Privatesector P U B L I C P O L I C Y F O R T H E Note No. 210 May 2000 Measuring the Impact of Energy Reform—Practical Options Vivien Foster Government interventions in energy markets have many effects on the poor. But there has been little measurement of these effects, making it hard to know exactly what the effects of a project have been, and hard to compare those of different interventions. This could be rectified by building impact indicators into energy projects at the design phase—and doing so consistently and systematically, across countries and over time. This Note discusses the development of suitable indicators. First, agreement is needed on workable definitions of poverty and what would constitute welfare improvements for the poor. Then there must be explicit hypotheses on how specific elements of energy projects, individually or together, affect the poor. Finally, the indicators must be based on data that can realistically be collected in real-life low-income communities, in real-life developing countries. Following a decade of energy sector reforms in The aim of this methodology is not only to make many developing countries, people are increas- it easier to answer questions about how energy ingly questioning how much these reforms have sector interventions have affected the poor. It is benefited the poor. That question has proved dif- also to help focus attention on poverty issues ficult to answer, in part because of the lack of a before interventions are made, encouraging the framework for thinking about the issue and in adoption of pro-poor features in the design. part because of a shortage of suitable data. The proposed approach has two stages. The first This Note proposes a methodology for measur- is to provide a set of welfare indicators suffi- ing the impact of interventions in the energy sec- ciently broad to capture the kinds of energy tor on the welfare of poor households. Here, issues likely to concern poor households. The energy sector interventions refer to any measure second is to calculate the value of these indica- that significantly affects the cost, quality, and tors for poor households before and after the conditions of access to energy services, whether intervention to gauge its effect on their welfare. wholesale sector reform or a small investment The process depends critically on the availabil- project. These interventions include restructur- ity of data sets that combine information about ing, privatization, and liberalization of traditional energy use with indicators of poverty (Gomez- electric and natural gas utilities. They also Lobo, Foster, and Halpern forthcoming; Lovei include policy decisions affecting the availability and others 2000). and relative prices of alternative energy sources, both traditional biomass and commercial fuels— Measuring the welfare impact of energy sector perhaps of more immediate relevance to poor interventions on the poor is not quite the same households. as measuring the impact on poverty. For The World Bank Group ▪ Private Sector and Infrastructure Network 2 Measuring the Impact of Energy Reform—Practical Options example, an energy pricing reform might reduce nomic sense to use electricity for lighting but the cost of electricity to poor households, LPG for cooking, for example. directly increasing their welfare. The same price change might indirectly take some of these All this means that any indicators measuring the households out of poverty—by releasing welfare impact of energy sector interventions on women and children from the time-consuming the poor need to consider a household’s full task of gathering traditional fuel, or by raising range of energy sources rather than focusing on productivity in household chores or in the oper- a single source. Many of the traditional indicators ation of home-based microenterprises. Though tend to concentrate narrowly on electricity—for measurable in principle, this ultimate effect is example, measuring the number of household much harder to gauge with any reliability connections or the share of household spending (Chong and Hentschel 1999). In particular, it is on electricity. This overlooks the fact that inter- difficult to attribute changes in poverty to one ventions affecting the prices and availability of intervention rather than another. Thus the more different fuels may affect the welfare of poor modest objective of examining how energy sec- households just as much as electricity sector tor interventions directly benefit the poor is reforms, if not more so, even after households probably also more useful for impact evaluation. obtain an electricity connection. The following section broadens some of the traditional electric- Stylized facts about energy and ity-based indicators of welfare to encompass the poverty full range of fuels used by households. A good place to begin is with a brief review of Indicators of the welfare impact some stylized facts on energy consumption and poverty (Albouy and Nadifi 1999). The energy To choose an appropriate set of indicators also literature has traditionally been dominated by a requires a working definition of human welfare theory of transition in which households gradu- as it relates to interventions in the energy sector. ally ascend an “energy ladder.” The ladder Consistent with the literature, this section takes begins with traditional biomass fuels (firewood three different perspectives on human welfare— and charcoal), moves through modern commer- basic needs, monetary, and nonmonetary (Lok- cial fuels (kerosene and liquefied petroleum gas, Dessallien 1999). or LPG), and culminates with electricity. The ascent of this ladder, though not fully under- For many of the indicators discussed here, it will stood, is thought to be associated with rising often be necessary to calculate the shares of total income and increasing urbanization. household energy consumption represented by different energy sources. In doing so, it is essen- But the empirical work on energy and poverty tial to take into account that different types of has found that reality is more complex than this fuel have different efficiency factors, ranging simple transitional theory suggests. At any given from 10 percent for fuelwood to 65 percent for time households tend to rely on a range of fuels electricity (Leach and Gowen 1987). Effective that typically encompass at least two of the steps energy consumption refers to the energy actually on the energy ladder (Barnes and Qian 1992; consumed by the household—after efficiency Hosier and Kipondya 1993; ESMAP 1994; factors have been taken into account—rather Eberhard and van Horen 1995). There are sev- than the energy purchased by the household. eral possible explanations for this. One is that unreliable supplies require households to rely Basic needs on diverse sources of energy. Another is that dif- ferent energy sources are more cost-effective in According to one traditional view, welfare some uses than in others, so it may make eco- relates to people’s ability to satisfy their most The World Bank Group 3 BOX 1 ENERGY IN THE BASIC NEEDS APPROACH Many Latin American countries have traditionally measured poverty using multidimensional indexes of unsatisfied basic needs. The indexes vary from country to country, but generally include basic material needs. While intuitively appeal- ing, this view involves subjectivity in defining a measures of sanitation, housing quality, and educational attain- basic need (Hicks 1998). For the energy sector, ment. A recent survey in Latin America found that among thirteen it raises two questions: To what extent can countries, only three—Bolivia, Panama, and Peru—had indexes energy be regarded as a basic need? And how that included an electricity connection as a basic need (Hicks should a basic energy need be defined? 1998). While policymakers have sometimes defined an electricity connection as a basic need (box 1), this view conflicts with households’ tendency to extent to which households have to rely on a use a wide range of fuels even when electricity diversity of fuels. Concentration indexes can be is available. A more plausible definition of a calculated as the sum of the squares of the shares basic energy need would be reliable access to of different energy sources in a household’s one or more sources of energy. effective energy consumption. But such indexes should be interpreted with caution because fuel The most basic indicator of access is coverage of diversity may simply indicate that different fuel energy services. This indicator is widely used for sources are more cost-effective in different uses, electricity infrastructure, but less so for other rather than reflecting reliability problems. energy sources, where it is potentially just as useful. Access to traditional biomass and mod- Monetary ern commercial fuels is by no means universal, but may be limited by local environmental fac- The standard economic view is that the pur- tors and deficiencies in commercial distribution chasing power of the household (whether mea- networks (Barnes and Qian 1992). In addition to sured by income or consumption) provides the looking at coverage rates for different energy best overall indicator of welfare. Energy sector sources, it may be helpful to sum the number of interventions might affect economic measures of types of energy to which each household has well-being in several ways. The most direct way access, keeping in mind that access covers fuel is by reducing (or perhaps increasing) the cost sources that a household may choose not to use. of satisfying energy requirements and thereby increasing (or reducing) the purchasing power The basic coverage indicator says nothing about of a given household income. Households might the reliability of the service, however. A house- respond to the increase in purchasing power by hold may have an electricity connection but using more energy or expanding their con- receive the service only a few hours each day. sumption of other goods, leading either way to Access to other types of fuel may be similarly an improvement in economic welfare. intermittent and uncertain. A reliability index could be constructed by asking poor households A traditional monetary indicator of welfare, what share of the time they are able to obtain widely used in the electricity sector, is the share energy from a particular source. This informa- of household income (or expenditure) devoted tion can be aggregated across fuel sources by to energy. A large share is taken to imply an taking a weighted average of the reliability score unacceptable economic burden of meeting for each energy source, with the weights corre- energy requirements. sponding to the share of each energy source in the household’s effective consumption. Although relatively simple to calculate, this indi- cator compounds several different effects, com- A more indirect—but less information- plicating its interpretation. For example, a large intensive—way of gauging reliability is to use a share of energy expenditure could be due to consumption concentration index to capture the high consumption (reflecting large household 4 Measuring the Impact of Energy Reform—Practical Options BOX 2 COST OF MEETING ENERGY REQUIREMENTS FOR COOKING IN DAR ES SALAAM, TANZANIA A study of the costs of using alternative cooking fuels in Dar The second comparison is between the capital and operat- es Salaam, Tanzania, is interesting because it compares ing costs of using different fuels. The ranking of fuels from alternative measures of unit costs (Hosier and Kipondya most to least expensive is very different for capital and oper- 1993). The first comparison is between the financial and eco- ating costs. The capital costs range widely, with electricity nomic costs of different fuels, where the economic cost being by far the most expensive. Summing the economic cost adjusts for the distortionary effect of subsidies and duties of a notional cooking budget of 320 megajoules a month with and also takes into account the foreign exchange compo- the associated capital cost yields the total financial and eco- nent of imported fuels. The financial and economic costs dif- nomic costs. While electricity is the cheapest cooking fuel in fer substantially, particularly for electricity, which is heavily terms of financial cost, it becomes the most expensive in subsidized. terms of economic cost. BOX TABLE 2 FINANCIAL AND ECONOMIC COSTS OF COOKING FUELS IN DAR ES SALAAM, 1990 (Tanzanian shillings) Fuel cost Amortized Total monthly cost (per effective megajoule) monthly of 320 megajoules Fuel Financial Economic appliance cost Financiala Economicb Firewood 3.94 5.27 n.a. 1,259.35 1,686.40 Charcoal (traditional) 3.59 5.64 22.22 1,169.81 1,827.02 Charcoal (improved) 2.39 3.76 125.00 890.06 1,328.20 Kerosene 5.24 9.13 33.33 1,709.52 2,954.93 LPG 3.17 4.49 208.33 1,224.21 1,645.13 Electricity 0.62 10.38 458.33 657.99 3,779.93 n.a. Not applicable. a. Financial cost is financial fuel cost of 320 megajoules plus monthly amortized appliance cost. b. Economic cost is economic fuel cost of 320 megajoules plus monthly amortized appliance cost. Source: Hosier and Kipondya 1993. size, high levels of discretionary use, or low effi- An affordability index could then be defined as ciency of use), high unit prices of energy, or the share of households whose effective energy exceptionally low income. Each explanation car- consumption per capita exceeds the subsistence ries very different policy implications. threshold. The same information could also be expressed as the ratio of each household’s effec- Perhaps a more useful way of thinking about tive energy consumption per capita to the sub- the affordability of energy is to examine sistence threshold. whether households are able to purchase enough energy to meet subsistence require- To complement the affordability index, fuel costs ments. The subsistence threshold would need and fuel subsidies could be tracked over time to to be externally defined, based on what would see how energy pricing policies affect the rich be required to perform basic functions such as and the poor. This exercise gives rise to two lighting, cooking, and (depending on climate) more indicators: the average fuel cost per effec- heating.1 And it should be expressed in per tive unit of energy consumption (total household capita terms to take into account differences in energy expenditure divided by total effective household size.2 energy consumption) and the average subsidy The World Bank Group 5 BOX 3 HEALTH EFFECTS OF DIFFERENT ENERGY SOURCES IN SOUTH AFRICA A recent study reviewed the empirical evidence on the health and wider social impacts of different energy sources in South Africa per effective unit of consumption (calculated by weighting the unit subsidy on each type of fuel (Eberhard and van Horen 1995). Examining small-scale research by the share of that fuel in each household’s total projects that measured the intake of particulates among children, effective energy consumption). the study concluded that children living in urban homes relying on coal inhale more than five times the daily limit recommended by An important drawback of the average fuel cost the U.S. Environmental Protection Agency. Children living in rural measure is that it overlooks the costs of comple- homes relying on fuelwood inhale more than nine times the limit. mentary capital investments (such as lightbulbs and stoves) required to use the fuel productively. A health survey conducted as part of the study revealed that This can create a misleading impression, since children from coal-using homes are 190 percent more likely to some energy sources have low fuel costs but high develop lower respiratory illness (pneumonia, bronchitis, asthma) capital costs, and others the opposite. To the than children from electrified homes. Acute respiratory infections extent that poor households are credit con- are the second most important cause of child mortality in South strained, high capital costs may prevent them Africa. from taking advantage of fuels with overall lower costs. An average total cost per effective unit of A larger-scale health and safety survey of nonelectrified house- energy consumption can be estimated by adding holds in South Africa showed that about 6.5 percent had experi- the amortized capital costs of the durables used enced (sometimes fatal) incidents of paraffin poisoning of children. for cooking, lighting, and heating, as a study of Burns resulting from exposed flames in the household are the cooking fuels in Tanzania did (box 2). This study fourth most important cause of death for children in South Africa. also shows how the incidence of subsidies varies across different types of fuel in Tanzania. To produce a more informative measure of eco- and even education outcomes. In households nomic burden, some of the types of information relying on traditional fuels, indoor air pollution described above could be combined. For exam- may cause respiratory illness, and paraffin poi- ple, it might be interesting to track how the cost soning of children and serious burns have also of subsistence-level per capita consumption been documented (box 3). Although the link changes as a percentage of per capita income (or between energy and education has yet not been expenditure), or how the total subsidy received studied in depth, recent findings suggest that at a subsistence level of consumption changes as electric lighting significantly increases the time a percentage of household income (or expendi- poor children are able to spend reading and ture). These measures hold consumption con- studying (Domdom, Abiad, and Pasimio 1999). stant at a level thought to represent a basic requirement and thus avoid confusing quantity Where health and education effects are impor- and price effects. tant, two types of indicators could be used to measure them. The first type aims to measure the Nonmonetary exposure levels of poor households, in terms of indoor air pollutants inhaled or hours of reading In recent years there has been a trend toward (the second being somewhat harder to capture). complementing economic measures of depriva- The second type of indicator tries to capture the tion with nonmonetary measures to obtain a consequences of these exposures, such as the inci- multidimensional view of human well-being, dence of respiratory illnesses in poor communi- particularly by tracking health and education ties or the rate of grade completion among indicators. school-age children. With the indicators of con- sequences, while theoretically of greater interest, There is some evidence that interventions in the it becomes more difficult to isolate the effects of energy sector could have direct effects on health the energy sector intervention from those of 6 Measuring the Impact of Energy Reform—Practical Options TABLE 1 SUMMARY OF PROPOSED WELFARE INDICATORS Indicator Comments Basic needs Coverage indexa Whether or not a household has access to a particular energy source; may be The indicator does not take into account aggregated to give the total number of energy sources available to each household. reliability of supply. Reliability index Percentage of time on average that an energy source is available for use by a The indicator requires a subjective household; may be aggregated as a weighted average. household assessment of reliability. Concentration index The sum of the squares of the shares of different energy sources in a Fuel diversity captures more than mere household’s effective energy consumption. unreliability of fuel supply. Monetary Affordability indexa Percentage of households whose per capita effective energy consumption exceeds Determining the subsistence threshold a subsistence threshold, or ratio of a household’s per capita effective energy often involves much subjectivity. consumption to a subsistence threshold. Average fuel cost per effective unit of energya Total household energy expenditure divided by the household’s total effective The indicator fails to take into account energy consumption. the capital costs of using fuels. Average subsidy per effective unit of energya Average of the unit subsidy for each energy source weighted by the share of that energy source in the household’s total effective energy consumption. Average total cost per effective unit of energy Total household energy expenditure, plus amortized capital cost of durables used Calculating the amortized capital costs for cooking, heating, and lighting, divided by the household’s total effective of durables for the full range of fuel uses energy consumption. is likely to be complicated. Economic burden Average fuel cost per effective unit of energy multiplied by the subsistence threshold, divided by per capita income (or expenditure). Nonmonetary Exposure rates Health: Twenty-four-hour exposure rates to indoor air pollutants. Education: Hours of reading by schoolchildren. Incidence rates Health: Proportion of households affected by energy-related incidents of ill health, It is difficult to isolate the impact of such as respiratory illness, burns, and paraffin poisoning. energy sector interventions on incidence Education: Grade completion rates of schoolchildren. rates, which may be affected by many other factors. a. Among the most essential indicators presented. The World Bank Group 7 other factors that might also influence health and holds according to their relative position in the educational attainment. overall distribution of income (or consumption), by dividing the population into income (or con- Summary of indicators sumption) quintiles or deciles. Separate welfare indicators can then be calculated for each quin- Among the indicators for measuring the impact tile or decile. of energy sector reforms on household welfare, the access and affordability indicators will be rel- This approach also allows an assessment of the evant in most cases, while the broader health equity of interventions in the energy sector, by and education indicators may be of more inter- examining how benefits are distributed across est in some cases than in others. Calculating all income groups. The analytical tools for measur- the indicators in all cases may be neither feasi- ing inequality are already well developed in the ble nor desirable. To aid selection, the most income distribution literature (Cowell 1995). essential—and easily calculated—of the indica- Standard measures such as the Gini coefficient tors are noted in table 1. can be readily adapted to the energy sector, giv- ing rise to concentration coefficients that mea- Combining energy and poverty sure the extent to which distribution of services information departs from an equitable benchmark (Kakwani 1986).3 Although widely used in analyzing pub- All the indicators discussed above provide gen- lic expenditure programs, these analytical tools eral information on the welfare impact of energy have rarely been applied to the energy sector. sector interventions on any household. To say Box 4 describes an interesting exception. something about the welfare impact on the poor, it is necessary to calculate the indicators sepa- Implementation issues rately for the poor and the nonpoor. But which is more useful for this type of analysis, an While conceptually straightforward, many of the absolute or relative concept of economic proposed indicators are relatively data intensive. poverty? The availability of suitable data from existing sources and the cost of gathering additional data Many countries have developed poverty lines, are likely to be the main constraints in applying typically based on the cost of acquiring a basic this approach to assessing the welfare impact of basket of food and nonfood requirements energy sector interventions on the poor. (Ravallion 1998; Lanjouw 1999). International benchmark poverty lines also exist, such as the The ideal data set would have these three char- $1 a day and $2 a day lines adopted by the World acteristics (Gomez-Lobo, Foster, and Halpern Bank for extreme poverty and poverty. Poverty forthcoming): lines allow absolute judgments about which ▪ It would combine information on energy- households are poor and which are not, and related behavior with information on income thus analysis of how energy sector reforms affect or consumption. these two groups. ▪ It would record such information both imme- diately before and some time after the energy But constructing poverty lines is far from sector intervention for the same households. straightforward because of the difficulties of ▪ It would contain information both for house- establishing the basic basket of goods. holds that had been affected by the interven- Moreover, dividing the population into the two tion and for a control group that had not been. broad categories of poor and nonpoor may con- ceal important gradations within each group. Under less than ideal circumstances—those that Perhaps a richer approach is to classify house- decisionmakers typically confront—there are 8 Measuring the Impact of Energy Reform—Practical Options BOX 4 INEQUALITY ANALYSIS OF ELECTRICITY CONNECTIONS IN COLOMBIA A recent study applied inequality analysis to electricity con- cross-subsidies in electricity pricing, which are based on nections in Colombia, looking at the change in electricity the characteristics of each neighborhood. Analyzing the inci- connection rates by income quintile between 1974 and 1992 dence of these cross-subsidies across income quintiles, it (Vélez 1995). The concentration coefficients for these two found a slightly progressive pattern, indicated by a concen- years indicate that the distribution of electricity connections tration coefficient of –0.033. And distinguishing between went from regressive (0.157) to virtually egalitarian (0.034). legal subsidies (those accruing to legitimate paying cus- The reason is that the new connections during the interven- tomers through the official tariff structure) and illegal subsi- ing period were somewhat skewed toward lower-income dies (those accruing implicitly to households with nonpaying, households, as indicated by the slightly negative concentra- clandestine connections), the study found that illegal subsi- tion coefficient of –0.031. dies are much more progressive, with a concentration coeffi- The study also looked at Colombia’s complex system of cient of –0.301 compared with –0.016 for legal subsidies. BOX TABLE 4 INCREASE IN ELECTRICITY COVERAGE BY INCOME QUINTILE IN COLOMBIA, 1974–92 Electricity coverage rate Increase in coverage, 1974–92 Income (percent) New connections Share of new connections quintile 1974 1992 (thousands) (percent) 1 (richest) 91.3 98.0 750 17.4 2 73.5 96.0 849 19.7 3 61.7 93.4 897 20.8 4 49.1 90.4 943 21.9 5 (poorest) 41.4 81.3 869 20.2 Concentration coefficient 0.157 0.034 –0.031 Source: Vélez 1995. shortcuts that may permit some approximation each of the fuels the household uses, from which of the indicators. household fuel shares can be derived. This infor- mation, rarely available in direct form, can gen- Spanning the full range of data requirements erally be inferred from data on household expenditure on different fuels, by applying unit The data set should contain comprehensive prices and efficiency factors to derive implicit information about both the energy-related deci- levels of effective consumption. This approach sions of the household (required to calculate the does not capture consumption of traditional bio- welfare indicators) and the poverty indicators mass fuels that households gather at no mone- required to examine the welfare impact on the tary cost, however, which may be a particularly poor. Only ten basic pieces of information are important energy source for the poorest. This required to calculate all the indicators on access information can be obtained only through a spe- and affordability (table 2). (The health and edu- cial survey. cation indicators are omitted from table 2 because they are much more complex and case The most important source of information will specific.) Moreover, many are parametric (such be household surveys, such as the World as subsistence thresholds and unit costs) and can Bank–inspired Living Standards Measurement therefore be derived from external sources. Study surveys or the general income and expen- diture surveys. These combine information on Perhaps the most critical input for these indica- energy expenditure with information about tors is the effective household consumption for household income and expenditure, from which The World Bank Group 9 TABLE 2 DATA REQUIRED TO CALCULATE INDICATORS, BY POTENTIAL SOURCE Data sources Engineering Household Indicator estimates Price surveys surveys Electric utilities Special surveys Coverage index • Household access • Household access by fuel by fuel Reliability index • Reliability of • Reliability of access by fuel access by fuel Concentration • Efficiency • Unit cost by fuel • Household index factors by fuel expenditure by fuel Affordability • Per capita • Unit cost by fuel • Per capita • Per capita index subsistence subsistence subsistence threshold threshold threshold • Efficiency factors • Household by fuel expenditure by fuel • Household size Average fuel • Efficiency factors • Unit cost by fuel • Household cost per effective by fuel expenditure unit of energy by fuel Average subsidy • Efficiency factors • Unit subsidy by fuel • Household • Unit subsidy per effective by fuel • Unit cost by fuel expenditure by fuel by fuel unit of energy Average total • Capital cost • Unit cost by fuel • Capital cost • Capital cost cost per effective of household of household of household unit of energy energy use energy use energy use • Efficiency factors • Household by fuel expenditure by fuel Economic burden • Per capita • Unit cost by fuel • Per capita • Per capita subsistence subsistence subsistence threshold threshold threshold • Efficiency factors • Household by fuel expenditure by fuel • Household size Povertya • Household income or expenditure a. Required in all cases to calculate indicators by income group. 10 Measuring the Impact of Energy Reform—Practical Options absolute or relative indicators of poverty can be holds in the pre- and postintervention samples, derived. In many cases household surveys com- ranging from matched pairs to multiple regression plemented by external price and engineering models (see Baker 1999 for a detailed discussion). parameters will be adequate for the analysis of the economic indicators of welfare. Obtaining data on treatment and control groups For indicators of access special surveys may be required, since household surveys typically con- A data set containing information both on sider access only to electricity. In some cases it households affected by the intervention and on may be possible to “piggyback” on an existing a control set of similar households not affected household survey by incorporating additional makes it possible to be sure that the impacts questions on energy consumption. observed are not in fact attributable to differ- ences in the pre- and postintervention samples Although household surveys increasingly record or to extraneous influences on energy con- the detailed expenditure information needed for sumption behavior unrelated to the intervention this type of analysis, many countries still lack such (Baker 1999). information. In these countries information on energy expenditures would have to be obtained One possibility is to compare different regions of from a special sector survey. Some countries may a country, some affected by the intervention and even lack reliable information on economic mea- the others not. But where the intervention had a sures of poverty. An alternative that is sometimes national reach, as is often the case, this option is available is the poverty map, which classifies unavailable. Moreover, constructing such a con- areas as poor or not poor according to an index trol on the basis of international comparators is of economic or noneconomic poverty indicators. likely to raise as many problems as it resolves. Where poverty maps are available, impact indi- cators can be calculated for a sample of house- To alleviate the problem of devising an adequate holds in the areas classified as poor. control, the indicators presented here tend to focus on outcomes directly linked to energy sec- Obtaining data before and after the tor parameters (such as consumption decisions) intervention and to avoid links with general levels of poverty (which may be sensitive to a wide range of deci- One of the main limitations of relying on exist- sions). Nevertheless, this problem is almost ing household surveys is that their timing is impossible to resolve entirely. unlikely to coincide exactly with the timing of the intervention. In some cases it may be possi- Conclusion ble to use a past household survey as the base- line for measuring impact, and then to repeat This Note began by arguing the need for a set of only the relevant sections of the survey on a sub- quantitative indicators for measuring the effect set of the original sample at a suitable time after of interventions in the energy sector on the wel- the intervention. fare of the poor. It developed three sets of indi- cators, covering access to energy services, their Even where timing is fortuitous, longitudinal sur- affordability, and effects on health and educa- veys (those following the same households over tion outcomes. time) are still extremely rare in developing coun- tries, so it is seldom possible to observe the same This set of indicators produces a holistic view of household before and after an intervention. But energy consumption rather than focusing nar- there are many statistical techniques that can be rowly on the electricity sector, as has too often used to control for differences between house- been done in the past. This approach is sup- The World Bank Group 11 ported by empirical studies of energy and Eberhard, Anton A., and Clive van Horen. 1995. Poverty and Power: Energy and the South African State. East Haven, Conn.: Pluto poverty, which find that the poor make limited Press. use of electricity even after obtaining a house- ESMAP (Energy Sector Management Assistance Programme). 1994. hold connection. “Ecuador: Energy Pricing, Poverty and Social Mitigation.” Report 12831-EC. World Bank, Washington, D.C. Gomez-Lobo, Andres, Vivien Foster, and Jonathan Halpern. The major challenge in implementing this Forthcoming. “Information and Modeling Issues in Designing approach is the need for household-level infor- Water and Sanitation Subsidy Schemes.” Policy Research Working mation about both poverty and energy use. But Paper. World Bank, Finance, Infrastructure, and Private Sector Development Network, Washington, D.C. this Note suggests shortcuts for deriving the Hicks, Norman. 1998. “An Analysis of the Index of Unsatisfied Basic information at relatively low cost from existing Needs (NBI) of Argentina with Suggestions for Improvement.” data sources. World Bank, Latin America and the Caribbean Region, Poverty Sector Unit, Washington, D.C. Hosier, Richard H., and W. Kipondya. 1993. “Urban Household Notes Energy Use in Tanzania: Prices, Substitutes and Poverty.” Energy 1 Subsistence energy consumption can also be defined empirically Policy 21 (5): 454–73. rather than normatively. This can be done by looking at the actual Kakwani, Nanak. 1986. Analyzing Redistribution Policies: A Study energy consumption of a reference group believed to be living in Using Australian Data. Cambridge: Cambridge University Press. a subsistence situation, for example, those whose total income or Lanjouw, Jesko O. 1999. “Demystifying Poverty Lines.” United consumption lies close to the extreme poverty line. Nations Development Programme, New York. 2 Where there are significant proven scale economies in energy con- Leach, Gerald, and Marcia Gowen. 1987. Household Energy sumption at the household level, this could be reflected by reduc- Handbook: An Interim Guide and Reference Manual. World Bank ing the weight attached to each marginal individual as household Technical Paper 67. Washington, D.C. size increases. Lok-Dessallien, Renata. 1999. “Review of Poverty Concepts and 3 The concentration coefficient ranges from +1 to –1, with positive Indicators.” United Nations Development Programme, New York. Viewpoint is an open values indicating a regressive distribution, negative values a pro- Lovei, Laszlo, Eugene Gurenko, Michael Haney, Philip O’Keefe, and forum intended to gressive distribution, and a value of zero a perfectly equitable dis- Maria Shkaratan. 2000. “Maintaining Utility Services for the Poor: encourage tribution. The formula for calculating the concentration coefficient Policies and Practices in Central and Eastern Europe and the dissemination of and is Former Soviet Union.” World Bank, Europe and Central Asia debate on ideas, 2 n  1 Regional Office, Washington, D.C. Draft. ∑ ix i −  1 +  innovations, and best n i =1  n Ravallion, Martin. 1998. Poverty Lines in Theory and Practice. Living practices for expanding Standards Measurement Study Working Paper 133. Washington, the private sector. The where n is the total number of groupings of the income variable D.C.: World Bank. views published in this used (for example, ten deciles) and xi is the share of connections Vélez, Carlos E. 1995. “Gasto Social y Desigualdad: Logros y series are those of the going to group i (not to be confused with the connection rate for Extravíos.” Departamento Nacional de Planeación, Misión Social, authors and should not that group). Bogotá, Colombia. be attributed to the World Bank or any of its affiliated organizations. References Vivien Foster (vfoster@worldbank.org), World Nor do any of the con- Albouy, Yves, and Nadia Nadifi. 1999. “Impact of Power Sector Bank, Latin America and the Caribbean clusions represent Reform on the Poor: A Review of Issues and the Literature.” World Bank, Energy, Mining, and Telecommunications Department, Region, Poverty Sector Unit official policy of the World Bank or of its Washington, D.C. Executive Directors or Baker, Judy L. 1999. “Evaluating Project Impact for Poverty Reduction: This Note originally appeared as a chapter in the countries they A Handbook for Practitioners.” World Bank, Latin America and the Caribbean Region, Poverty Reduction and Economic Management Energy Sector Management Assistance represent. Sector Unit, and Poverty Reduction and Economic Management Programme (ESMAP), Energy and To order additional Network, Poverty Division, Washington, D.C. Development Report 2000: Energy Services for copies please call Barnes, Douglas F., and Liu Qian. 1992. “Urban Interfuel Substitution, Energy Use and Equity in Developing Countries: Some Preliminary the World’s Poor (Washington, D.C.: World 202 458 1111 or contact Suzanne Smith, editor, Results.” Industry and Energy Department Working Paper, Energy Bank, 2000). For more Room F11K-208, The Series, no. 53. World Bank, Washington, D.C. information on ESMAP go to World Bank, 1818 H Chong, Albert, and Jesko Hentschel. 1999. “Bundling of Basic Services, Street, NW, Washington, Welfare and Structural Reform in Peru.” World Bank, Development www.esmap.org. D.C. 20433, or Internet Research Group, Washington, D.C. address ssmith7@ Cowell, Frank A. 1995. Measuring Inequality. 2d ed. London: Prentice worldbank.org. The Hall/Harvester Wheatsheaf. series is also available Domdom, Aleta, Virginia Abiad, and Harry Pasimio. 1999. “Rural on-line (www.worldbank. Electrification Benefit Assessment Study: The Case of the org/html/fpd/notes/). Philippines.” World Bank, Energy Sector Management Assistance Printed on recycled Programme (ESMAP), Washington, D.C. Draft. paper.