willing to pay the full cost of obtaining telephone service, but the monopoly provider was not allowed to charge them more than it charged urban customers, with the result that the firm provided no service in these areas at all. Similarly, Menard and Clarke (2002a) noted that the national water supply enterprise in C6te d'lvoire expanded service in Abidjan, the low-cost area, far more rapidly than it expanded service in higher-cost secondary centers in the late 1980s and early 1990s. Subsidies for high-cost areas are problematic for both efficiency and equity reasons. Any program that redistributes wealthl between groups will increase the welfare of the group receiving the subsidy at the expense of 33 GEORGE R G CLARKE AND SCOTT J WALLSTEN the other group. Cremer and others (1998a,b) showed that traditional meth- ods of universal service provision in the telecommunications sector have been net welfare reducing. In addition to being inefficient, mechanisms to support high-cost areas do not necessarily promote e(quity. Unless low- income households happen to be concentrated in highl-cost areas and lhave infrastructure connections, they will not berefit from subsidies for high- cost areas. For example, in a study of universal service in the telecommu- nications sector in the United States, Rosston and Wimmer (2000) found that cost-based programs do a poor job of targeting subsidies to low-income households. Even though rural poverty is a serious problem in many devel- oping countries, the extremely low level of infrastructure coverage in rural areas makes it highly unlikely that poor households will be the main bene- ficiaries of redistribution to rural areas. Targeting the Poor In addition to subsidies for rural areas, universal service laws typically also aim to make connections affordable for the poor. One problem with reach- ing this goal is finding ways to identify those eligible for subsidies. Several mechanisms can be used to target the poor, ranging from the precise to the broad (see Foster, G6mez-Lobo, and Halpem 2000 for more details). Each has its own advantages and disadvantages in terms of precision, cost, and unintended consequences. Common methods include identifying house- holds and neighborhoods and using block tariffs, where initial use is charged at lower rates than higher use. Subsidies targeted at the poor can be much more effective at increasing the number of people connected than subsidies for entire regions. Eriksson, Kasermain, and Mayo (1998), for example, found no evidence that geographically based subsidies affeci telephone coverage in the United States. In contrast, they found that tar- geted programs-that is, programs meant to help people who would have trouble paying for service-seem to positively affect coverage. They con- cluded that targeted programs are much more effective at increasing net- work connections than geographically based subsidies. The following sec- tions discuss various methods of targeting the poor. Targetug Households. One common method of providing subsidies is to base them on households' sociodemographic characteristics. For example, 34 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE in Chile households receive subsidies based on the size and composition of the household, the occupation and education of the head of household, the household's assets, the houselhold's income, and the characteristics of the dwelling.5 Although using household characteristics to target subsidies can identify the poor precisely, the mechanism hias two main drawbacks. First, it is expensive. Foster, G6mez-Lobo. and Halpern (2000) estimated that having social workers collect enough information (such as education level of the head of household, the materials used in housing construction, and the presence of other infrastructure services) to identify low-income households in Panama could cost about US$10 per household. This could become extremely costly in low-income countries, because these inter- views would need to be performed every few years. The cost of data collec- tion depends, of course, on the type of information collected; for example, relying on a house's observable characteristics would be much cheaper than relying on information that would necessitate a household interview, espe- cially as the characteristics of the hiouse might change less frequently than the socioeconomic characteristics of the house's occupants.6 The drawback is that if fewer data are collected, this decreases the precision of the target- ing. Second, collecting accurate data on household characteristics might be difficult when interviewees know that their answers will affect the price they pay for infrastructure services. The problems with collecting accurate data are likely to be magnified in countries where corruption is a problem. Targeting Neighborhoods. An alternative to basing subsidies on the house- hold's socioeconomic characteristics is to base theni on the socioeconomic characteristics of the neighborhood as a whole. Althoughl this method relies upon some knowledge of the socioeconomic characteristics of households within the neighborhood, this could often be derived using data from a cen- sus or other household survey without having to perform household inter- views. Even if recent census data or other household-level data are not available, this method will still generally be less data intensive thian inter- viewing every household applying for connections. Although in principle the approach could be used for both metered and unmetered connections, in practice it has generally been used for unmetered households. It was used, for example, for water supply for unmetered properties in Mexico City in the early 1990s (Haggarty, Brook, and Zuluaga 2001) and in Dar es Salaam, Tanzania. 35 GEORGE R G CLARKE AND SCOTT J WALLSTEN One drawback of this approach is that it will generally target households less precisely than household surveys, and some wealthy households liv- ing near poor neighborhoods will receive subsidies while some poor house- holds living in wealthy neighborhoods will not. For example, Foster, G6mez-Lobo, and Halpern (2000) found that if subsidies were targeted toward extremely poor urban households in Panama by paying subsidies for water use to all households in zones where more than 50 percent of the population was in extreme poverty and where telephone coverage was below 30 percent of the population, only 6 percent of households in extreme poverty would receive subsidies and 31 percent of subsidies would be paid to households not in extreme poverty.7 Furthermore, targeted sub- sidies based on census data become more imprecise between censuses, especially when they take place infrequently. This source of imprecision is likely to be an especially large problem in rapidly growing cities in devel- oping countries. A second problem is that subsidies based on geographic location are often opaque. This makes manipulating subsidies for political reasons easy, a problem that is more likely to be serious in countries where corruption is widespread. Block Tariffs. A final way of subsidizing infrastructure services, at least when service is metered, is through block tariffs. Under this system users are charged a low rate for the first units of consumption and progressively more for additional consumption. For example, users might be charged a low rate for initial units of electricity each month, but progressively more for additional kilowatts. The idea is that if infrastructure services are nor- mal goods, then block subsidies will be targeted to low-income households. Although used in many sectors, these tariffs are especially common in the water sector in developing countries. For example. 20 out of 28 utilities in Asia that used volumetric charges used block tariffs (Asian Development Bank data cited in Boland and Whittington 2000).8 While block tariffs are inexpensive to adminiister and benefit poor peo- ple who are connected to the network, they face many problems. First, everyone who is connected-poor and rich-receives the low rate on ini- tial usage, meaning that some portion of the subsidy will go to high-income households. Furthermore, when the initial block is large, middle-income households might actually benefit more than low-income households. Con- sider, for example, a two-part block tariff with the initial rate set below mar- 36 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE ginal cost and the higher rate set above marginal cost. Under this scenario households that consume the full allocation at the initial rate will receive the largest subsidy. As the initial blocks are often quite large in practice, low-income householcls that consume relatively modest amounts might actually receive smaller subsidies than middle-income households that consume the full amount. Second, these tariffs can, perversely, even hurt the poorest households (Boland and Whittington 2000; Whittington 1992). In the water sector, households not connected to the piped water grid often buy water from neighbors who are connected. Because these neighbors supply many households, they purchase large amounts of water, meaning that they are likely to far exceed the initial, subsidized, block. They will pay a high average price for water and will charge nonconnected families who buy from them accordingly. As a result, the poorest households, who are more likely to share connections or buy water from neighbors, can end up pay- ing the higher rate, while high-incomiie users with single house connec- tions pay the lower rate. The possibility that block tariffs end up hurting the poorest is not merely theoretical. Whittington (1992) founcd that relatively high-income users in Kumasi, Ghana, paid the lowest average price for water while relatively poorer households paid higher average prices. Finally, large households will tend to pay a higher average price than small households, because they generally consume more. To the extent that household size is a poor proxy for per capita income, subsidies will be misdirected. 2.5 Incorporating Universal Service into Reforms: Regulation, Privatization, and Access When low-income households have low willingness to pay for infrastruc- ture service, especially in sectors where externalities arise from universal coverage, for instance, public health externalities associated with water supply and sewerage, it is important to consider ways to boost coverage among low-income households and in high-cost areas without losing the efficiency benefits associated with privatization and increased competition. In this section we dliscuss several issues, including the privatization process, the role of regulators in setting quality standards, the ways to 37 GEORGE R G. CLARKE AND SCOTT J. WALLSTEN encourage service providers to provide services that are affordable for low- income households, and the approaches for cross-subsidizing service after the introduction of competition. Incorporating Universal Service in the Privatization Process Although privatization itself could help or hurt low-income households, the privatization process can be designed to affect access positively. Notably, licenses sold to private investors can mandate certain types of investment, including investment to increase access in low-income or rural areas. Although additional obligations included in the license will reduce the price an investor is willing to pay for a formerly state-owned firm, these price reductions need to be weighed against the positive impact on policy goals. Furthermore, assuming a fair bidding process, price reductions aris- ing from such obligations would be the implicit subsidy that would have been necessary to achieve the policy objective. For example, in Mexico, as part of its privatization, Telmex was required to install payphones in 20,000 mural areas over a five-year period to meet the policy goal of ensuring some telephone access in all villages with at least 500 residents (Wellenius 2000). Although price regulation is one important aspect of regulation, many reg- ulations also target quality. Although quality standards typically exist when public operators provide service, private operators might be more affected by quality regulation, both because privatization is often associated with the establishment of new regulatory authorities and because private operators are less able to ignore quality standards. Regulators and policymakers are generally tempted to promote extremely high quality standards, and engi- neering and design specifications are typically imported from advanced industrial countries (see chapter 6 by Warrick Smith and chapter 7 by Bill Baker and Sophie Tremolet). As a result, service may be costly and afford- able only to the elite. Furthermore, since the existing infrastructure may have focused on large-scale operations, providers, especially foreign own- ers from industrial countries, might not offer low-cost options. While high- income households probably value high-quality service, that is, service sim- ilar to standards in industrial countries, low-income households might prefer more flexible quality standards, especially if they are unable to afford high-quality, high-cost service. For example, if postal regulations mandate daily delivery everywhere, then the costs of service in rural areas will be 38 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE higher than they would othenvise be (Cremer and others 1998a), and these higher costs will need to be financed either through higher prices or through general tax revenues. Allowing people to decilde on the levels of price and quality should improve economic outcomes; for example, low-income households might prefer a low-quality telephone line or a telephone that only receives incoming calls to an expensive, high-quality digital line or to no service at all. Rate Rebalancing, Subsidies, and Competition Although reforms, especially those that increase competition, might lower the cost of service and, in so doing, reduce the need for subsidies, compe- tition also makes cross-subsidizing service more difficult. If competitors entering the market generally try to serve the most profitable customers first (cream-skimming), the profits needed to subsidize unprofitable areas will disappear and rates might need to be rebalanced, with some prices iis- ing toward cost and others falling. Consequently, low-income consumers might be hurt even if reform leads to lower average prices. Although this is a theoretical possibility, there are reasons to doubt that rate rebalancing resulting from increased competition will pose as serious a problem as some observers have suggested. For example, wlhile tariff rebalancing in the telecommunications sector has often led to increases in local residential tariffs, which were kept artificially low under monopoly provision, whether this has harmedl low-income consumers is not clear. First, households benefit from low prices for local service only if they have a telephone, something that is relatively uncommon among low-income households in most developing countries. Even when penetration is high, it has not been demonstrated that the poor value local service more highly than long (listance service. For example, in the United States rate rebal- ancing seems to have positive benefits for the vast majority of households, because long distance prices have fallen by far more than local prices have risen (Wolak 1996). Combined with evidence that competition substan- tially improved coverage, this suggests that the net impact on low-income households was positive even with rate rebalancing. Second, considering the impact of technological change is also important when thinking about the effects of competition on prices. Even if competi- tion appears to negatively impact low-income households in the near term 39 GEORGE R G CLARKE AND SCOTT J WALLSTEN because of rate rebalancing, competition can affect technological change, and thus prices, in the medium or long term. New technologies provide new options for serving rural and remote locations, potentially lowering the costs of serving some high-cost areas and reducing the need for subsidies. For example, while stringing wires over long distances may have been costly, fixed wireless systems may permit rural telecommunications serv- ice at relatively low cost. Especially in the telecommunications sector, competition may allow creative entrants to provide service in ways the incumbent never imagined; thus barring competition can prevent these advances from ever appearing. Third, although many subsidies are focused on usage prices, focusing on connection fees might be more appropriate, especially in countries where coverage among low-income households is initially low. While usage prices have often been low, connection prices have often been quite high, and in many cases actual connection prices are much higher than listed prices when bribes are required to actually get service. While long waiting lists for service demonstrate that demand for service exists even at high prices, extremely high connection charges make a mockery of any policy intended to connect the poor. In Nigeria in 1999, for example, where annual per capita income was about US$260 (World Bank 2002a), the connection charge for a telephone line was US$210 (Onwumechili 2001), high even by industrial country standards. As coverage is generally far higher among high-income households, and considering that low usage prices such as block tariffs benefit only connected users, focusing on connection fees is attractive, because it will benefit nonconnected households, many of which are relatively low income. 2.6 Empirical Analrysis of Histor:ical Universal Service Provision and the Effects of Liberal!ization Opponents of liberalization worry that reforms will hurt the poor, even if they improve efficiency. If new entrants are interested in providing service only to profitable high-income and business consumers, competition might force the incumbent provider to either abandon cross-subsidies or be left serving only unprofitable low-income and high-cost consumers. Further- more, critics claim that competition will erode monopoly profits, forcing gov- 40 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE ernments to find new sources of funds to finance access for high-cost and low-income consumers, something that could be difficult in developing coun- tries with inefficient and distortionary tax regimes.9 The implicit assumption behind these arguments is that countries have successfully managed to pro- mote access for vulnerable groups and to target cross-subsidies toward them prior to reforms; however, with the exception of Eastern Europe and Central Asia, the evidence suggests that monopolies have not used subsidies to serve the poor. In this section we use household data from around the world to investigate how well monopolies have served rural andl low-income con- sumers and how those consumers have fared under liberalization. Evaluating access by the poor to infrastructure utilities is difficult, as lit- tle consistent data on the subject are available. Cross-country data on telecommunications and electricity, such as those from the International 'relecommunications Union and the U.S. Energy Information Agency, respectively, for example, do not track connections by income group or regions of countries, and databases kept by utility conipanies generally do not provide the information needed to assess the impact of privatization on the poor (G6mez-Lobo, Foster, and Halpern 2000b). Even if the companies are willing to make their data available, they typically have information only on numbers of customers and do not collect detailed information on households' sociodemographic characteristics.iU Moreover, for obvious rea- sons, utility companies generally do not have detailed information about informal or illegal connections. The only way to obtain a consistent picture of access to infrastructure by the poor is through household surveys. Such surveys will generally have more information about houselholcls' socioeconomic characteristics and are less likely to omit individuals with informal or illegal connections. Unfortunately, household surveys are usually not designed to measure infrastructure use, meaning that they typically have limited information in this regard (see G6mez-Lobo, Foster, and Halpern 2000a,b for several recommendations that might make surveys such as the Living Standlards Measurement Study sur- veys more useful for analyzing infrastructure reforms). Furthermore, finding household surveys with similar information for years both before and after reforms took place can be difficult.'I Consequently, even case studies often have only limited information on the impact of reform on the poor. To address this gap in our empirical understanding of the subject, we exploit the MEASURE DHS+ demographic and health surveys (henceforth 41 GEORGE R G. CLARKE AND SCOTT J WALLSTEN referred to as the DHS surveys) to glean relatively consistent cross-country information. These household surveys, funded by the U.S. Agency for International Development, have been carried out around the world prima- rily as a tool for measuring changes in health status and the effectiveness of health-related initiatives (for more information, go to http://www. measuredhs.com/). As these surveys provide comparable information on a relatively large number of countries over time, especially in Africa, they allow us to compare coverage among low-income households in reforming and nonreforming countries and to look at how coverage has changed after reform. The main drawbacks to these surveys are that they contain only limited information about coverage, and as data on income are not avail- able, the education level of the head of household must proxy for income. In particular, we assume that households headed by someone with no edu- cation tend to be poor, while households headed by someone with at least a secondary education tend to have higher incomes. 12 Have Cross-Subsidies Supported the Poor? Despite purported attempts to subsidize services to low-income and high- cost users and the maintenance of state-owned or regulated private monop- olies to ensure that cross-subsidies are possible, there is little evidence that these attempts have been successful. In general, coverage of rural and low-income urban households is extremely low-considerably lower than coverage of higher-income households-especially in low-income coun- tries in Africa and Latin America. Table 2.1 and figure 2.1 show coverage for urban households headed by men with a secondary education or higher and men with no education in the late 1990s. Figure 2.2 and table 2.2 show general coverage data for urban and rural households. Several patterns-generally consistent with those observed in chapter 3 by Kristin Komives, Dale Whittington, and Xun Wu-are evident. First, urban households in Africa and Latin America are generally more likely to have electricity or piped water than they are to have a telephone or a flush toilet (table 2.2). In both regions, as well as in Europe and Central Asia, urban households are more likely to have electricity than piped water and more likely to have flush toilets than telephones. However, some variation from country to country is apparent; for example, in 7 of the 21 African countries, households are more likely to have water connections than elec- 42 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE tricity connections. Second, urban householdls in both low- andi middle- income countries in Eastern Europe and Central Asia were more likely to have infrastructure connections than households in Africa and Latin Amer- ica. Although urban households in Eastem Europe and Central Asia were more likely to have electricity than other infrastructure services, the pat- tern for other services was less clear than for Africa or Latin America. In general, householcls headed by indlividuals with a secondary educa- tion or higher were far more likely to have infrastructure connections than households headed by individuals with no education (figure 2.1), with the difference being especially large in low-income countries in Africa and Latin America. In low-income countries in Africa, about 80 percent of urban households headed by an individual with a secondary education had access to electricity, 63 percent had access to piped water in either their houses or yards, 20 percent had a telephone, and 38 percent had a flush toilet. In comparison, only 32 percent of urban hiouseholds headed by indi- viduals with no education had electricity, 27 percent had piped water, and only 10 percent had a flush toilet. Telephone coverage among urban house- holds in Africa headed by individuals with no education was especially low: less than 2 percent on average and less than I percent in most coun- tries (table 2.1). Although coverage was higher in low-income countries in Latin America, the basic pattern was simllar. The differences in coverage were not due to differences in only a few countries. For electricity, piped water, ancd telephones, coverage was lower-and in most cases much lower-for houses headed by incividuals with no edlucation than it was for households headed by individuals witlh a secondai-y education or higher in all low-income countries in Africa and Latin America for which data were available. [3 In middle-income countries in Latin America, similar patterns were observed for electricity, flush toilets, and telephones, althouglh on average, urban households headed by individuals with no education were slightly more likely to have access to piped water than urban households headed by individuals with a secondary education or higher. Eastern Europe and Central Asia appear different, with higher overall coverage in most sectors and less noticeable differences between households with heads with differ- ent education levels. Overall, this suggests that cross-subsidies have been relatively ineffective in targeting service for poor households. Several empirical studies support 43 O_0 00 .7 .7 as_ n CO zt O~ O Cl o C CX O D C o 0 N t -' o IIIXI ~~lo co m N x @o v o N cn ct co c X vi 0 X~~~ Ct 04 X CL, CD -r CC Oto to trcvO Ni - _ Li, X ISJ;oe cD c o, 2 c0=1 Y=° t > - ct1, nt Z~ ::5 c) m~ _ - B ° *U ° 'N 0 o A 0 t - N i N 0 t* m 0 . ~ 0 LtA , A .2 _ .CQi- -L i00> 0Oa , N oo i Ic l v ; i 2:, _=t A , o, o too200 0 o0 20 3 m - - - ,- - - o x w -C Li . e E_ E E*,'0 44, 0 Ct - - - - - - i, 0 0- -'N - c- m mm m ', m - o ii-~~~~~~~~~~~~~~~~~z > o m.2 QL c C' m l Cc0 m .i LiC . to…to1 c clcD c l > c o m~~~oN. m t i ~t m iL o N lo N.00 to L m 0 0 -C 0~~~~~ 0~~~~~~~~~0 oj le E - _j ) 0t, _~ co > :. * O 00 0 L, N N 0 m 20 mEi 0 0 ii t i , . 0 UtWc E . -= ' -o ~ ~ ~ z6E - jk 440 t l-E = L X , ,_ s ° ° R t X t ° s a X~~~~~~~C 0 C, 00 ON. - 0 'n .n 'o 0- 0 nC CL m CO NO C N N on 0 0 0 II I_ _-O _ rON )N . rs v r sroo a, 1o m °O £ Ob£ > -o C o _E - N N N. N - mmo CO I m 'n - - n -ED ~~~~O~~ N0 n N) _. -N O CO N. 0 ON - S .10 N. 0 - ON N XO n I N.. On nh-O ON- O -nN * nn C N . OLnN . CO 00 . 10c CO ,, o O N N. , N m 00 O 1 0 oE J-1 _o C .0m~ ~~ ~ ~~~~~~~~~~~~ r5 _0'o O - O 0 I n N 1,1 C n~ m 1,1 . O 0 m NO 0o o -0 .0 .0 ~ ~ ~ ~ ~ ~ ~ ~ ). N. In ao o o - u n ° N 0 - 0N' m N. ° 00 - In O '° 0 Cn I o 0O a o -N mD 00 N N- 0 -O N. Nh-0 o CO In ON CO C - a, r e a, a, 0 a, o, a, a) a, a,SU 5U w 0~1 E Z 00 NC. 0 N N. N. _ n N In In 0 h C -In CI 0 0 0 N n 0 mN mN 0 N. C 0 C N O 0 NO m n 0 0 C a .0 0~~~~~~~~~ - 'm~~~~~~~~~~~ - - >1tsC N. CO O o a, n a, a, m h- - N. 100 0- g i gi g , 98gg T_o _ ~~~~~~~~~- -O - s. - m 0 0 0 oS m O m Q) m 0 c~~~~~~~~~~~~~~0)0 2 ac T B, 'r m 00 00 00 00o 00 0 -'1 o o .> N. N.m 0 N. - 0" OnE cN. E 0 o 0 0 o 00Jk< -Q, M E 0 00 0 - 0 00 00 00 00 000E m00Z 10 10 0) 20 .0 = E ~ ~ ~ I EE C ~ ~ ~ ~ ~ ~ ~ ~ ~ 9 0 - -oO, 0) 0) C~~~~~~~~~~~~w 4 GEORGE R G CLARIKE AND SCOTT J WALLSTEN FIGURE 2.1. Infrastructure Access for Urban Households with Heads with Different Education Levels, Developing and Transition Economies,1990s Electricity (percentage of households) 100 __ 80- 70- 60 50 40 30 20 1 0 Sub-Saharan Latin America Eastern Europe Latfn America Eastern Europe African and Caribbean and Central Asia and Caribbean and Central Asia Low Low Low Middle Middle Water (percentage of households) 100 - 90 80- 70 - 60 - 50- 40- 30- 20- 1 0 Sub-Saharan Latin America Eastern Europe Latin Amerca Eastern Europe African and Caribbean and Central Asia and Caribbean and Central Asia Low Low Low Middle Middle 7 Secondary education ] No education 46 Telephone (percentage of households) 100 90 80 70- 60 - 50 40 - 3 0 2 0 1 0 0 Sub-Saharan Latin America Eastern Europe La in America Eastern Europe African and Caribbean and Central Asia and Caribbean and Central Asia Low Low Low Middle Middle Flush toilet (percentage of households) 100 90 - 80 - 70 - 60- 50 - 40- 30- 20- 10 lo- It Il I I Sub-Saharan Latin America Eastern Europe Latin America Eastern Europe African and Caribbean and Central Asia and Caribbean and Central Asia Low Low Low Middle Middle * Secondary education LZ No education Note Data are for urban households headed by men, which are classified based on the educational level of the household head Coverage implies that the household has a connection to that service in its house (or yard for water) Regional averages are computed as simple averages (no weight- ing) Income classifications for countries are based on classifications in (World Bank 2002b) Data are for all countries in these regions for which data were available for various years between 1994 and 2000 Source Authors' calculations based on raw data from MEASURE DHS+ Demographic and Health Surveys 47 GEORGE R G. CLARKE AND SCOTT J WALLSTEN FIGURE 2.2. Infrastructure Access for Households in Urban and Rural Areas, Developing and Transition Economies, 1990s Electricity (percentage of households) 100- 90- 80- 70- 60- 50- 40- 30- 20- 10 I m_-1 X- Sub-Saharan Latin Amerca Eastern Europe Latin America Eastern Europe African and Caribbean and Central Asia and Caribbean and Central Asia Low Low Low Middle Middle Water (percentage of households) 100- 90- 80- 70- 60-I 50- 40 30- 20- 1 0 0- Sub-Saharan Latin America IEastern Europe Latin America Eastern Europe African and Caribbean and Central Asia and Caribbean and Central Asia Low Low Low Middle Middle D Urban g Rural 48 Telephone (percentage of households) 100 - 90 - 80 - 70- 60- 50- 40 30 2 0 1 0 Sub-Saharan Latin America Eastern Europe Latin America Eastern Europe African and Caribbean and Central Asia and Caribbean and Central Asia Low Low Low Middle Middle Flush toilet (percentage of households) 100 - 90 - 80 70- 60- 50- 40 - 30 - 20- 10 0- - Sub-Saharan Latin America Eastern Europe Latin America Eastern Europe African and Caribbean and Central Asia and Caribbean and Central Asia Low Low Low Middle Middle [2 Urban :2 Rural Note Low are low-income countries, Middle are middle-income countries Income classifications for countries are based upon classifications in (World Bank 2002b) Regional averages are computed as simple averages (no weighting) Classifications of urban and rural households are based on original classifications in the DHS+ datasets Coverage implies that the household has a connection to that service in its house (or yard for water) Data are for all countries in these regions for which data were available for various years between 1994 and 2000 See Table 2 2 for more information Source Authors' calculations based on raw data from MEASURE DHS+ Demographic and Health Surveys 49 E j a 2. _ o ~ ~ ~~~C ° CZ In CC U~~~~~~ ,,, CS C> CD C> ,~ C 80t- D ov E~~~~ 0 0 _x C Q o _0 N..-0r'~ 0.-CL 000 Li, 40000 o Li s - 0 a 0c. o CL > to_ H CG C. - ur O CZ, er - m .0 _6 _00.D 0 0- tw o - - _ - Ro_ __ c O5 c, B N t L £ 0 L _ -6 X so ff XQX- .C C D 0 t 4- - Li, b~~~~~~~~~a 0)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~f CU 0…Li, L< O ~ ~ e CD ec fu M~~~~~~~i - oi L i 0 Li > d~~~~~~~ - 040- -e50 . ODN 0 (N CO 0O '0 0 _ a ( 0, 1 0 ° C1 0'-. 'oo° cC ," Cl ~00 N 0~- '0,- a: .- 0 om r 00,_ _ - I- O c oI 00,0 _-- N U' I) a,y, ,' E "I oo c, In c- _- CV 's( 00 - N 00s -(0 '0vu °C o ur~~~~~ o v _ v o o r 9_Or . rvC'X _ C- _o V o o o cn _so otrn oD c m a m co c> -o t r- 0- lo,0 . a Nn c N N a, c~~~~~~~~~~~ o _ o(0 0 U' 0 , c In 0 No N N 000 (N 000 0 a, S 0 3,0000 R' lo ( 0N - - -rC0 N 00 O coP lo E o 0 , a, a, N a, 0 a, a, 0a, ,0C o0 S o 2 s:~~~~q a c D O 3 ° ° 5 m ' XC E 76Ee< ' eX uoE z E - n En 2 0 51 Q-j ~~~~~~~~~~'0C'0 o Q- 11 o o~~~~~~~~~~~0 f 0' 000 00 00 00 0 0 0 0 0 0 -~~~~~ E0 00 N N ,U' N N - 'U0 U'N.0 U'00 0 N- ( N( 0 U' 00 (N20 ' Nd ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 0. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 5 1 GEORGE R G CLARKE AND SCOTT J WALLSTEN this observation. For example, even the subsidy scheme used in Chile, which was based on household characteristics and which observers have perceived as quite successful (see, for example, G6mez-Lobo, Foster, and Halpern 2000b) has failed to target low-income households effectively. Only one-third of households receiving subsidies in 1996 were in the lowest quintile (Shirley, Xu, and Zuluaga 2002), and about 23 percent of subsidies go to households with income higher than the median (G6mez- Lobo 2001). Similarly, Walker and others (2000) estimated average monthly subsidies for piped water for consumers with different incomes in six cities in El Salvador, Nicaragua, Panama, and Venezuela, and found that the subsidies appeared to benefit rich and poor consumers to similar degrees in Nicaragua, Panama, and Venezuela. In El Salvador they found that all consumers appeared to be overcharged, although poor consumers appeared to be overcharged less than high-income consumers. In addition, Whittington (1992) found that an increasing block tariff for water in Kumasi, Ghana, resulted in the poor paying higher average prices than better-off households.14 Finally, Waddams-Price (2000) lists several examples, including kerosene subsidies in Ecuador and Indone- sia, electricity subsidies in the Republic of Yemen, and public transport and water subsidies in Hungary, where subsidies failed to effectively tar- get the poor. Have Cross-Subssidies Supported Rural Areas? The data presented demonstrate only that coverage of the poor is extremely low for most infrastructure services, suggesting that subsidies have not been effectively targeted. We can explore the question of whether subsi- dies have benefited rural areas somewhat more directly. In addition to differences in coverage for low- and high-income house- holds, similar differences in coverage are apparent between urban and rural areas in most developing countries (figure 2.2). Rural coverage is generally lowest in low-income countries in Africa. For example, about 47 percent of urban households in Africa had electricity, 37 percent had piped water, 18 percent had flush toilets, and 6 percent had telephones. In com- parison, only 7 percent of rural households had electricity, 4 percent had piped water, 1 percent had flush toilets, and 0.3 percent had telephones. In almost half of the countries in Africa, fewer than 1 in 1,000 rural house- 52 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE holds had a telephone and only in Zimbabwe did more than 1 in 100 rural householdls have a telephone. Although some households might have access to telephones outside their homes, many households probably do not have any access to telephones. Onwumechili (2001) estimated that by 1999 nearly 75 percent of the world's population had never made a tele- phone call. Although coverage rates were higher among rural households in Latin America than they were in Africa, the basic patterns were similar (figure 2.2). Few rural households in the two low-income countries in Latin Amer- ica for which data were available had electricity, piped water, telephones, or flush toilets, with coverage especially low for telephones and flush toi- lets. Although coverage appears higher in midclle-income countries in Latin America, especially for electricity, where more than 50 percent of rural households had coverage, coverage among rural households is far lower than among urban households. Once againi, the countries in Europe and Central Asia for which data wvere available generally appearedl to have significantly higher coverage for rural households than in Africa and Latin America. In particular, coverage for electricity was higher than 90 percent in rural areas for all three countries for which data were available, compared with close to 100 percent for urban areas. However, coverage for otller services-although generally higher than in other regions-was far lower in rural areas than in urban areas. We can also test somewhat more directly whether cross-subsidies tendled to be used to fund high-cost areas. In particular, we know in which coun- tries a single firm provides water in all urban areas and in which countries service is provided at the local or state level. Moreover, in these countries the capital city tends to be the largest city with the highest concentration of high-income consumers.i5 If cross-subsidies were used to support cov- erage in high-cost areas, *ve wouldl expect to see less variation in coverage between the capital and other urban areas in countries witlh a smgle water firm, hecause that firm wouldl be able to use income from the capital to fundl water provision elsewhere. Table 2.3 shows coverage in African capital cities and all urban areas by sector structure when service is provided by one firm throughout the country and when it is organized at the subnational level. The table yields two interesting results. First, coverage in the capital andl in all urban areas in countries with a single provider is, on average, lower than in countries 53 GEORGE R G CLARKE AND SCOTT J WALLSTEN TABLE 2.3. Access to Piped Water Where Urban Water Supply Is Provided at the National and Local Levels, Africa, Various Years 1994-2000 Access to piped water (percentage of households) All urban as share Country Capital All urban Difference (%) of capital (%) Local or state 52 40 37 77 14 6 74 38 Kenya 82 6 61 3 21 3 74 21 Madagascar 31 7 17 7 14 0 55 84 Mozambique 28 6 23 4 5 1 81.82 Nigeria 25 2 24 4 0 8 96 83 Tanzania 78 8 48 2 30 6 61 17 Togo 67 5 51 6 15.9 76 44 National 42 77 30 88 11 9 74 08 Benin 981 56 5 41 7 57 59 Burkina Faso 27 3 25 2 21 92.31 Cameroon 431 28 5 14 6 6613 Central Affican Republic 10 2 5 0 5 2 49 02 Chad 211 11 6 9 4 54 98 C6te d'lvoire 638 51 0 12.8 79.94 Ghana 66 0 41.5 24 5 62 88 Guinea 39 6 30 0 9 6 75 76 Mali 174 158 16 9080 Niger 33 9 27 2 6.7 80 24 Senegal 79 4 65 4 14 0 82 37 Uganda 13 3 12 9 0 4 96 99 Note Access to piped water implies that the household has piped water in either its house or compound Information on decentralization is from Water Utilities Partnership (http l/www wupafrica org/) Nigeria has state-level provision of its water supply Source Authors' calculations based on DHS surveys where provision is organized at the subnational level. This result is consistent with the notion that competition improves service, even if it is only bench- mark competition. Second, coverage outside the capital city relative to cover- age in the capital is not, on average, higher in countries in Africa where the water supply is organized at the national level than it is in countries where the water supply is organized at the local level. (1 This result suggests that cross- subsidlies have not been used to support service provision in high-cost areas. 54 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE Effects of Privatization and Competition The fact that subsidies did not appear to serve rural and poor consumers does not by itself imply that reforms will automatically benefit these groups. Although on the one hand case study evidence suggests that pub- lic monopolies have often been overstaffed and inefficient and have lacked the resources needed for investment, on the other hand tariffs have often been heavily subsidized from general government revenues and companies have often cross-subsidized certain consumers or services. The net impact that reform has on coverage will therefore depend on whether it removes constraints on investment (supply) and how it affects prices paid by low- income consumers (demand). If coverage is low because enterprises in developinig countries lacked the resources needed to expand the system, then low-income consumers might benefit from reforms even if they result in higher prices. If coverage of low-income households is low because the poor have low willingness to pay rather than because service is rationed, then ignoring any impact that reform has on quality, reform will benefit the poor only if it results in lower prices. Prior to reform, water utilities in developing countries, which were mostly publicly owned, often charged prices far below costs (see, for example, World Bank 1994). Many case stuclies have noted that the poor financial performance of many public utilities, combined with governments' poor fis- cal situation, resulted in utilities having insufficient financial resources to finance investment and maintenance. As a result, utility companies heavily rationed service (in relation to water supply, see, for example, the case stud- ies in Shirley 2002, especially Menard and Clarke 2002b and Alcazar, Xu, and Zuluago 2002, and the case studies in Savedoff and Spiller 1999). When services are rationed, that is, when many households want service but are unable to get it, low-income households might be especially unlikely to get service, because several market and nonmarket mechanisms direct serv- ice toward relatively wealthy and politically connected individuals. First, houses connected to the system will generally command higher sales prices or rents than nonconnected houses. People willing or able to pay the highest sales prices or rents will be more likely to acquire resi- dences with connections. Second, bribes and other side payments are often necessary to get a connection when service is rationed. Again, wealthier individuals will be the ones most likely to get a connection regardless of thie official tariff rate. Although little empirical evidence is available at the 55 GEORGE R G. CLARKE AND SCOTT J WALLSTEN household level, Clarke and Xu (2002) found evidence for enterprises that is consistent with this hypothesis.' Finally, high-income households will generally have greater political power than other households and might be more willing to make campaign donations or informal payments to politi- cians to ensure that infrastructure services are provided in high-income neighborhoods first. Consequently, when services are rationed we might expect low-income households to be less likely to receive service even if they are willing to pay official tariffs and connection fees. Addressing the question of why the poor tend not to be connected to infra- structure utilities is crucial, as full cost recovery has been a cornerstone of infrastructure reforms. The hope was that cost recovery would allow utilities to become self-supporting rather than having to rely on government subsi- dies, and given many governments' poor performance in providing subsidies consistently-and in many countries even to pay their own utility bills con- sistently-full cost recovery was often seen as a precondition for introduc- ing private sector participation (see, for example, Menard and Clarke 2002a,b and Clarke, Menard, and Zuluaga 2002). If low-income households are generally willing and able to pay for infrastructure connections, then even if pnces increase, reforms that remove constraints on investment and allow nonconnected households to connect to the system will benefit non- connected households that were previously unable to get connections. In contrast, if low-income households are willing to pay only relatively modest prices, full cost recovery might not be consistent with universal service goals without massive cross-subsidization. In the telecommunications sector, the sector on which most cross-country empirical work has focused, strong evidence indicates that reforms that increase competition and privatize state-owned utilities increase service availability.'8 Almost without exception, cross-country empirical research in both industrial and developing countries has found that competition increases the number of telephone connections (Li and Xu 2001; Petrazzini 1996; Ros 1999; Wallsten 2001a). The evidence on privatization is less conclusive, with some studies finding that privatization is associated with improved coverage (see, for example, Ros 1999), and others finding that it has little impact (see, for example, Wallsten 2001a).19 The evidence from cross-country studies is generally consistent with the experience documented in country case studies, which have often found that reform increases penetration and reduces retail prices (see, for example, Galal and Nauriyal 1995; Wellenius 1997). 56 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE These results are not surprising when considered in a historical context. In the late 19th and early 20th centuries in the United States, Mueller (1997) found that telephone service expanded at 5 percent per year under a Bell monopoly and 40 percent per year under competition. Similarly, Wall- steni (2001b) found that telephone penetration and rural service in early 20th century Europe increasedl more quickly Linder competitive regimes than under monopoly provision. Recent experience following reform has been similar. Mueller (1997) found that telephone penetration increased after the 1984 AT&T dlivestiture, ancd that the growth rates were highest among low-income groups and in regions with low telephone densities. When privatization and liberalization result in price decreases and serv- ice expansion, low-income households should benefit from reform. Low- income households with connections will benefit from lower prices and low-income households that connect to tlhe system, either because of price dlecreases or because of decreased rationing, will also benefit (households might also benefit from changes in service quality). Even nonconnected households might benefit indirectly if the drop in tariffs affects prices charged by resellers, for example, by water vendors or at payphones. In contrast, when privatization and liberalization result in price increases or the regularization of unofficial or illegal connections (whichi effectively increases prices for these consumers from zero to the official tariff), the impact on low-incomiie households is less clear. Although price increases will generally hurt low-income households that alreacly have connections, increases in coveiage will benefit those households that are able to get con- nections. That is, if privatization and sector liberalization remove con- straints on investment tlhal were the result of either the poor performance of a public utility or prices being set below cost, those households that were unable to get connections under public ownership might benefit despite price increases. Note, however, that increases in coverage following reform might not ben- efit consumers at all income levels. If price increases cause some house- holds to dlisconnect, something that might be common among low-income households with low wvillingness to pay, disconnected householdls will lose from reform.211 If many middle- or high-inconme households willing to pay high prices for utility services are able to connect because of the removal of investment constraints while a smaller numiibei of low-income house- holds disconnect because of the price increases, low-income households 57 GEORGE R G CLARKE AND SCOTT J WALLSTEN might suffer even if total coverage increases, that is, if more middle- and high-income households connect than low-income households disconnect. While household data that cover countries and times where reforms have occurred are rare, the DHS data allow us to test the effects of reforms in several sectors, countries, and time periods. In particular, we can compare access in African countries with public and private water operators, com- pare public and private telecommunications providers in Africa and Latin America, and make time series comparisons of the effect of private sector participation in the electricity sector in Latin America. Cross-country evidence from the DHS surveys companng countries with public and private operators does not generally support the assertion that public operators are better at serving low-income households than private operators. Table 2.4 compares coverage for low-income households in the mid- to late-1990s for countries with public enterprises in the water sup- ply sector, countries that had recently introduced private sector participa- tion (within two years of the survey date), and countnes that had had pri- vate sector participation for many years. On average, coverage among households headed by an individual with no education appears slightly lower in countries with public operators (25.4 percent) than it is in coun- tries with established private operators (30.6 percent). Coverage for house- holds headed by individuals with no education was higher in C6te d'lvoire than in 11 of 17 countries with public operators and higher in Guinea than in 9 of 17 countries. Conclusions are similar when comparing countries based on the share of connected households with no education as a per- centage of the share of connected households with a secondary education. Cross-country evidence on access to telecommunications services in Africa and Latin America leads to similar conclusions (table 2.5). Cover- age for households headed by individuals with no education is similar in African countries with public operators and privatized operators. In Latin America coverage actually appears to be lower in countries with public operators than it is in countries with private operators. Coverage among households headed by individuals with no education is lower in both coun- tries with public operators than in any of the four countries with privatized operators. Thus there is little evidence that public ownership benefits low- income households in terms of coverage, and in some cases coverage appears lower among households headed by individuals with no education in countries with public operators. 58 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE TABLE 2.4. Piped Water Coverage for Urban Households Headed by Individuals with Different Levels of Education, Africa, Various Years Percentage of households No education with piped water as share of Secondary secondary Country Year education No education Difference (%) education (%) Public 62 6 25 4 372 40 58 Zimbabwe 1999 96 5 85 8 10 6 88 91 Kenya 1998 716 494 22 3 6899 Tanzania 1999 65 9 45 9 20 0 69 65 Comoros 1996 521 321 200 61 61 Nigeria 1999 34 9 18 3 16 6 52 44 Togo 1998 77 8 39 5 38 3 50 77 Ghana 1998 64 7 25 6 391 39 57 Benin 1996 86 1 31.8 54 3 36 93 Zambia 1996 81 4 240 574 2948 Uganda 1995 25 5 3 5 22 0 13 73 Cameroon 1998 66 0 13 2 52 8 20 00 Niger 1998 70 S 17 4 53 1 24 68 Burkina Faso 1998 62 2 11 4 50 8 18 33 Chad 1997 34.7 9 6 251 27.67 Mozambique 1997 636 118 518 1855 Madagascar 1997 52 3 3 0 49 3 5 74 Mali 1996 58 5 9 7 48 8 16 58 Private sector participation (recent) 61 5 32 7 28 8 53 17 Senegal 1997 91 2 60.3 30 9 6612 Central African Republic 1994 31 7 5 1 26 6 16 09 Private sector participation (established) 65 7 30 6 35 1 46 58 Guinea 1999 47 0 23 8 23 2 S0 64 C6te d'lvoire 1994 84 4 37.4 47 0 44 31 Note Access to piped water implies that the household has piped water ir either its house or com- pound Private sector participation includes lease contracts (C6te d'lvoire, Guinea, and Senegal) and management contracts (Central African Republic) Central African Republic and Senegal had private sector participation for one year before the survey was taken Guinea had private sector participa- tion for 10 years and C6te d'lvoire for 35 years Information on private sector participation is from Water Utilities Partnership (http l/www wupafrica org/) Source Authors' calculations based on DHS surveys 59 GEORGE R G CLARKE AND SCOTT J WALLSTEN TABLE 2.5. Telephone Coverage for Urban Households Headed by Individuals with Different Levels of Education in Africa and Latin America Percentage of households No education with telephones as share of Secondary secondary Country Year education No education Difference (%) education (%) Africa, privatized 13 1 1 7 11 50 12 98 Guinea 1999 14.8 2 7 1210 18 24 Ghana 1998 147 1 9 12 80 12 93 Madagascar 1997 9 9 0.3 9.50 3 03 Africa, pubhlc 21.5 1 7 19.80 7 91 Uganda 1995 5 8 0 0 5 80 0 00 Kenya 1998 15 4 5 8 9 60 37 66 Nigeria 1999 10 7 0 9 9 80 8 41 Mali 1996 13 4 10 12 40 7 46 Niger 1998 14 6 0 9 13 70 6 16 Mozambique 1997 25 0 04 24 60 1 60 Zimbabwe 1999 45.8 00 45 80 0 00 Chad 1997 10 4 0 3 10 00 2 88 Cameroon 1998 16 4 3 6 12.80 21 95 Central African Republic 1994 29 9 1 3 28 60 4 35 Burkina Faso 1998 34.9 3 0 31 90 8.60 Comoros 1996 35 4 2.9 32 50 8.19 Latin America, pnvatized 58.9 19 0 39.90 32 26 Colombia 2000 81 2 39 4 41 80 48 52 Dominican Republic 1996 60 5 20 0 40 40 33 06 Peru 1996 38 0 8 6 29 30 22 63 Bolivia 1998 56 1 8.0 48 20 14 26 Latin America. public 502 41 4610 8.17 Nicaragua 1998 39.5 5 3 34 20 13 42 Guatemala 1998 60 8 2.8 58 00 4 61 Source Authors' calculations based on DHS Surveys Note Coverage is for urban households, which are classified based on the educational level of the household head Privatization information is provided by the International Telecommunications Union. All privatizations other than in the Dominican Republic are recent (within the past five years) Data are for all countries in these regions for which data were available for various years between 1994 and 2000 60 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE Time series evidence from the DHS surveys is also generally consistent with the hypothesis that private sector participation does not harm, and may actually help, low-income households. Figure 2.3 presents some evi- clence on the impact of reform on service to poor households in the elec- tricity sector in Latin America for the four countries for which pre- andi postreform data were available. In three of the four cases, Brazil, Colom- bia, and Peru, prices, which were significantly lower than similar prices in countries of the Organisation for Economic Co-operation and Development before reform, increased significantly. In the fourth country, Bolivia, which had the highest prices before reform, prices fell slightly. Coverage for urban consumers, which was already high prior to reform, increased following pri- vatization in all four countries. As this occurred despite increased prices in three of the four countries, this suggests that capacity was constrained prior to reform. Coverage also increased for the poor in three out of four cases, despite large price increases in two of the three countries. Although total coverage increased slighLly in Colombia-it was close to 100 percent even before reform-coverage of households with heads with no education fell slightly. These results are interesting for two main reasons. First, despite large price increases in three of the four countries, coverage actually increased in all four countries, suggesting that supply constraints playedl an impor- tant role in blocking service to the poor under public ownership. This is consistent with cross-country evidence from the telecommunications sec- tor. Ros (1999) found that higher residential subscription prices were cor- related with higher coverage in a sample of 110 industrial and developing countries. He interprets this as indicating that supply-sidle constraints were more important than demand-side constraints. Second, increases in cover- age for the total urban populationr might not always imply increased cover- age for the poor. In Colombia, while coverage in urban areas appears to have increased slightly overall following reform, coverage appears to have fallen slightly among households headed by individuals with no eclucation. Figure 2.4 presents similar evidence for the effect of private sector partic- ipation in the water supply in Africa. In December 1995, following unsuc- cessful reform of the public sector water utility in Senegal, the Senegalese government signed a lease contract with S6n6galaise des Eaux, a private sec- tor company with Saur International as its majority shareholdler, under which Sen6galaise des Eaux would become responsible for maintaining water 61 0 0 e -> -U t N 0 o o o to o (N o0 N - _ _ - B 0 > X I - , _ _ - 00 _ - 0 CDc 0~~~~~~~~~~~~~~~~~~~~~ (U - v .... _ o_C n F _ W- Wc E -- E, 0 *, _ _ _ _ No 2 00 2 a~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1 C~~~~~~~~~~~~~~~~~~~~~ (U - _W iD, * m- X O X5 0 0 - o __ o~ --4 [_ o I 0 % 0 0 0 0 0 0 0 0 0 0 0 C C,o -~ £ 62 ~ ~ ~ ~ ~ ~ ~ ~~~ (U 0~~~~~~~~~~~~~~~~~~~~~~~~~~' 0)- __ -__ _- I0 -~~ ~~ o - __--- ---~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~> 0 ~~~~~~~~~~~ -00 ~ C 0 cm 0 -j ~ ~ ~ ~ - LU -~~~~~~~~~~~~~~~~~~~~~~ ~~~~< ~~~~~~~~~~',0 0~~~~~~~~~~~~ 62 .0 0~~~~~~~~~~~~~~~~~~~~~~ cl >~~~~~~~~~~~~~ E0, o a)~~~~~~~~~~~~~ a) 10.0 o U '-4-b 0,B c 2T > 2 ~~~~~~~~~~0 ~ ~ ~ ~ ~ ~ ~ ~ U _ ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~ o ' > la> 0 '0 o ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ I- - E0 X~~~~~~~~~~~~~,, - c D 0) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 10 0-0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1 0 0 - I - - - - ~ ~ ~ ~ ~ ~ ~ I j' tO t C ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~o < 63 GEORGE R G CLARKE AND SCOTT J WALLSTEN services in 54 towns in Senegal. As the government was concerned about the political sensitivity of transferring assets to a private operator, and believed that it would have been difficult to convince private sector companies to assume the risk associated with debt service, the government opted for a lease contract (Kerf 2000), and in addition, the private operator agreed to under- take some investment in the distribution network. As one of the main goals of the reform was to make the utility financially self-sufficient, modest price increases of about 2.4 percent per year were planned between 1996 and 2003. Following the introduction of private sector participation, the number of connections in Dakar appeared to increase modestly, from about 135,414 in 1995 to 147,212 in 1997 to 157,429 by 1999. The expansion was accompanied by modest price increases; for social and political considera- tions prices were increased by only about 3 percent per year in the early years following reform (Kerf 2000). D)espite the price increases, coverage appears to have increased for botlh high- and low-income households. Between 1992 and 1997 coverage for urban households headed by a man with secondary education or higher increased by about 1.4 percent per year, while coverage for urban households headed by a man with no educa- tion increased by about 3.2 percent per year (figure 2.4). These increases compare favorably witlh the annual increases observed in other African countries between the early and late 1990s. Of the eight countries with public utilities where similar data were available, coverage grew more slowly for low-income households in seven of the eight countries and more slowly for high-income households in all eight countries. Although these results suggest that private sector participation can ben- efit the poor even when reform is combined with modest price increases, the benefits are not automatic. For example, in 1992 a 30-year concession contract to supply water and sanitation services in Buenos Aires was awarded to a private company. The contract was awarded to the consortium that agreed to the largest price reductions for connected customers (see Alcazar, Abdala, and Shirley 2002 for details). Although it seems plausi- ble that consumers would benefit when contracts are bid based on price reductions, the benefits of the large price decreases accrued to mostly mid- dle- and upper-income customers who were already connected at the time of reform. In contrast connection fees, which included the cost of expand- ing the secondary network, remained high: between US$1,107 and US$1,528. As the average monthly income in the poorest sections of the 64 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE city wvas some US$200 to US$245, many poor households were unable to afford the high cost of a connection. Consequently, despite the decreases in monthly subscription rates and the expansion targets based on geo- graphical location with poor areas prioritized, the reform initially failed to benefit poor households as much as originally intencded. This eventually led to a renegotiation of the contract, under which the cost of new connec- tions was passed onto both new and existing customers through an addi- tional surcharge: the universal service and environment improvement fee. In summary, the evidence from cross-country comparisons of reformer-s and nonreformers and fiom comparisons before andl after reform fails to support the hypothesis that reform harms low-income consumers. In many cases the poor seem to benefit, at least in terms of being connected to the network. Note, however, that the impact of reform will vary from country to country and city to city. In countries and cities where coverage among poor households is alreadiy high or where many poor consumers have informal or illegal connec- tions, significant price increases and regulanzation of customer accounts might lead to a reduction in coverage among low-income households even if total coverage increases.21 In contrast, in countries where service was heavily rationed pnor to reform, privatization and liberalization might result in increased coverage for low-income households, even if prices increase. 2.7 Conclusions Most countries have an explicit policy goal of promoting universal access to certain infiastructure utilities. When service was providledl by monopo- lies (typically state-owned, but occasionally private), these obligations were, in theory, funded through cross-subsidies: hligh-income andc low-cost consumers were charged prices above cost to finance service to low-income and high-cost consumers, who paid prices below cost. While this arrange- ment soundls simple, in practice it has not workedl well. Cioss-subsidies have often been poorly targeted and have typically failed to reach poor con- sumers. Although low prices might increase clemand for infrastructure services by poor and rural consumers, they also lead to supply-side distor- tions that might lessen or nullify their impact. Moreover, the opaque nature of cross-subsidies also makes it difficult to determine who pays and who benefits from them. In practice, strong evidence indicates that public and 65 GEORGE R G CLARKE AND SCOTT J WALLSTEN private monopolies failed to ensure access for rural and low-income urban consumers, especially in Africa. Indeed, the relatively wealthy appeared to benefit from subsidies far more than the poor. Despite these failings, many observers worry that reforms such as privatiza- tion and competition in infrastructure utilities could harm the poor, making cross-subsidies unsustainable and raising prices beyond the reach of the poor. The limited evidence that is available suggests that overall this has not been the case. Most likely this is because, outside Eastern Europe, state-owned monopolies did a poor job of directing cross-subsidies to poor or rural con- sumers. Even when official tariffs rise under rate rebalancing, the real cost of connecting may fall, allowing more poor people to connect to the networks. Moreover, entry and competition allow entrepreneurs to discover and thy new methods of providing service to poor and rural areas, generating a wealth of service, price, and quality options. Maintaining state-owned or -regulated private monopolies might stifle innovative solutions to provid- ing access to the poor. Indeed, if competitive entry and privatization increase efficiency, areas and customers that monopolists found unprof- itable might become profitable, or at least might require smaller subsidies. Some regions and users thought to be unwilling or unable to pay for serv- ice have turned out to be profitable customers, as evidenced by creative entry mechanisms from new competitors.22 Nonetheless, not everyone is willing or able to pay the cost of utility serv- ices, meaning that some regions and users will require subsidies if society wants them to be connected. Reforms mean that new methods are necessary to raise subsidies, including competitively neutral financing mechanisms such as universal service funds and subsidy auctions. While reforms pre- sent a challenge to ensuring access to the poor, in light of the almost com- plete failure of service provision to the poor under monopoly provision in many developing countries, reforms also provide an opportunity to com- pletely rethink the role of subsidies and of how to ensure access by the poor. NoVtes 1. In practice, the evidence for this is somewhat mixed. For example, Esrey (1996) found that the health benefits (in tenrs of diarrhea, child health, and child weight) froni improved sewerage are greater than the health benefits of improved watcr in Africa, Asia, and Latin Amenca. Furthermore, although he did find some benefits 66 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE associated with on-site (that Is, in-house or in-comilpounid) piped water, he foulid no benefit from access to interriiediate facilities (public taps, hand punips, wells) Jalan arid Ravillion (2001) found that piped vater reduces the pievalenice arl(l duration of diarrhea for childireni in India. Howvever, they also found thdt the benefits largely bypasscd pooi families and families witlh poorly cducated women. They interpretedl their results to suggest that public actioni to promote health knowledge aricl recluce povety should be combinecl with subsidies for infrastructurc access. 2. Note that to the cxtent that healthi benefits arc interinalizecl, that is, thc costs of pool health ate piimarily borne by the individual or householdl that beconies sick, and that individtials are infonired about the potential hcalthi costs of conslUImI- ing nintieated watcr, the healthi exteiiialities associated with access to piped water ate piobably sirraller thiani the total health benefits 3. Estimilates fi-om the late 1980s suggest that the long-run maiginal cost of a cubic meter of water was about four times higher In the rest of the Cote d'lvoire thani in Abidljan (World Barik 1990). 4. Of course, cross-subsidies are possible between high- and low-cost areas even with regional companies, andCI some nationial companiies charge differenit prices In dlifferent regions. I-lowever, in geneal cross-subsidies arc hiar(ler to obscive andl easier to provide when a single companiy piovides ser-vice to the entite country. 5 The actual process of assignitig subsidies, paid directly to the water comipa- nies withinl each iegion that then subtract the amounIt flom houselhold bills, is corn- plicated, involving allocating subsidies fiist between regions based on hotuselholdl inconiie in the ICgion andl their witiin regions, between municipalities based on a points system (see Shirley. Xu, andl Zuluaga 2000 for a corirplete (lescription). 6 For example, In the 1990s In Buenos Aires, Argentiia, unmrietered residlential households, which accounted for most households, vere billed for water and sew- erage based upon the location of the property, the area of the property, the area of property withi coiistrLctioni on it, the type of coisltiuction (six categories), and(l tlC age of the property (Alcazar, Abdala, and Shirley 2002). 7 Foster, G6niez-Lobo, ancl Halpein, (2000) noted that clianiginig the cilgibility ciiteria Would i ncrease the number of hotiseholdis in extrenie poverty that receive subsidies, but vould also increase the hiouseholds not rn poverty receiving subsr- dies Similarly, Estache, Foster, arid Wodoni (2002) founid that most beneficiaries of a similar plan in Colonibia were middle-class, not poor. 8 Boland and Whittington (2000) suggest that one of tire main reasons for the popularity of block tariffs is that muIltilateral donors, interrational finanicial and engineering consultants, and water sector professionials encourage their uise 9. In line with this thinkiig, a recent report on liberalization In the telecomriiu- riicationis sector suggests- "[Als treil(ls towardl pilvalizationl and liberalization of basic telecornmunicationis services accelerate vorldwile, concerns about univer- 67 GEORGE R G CLARKE AND SCOTT J WALLSTEN sal service, particularly rural service, are increasingly raise(l among policymakers, user groups and industry participants. This concern stems from tile possibility that, when the state relinquishes ownership and management of telecommunications networks, and when competing private operators seek to gain maximum profit, service requireimients in costly areas will be overlooked" (Pyramid Research 1997) 10. Gomez-Lobo, Foster, and l-alpern (2000b) notc that this remains true even when a water company has a special tariff for vulnerable households, because the critena for qualification rarely coniespond closely to objective definitions of low- income or disadvantaged households. 11. Because it often takes several years for surveys collected for other reasons to be made available to researchers other than those who perfonned the survey, ana- lyzing recent reforms using publicly available data canl be difficult. In addition, there are relatively few easily accessible, publicly available household surveys with detailed data on infrastructure use for developing countries for recent years. As most reforms are quite recent, this makes getting postrefonn data difficult. 12. Although the education of the head of household is an imperfect proxy for household inconie oi consumption, it tends to be highly correlated with the vari- able of interest. For example, in a simple regression of household expeiiditures on five education dummies representing the education level of the head of household for a sample of households from Abicljan, Cote d'lvoire, each dummy variable is statistically significanitly' different from tlhe next level. In addition, the difference in expenditure levels appears laige: average annual household expenditures for households witlh heads with no education was CFAF 1.3 million, compared with CFAF 2.4 million foi households with heads with a secondary education or higher. Data on Abijan are foi 1996 from the Enquete suir les Depenses des Menages d'Abijan, carried out by the Institut National de la Statistique, Abidjan, Cote d'lvoire, and were provided by the African householdl database at the Wlorld Bank 13. In 3 of the 21 African countries, households headed by individuals witlh no education were about as likely to have flush toilets dS households headed by indi- viduals with a secondary education. 14. One reason for this is that under the block tariff scheme used in Ghana, poor households tended to share single connections, meaning that they ended up pay- ing the highest rate, while wealtlhier households had their own connections, put- ting theni in lower consumption brackets. 15. In every sector and for almost all the countries for which DHS data were available, infrastructure coniiections were more common in the capital than they were in urbani areas outside the capital. 16 The mean difference in coverage between the capital city and other urban areas is statistically insignificant for the countries shown in table 3.3. This remains true when Nigeria, where state governments are responsible for water supply, is excluded from the group of countries where local governments are responsible for 68 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE water supply Nigeria has state-level provision, compared witti local muni.cipal- level provision itn other countries. As this is soniewliere between natiolial-level and municipal-level provision, we liave singled Nigeria out. This is especially impior- tanit given the laige size of some Nigerian states (36 states in total for a country of 126 mnillioni people), some of which are the size of small couLitries (up to some 6 million to 7 million people In 1995). 17. Clarke arid Xu (2002) foulid that firms ttiat are miore plofitable pay hIigher bribes to infrastructure enterprises thiani less profitable enterpiises. They noted that this is consistent withi the "speed money" hypothesis, which suggests that bribes operatc as a price mechianiism ensuirinig that those most willing to pay gain access to infrastructure services A laige literatutre on hedonic pricing is available that suggests ttiat rents are higher for houses with iiifrastrlicture conniections (see, for example, Northi and Cnffini 1993) 18. Tbe laige aniounit of researchi on teleconimimunications partly reflects the bet- ter avdilability of Closs-couLIntry data for developing countlies. For example, sev- eial studies have used the data collected by the Interinationial TelecomimiLiunications Union 19. Li arid Xu (2001) found that shiare issue piivatizationi appears to have a pos- itive impact on covelage, but founl nio evidence foi otber types of privatizatron 20. F'or examlple, nearl) one-thlirdl of water coninectionis In Conakry, Guinea, were inactive because of nonpaynient five years after the large price increases that fol- lowedl the intioductioni of private sectoi participation in 1989 (Brook Cowen 1996) 21. An additional point is that if low prices aiid high numiubers of coninectiolis meant that the utility needled to be subsidized from general tax revenues, the over- all impact on low-inicomiie households wvill also depeend on lhow refol Ill affected tlese subsidies. 22. For examiple, condlomiiial sewerage systems reduce Costs over traditionial systenis by using smaller pipes, being installed In shallow trenches, and being installed under household yards rather thani undler roads (Komlives andl Blook Cowen 1998) References AlcAzar; Lorena, Manuel Abdala, and Mary M. Shirley. 2002. "The Btienos Aires Water Concession." In Mary M. Shirley, ed., Thlrstnntgfor Efficiency: The Economics and Polttics oJ Urban WVater System Reform. Oxford, U.K.: Pergamon Press. Alcazar, Lorena, L. Colin Xu, and Ana Maria Zuluaga. 2002. "Institu- tions, Politics, and Contracts: The Privatization Attempt of the Water 69 GEORGE R G CLARKE AND SCOTT J WALLSTEN and Sanitation Utility of Lima, Peru." In Mary M. Shirley, ed., Thirsting for Efficiency: The Economics and Politics of Urban Water System Reform. Oxford, U.K.: Pergamon Press. Boland, John J., and Dale Whittington. 2000. "The Political Economy of Water Tariff Design: Increasing Block Tariffs Versus Uniform Price with Rebate." In Ariel Dinar, ed., Political Economy of Water Pricing Reforrms. Oxford, U.K.: Oxford University Press. Brook Cowen, Penelope J. 1996. "The Guinea Water Lease-Five Years On." Public Policy for the Private Sector no. 78. World Bank, Washington, D.C. Brook, Penelope, and Warrick Smith. 2000. "Improving Access to Infra- structure Services by Low-Income Households: Institutional and Policy Responses." World Bank, Washington, D.C. Cannock, Geoffrey. 2001. "Telecom Subsidies: Output-Based Contracts for Rural Services in Peru." World Bank, Washington, D.C. Chisari, Omar, and Antonio Estache. 1999. "Universal Service Obliga- tions in Utility Concession Contracts and the Needs of the Poor in Argentina's Privatizations." Policy Research Working Paper no. 2250. World Bank, Washington, D.C. Chisari, Omar, Antonio Estache, and Jean-Jacques Laffont. 1999. "The Needs of the Poor in Infrastructure Obligation: The Role of Universal Service Obligations." World Bank, Washington, D.C. Clarke, George R. G., and L. Colin Xu. 2002. "Ownership, Competition, and Corruption: Bribe Takers Versus Bribe Payers." Policy Research Working Paper no. 2783. World Bank, Washington, D.C. Clarke, George R. G., Claude Menard, and Ana Maria Zuluaga. 2002. "Measuring the Welfare Effects of Reform: Urban Water Supply in Guinea." World Development 30(9): 1515-35. Crandall, Robert, and Leonard Waverman. 2000. Who Pays for Universal Service? When Telephone Subsidies Become Transparent. Washington, D.C.: The Brookings Institution. 70 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE Cremer, Helmuth, Farid Gasmi, Andre Crimaud, and Jean-Jacques Laf- font. 1998a. "The Economics of Universal Service: Practice." Discus- sion Paper. World Bank Institute, Washington, D.C. . 1998b. "The Economics of Universal Service: Theory." Discussion Paper. World Bank Institute, Washington, D.C. Eriksson, Ross, David Kaserman, and John Mayo. 1998. "Targeted and Untargeted Subsidy Schemes: Evidence from Postdivestiture Efforts to Promote Universal Telephone Service." Journal of Law and Economics 41(2): 477-502. Esrey, Steven A. 1996. "Water, Waste and Well-Being: A Multicountry Study." American Journzal of Epiderniology 43(6): 608-23. Estache, Antonio, Vivien Foster, and Quentin Wodon. 2001. "Making Infrastricture Reform Work for the Poor: Policy Options Based on the Latin American Experience." LAC Regional Studies Program. World Bank, Washington, D.C. . 2002. Accountingfor Poverty in Infrastructure Reform: Learning from Latin Amnericas Experience. Washington, D.C.: World Bank. Foster, Vivien, Andres G6mez-Lobo, and Jonathan Halpern. 2000. "Designing Direct Subsidies for Water and Sanitation Services in Panama: A Case Study." Policy Research Working Paper no. 2344. World Bank, Washington, D.C. Galal, Ahmecl, and Bharat Nauriyal. 1995. "Regulating Teleconimunica- tions in Developing Countries." Policy Research Working Paper no. 1520. World Bank, Washington, D.C. G6mez-Lobo, Andr6s. 2001. "Incentive-Based Subsidies." Public Policy for the Private Sector no. 232. World Bank, Washington, D.C. G6mez-Lobo, Andr6s, Vivien Foster, and Jonatlhani Halpern. 2000a. "Better Household Surveys for Better Design of Infrastructure Subsidies." Public Policy for the Private Sector no. 213. World Bank, Washington, D.C. 71 GEORGE R G. CLARKE AND SCOTT J WALLSTEN . 2000b. "Information and Modeling Issues in Designing Water and Sanitation Subsidy Schemes." Policy Research Working Paper no. 2345. World Bank, Washington, D.C. Haggarty, Luke, Penelope Brook, and Ana Maria Zuluaga. 2001. "Water Service Contracts in Mexico City, Mexico." In Mary M. Shirley, ed., Thirstingfor Efficiency: The Economics and Politics of Urban Water Sys- tem Reform. Oxford, U.K.: Pergamon Press. International Telecommunications Union. 1998. World Telecommunica- tions Development Report: Untversal Access. Geneva. Intven, Hank, and McCarthy Tetrault. 2000. Telecommunications Regula- tion Handbook. Washington, D.C.: World Bank. Jalan, Jyotsna, and Martin Ravillion. 2001. "Does Piped Water Reduce Diarrhea in Rural India." Policy Research Working Paper no. 2664. World Bank, Washington, D.C. Kerf, Michel. 2000. "Do State Holding Companies Facilitate Private Par- ticipation in the Water Sector? Evidence from C8te d'lvoire, The Gam- bia, Guinea, and Senegal." Policy Research Working Paper no. 2513. World Bank, Washington, D.C. Komives, Kristin, and Penelope Brook Cowen. 1998. "Expanding Water and Sanitation Services to Low-Income Households." Public Policy for the Private Sector no. 178. World Bank, 'Washington, D.C. Li, Wei, and L. Colin Xu. 2001. "Liberalization and Performance in Telecom- munications Sector around the World." World Bank, Washington, D.C. Menard, Claude, and George R. G. Clarke. 2002a. "Reforming Water Supply in Abidjan, C8te d'Ivoire: A Mild Reform in a Turbulent Environment." In Mary M Shirley, ed., Thirstingfor Efficiency: Tlhe Economics and Politics of Urban Water System Reform. Oxford, U.K.: Pergamon Press. 72 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE . 2002b. "A Transitory Regime: Water Supply in Conakry, Guinea." In Mary M. Shirley, edl., Thirsting for Efficienrcy: The Econornics and Politics of Urban Water System Reform. Oxford, U.K.: Pergamon Press. Mueller, Milton Lawrence. 1997. Universal Service: Competition, Intercon- nection, and Monopoly in the Making of thte American Telephone System. Cambridge, Massachusetts: MIT Press. North, J. H., and C. C. Griffin. 1993. "Water Source as a Housing Charac- teristic: Hedonic Property Valuation and Willingness to Pay for Water." WVater Resources Research 29(7): 1923-29. Onwumechili, Chuka. 2001. "Dream or Reality: Providing Universal Access to Basic Telecommunications in Nigeria?" Telecoinmunications Policy 25(4): 219-31. Petrazzini, Ben A. 1996. "Competition in Telecoms-Implications for Universal Service and Employment." Public Policy for the Private Sec- tor no. 96. World Bank, Washington, D.C. Pyramid Research. 1997 Prnvatising Telecoms Markets. Economist Intelli- gence Unit: London. Ros, Agustin J. 1999. "Does Ownership or Competition Matter? The Effects of Telecommunications Reform on Network Expansion and Effi- ciency." Journal of Regulatory Economics 15(1): 65-92. Rosston, Gregory, and Bradley Wimmer. 2000. "The 'State' of Universal Service." Injormnation Econornics and Policy 12(3): 261-83. Savedoff, William D., and Pablo T. Spiller, eds. 1999. Spilled Water: Insti- tutional Comnmitmnent In the Provision of Water Services. WVashington, D.C.: Inter-American Development Bank Shirley, Mary M., ed. 2002. Thirstingfor Efficiency: The Econotnics and Politics of Urban Water System Reform. Oxford, U.K.: Pergamon Press. 73 GEORGE R G CLARKE AND SCOTT J WALLSTEN Shirley, Mary M., and Claude Menard. 2002. "Cities Awash: A Synthesis of the Country Cases." In Mary M. Shirley, ed., Thirstingfor Efficiency: The Economics and Politics of Urban Water System Reform. Oxford, U.K.: Pergamon Press. Shirley, Mary M., L. Cohn Xu, and Ana Maria Zuluaga. 2002. "Reforming Urban Water Supply: The Case of Chile." In Mary M. Shirley, ed., Tlhirsttngfor Efficiency: The Economics and Politics of Urban Water Sys- tem Reform. Oxford, U.K.: Pergamon Press. Waddams-Price, Catherine. 2000. "Subsidies and the Reform of Infra- structure Services." University of Wfarwick, Warvick, U.K. Walker, Ian, Fidel Ordonez, Pedro Serrano, and Jonathan Halpern. 2000. "Pricing, Subsidies, and the Poor." Policy Research Working Paper no. 2468. World Bank, Washington, D.C. Wallsten, Scott. 2001a. "An Econometric Analysis of Telecom Competi- tion, Privatization, and Regulation in Africa and Latin America." Jour- nal of Industrial Economics 49(1): 1-20. . 2001b. "Ringing in the 20th Century: The Effects of State Monop- olies, Private Ownership, and Operating Licenses on Telecommunica- tions in Europe, 1892-1914." Research Working Paper. World Bank, Washington, D.C. Water and Sanitation Program. 2001. "The Buenos Aires Concession." Water and Sanitation Program-South Asia: New Delhi, India Water Utilities Partnership. 2000. "Draft Report on Status of Reforms of the Water and Sanitation Sector in Africa." Abidjan, Cote d'Ivoire. Wellenius, Bjorn. 1997. "Telecommunications Reform-How to Succeed." Public Policy for the Private Sector no. 130. World Bank, Washington, D.C. . 2000. "Extending Telecommunications beyond the Market: Toward Universal Service in Competitive Environments." Public Policy for the Private Sector no. 206. World Bank, Washington, D.C. 74 UNIVERSAL SERVICE EMPIRICAL EVIDENCE ON THE PROVISION OF INFRASTRUCTURE Whittington, Dale. 1992. "Possible Adverse Effects of Incieasing Block Water Tariffs in Developing Countries." Economic Development and Cultural Change 41(1): 75-87. Wolak, Frank. 1996. "Can Universal Service Survive in a Competitive Telecommunications Environment? Evidence from the United States Consumer Expenditure Survey." Informatiotn EconornEcs and Policy 8(3): 163-203. World Bank. 1990. "Report and Recommendation of the President of the IBRD: Water Supply and Sanitation Sector Adjustment Program." World Bank, Washington, D.C. . 1994. World Development Report 1994: Infrastructure for Develop- ment. New York: Oxford University Press. . 2002a. World Development Indicators. Wasihington, D.C.: World Bank. . 2002b. World Developmeint Report 2002: Building Institutions for Markets. New York: Oxforcl University Press. 75 3 Infrastructure Coverage and the Poor: A Global Perspective Kristin Komives, Dale Whittington, and Xun Wu 77 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU 3.1 lntroduction This paper presents a global perspective on infrastructure coverage and the poor that many people will think they have seen before, but have not.' It is widely assumed that the poor in developing countries have fewer infrastruc- ture services than middle- and upper-income households, but surprisingly little information is available on the actual empirical relationship between household income and infrastructure service coverage in different countries. The available coverage statistics are typically countrywide averages. These are widely used to assess the scope and magnitude of infrastructure prob- lems in developing countries, and are often the only global, cross-country data available about infrastructure services. When such coverage statistics reveal that many households do not have service, that is, they are not cov- ered, the general assumption is that such houselholds are poor. Global cov- erage statistics are often compiled by international organizations such as the World Health Organization and the World Bank. and have profoundly shaped the way many people conceptualize infrastructure policy problems.2 Numerous problems are associated with the countrywide infrastructure coverage statistics currently available, despite their widespread use and influence. The data on household coverage typically come from general purpose household surveys, such as censuses, that include a few questions designed to determine whether a household has various infrastructure serv- ices. For example, a member of a household may be asked whether the house has an in-house piped water connection or electricity. The global sta- tistics from such surveys are usually self-reported by countries and are of varying quality. In many cases the wording of questions in the different sur- veys is not the same, and the surveys may have been carried out in differ- ent years and with different sampling procedures. Such general purpose surveys typically ignore informal service options, such as water vending or the provision of electricity from a private genera- tor. In addition, different surveys may use different definitions of some infrastructure service options. Countries generally report summary statis- tics that cannot be related to the income of individual households, so determining how coverage of the poor differs from coverage of other income groups is impossible. Moreover, the international agencies that compile A shoiter version of this paper, with an emphasis onl infrastructure coverage in Latin Ameri d, appears in Cecilia Ugaz and Catherine Wadd(lams Price, eds., Utlht) Pnratztuanti and Reguliaioiil: A Fair Dealfior Consumers?, Edward Elgar Publishing, 2003. 78 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE coverage statistics for one infrastructure service, for instance, water, rarely coordinate their efforts with other agencies (or even with other divisions in the same organization) interestedI in dlifferent infrastructure services, so seeing comparal)le coverage statistics for multiple infrastructure services is unusual. This chapter introdluces a new data source for global coverage statistics, the World Bank's Living Stanclards Measurement Study (LSMS) surveys, which addresses some, but not all, of these limitations. These surveys on multiple topics gather extensive socioeconomic and expenditure informa- tion from households, as well as limited information on households' use of selected infiastructure services. The data used here are drawn from LSMS surveys conducted in 15 countries. The pooled sample includes more than 55,500 households in Asia, the Americas, Eastern Europe and Central Asia, and Sub-Saharan Africa. The LSMS surveys enable us to examine coverage for several infrastructure services among various income groups in many different countries using household-level data. The results of our analyses show that all income groups throughout the world have much higher levels of coverage for electricity than other formal infrastructure services (in-house piped water service, sewer service, and private telephone service). In many countries most households in urban areas now have electricity service. The relationship between income and coverage is remarkably similar for electricity, in-house water connections, andl sewer service. As monthly household incomes increase from US$100 to US$250, coverage of all these infrastructure services rises rapidly. As expected, coverage is much higher in urban thani in rural areas for elec- tricity, water, sewer service, an(d teleplhone service. The findings confirm that, with some exceptions, the very poor rarely have these infiastructul-e services, althoughi they often do have electricity if they live in urban areas. The very poor in Eastern Europe andl Central Asia have much higher levels of coverage than elsewvhere in the worldl and often have electricity, water, sewer service, and telephonie service. The results also suggest that if the poor have access to services in their com- munities, many will decide to connect.4 Where the very poor do not have form-lal infrastructure services, informal, private, and community infrastructure solutions fill the gap for many hiouseholdls. Few households in any of the 15 countries in our sample report using unimproved water sources or candles for lighting; however, many householdIs at all income levels and in both rural and urban areas use 79 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU wood, thatch, or dung for cooking fuel. Few poor households without pri- vate telephones have public telephones in their communities, and the vast majority of the poorest rural households have no toilet, sewer, or septic facilities in their homes. 3.2 The ]Data: Living Standards M/easuremenr Study Surveys in 15 Countries The World Bank initiated the LSMS program in the 1980s to improve the qual- ity of survey data available for policy research and analysis in developing countries. Since then more than 20 countries have administered nationally representative household surveys based on thle LSMS model of questionnaire design and quality control. The multicountry dataset used in this analysis con- sists of surveys from 15 of these countries (table 3.1).5 The pooled sample includes households on four continents in both low- and middle-income countries. The 15 surveys were administered between 1988 and 1997. This multicountry LSMS dataset is unique in five important respects. First, it enables us to look at multiple infrastructure services for the same household. Second, because the LSMS surveys are primarily designed to measure households' economic well-being, that is, their standard of living, the dataset arguably contains the best information available on household expenditures, consumption, and income for multiple developing countries. This enables us to clearly identify the poorest households in our sample and their use of infrastructure services. Third, the LSMS surveys generally use similar survey administration protocols, quality control procedures, and survey questions across countries. Fourth, the ISMS surveys have been implemented in many developing countries, which enables us to con- struct a global perspective on infrastructure coverage and the poor that is not possible with a survey in a single country. Note, however, that the coun- tries in our sample are not in any sense a random sample of the developing world.6 Fifth, some LSMS household surveys were accompanied by commu- nity surveys that gathered information about the availability of infrastruc- ture and other services in the areas where sample households live. The com- munity surveys enable us to distinguish between (a) households that do not have infrastructure services and could not have such services because they do not have access to them in their neighborhoods, and (b) households that 80 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE TABLE 3.1. LSMS Datasets Used in This Study 1998 Community-level survey Region and per capita Survey Number of available and used in the country GNP (USS) year households analyses for this chapter Asia Pakistan 480 1991 4,800 No Vietnam 330 1992-93 4,800 No Nepal 210 1996 3,373 Yes. Eastern Europe and Central Asia Russia 2,300 1994-95 3,973 No Kazakhstan 1,310 1996 1,996 Yesb Bulgaria 1,230 1995 2,468 No Albaniac 810 1997 1,503 No Kyrgyz Republic 350 1993 1,937 Yes Latin America and the Caribbean Panama 3,080 1997 4,938 Yes Jamaica 1,680 1997 2,016 No Ecuador 1,530 1995 5,661 Yes Nicaragua 390 1993 4,454 Sub-Saharan Afrlca South Africa 2,880 1993 8,850 No COte d'lvoire 700 1988 1,584 No Ghana 390 1988-89 3,193 No a Information on community access to private telephone service is not available b Information on sewer access is not available c Does not include households in Tirana Source World Bank data do not have infrastructure services, but do have access and could have cho- sen to have such services if they had the resources and desire to do so. The 15 LSMS surveys in the multicountry dataset include roughly simi- lar questions, but the answer categories, exact question wording, and cur- rency units are often different from country to country. For this analysis we have creaLed new income, expenditure, and infrastructure variables that 81 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU can be compared across countries. There are inevitably conceptual and measurement problems in the creation of such global variables. Our pur- pose here is to look for broad patterns of infrastructure use by households at different levels of economic well-being. We therefore caution readers not to make too much of individual results.7 WTe created cross-country income and expenditure variables by convert- ing local currencies to 1998 U.S. dollars using first the official currency exchange rate in the survey year, and then the U.S. consumer price index.8 Our cross-country infrastructure variables classify infrastructure options in each sector as "advanced," "intermediate," or "basic" solutions (see the appendix). Sections 3.3, 3.4, and 3.5 of this chapter examine advanced solu- tions, which are typically provided by a utility (electricity, in-house water taps, sewer connections, and telephones). Section 3.6 looks at intermediate and basic solutions, which tend to be more informal or private forms of infra- structure service; for example, in the energy sector kerosene would be an intermediate energy source and wood would be a basic energy source. We use monthly household consumption aggregates as income proxies in this analysis because we consider the consumption data to be more accu- rate and reliable than the self-reported income data. For the purposes of this analysis, the poorest households are those with the lowest per capita income proxy.9 The pooled sample of households from all countries is divided into 20 quantiles of 5 percent each. We divide households in the urban and rural areas of each country into income deciles by per capita consumption. We present the results by decile or quantiles of 5 percent as appropriate, with special emphasis on infrastructure coverage among households in the poorest deciles and quantiles. 3.3 Who Has infrastructure Services? Global Infrastructure Coverage More than 65 percent of households in the pooled cross-national sample had electricity in their homes at the time of the LSMS survey. "1 By contrast, only 38 percent of households had in-house water taps, 36 percent had sewer connections, and 24 percent had telephones." The distribution of these utility connections among households is highly correlated with our income proxy, that is, monthly aggregate household consumption: a higher 82 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE percentage of wealthy than poor households have electricity, in-lhouse taps, sewer connections, andl telephones in thieir homes. Figure 3.1 shows how coverage of these services varies by income level in the cross-country pooled sample. Each dot on the graph represents one quantile of 5 percent of lhouseholdis. The dots plot the quantile's median income against the coverage of electricity, in-house water taps, sewer con- nectioris, and telephones within that group. Aggregate consumption among households in the poorest 5 percent quan- tile of the pooled sample was less than US$1 per household per (lay (US$27 per month, on average). These householdls came from all countries in the sample, but the majority live in the Kyrgyz Republic, Nepal, and Vietnam, the poorest of the 15 countries (most of the richest households are in Ecuadlor, Jamaica, Panama, Russia, and South Africa, the wealthiest coun- tries in the sample). Electricity was the only service with significant pene- tration among the poorest 5 percent quantile of households: nearly 32 per- cent had electricity in their homes. Few had in-house water taps (6 percent), sewer connections (3 percent), or telephones (3 percent). Telephone coverage remains at 3 or 4 percent among householdis in the first five 5 percent quantiles, that is, the poorest 25 percent of the sample households. T'elephone coverage begins to rise only when the median income proxy reaches US$120 per household per month. By contrast, cov- erage of electricity and in-house water taps begins to rise immediately and increases sharply from 5 percent (qLuantile to 5 percent quantile. By tile 10th 5 percent quantile, that is, at the median income proxy, whiich equals US$225 per household per month, 66 percent of the sample households had electricity and 33 percent hadl in-house wvater taps. Above US$225 per houselhold per month the use of electricity and in-house taps continues to rise, but at a slower rate. Nearly all the households in the wealthiest 5 per- cent quantile (US$1,300 per household per month) had electricity, 88 per- cent hiad in-house water taps, and 72 percent had telephones. Electricity was the most widespread of tilese three services at all income levels and telephone service was the least common. The coverage lines for these three sectors never cross in figure 3. 1, and thie slope of the three lines is remarkably similar among households with incomes (as approximated by the consumption aggregate) above US$250 per month. Figure 3.1 does show one puzzling result. One woulcl generally expect more households, and particularly more poor householdls, to have modern 83 0w 0 G 0. ~ ~~~~~L0. V 0 - r, e1 \a C_ U I \ o o E, \ \ o 0~~~~~~~~~~~~~~~~~~~ CL 0 0 0 0 ' -5 75 0' 000 0' C' ° S S - CY c° o 8o S 0 R ° C.0 0 0 0D C S 0 0 O 2 0~~~~~~~~~~~~~~~~~~~ 0Q a cE E m ,~~~~~~~~~~~~~~~~~~~~~~~~~L C E 84 0 = 0 Oo E liii~~~~~~~~~~~~~~~~~~~~ o~~~~~~~~~~~~~~~~~~~~~~~~az- > Cii -> 0 C a, a, -0 0 a .0 o 0 E 0~. 0 0 -..~~~~~~~~~~~~~~0 0 LU 0E C C a, *5~~~~~~~~~~~~~~~~~~~~~~~~~C ~~~~ o a~~~~~~~~~M - *.' ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~r 7 84~.. INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE water services than advanced sanitation solutions, but tIhe coverage of in- house water taps and sewer connections appears to be virtually identical up to US$300 per household per month. Two shortcomlnlgs in the LSMS data used for this analysis cause this result. First, in-house water taps are Just one form of pEivate household water connectLion. In many types of dwellings in-house taps might not be feasible or desirable to install, or households may not initially want to invest in indloor plumbing facilities. In these cases households could choose to install a yard tap rather than an in-house tap. The in-house tap variable reported in figure 3.1 therefore undlerstates the number of households with private water connections. Identifying households with yard taps is possible in only 7 of the 15 coun- tries in this sample. In those seven countries, almost none of the poorest househIolds had sewer connections or in-house connections and yard taps, but at higher income levels in-house connections and yard taps were much more prevalent than sewer connections (figure 3.2). Second, information on sewer connectLiois is available only for 12 of the 15 countries. When households in only those 10 countries are pooled and (livided into quantiles of 5 percent, as expected, sewer coverage clearly lags behind coverage of in-house water taps (figure 3.3). Sewer coverage is consistently about 10 percent lower than in-house water tap coverage for households with incomes (as approximated by the consumption aggregate) of less than US$400 per month. Above US$400 per month, the gap between in-house water service and sewer connections actually widens. As in fig- ure 3.1, electricity coverage is higher than the coverage of otlier infrastruc- ture services at all income levels. In the remainder of this chapter we present results for the pooled sample of households from all 15 countries (as in figure 3.1) and, except wvhete noted, we use coverage figures for in-house water taps only rather than both in-house and yard taps.'2 Coverage in Urban and Rural Areas As anticipated, a smaller percentage of rural than urban residents had infra- structure services in their homes."' Fewer rural households had electricity (46 percent versus 89 percent in cities), in-house water taps (12 percent ver- sus 59 percent), sewer connections (7 percent versus 61 percent), and tele- phones (8 percent versus 38 percent). The poor live disproportionately in 85 02 -~~~~~~~~~~~~~~~~~~ 02~~~~~~~~~~~~~~~~~~~~ C 0.~~~~~~~~~~~~~~~C -~~~~~~~~ ~~o 0 4E 0A - . 4 o~~~~~~~~~~~~~~~~~~~~~~~~C' - S * ~~~~~~~~~~~~~~~~~~~ cu 0~~~~~0 02 u :2~~~~~~~~~ .0 o 'A ~~~~~~~~~~~E ' 4, 0 E 02~~~~~~~~~~~~~~~~~~~~ o = ~~~~~~~~~~~~~~~0.~~~~~ c ~~~~~~~~~~~~~00 -i 0 0 4 cc -Cm~~~~~~~~~~~~~ -c '0 E o~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~t 86 t;~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~: 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. C 0 a) ~ ~ ~ ~ ~ ~ ~ 02 02 2 02 : W Li~~~~~~~~~~ o ~~~~~~~~~~~~~~~~~~~~~~~~~~~o' o 0 0 C,~~~ 0 0 0 ( U) .02 .02 .02 .02 02~~~~~~~ fo 2~~~~o a 2 020 w -~~~~~~. . ~C: a 0 C & 0.0.0. E Li 02~~~~~~~~~~~~~~~~~~~~~~~~~C ~~III ~~~ 02~~~ l i i i 02 .0202~~~~~~~a (Yi~~~~~~~~~~~~~~~~~0I 0'1110 .0 03~~~~~~~~~~~~~c 8702 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU rural areas, but urban or rural location alone does not explain the infra- structure gap between urban and rural areas.14 Figures 3.4 and 3.5 show that a smaller percentage of the poor than the rich in both urban and rural areas had electricity, in-house taps, sewer connections, and telephones. Very few of the poorest rural households had in-house water taps (2 per- cent), sewers (1 percent), or telephones (2 percent). Rural coverage of these three services remains less than 10 percent up to income levels of US$200 per household per month. Perhaps surprisingly, electricity is reaching a substantial number of the rural poor: 27 percent in the poorest quantile. By contrast, a significant number of the poorest urban households had in-house water taps (31 percent), sewers (28 percent), and telephones (14 percent). Coverage of these services rises steeply from one 5 percent quan- tile to the next 5 percent quantile. Electricity coverage in urban areas is surprisingly similar across income groups. Nearly 80 percent of the poor- est urban households had electricity, and coverage rises further among higher income groups. 3.4 'Who JRas Access to Services? W/ho Has Access and Chooses Not to Connect? One reason why many households do not have infrastructure connections in their homes is that they live in places where they do not have the option of connecting to a utility network, that is, no network service exists in their neighborhoods.'" Information on community access to infrastructure net- works is available for most households in the urban and rural areas of five countries in our sample (see table 3.1). Where this information is available, we can begin to isolate the role that household choices play in creating the observed coverage patterns, that is, which households have access but choose not to connect.16 In these five countries community access to infrastructure is high in urban areas and low in rural areas (figures 3.6 and 3.7). Households of all income levels in both urban and rural areas were most likely to have electricity serv- ice and least likely to have sewer service available in their communities. In urban areas infrastructure access was not highly dependent on household income, and the percentage of households with access to services was sim- ilar across income levels, but in rural areas the wealthy were much more 88 -0 0) ~~~~~~~~~~~~~~~~~~~~~~~~~~-~~~~~a C~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~- 0f -~ e ~4 -5 4 4 . 44 ' 0~ 4- -c c~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~L -~~~~~~~~~~ c o o o~~~~~~~~~~~~~~~~~ 4 4 4 4 ~~~~~~ CC , 41 4. 4.4. ~- m44o ~~~~~~~0 42~~~~~~~~~~~~~~~~~~~~0 - E 111 D 0) 0 4 c 2 -C 0) -~~~~~~~~~~~~~~~~~~, 0) -0 0 5! cc. E o ~ ~ ~ ~~~~~~~~Eo0~~~~~~~~~~ 0) ci - 4444~~~~~~~~E 4C 40~ 0~~~~~~~~~~~~~~~~~~~~ C~~~~~~~~~~~~~~~~~~~~~~c ~ U-o0 4- 4 .4.4 .4~~~~~~~~~~~8 -~~~~~~~ 0-c~~~~~~~~~~~ o~~~~~~~~~~~~~~~~~z4 -~~~~~~~~ 4-~~~~~~~~C 4i 0) ~~~~~~~~~~~~~~~~~~~~50 4- ;4 ~ ~ ~ ~ ~~- --5 -5 ~ 4-- >4-4 -c a) 2C4-4 0JL 04' LA -, )L~~~~~~~~~~~~~~O UO 0 4- 0 4-Li~~~. O~~4- LEA 0 0 0 ~~~~~~' 4-P C 4- 4- ~~~~ U, ~ CC 0 4-4-4-~~~~~~~~~~~~~~~ - '- - - ~ >C o~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~2 >4 4- 2 2 ~~~~~~~~ ~ ~ 4-4-0>~a 0.0.0.0. 4-~; : C t 0 F) >4 liii - _ 0~~~~~> X AAAA ~~~~~~~~0 fTTT~~~~~~~~~~- 0 f0 0 2 _ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ - ~~~~~~~~~~~- o 0 - o 0 76 - c J-Z.~ 0 .4- ~~~~~~~~~~~~~~~~~ oU - maE- a O cl4-* tD~~~~~~~~~~~~~C40 UA0-V , L 90C 3 j L E C~~~~~~~~~~~~~~~~~~~C ~~~~~~~~~~~0~~~~~~~~~ o 2~~~~~~~~~~~~~~~~~~~ a,~ o ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 -E C a, ~~~~~~~~0~- O o a,~~~~~~~~~~~~~~~~ ~0' -o c a,- 1a,Ja, - w 7 a, a, a, a, ~~~~a,22gE CL aE a,a,a C. 000- QL a, a, a, a . E 'a -C cl 75 wC~~~~CC~Ca, ,- m E ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -a, -a, a, ~~~~~~~~~~~~~~~~~~~~ E ~~ C ~ ~ , a -0 a,C C O~~~~~~~~~C~~~~~ U, ~ ~ C C a, - - a,3 0-a,a a E 0 '' '0 W - ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ a,~~~~~~~~~~~~~~~r~ C 0~~~~~~~~, 0 O , 0W E WC a. ca o ~~~~~~~~~~~~~~~~~~~~~~~~~2 L- o-. CC~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0M a-- 0 - !q -e !3~~~~~~~~~CaC a, C0 O aF6 Oz - ~~~~~~~~-W a-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Win EC~~~~~~~~~~~~~Oa,a- 0~~~~~~z ~2 C a) -~~~~~~~~~! W~~~~, a, 0 8 0 a0 % a, WC O , j aa,a, - aO a) - = OO 0)C w~~~~~~~~~~~~a0faC a U - 0W 91 0 -c~~~~~~~~~~~~~ C~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C o 2 2 2 E 2H =0 o 6- 0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~4 C)C c 0 ° C =: C o 0 4, .) ' ' ' t0 v 0 0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~40 0> = C, CCC°V F- 0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 4- 0 c 0 VC C o o0 0 0 @~~~~~~~~~~~~ ° °-E > 0303~~~~ -C c R<>C CU ~ ~ ~ ~ ~~ ~~~ ~~~~~~~~~~~~~~~~~~~~- 0) C - e 0 E c lu~~~~~~~~~~~~~- 0 ° 0 u- E ., .w 04 .0 04~~~~~~~~~~~~~~~~~~~ a, C 0) t ~~~~~C ,.c~~~~~~~~~~~~~~~ X 0 . - 0r uo a o a 0~- 0404 0 LLJ -C~~~~~~~~~~~~~~~~~~~~~~- .0)~~~~~~~~~~~~~~~~~~~~ 95 *C 4 _ 2 o - E r v- __ _ _ 0)r 0~~~~~~~~~~~~~~~~~~~~~ 4-'~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~~~~~- o i _ O, , _o E ' -~~~~ ~ ~~~ ~ ~ ~ e 'J 2> -: 02 0~~~~~~~~~~~~~~~~~~~ ci 0~~~~~~~~~~~~~~~~~~~~~~~c ! -- - --E~ C r f= W~~~~~~~~~~~~~~ s 2 - E_ 96 C~~~~~~~~~~~~~~~~~6 ~~~~~~~~~~~0~~~~~~~~~~ ci ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ > ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~44 '0) * : 8 IE 4' - c02.E 0 0- - -4- - I-- - - -0 -- 0-44 CiE - -a~~~~~~~3 ~U4 - Et C, C a a 0 a 0~ - 0 o~~~~~~~~~~~~0 aCO 96 - _ E ~~~~~~~~~~~~~~~~~~~~~~~~~~~~C- U r -X -°B X °~~~~~~~~~~~~~~~~ I 4 3.3z. _ _ o~~~~~~~~~~~~~~ _ ti {t<3 =e,__- _~~~~~~~~~~~~ _: -s, ~~~~~~~~~~~~~~~~~E < -0 s_ m | 5, ,' > ,3tR ;. _ _ = D $-~~~~~~~~~~~~~~~~~~~~~~~ > a o _ Sos _~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~oU D I , , ;X.s , -, h W ; n ,, _ _ V~~~~ tt: ' o m ^ V o @~~~Z >s ° ec 2 s <,, o o~~~c C E ° @ c _ c c~~~~~~~~~~~~~~~ CJ - _ : ° z E u~~~~~~~~- - C _ =p , c :- s 3 7 w-a 2 E- DoC E C, sS E° 3 S R R ° ', e n I E! mSn C, E° 2~~~ ~~~~~~~~~~~~~~~~~~~~~~ 2) El TO 3 97c t ,~~ ~ ~~~~~~~~~~~~~~~ 0 : VS [ _ ~~~~~~~~~~~~~~~~~~~~~~~~o o C o -g: i E o C C O , v[ o s0 c .~~ B _ - - -- ~~~~~~ ~~ ~ \ v S _ v:c c < m C v,° ~~~~~~~~~~~~~~~~~~~~~~~~~, C E GT,U E F ~~~~~~~~~~~~~~~~~~E l 2 : >980 U 4 c - o O _~~~~~~~~~~~~~ E , _ _ U_ °t 0)~~~~~~~~~~~~~~~~~~~~ > _ ~~~~~~~~~~~~~~~~C 4C _ (0 5~~~~~~~~~~~~~~0 0 I 0g I 2 I -,4 I 4) ^ ( e EE -~~~~~~~~ 4,~~~~~~~~~~0 I- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 ,,E E CL o E _ -E fo 0 E4 C -0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ , o E E E ,E U Q -o 4,~~~~~~:E o C tt: u F E 5E w vs 2 _ E X v,~~~~~~~~~~~w E 0 0 : (0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 0) ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~) 0) 22~~~~~~~~~~c o && 042 a 0 0 -6~~~~~~~~~~~2 4,- 4 (0V 0c0 E II ,4 m -6~~~~~~~~~~ - 0 4 4,0(~~~~~~~~~~~~~~~~~~~ U. j,E~ 99 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU or sewer connections, but by the richest quintile, coverage levels in Nepal approximate coverage in quintiles with similar median income proxies in the other countries, for instance, the poorest quintile in Ecuador.24 3.5 A Miultivariate Analysis of Household Coverage The results presented in sections 3.3 and 3.4 demonstrate that infrastruc- ture coverage varies with household income, by country of residence, and between urban and rural areas. In this section we employ logistic regres- sion models to examine the relative importance of these variables after sta- tistically controlling for a number of other factors. We hypothesize that the likelihood that a household will have a connec- tion depends on the following seven variables: o The monthly household income proxy (INCOME PROXY) o Whether or not the household lives in a rural area (RURAL) o Whether or not the household lives in a low-income country with gross national product per capita less than US$760 (LOWINCY) o Whether or not the household is among the poorest 30 percent of the population in its own country (POOR) o Whether or not the household owns its home (HOMEOWNER) o The size of the household (HHSIZE) o Whether or not the household lives in an Eastern European or Central Asian country (EEUROPE). Table 3.2 presents the results of logistic regressions for five different binary dependent variables: electricity, in-house tap, house/yard tap, sewer con- nection, and telephone. All models are estimated with the pooled cross- country dataset. (A country is omitted from these models only if informa- tion on the dependent variable is not available for that country.) Among the seven independent variables, three measure how income affects the likelihood that households will be connected to these services: the household INCOME PROXY (which measures household wealth across countries), LOWINCY, and POOR. The results show that the household INCOME PROXY has a significant and positive influence in all five mod- els, and the magnitude of its effect is largest in the model for electricity and smallest for the sewer model. While the household INCOME PROXY measures the differences in wealth for households across countries, the second income-related variable. POOR, measures such differences within 100 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE TABLE 3.2. Logistic Regression Coefficients and Standard Errors from Multivariate Analysis of Infrastructure Coverage in Pooled Sample of Households from 15 LSMS Surveys Dependent variable (yes/no) Independent variable Electricity In-house tap House/yard tap Sewer Telephone INCOME PROXY 0 271- 0 226* 0129* 0075' 0 217* (USS hundreds) (0 008) (0 005) (0 007) (0 004) (O 004) RURAL -1 981 ^ -2 211 ^ -1 928^ -3 003^ -1 580^ = 1 if in rural area (0 027) (O 025) (O 034) (O 039) (0 032) = 0 if in urban area LOWINCY -0 068' -0189^ -1 853^ -0 735^ -1 059' = 1 if low-income country (0 029) (O 028) (0 038) (0 036) (0 041) = 0 if not POOR -0 573' -0 502^ -0 427^ -0 634^ -0 582' = 1 if household dectle ranking is 3 and below (0 033) (0 037) (0 042) (0 046) (0 049) = 0 if household decile ranking is 4 and above HOMEOWNER 0135' 0 282' 0140* -0 527' 0 660' = 1 if owner (O 029) (0 027) (0 037) (0 036) (0 036) = 0 if renter or other HHSIZE -0 038' -0 082' -0 021^ -0 038 -0 086' Size of the household (0 004) (0 004) (O 005) (0 006) (0 006) EEUROPE n a. 1 555' n a 1.477' 1 301' = 1 if in Eastern Europe or Central Asia (0 030) (0 037) (0 033) = 0 otherwise Pseudo R2 0 28 0 28 0 31 037 0 32 n a Not applicable 'Significant at the 95 percent confidence level Note Eastern European countries were left out of the electriCity equation because virtually all households in these countries have electricity Sources LSMS surveys, World Bank for 1998 GNP per capita each country. The coefficients on POOR are consistent for the five models, ranging from -0.43 for a yard tap to -0.63 for sewer. Clearly being poor in one's own country will lower the chance of being connected to these serv- ices at all income levels (note that the poor in some countries have much lower incomes than the poor in other countries). The third income-related 101 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU variable, LOWINCY, attempts to measure whether or not living in a low- income country would have an effect on the likelihood of having a connec- tion. The results show that residing in a low-income country has a negative impact on infrastructure connection, and for house/yard taps and tele- phones this influence can be quite substantial. The other four independent variables are also statistically significant in most of the regressions. As one might expect, rural households are less likely to have a connection. The coefficients for RURAL are statistically significant and negative across the five models. Of all the independent variables, RURAL has the largest impact on the dependent variable across all the models. Home ownership is statistically significant and positive in the electricity, in-house tap, yard tap, and telephone models, but negative for the sewer model. Household size has a small but negative effect on con- nection. Last, households in Eastern European and Central Asian coun- tries are more likely to have connections than households in the other countries. Figure 3.14 depicts the relationship between household income and the predicted probability of having a connection based on the results of the regression models presented in table 3.2. We used the mean value for all the independent variables except for the household income proxy, which we allowed to vary from zero to US$1,300. The probability of having an in- house water tap shows the largest increase across the household income range. The predicted probability of having a sewer connection is the flat- test of the five curves, suggesting that connections to sewers are the most invariant to household income. For both in-house taps and sewer service, the marginal effect of income on the predicted probability of having a con- nection is fairly constant across the income range. This is not true for elec- tricity and telephone service. The electricity curve in figure 3.14 is con- cave, while the telephone curve is convex. This means that the marginal effect of income on the predicted probability of having an electricity con- nection declines as income rises. In the telephone model, the marginal effect of income is rising. Figures 3.15 and 3.16 present predicted probability curves for urban and rural households. As one would expect, the predicted probabilities for all five infrastructure services at all income levels are much higher for urban areas than for rural areas. In both rural and urban areas the probability of having a sewer connection or a house or yard tap is fairly flat across income 102 >0 Ca ~~~~,-? ~~~~ 0~~~~Zs7 0~~- '-'~~ Wa :1 Ga~~~~c -C> C c ) c 0 I 'am~~~~~c 01 E C,~~~~~~~~~~~' 0so a -=~~~~~E- 0 ua I I~~~~~~~~~~~~~~~~~~ o (Mc C 0 Ca >~~~~~~~~~~~~~W0 Li~~~~~~~~~~~~~~~~~~~C 0D 0 .5 I . 'I a)~~~~~~~~~~~~~~ M 0~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~0 ~ ~ ~ ~ ~ ~ ~ 0 00 4) ~~~>0)04 I ~~~~~~~~~~~~~~~ I~~~~~~~~~~~~~~ - I ~~~~~~~~~~~~ * t 0 U~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~ C.00 I ~ ~ ~ ~ ~~~ ~~~~~0~ E S~C 0 " 4o E Ec 0 I ~~~~~~~~~~~~ cm 4) = E - o g r- E-O 0~~~~~~~~~~~~~~~~~~~~~ U~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 0) >= Et -0 s 0 75 -0 0 0~~~~~~~~~~~~i 5.4))4 I'~~~~~~~~~~~~~~~~~' I o ~~~~~~~~~E~ Lei 0 E) 0 4 U. -~~~~~~~~~~~~~~~~~~~oc 104 0 >C00 C) m >~ 0 ,' 0 cC * ~~~~~0~ * ~~~~> I - u,~~ > .C I ~~ E~~wfl * 2 ~~~~~~~~C * *-~~~~~~C 0 0 * ~~~~~~~~~~~~~~~ ~~ 0 W:5 * 0 W~0 * -~~~~~~ - o6 O- IO E~O U a~~~~~~~~~~~~~~~~~~~~~~c~ o 0 0 WW - E ~~~~~~~~~~~~~~~~~~~~ c E W oD 0,0 -o~~~~- I -~~~~~~~~~~~~~~' o ca. 0 C 0 a O I. ~~~~~~~~-2.0c -~~~~~~> m 00 '--: 'R i*- --' - I*~~~~~~ 0,nE KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU levels. For the other sectors income effects differ between rural and urban areas. Income has a greater effect on the probability of having an electric- ity connection in rural areas than it does in urban areas. The case of tele- phone service appears to be exactly the opposite: income has a greater effect in urban than in rural areas. 3.6 Are Other Service Op;ions FiFlGiag the Gap? Formal sector utilities providing electricity, in-house water taps, sewer connections, and private telephones are just one means that poor house- holds can use to meet their demand for infrastructure services. In each of these sectors a number of other options also exist, such as private electric generators, public water taps, private wells, septic tanks, and public tele- phones. These alternatives may be more cost-effective solutions for serv- ing some areas, for instance, septic tanks in rural areas, or they may be more desirable for other reasons, for example, formal utility service may be unreliable. This section examines the extent to which poor households that do not have electricity, in-house taps, sewer connections, and private tele- phones rely on informal service providers such as water vendors; private sources, for instance, private wells; or community service options, for example, public telephones. Are these services filling the infrastructure gap for poor households? How many poor households are left relying on basic or unimproved sources of services? The Energy Sector Electricity is one of several energy sources that households around the world use in their homes. Most households rely on more than one energy source, choosing different fuels for different purposes or substituting one fuel for another as price, availability, or quality changes. Most households in the pooled LSMS sample used electricity for lighting, but few-and even fewer of the poor-relied on electricity for their cooking needs. Households without electricity used other fuels for lighting, cooking, and all other energy needs. Virtually all households without electricity connec- tions used kerosene, gas, or oil lamps for lighting. Few households used candles or flashlights, and even fewer reported having no source of light- 106 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE ing in the home. In Ghana, Nepal, Nicaragua, and Vietnam only 4 percent, 7 percent, 1 percent, and 2 percent of households, respectively, used can- dles, flashlights, or something other than electricity, gas, oil, or kerosene lamps for lighting. In 8 of the 10 countries where dlata about households' cooking fuel are available, less than 2 percent of all households used electricity as cooking fuel.2~' Households that did not use electricity for cooking chose from a range of possible fuels. Other modern fuels include bottled gas or natural gas; at the opposite end of the spectrum are wood, straw, dung, and thatch; and in between are a number of intermediate energy sources such as kerosene and charcoal. Wood, straw, dung, and thatch were overwhelm- ingly the most common cooking fuels arnong both the urban and rural poor in most countries. 26 Not surprisingly, the use of these fuels was higher in rural than in urban areas. Most, andl in some countries virtually all, of the poorest rural householdls use thlese basic cooking fuels, but poor house- holds were not the only ones using wood, dung, thatch, or straw for cook- ing. In the poorest countries in the sample (Cote d'lvoire, Nepal, Nicaragua, and Vietnam), the vast majority of the richest rural households also relied on these fuels. The rural rich in the wealthier countries (Ecuaclor, Panama, ancl South Africa) were, however, mLich less likely to cook wilth wood, dung, thatch, or straw than the rural poor. The urban areas of Bulgaria, Ecuador, and Panama were the only excep- tions to the widespread use of wood, straw, clhng, ancl thatch by poor house- holds. Only 13 and 10 percent of the poorest urban deciles in Ecuador and Panama, respectively, used wood, dung, straw, or thatch for cooking fuel (table 3.3). In Bulgaria, less than 7 percent of the poorest urban households cooked with these fuels.27 Water Many households in developing countries obtain wvater from more than one source. LSMS surveys generally ask for only the primary water source or drinking water source. Households without in-house connections cited a range of other water sources as their primary or drinking water source. Even households with water connections may obtain water from more than one source Some households used unimproved water sources, such as rivers and streams. Others chose from a range of informal, private, or 107 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU TABLE 3.3. Use of Wood, Dung, Thatch, and Straw as Coolcing Fuel among the Poorest and Richest Urban and Rural Deciles Urban areas Rural areas Country Poorest 10% Richest 10% Poorest 10% Richest 10% Low-income economies Cote d'lvoire 92 4 100 94 Ghana 69 20 100 82 Nepal 85 4 100 86 Nicaragua 95 28 99 87 Vietnam 88 27 99 88 Middle-income economies Ecuador 13 0 56 22 Panama 10 0 99 11 South Africa 7 0 84 4 Note Countries are classified by 1998 GNP per capita Low-income economies had GNPs less than US$760, while the middle-income economies all had GNPs of less than US$3,080 Sources Sample households from LSMS surveys, 1998 GNP per capita from World Bank data improved community water sources such as yard taps, public taps, wells, water vendors, or rainwater collection. The use of unimproved water sources was most prevalent in Albania (not including Tirana), Ghana, Nicaragua, and Vietnam. In Bulgaria, Jamaica, Kazakhstan, the Kyrgyz Republic, and Pakistan few households, even in rural areas, obtained water from rivers or streams. Note that in countries where the coverage of in-house taps was high, the number of households still relying on rivers, streams, or springs was not necessarily low. In Alba- nia, for example, 32 percent of households used in-house water taps, yet 42 percent still relied on basic water sources. In Cote d'Ivoire, by contrast, most households obtained water from informal, private, or community sources. Although only a minority of households in Cote d'Ivoire had in- house taps at the time of the LSMS survey, few households relied on rivers or streams as their primary water source. Figures 3.17 and 3.18 examine the relationship between income and household water source choice in the pooled urban and rural samples from all the countries except Nepal (for which the data do not permit such an 108 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE FIGURE 3.17. Primary Water Source in Urban Areas by Decile of Urban Households Percentage of households in decile using source as primary water source 100 90 80- 70- 60- 50- 40- 30- 20- 10 0 86 199 290 448 745 Median monthly household aggregate consumption (1998 US$) * Percentage with in-house tap C| Percentage using other improved sources j Percentage using unimproved sources Note Some LSMS surveys ask for respondents' primary water source (Albania, Bulgaria, Ecuador, Nicaragua, South Africa) The remaining surveys ask for primary drinking (or drinking and cooking) water source The urban/rural definitions used by LSMS researchers were adopted for this analysis The households were divided into deciles according to their per capita consumption Other improved sources include yard taps, standposts, wells, vendors, and rainwater collection Unimproved sources include rivers, streams, and springs Median monthly household aggregate consumption is used as a household income proxy The consumption aggregates prepared by the LSMS survey research teams were adopted for this analysis Source 25,458 urban households in a pooled dataset of LSMS surveys from 14 countries 109 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE FIGURE 3.18. Primary Water Source in Rural Areas by Decile of Rural Households Percentage of households in decile using source as primary water source 100 - 90 80 - 70- 50 40 - 30- 20- 10 __ I 32 108 173 235 340 Median monthly household aggregate consumption (1998 US$) []Percentage with in-house tap [ Percentage using other improved sources D Percentage using unimproved sources Note See note for figure 3 17 Source 26,104 rural households in a pooled dataset of LSMS surveys from 14 countries analysis). In both urban and rural areas a smaller percentage of the poorest households had in-house taps than households in other income deciles and a greater percentage of the poor used informal, private, or community sources. In urban areas few households at any income level were using rivers or streams as their primary water (or drinking water) source. In rural 110 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE areas between 20 and 30 percent of households in all but the richest deciles relied on unimproved water sources. Water vendors are an informal source that has recently attracted much attention in discussions of water service and the poor. Information about water vendors is available in 4 of our 15 sample countries: Cote d'lvoire, Ghana, Nicaragua, and Pakistan. More tlhani 15 percent of households in Cote d'lvoire used vendors as a primary source of water, compared with 1 percent of households in Ghana and less than I percent in Nicaragua and Pakistan. In all four countries a greater percentage of rich households than poor households used vendors Fewer than 1 percent of householcls using vendors were in the poorest decile, whereas 20 percent were in the richest decile. In three of the four countries, households using water vendlors spent, on average, more per month than households with in-lhouse water taps or those using other improved sources (table 3.4); however, only in Pakistan, vhere households with in-house service were spendling very little per month on water, were the median expenditures of those using vendlors significantly higher than those with in-house taps. Figures 3.19 and 3.20 present the distribution of monthly water expenditures by those households in these four countries that rely on vendors and those with in-house taps (note that only households wfith nonzero expenditures on water are included). Two aspects of these figures and table 3.4 are striking. First, average monthly expenditures on water from vendors are not higher than the likely full cost of in-house piped water service. Although the per Linit price of water bought from vendors is certainily higher than the per unit price of water from in-house service, total household expenditures oni water were TABLE 3.4. Median Monthly Household Expenditures on Water by Households Relying on Different Primary Drinking Water Sources in Four Countries (1998 US$) Country In-house water tap Vendor Other improved COte d'lvoire 12 40 13 90 6 90 Ghana 4 90 4 40 1 90 Nicaragua 4 60 6 00 2 40 Pakistan 1 00 7 50 0 80 Source Sample households from LSMS surveys in these four countries KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU FIGURE 3.19. Frequency Distribution of Mionthly Expenditures on Water by Households in Four Countries Using Water Vendors as Their Primary Drinking Water Source Fraction of households using vendors 0.4 - 03 - 02 - 01 H K ] r I _,_ 0 - 0 10 20 30 40 50 60 70 80 90 100 Monthly expenditures on water (1998 US$) Note Nicaraguan households were asked about their primary water source rather than their primary drinking water source Source 328 households in a pooled dataset of LSMS surveys from C6te d'lvoire, Ghana, Nicaragua, and Pakistan 112 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE FIGURE 3.20. Frequency Distribution of Monthly Expenditures on Water by Households in Four Countries Using In-House Taps as Primary Drinking Water Source Fraction of households with in-house taps 04 - 03- 02 - 0.1 0 10 20 30 40 50 60 70 80 90 100 Monthly expenditures on water (1998 US$) Note Nicaraguan households were asked about their primary water source rather than their primary drinking water source Source 3,073 households in a pooled dataset of LSMS surveys from Cote d'lvoire, Ghana, Nicaragua, and Pakistan 113 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU smaller than what one might expect from the water vending literature (see, for example, Crane 1994; Fass 1988; Whittington, Lauria, and Mtu 1991; Whittington and others 1989, 1990; Zaroff and Okun, 1984). These find- ings should not, however, be considered definitive, because such a small percentage of sample households in these countries use water vendors. Second, a large percentage of households with in-house taps were spend- ing almost nothing for water. Sanitation Some of the LSMS country datasets have information on two aspects of a household's sanitation situation: (a) whether a household had a toilet or latrine and (b) whether a household had a means of removing wastewvater from the house, that is, either a sewer connection or a septic tank. Infor- mation on septic tank usage is available in 6 of the 15 countries (Bul- garia, Ecuador, Kazakhstan, Nepal, Nicaragua, and Pakistan). In these six countries more households had sewer connections than septic tanks, but septic tanks nonetheless made a significant contribution to san- itation infrastructure. More than half of all households in Bulgaria, Ecuador, and Kazakhstan had eithcr a sewer connection or a septic tank. By contrast, most households in Nepal (84 percent), Nicaragua (74 per- cent), and Pakistan (63 percent) %Nere without either sewer connections or septic tanks. With the exception of Bulgaria, the poorest households in eachi country had lower rates of coverage of sewer connections and septic tanks than the population as a whole. In Bulgaria nearly all households had either a sewer connection or a septic tank, but the poorest households were more likely to have septic tanks than sewers (table 3.5). Rural households at all income levels had lowver rates of coverage of all sanitation facilities than urban households (figures 3.21 and 3.22). Few urban households were without a toilet or latrine in their home. By con- trast, nearly 30 percent of each rural decile lacked any sort of sanitation facillties. Not surprisingly, the greatest sanitation deficit was among the poorest rural households. Between 80 and 90 percent of households in the poorest two deciles in the pooled rural sample had no latrine or toilet in their homes. Approximately one-quarter of households in the poorest urban decile of the sample had no sanitation facilities. 114 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE TABLE 3.5. Percentage of Poorest Urban and Rural Households with Sewer Connections or Septic Tanks in Six Countries Poorest urban decile Poorest rural decile Percentage Percentage Percentage Percentage Country with sewer with septic with sewer with septic Bulgaria 86 12 18 70 Ecuador 41 14 5 13 Kazakhstan 70 8 8 4 Nepal 7 4 <1 0 Nicaragua 9 3 0 <1 Pakistan 20 15 <1 2 Source Sample households from LSMS surveys in these six countries Telephone Service For households without a private telephone in their homes, having access to a public telephone in their community can be a real advantage. In the absence of a public phone, the presence of at least some private telephone connections in the community may still give households wilthout a tele- phone a means of communication. Telephone ownvers may rent out their phones or allow others to use their telephone for emergency communica- tions. Information on such uses of private telephones is not available in the LSMS surveys, but the community questionnaires in three countries (Ecuador, the Kyrgyz Republic, and Panama) clo ask about access to pub- lic telephones. In these three countries, poor householdls were less likely than the popu- lation as a whole to have access to public telephones in their communities. In Nepal and Panama access to public telephone service increases with aggregate household income. In Ecuadlor access to public telephones is fairly uniform across income deciles. Most of the poorest urban and poor- est rural households in these three countries did n1ot have either their own private telephone or access to a public phone in their community. The only exception is urban areas in the Kyrgyz Republic, vhere Just over half of the poorest urban households had access to a public telephone in their communities (table 3.6). 11 5 +c di ,' 0- s Cdi z~~~~~~~~~~~~~~~~~~~~~~~L o 0~~~~~~~~~~~~~~~~~~~~~~~0 0~~~~~~~~~~~~~~~~ 1 0 C _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ o~~~~~~~~~~~~~~~~~~, 5 E u 4) , CL O~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~- g -= 0 a,~~~~~~~~0 I - rz~~~~~~~~~~ I ° C ! C 0. I 0 * . , C di _ XN o ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~t o C) 00 ~~~~~~~~~~~~0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 0 (.0 E~~~~~~~~~~~~~~~~~~~~~~~~~~~C CL E a, 00 w Ln 0~~~ ..w . M 0 CC I = = UU 00 0 o di ___ _A_ 116 - w C~~~~~~~~~~~~~~~~~~~~~C 0 0 1 1 co 0 0~~~~~~~~~~~~~~~ o~~ ~ - o S 0 UN C u ~~~~~~~~~~~- O 'c E _ ~0 0 w -~ (U -'- o - @ :2 E a c 4t L (U 00m~' 0 c 4, cD c: ~~~~co C. CD CD CD " "CD'' -° cn Z CU ( CL 0 - 0- cc a, ' I u I t - * 0~~~~~~~~~~~~~~~~~~~~~~ 117 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU TABLE 3.6. Percentage of Poorest Urban and Rural Deciles with Access to a Public Telephone in Their Community in Three Countries Country Percentage of poorest urban decile Percentage of poorest rural decile Ecuador 15 12 Kyrgyz Republic 60 29 Panama 33 4 Note Community is defined as the primary sampling unit in which the household lives In urban areas this is typically smaller than the entire city, and in rural areas the community may consist of more than one village Source Sample households from LSMS surveys in these three countries 3.7 Conclusions Coverage statistics are widely used to paint a picture of infrastructure con- ditions in developing countries, and they are often the only global, cross- country data available for infrastructure services. It is thus important to use coverage statistics to their fullest advantage, while at the same time being careful not to read more into the data than they can actually reveal. In this chapter we have used a new data source, the World Bank's LSMS surveys, to construct infrastructure coverage statsstics for a pooled sample of house- holds from 15 countries. Several of the results from our analyses using these LSMS datasets are worth recapping. First, electricity coverage was higher than coverage of other infrastructure services at all income levels: 65 percent of the house- holds in the sample had electricity in their homes. By contrast, only 38 per- cent of households had in-house water taps, the infrastructure service with the next highest level of coverage. The relative ranking of coverage rates among the four infrastructure services (electricity -4 water -4 sewer -4 telephone) held across all income levels. Second, infrastructure coverage for electricity, water connections, and sewer connections rises rapidly as household income, as measured by a con- sumption aggregate, increases from about US$100 to US$250 per month. We want to emphasize again that the 55,500 households in this pooled dataset are not representative of the global population in developing coun- tries. We believe, however, that our findings regarding these relationships between infrastructure coverage and household income are relatively robust 118 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE with respect to the countries in the pooled sample and the sampling proce- dures used within countries. Third, electricity was the only infrastructure service wvith significant pen- etration among the poorest 5 percent of the sample households, 32 percent of which had service. Only 6 percent of the poorest households had an in- house water connection and only 3 percent had a sewer connection. Almost 80 percent of the poorest households in urban areas had electricity serv- ice. Even in rural areas, 27 percent of the poorest households in our sam- ple had electricity service. When a household hacl the opportunity to con- nect to the electricity network, the vast majority did so regardless of their income level. This was not true for the other three infrastructure sectors. Moreover, when households had a choice among all four infrastructure services, thiey appeared to choose electricity first. Fourth, few households in the sample relied on electricity as a cooking fuel. The vast majority of poor households in both rural and urban areas used wood, straw, dung, and/or thatch as their primary cooking fuel. In the poorest countries in the sample, even the majority of the richest rural households also relied on these basic fuels. Fifth, although the majority of households in the pooled sample dld not have an in-house water connection, relatively few households were using unimproved water sources, such as rivers or streams, as their primary source. In urban areas few houselholds at any income level were using unimproved water sources. In rural areas between 20 and 30 percent of houselholds in all except the richlest income deciles relied on unimproved water sources. Water vendors wvere not a major water source for househiolds in the four countries in the sample in which these data were collected; however, those households that purchased water from vendors were usually not paying much more per month than the likely full cost of private in- house water service (although the price per unit of water purchased from vendors is almost always higher than the price per unit of water from piped distribution systems). Sixth, in those countries in which the LSMS surveys collected informa- tion on toilets, latrines, and septic tanks, most urban households had a toi- let or latrine in their home. The greatest sanitation deficit existed among the rural poor: 80 to 90 percent of poor, rural households had no sanitation facilities of any kind. This will come as no surprise to thiose working in the water and sanitation sector. 119 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU Appendix: Construction of Cross-National! Infrastructure Variables LSMS surveys generally ask for each household's primary source of drink- ing water, the energy source used for light and cooking, and whether or not the household has a telephone. The surveys ask infrastructure questions in similar ways across countries, but the answer categories are usually differ- ent from country to country. Therefore much detail was lost in creating the cross-country variables for this study. In general, it was possible to create three categories of infrastructure use for each sector: "advanced," "inter- mediate," and "basic" infrastructure solutions. Variable Advanced Intermediate Basic Water sector Water source In-house water tap Other improved sources, River, stream, spring such as yard tap, public tap, well, rainwater, vendor Sanitation sector Toilet Flush toilet Latrine, no-flush toilet or No toilet or latnne other toilet (for example, (includes bucket toilet chemical) open hole) Sewer/septic Sewer connection or septic tank Energy sector Electricity Electricity (from grid or generator) Cooking fuel Electricity, bottled gas, Kerosene, charcoal, coal Wood, dung, thatch, straw natural gas Telecommunications Telephone Private telephone Access to public telephone No access to public in community telephone and no private telephone Source Authors Notes 1. By coverage we simply niean whether or not a household has an infrastructure service, such as electncity or a piped water supply. If a household cloes have a par- ticular service, it is said to be covered. 2. Coverage data can aid in describing an existing infrastructure situation, but they cannot be used to determine why such a situation exists, even if one were able 120 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE to go back to thie original clatasets. This is because most surve)s on which the cov- ciage sumimiiiaries are based (lo not ask respondents wlhat services they couldl have choseni (but did not) and about thie attributes of such service optionIs, for example, price, quality, reliability. What wve see in the coverage statistics is the outcome of bothi supply and (lenialid factors that bear on a lhouselhold's infrastruCture choices, but policy analysts generally cannot disenitaigle such factois from the coverage sta- tlstlcs. 3. For exarirple, a responidenit may be asked, "What is the househiold's principal water souice for drinking and cooking9" Sonie surveys may use precoded answers that chstinguishi between in-hotise corriiectionis andi yardl taps, but others ma) iiot. 4. Wle reserve the terimi access to refei to a lhouselhold's ability to obtain an infia- structure coninectioni should the household decide to do so. Foi example, a house- hIold hlas access to sewer service if theie is a sewer nietwork in the houselhold's neighborhood 5. We chose these 15 LSMS surveys because tire data anidl suppoltinig docu- reniiLtioni were readily available, andc because they contain all or most of the unfrastiLuctUre and( hou.sehold consumptioni variables of interest Wheni niore thani one suivey year was available for a particular counitry, we usedl the most recent survey year 6. -louseliolcl selection in eacih sur-vey wds random, hut that does not mean thiat the 15 suiveys togethier give uis a random selection of all lhouselholcds in thle devel- oping wo Ild. 7. Tiis note of caution is particularly important for counitry-specific results. Sonie LSMS surveys are self-weighting In this analysis no weighits have been usecd to a(hjust for sample lesigin or nonresponse. Thus thic results are applicable for the sample populationi only. 8. Purchasing power parity conversion wouldl have been preferable, but COnlSUrllp- tion-heading coniversioni factors, which WoUld liave been used to convert infonimationl on expenditures, were not available for all sectors, all couLitries, or all survey years. 9 The consumiiptioni aggregates usecd hele vere prepared by LSMS survey reseaclh teanis. The aggregates combine inforimiation collected fronri houselholdIs about thieir expenditures on andl conisumilptioni of a host of foocl arid nonfood items. 10. These hoLiseliolds may obtain electricity fioni a utllity coninectioni or fronii an electrical generator. Unfortuiriately, differentiating between tIlese soLices is possible in only 4 of the 15 countries. Ecuador, Nepal, Nicaragua, ancd Panamila In these four countiles electrical generators for lhouselhold use are rare. In Nepal ancI Nicaragua fewer than 1 percent of hIousehiolds withi electiicity reported obtaining the electricity from a generator. In Nicaragua arid Panamila 1.5 andl 3 5 percent, respectively, of hlouseholdis withi electricity rely on generators. In Paiianila poor households are soriewivat more likely thani rieci househioldls to rely on generators for electricity, nearly 12 percent of those in the poorest quriitile 121 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU who have electricity versus 2.3 percent of the richest quintile use generators. In Nicaragua few of the richest or poorest obtain electricity from generators. The Nicaraguan generator users are disproportionately concentrated in the middle- income quintiles. 11. Information on telephone service is available in onily 10 of the 15 countries, sewer connection information is available in 12 countries, and electricity data in 14. These coverage figures reflect thie percentage of sample hiouseholds for whom data on the service are available who lhave the service in their homes. 12 As a result, the coverage differences between sewver and water service and the important role of yard taps in water service coverage, both of which are appar- ent in figures 3.2 and 3.3, will be obscured. 13 Thiis analysis uses the urban versus rural classifications made by LSMS sur- vey teams. 14. More than 91 percent of hlouLseholds in the poorest quantile of the pooled saniple live in rural arcas, vhereas only 13 percent of the richest househiolds are rural residents. 15. Note that some sucli households may have consciously chosen to live in a place without access, because rents or land values wvere cheaper there. 16. The LSMS community surveys provide an imperfect measuLre of access. The surveys make it possible to deterniine whether an infrastructure network is avail- able wvithin each respondent's community, but having a network in the area does not necessanly meani that extenidinig the network to all homes in the area is techni- cally or financially feasible. Nevertheless, the community data roughly dlivide households into two groups: those with no possibility of connecting to a network and those that may have a chance of connecting 17. Figure 3.11 shows a strange result foi telephone access. A highi percentage of the poorest households hiave access to private telephone services in their com- murnities. As income increases, the percentage with access first drops and then rises agaln This result is due to the influence of the Kyigyz Republic. Many of the poorest households represented in the figure live in the Kyrg)z Republic. where access to private telephone service is high even for the poor. Whenl the Kyrgyz Republic is ornitted, access to telephone service in the remailling four countries is extremely low for the poorest hiouseholds and rises among higher- income groups 18. The low connection rates in rural areas may reflect problems with the access data more than poor households' willingriess to connect. Rural commullities and primar sampling units cover a larger land area than urban communities. The fact that there is access to a particular service somewlhere within a rumal community does not necessarily mean that installing a connlection in every home is techbmcally or financially feasible While this is true in urbani areas as well, this weakness in the community survey data is especially problematic in rural areas 122 INFRASTRUCTURE COVERAGE AND THE POOR A GLOBAL PERSPECTIVE 19. In Ecuadlor and Panaiia, where informationi on yard taps is available, a greater peicentage of urban and rural housellois of all quatitiles cliose an In-lhouse or yar l tap thani a sewer coninectiorn (giveni access to cach service) 20 Nearly 2,500 of thie 3,000 households with access to all four services live in Ecuador adni Panania In these Cou0ntries inforimiationi on yard taps is available. Nearl 95 percent of the 2,500 househiolds have an rn-house or yardl tap. 21 Because thie LSMS suiveys lack infoniiation on service piIces, dletermininiig how vdifferences in prices andl connectioni fees for these services contribute to this outconrie is not possible. 22 Inforimiationi on telephone access is not available Ii Nepal, so figure 3.12 slhowvs coverage levels b) quintile arid sector for households wvith access to just thlee services 23. The water sector variable for thle EcLrador and Panamila figurcs Is house or yar d taps Thle Nepal and Kyrg z Republic figures include onl) in-house taps 24. IT Kazakhstani a greater percentage of house0- ollds also chose clectricit) than in-1ioLise water taps or telephones; hiowvever, Kazakhistan departs from the pattern In figures 3 10-3.13 in that coverage rates for electricity, in-louse taps, andl tele- phonies are fairl) flat acIoss all quintiles. 25. Bulgaria and South Africa are the only two coLintries in the sample whele a significdalt numliber of hoLIseholdls reported using electricity for cooking (75 percent In Bulgaria aidn 43 percent in Southi Afiica). In Southi Africa alirrost no houselhol ls in the poorest ilecile cook with electricity (3 percent), bLIt 66 perceilt of the poor- cst decile of 13ulariai1 IroursClIolds relh on clectricity for cooking. 26. Inifoiriationi about cooking fuels is available fou a subset of the 15 coulitries Bulgaria, CrIte dIlvoiie, Ecualoil Ghana, Nepal, Nicaragua. Panamia, Soutih Africa, arid Vietnarri. 27 In Bulgaria (hifferenitiatinig between ifiteriniedliate adnl basic fuels is not possi- ble. but 7 percent of the poorest urban decile use eithier ruterilechiate fuels or wood, duig, thatch, or strawv References Crane, Randall. 1994. "Water Markets, Market Reform, and the Urban Poor: Results from Jakarta, Indonesia." World Development 22(1): 71-83. Fass, Simon. 1988. "Water." In Political Economy ti Haiti: The Drama of Survival. New Brunswick, New Jersey: Transaction Publishers. Whittinglon, Dale, Donald T Lauria, and Xinming Mu. 1991. "A Study of Water Vending andl Willingness to Pay for Water in Onitsha, Nigeria." World Development 19(2/3): 179-98. 123 KRISTIN KOMIVES, DALE WHITTINGTON, AND XUN WU Whittington, Dale, Donald T. Lauria, Daniel Okun, and Xinming Mu. 1989. "Water Vending Activities in Developing Countries: A Case Study of Ukunda, Kenya." Internattonal Journal of Water Resources Develop- ment (September): 158-68. Reprinted in Richard Layard and Stephen Glaister, eds. 1994. Cost-Benefit Analysis. Cambridge, U.K.: Cambridge University Press. Whittington, Dale, Apia Okorafor, Augustine Okore, and Alexander McPhail. 1990. "Strategy for Cost Recovery in the Rural Water Sector: A Case Study of Nsukka District, Anambra State, Nigeria." Water Resources Research 26(9): 1899-1913. Zaroff, Barbara, and Daniel A. Okun. 1984. "Water Vending in Develop- ing Countries." Aqua (5): 289-95. 124 4 Measuring the Impact of Energy Interventions on the Poor-An Illustration from Guatemala Vivien Foster and Jean-Philippe Tre 125 VIVIEN FOSTER AND JEAN-PHILIPPE TRt 4.1 lntroduction Following a decade of energy sector reforms in many developing countries, people are increasingly beginning to question the extent to which these reforms have benefited the poor. This question has proved difficult to answer, in part because of the absence of a framework for thinking about the issue, and in part because of a shortage of suitable data. Yet ensuring that energy sector interventions are designed in such a way as to benefit the most vulnerable in society is important from the perspective of social equity and also because this is likely to affect the social acceptability- and hence the ultimate sustainability-of reforms. This chapter proposes a methodology for measuring the impact of inter- ventions in the energy sector on the welfare of poor households. The aim of this methodology is not only to make it easier to answer questions about how energy sector interventions have affected the poor, but also to help focus attention on poverty issues ex ante, thereby motivating the adoption of pro-poor features in the design of interventions. Here we take energy sector interventions to refer to any measure that sig- nificantly affects the cost, quality, and conditions of access to energy ser- vices, whether wholesale sector reform or a smaller-scale investment project. These interventions include restructuring, privatization, and liberalization of traditional electric and natural gas utilities. They also include policy deci- sions affecting the availability and relative prices of alternative energy sources such as traditional biomass (firewood, charcoal, and so on) and com- mercialized fuels (kerosene, liquefied petroleum gas [1.PG],and so on) that are perhaps of more immediate relevance to poor households. The proposed approach is based on a set of welfare indicators that are suffi- ciently broad to capture the kinds of energy issues that are likely to be of con- cern to poor households. The values taken by these indicators for poor house- holds should be calculated prior to any energy sector intervention. The patterns revealed by the indicators should help to identify which types of energy sector interventions are most likely to benefit the poor, and also pro- vide a baseline for subsequent evaluation. Once the intervention has been completed, the indicators should be recalculated and compared against the baseline values. The poverty impact of the intervention can then be gauged in terms of the movement of these welfare indicators for the lowest income groups. This overall process depends critically on the availability of datasets that combine information about energy use with poverty indicators (G6mez- 126 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR Lobo, Foster, and Halpern 1999; World Bank 2000). To clemonstrate the feasibility of the methodlology, the calculation of the indicators is illustrated using standardl household survey data fromii Guatemala. Measuring the welfare impact of energy sector interventions on the poor is not quite the same as measuring the impact of energy sector interventions on poverty itself. For example, an energy pricing reform in a particular country may reduce the cost of electricity to poor households, leading to a direct increase in their welfare. This very price change mighit indirectly also serve to take some of these households out of poverty altogether, for example, by releasing women and children from the time-intensive task of gathering tra- ditional biomass fuels, by raising the productivity of women in household chores, or by facilitating the operation of home-based microenterprises. How- ever, this ultimate effect, though measurable in principle, is much harder to gauge with any degree of reliability (Chong and Hentschel 1999). In particu- lar, attributing overall changes in poverty rates to one intervention versus another is difficult. Thus the more modlest objective of examining how those who are currently poor benefit directly from energy sector interventions is probably also more useful for the purposes of impact evaluation. Although this chapter focuses on the energy sector, many of the ideas devel- opecl are potentially applicable to the other infrastruLcture sectors-water, telecomrriunications, ancl transport-with relatively minor adlaptations. The remainder of this chapter is organized as follows. Section 4.2 reviews some of the stylized facts of energy consumption and poverty, with a view to identifying what kinds of indicators might be relevant to undler- standing the energy situation of poor householdls. Section 4.3 presents a series of indlicators for the welfare impact of energy sector interventions. Section 4.4 presents a number of poverty measures and shows how they might be combined with the energy sector indicators. Section 4.5 dis- cusses the data issues that arise in implementinig this methodology. Sec- tion 4.6 illustrates the methodology using data from Guatemala Section 4.7 discusses the applicability of the method to other infrastructure sec- tors. Finally, section 4.8 concludes. 4.2 Stylized Facts about Energy and Poverty Before we outlinie the pioposed approach, a brief review of some of the styl- izecl facts about energy consumption and poverty is helpful (Albouy and 127 VIVIEN FOSTER AND JEAN-PHILIPPE TRt Nadifi 1999). The energy literature has traditionally been dominated by a theory of transition, whereby households gradually ascend an energy lad- der that begins with traditional biomass fuels (firewood and charcoal), moves through modern commercial fuels (kerosene and LPG), and culmi- nates with the advent of electricity. The ascent of the energy ladder, though not fully understood, is thought to be associated with rising incomes and increasing levels of urbaniza- tion. However, one of the central findings of the empirical literature on energy and poverty is that reality is rather more complex than this sim- ple transitional theory would appear to suggest. In particular, at any given time households tend to rely on a range of different fuels that typically encompasses at least two of the steps on the energy ladder (Barnes and Qian 1992; Eberhard and van Horen 1995; ESMAP 1994; Hosier and Kipondya 1993). A number of explanations can be advanced to account for this phenomenon. One view is that low levels of reliability require households to adopt a diversified energy portfolio to assure security of supply. Another possibility is that different sources of energy are more cost-effective for some uses than for others; therefore it may make eco- nomic sense to use electricity for lighting, but LPG for cooking. Thus conceiving of different energy ladders for different types of applications may be more appropriate. All this means that any indicators purporting to measure the welfare impact of energy sector interventions on the poor need to consider a house- hold's full portfolio of energy sources rather than focusing solely on a sin- gle source of energy, such as electricity. Thiis is important, because many of the traditional indicators of the welfare impact of energy sector interven- tions tend to concentrate narrowly on the electricity sector, for example, by measuring the number of household connections or the proportion of household expenditure being spent on electricity and so forth. This over- looks the fact that interventions affecting the prices and availability of liquid and solid fuels may have just as much, if not more, of an impact on the welfare of poor households as electricity sector reforms. Indeed, as the previous literature review illustrates, this remains true even after house- holds have been provided with an electricity connection. Hence the aim of the following section will be to broaden some of the traditional electricity- based indicators of welfare to encompass the full range of fuels used by households. 128 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR 4.3 Indicators of the Welfare Impact of Energy Sector Interventions To choose an appropriate set of indicators, a working definition of human welfare as it relates to interventions in the energy sector is also necessary. In common with the literature (Lok-Dessallien 1999), this section takes three different perspectives on human welfare-the satisfaction of basic needs, monetary measures, and nonmonetary measures. Many of the indicators presented in this chiapter are based on informa- tion about physical energy use. In this context distinguishing between gross energy usage and net energy usage is important. Gross energy usage refers to the total amount of energy embodlied in the diverse fuels purchased by the household, that is, it represents the thieoretical maximum amount of energy that couldI be derived from these fuels in a laboratory context. In practice, extracting this full energy content in household uses is not tech- nically possible, and a significant proportion of the embodied energy con- tent will be dissipated in the process of use. The extent to which house- holds are able to extract useful energy from embodied energy will dlepend both on the fuel concenecl and on the technology with which it is used. Net energy usage refers to the useful energy that the houselhold manages to extract from the energy embodied in its fuel purchases. Where relevant, all indicators can be expressed on either a gross or net basis. Wlhile in some instances a comparison of gross and net values of the indicators may be of interest, in general it is the net value of the indicator that is of greatest policy interest because it is the closest reflection of the energy Lisage sitiation of the household. The ratio of household net to gross energy consumption will be referred to as the efficiency factor, and is itself a useful indicator of the quality of the energy used by the householdl. Bastc Needs Welfare Measures According to one traditional view, wvelfare relates to wilether or not people are able to satisfy their most basic material needs. While this view is intu- itively appealing, its main drawback is the subjectivity involvedl in defin- ing a basic need (Hicks 1998). As far as the energy sector is concerned, it raises the question of the extent to which energy can be regardedl as a basic need and how a basic energy need might be defined. 129 VIVIEN FOSTER AND JEAN-PHILIPPE TRE While policymakers have sometimes defined an electricity connection as a basic need (box 4.1), this view appears to jar with the stylized fact pre- sented earlier that households tend to use a wivde range of fuels even when electricity is available to them. Thus defining a basic energy need in terms of reliable access to one or more sources of energy, electricity or something else, would seem to be more plausible. From this particular perspective, the key issue is whether or not house- holds have adequate access to energy services. The most basic access indi- cator is coverage of energy services. Coverage is defined here as potential access to a given source of energy, regardless of whether or not a house- hold chooses to make use of that energy source. In this sense, coverage defines the household's opportunity set of fuels. Knowing which fuels a household could potentially use as well as which fuels the household actually chooses to use is relevant from a policy point of view, because it helps us understand the extent to which observed patterns of energy usage reflect demand decisions or supply constraints. Coverage indicators are already widely used for electricity infrastructure, but less so for other energy sources where they could be just as useful. The reason is that access to traditional biomass and modem commercial fuels is by no means universal, but may be limited in the former case by local environmental factors and in the latter case by deficiencies in commercial distnbution net- works (Barnes and Qian 1992). In addition to looking at coverage rates for different energy sources individually, summing the number of different types of energy sources to which each household has access may be helpful. A limitation of the basic coverage indicator is that it says nothing about the reliability of the service. A household may have an electricity connec- Box 4. U Energy in the Basic A;escs App o,ch| A'Vunerou.s lAtin Amerikan countrieh traditionn all/i measured poureri hI using multidimenisional indexes oJ uasaifir/Wd ho.~s nerd%. The e.art cOtt'eti of the indexe.s raries from countri ta couninrir ho vi Pr. I/let, gcnerallb inr(lude measures ojfhousing qualitj. saiIilionii. and .lducalionI all inil ciii. recent surrel o f husic needs inderve in .lotin imerfra t1!,rli; 1998) fonnd that out of the 1'3 countries sur7e! ed. onl' 83 1 ol i,a. Panamu, anrd Peri) considered an electricity connection to be a uaic fneedi. 130 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR tion, but receive the service for only a few hours each day. Similarly, access to other types of fuels may be intermittent and uncertain. As noted earlier, it is this very unreliability of energy sources that leads many households to maintain a diversified portfolio of fuels. Thus having some way of measur- ing the reliability of fuel supply would be useful. A reliability index could be constructed by asking poor households to assess what proportion of the time they are able to obtain a particular source of energy. Once again, aggregating this information across fuel sources may be helpful, and can be done by taking a weighted average of the reliability score for each energy source, with the weights corresponding to the shares of each energy source in the household's overall effective consumption. The extent to whlch householdIs' energy usage is spread across cifferent types of fuels or concentrated in any given one is also of interest. In com- bination with previous indexes, thlis reveals the extent to which households may be particularly vulnerable to shocks in the availability or pricing of one particular fuel. A useful way to summarize the degree of diversifica- tion in a household's energy portfolio is through a consumption concentra- tion index measure Concentration indexes can be calculated as the sum of the squares of the shares of different energy sources in a household's effec- tive energy consumption. From a public policy perspective, the value of the various access indlica- tors is that they help define an overall picture of the availability of differ- ent fuels to different socioeconomic strata of the population, and thus the options that different groups face. They therefore reveal the extent to which there may be supply bottlenecks that prevent poor households from using particular fuels. Economic Welfare Measures 'rhe standard economic view is that the household's purchasing power, whethet measured by income or consumption, provides the best overall indicator of welfare. Energy sector interventions might affect economic measures of well-being in various ways. The most direct influence is that they may reduce (or perhaps increase) the cost of satisfying energy require- ments, and thereby increase (or reduce) the purchasing power of a given level of household income. The household may use such an increase in pur- chasing power either to increase its energy use or to expand its consumption 131 VIVIEN FOSTER AND JEAN-PHILIPPE TRE of other goods, with both outcomes resulting in an improvement in eco- nomic welfare (or vice versa). A traditional monetary indicator of welfare widely used in the electricity sector is the proportion of household income or expenditure devoted to energy. A high share of household expenditure on energy is taken to imply an "unacceptable economic burden" of meeting energy requirements. Although relatively simple to calculate, this indicator confounds a number of different effects, which in turn complicates its interpretation. For exam- ple, a high share of energy expenditure could be due to a high level of con- sumption as a result of large household size, high levels of discretionary use, or low efficiency of use; it could be due to high unit prices of energy; or it could be due to exceptionally low levels of income. Each of these potential explanations has different policy implications. Perhaps a more useful way of thinking about the affordability of energy is to examine the extent to which households are able to purchase enough energy to meet subsistence requirements. The subsistence threshold would need to be externally defined, and should be based on what the household would require for basic functions such as lighting, cooking, and, depend- ing on the climate, heating.' T 1he corresponding affordability index could then be expressed as the proportion of households whose energy consump- tion exceeds the subsistence threshold. The same information could also be expressed in continuous terms as the ratio of each household's energy consumption to the subsistence threshold. In addition to overall affordability, identifying any distortions in the pric- ing of fuels is also important to gauge how these distortions may have a dif- ferential effect on the rich and the poor. This gives rise to two further indi- cators. The first is the average fuel cost equal to total household energy expenditure divided by total energy consumption. In the case of biomass fuels, which tend to be gathered directly from nature, some shadow price must be found. A standard approach is to consider the value of the time spent gathering the fuel. Alternatively, where parallel markets exist, as may be the case for fuelwood, the market price of the fuel may be used. The sec- ond key pricing indicator is the average subsidy. which can be calculated by taking the unit subsidy on each type of fuel and weighting it by the share of that fuel in each household's total energy consumption. An important drawback of the average energy cost measure is that it over- looks the costs of complementary capital investments, such as light bulbs, 132 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR stoves, and so on, required to make procluctive use of the fuel. This can create a misleading impression, because some energy sources have low fuel costs but high capital costs, and vice versa for others. To the extent that poor households are credit constrained, high capital costs may represent a barrier to entry that prevents them from taking advantage of cheaper fuels. Although cumbersome, estimating the average capital cost of energy use by adding together the amortized capital expenditure on energy-relatecl durable goods may sometimes be feasible and desirable. Box 4 2 provides an Interesting example for the specific case of cooking fuels in Tanzania. The example also illustrates how the incidence of subsidies varies across different types of fuels in 'Tanzania. When both energy cost and capital cost measures are available, the average total cost of energy should also be calculated. Computing an index of capital intensity, definedl as the household's total capital expen- ditures divided by its overall capital and operating expenditures, may also be of interest. Furthlermore, one can combine some of the types of information described earlier to produce a more informative measure of economic burden. For example, it might be interesting to track how the cost of acquiring a subsis- tence level of per capita consumption changes as a percentage of per capita income (or expenditure), or how the total subsidy received at a subsistence level of consumption changes as a percentage of household income (or expenditure). These measures have the advantage of holding consumption constant at a level thought to represent a basic requirement, and thereby avoid confounding quantity and price effects. Broader Measutres of Welfare In recent years there has been a trendl toward complementing economic measures of welfare with nonmonetary measures to obtain a more multi- dimensional view of human well-being, in particular, by tracking health and education indicators. Some evidence suggests that interventions in the energy sector could directly influence human healtlh and even educa- tional outcomes. As regards health, indoor air pollution is a potential cause of respiratory illness in households that rely on traditional fuels. Furthermore, paraffin poisoning of children and serious burns have also been documented as problems in households lacking access to electricity. 133 VIVIEN FOSTER AND JEAN-PHILIPPE TR8 Boz 4.2 Cost of Meeting Energy 1Requiremeni3 1'or Cooking In Dar es Salaam, Tanzania HIosier and Kipondia (199.3j condu tel (ifi intrcrrs'tiigi o(mpuratil e anfnli i% of the costs of using alternative cooking /tels to nor 's .Slorinm. Thn:oaln. The) took into a cooint the differing efficienc! rates aft v lii h gross energi purchases can be transfirmed iinto nn t eniergi se. rangingfronm l /percfnt forfueluwood to65 perf eottfoirelectri' iti. 1/ic%f U/fl /ic flnano ul c per effective megajoule of energ? reported in tile table heluoi. These cfin be com- pared with the trme eronomic costs, repoteied in the ne.tt ( oltomn, which adjus Jor the distortionar) effect of sfh.sifc' fifi( dotites and i/.so tfake intof oicifoiint theforeign exthange (oiiiponent oJ imported!fiel%. 7e lflerene beeeten flinancial anid economu costs (ire ,hs%tantial, purticilarbh in the ( oge of elem- tricity. which is heavilY subsidized. Tile seconds tep in the analhsis is to tahke inlo off loiont the cost of the corn- plementari appliances that are required to (oo) titi/ thr sr / ariousufuls, which is presented in the tuble as uil aniorltized miont/ilii ist. Thle rlatn thoue a uid(le raoge of capital costs, uith e0/( tnt (t e hoing hi far the mist e tp0v0110. Summing the economic toit oj a notioinul (oof kiiig bodget of .320 megajoiles per month and the associated capital cost vieldt the fotuls reportied in the lnst tuo (olumns on both afinancial (rl] fr u (n 'colomorc (ost bJiis^. 1L refis eler- tricity is the cheapest (oo/.ingfiel in terms of frinioctol (fst, it is the most expensive in terms of economnic cost%. | otal mon?hly Fuel cost c th'e economic rost (T Sh/eftectieve fiJ o of 320 i./nimonth (35 J7) Fuel Financial Economic (T 57/mont: aaiil Econcmic Firewood 3 94 5 27 n.a 1,259 35 1,686 fiO Charcoal (traditional) 3 59 5 64 7722 1,169 81 1,877 02 Charcoal (improved) 2 39 3 76 125 00 890 06 1,328 20 Kerosene 5 24 913 33 33 1,709 52 ),95fi 93 LPG 3 17 4 49 708 33 1,224 21 1,65 13 Electricity 0 62 10 38 45i 33 657 99 3,779 93 n a Not applicable. MJ Megajoule T Sh Tanazania shilling Snf,,u.. llo r and Kplrpx,,sa (1'')J91 134 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR Box 4.3 reviews the evidence from South Africa on each of these points. The link between energy and education has not been studied in great depth so far, but some recent findings suggest that electric illumination significantly increases the number of hours that poor children are able to spend reading and studying (Domdom, Abiad, and Pasimio 1999). Whether or not such health and education impacts are of interest in any particular instance will depend on the specifics of each case. Trying to col- lect baseline evidence on the potential severity of any existing problems of this kind would certainly be desirable. Where the issues prove to be impor- tant, two types of indicators could be used. The first type aims to measure the exposure levels of poor households, whether in terms of indoor air pol- lutants inhaled or the number of hours of reading undertaken (with the lat- ter being somewhat harder to capture). The seconcl type of indicator tries to capture the consequences of these exposures, which might be in terms of the incidence of respiratory illnesses in poor communities or the rate of grade completion among school-age children. While theoretically of greater interest, the problem with this latter type of indicator is that isolat- ing the impact of the energy sector intervention from other factors that might also contribute to health and educational attainmenit becomes difficult. Box 4.3 South African Evidence on the Health Impacts of Different Energy Sources Eherhard and( van Horen (7995) reviewed the ernpLrical evidenice on the health and wider social rinpacts of different energy sources In South Afrtca. Theyfirst considered a numl)er of small-scale studies that used personal exposuLre monitors to 1e7asure tie tintake ofpartictlates among children. They conelided that children liviiig in urban homes reliant on coal inhale rnore thanfifve times the U.S. Environmental Protection Ageney recomi- mrended 24-hour limit of 260 nigm-¢, while children living in riiral hiomes reliant otnfuelwood inhale more thiati nine times this linit. A health strvey conduteted as part of the same study revealed that children froni homes using coal were 190 percenit iiore ltkel) to develop lower respiratory illness (piieit- monia, bronchitis, astlma, and so on) than childrentfrom electrified homes. They noted that acute respiratory infections are the second most important cause of child mortality in South Africa. 135 VIVIEN FOSTER AND JEAN-PHILIPPE TRt Summary This section presented a number of possible indicators for measuring the impact of energy sector reforms on household welfare, which are summa- rized in table 4.1. As noted, all the monetary indicators can be calculated on the basis of either gross energy usage or net energy usage. While the latter are of greater policy relevance, comparisons between gross and net indicators may be helpful in certain contexts. The access and affordability indicators will be relevant to most interventions, while the broader health and education indicators may be of greater interest in some cases than others. Calculating all the indicators in all cases will not necessarily be feasible or desirable. To aid in selection, the most important-and easily calculated-of the indicators shown in table 4.1 are marked with an asterisk. 4.4 Combining E1nergy Sector ][nformation with Poverty Information All the indicators presented in the previous section provide general infor- mation about the welfare impact of energy sector interventions on any household. To be able to say something about welfare impact on the poor, the indicators must be calculated separately for poor and nonpoor groups in a society. For these purposes, thinking of poverty in economic terms is helpful, that is, relative to the level of overall household income or con- sumption; however, asking whether an absolute or relative concept of eco- nomic poverty is most useful for this type of analysis is relevant. Many countries have developed poverty lines, which typically purport to capture the cost of acquiring a basic basket of food and nonfood requirements (Lanjouw 1999; Ravallion 1998).2 Using a poverty line in combination with some overall measure of economic welfare, such as aggregate income or con- sumption, it is possible to make absolute judgments about which households are poor and which are not and to examine the differential impact of energy sector reforms on these two groups. However, the construction of poverty lines is far from straightforward because of the difficulties of establishing the basic basket of goods. Moreover, dividing the population into two such broad categories may conceal important gradations within each group. Perhaps a richer way of analyzing this question is to classify households according to their relative position in the overall distribution of income (or consumption) 136 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR TABLE 4.1. Summary of Proposed Welfare Indicators Type of indicator Indicator Comments General Effioency factorf-ratio of net to gross total household energy consumption Basic needs Coverage index*-whether or not a household Coverage does not necessarily mean has access to a particular energy source May be that the household uses that fuel, but aggregated to give total number of available that they have the option to use it energy sources for each household should they so desire Does not take reliability of supply into account Reliabilty index-percentage of time, on Requires a sublective household-level average, that a particular energy source is available assessment of reliability for use by a particular household May be aggregated as a weighted average Concentration index-the sums of the squares of the shares of each energy source in a household's overall effective energy consumption Monetary Affordability index -percentage of households There may be a high degree of whose energy consumption exceeds a subsistence subjectivity in determining the threshold Can also be measured as the ratio of a subsistence threshold household's energy consumption to a subsistence threshold Average energy cost-total cost of a household The indicator fails to take into account energy basket divided by the household's total the capital costs of using fuels energy consumption Average energy subsidy'-average of the unit subsidy for each energy source weighted by the share of that energy source in the household's total energy consumption Average capital cost-total amortized capital Requires a considerable amount of cost of energy-related durables divided by the information about household capital household's total energy consumption goods Average total cost-total energy and related Calculating the amortized capital costs capital expenditures divided by the household's of durables for the full range of fuel total energy consumption uses is likely to be complicated Capital intensity-ratio of total capital costs to total capital plus energy costs Economic burden-average energy cost multiplied by subsistence threshold divided by per capita income (or expenditure) Nonmonetary Exposure rates-24-hour exposure rates to indoor air pollutants such as particulates (hours of reading by schoolchildren) Incidence rates-proportion of poor households It is difficult to isolate the impact of affected by energy-related incidents of ill health, energy sector interventions on incidence such as respiratory illness, burns, and paraffin rates, which may be affected by many poisoning (grade completion rates among poor other factors households) * Highlights the most important-and easily calculated-of the indicators Source Authors 137 VIVIEN FOSTER AND JEAN-PHILIPPE TRE by dividing the population into income (or consumption) quintiles (or deciles). Separate welfare indicators can then be calculated for each quintile (or decile). Furthermore, this approach has the additional advantage that it allows an assessment of the equity of interventions in the energy sector by examining how benefits are distributed across income groups. The analytical tools for measuring inequality are already well developed in the context of the income distribution literature (Cowell 1995). Standard measures such as the Cini coefficient can be readily adapted to the energy context, giving rise to concentration coefficients that measure the extent to which the distribution of services departs from an egalitarian benchmark (Kakwani 1986). Although already widely used in the analysis of public expenditure programs, attempts to apply these analytical tools to the energy sector have been relatively rare. Box 4.4 provides an interesting exception, showing how concentration coefficients can be used to analyze the distrib- utional incidence of access to and use of service subsidies. 4.5 Implementation Issues While conceptually straightforward, many of the indicators presented are relatively data-intensive to compute. The availability of suitable data from existing sources and the cost of gathering additional data are likely to be the major constraints when applying this approach to assess the impact of energy sector interventions on the welfare of the poor. This section outlines the requirements of the ideal data source for this type of exercise; however, it also suggests a number of shortcuts that may permit some approximation to the indicators under less than ideal circumstances, that is, those that typically confront the decisionmaker. The ideal dataset for assessing the poverty impact of energy sector reforms would have the following characteristics (G6mez-Lobo, Foster, and Halpern 1999). First, it would combine information on energy-related behavior with information on overall econoniic welfare. Second, it would record such information both immediately before and some time after the energy sector intervention for the same households. Third, it would contain information for households that had been affected by the intervention, as well as for a control group that had not been affected. Each of these issues is now discussed in turn. 138 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR Box 4.4 Inequality Analysis of Electricity Connections in Colombia Velez (1995) provides an interestmig example of the application of tnequality analysis to electricity connectiorns in Colombia. The study looked at the change Lf electricity connlection rates by income quinttle betiveeni 1974 anud 1992. The concenitratiotn coefjicients (CC)for these two jears indicate that the distributtion of electricity connections weiitfrorn regressive (CC = 0 157>0) to virtually egalitartain (CC = 0.0.34 =0). The reason for this is thltt the new connections made during the intervening period were somewhat skewved toward lower-income hunseholds, (is indicated by the slightly negative CC of -0.031 shown in thle table. Electricty coverage coverage increment 1974-92 Income rates (A) No of new Share of new quintile 1974 1992 connections connections 1 414 81 3 869,000 0 202 2 491 90 4 943,000 0 219 3 617 93 4 897,000 0 208 4 73 5 96 0 849,000 0 197 S 91.3 980 750,000 0 174 Concentraton coeffwient (CC) +0157 +0 034 -0 031 Colombia has a relatively cormiplex system of cross-subsidies int electricity pricing based on the characteristics of each neighborhood. Ve'lez also ana- lyzed the iriclhenice of these cross-subsidies across tncome quintiles alldJouild a slightly progressive pattern, indicated by a CC of'-0. 033. The pattern oJ incidence ts furthet broken down betwveen. legal subsidies (those accruing to legittiiiate paying customers via the official tariff structure) and illegal sub- sidies (those accruing implicitly to householdks wi th nonpaying clandestinie connections). This di aggregated analysis revealed that illegal subsidies are much niore progressive than legal ones, with a CC of -0.301 versus -0.016. Spannin,g the Full Range of Data Requirements The dataset should contain comprehensive information bothi about house- holds' energy-related decisions (requirecl to calculate the welfare indica- tors) and about the poverty indicators required to examine the welfare impact on the poor. Table 4.2 identifies the types of inforimiation required 139 VIVIEN FOSTER AND JEAN-PHILIPPE TRt TABLE 4.2. Data Required to Calculate Indicators by Potential Source Data sources Engineering Price Household Electric Special Indicator estimates surveys surveys utilities surveys Efficiency Efficiency Unit cost by Household factor factors by fuel fuel expenditure by fuel Coverage Household Household index access by fuel access by fuel Reliability Reliability of Reliability of index access by fuel access by fuel Concentration Efficiency Unit cost by Household index factors by fuel fuel expenditure by fuel Affordability Subsistence Unit cost by Subsistence index threshold, fuel threshold efficiency factors by fuel Average Efficiency Unit cost by Household energy cost factors by fuel fuel expenditure by fuel Average Efficiency Unit subsidy Household Unit subsidy energy factors by fuel by fuel expenditure by by fuel subsidy fuel Average Capital cost of Unit cost by Capital cost of Capital cost capital cost household fuel household of household energy use energy use energy use Average total Capital cost of Unit cost by Capital cost of Capital cost cost household fuel household energy of household energy use, use; household energy use efficiency expenditure by factors by fuel fuel Capital Capital cost of Unit cost by Capital cost of Capital cost intensity household fuel household energy of household energy use, use; household energy use efficiency expenditure by factors by fuel fuel Economic Subsistence Unit cost by Per capita Subsistence burden threshold, fuel subsistence threshold efficiency threshold, factors by fuel household expenditure by fuel Poverty Total household measure income or expenditure Source Authors 140 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR to calculate each of the proposed indicators, and also shows the most likely sources for this information. Close inspection of the table reveals that only 10 basic pieces of information are required to calculate all the indicators, some of which are parametric in nature (subsistence thresholds and unit costs), and so can often be (lerived from beyond the survey data. Perhaps the most critical and ubiquitous input into the calculation of these indicators is the physical measure of consumption for each of the fuels that appear in the household energy portfolio. This information, which is rarely available in direct form, can generally be inferred from data about household expenditure on different fuels by applying unit prices and effi- ciency factors to derive implicit levels of gross and net consumption. One important limitation of this approach is that it does not capture the con- sumption of traditional biomass fuels that the household may gather directly at no pecuniary cost, although some surveys require households to estimate the value of the fuels they collect or to provide information about the time used for fuel collection. Table 4.2 shows that the most important source of information will be household surveys, such as the Living Standards Measurement Study (LSMS) surveys or general income and expenditure surveys.4 Household surveys are particularly valuable because they combine information about energy use with information about overall household income and expendi- ture, from which absolute or relative indicators of poverty can be derived. In many instances household surveys complemented by external price and engineering parameters will be adequate to undertake the proposed analy- sis. This is precisely the case for the Guatemala illustration developed in the following section. For indicators of access, special surveys may well be required, because household surveys typically consider only access to electricity. In some cases, however, "piggy backing" onto an existing household survey may be possible at relatively low marginal cost by incorporating additional questions on energy consumption habits to fill in the missing pieces of information. Although there is an increasing trend toward conducting household sur- veys tlhat record the detailed information on expenditure patterns necessary to conduct this type of analysis, a significant number of countries still lack such information In these cases information on energy expenditures would have to be obtained directly from a special sector survey. Other countries may not even have reliable information on economic measures of poverty. 141 VIVIEN FOSTER AND JEAN-PHILIPPE TRE An alternative to economic measures of poverty that is sometimes available is the so-called poverty map, which classifies geographical areas as poor or not poor according to an index of economic and/or noneconomic poverty indicators. Where these are available, the calculation of poverty impact indicators for the energy sector can be undertaken for a sample of house- holds from those geographical areas classified as poor by the poverty map. Obtaining Data before and after the Interventiton As noted, the ideal dataset should contain data for the same households both immediately before and some time after the intervention in the energy sector. One of the main limitations of relying on existing household surveys is that their timing is unlikely to coincide exactly with the timing of the intervention. In some cases using a past household survey as the baseline for impact measurement may be possible, and then repeating only the relevant sections of the survey on a subset of the original sample at some suitable time after the intervention. Even where the timing is fortuitous, longitudinal surveys (surveys that follow the same households over time) are still extremely rare in develop- ing countries, so that observing the same household before and after the intervention is usually not possible. In practice a number of statistical techniques can be used to control for differences between households in the pre- and postintervention samples ranging from matched pairs to mul- tiple regression models (see Baker 1999 for a detailed discussion). Obtaining Data on 7reatment and Control Groups Finally, the dataset should contain information on both households affected by the intervention and a control set of similar households that were not affected. The importance of this is that it makes it possible to be sure that the impacts observed as a result of the intervention are not attributable either to (a) differences in the pre- and postintervention samples or (b) extraneous influences on energy consumption behavior unrelated to the intervention itself (Baker 1999). One possible way of doing this is to compare different regions of a coun- try, some of which have been affected by the intervention and others that have not been affected. However, where the intervention in the energy sec- 142 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR tor is national in its reach, as is often the case, this option is no longer available. Moreover, constructing such a control on the basis of interna- tional comparators is likely to raise as many problems as it resolves. To attenuate the problem of devising an adequate control, the indicators presented in this chapter have tended to focus on outcomes directly linked to energy sector parameters, such as consumption decisions, and to avoid indicators related to general levels of poverty, which may be sensitive to a wide range of decisions. Nevertheless, this problem is particularly intrac- table and almost impossible to resolve entirely. 4.6 Illustrationfrom Guatemala The purpose of this section is to dlemonstrate the relative ease with which the proposed indicators can be calculated using fairly standard and widely available householdl survey data. The disctIssion is limited to the basic needs and monetaiy welfare indicators, because the information required to calculate the nonmonetary measures of the impact on health and educa- tion was not available. Given the limited tradition of household surveys in Guatemala, we coulcl only calculate the indlicators for a single year, 1998/99. The absence of two points in time across which to compare the performance of the indicators to some extent limits the illustration of the proposed methodology. Never- theless, the exercise of calculating them at one given point in time is help- ful to demonstrate the practical steps involved and to show how the indica- tors provide a useful diagnostic of the starting point for any subsequenL intervention. Furthermore, since a major reform of the electricity sector took place during the survey period, the indicators could be regarded as baseline values for a potential future assessment of the impact of electricity reform on poverty in Guatemala. Methodology The exercise is based on a household income and expenditure survey for Guatemala known as the Encuesta Nacional de Ingresos y Gastos Famil- iares or ENIGFAM (national household income and expenditure survey) covering the period April 1998 through March 1999. While ENIGFAM 143 VIVIEN FOSTER AND JEAN-PHILIPPE TRt shares certain characteristics of the LSMS surveys, the range of information available is not as complete. Consequently the same exercise should, in prin- ciple, be more straightforward for countries for which LSMS data are avail- able. The particular value of ENIGFAM for the task at hand is that it com- bines information on poverty with information on household energy decisions. ENIGFAM provides a suitable empirical basis for establishing the rela- tive poverty of different households in Guatemala (Foster and others 2000). Specifically, the general household expenditure data contained in the sur- vey can be used to derive a consumption aggregate. The consumption aggregate is a measure of the economic welfare of the household that reflects its actual use of resources over the reference period of a year. The consumption aggregate incorporates expenditures on food, clothing, house- hold items, transport, leisure, health, education, housing, and consumer durables. Housing consumption is captured in terms of the rental value of the home, while durables are amortized so as to capture only that portion of the goocd consumed in the current year. I'he consumption aggregate pro- vides a basis for sorting households into deciles ranging from the poorest IO percent of the population to the richest 10 percent of the population. ENIGFAM also contains a wealth of information on household energy behavior. Specifically, it contains data on monthly household expenditure on energy broken down by batteries, candles, electricity, fuelwood, kerosene, and propane gas, and clarifies which fuels are used for cooking and which for lighting. In addition to data on fuel expenditures, ENIGFAM provides an inventory of household energy-related durables, together with information on the purchase costs of these appliances. As mentioned earlier, a potential practical problem with fuelwood is the absence of any recorded monetary expendlture when this is collected directly by the household. Fortunately ENIGFAM gets around this problem by requir- ing households that gather their own wood to provide an estimate of the cost of purchasing an equivalent amount of fuelwood in the marketplace. To convert data on fuel expenditures into physical units of consumption, data on energy prices were required. As ENIGFAM did not incorporate a price survey, we took regional unit prices directly from the Guatemalan consumer price index. Dividing unit prices into expenditures yields an estimate of the physical consumption for each energy source; however, as each fuel is measured in different units, we had to convert them all into a common unit of kilowatt-hours (kWh) for the purposes of aggregation. We 144 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR did this by applying standard conversion factors (Unite(d Nations 1987). The standlarclized measures of physical energy use were summed together to obtain an estimate of total gross energy consumption at the household level. Cross household energy consumptioni can, however, be a misleading figure. Different cooking, fuels differ dramatically in terms of the efficiency with which they generate heat from a given kWh of fuel input, wvhile different lighting fuels vaiy tremendously in terimis of their luminous efficacy, that is, the amount of light they put out from a given kWh of fuel input. Furthermore, a particular fuel may exhibit a greater or lesser degree of efficiency (or luminous efficacy) depending on the complementary capital good with which it is used. For exam- ple, kerosene procluces eight times more luminosity when burnt in a hurricane lamp than in a plimitive wick lamp (Van der Plas and De Graaff 1988). How- ever, comparisons of gross energy consumption between households do not give a clear picture of households' level of energy comfort, because they fail to take into account the dlifferential peiformance of alternative energy sources. 'ro get around this problem, we acljusted gross energy consumption fig- ures to reflect the efficiency of each fuel relative to the efficiency of elec- tricity in the same use, that is, the efficiency factors reported relatedl to rel- ative efficiency rather than absolute efficiency (table 4.3). Consultation wvith energy experts in Guatemala revealed that open fires were the most widely used technology for burning wvood, whille primitive wick lamps were the most widely used technology for lighting using kerosene. The efficiency factors show that propane gas is 77 percent as efficient as electricity for TABLE 4.3. Relative Efficiency and Luminous Efficacy Factors Used to Adjust from Gross to Net Energy Consumption Cooking Lighting Appliances Relative Relative Relative Fuel efficiency Fuel luminous efficacy Fuel efficiency Electricity 1 00 Electricity 1 00 Electricity 1 00 Propane 0 77 Kerosene 0 01 Batteries 0 90 Fuelwood 0 15 Candles 0 02 Car baneries 0 90 Note Electricity provides the baseline against which the efficiency and luminous efficacy of other fuels are expressed, hence the factor for electricity is always 1 Sources Leach and Gowen (1987), Van der Plas and De Graaff (1988) 145 VIVIEN FOSTER AND JEAN-PHILIPPE TRE cooking, whereas wood is only 15 percent as efficient as electricity. For lighting, the divergence is far greater. Kerosene wick lamps produce only 1 percent as much luminosity as electricity, whereas for candles the value is a mere 2 percent. As regards appliances, batteries are believed to be about 90 percent as efficient as electricity. Multiplying gross energy consumption by the efficiency factors yields an estimate of net energy consumption in kWh. As the efficiency factors have been normalized against electricity, this value can be interpreted as the total amount of electricity that the household would have to consume to obtain its current level of energy services exclusively from electricity. Thus the net energy consumption figure gives a far more accurate impression of the house- hold's level of welfare in terms of the output of energy services. It also makes comparisons of energy consumption between households far more meaningful. A comparison of energy expenditure together with gross and net energy consumption across deciles reveals some interesting conclusions (table 4.4).'5 On average, households in Guatemala spend US$238 on energy per year, obtaining 13,090 gross kWh, which translates into 3,236 net kWh. TABLE 4.4. Comparison of Energy Expenditures with Gross and Net Consumption across Deciles, 1998/99 Energy Gross N1et Efficiency expenditure consumption consumption factor Deciles (US$/year) (kWh/year) (kWh/year) (INeVgross kWh) 1 120 10,060 1,658 0 16 2 161 13,320 2,217 0 17 3 194 15,506 2,692 0 17 4 190 13,799 2,607 0 19 5 229 16,216 3,173 0 20 6 219 14,160 3,055 0 22 7 254 13,627 3,516 0 26 8 279 13,387 3,798 0 28 9 301 11,448 4,078 0 36 10 434 9,399 5,539 0 59 Total sample 238 13,090 3,236 0 25 Decile 101 3 61 0.93 3 34 3 58 Source ENIGFAM data 146 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR This represents an average efficiency factor (ratio of net to gross energy consumption) of 25 percent. The level of household expenditure on energy rises steadily through the deciles, with the 10th decile spending 3.61 times as much as the first. However, it is striking that gross energy consumption peaks at aroundl the fifth decile. Indeed, the gross consumption of the 10th decile is actually slightly lower than that of the 1st decile. When expressed in net terms, con- sumption rises steadily through the deciles from 1,658 to 5,539 kWh per year. In other word, the top decile consumes more than three times as much net energy as the bottom decile. The efficiency factor, or the ratio of net to gross energy consumption, also rises dlramatically from 16 percent in the J sl decile to 59 percent in the 10th decile as households switch to more efficient forms of energy, such as propane gas and electricity. It is precisely this interplay of rising energy demand and increasing energy efficiency that explains the bell-shaped pattern of the gross con- sumption curve (figure 4.1). On the one hand, as households become richer they demand more energy, as indicated by the rising net consumption curve. On the other hand, as householdls become richer they switch to more efficient fuels, as indicated by the rising efficiency factor. Over the first three deciles energy demand rises more rapidly than energy efficiency, and so gross energy consumption increases. Over the middle dleciles rising energy demandl keeps pace witli improving energy efficiency, and so gross energy consumption is relatively flat. Finally, over the last three (leciles efficiency increases more rapidly than energy demand, and thus the gross consumption curve starts to fall back. It is interestinc to see how the efficiency factor affects the relative costs of different types of fuels. Table 4.5 compares the original average prices per gross kWh for each fuel derivedl from the consumer price index with the average prices per net kWh.6 For cooking, the net prices show that fuel- wood, far from being the cheapest source of energy, is actually just as costly as propane gas; however, both fuels remain slightly cheaper than electricity, even in net terms. For lighting, the conversion from gross to net terms dra- matically wipes out any apparent cost advantage of kerosene. Canclies remain by far the most expensive source of lighting, wvhether in gross or net terms. For appliances, the gross to net conversion is irrelevant, because all alternatives are based on electricity; however, the figures show that batteries are substantially more expensive per kWh than mains electricity. 147 VIVIEN FOSTER AND JEAN-PHILIPPE TRP FIGURE 4.1. Gross and Net Energy Consumption and Efficiency Factors by Decile, 1998/99 Index relative to first decile 400 - 3 50 - Gross energy consumption - - - - Net energy consumption 3 00 . ... Efficiency .7 2 50- 1 50- 1 00 -- 0 50 0 00- I 000- 1 1 2 1 3 4 5 6 7 8 9 1 0 Decile Source ENIGFAM data TABLE 4.5. Gross and Net Unit Prices for Different Fuels, 1998/99 Cooking Lighting Appliances Fuel Gross Net Fuel Gross Net Fuel Gross Net Electricity 0 08 0 08 Elearicity 0 08 0 08 Electncty 0 08 0 08 Propane 0 05 0 06 Kerosene 0 05 5 87 Batteries' 0 59 0 53 Fuelwood 0 01 0 06 Candles 0 26 13 00 Car battenesb 2 57 2.31 a The unit price is based on the assumption that the batteries are used to power a 16-watt radio b The unit price is based on the assumption that the batteries are used to power a 16-watt black and white television set Sources Leach and Gowen (1987), Van der Plas and De Graaff (1988), consumer price index for 1998 148 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR An important conclusion of this analysis is that the technologies used predominantly by the poor for lighting and powering appliances are orders of magnitudle more expensive per kMh than electricity, suggesting tlhat poor householdls could realize considerabile savings in energy costs by switch- ing to electricity. Bastc Needs Inidicators This section derives the basic needs indicators for the Guatemala data. Calculating the coverage index is somewhat problematic insofar as the household survey asks people about what fuels they currently use, but does not inquire what other fuels they might potentially have access to. Conse- quently actual use must be taken as a proxy for access. This is unfortunate in the sense that it makes it difficult to undlerstand why particular house- holds fail to make use of specific fuels, specifically, whether this is due to availability constraints or to other factors such as affordability. The coverage of different energy sources varies substantially across dleciles (table 4.6). On average, each Cuatemalan household uses 2.6 different types TABLE 4.6. Coverage Index and Total Number of Available Energy Sources by Decile, 1998/99 Car Deciles Electricity Propane Fuelwood Kerosene Candies Batteries batteries Total 1 033 002 090 047 043 034 001 250 2 041 003 094 046 047 035 004 270 3 041 011 088 041 049 045 006 280 4 055 018 081 037 046 033 006 276 5 0 63 0 29 0 78 0 25 0 50 0 27 0 09 2 81 6 066 035 070 027 047 029 009 283 7 074 056 058 018 033 019 007 265 8 080 064 051 017 033 016 007 269 9 090 072 040 006 025 009 006 247 10 095 078 015 004 018 004 002 216 Total sample 0 64 0 37 0 67 0 27 0 39 0 25 0 06 2 64 Source ENIGFAM data 149 VIVIEN FOSTER AND JEAN-PHILIPPE TRt of fuels. The total number used is relatively constant across deciles, although it peaks in the middle deciles. Fuelwood and propane gas are the two major fuels used for cooking. The use of fuelwood is virtually universal among the lowest deciles and remains dominant as far as the sixth decile (figure 4.2). Thereafter propane gas takes over as the main cooking fuel. The use of electricity for cooking is extremely rare and is confined to the 10th decile. Given that the net costs for fuelwood and propane are roughly the same (table 4.5), the greater use of wood among poorer households cannot be attributed to any fuel cost advantage. A number of possible explanations account for this. One is that because of distribution problems propane gas may not always be available to poor households, more than 90 percent of which live in the rural areas (Foster and others 2000). Even though ENIG- FIGURE 4.2. Proportion of Households Using Different Fuels for Cooking by Decile. 1998/99 Percentage of households using the fuel for cooking 100 - 90 - 80 - \ ,, 70 - 60 - 50 - Fuelwood - - - - Propane 40 . ........ Other , ' \ 30- 20 .- , 10 - - .................................... .................... -- -... 1 1 2 1 3 1 4 1 5 1 6 T 7 T 8 1 9 1 10 Decile Source ENIGFAM data 150 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR FAM does not provide information on whether houselholds have access to fuels that they choose not to use, we can make some inferences by compar- ing the use of propane gas between urban and rural areas among house- holds in the higher-income deciles, which are uikely to face any finan- cial constraints in purchasing fuel. Figure 4.3 shows that the coverage of propane gas in rural areas lags behind that in urban areas by about 40 to 50 percent among the higher-income deciles. This suggests that limited availability of propane gas is an issue in rural areas. Second, propane is actually more expensive to use than wood. This IS because even though the net fuel costs are virtually identical, propane gas requires an additional investment of US$205 for a propane stove, whereas wood can be burned at zero capital cost using a primitive three-stone stove. The investment required for a propane stove is large relative to the average total monthly expenditure of US$240 for households in the first quintile. Fur-therimiore, propane is sold in relatively large quantities-35-pound cylinders costing around US$9 each-compared with fuelwood, which houselholds can gather in the amount requiied often without any monetary FIGURE 4.3. Coverage of Propane Gas across Urban and Rural Areas by Decile, 1998/99 Percentage of households 100- 90- 80- Urban - - - -Rural 70- 60- 50- 40- 30 - / , - - - 20 - 10- 0 1 2 3 4 5 6 7 8 9 10 Decile Source ENIGFAM data 151 VIVIEN FOSTER AND JEAN-PHILIPPE TRE outlay. Consequently, budgeting to buy propane cylinders may be more difficult for poor households than richer households. Once again figure 4.3 provides some support for the view.that affordability of propane is an issue among lower-income households. This can be inferred because in urban areas, where the availability of the fuel does not appear to be a major problem, coverage of propane among urban households in the first quintile is almost as low as for rural households in the same quintile. A similar phenomenon can be observed for electricity (figure 4.4), although the differences between high- and low-income households are much less pronounced. The two most important lighting technologies are electricity and simple wick kerosene lamps. The use of kerosene (and of other alternatives such as candles) is confined to the lowest deciles, declining sharply thereafter (figure 4.5). Indeed, even in the first decile electricity just dominates kerosene as the main source for lighting. Many poorer households own audiovisual appliances such as radios, cas- sette players, and televisions. Where the household lacks an electricity FIGURE 4.4. Coverage of Electricity across Urban and Rural Areas by Decile, 1998/99 Coverage rate for electricity 100 90 Ura 90- - - - - Rural ~ ~ 80 - 70 - 60 - 50 -- 40 - , ,," 30 - 20 - 10 0 1 2 3 4 5 6 7 8 9 10 Decile Source ENIGFAM data 152 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR FIGURE 4.5. Proportion of Households Using Different Household Fuels for Lighting by Decile, 1998/99 Percentage of households using the fuel for lighting 100 - 90 - Electty - - - - Kerosene 80 - ........ Candles 70------Other 70- 60 - 50 - 40- 30 - 20 __ 10 - 11 2 3 1 4 1 5 1 6 7 8 9 10 I Decile Source ENIGFAM data connection, these appliances are powered by batteries or, In the case of tel- evision, by car batteries. The data show that batteries are the most wide- spread source of energy for audiovisual appliances until the fourth decile (figure 4.6). A striking result is that the switch to electricity for use for lighting and appliances takes place at a much lower point in the income dlstribution (at around the second decile) than the switch from fuelwood to propane gas for use in cooking (at around the sixth decile). Thus many households in deciles three to five combine the use of a modem fuel such as electricity for one set of activities with a traditional fuel such as woodl for cooking. This could reflect a number of factors. First, electricity may be niore widely available than propane gas. Second, whereas a switch to propane gas for cooking does not entail any cost savings, a switch to electricity can potentially result in 153 VIVIEN FOSTER AND JEAN-PHILIPPE TRE FIGURE 4.6. Proportion of Households Using Different Household Fuels for Audiovisual Appliances by Decile, 1998/99 Percentage of households using the energy source for audiovisual appliances 100 90 Electricity - - - - Batteries 80 70 60- 50 40 30~~~~~~~~~~- - ' _ __ 30 20 10 - 0 1 2 3 4 5 6 7 8 9 10 Decile Source ENIGFAM data quite considerable cost savings given the high unit costs of kerosene lamps and batteries. For example, a household running the equivalent of one 60- watt light bulb and one 16-watt radio for four hours a day can reduce its daily energy bill from US$1.44 to US$0.02 by switching to electricity. As ENIGFAM does not collect information about reliability of access, cal- culating the reliability index using this dataset is not possible. However, the concentration index-which measures the extent to which each household's energy portfolio is spread across different fuels-can be readily calculated using the share of each fuel source in the household's total net energy con- sumption. The index takes a value of 0.89 in the 1st decile, falling to 0.65 in the 10th decile, with an overall average of 0.74 (figure 4.7). The higher values of the concentration index for the poorer households demonstrate 154 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR FIGURE 4.7. Concentration Index by Decile, 1998/99 Concentration index 1 00 - 090 - 0.80 0 70- 0.60- 0.50 - 0 40- 0 30- 0 20- 0.10 1 2 3 4 5 6 7 8 9 10 Decile Source ENIGFAM data their heavy dependence on fuelwood, which provides more than 80 percent of their net energy consumption (figure 4.8). By contrast, richer households have their energy portfolios more evenly spread among electricity, propane gas, and fuelwood. Finally, the relative importance of cooking versus other uses in absorb- ing households' overall energy resources is also interesting. On average, 69 percent of net energy consumption is used for cooking and the remain- ing 31 percent for lighting and energy-related appliances. The share of these two uses varies dramatically across deciles (figure 4.9). Households in the Ist decile use more than 90 percent of their net energy resources for cooking, a fraction that falls steadily to about 30 percent of net energy resources for households in the 10th decile. Only households in the 10th decile dedicate less than half of their net energy resources to cooking. 155 VIVIEN FOSTER AND JEAN-PHILIPPE TRE FIGURE 4.8. Shares of Net Household Energy Consumption by Fuel, 1998/99 Percent 100 7 90 I 80 70- 60- 50- 40- 30- 20- 10 b 0 1 2 3 4 5 6 7 8 9 10 Decile C] Other g Electricity D2 Propane gas D Fuelwood Source ENIGFAM data Monetary Indicators This section derives the monetary welfare indicators from the Guatemala data. These include the measures of affordability and economic burden, as well as the various average cost measures. A first step is to estimate the household energy subsistence threshold to permit the calculation of the affordability index. We used two different meth- ods. The first approach was to examine the average total net energy con- 156 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR FIGURE 4.9. Shares of Net Household Energy Consumption by Use, 1998/99 Percent 100- 90 - 80 - 70 - 60 - 50 - 40- 30 - _ _ 20 - ___ 20 …-~~ - - - -Cooking 10 - - - - - Lighting and appliances 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 1 l0 Decile Source ENIGFAM data sumption of those households whose overall consumption levels fell within plus or minus 10 percent of the US$1 (purchasing power parity adjusted) extreme poverty line. This gives a subsistence threshold of 2,125 kWh per year, equivalent to 5.8 kWh per day. The seconcl approach was to define a basic set of energy needs in consultation with energy experts in Guatemala. This led to an estimate of 2,154 kWh per year, equivalent to 5 9 kWh per day, basecl on local knowledge of energy consumption patterns among low- income rural households.' Given the similarity between the two estimates, we adoptedl the first value-derived directly from the household survey dataset-as the subsistence threshold. Using this subsistence threshold of 2,125 kWh per household per year, we can calculate the affordability index and affordability ratio clefined ear- liei, as well as the measure of economic burden (table 4.7). All three of the new indicators suggest that energy affordability problems are confined to households in the first quintile and are particularly acute among house- holds in the first decile. Only a quarter of first decile households consume above the subsistence threshold, and on average they consume only 78 per- cent of the subsistence amount. The economic burden measure indicates 157 VIVIEN FOSTER AND JEAN-PHILIPPE TRE TABLE 4.7. Measures of Energy Affordability, 1998-99 Subsistence Expenditure threshold-based measures share-based measures Deciles Affordability index Affordability ratio Economic burden Expenditure share 1 0 24 078 079 013 2 047 1 04 0 19 0 11 3 0 50 1.27 0.10 011 4 053 1 23 013 009 5 0 60 1 49 0 06 0 09 6 0.59 1.44 0 05 0 08 7 0 70 1.65 0 07 0.08 8 0 74 1.79 0.04 0.07 9 0 83 1 92 0 02 0 06 10 094 261 0.01 005 Total sample 0.61 1.52 0.15 0 09 Source ENIGFAM data that these households would have to devote 79 percent of their budget to energy in order to purchase the subsistence threshold. The situation is sub- stantially better for households in the second decile, which on average con- sume around the subsistence threshold, with the economic burden of this expenditure falling to 19 percent of the household budget. These conclusions can be compared with those derived from the more traditional measure of energy expenditure as a proportion of total house- hold expenditure. It is striking that the profile of the expenditure share indicator over deciles is considerably flatter than that of the economic bur- den measure. Based on expenditure share, households in the first quintile do not look that much worse off than other households in the bottom half of the distribution; however, this measure conceals the fact that these house- holds keep their expenditure down by curtailing their energy consumption. A comparison of these affordability indexes for households with and with- out access to modern fuels is interesting. In the case of electricity, the results are particularly dramatic (table 4.8). The economic burden for households in the first decile falls from 110 percent for households without an electricity connection to 14 percent for households with an electricity 158 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR TABLE 4.8. Comparison of Economic Burdens of Households with and without Access to Electricity, 1998-99 With electricity Without electricity Deciles Economic burden Expenditure share Economic burden Expenditure share I 014 015 1 10 013 2 009 013 026 010 3 007 012 012 010 4 006 0 10 022 008 5 0.05 0 10 0 07 0 09 6 004 008 006 007 7 003 008 018 007 8 003 007 007 006 9 002 006 003 005 10 001 005 004 004 Total sample 0 04 0 09 0 33 0 09 Source ENIGFAM data connection, but this effect hardly shows up at all in the expenditure share measure, which is virtually the same for both groups of households. This result indicates that electrification can have a powerful impact in making energy affordable to low-income households, and that once households are electrified, affordability is no longer a serious issue. In the case of propane gas, the results are similar, but less dramatic (table 4.9). The economic bur- den for first decile households with access to propane falls from 80 to 25 percent, whereas the expenditure share actually rises from 13 percent to 22 percent. Moving on to average cost indicators, table 4.10 presents average fuel cost, average capital cost, and average total costs in gross and net terms by decile. The calculation of capital costs merits a word of explanation (Foster and oth- ers 2000). ENIGFAM provides information on the purchase costs of new durables for those who bought them during the survey period and an inven- tory of the durables owned by each household regardless of when they were purchased.8 The replacement cost of durables is assumed to be equal to the average purchase cost. As the survey cloes not provicle any information about the life of durables, the depreciation periods stipulated in Guatemalan tax leg- 159 VIVIEN FOSTER AND JEAN-PHILIPPE TRE TABLE 4.9. Comparison of Economic Burdens of Households with and without Access to Propane Gas, 1998-99 With propane gas Without propane gas Deciles Economic burden Expenditure share Economic burden Expenditure share 1 0.25 0 22 0 80 0 13 2 009 015 0.19 0 11 3 007 0.13 0.10 011 4 0 07 0.12 0 14 0 09 5 005 009 006 0 10 6 0 04 0 09 0.05 0 08 7 008 008 006 008 8 004 0.07 0.04 007 9 0 02 0 06 0 02 0.06 10 001 005 002 005 Total 0 04 0.07 0 21 010 Source ENIGFAM data TABLE 4.10. Comparison of Gross and Net Energy Prices and Efficiency Factors across Deciles, 1998-99 Average fuel cost Average capital cost Average total cost Deciles Gross Net Gross Ilet Gross IMet 1 0002 008 000 001 002 0 10 2 001 0.08 000 002 002 0 10 3 002 008 001 002 003 010 4 002 008 001 004 004 011 5 0 02 0 08 0 02 0.04 0 04 0 12 6 003 007 0.02 004 005 0 12 7 0.04 007 003 005 007 0.12 8 004 0.08 0.03 0.05 008 0.13 9 005 008 004 005 0.09 0 13 10 007 008 006 007 013 0.15 Total 0 03 0 08 0 02 0 04 0.06 012 Decile 101 3 50 0 98 21 00 5.33 6 21 1 59 Source ENIGFAM data 160 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR islation are used to annualize the capital cost of the durable goods. In addi- tion, the capital cost measure incorporates expenditures on light bulbs. The average fuel cost figures are derived by dividing total fuel expendi- ture by total (gross or net) consumption. The average gross fuel cost figures show that the poor purchase fuel that appears to be comparatively cheap in gross terms. Indeed, the 1st decile pays only US$0.02 per kWh, compared with US$0.07 per kWh for the 10th decile. However, the average net price is remarkably constant across deciles, averaging US$0.08 per kWh. The reason is that the 10th decile buys fuel that is 3.5 times as expensive in gross terms as that purchased by the 1st decile, but also happens to be 3.5 times more efficient; hence the gross price differential almost exactly off- sets the efficiency differential. The average capital cost figures come from dividing the amortized capital costs of all the consumer durables by total (gross or net) consumption. As might be expected, capital costs rise steadily by decile. Indeed, the 1 st decile devotes only 14 percent of its total energy-related expenditure to capital goods, whereas the 10th clecile devotes 44 percent (table 4.] 1). Putting fuel and capital expenditures together, the gross differential in average total costs TABLE 4.11. Relative Value of Fuel and Capital Expenditures across Deciles, 1998/99 Absolute expenditure Expenditure share Deciles Fuel Capital Total Fuel Capital Total 1 782 130 911 086 0 14 100 2 1,048 226 1,274 082 018 100 3 1,259 291 1,550 081 0 19 100 4 1,234 435 1,668 0 74 0 26 100 5 1,487 541 2,028 0 73 0 27 100 6 1,423 594 2,017 0 71 0 29 100 7 1,649 810 2,459 067 033 100 8 1,813 960 2,773 065 035 100 9 1,957 1,172 3,130 063 037 100 10 2,824 2,208 5,032 056 044 1 00 Total sample 1,549 737 2,286 0 68 0 32 1 00 Decile 10 1 3 61 17 03 5 52 0 65 3 09 100 Source ENIGFAM data 161 VIVIEN FOSTER AND JEAN-PHILIPPE TRt between the poorest and richest households increases to a factor of more than sixfold. In net terms the differential reduces to only 50 percent. Calculating the average subsidy indicator was unnecessary for Guatemala, because none of the fuels considered are subsidized by the government. Electricity had been subsidized for many years, but in the run-up to privatization in 1998 all histoncal subsidies were phased out. Summary and Implications On the basis of the foregoing exercise we can conclude that many of the indicators proposed in this chapter are relatively straightforward to calcu- late using information that is typically available in household surveys. In particular, the set of indicators relating to monetary measures of well-being was straightforvard to calculate. Those relating to access presented greater problems because of the absence of information on the availability and reli- ability of fuels that the household chooses to use. Information relating to nonmonetary impacts such as health could not be derived from standard household survey data. We argued at the outset that the value of calculating indicators such as those proposed in this chapter is not only as a baseline against which to measure future progress, but as a diagnostic tool that may help improve the poverty focus of energy sector interventions at the design stage. In the Guatemala case, the absence of data for two points in time prevents an illustration of the comparison of indicators before and after a sector inter- vention; however, it is possible to show how the analytical exercise per- formed here can help shape the nature of policies toward the sector. Consider the main findings on access to alternative energy sources. An important deficit in electricity coverage is clearly apparent: electricity reaches less than half the households in the first quintile. Moreover, given the sub- stantial net price differential between electricity and alternative fuels for light- ing and powering appliances, the potential cost savings from electrification are extremely high. Our estimates indicate that with electrification, the costs of meeting a basic set of household requirements would fall by a factor of 70. Our results also provide indirect evidence of a considerable deficit in coverage of propane gas in rural areas; however, the case for promoting access to propane is not as clear. The net energy costs of cooking with propane are no lower than those of cooking with fuelwood. In addition, 162 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR propane cooking may be financially problematic for poor households both because of the necessary capital investment in a propane stove and because of relatively infrequent but expensive purchases of propane cylinders. A case for increasing the coverage of propane might nonetheless be made either in terms of the time savings entailed by the avoidance of gathering fuelwood, or in terms of the environmental and health gains of avoiding deforestation and the emissions associated with burning wood. However, any attempts to increase propane coverage would need to address the financing problems on the demand side as well as any distributional bottlenecks on the supply side. The complete absence of energy subsidies in Guatemala means that rela- tive prices are not distorted either in favor of or in prejudice to the poor. Indeed, despite the relatively expensive technologies that the poor may be forced to use for lighting and for powering appliances, the average net energy cost is similar across deciles. These findings suggest that the pricing of fuels seems to be on a sound basis, with no apparent case for a policy change. As regards subsidizing the use of energy, the indicators suggest that serious problems of affordability are confined only to households in the first decile, but even households in the first decile do not face affordability problems when they are connected to the electricity grid. This suggests that electrification is a more appropriate policy response than subsidizing the use of electricity substitutes. 4.7 Wider Applicability Thus far the discussion has been couched purely in terms of the energy sector; however, the need to measure the impact of policy interventions on the poor is one that cuts across all the infrastructure sectors: energy, water, telecommunications, and transport. The methodology we have proposed is of wider relevance and could fairly readily be adapted to other infrastruc- ture sectors. This section briefly reviews the issues that might arise in applying the proposed indicators to other infrastructure sectors. Water Like energy, water can be provided from a number of sources, such as con- ventional piped distnbution, private wells, public standpipes, or private ven- dors. Many households may lack access to one or more of these alternatives, 163 VIVIEN FOSTER AND JEAN-PHILIPPE TRE in particular, a connection to the piped water network. Whereas households connected to the electricity network will continue to make extensive use of alternative fuels, piped water is typically more of a substitute for than a complement to nonpiped sources. Nonetheless, households may sometimes choose to combine alternative sources of water, perhaps as a way of deal- ing with poor reliability or quality of public supply, for example, using a private well as a backup when piped water services are suspended or buy- ing bottled water if piped water is of inadequate quality. All of these fac- tors make it important for the water sector to have welfare indicators that embody information on all alternative sources and not simply on those sup- plied by the network utility. A key issue in calculating the energy indicators was the need to correct for the differing efficiencies of alternative fuels. While the same concern may be present in the water sector, in practice it is much harder to capture or quantify. Anecdotal evidence suggests that water purchased in small quantities at high unit costs may be much more efficiently used than piped water supplied under pressure; however, no simple way to correct for this is available as was the case for energy, largely because there is no objec- tive physical measure of the hygiene output provided by water services as there was of the energy output provided by fuel. The basic needs indicators of coverage, reliability, and concentration would therefore appear to carry across to the water sector. Turning to the monetary indicators, the concept of a subsistence threshold is fairly com- monplace in the water sector, facilitating the calculation of the affordability index and the economic burden. The average cost of water and the average value of subsidies remain relevant. For water that is collected directly from nature, as for fuelwood, costing would have to be either on the basis of the time expended or in terms of the opportunity cost of acquiring water from the next best alternative source. The consideration of capital costs may also be of interest, in the sense that the cost of indoor plumbing needed to take full advantage of piped water may be an important barrier to access. Esti- mating the unit costs of interior plumbing based on information about san- itary facilities in the household should, in principle, be possible; however, in practice this is unlikely to be as straightforward as estimating the capi- tal costs of standard energy durables. Finally, the nonmonetary aspects of welfare are likely to be particularly important for water and sanitation, which are known to have major impacts 164 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR on health. In this case some measure of the incidence of wvaterborne dis- eases in the target population would be the key parameter to consider. Although the indicators requiredl for the wvater sector are in some respects simpler than those for the energy sector, in practice they may be harder to calculate on the basis of standard household survey data. I'his is because hiouseholdl surveys rarely allow households to report more than one main source of water, nor do they break down expenditure data by type of water source or provide adequate information about the time spent collecting water, the prices of alternative water sources, or the extent of household sanitary installations. Nonetheless, with a number of relatively modest changes to the stanclard water survey module, applying the framework dis- cussed would be relatively straightforward. Telecomn,munications Households face at least three basic options for telecommunications ser- vices: a private fixed line; a private cellular telephone; or a public tele- phone service, which could be anythling from a public booth to a microen- terprise renting out the use of a mobile telephone. In addition, a range of other technologies is available, from beepers, to faxes, to email. For the purposes of this section, only the three basic alternatives for voice teleph- ony are considered. The availability of alternative forms of supply at dif- fering costs once again justifies the use of a holistic approach that com- bines information about different sources of service. The major challenge of applying the proposed indicators to the telecom- munications sector is likely to be the problem of converting monetary expen- ditures into physical units, which is complicated by the following factors. First, dlifferent types of phone calls (local, long distance, international) attract different unit charges, and survey data rarely break down expenditure across different types of calls. This form of heterogeneity might potentially be ignored altogether by simply converting all expenditures into minutes of local call equivalent. Second, different types of telephones have varying charges for calls of equivalent distances. To the extent that expenditure can be bro- ken down across the type of telephone used, adjusting for the relative costs of different types of telephones might be possible, so that all expenditures can be converted to minutes of fixed private line telephony equivalent. Third, the use of value added network services makes converting expenditures into 165 VIVIEN FOSTER AND JEAN-PHILIPPE TRE minutes of call time harder. It is difficult to see how this distorting effect might be taken into account. As the discussion makes clear, the conversion from monetary to physical units will be possible only to the extent that expenditures are broken down across the three different types of telephone services and that the impact of value added services is relatively modest. Where these conditions are met, calculating the full range of indicators presented in this chapter would be relatLively straightforward. A possible exception is the affordability index, in the sense that the concept of subsistence consumption may be difficult to apply to the case of telecommunications. As in the case of water, however, the indicators would not be as easy to derive as for the energy sec- tor, because conventional household surveys tend not to break down expen- diture across different types of telephone calls. Finally, whether any nonmonetary welfare measures are directly related to telecommunications is not clear. Transport The transport sector obviously fits the pattern of a range of alternative types of service available to the household with a high degree of substitutability between alternatives. Thus indicators of coverage, reliability, and concen- tration would appear to have a direct application. As with telecommunications, the concept of a subsistence threshold does not appear to be as relevant as for water and energy, although if desired it could be interpreted in terms of the distance traversed for essential travel, such as commuting to work, traveling to school, and visiting the local market. Standard household surveys are more likely to contain information about travel expenditure than distances traveled. Conversion to physical units is therefore contingent on the availability of information about typical costs per mile traveled by different modes. An important limitation is that esti- mating distances traveled by such modes as bicycle and foot that do not entail direct operating expenditures may not be possible. As with energy, a significant issue is the relative efficiency of alternative forms of service. Hence a bus may be much cheaper per mile traveled than an automobile, but it is also likely to take much longer. One possible means of taking this into account is to add the time costs to the financial costs of each mode of travel to come up with a time-adjusted overall cost of travel; 166 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR however, this would require information about the average speed traveled by different modes of transport. The relevance of nonmonetary welfare measures is not as clear as in the case of water and electricity, except perhaps in terms of travel safety. Summary Table 4.12 summarizes thle overall conclusions of the discussion, broken down by the special methodological issues that arise in each sector and the relevance of the indicators to each sector. TABLE 4 12. Methodological Issues and Relevance of Proposed Indicators to Other Infrastructure Sectors Category Water Telecommunications Transport Methodological issues Does expenditure capture Excludes water Yes Excludes travel on foot all uses of the service? gathered from nature and bicycle Information needed to Unit prices of all water Unit prices for different Unit prices of alternative convert from expenditure sources types of telephone services modes of transport to physical units7 Need to adjust for Desirable but not No Should adjust for relative efficiency of possible different speed of different modes7 alternative modes Indicators Coverage index I / / Reliability index / / / Concentration index / / / Affordability index / X,i Average variable cost / / / Average subsidy / / / Average capital cost / / / Average total cost / / / Capital intensity / / / Economic burden / K Nonmonetary indicators Health impacts n a Traffic safety n a Not applicable Source Authors 167 VIVIEN FOSTER AND JEAN-PHILIPPE TRE 4.8 Conclusions This chapter began by arguing the need for a set of quantitative indicators as a means of measuring the impact of interventions in the energy sector on the welfare of the poor. The value of these indicators is twofold. First, they help to provide a diagnostic tool of the sectoral situation prior to any policy intervention. Second, they provide a means of gauging the impact of any particular intervention by comparing the value of indicators before and after the change took place. We developed three sets of indicators covering access to and affordabil- ity of energy services and wider impacts on health and educational out- comes. An important feature of these indicators is that they seek to take a holistic view of energy consumption rather than focusing narrowly on the electncity sector, as has too often been the case in the past. This philoso- phy is supported by the findings of empirical studies of energy and poverty, which find that the poor make only limited use of electricity, even once they have been provided with a household connection. The major challenge of implementing this approach is the need to have household-level information both about poverty and about energy usage. To assess how onerous the data requirements might be in practice, we cal- culated the indicators using household survey data from Guatemala. This exercise showed that most of the indicators could be calculated relatively quickly and easily on the basis of a fairly standard survey, together with some complementary information derived from the national consumer price index. The case study also demonstrated the value of the indicators in iden- tifying those policy interventions that are most likely to benefit the poor. With minor adaptations, the basic principles of this framework could be applied to any of the other infrastructure sectors, and would be of greatest direct relevance to the water sector. However, for a number of reasons the type of information needed to calculate such indicators would probably be harder to obtain in the case of water than in the case of energy because of the way that questions about water are typically structured in household survey questionnaires. Nonetheless, some relatively modest changes in questionnaire design would permit the indicators to be calculated for the water sector. 168 - 0 Cr.- CNO n Nn - v K D R o °- 0 C 8 o I~~ 05 00 _ _ 05 0 o n cl 1 _ " o o _ IR 7 C O - 0 N- to - o r to Cr tO ONtO O G- Cr UNC ON -h _ _ Cr Un ON C -CUONv Ln C n w N-CU0OU CUCU ONONO L n 0 5 a UNOIn rJ 0 r 0 ¢ o > R - v NON t00 LO -C No N -ON v~~~t. C: _ NCN _ o C_ = __ __ c c 6.- -C --C _ _-~~~~~~~~~~~~~~~~~~~~~- s - - - 0) 4 o _ na }~t aC C O _ ~ _t = _ n to _ _rr 6 _- o Q _ m O U-0. -nUN OC r tNo __ot_ tOn.-uNr 0mm C _n -orn _U UN COC _ Cv 0rC UN S U . t C-_ aa C ) Q _ n _co r L n ON 0 00 Dsr__v no O :t: ~~q oo_I_ qt U rs re _ rs~~~~~~i r o i 0L m _ Zrs _ sv,* 4) mo tO 00 *Deo NJ _ U to O N- O aO _N N- Cu _o _o CU to UNCN CU U N :3 _ e CO L moi Ro-- 40 _ 0 z - 2 uN - aU °° o~~~ C 0~~C 0 09 ~~~~~~~~~~~~~~ CC~~~~~~~~~~~~~~~~C C-, - ~ ~ ~ - 4~~~~ CC C c S < e ~C = C 00 E 169 ON2 OP.; ~ - mm - LA LA 0m Im 00 JO E -0 0 L g m LA 00 ON - .C Si 2 ZS o 8'ooS s o U'l _ .n a o _ ~~~~~~~~~~~~~> S c o = 8 8 S88°S O@88; 0 O O O O O0 N 00 - (N A 00 mOO 0 m 0 A - 8L8 - N C mm 8 8- 8o8 _c 0 .ONC OO O O O ONO O L AO s ~oo oo 6c 00 0000 00 8gg E 80 _ _ _ _ _ 8 _ _ _ 8 _ _ _ _ _ _ _ _ _ _ _ 8 _ _ _ 8 _ _ _ 8 _ _ _ 8 _ _ _ _ _ _ _ IuC m O 0 0 ON0 0 - O ON 0 0 0 0 m O E Cm 0 o 0 m mm 0 m 00 A C~L ON … 0 O N . m - _-e 00 Of C m LA - - ON C O ON 0 m 00 A- 00 - ON { ON mOO mo o m m oe o 0D 0 _m Q 0 0 0 0 6 6 8 6 o 0 0 0 0 u ~~O mc - *80 00 mA-X ON LA Cor 00 LA@ 8 O. mO O mO O A- m mm 0C 00 m c .0 ID C> M a r _o _ _u V c 6 o o 0 o o 0 o 0 o o o o 0 o o 6 o CL 0 -0 0 ON mLC w~~~~~~~ o6o o o6 o o6 0060 o00 ooo Z .° - C oo6 oo 0 C0 00 0 00 U 0J A- E '-0 mo A NCO N m A- C O O°- 00 ON ° 0 0 0 0.- 00 A- m 217 0 C Li A NONm O C-m-m A '0 o 0 ON '0~~-0 C --0 LA> C0 I 0000000 ~~~ ~~~~~~~CC ,C >C 000 00 06 *00 2 0 N0 0 00 O N 0C 0 O 0,0~~~0 CL CDQC 20c 00 40 w L AA0 170 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR Appendix 4.2: Methodology Units of Mieasure and Energy Efficiency Conversion Factors Table A2.1 presents measure and efficiency conversion factors for the var- ious fuels considered in this analysis. The following describe the method- ology used in estimating these parameters: * Electrictty. Electricity is expressed in kilowatt-hours (kWh) and is used as the reference in assessing the efficiency of all types of householdl fuels used. Units of all other fuels are converted into kWh, and their efficiency factors are estimated relative to that of electricity in lighting and cookin, as well as in the use of other elec- tric appliances * Kerosene. The adoptedl unit of measure for kerosene is the British gal- lon, which is equal to 3.78 liters. The United Nations (1987, pp. 23-24) TABLE A2.1. Units and Efficiency Conversion Factors Conversion factor from unit of measure Efficiency Effective Category Unit to kWh conversion factor kWh Lighting and appliances Electricity kWh 1 00 1 00 1 00 Kerosene Wick lamp British gallons 36 29 0 01 0 30 Mantle lamp British gallons 36 29 0 07 2 42 Candles 30 grams 0 30 0 02 0 01 Batteries Pack n a 1 00 n a Car batteries Single battery n a. 1 00 n a Cooking Propane Pound 5 24 0 77 4 03 Wood 2 kg 7 00 015 1 07 n a Not applicable Source Authors' calculations using end-use efficiency factors provided in Leach and Gowen (1987) and United Nations (1987) 171 VIVIEN FOSTER AND JEAN-PHILIPPE TRC provides a liquid fuel equivalent estimate of 12 megawatt-hours per metric ton of kerosene. Given a density of 0.80 gram/cubic centimeter (0.80 kilogram/liter or 0.80 x 10-3 tons/liter), 1 ton of kerosene corresponds to a volume of 1,250 liters or 330.69 British gallons. Thus the factor of conversion from gallon of kerosene to kWh is obtained by dividing the liquid fuel equivalent of 12 x 103 kWh/ton by 330.69. The resulting factor of conversion equals 36.29. Efficiency conversion factors are derived from Van der Plas and De Graaff (1988, pp. 12, 55). The luminous efficacy of kerosene wick and mantle lamps (0.1 and 0.8 lumen/watt, respectively) is divided by that of a standard 60-watt ref- erence lamp (12 lumens/watt). Efficiency conversion factors of 0.00833 and 0.06667 are obtained for wick and mantle lamps, respectively. o Candles. A candlestick is assumed to weigh 30 grams (0.030 kilogram). Given a calorific value of 36 megajoules per kilogram (Van der Plas and De Graaff 1988), the conversion factor from kilograms to kWh is obtained by multiplying 36 by 0.030 divided by 3.6 (equivalence fac- tor between megajoules and kWh). The efficiency conversion factor of candles is calculated as that of kerosene with a luminous efficacy of 0.2 lumen/watt. o Propane. Most purchases of propane in Guatemala are made in contain- ers of 25 or 35 pounds, and we therefore adopt the pound as the unit of measure for propSane. A liquid fuel equivalent of 12.67 megawatt-hours per metric ton of propane is given in United Nations (1987). This corre- sponds to a conversion factor of (12.67 x 10-3)/(2.42 x 10-3), or 5.24 kWh per pound of propane. The efficiency factor for propane relative to electricity is derived from end-use efficiency factors provided in Leach and Gowen (1987, p. 101). Given an average factor of 65 percent for electricity and 50 percent for propane, an efficiency factor of 0.769 (0.50/0.65) is obtained for propane. oWood. Fuelwood is assumed to be sold by log of 2 kilograms. A liq- uid fuel equivalent of 3.5 megawatt-hours/metric ton for wood is given in United Nations (1987). Hence the conversion factor from kilograms to kWh is obtained by multiplying 3.5 by 2. The effi- ciency factor of fuelwood is calculated as that of propane using an end-use factor of 10 percent for air-dried wood and 65 percent for electricity. 172 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR Effective Costs of Fuels Per Kilowatt-Hour * Batteries. The effective price of batteries is estimated by dividing the total monthly expenditure on batteries by the total amount of energy consumed during the same period of time. Households not connected to the electricity grid and owning only a radio consume, on average, 1.25 packs of batteries per month. Assuming that each pack of batteries costs Q 5.3 and that a 16-watt radio is used for four hours every day, the effec- tive price is calculated as follows: (1.25*5.3)/(0.016*4*30) = Q 3.5. * Car battertes. The effective price of car batteries is calculated similarly to that of regular batteries. We estimate that a car battery must be recharged weekly at a cost of US$1 (or Q 6.5) to power a 16-watt tele- vision for four hours of use every day. Hence the effective price is obtained as follows: 6.5/(0.016*4*7) = Q 15. * Other faels. Effective prices of propane, fuelwood, kerosene, and can- dles were calculated by dividing their retail prices by their correspon- ding energy conversion factor. The retail prices shown in table A2.2 were used to calculate effective prices. For example, the retail price of a 35-pound container of propane is Q 63.56 in the metropolitan area, or Q 1.82 per pound of propane. Given an energy conversion factor of 5.24 for propane, its effective price is obtained by dividing 1.82 by 5.24, which equals Q 0.35 (US$0.05) per pound of propane. TABLE A2.2. Retail Prices for Different Fuels 1998/99 (Q/unit) Propane Fuelwood Kerosene Candle Batteries Region (Q/35 pounds) (Q/log) (Q/gallon) (Q/stick) (Q/pack) Metropolitan 63 56 0 47 12 05 0 60 5 89 North 57 74 0 59 11 05 0 56 5 05 Northeast 57 32 0 44 11 15 0 32 5 72 Southeast 57 25 0 57 12 84 0 36 4 85 Central 57 95 0 40 10 94 0 42 4 85 Southwest 57 82 0 43 10 91 0 59 4 75 Northwest 57 30 0 36 11 52 0 50 5 50 Peten 57 30 0 36 11 58 0 50 5 50 Q Quetzales Source National Institute of Statistics, Guatemala, data 173 VIVIEN FOSTER AND JEAN-PHILIPPE TRt Notes 1. Anothier possibility is to define subsistence energy consumptioni empiricall) rather than normatively. This can be done by looking at the actual energy con- suinption of a reference group that is believed to be living in a subsistence situa- tion, for example, those whose total income or consumllption lies close to the extreme poverty line. 2. International benchmark poverty lines also exist, for example, the US$1 pur- chasing power parity per day or USS2 purclhasing power parity pet day povert) lines adlopted by the World Bank as thresholds of extreme poverty and poverty, respectivel). 3. The concentration coefficient is bounded between plus and minus one, with positive values indicating a regressive distribution, negative values indicating a progressive distribution, and a value of zero indicating a perfectly egalitarian dis- tribution. Thc formula for calculating the conicentration coefficient is it I]-(+ n) where it is the total number of groupings of the income variable used (for example, 10 deciles) andl x, is the shaie of the total number of connectionis going to grouping i (not to be confused with the connection rate for that grouping). 4. The World Bank started the LSMS surveys in 1980 to explore ways of iniprov- ing the types and quality of household (lata collected by statistical officcs in devel- oping countries. LSMS surveys have now been conducted in sonie 50 countries woildwide Although the specific content of individual surveys varies somewhat, they all tend to provide a comprehensive picture of the quality of life by covering a wide range of socioeconomic and demographic issues. All surveys must meet cer- tain quality standards in relation to the design of questionnaires and the adminis- tration of fieldwork. 5. Appendix 4.1 provides a detailed comparison of gross and net hiousehold energy consumption denving from different fuels across deciles. 6. Appendix 4 2 describes the methodology used in estimating average prices per gross and net kWh for different fuels. 7. This threslhold provides enough energy to rmn two 60-watt light bulbs and one 16-watt radio for four hours each day, and incorporates a cooking requiremllenit of five 2-kilogram logs of fuelwood each day. 8. The durables the survey considered are electric heaters, electric cookers, refrigerators, microwave ovens, food processors, irons, electric showers, sewing machines, washing machllines, radios, televisions, stereos, video recorders, and computers. 174 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR References The word "processed" describes informally reproduced works that may not be commonly available through libraries. Albouy, Y., and N. Nadifi. 1999. "lmpact of Power Sector Reform on the Poor: A Review of Issues and the Literature." World Bank, Washington, D.C. Processed. Baker, J. L. 1999. "Evaluating Project Impact for Poverty Reduction: A Handbook for Practitioners." White Cover Report no. LCSPR/PRMPO. World Bank, Washington, D.C. Barnes, D., and L. Qian. 1992. "Urban Interfuel Substitution, Energy Use and Equity in Developing Countries: Some Preliminary Results." World Bank, Washington, D.C. Processed. Chong, A., and J. Hentschel. 1999. "Bundling of Basic Services, Welfare, and Structural Reform in Peru." World Bank, Development Economics Research Group, Washington, D.C. Processed. Cowell, F. A. ] 995. Measuring Inequality, 2nd ed. London: Prentice Hall Harvester Wheatsheaf. Domdom, A., V. Abiad, and H. Pasimio. 1999. "Rural Electrification Ben- efit Assessment Study: The Case of the Philippines " Draft report. World Bank, Energy Sector Management Assistance Program, Washing- ton, D.C. Eberhard, A., and C. van Horen. 1995. Poverty and Powver: Energy and the South Afrtcan State. East Haven, Connecticut: Pluto Press. ESMAP (Energy Sector Management Assistance Program). 1994. "Ecuador: Energy Pricing, Poverty, and Social Mitigation." Report no. 12831-EC. World Bank, Washington, D.C. Foster, V., J.-P. Tre, K. Lindert, and C. Sobrado. 2000. "Nota Preliminar sobre la Pobreza en Guatemala con Base en la ENIGFAM 1998/99." 175 VIVIEN FOSTER AND JEAN-PHILIPPE TRt World Bank, Latin America and Caribbean Region, PQverty Reduction and Economic Management Division, Washington, D.C. Processed. G6mez-Lobo, A., V. Foster, and J. Halpern. 1999. "Information and Mod- eling Issues Related to Water Subsidy Design." World Bank, Washing- ton, D.C. Processed. Hicks, N. 1998. "An Analysis of the Index of Unsatisfied Basic Needs (NBI) of Argentina with Suggestions for Improvement." World Bank, Latin America and the Caribbean Region, Poverty Unit, VWashington, D.C. Processed. Hosier, R., and W Kipondya. 1993. "Urban Household Energy Use in Tanzania. Prices, Substitutes, and Poverty." Energy Policy 21: 454-73. Kakwani, N. 1986 Analyzing Redistribution Policies: A Study Using Aus- tralhan Data. Cambridge, U.K.: Cambridge University Press. Lanjouw, J. 0. 1999. "Demystifying Poverty Lines." United Nations Development Programme, New York. Processed. Leach, G., and M. Gowen. 1987. Household Energy Handbook: An Interim Guide and Reference Manual. Technical Paper no. 7. Washington, D.C.: World Bank. Lok-Dessallien, R. 1999. "Review of Poverty Concepts and Indicators." United Nations Development Programme, New York. Processed. Ravallion, M. 1998. "Poverty Lines in Theory and Practice." Working Paper no. 133, Living Standards Measurement Study. World Bank, Washington, D.C. United Nations. 1987. Energy Statistics: Definitions, Units of Measure, and Conversion Factors. Studies in Methods no. 44. New York: Depart- ment of International Economic and Social Affairs. Van der Plas, R., and A. De Graaff. 1988. A Comparison of Lampsfor Domestic Lighting in Developing Countries. Energy Series Paper no. 6. World Bank, Industry and Energy Department, Washington, D.C. 176 MEASURING THE IMPACT OF ENERGY INTERVENTIONS ON THE POOR Velez, C. E. 1995. "Gasto Social y Desigualdad: Logros y Extravfos." Department of National Planning, Social Mission, Bogota, Colombia. World Bank. 2000. "Maintaining Utility Services for the Poor: Policies and Practices in Central and Eastern Europe and the Former Soviet Union." Draft report. World Bank, Europe and Central Asia Region, Washington, D.C. 177 5 Impact of Market Structure on Service Options for the Poor David Ehrhardt 179 DAVID EHRHARDT 5.1 Pro-Poor Structural Checklist for Privage Participation in Infrastructure Governments should consider a number of key pro-poor structural issues when planning to introduce private participation in network utility indus- tries. Decisions made about these issues will, to a large extent, determine the impact of private participation on poor communities. Reforms aimed at assisting the poor must be considered explicitly before any major reform is actually carried out. They should begin with the col- lection of data to assess the specific conditions that affect the poor and their specific needs and preferences rather than relying on generalizations. Pro-poor issues should be considered as part of an integrated approach to structural reform. This will ensure that pro-poor structural reforms are compatible with the overall approach; for example, they should result in regulation that focuses on network access and is not too onerous for infor- mal vendors and on a subsidy regime that is robust to the degree of entry and competition envisioned. The most important factor in ensuring that poor communities receive service is establishing a competitive, efficient, and well-capitalized indus- try. This will benefit all consumers to the extent that utilities are able to provide good service at least cost and are able to fund capital expenditures to meet demand. It is also more likely to benefit the poor, because ineffi- cient, undercapitalized utilities are unlikely to be able or willing to take risks or expend capital to meet the needs of the poor. To ensure that all customers benefit from private participation, certain reforms should, in general, be considered in relation to utility market struc- ture, namely: o Horizontal unbundling, which introduces competition in each sector of the industry o Vertical unbundling, which separates the ownership or operation of dif- ferent sectors of the utility from one another o Private participation, which may range from build-own-operate-trans- fer ventures to privatization o Free entry into an industry. The manner in which utilities are reformed along these lines will affect subsidy design and regulation. Administrative capability is also important in determining the types of reforms implemented. 180 IMPACT OF MARKET STRUCTURE ON SERVICE OPTIONS FOR THE POOR In addition to the reforms aimed at improving service and efficiency for all customers, the most important reforms specifically targeted to the poor are those that would allow small-scale and informal vendors to provide service, even if they are directly competing with the dominant provider To facillitate this, the major reforms necessary are as follows: * Allowing entry by removing exclusivity laws and licensing require- ments * Allowing vendors to access bulk supplies of water and electricity from existing suppliers at fair prices * Allowing interconnection with existing telecommunications networks. When implementing these reforms, subsidy dlesign should also be taken into account. Pro-poor subsidy design may be enlhanced by the following considerations: * General income subsidies financed by taxation are likely to be the most efficient and have the least impact on market structure. * If the government is fiscally constrained, then a universal service fund should be considered; however, this may be difficult to administer. * If a subsidy is required but administrative capacity is limited, then a cross-subsidy with restrictions on free entry and unbundling may be necessary. This should be assessedl on a case by case basis to deter- mine whether competition or cross-subsidlies are more likely to be pro- poor. An important factor in this decision is whether the monopoly provider is already providing some service to the poor or can be com- pelled to do so. If this is not the case, then the poor would not benefit from the cross-subsidy and are likely to be better off with new, innova- tive entrants to the market. * The administration of a subsidy through several vendors may be com- plex and may prevent the government from subsidizing the customers of new entrants; howvever, free entry and unbundling should not be restricted because of this concern. If an entrant is able to serve the poor without a subsidy, then this should be welcomed. Regulation wvill be affected by the recommended pro-poor reforms. As the structure of the industry changes with the introduction of pro-poor competition into various unbundled sections of the utilities, appropriate regulatory responses are likely to include the following: * Reducing the scope of regulation so that competitive areas are no longer regulated 181 DAVID EHRHARDT o Changing the focus of regulation from controlling retail prices to regu- lating the price at which bulk service or network access is provided to competing providers o Adding an antitrust or competition law element to regulation to prevent providers that have a dominant position in a market from using that position to prevent competition in that or related markets. At the same time, small entrants should not be regulated even if they have a local monopoly. 5.2 Problems with Service Delivery to the Poor According to Kerf and Smith (1996) in 1996, on average, only about 48 percent of households in Sub-Saharan Africa had access to electricity, ranging from 1 percent in Burundi to 96 percent in Senegal. Only about 42 percent of Sub-Saharan Africa had access to safe water. In East Asia and the Pacific the figure for access to safe water was 68 percent, and in Latin America about 76 percent had access. Those left without access to serv- ices were undoubtedly the poorest people in these regions. Those living in rural communities generally have less access to utility services than urban residents. In a survey of electricity coverage in devel- oping countries, Komives, Wittingham, and Wu (2000) found that outside Africa, while urban coverage exceeded 85 percent, rural electrification varied widely among the countries surveyed. In Kenya, by contrast, fewer than 2 percent of the 3.7 million rural households have access to grid elec- tricity. Even if the annual connection rate was 10,000 connections, con- necting the existing rural population would take almost 400 years (Hank- ins 2000). These conditions tend to apply in other utility industries. Lack of access is usually a function of local geography in both rural and urban areas. Areas where poor people live often do not have utility serv- ices, because these services are provided over networks that do not extend to those areas for several reasons. First, service providers may simply find it uneconomical to supply poor areas. This may be because of concerns about revenue collection or additional costs. The result is that the infra- structure is not put in place to provide the service. This problem may be compounded if the utility is a state-run monopoly provider. Such utilities have tended to be inefficient and unprofitable. Consequently, it may be 182 IMPACT OF MARKET STRUCTURE ON SERVICE OPTIONS FOR THE POOR unable or unwilling to make the necessary capital expenditures to provide service to the poor. In some countries the percentage of poor people with access to services is actually falling, because state-run utilities are unable to keep up with expanding demand. Second, the traditional sectors of society have virtually abandoned many poor urban areas. Some areas have levels of violence that make the staff of the service provider reluctant to enter the area; for example, in some areas of downtown Kingston, Jamaica, technicians of the water and electricity utilities lhave been physically attacked and shot at. Such concerns may prevent utili- ties from investing at all in these areas or from maintaining existing networks. Third, poor communities often grow up on marginal lands, such as steep hillsides or riverbanks. Such terrain is more difficult to service, and the engineering issues involved in delivering the service may be costly to resolve. In other cases poor communities may develop along urban periph- eries, which may mean that costly new trunk mains or transmission lines need to be extended to the edge of the metropolis and new distribution sys- tems installed. Utilities may not make the investment, having little hope of recovermg the cost through additional tariffs. Fourth, rural areas may be more difficult to serve for several reasons, includ- ing that rural communities are more sparsely populated and also tend to have a greater proportion of poor to rich consumers than urban communities. Fifth, in both poor urban and rural areas, inhabitants often do not have formal title to the land occupied. Where land tenure is formalized, poor people will often be tenants, not owners. Lack of a freehold title can pre- sent various problems for service providers, in that utilities may be legally required to contract with the actual owner of the property or may require that customers have title to the property being served as security against nonpayment of utility bills. Sixth, in many slums and rural communities, liability for utility bills may also be uncertain. Poor communities often have more than one family liv- ing in a single house, which can pose a problem for bldling and collection. Finally, the mind-set of traditional utilities sometimes results in lowv- quality service to poor communities. Most large utilities tend to value structure and order and are unable to deal effectively with the sometimes disordered environment in poor communities. The special characteristics of poor com1munities often require alternative approaches that utility com- panies may not be comfortable with. 183 DAVID EHRHARDT 5.3 Structural Options In response to disappointing service and financial performance by conven- tional, state-owned monopolies, many governments have undertaken seri- ous reforms. These reforms typically change the structure and ownership of the industry, introducing competition and private capital. This section outlines structural options available to governments as they reform their utility sectors and examines the following options: o Unbundling an integrated monopoly provider o Changing the ownership structure o Allowing entry by new providers in some market segments. The possibilities for new and innovative structures are myriad, and the aim is not to document them all, but rather to provide a framework or typology within which the main options can be identified and their relationship with each other can be seen. Unbundling an Integrated Monopoly Provider Until the 1980s the accepted structure for a utility provider was a verti- cally integrated geographic monopoly. In the industrial countries this started to change in the 1980s. One of the first high-profile cases was the break-up of AT&T in 1984 in the United States. The trend spread to the energy industry, notably with unbundling in the United Kingdom in 1989, and later to the water industry as well. During the 1990s developing countries started to pursue similar strate- gies. Across Latin America, in particular, restructuring and privatization of utility companies accelerated. Competition was also introduced into many of the unbundled industries; for example, in Argentina the three major vertically integrated electricity companies were broken up and pri- vatized in 1992. This has resulted in 25 generation companies, 22 distri- bution companies, and 1 transmission company. These reforms have often succeeded in increasing the efficiency of utilities and making them more responsive to customers. A range of possible options for unbundling utility industries is available. Figure 5.1 presents a typography of the options available. It illustrates full unbundling of a utility such as a water, electricity, or gas utility. Produc- tion, transmission, distribution, and retail supply of the service are sepa- 184 IMPACT OF MARKET STRUCTURE ON SERVICE OPTIONS FOR THE POOR rately operated by clifferent companies. Production refers to the abstrac- tion and treatment of water or gas or the generation of electricity. Trans- mission includles main trunk lines through which water, gas, or electricity are fecl from the production plant to the distribution network, which carries the service to individual homes. Retailers contract with distributors and customers to sell the service from the distributor's network to the consumer. In figure 5.1 competition is present in all markets except for transmission and distribution, where competition is seldom feasible. The two companies shown as distributors are regional monopolies. Unbundling in telecommunications or transport looks a little different. This is because the network is not configured to deliver a product to con- sumers, but rather to connect together customers who wish to communicate with each other. In telecommunications unbundling has typically involved the separation of local telephony, international telephony, wireless teleph- ony, and value added services. In transport the equivalent could be having FIGURE 5.1. An Unbundled Industry Production company 1 Production company 2 Production company 3 F TransmIssion company Distribution company 1 | Distribution company 2 | ~~~~ ~~Retail Rtl Retail Rtl 0 company 1 ~company 2 cmay3company 4 CUSTOMERS Source Author 185 DAVID EHRHARDT para-transit operators provide feeder services to trunk routes operated by larger buses and allowing competition on some or all routes. Changing the Ownership Structure for Existing Providers The wave of utility reforms of the 1980s and 1990s included not only unbundling and competition, but also increased involvement of the private sector in the ownership and management of utility companies. In some industrial countries private participation in infrastructure has long been the norm. In the United States, for example, telecommunications and elec- tricity companies have always been privately owned, as has a significant proportion of the water industry. Similarly in France about a third of the water systems are privately managed, reflecting a long-standing tradition of concessioning infrastructure provision to private enterprise. Nevertheless, in many industrial countries and in most developing coun- tries, infrastructure provision has been the preserve of the public sector, at least from the 1950s to the mid-1980s. This often led to problems, includ- ing low levels of efficiency and a lack of capital with which to expand serv- ices as demand grew. Private participation has generally succeeded in reducing costs and increasing investment in utilities. More and more industrial and develop- ing countries have since followed this route. For example: o In 1993 in Buenos Aires, Argentina, a concession was awarded to a pri- vate consortium for the provision of water and sanitation services. o In 1996 the sale of generating companies in the electricity sector began in Brazil, with the sale of distribution companies following in 1997. o In the Czech Republic shares in municipal water and sanitation operating companies were sold as part of the voucher privatization process. This trend accelerated during the 1990s. In 1990 US$15.6 billion was invested in privately financed infrastructure projects in developing coun- tries, and by 1998 this figure had increased to US$95.3 billion. From 1990 to 1998 the value of such investments totaled US$496.2 billion, with the largest amount of such investment going to the telecommunications and energy sectors (Neil 1999). Diverse experiences, ranging from the Anglo-Saxon approach of private ownership of assets to the French model of concessioning or contracting 186 IMPACT OF MARKET STRUCTURE ON SERVICE OPTIONS FOR THE POOR out infrastructure management, have led to the evolution of a diversity of possible structures involving the private sector. 'Table 5.1 summarizes the main models for private sector participation. A continuum of options is available, ranging from contracting out discrete services to three-year man- agement contracts, privatization, and asset sales. Typically, the longer the contract, the more responsibility for investment the private sector takes, and the greater the risk, while shorter contracts tend to imply more fre- quent competition for the market and greater flexibility. Allowing Entry by New Providers in Some Market Segments Often reforms have focused on restructuring an existing utility, but changes in industry structure and ownership can also result from the entry of new providers into the business. These new providers may compete with the TABLE 5.1. Options for Private Participation Management Maintenance Investment Ownership Option Operation of system (a) (b) (C) Planning Financing of assets Service / X x J x X Public sector contract Management / / / X XK K X Public sector contract Lease / / / / / / X Public sector Concession / / V / Public sectori Asset sale / / / / / / / Company BOOT (new / )t . / / / . Company/ assets) public sector / Responsibility lies with the private operator St Responsibility lies with the public sector BOOT Build-operate-own-transfer a This incorporates three different functions planning (a), carrying out the work (b), and financing the maintenance (c) b The assets are transferred to the concessionaire for a fixed period of time, but are owned by the state Source Author 187 DAVID EHRHARDT incumbent, or they may serve market niches that were previously unserved. New entrants can be a powerful force for change in their own right and can act as a catalyst for further structural change in the industry; for instance, MCI's entry into the U.S. telecommunications market catalyzed the sector's restructuring and liberalization. Thus the extent to which entry is encour- aged is another important policy tool. If competitors are able to enter the incumbent's market, this will tend to drive prices down in line with costs. Even if competition is limited, the threat of increasing entry will encourage incumbents to lower prices. This may also lead to further unbundling in the future. Box 5.1 describes an example of this process. In the telecommunications sector, in particular, developing countries are increasingly permitting competition in cellular telephony. Where competi- tion exists, access has increased and prices have also tended to fall, thereby improving services to all customers, including the poor (Rossotto, Kerf, and Rohlfs 1999). Poor customers are likely to benefit to a greater extent from these changes in the market than other groups, because they are hit the hardest by the inefficiencies of monopoly providers. New entrants may also improve service simply by providing it where it is not currently offered. This is more likely to be the case in poor commu- Box 5.1 Liberalizing TelecommunicaLions In Jamaica In Jamaica. the entrance of -sat lif enses (operators /rofi iing i'tterna- tionial communication senuce using nini satellite dishes) and call-b,e I services into the teleoommuniutnoos market hegan to bhip ouia) at Ca/ile and WVireless:s monopolb in long dictancee telephont. These newx entrants were encouraged b) the existence of cross s.ubsidis../for lot al telephony from long distanice cailers. which raised the price of lon/,r distance sert res andl reated price distortions. Partly in response to this eompetiu e threat. Coahb/i' im Ilireless recenth agreed to give up its exclusitii in morni arnos. The consequence of this iill be the unbonldling of the industr's rfni competiors emerge in (eliilar teleph- ony, talue added sencices. and broadblantd andi datoi transmission. 188 IMPACT OF MARKET STRUCTURE ON SERVICE OPTIONS FOR THE POOR nities. In villages in Bangladesh, for example, women entrepreneurs pro- vide payphone service at a profit using mobile cellular telephones (Welle- nius 2000). Poor customers throughout the developing world rely on infor- mal entrepreneurs to provide utility services when the traditional utility fails to deliver. Facilitating new entry is one of the most important ways in which stnic- tural reform can improve service to poor people. To date this area has not been given enough attention. 5.4 Structural Reforms Targeted to the Poor Such local characteristics as physical conditions, economic capabilities, social patterns, and land tenure arrangements mean that providing appro- priate service to the poor will often reqliire nonstandard service dlelivery mechanisms, types of service, and tariff andl paymnent mechanisms; how- ever, utilities tend to adopt a one size fits all approach to service and charging. Private sector participation can help, but by itself may be insuf- ficient, and the tendency to ignore poor and marginal areas may continue. Few utility managers will have much contact with poor areas or much real understanding of the needls of potential customers In such areas. Thus, even if an existing utility is disaggregated and privatized, problems of lack of access, inappropriate types of service, and to some extent inappiopri- ate pricing for poor people are likely to continue. Therefore making it eas- ier for other providers to serve markets out of the mainstream, such as poor areas, is imperative. This section examines specific structural reforms targeted to the poor. It discusses structural changes that could facilitate innovation in service pro- vision and provides examples of such innovations. The essence of the argument is that flexible and innovative approaches are needed if service is to be improved. While some traditional utilities try to develop approaches specifically designed to provide the poor with serv- ices, many mainstream utilities are often neither flexible nor innovative. Thus allowing new entrants, which will tend to offer alternative solutions, will frequently be beneficial. To make the case for structural reform, it is important to illustrate the types of innovations new entrants could offer and 189 DAVID EHRHARDT address the poor reputation that small and informal providers have in many areas. In the following sections we argue that o In some cases small operators may be able to provide a basic needs level of service more cheaply than formal network operators. o Small operators and new entrants may offer cost-quality combinations better suited to poor people's willingness to pay. o New entrants can offer innovative tariff and payment systems that make it easier for poor people to access service. In general, facilitating entry by small and innovative operators will increase choices for poor people. Price and Quantity Options Poor people often receive service from informal service providers rather than from formal utilities. For example, poor communities may be served by informal para-transit systems such as minibuses or route taxis rather than formally scheduled bus or metro services. Similarly, poor households often buy water from water carriers or vendors rather than obtaining it from a piped utility system. In many cases the service provided by informal providers is of lower quality than that of the formal network service and can also have higher unit costs. For example, in Guasmo Norte, a squatter community in Guayaquil, Ecuador, residents were paying about US$9.50 per month for an average of 25 tanks of often contaminated water. This was more expen- sive than the clean piped water being received by wealthier residents of the city (Salmen 1987, p. 39). Similar examples exist in most developing countries for a variety of services. People often use inferior services because the formal network does not extend to their communities. There is a widespread assumption that where it is available, a high-quality, formal network service is always preferable. This assumption should be questioned. The high fixed costs of formal net- work services mean that, in some cases, alternative approaches will offer poor people better value for money. The limited quantity poor customers demand may mean that an informal supplier's overall costs are less than those of a formal supplier. The apparent paradox of higher unit prices but lower monthly bills fol- lows from the cost structure of service providers. Network utilities provide 190 IMPACT OF MARKET STRUCTURE ON SERVICE OPTIONS FOR THE POOR a service that requires heavy capital expenditures on the distribution net- work, which means that fixed costs are high. The infrastructure does not require large subsequent expenditures per unit to deliver the service. By contrast, an informal vendor may provide service by trucking water door to door, reselling mobile phone use, or recharging 12-volt batteries. These approaches do not entail large up-front expenditures, but are more costly per unit delivered. So while purchasing a network service will be cheaper for most people, poor people who consume only small quantities may find that using alternative sources is cheaper. Table 5.2 provides a stylized example of the cost strictures of two dliffer- ent approaches to service provision for providing service to an area in which no fixed network is available. It could represent telephone service in a Latin American favela, energy supply for an African village, or water provision for an Indian slum. Table 5.2 assumes that service to the area can be provided either by a formal utility network or by informal suppliers. Fixed costs for the formal network primarily represent the cost of installing the network. Variable costs are those that occur each time a unit of the serv- ice is produced or delivered, and could include the cost of fuel for elec- tricity generation or pumping and treatment costs for water supply. Figure 5.2 is a graphical representation of table 5.2. It shows that at I 1 units per month, the costs for the formal network andl the informal provider are identical; however, if a consumer uses more than l1 units per month, service from the formal network is less expensive, and conversely, if a con- sumer uses less than 11 units per month, the service from the informal sup- plier is less expensive. A consumer may choose to consume any price and quantity bundle along either the informal supplier or the formal network supply curve. ff a customer cannot afford the formal network's minimum bill TABLE 5.2. Costs for Different Types of Services Variable cost Total cost Total cost Supplier Fixed cost per unit for 5 units for 15 units Formal network 100 1 105 115 Informal supplier 0 10 50 150 Source Author 191 E - D~~ \ 0E C'-~~~~ z E N - Qi c7 - C92 CL~~~~~~~~~~~~~~~~~~ 0~~~~~~~~~~~~ 'O~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. 0~~~~~~~~~~~~~~~0 C~~~~~~~~~~~~~~~~~~~~~~ (U 192N IMPACT OF MARKET STRUCTURE ON SERVICE OPTIONS FOR THE POOR of aroundl 100 units per month, then the only option is the informal supplier. In figure 5.2 a poor customer is consuming on the informal supplier's curve. The challenge is for service options to be provided at an intermediate level of price and quantity with flexible payment options. This would be possible if the advantages of a formal provider, including lower cost of cap- ital and better management, could be combined with the informal system's entrepreneurship andl responsiveness to actual demand. A novel approach can be found in Bolivia, where the water service providter, Aguas del llli- mani, offers households the option of paying a reducecl connection fee in exchange for supplying labor during the connection process. The provider also offers financing for the connection fees (Komives 1999). Using the framework in table 5.2 and figure 5 2, this would be a shift down in the fixed cost of supply from the utility and resLilts in a lower bill for any given quantLity purchased. More generally, improved access to finance can help to overcome the problems imposed by high connection costs This issue is dliscussed later in this section, which also outlines how structural changes can be instru- mental in increasing flexibility and the range of options available. Price and Quality Optiotns The discussion in the previous section can be expanded to take the quality dimension into account. Formal utility service is generally of a high qual- ity. Poor consumers may be purchasing services from informal suppliers at a lower quality than they would like and at a lower price than they are will- ing to pay. Foor example, water provided by informal vendors may be con- taminated witlh harmful bacteria, and poor customers would probably be willing to pay more for a level of service that is at least safe. If no service is offered in the area, the choice poor customers face is especially stark. The dominant utility may need to invest heavily to pro- vide high-quality service, but an informal provider is likely to have lower costs for providing service of lower quality. These options are illustrated in figure 5.3. Two possible price and qIuality bundles are available to all cus- tomers, one provided by the utility and the other by an infor-nmal supplier. Poor customers may be forced to choose the option providled by the vendor, because they cannot afford the option provided by the utility. The preferred space of a customer buying from the vendor is clearly in area P on the 193 CY~~~. o~~~~~~~~~ \ t-~~~~~~~~~ \ x~~~~~~~~~ I re~~~~~~~~~ I u~~~~~~~~~ I D \ ~* i \ i 0~~~~~~~~~~~ 0~~~~~~~~~~~~~~~ CY 0 (U (Vi\ \' ' I 0~~~~~~~~~~~~~~~~~~~ \ : I~~~~~~~~~~I 194 IMPACT OF MARKET STRUCTURE ON SERVICE OPTIONS FOR THE POOR graph. However, as noted, poor customers may be willing to pay a higher price for higher quality, though not at the level provided by the utility. This may be captured in area Q shown on the graph. The challenge is to provide a more appropriate position on the price and quality spectrum to the poor. The poor would almost certainly want to be provided with certain aspects of quality, such as stable electrical current or safe water; however, a price versus quality tradeoff on other aspects of provision is likely, such as water pressure, hours of electricity supplied, and indoor plumbing. Changes to traditional structures can help open up the possibility of serv- ices whose price and quality are better matched to poor people's needs and ability to pay. Allowing entry by alternative suppliers will open up new delivery mechanisms. Providing ways for informal suppliers to intercon- nect with aspects of the formal service can result in improved service qual- ity and access to finance, as illustrated by the following examples: * Teshie, a low-income, unplanned community in Ghana, grew beyond the dominant water provider's capacity to serve it. The network of the utility, the Ghana Water Company, extended to roughly half of the area, and the remaining area received no service at all. Entrepreneurs began to serve this latter area using tankers. An association of water tankers was formed that purchased bulk water from the Ghana Water Company and resold the water to individual tankers. Consequently, customers without a piped connection were now provided witlh the same quality of water as the rest of the community (Kariuki and Acolor 2000). * In Guatemala City a community group, the Community Action Neigh- borhood Association, operates the water supply system for the area. Water is provided by two systems, one older system handed over to the association by the municipality and a newer system constructed by the association. Customers are charged a flat rate of US$1.50 per month if they are connected to the old system, which provides water twice a week for two hours. Those connected to the new system, which provides water for about two hours each day, pay a variable rate of US$0.30 per cubic meter for the first 30 cubic meters and US$0.55 for every cubic meter after that (Snell 1998). Customers may choose to connect to either system. This gives customers with different incomes the ability to choose the quality of service appropriate to their needs and ability to pay. 195 DAVID EHRHARDT o The Dar es Salaam City Commission, which is responsible for providing sewerage services to the city, operates a sewerage system that serves about 20 percent of the city's demand. It also operates a fleet of four pit emptying trucks that serve the remainder of the city, which relies on septic tanks and pit latrines. This is insufficient to serve the entire area, resulting in waiting lists for the service. Private operators, initially ille- gal, began supplying cesspool emptying services by truck. Though this is of lower quality than piped sewerage service, previously unserved customers received a more hygienic option from the informal operators. Innovations in Payment Mechanisms New entrants may also offer tariff and payment mechanisms more suited to the needs of poor customers, because the cost structure of informal opera- tors may allow flexibility for poor consumers. Traditional utilities usually require up-front connection fees and reconnection fees, which can be oner- ous for the poor. By contrast, informal vendors do not usually require such large, one time expenditures. Poor customers are therefore able to control their expenditures directly by controlling their consumption. In addition, the poor often operate outside the traditional cash economy and may engage in bartering activities to meet their needs, and informal providers are more likely to be able to design payment mechanisms that can accommodate noncash transactions. A combination of privatization, competition, and regulation can push even the dominant utility to implement flexible payment mechanisms in response to the needs of the poor. British Telecom, for instance, offers such payment options as lower up-front fees for irregular users and prepaid call- ing facilities so that consumers can control their expenditure ahead of con- sumption (Wellenius 2000). Other innovations may also be expected as entry is liberalized. The fol- lowing sections review possibilities in the areas of optional tariff structures and improved payment security. Optional Tarigfs. Utilities generally charge a single rate for household use. An optional tariff refers to a rate structure that allows consumers to choose from two tariff options. For example, a customer could choose to pay a lower monthly fee but a higher per unit charge, or to pay a higher monthly 196 IMPACT OF MARKET STRUCTURE ON SERVICE OPTIONS FOR THE POOR fixedl fee and a lower per unit charge. If well designed, optional tariffs can lower customers' bills while at the same time increasing the utility's prof- its. Monopoly providers seldom implement such services, because they seem risky; however, in competitive markets the practice is common, for example, in the telecommunications market in the United States. Figure 5.4 shows a theoretical example of an optional tariff. The solid line represents the standard tariff, which is based on a fixed price per unit. The dashed line represents the tariff option designed for low-income con- sumers Here the unit charge for low levels of consumption is below the sLandard tariff, but after an initial lifeline block, the unit charge is signifi- cantly higher than the standard tariff. Such optional tariff mechanisms can be an incentive-compatible, viable alternative to subsidized tariffs. Monopoly distributors could offer them, but seldom do. Where competition between retailers exists, optional tariffs are more commonly offered, as entrants and innovators use them to com- pete with the incumbent. Improved Payment Security. Utilities are often unwilling to provide service to predominantly poor areas because of the risk of nonpayment. Alterna- tive payment security mechanisms would improve the likelihood of pay- ment and reduce the risk to the service provider. Monopoly utilities seldom offer innovative security mechanisms, but new entrants could. For example, a large furniture chain with a strong system for extendling and collecting installment payments could also become an elec- tricity retailer and could use household chattels as security. This allows more financial security in supplying those who dlo not own their homes. Prepayment arrangements also reduce the risk to the service provider. In adldition to increasing payment security for the provider, they simplify budg- eting for the poor household. Again, allowing new entrants is often necessary to drive this kind of innovation. For example, dominant providers of cellular telephone setvices in the United States would not sell to ghetto areas, but competition allowed innovative providers to develop a strategy that provided service and was profitable. These companies sold cellular handsets to cus- tomers below cost, taking a risk that the customers' use of the service would compensate for this. They then presold blocks of calls to customers in afford- able amounts. As a consequence, the vendors had security by means of pre- payment and the customers had a service that was affordable and flexible. 197 '.~~~~~~~~~~~~~~c .... .... . .\.... ..... .... EGb CO _ ON CONu e E 0 o 60 ° - NJ 00 °o °CoO° NJ rto u nvvO rN o CD .0 m O 0 - q NJ 00 O 0 U J O ON to . CON 00 N - O - ON _ N. o _ o 00 Co Coo _CSo ON 0 -0o o X Ch C lo N o U N o w ~ ~ a, Ch _ _ N- N_ OC C C C C Co to 09 W to o Co 00 _ COo 00 t5oO_ o Q1 rq CD 0 C> C ON E- - C. ON oN 0 NJ ' ° 0 0 to R - C u) UN tn 00 b to 00 0 ON 0 0 I. c noo c ooo oc UN u N N- ON O U r_N o (0 V~~~~~~~~~~~~~~~~~~~ C~~~~~~~~~~~~~~~~~~ °c-O 0rs o% UN N UN INr 0 N- IN 0 Co 0 to tnoo D~~~N UN 00 ON 0 0Co 00 o t O IN to nN 0 0 0 0 N 0 C? UN z ON UN O _ O _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _2 I o - c U lU : C to >n Co N-o NJ r >1 I ° D z Z W ~ °-, 4-' - I,~~~~~~~~~~~~8 QUENTIN WODON, MOHAMED IHSAN AJWAD, AND CORINNE SIAENS households. For instance, 24.5 percent of all households in the 1st or poor- est decile have access to electricity provided by ENEE compared with 93.3 percent of households in the 10th richest decile. Also as expected, house- holds in urban areas have higher access rates than households in rural areas. The fact that richer households have higher rates of access to elec- tricity than poorer households is one of the reasons why from the point of view of poverty reduction, and in comparison with other programs that could be implemented to fight poverty, electricity subsidies may not be well targeted. The design of the electricity subsidy in Honduras follows a basic self- targeting structure, that is, households self-select themselves for inclusion in the subsidy scheme by not consuming more than a certain amount. This implies that neither the utility nor the government needs to design a more complex targeting mechanism based on means-testing, whereby they use household characteristics as proxies to identify who is poor and who is not, and thereby who is eligible to receive the subsidy and who is not. As of July 2002 the subsidy was provided to all households that purchased electricity from ENEE and consumed less than 300 kWh per month, the electricity consumption lifeline. Specifically, households consuming less than 300 kWh paid the price of electricity at the 1994 rate plus an increase of 16.31 percent approved in 2000. Households that consumed more than 300 kWh per month paid the full price at the 1997 rate plus the same 16.31 percent increase. The government paid the difference between the 1994 and 1997 prices (L 23.8 million per month in mid-2002) directly to ENEE.' To assess the targeting performance of the subsidy, that is, what share of the subsidy goes to the poor, we combine administrative information with household survey data. On the administrative data side we rely on infor- mation provided by ENEE. This information includes (a) the percentage of residential clients by consumption level, (b) the average amount of elec- tricity consumed by clients at each level, (c) the total electricity bill with- out subsidy by level, and (d) the total subsidy paid by level. On the household data side, to estimate the proportion of poor households at the various consumption levels we use data from a survey implemented in 1999 by the Family Assistance Program (Programa de Asignaci6n Familiar or PRAF), with financing from the Inter-American Development Bank and assistance from the International Food Policy Research Institute. The survey was implemented in the bottom half of Honduras's municipali- 282 LIFELINE OR MEANS-TESTING' ELECTRIC UTILITY SUBSIDIES IN HONDURAS ties in terms of malnutrition as measured by the Ministry of Education's annual census of weight and height luiring the first year of primary school. The survey has modules on expenditures, eclucation, health, and the impact of Hurricane Mitch. The advantage of this survey is that it includles data on electricity consumption, which are not available in the labor force sur- vey. 'lThe disadvantage is that the survey is not nationally representative, but we assume that the results obtainedl lere would also apply if the survey were nationally representative.2 'Table 8.2 provides the results. For example, according to ENEE's admin- istrative records, the share of households coniected to its netvork with monthly consumption below 20 kWh is 20.31 percent (I 15,723 house- holds). Of these wve estimate from the PRAF survey that 44.93 percent are poor. That is, among all households that consume less thani 20 kWh accord- ing to the PRAF survey, about 45 percent have a level of per capita con- sumption that is below a reasonable poverty line for Honduras. With aver- age consuiiiption of 3.36 kWh per ouLiseliold per month, the total consumption for this group is 388,626 kWh. Without the subsidy this group would have to pay a total bill of L 929,256, but this bill is reduced to L 333,282 when the subsidy of L 595,973 is taken into account. Presenting statistics on errors of inclusion and exclision is conimon when assessing the targeting performance of a subsidy (see, for example, Cornia and Stewart 1995). There are various wvays to present these two types of errors. The simplest wvay, used in table 8.2, consists of computing the share of householdls in poverty and receiving the subsidy as well as the share of householcls not in poverty that are also receiving the subsidy. In general, as errors of inclusion increase, errors of exclusion decrease, ancd vice versa. If all households receive the subsildy, there are no errors of exclusion, but the errors of inclusion will likely be large because all non- poor householcls benefit from the subsidy. In table 8.2, among those households that are connectedl to ENEE, 60.19 percent are nonpoor and receive the subsildy. l3y contrast, only 1.68 percent are poor and do not receive the subsidy (nationally, the share of the population in poverty not receiving the subsidy is larger, because many poor householdls are not connectecl to the grid). In addition, 23.28 percent of households that are poor receive the subsidy and 14.85 percent of hiouseholds that are nonpoor do not receive the subsidly. From these results we can estimate that thie ratio of poor versus nonpoor beneficiaries 283 DC °- 1- ao o ° ° - V - C C CL0 0 . - 0 CD O _ . , N 0~ D a. °C o °'- a, | -6 C - ~~~~0 ., CD. N~~~ 0~ C :BOX S bI o 0 c E _ _ N0 _ V} I e = I =X U '0 .2'z_v rg 0_n LA i- C 00 a, E~~~~~~~C m 0 a0 ( a, a, w - CL O p_ D V c- c a . _=- ° O NV (Vi U ,U, E28 0~~~~. c U, 0 (N~~C N~ CD, C> 40 E > >~~~~~~ I- C 3' X , 0 4 0 U 284 LIFELINE OR MEANS-TESTING) ELECTRIC UTILITY SUBSIDIES IN HONDURAS is 0.39 (23.28/60.19), andl so the number of nonpoor householcis receiving the subsidy is more than twice as large as the number of poor households receiving the subsidy.I The most important statistic, however, is the share of the subsidy given to the nonpoor. This tells us how much poverty reduction is obtained (in terms of the poverty gap as dlefined in appendix 8.1) for each lempira spent on the subsidy. As computed in the last column of table 8.2, this share is above 80 percent. Part of the reason for the subsidy's poor targeting per- formance is related to the level of the lifeline threshold, which is set too high. Some 83.5 percent of households with access to electricity consume less than 300 kWh per Enonth, and hence qualify for the subsidy. Further- more, while the level of poverty is higher among households that consume less than 100 kWh, most of the subsidy is spent on households that con- sume between 100 andl 300 kWh per month, but these householdls are less likely to be poor. In part because the subsidy is not well targeted, its impact on poverty is small. This is Illustrated in table 8.3 wilich provides poverty measures with and without incorporating the value of the electricity subsidy in the overall consumption aggregate. The table provides three measures of poverty: the headcount inclex (the share of the population in poverty), the poverty gap (the distance separating the poor from the poverty line), and the squareti poverty gap (appendix 8.1 provides a formal definition of these poverty measures). Trie measures are proviclecl for two alternative poverty I ines cor- responding roughly to the extreme poor (L 400 per person per monthi) and the poor (L 600 per person per month). Overall the changes in poverty when subsidlies are taken into account are small, and these changes are likely to be slightly overestimated, because we do not take substitution effects due to the subsidy into account (if the subsidy were eliminated, electricity prices would go up and householdls would substitute consump- tion toward other goods). Althoughi this is not reported here, we carned oult additional simulations to assess if the results are sensitive to the choice of the PRAF suivey for the analysis. That is, using a large set of variables common to both the PRAF and the nationally representative EHPHM survey, and fitting a predictive model of electricity consumption in the PRAF survey, ve obtained a predic- tion for electricity consumption in the EHPHM survey, and redidl estimations regarding targetinig performance and the impact of the subsidly on poverty 285 QUENTIN WODON, MOHAMED IHSAN AJWAD, AND CORINNE SIAENS TABLE 8.3. Impact of the Electricity Subsidy on Poverty, Honduras, 1999 Without subsidy With subsidy kWh consumed/ Headcount Poverty Squared Headcount Poverty Squared month (%) gap (%) povy gap (%) gap (%) pov. gap Poverty line of L 400/personl month 0-20 44 93 12 42 5 99 44 93 12 29 5 92 20-100 36 00 10 45 426 35 66 1019 4 11 100-1 50 20.57 6 06 2 61 16 82 5 60 2 35 150-200 10 98 2 67 0 93 1098 2 24 0 72 200-250 15 64 5 32 2.00 15 64 4 56 1.51 250-300 17 09 3 06 1 02 17 09 2 38 0 79 More than 300 1015 2 70 112 1015 219 087 Poverty line of L 6001personl month 0-20 71.01 29 00 14 56 71 01 2885 14 44 20-100 63 47 23 70 11 60 63 47 23 36 11 36 100-150 44 74 13 97 675 43 39 13 29 6 34 150-200 31 26 819 335 2745 747 295 200-250 35 80 1315 6 05 3416 11 88 5 21 250-300 2915 10 41 437 2915 9.36 3 73 More than 300 17 97 6.33 3 03 17 97 5 65 2 60 Source Authors' estimates using 1999 PRAF survey data with these predictive values for electricity consumption in the EHPHM sur- vey. The results obtained using this procedure were fairly similar. 8.3 Alternative Targeting Indicators The evidence suggests that Honduras's lifeline subsidy is badly targeted and therefore fails to benefit the poor very much. In this section we com- pare the lifeline targeting technique to other means of targeting that could use more and better information to determine eligibility for utility subsidies. 286 LIFELINE OR MEANS-TESTING? ELECTRIC UTILITY SUBSIDIES IN HONDURAS As mentioned in the previous section, investigators often analyze the tar- geting performance of any given indicator, such as the lifeline level of elec- tricity consumption usedi in Honduras, using simple summaiy statistics such as the errors of inclusion and exclusion for a given targeting mecdia- nism. A generalization of this approach consists of using ROC curves to assess which indicator-in our case lifeline versus various potential means-testing mechanisms-has the best performance in identifying the poor. More precisely, the idea is to use simple categorical regressions to assess how various targeting indicators predict the probability of being pooi; and to see how the two types of errors (exclusion of some poor house- holds and inclusion of some nonpoor households) vaiy with the choice of a particular level of thie indicator to determi-ne eligibility. In some cases one can find a best overall eligibility criteiion independently of the weightirig of the two types of errors in policymakers' objective function. In other cases some weighting scheme is needed, and for any given weightling scheme, the ROC curve can help select the best indicator. Our objective here is not to discuss the method in detail (see appendlix 8.2 for an outline of the basic idea behind the ROC curve). Rather, we focus on the empirical results for Honduras's electricity subsidy. For each indicator that can be used for targeting (lifelinie ot other), one associates a curve that plots the probability that a poor houselhold will be classilied as poor against the probability that a nonpoor househiold will be classified as poor for every possible value given to the indicator. Note that the indicator can be complex, that is, it can consist of a combination of indlicators, as the regiession can be multivariate. If the ROC curve lies on the 45 degree line, the model has no predictive power, because the probability that a poor household would be classified as poor is no higher than the probability that a nonpoor hiousehold would be classified as poor. The more the ROC culve bows upward, the greater the model's predictive power. A summary meas- ure of predlictive power is the area underneath the ROC curve. If the area is above 50 percent, then the model has some predictive power. An area of 100 percent implies that the model predicts poverty perfectly. We used the methiodology to assess how well various indicators per- formecd for identifying thle poor among the sample of households with a connection to the electricity grid in Honduras. We employed poverty lines of L 400 (extreme poverty) and L 600 (poverty) to define the poor. The first model in table 8.4 (household characteristics) combines infor- 287 QUENTIN WODON, MOHAMED IHSAN AJWAD, AND CORINNE SIAENS TABLE 8.4. Areas under ROC Curves for Alternative Targeting Mechanisms, Honduras Performance in identifying Performance in the extreme poor identifying the poor (area under ROC curve. (area under ROC curve, Model percent) percent) Household charactenstics 87 83 Demographics 72 71 Educational attainment 71 72 Employment status 69 66 Geographic location (department) 66 63 Housing characteristics 82 81 Size of house 77 77 Quality of house 72 72 Access to water and sanitation 61 58 Electricity consumption 70 73 Note A larger area indicates better targeting performance Source Authors' estimation mation on a number of household characteristics, including cleniograph- ics, education, employment status, and geographic location. The model is better at identifying the extreme poor (area under the ROC curve of 0.87) than the poor (area of 0.83). Withlin these household characteris- tics (that is, with separate models with subsets of variables), demograph- ics and education variables are better than employment and location variables at identifying the poor. Housing characteristics can also be used to identify the poor, with a similar level of performance (area under the ROC curve of 0.82 for the extreme poor and 0.81 for the poor). Within housing characteristics, the size and quality of the house are better at identifying the poor than other characteristics. Finally, the lifeline threshold (related to the level of energy consumption in the household) has some predictive power (the area under the ROC curve is above 0.5), but less so than some other easily identifiable variables. The bottom line is that if the objective is to target the poor, variables are available that are better at doing so than the level of energy consumption (see appendix 8.2 for examples of actual ROC curves). 288 LIFELINE OR MEANS-TESTING7 ELECTRIC UTILITY SUBSIDIES IN HONDURAS While it is not surprising that hoLisehold or housing-based targeting indi- cators wouldl be better al identifying the poor than households' level of energy consumption, one mighlt believe that for a service provider or utility to gather such information could be difficult or expensive. Clearly means- lesling (using correlates of poverty for targeting) requires information, and gathering this information requires effort. Howvever, experiences in other countries suggest that the cost of doing so need not be very high if the same type of information is used for targetinig a range of social programs rather than utility subsidies only. More specifically, one clear possibility for reducing the administrative cost of means-tesling is lo use a single system of means-testing at the national level for many different programs. In Latin America this has been done witlh some success in Colombia (the System for Selection of Benefi- ciaries) and in Chlile (the Committee for Social Municipal Assistance or CAS), among others. In Chile, for example, as documentedl by Clert and Wodon (200l), the CAS system is used as a targeting instrument not only for water subsidies, but also for the family income subsidy, the social hous- ing subsily, and the pension subsidy scheme. Because the fixed adminis- tratLive costs are spread across several programs, the CAS is cost-effective. In 1996, for example, administrative costs represented a mere 1.2 percent of the benefits distributed using the CAS score. If the administrative costb of the CAS system hadl hadl to be borne by the water subsidy scheme alone, they would have represented 17 8 percent of the value of the subsidies. The cost of interviews for determining eligibility for the subsidies was US$8.65 per household, ancd the Ministry of Planning estimates that 30 percent of Chilean householdls undlerwent interviews, which seems reasonable given that the target group for the subsidy programs is the poorest 20 percent of the population. 8.4 Conclutsion Governments have a range of options foi helping to reduce poverty. Simi- larly, private uLilities have a number of optiotns for helping their low-income customers. While decisions about the choice of a specific instrument are based on various criteria, intervenitions whose benefits are immediate, visi- ble, andl acministratively easy to implemenit and are supported not only by 289 QUENTIN WODON, MOHAMED IHSAN AJWAD, AND CORINNE SIAENS the poor, but also by the nonpoor or not as poor, are attractive from a politi- cal economy point of view.4 Lifeline subsidies for basic infrastructure or utility services have all these characteristics. The subsidies may take various forms, but their key characteristic is that they are provided to all customers with a consump- tion level below a minimum threshold considered necessary for meeting basic needs, hence the use of the "lifeline" expression. Lifeline subsidies have an immediate impact by reducing beneficiaries' expenditures for a given level of provision. T'he benefits of the subsidies are easily understood (even though their costs may not be). Lifeline subsidies often enjoy wide- spread political support, especially when the lifeline threshold is set suffi- ciently high so as to benefit the less poor as well as the poor or the median consumer as well as the low-income customer. The subsidies are easy to implement at relatively low administrative cost because no means-testing is involved. For poverty reduction, however, while the characteristics of lifeline sub- sidies help to muster support, they may not ensure effectiveness or a good cost-benefit ratio. Indeed, one of the major drawbacks of lifeline subsidies is that they may not be well targeted. When they are well designed, life- lines can reach the poor through self-targeting. That is, if the lifeline threshold is low enough, only those who consume little will be eligible, and these customers may be comparatively poor. In many instances, however, the leakage of lifeline subsidies to the nonpoor is such that it dilutes the effectiveness of the policy for poverty reduction. In this chapter we provided a partial evaluation of the lifeline or increas- ing block tariff electricity subsidy in Honduras. With funding from the gov- ernment, the public utility is offering electricity at greatly subsidized rates for those households with monthly consumption below 300 kWh. Because the lifeline threshold is set so high, 83.5 percent of the utility's residential clients benefit from the subsidy. At the same time, 81.8 percent of the sub- sidy may well be spent on nonpoor households. While this last statistic could be lower if we were using a different method for measuring poverty, it remains true that the impact on poverty of the subsidy is rather small in comparison to its cost. The fact that the current subsidy is badly targeted does not mean that it could not be improved by reducing the lifeline thresh- old. A lower lifeline subsidy as currently being considered by the govern- ment would have the potential of being more effective. Alternative proxy 290 LIFELINE OR MEANS-TESTING? ELECTRIC UTILITY SUBSIDIES IN HONDURAS means-testing targeting mechanisms based on household or housing char- actelisties could also be used to improve targeting. Nevertheless, experi- ence in other countries such as Chile suggests that even witlh better means- testing, other types of interventions would probably have a better impact on poverty per dollar spent than utility subsidies. Appendix 8.1: Defirnition of Poverty Measures Tlhis appenhdx, whilch is reproduced with minor changes from Coudouel, Hentschel, and Wodon (2002), provides mathematical expressions for the poverty measures useci in table 8.3. Poverty Headcouitt This is the share of the population that is poor, that is, the proportion of the population for whom consumption or income y is less than the poverty line z. Suppose we have a population of size n in which q people are poor. Then the heaclcount index is defined as H = n . n Poverty Cap The poverty gap, whlich is often considered as representing the clepth of poverty, is the mean distance separating the population from the poverty line, with the nonpoor being given a distance of zero. The poverty gap is a measure of the poverty deficit of the entire population, where the notion of poverty deficit captures the resources that would be neededl to lift all the poor out of poverty through perfectly targeted cash transfers. It is (lefinedl as follows: G 1 [z-Yj where y, is the income of inclividual t, ancl the sum is taken only on those individuals who are poor. The poverty gap can be written as being equal to the product of the income gap ratio and the headcounit index of poverty, where the income gap ratio is itself defined as PC = I *H, with I= ZYq qwhere yq =-E y, is the average inconte of the poor. L= q 291 QUENTIN WODON, MOHAMED IHSAN AJWAD, AND CORINNE SIAENS Squared Poverty Gap This is often described as a measure of the severity of poverty. While the poverty gap takes into account the distance separating the poor from the poverty line, the squared poverty gap takes the square of that distance into account. When using the squared poverty gap, the poverty gap is weighted by itself, so as to give more weight to the very poor. Said differently, the squared poverty gap takes into account the inequality among the poor. It is obtained as follows: P2= I Z Y'] It is important to use the poverty gap or the squared poverty gap in addi- tion to the headcount for evaluation purposes, since these measure differ- ent aspects of income poverty. Indeed, basing evaluation on the headcount ratio would consider as more effective those policies that lift the richest of the poor (those close to the line) out of poverty. Using the poverty gap PC and the squared poverty gap P2, on the other hand, puts the emphasis on helping those who are further away from the line, the poorest of the poor. Appendix 8.2: [dentifying &he Targetiing Performance of Various I[ndicators 'Using ROC Curves Following Wodon (1997), denote by P, P-, and P+ the number of the poor, the number of the poor classified as nonpoor, and the number of the poor classified as poor by a model. Also denote by NP, NP-, and NP+ the num- ber of the nonpoor, the number of the nonpoor classified as nonpoor, and the number of the nonpoor classified as poor. Sensitivity SE = P+/(P- + P+) = P+/P is the fraction of poor households classified as poor. Specificity SP = NP/(NP- + NP+) = NP-INP is the fraction of nonpoor households clas- sified as nonpoor. The errors of inclusion and exclusion can be defined as 1 minus SP and I minus SE (other definitions could be used as well, but ROC curves are based on these definitions). Nonpoor Poor Predicted nonpoor SP = NPJ (NP- + NP') 1 - SE = P1(P + P1) Predicted poor 1 - SP = NPI (NP- + NP+) SE = P+I(P- + P+) 292 c 0 E Q 0, c C1 o ~~~~~~~5~~~~~- O o C o C~~~~~~~~~~~~~~~~~~~~~~~a Ci,CD X o C o o o o o , 0O -~ VM ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~C o E _ C 0~~~~~~~~~~~~~~~~~~~~~~~C (UJ o Li C 2 U.S ..* 0C C: 0 - CQ Li Li LL 0I 293 QUENTIN WODON, MOHAMED IHSAN AJWAD, AND CORINNE SIAENS When using a statistical package and running a probit or logit regression for poverty, each observation is given an index value equal to the predicted right-hand-side of the regression. This predicted value is used to classify the households as poor or nonpoor, with the computer typically using one- half as the cutoff point, which we will denote by c (those above the cutoff point are classified as poor). However, this cutoff point can be changed. A ROC curve is a graph that plots SE as a function of 1 - SP for alternative values of the cutoff point. Figure A2.1 shows the ROC curves estimated for those with access to electricity in Honduras with two different models: the housing model and the level of electricity consumption of the households (which is a generalization of the lifeline used for targeting the subsidy by ENEE). At the origin, c = 1, SE = 0, and SP = 1. At the upper right corner, c = 0, SE = 1, and SP 0. The higher the ROC curve, the better its pre- dictive power (a 45 degree line has no predictive power while a vertical line from the origin to the top of the box followed by a horizontal line until the upper right corner has perfect predictive power). Clearly the housing model performs better than the level of electricity consumption of house- holds in identifying the poor. The area below a ROC curve provides a summary statistic of the predic- tive value of the underlying model. An area of 0.5 corresponds to the 45 degree line, which has no explanatory power. An area of 1 corresponds to perfect prediction. lf the ROC curve of one targeting indicator or set of indicators lies above the ROC curves of all the alternatives at all points, that indicator will typically be the best to target the poor for the class of social welfare functions based on the two types of errors that can be com- mitted through targeting. If two ROC curveb intersect, the choice of the best indicator will depend on the normative weights the policymaker attaches to the two types of errors. Notes 1. The tariff shucture for 1997 distinguishes betweeni households consuming more or less than 500 kWh per month For those households that consume less than 500 kWhl per monitlh, the price is a flat ratc of 1. 6.9 for the first 0 to 20 kWh. Thereafter the unit price per kWh is L 0.6979 for 20 to 99 kWh, L 1.0173 from 100 to 299 kWh, and L 1.1829 from 300 to 499 kWh For lhouselholds consum- ing more thain 500 kWh per mointlh, the flat rate for the first 20 kWh is 7 0800 Lenipiras. Then the uiiit rate per kWh is L 0.7161 for the next 80 kWh, L 1.0438 294 LIFELINE OR MEANS-TESTING7 ELECTRIC UTILITY SUBSIDIES IN HONDURAS for the next 200 kWh, l, 1 2137 for the next 200 kW'h, and L 1 3352 above 500 kVWh. 2. Using variables comimiilon to the PRAF stirvey adli the nationially iepresenta- tive EHPHM labor force survey, wve pieclicted energy consumption in the ElIPHM tisirg a model fitted In the PRAF survey. The results obtained wvith tile EHPI-IM suIvey for the assessnment of the targetinig perforimanice of the subsidy andI its impact on povert) vere siiilar 3. Anothier wvay of dlefiniEig the errors of inClUsioii ailI excIUsioii consisIs of con- siciering the fraction of subsidy recipieiits that are iionpoor as errors of inclusion and the fraction of houIseholcis that are not recipients but are poor as eriors of exclusion. According to this alterinative dlefiiiition, the eriors of inclusioli ate equLal to 0 72 [60.19/(60.19 + 23.28)], whiile the errois of exclusion are equal to 0.10 [1.68/(1.68 + 14.85)]. 4 For Issues relating to targeting, its costs, and( the Interplay wvilth the political econonily, see, for iiistanice, Besle) and KaiibuL (1993); Sell (1995). References The word "processed" describes informally reprodclued works that may not be commonly available through libraries. Besley, T., andc R. Kanbur. 1993. "The Principles of Targeting." In M. Lipton andl J. Van cler Gaag, eds., hIcludinig the Poor. Washington, D.C.: World Bank. Clert, C., and Q. Wodon. 2001. "The Targeting of Government Programs in Chile: A Quantitative and Qualitative Assessment." In E. Gacitua- Mario andl Q. Wodon, ecds., Measuremenit and Meaning: Combtr2imng Quantitative and Qualitative Methods for the Analysis of Poverty and Soctal Exclusion in Latin America. Technical Paper no. 518. Washing- ton, D.C.: World Bank. Cornia, G. A., and F. Stewvart. 1995. "Tvo Errors of Targeting." In D. Van de Walle and K. Nead, eds., Public Spending and the Poor: Theory and Evidenice. Baltimore, Maryland. The John Hopkins University Press. Coudouel, A., J. Hentschel, and Q. Wodon. 2002. "Poverty Measurement and Analysis." In J. Klugman, ed., Poverty Reduction Strategies Source- book Washington, D.C.: World Bank. 295 QUENTIN WODON, MOHAMED IHSAN AJWAD, AND CORINNE SIAENS G6mez-Lobo, A., and D. Contreras. 2000. "Subsidy Policies for the Utility Industries: A Comparison of the Chilean and Colombian Water Subsidy Schemes." University of Chile, Department of Economics, Santiago. Processed. ENEE (Empresa Nacional de Energfa Electrica). 2002. "Analisis del sub- sidio que ortoga el Gobierno Central de Honduras a los Abonados Resi- denciales con consumos iguales o menores a 300 hWh/mes." Teguci- galpa. Processed. Estache, A., V. Foster, and Q. Wodon. 2002. Accounting for Poverty in Infrastructure Reform: Learning from Latin America s Experience. Devel- opment Studies. Washington, D.C.: World Bank Institute. Foster, V., and C. Araujo. Forthcoming. "A Case Study from Guatemala." In V. Foster and Q. Wodon, eds., Energy and Utiltty Reform: Lessons from Latin America for Poverty Reduction. Technical Paper. Washington, D.C.: World Bank. Sen, A. 1995. "The Political Economy of Targeting." In D. Van de Walle and K. Nead, eds., Public Spending and the Poor: Theory and Evidence. Baltimore, Maryland: The John Hopkins University Press. Wodon, Q. 1997. "Targeting the Poor Using ROC Curves." World Develop- ment 25(12): 2083-92. World Bank. 2001. "Honduras: Poverty Diagnostic 2000." Report no. 20531-HO. Washington, D.C. 296 frim woA. Wa ho7lx-w ^>o) (vw)) "ml-o), ('r,"xmiro r1pfow Edward Atm Strud wackeI 00o0 ffSN 1 op0o Om@4 qjEmbwEr rX ',IjTQ Frot2pft'.-Oor C' ThoXf i(moAdi pmMC 'llm- '0330 aSHI NGT dto C1 wt 1-4 __Max_ ) a2l 153 amwo&0 VCgtt3 ma;wep C'ww+ wq6.a ',woo4 g.Mr Br vb 8J8,'I1 9qf,g 7.^0 ,C58082 35t2 f,r __I U 'Iro 'tilUIG^ # t(-IWVv *01 M K) reoro,# 10 fMuc d w. xwcrscoG Ot mt,/ (0,4Mt%S fMo pti7csmImo 0'$wtt fr >t.t.'I4-/'."f-h3: v" tq,C t.t o@ f-F$wl6>3: 4@ilt lnl.lg; - \#,qnKle= d3 Orfl 4:t't c.T4$¢3(( Om>5t, Lw rfenx owav"fg5)7aY%o Mv5p l.moul° mf W" (0§.5 tib-m¢i "'oJ F''8(x*,^t5WR,pft t.T,-, tfoJ .t5tq.t Edward A Strudwicke 00803 ISN 4 fm >j ' ( <>9 780821 353424 _ I[~ ~ ~ ~ ~~~M '11 (M5 , ,0 "51>ut.,C,5p