37480 ENERGYWorkingNotes ENERGY AND MINING SECTOR BOARD NO.3 DECEMBER 2004 Willingness to Pay for Water and Energy: An Introductory Guide to Contingent Valuation and Coping Cost Techniques Francesco Devicienti, Poverty Reduction Group, PREM Irina Klytchnikova, The World Bank and Stefano Paternostro Washington, D. C. Comments and suggestions welcome ABSTRACT TABLE OF CONTENTS In order to determine the impact of many water and 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 energy reforms, analysts need to elicit the preferences 2 Overview of the paper . . . . . . . . . . . . . . . . . . . . . . . . . 3 of users and their demand for the goods in question. 3 Willingness to pay for (improved) water services. . . . . . . 6 3.1 Averting expenditure method . . . . . . . . . . . . . . . . . . . . 6 When these goods are not routinely bought and sold 3.2 Travel cost methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 in the market, the standard approach of demand 3.3 Hedonic pricing methods . . . . . . . . . . . . . . . . . . . . . . . 9 estimation based on observed prices and quantities is 3.4 Contingent valuation methods . . . . . . . . . . . . . . . . . . . 9 not viable. Instead, analysts must 3.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 KEYWORDS resort to the types of non-market 3.4.2 CV biases and the various elicitation formats for Willingness to pay, methods described in this paper. non-marketed goods . . . . . . . . . . . . . . . . . . . . . . . . . . 10 contingent valuation, The two methods most widely used 3.4.3 Validation of the CV results . . . . . . . . . . . . . . . . . . . . . 14 coping cost methods, in the context of water and energy 3.4.4 Other recommendations . . . . . . . . . . . . . . . . . . . . . . . 15 cost-benefit analysis, 3.4.5 Theoretical underpinnings of the WTP measurement are contingent valuation and water, energy. for improved services. . . . . . . . . . . . . . . . . . . . . . . . . . 15 coping cost. These methods are 3.4.6 Econometric modeling of the WTP using CV methods . . 16 described in the paper with 3.4.7 Example: Estimating an average willingness to pay particular emphasis on their application in developing for water service improvements in Famagusta (Cyprus) . . 17 countries. The paper suggests that both the contingent 3.5 Stated choice experiments . . . . . . . . . . . . . . . . . . . . . . 18 valuation and coping cost methods are useful tools 4 Willingness to pay for energy services . . . . . . . . . . . . . . 18 for the evaluation of water and energy projects, 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 4.2 Consumer Choices and the Demand for Energy . . . . . . 19 particularly if they are used together to validate 4.3 Willingness to pay for improved energy services. . . . . . . 21 results. Moreover, it is critically important that the 4.3.1 WTP of firms for electricity service improvement . . . . . . . 22 analyst carefully confront a number of technical and 4.3.2 Willingness of households to pay for energy practical issues before the results of these non-market service improvements. . . . . . . . . . . . . . . . . . . . . . . . . . 27 approaches to preference elicitation may be 5 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . 28 validated. These issues are also described in the paper. Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Energy and Mining Sector Board · Infrastructure Network · The World Bank Group www.worldbank.org/energy For invaluable help, we wish to thank Vijdan Korman, Julian Lampietti, Dale Whittington, Lucio Monari, Caroline Van Den Berg, and Kinnon Scott. This paper draws heavily on their work and suggestions and those of many others. Yet, the usual disclaimers apply. 2 Willingness to Pay for Water and Energy 1 INTRODUCTION provided directly by the state, or indirectly through regulation (for example, water supply, sanitation, Policymakers committed to economic development energy, and so on), is often made more difficult when and poverty reduction often have to prioritize the an explicit market is absent. types and number of public actions that should be undertaken in a specific area, context, and country. Economists have nonetheless developed a variety of As with most policy actions, a sound valuation is techniques to value non-market goods. This paper usually necessary to decide which among the set intends to convey a general introduction to these of feasible alternatives should be recommended techniques--both revealed and stated preference-- and adopted. In principle, this requires that all and practical guidance to analysts interested in the benefits for each proposed action be compared to applications. Reflecting its intent to offer practical help, all the costs of the action in order to determine the paper restricts its focus to the theoretical and the action with the highest net benefit score. empirical issues arising in a particular range of applications: reforms aimed at improving the supply of While the choice is generally based on a water and energy in developing countries. While the comparison between the average level of the costs broad methodological aspects discussed are likely to be and benefits in the population, the distribution of relevant in other contexts, the specifics of these types of the gains and losses among the said population is reforms and of the technical and practical issues arising also crucially important. Careful evaluation not in the relevant evaluation justify this focus. only allows the analyst to judge whether the project is worth implementing, but also assists in Although the paper concentrates on techniques for determining ways in which final users and valuing non-market goods, this is not to say that the beneficiaries should be required to contribute more traditional, relatively well-known techniques for (user fees, accession costs, and so on). It also estimating a demand curve for goods and services enables an assessment of the risks faced and the based on market price-quantity data (and sometimes potential social and political opposition on supply functions, too) are less useful. On the contrary, account of the perceived "unfairness."1 when the analyst is in a position to use these techniques, for example, when a rich, detailed dataset There are, however, certain types of public actions is available, traditional techniques are likely to offer a for which it is not simple to compute the more consistent and reliable and less controversial associated benefits and costs in monetary terms, framework of analysis. However, the requisite nor to provide an evaluation of the welfare impact conditions for the use of these methods (mainly, but on the targeted population. The welfare not only a required information base) are unlikely to assessment of changes that occur in the markets be met in practice in most developing countries. For for private goods and services (for example, due example, electricity demand models are seldom to the implementation of price-subsidy reforms) is employed for measuring willingness to pay in most easy to carry out in principle, as it can be based developing countries because of insufficient data, on the observed changes in quantities and prices. particularly data relating to the prices and quantities However, the estimation of the costs and the of all the energy sources relied on by households. benefits of environmental resources whether Energy Working Notes 3 The rest of the paper is organized as follows. Section 2 OVERVIEW OF THE PAPER 2 provides a brief classification of the main non- market methods that have been used to elicit the Public actions aimed at improving the supply of consumer preferences relating to water and energy water and energy are recognized as a growing provision, anticipating some of the main lessons priority in developing countries. Policy interventions learned in the paper. The distinction between market may take the form of infrastructure building (for and non-market goods and the related implications example, pipelines or electricity networks), the for the evaluation strategy are also discussed. privatization of loss-making public utilities, the modification of existing government regulations Table 1 contains a summary description of existing for private provision of the services (for example, demand assessment techniques. These include the changes in the tariff structure), or improvements in quantitative approaches discussed in the paper service quality (reliability, safety, and so on) and (stated preferences and revealed preferences), as access (for example, coverage of remote areas). well as some qualitative approaches. The latter The evaluation of private projects may also be have been discussed elsewhere (for example, undertaken through the techniques discussed in Bamberger 2000) and are beyond the aims of the following sections. this paper, though it should be noted that they complement the techniques detailed in this note In each of these cases, a sound cost and benefit and work as stand-alone tools. analysis can inform a desirable course of action and potential tradeoffs. In general, one may argue Section 3.1 reviews coping cost methods (a type that the main difficulty lies in estimating the of revealed preference approach also known as (monetary) value of the benefits accruing to averting expenditure) that have been used for consumers and beneficiaries or their willingness evaluating water sector improvements; other revealed to pay (WTP) for the good or service provided. preference approaches for water are contained in The assessment of the costs,2 on the other hand, Sections 3.2 and 3.3. Stated preferences methods, poses fewer conceptual and practical problems, or contingent valuation, for water are reviewed in at least as far as infrastructure and operation costs Section 3.4, which focuses on practical and theoretical are concerned.3 considerations. A prototypical questionnaire to determine household willingness to pay for improved If the analyst can estimate a demand curve for a water supply has been included in Box 1. particular service because accurate price-quantity data are available through specialized surveys or, to Section 4 deals with willingness to pay for improved a lesser extent, through a Living Standard energy services and briefly discusses the production Measurement Study (LSMS; see Whittington, loss method, the captive generation method, and O'Sullivan, and Barnes 2004), then this demand contingent valuation. Box 2 contains an example curve can be used not only to predict changes in of a prototypical questionnaire to examine the behavior in response to changes in pricing policy, willingness of firms to pay for improved energy supply. but also to calculate estimates of welfare change that rely on standard cost-benefit methods.4 If the Section 5 concludes. analyst is unable to estimate the demand from 4 Willingness to Pay for Water and Energy market data, there are two broad categories of service provided by, say, utilities that impose major approaches that can be used to compute the value coping costs on consumers. These usually take the to consumers (and other beneficiaries) of reforms in form of investments in alternative supplies (water the provision of water and energy services: storage tanks, water treatment equipment, electricity generators, candles, and batteries) to deal with supply a. The "revealed preference in surrogate market" interruptions and inadequacies. If consumers are not approach is based on what people "do" to able to mitigate the consequences of inadequate cope with the absence of a market for the supplies, they may suffer from lost production or good they need; hence, it is also named the reduced household welfare. In general, the coping coping cost method or the averting costs avoided will be a lower bound on the expenditure method, and magnitude of the economic benefits of service b. The "stated preference approach," which is improvements. If a lower bound estimate of the based on what people say when they are benefits is greater than the cost of the intervention, asked directly about the good in question; this this benefit estimate can prove particularly useful. is often referred to as contingent valuation (CV), especially when used in the context of The CV (stated preference) approach has recently environmental amenities. been employed to evaluate water and energy projects in developing countries. A CV survey Revealed preference approaches rely on the constructs scenarios of possible future government observed behavior toward some market good that action. Survey respondents are asked to state their possesses a connection to the non-marketed good preferences concerning these actions. The choices of interest, for example, when the demand for made by respondents are analyzed in a manner improved public water is inferred by the purchase of similar to the choices made by consumers in actual bottled water or the purchase of water from private markets. In both cases, the economic value is vendors. Stated preference approaches, meanwhile, derived from the choices observed either in an rely on answers to specifically designed surveys actual market, or in a hypothetical market created relating to the non-market good in order to in the survey (Carson 2000). For example, understand the basis of the demand. respondents may be asked if they would agree to pay a specified monthly increase in their electricity Examples of revealed preference techniques that bill in exchange for improved service reliability. have been used in the evaluation of water and energy projects in developing countries include The use of contingent valuation has engendered travel cost methods, hedonic pricing methods, much debate among proponents (for example, averting expenditure methods, production loss Hanemann 1994) and critics (for example, Diamond methods, and captive generation methods. The and Hausman 1994). Each of the revealed preference names of the various methods sometimes relate methods is also subject to support and criticism on more to the specific area in which the methods account of the respective strengths and weaknesses, have been applied rather than to fundamental that is, each method has its advantages and differences in approaches. They are all based on disadvantages that relate to the nature of the good the same broad principles: the deficient quality of involved, the type of project to be assessed,5 and the Energy Working Notes 5 type of available information.6 While the following 2000; Asian Development Bank 1999). The pages are not meant to provide a full account of WTP methods described in the water section the merits and difficulties of the various approaches, therefore refer mainly to the household use of they briefly describe these in the context of water water. The section on energy, on the other and energy supply improvements, and they reflect a hand, focuses mainly on the willingness of deliberate attention to the subset of approaches that firms to pay both in manufacturing and in have found more extended application in agriculture (where the main energy use is for developing countries. irrigation). The same underlying principles discussed for the willingness of households to Two qualifications apply to the remainder of the paper. pay for improved water services can, when suitably modified, be applied to the analysis a. A larger number of WTP applications in of the willingness of households to pay for developing countries refer to water rather improved energy services and, to a lesser than energy. To some extent, this relates to the extent, the willingness of firms to pay. specific characteristics of water and energy. Water is a household necessity that, for the Before the various approaches for measuring the most part, cannot be substituted. Energy, on WTP are described in detail, it is helpful to highlight the other hand, can be seen as a production important features of the goods in question, as input for firms and households, and one type these are central to the methods that analysts can be replaced by a number of alternative employ in their cost and benefit analyses. sources (electricity and various types of fuels). As a result, models explaining energy choices First, water and energy services are not pure public and use are likely to be more demanding for goods, and the basis of the cost-benefit analysis the analyst, not least in terms of data therefore relates to market transactions that are requirements. Despite differences, most of the observed in principle. The goods are provided either general methodological issues are shared by the state, which is motivated predominantly by between the two sectors. These general issues natural monopoly considerations, or by the private will be discussed first in relation to water given sector within a detailed regulatory framework the higher number of studies on this sector. established by the state. In each case, the final b. A distinction is made between the WTP of consumer is usually asked to pay a price for service households and the WTP of firms based on access and use. In some cases, the quantities the different ways in which these goods are consumed and the price paid by final users can be used by these two types of economic agents. retrieved. The welfare evaluation of price changes Households mainly use water for drinking, (a modification of the tariff structure) can therefore cleaning, cooking, and heating purposes, be undertaken as in the case of pure private goods, while farms (and small producing households) that is, estimating consumer demand and deriving employ water mainly for irrigation. The the welfare impact of the proposed changes.7 literature on the pricing of irrigation water has been surveyed elsewhere and will therefore Second, even when price and quantity information not be covered at length here (see Johansson may not be readily available (because existing utility 6 Willingness to Pay for Water and Energy data are unreliable or unavailable to researchers) or households for these goods. Since these goods are, when there is no market for the goods in question, effectively, substitutes for a better service, their it may still be possible to look at data relating to purchase reveals the willingness of buyers to pay for substituted goods (for example, bottled water when improved service, which may in turn be used as a public pipe drinking-water does not exist) for an measurement of the WTP. assessment of demand. This is a simple example of the techniques described in this paper to value non- The theoretical underpinnings of the AE method market goods, that is, coping cost approaches. derive from a sort of household production model (Courant and Porter 1981; Bartik 1988; Lee and Certain circumstances require different techniques Moffitt 1993). Households are assumed to have for valuing non-market goods, and this is likely to preferences for the quality Y of a given public good occur when (a) the service is absent altogether in a (for example, municipal water) and a composite particular area, and, consequently, no transactions good Z, including "all other goods." The are observed from which consumer preferences may household's budget constraints state that total be understood (for example, remote mountainous expenditure on the composite good, plus the areas); (b) there are no close substitutes that can be "minimum averting expenditure," should not exceed used to understand consumer preferences indirectly; available income. The minimum averting and (c) new dimensions are added to existing goods expenditure is the minimum amount of money spent (for example, fundamental quality changes). In the on averting inputs that allow the household to last case, the new good may be viewed as non- obtain any given quality level Y from the existing marketed to the extent that current consumers have quality level X and averting inputs A. The household not yet engaged in complete transactions for this optimal choice of the AE is shown to depend on the good (or the new variety) and can only be asked price of the composite good (pz), the existing quality their preferences for the new good in the context of X, household income, the price of the averting a mental experiment. These techniques will be inputs (pA), and a vector of household described at some length in this paper under the characteristics reflecting tastes and preferences. heading of the stated preference or contingent valuation approaches. A number of important assumptions have to be satisfied in order for the AE to produce good measures of the WTP for service improvements. 3 WILLINGNESS TO PAY FOR (IMPROVED) WATER SERVICES a. Averting expenditures must not generate direct benefits to the household (that is, they should 3.1 Averting expenditure method not enter the household's utility function); in other words, they are only incurred as a The averting expenditure method (AE) is based on means of improving the existing quality of the the proposition that, when other goods are used to public service. For example, time spent compensate for the adverse effects of deterioration collecting alternative sources of water may in water systems, the value of the improved water have a social dimension that is valued for supply can be measured by the total expenditure of itself by the household, rendering the AE that Energy Working Notes 7 Table 1. Demand Assessment Techniques for Water Supply: A Summary "REAL" DETAILED PARTICIPATORY PARTICIPATORY OPTIONS ELICIT RELATIVE RAPID APPRAISAL RAPID APPRAISAL CONSIDERED DEMAND OPTION SELEC- OPTION SELEC- CONTINGENT BY COMMUNITY AMONG DIFFER- TION: INTERNALLY TION: EXTERNALLY REVEALED PREF- VALUATION GROUPS OR ENT SERVICES FACILITATED FACILITATED ERENCE SURVEYS METHOD BALLOT DESCRIPTION Improvements to a wide Community volunteers are A variety of participatory rapid Revealed preference surveys A questionnaire survey to deter- Detailed options and their variety of different services, encouraged and trained to appraisal techniques are used by estimate the time and financial mine the maximum willingness implications (costs, operations such as water, drainage, roads, undertake a participators trained researchers or facilitators costs of current household of individuals to pay for various and maintenance, institution- and so on, are considered by survey in their own community. to triangulate and confirm the behavior (for example, pay- options for level of service al, and so on) are considered the communities, which express Preferences and commitments preferences of different commu- ments to water vendors and the (including improved reliability) by communities using partici- OF their relative demand for these are then agreed on in nity groups, which are also time saved in collecting water, and payment arrangements patory rapid appraisal or the TECHNIQUES services. The total funds meetings. involved in the analyses. expenditures on substitute within the context of the current ballot. available for each community energy sources, time saved in or specified institutional regime. area should be reasonably collecting fuelwood, and so on). fixed. · Simple and easily · Good community sense of · Can provide reasonably · Provides good data for · More precise cost estimates understood · Good community sense of ownership accurate estimates of project appraisal lead to less confusion · Expresses "real" demand if ownership · Extension staff can assess current time and cost · Good data on the WTP and · Institutional charging of PO only in relative terms · Enhances empowerment appropriate time to elicit expenditure and hence potential revenues for differ- operations and mainte- TENTIAL · Preferences can be refined · Useful if demand assess- demand possible willingness to pay ent service levels, assuming a nance implications can be during micro-planning ment involves ongoing · Can enhance for service improvements thorough survey is thoroughly assessed negotiation empowerment · Data and analysis undertaken · Can be used in a changing BENEFITS · Can be used in changing the requirements are · Can guide tariff subsidy and institutional environment institutional environment modest cost recovery policy · Good baseline data for · Similarity to public impact assessment opinion polls means results · Compatible with conceptually easy for participatory rapid appraisal non-specialists and the understanding of politicians PO · Possible group or strategic · Possible group bias · Possible group bias · Cannot estimate the house- · Risk inhibiting community · Risk that key decisions will TENTIAL bias · Liable to lack technical and · Process can be manipulated hold response to price decisionmaking and be based on misleading · The WTP for different serv- financial rigor by extension workers who do increases (including for new ownership by, for instance, results from an unrepre- ice levels not readily known · Reliant on skills being not use sufficient technical or levels of service options) raising expectations about sentative group unless care RISKS · Process can be manipulated present in the community financial rigor if they are not · Poverty may constrain the particular options is taken to avoid group by extension workers who · Requires substantial flexibil- adequately supervised ability of poor people to · Relatively high cost and bias AND do not use sufficient ity by external funding · Extension workers with good convert the time savings requires specialized consulta- · Requires detailed cost technical or financial rigor agencies and local support facilitation skills are required resulting from service tion for reliable results information, so the earlier CONSTRAINTS institutions improvements into cash · Inaccuracies may occur in demand assessment may payments for services a changing institutional need to use other methods environment · Detailed work on some options can be redundant · Requires flexibility by funding agency Suitable for village or slum More suitable where low-tech- Suitable in most situations, Suitable where substantial Suitable for informing strategic Suitable where difficult choices general improvement projects. nology, low-costs solutions are possibly complemented by other water supply problems exist; decisions on levels of service, are to be made between Nongovernmental organiza- definitely viable, for example, methods to be used in conjunction with cost recovery policy, and so on different options TYPICAL tions often use these hand pumps participatory rapid appraisal in large investment programs, techniques methods for example, urban systems or policy frameworks for small USA rural supply schemes GE Source: DfID 1998. See also Klugman 2002, "Water and Sanitation," Chapter 23. 8 Willingness to Pay for Water and Energy includes such time costs a potentially biased household labor and monetary resources to avoid estimate of household WTP. exposure to the bad quality of municipal water (for b. Averting expenditures must be a perfect example, time for boiling water), the cost of filters, substitute for a change in the quality of the the annualized costs of required equipment, public good. Some averting actions do not purchases of bottled water, and the like. The meet this condition; for example, a point-of- opportunity cost of time is often computed on the use water treatment system installed in the basis of a minimum wage or an hourly wage kitchen sink will not reduce adverse health estimated from family income. effects resulting from inhaling chemicals evaporating from water in the shower Since AEi is a continuous variable, ordinary least (Abrahams, Hubbell, and Jordan 2000). square regressions are used (often supplemented c. If large fixed costs have already been by median regressions to check for the robustness incurred, the AE should include an of the results in the presence of outlier annualized stream of such costs. Averting observations). Once estimates of have been behaviors such as installing home water obtained, the model can be used to predict the AE treatment systems have significant fixed or (which, in this approach, is a measure of the WTP) sunk costs. In such cases, the sunk cost may for any profile of households and local or service- alter household behavior regarding the related characteristics. The project is said to pass provision of improved service. If households the cost-benefit test if the estimated total WTP solve their water problem through these exceeds the project's cost. Note that, in principle, investments, they may not need the distribution of the benefits and, therefore, the improvements. Therefore, using these distributional and poverty impact of the project averting costs may be misleading. itself could be provided by estimating the WTP for different profiles of consumers, for example, the The theoretical models outlined above also provide poor versus the non-poor (as defined in terms of the basis for the econometric approaches that have total household expenditure or any other chosen been used to model the AE and obtain WTP indicator). Gender and social dimensions could estimates. A linear regression model is generally also be explored by comparing the WTP for men assumed to be of the following form: and women, young and old, and so on. AEi = Wi'+i 3.2 Travel cost methods where AEi denotes the total averting expenditure of Analysts rely extensively on the travel cost method household i, Wi includes the determinant of the AE to place a value on outdoor recreational locations as suggested by the theoretical model (for example, such as national parks and fishing and hunting pz, pA, X, and so on), and i is the usual error term. sites. This method relies on observations of people's behavior, particularly expenditure for In the studies that have used averting expenditure transportation and other trip-related expenses methods to estimate the WTP for water incurred when traveling to and from the site of improvements, AEi has included measures of interest. Moreover, it calculates the benefits derived Energy Working Notes 9 from the specific site and uses this information to neighborhood characteristics, quality of neighborhood find optimal levels of service provision (Bishop and schools, accessibility to work, crime rates, and perhaps Heberlein 1990). some measure of the quality of local drinking water, air quality, noise pollution, and aesthetic views. The method is, however, more appropriate for valuing Assuming each of these attributes has an associated recreational sites than for estimating the WTP for implicit price, the market price of the property is improved water services. In fact, households often equal to the sum of the implicit prices, multiplied by use various alternative sources to maintain a certain measures of the attribute. The application of the level of water quality, for example, municipal and hedonic pricing method to capture the effect on prices private tankers for non-drinking water and private of improvements in environmental quality requires vendors and bottled water for drinking purposes. observations of sufficiently varying quality levels Measuring the value of the time spent carrying within the confines of a single housing market water from the specific site may not provide a (Leggett and Bockstael 2000). complete picture, and other non-transportation costs (for example, investments in storage and pumps to Korman (2002) notes the limitations of the hedonic cope with intermittent and poor quality public water) pricing method for valuing water quality improvements should be considered. In this respect, the travel cost in Famagusta, Cyprus, mainly on account of the fact method can be seen as a special case of averting that water quality did not vary sufficiently across the expenditure in that it focuses on a particular type of single markets or locations. The method could still be expense only, namely, travel. Though likely to provide applied if the geographical domain of the analysis were a lower bound (or conservative) estimate of the extended to capture more variation in water quality. WTP only, the method may still be a practical and Water is, however, not location specific (Whitehead easier alternative to the more demanding methods and Van Houtven 1997), and everyone generally has discussed in the following sections (see Whittington, access to the same level of water quality at the village Mu, and Roche 1990 for an application to water or even regional level. provision in Kenya). 3.4 Contingent valuation methods 3.3 Hedonic pricing methods 3.4.1 Introduction The hedonic pricing approach has been used extensively to estimate the value of property and While previous methods for estimating the WTP are housing prices (Blomquist and Worley 1981; Rosen based on the actual behavior of the subject involved, 1974). It assumes that a good (or service) can be the CV method relies on the subject's responses to a fully characterized by its attributes, that consumers survey questionnaire (O'Doherty 1998). Moreover, have very good information on the attributes of unlike travel cost, hedonic pricing, and averting goods, and that price differentials reflect the values expenditures, the CV method does not require public of different attributes of goods. For example, in goods or services to be linked to actual market housing markets, the price of a house can be transactions. If a market for the public good in question broken down into the value of its main attributes, does not exist, a hypothetical market can be created that is, physical structure, age, number of rooms, by a mental experiment using a properly devised 10 Willingness to Pay for Water and Energy questionnaire. It is well reported that, with stated than overestimated; (b) the use of the WTP preference techniques, researchers can design surveys questions rather than the willingness to accept to elicit references for goods with attributes that are questions; (c) the use of the referendum form rather not currently available in the market (Rubey and than open-ended questions; (d) an accurate Lupi 1997). description of programs and policies; (e) a reminder of substitute commodities; (f) the use of yes-or-no However, the hypothetical nature of the market, the follow-up questions; and (g) checks on the description of the public good in question, the respondent's understanding of the scenario (Haab elicitation format, and the payment routines of the and McConnell 2002). CV method may cause biases in the estimation of the WTP. Diamond and Hausman (1994) state that The CV method has been employed by several authors lack of experience in the markets for environmental to measure the value of water quality improvements commodities and in the consequences of such (Carson, Flores, and Meade 2001; Alberini and decisions can make CV questions difficult to answer Cooper 2000; Ready, Malzubris, and Senkane and raise uncertainties about the responses to the 1997; Hanemann 1994; Carson and Mitchell questions. Research in this field has focused on the 1993; Green and Tunstall 1991; Whittington and hypothetical nature of the CV method and calls into others 1990; Schultz and Lindsay 1990; Mitchell doubt the accuracy and reliability of the method's and Carson 1989; Edwards 1988; Korman 2002). value measure. Decades of experimentation and research have, however, gone a long way toward In these studies, the CV method has been helpful in mitigating or eliminating weaknesses and enhancing predicting household behavior as hypothesized by the strengths of the CV method. Researchers in this economic theory. For example, CV analysis has field (for example, Smith 2000; Hanemann 1994; enabled researchers to predict the number of Arrow and others 1993; Freeman 1993; Portney connections to water supply systems after conditions 1993; Mitchell and Carson 1989; Cummings, have improved, as well as the amount of additional Brookshire, and Schulze 1986) suggest various revenues for water authorities (Alberini and Cooper measures for improving the reliability of the 2000). Similarly, Griffin and others (1995) report estimates and minimizing the potential biases that, based on results in Kerala State in India, the arising from the implementation of a CV study. actual behavior of households can be predicted with remarkable precision using the CV method. They In response to the debate about the shortcomings conclude that a carefully designed and conducted of the CV method due to its hypothetical nature, the CV study can supply reliable information on how U. S. National Oceanic and Atmospheric people value goods and services. Administration assessment panel issued a set of guidelines for conducting CV surveys to assure the 3.4.2 CV biases and the various elicitation reliability and usefulness of the information that is formats for non-marketed goods obtained using the method (see Arrow and others 1993). Among others, the panel recommended the Three potential types of bias may arise during the following approaches: (a) a conservative survey implementation of a CV study (Mitchell and Carson design so that the WTP is underestimated rather 1989; Cummings, Brookshire, and Schulze 1986). Energy Working Notes 11 a. Strategic bias arises when a respondent thinks of the improvements (Griffin and others that s/he may be able to influence an investment 1995; Whittington and others 1990). or policy decision by not answering the interviewer's questions truthfully (Whittington The CV methods elicit consumer preferences and, and others 1990; Whittington, Mu, and Roche therefore, the WTP. The elicitation questions take 1990). Such strategic behavior may influence numerous formats: payment card, dichotomous results in one of two ways. Suppose an choice, open-ended format, and stated-choice individual is asked how much she would be experiments. The different formats, each with its willing to pay for potable water from a public own advantages and drawbacks, have been shown water supply. If she thinks that the water utility to yield statistically different responses (Welsh and will provide the service only if responses are Poe 1998; Brown and others 1996). Bohara and positive, but that someone else will ultimately others (1998) and Carson (2000) suggest that it is pay for the service, then she will have an unreasonable to expect different elicitation formats incentive to overstate her willingness to pay. to yield the same answers since they are not On the other hand, if she believes that the strategically and informationally equivalent. water utility has already made the decision to provide the public with water and the purpose of Payment card method: The interviewer presents the survey is to determine the amount people respondents with a range of WTP amounts ranging will have to pay for the improved services, she from $0 to a large sum and asks which amount will have an incentive to understate the extent presented on the card is the most they would be of her true willingness to pay. willing to pay for the level of the proposed change b. Starting-point bias may result from the bidding (see the example in Box 1). Although the payment game or dichotomous choice elicitation card method reduces a starting point bias, it may formats. A specific price is asked, and, when have the following drawbacks. (a) Positioning a respondent is unsure of an appropriate amounts at the top or bottom of columns tends to answer and wants to please the interviewer, draw "benchmark biases," and (b) the range of s/he may interpret this initial price as a clue amounts offered may compel respondents to stay to the "correct" bid. Starting-point bias exists within certain bounds. if the initial price affects an individual's reported final willingness to pay (Boyle, Dichotomous choice method: respondents face a Bishop, and Welsh 1985). single question of the form: "Are you willing to pay c. Hypothetical bias may result from a lack of $x?" The bid level may be constant for the whole clear understanding or correct perception of sample or may be varied across groups within a the characteristics of the proposed changes sample. Respondents accept or reject the WTP described by the interviewer. This source of amount. This elicitation method has the advantage bias is not likely to be significant for most of being simple and familiar because respondents public services, such as water, based on the make a yes or no decision based on a given price strong general familiarity with such services. in a manner that is similar to everyday market Respondents usually know the nature of the decisions (Freeman 1993). Schultz and Lindsay problem and comprehend the characteristics (1990) state that the dichotomous format minimizes 12 Willingness to Pay for Water and Energy both hypothetical and strategic biases, which are responses (Carson and others 1998). Those who greater in an open-ended format. This method is also have agreed to pay the amount used in the first considered incentive compatible, because it is in a discrete WTP question are asked if they are willing to respondent's strategic interest to say "yes" if his/her pay a higher amount, 1.5 times the initial bid, and WTP is greater than or equal to the price asked those who refuse to pay the amount in the first bid (Mitchell and Carson 1989). Diamond and Hausman are asked if they would agree to a lower amount in (1994), however, disagree with this view, arguing the second discrete WTP question. Hanemann (1984) that, since the results of the CV method research indicates that using the double bounded format are not a real obligation, strategic behavior may could improve the statistical efficiency of the occur with the use of a referendum approach as well. estimated mean and median WTP. Hanemann, Loomis, and Kanninen (1991) have found that the The dichotomous choice approach faces a "starting double bounded format provides a substantial gain point bias." This bias refers to the clustering effect of in the variance-covariance matrix of the coefficient WTP bids around an opening bid by an interviewer. estimates. This gain leads to better confidence To minimize this starting point bias, the CV literature intervals for the estimated mean and median WTP. recommends a procedure that involves randomly Calia and Strazzera (2000), however, show that the assigning survey respondents to groups or subsamples relative differences in estimates between the simpler and then asking each subsample to respond to dichotomous choice format and the double different starting bid amounts. Alternatively, one bounded-dichotomous choice format tend to should include follow-up questions to improve the decrease as the sample size increases. efficiency of the dichotomous choice format questions. However, studies that have specifically tested for the Open-ended questions: This format requires starting point bias have found no evidence that a respondents to state a WTP amount without the aid clearly defined referendum question is subject to of an explicit context or process. The open-ended such a bias (Herriges and Shogren 1996; McPhail question alleviates the starting-point bias, and the 1993; Whittington and others 1990). This format WTP can easily be estimated. Respondents may, also requires many more responses in order to however, find it difficult to set a value for the estimate the maximum WTP and involves relatively proposed change and therefore register a "protest more complicated econometric analysis (Bishop and zero" offering a $0 WTP amount, despite the good Heberlein 1990). An additional drawback is the having some value to them, as a means of refusing to possibility of a relatively higher percentage of yea- answer (Mitchell and Carson 1989). The unacceptability saying responses to please or agree with the of a large number of protest zeros that are interviewer, regardless of the true valuation (Bishop indistinguishable from true zeros (that is, zero bids that and Heberlein 1990; Mitchell and Carson 1989). are honest responses of low-income respondents or of people who do not positively value the proposed Double bounded-dichotomous choice format. To change) has led many CV researchers away from minimize the disadvantages of the dichotomous open-ended questions. choice method, respondents are asked, after the first question, to answer a follow-up supplementary In summary, the open-ended format theoretically dichotomous question based on their prior provides relatively unbiased responses as there is Energy Working Notes 13 Box 1. A Prototypical Questionnaire for Household WTP for Improved Water Supply Below is an example of a CV scenario devised by Whittington (2000) for the case of tap water in urban areas. The questionnaire is usually inserted within the larger set of survey questions. Interviewers first describe the set-up, ensuring that the interviewee fully understands the scenario, with which s/he might be unfamiliar. Interviewers might start by describing the characteristics of water supply systems in many developed countries, that is, that households have metered private connections and only pay for the water they actually use, that the water is safe to drink from the tap without further treatment, that the service is continuously available, and so on. The interviewee is then briefly reminded of the local situation. Such a system does not exist in [NAME OF TOWN]. Many households only have water for a few hours each day. Sometimes the water supply breaks down, and some people do not have service for several days. Many people in [NAME OF TOWN] treat their water before they drink it. Today, the typical household in [NAME OF TOWN] has to pay between $x and $y per cubic meter for the water they use if they have a metered connection. The average household water bill is about $z per month. Interviewers then begin to describe the alternative scenario in detail. Water would be available 24 hours a day, 365 days a year, with sufficient pressure, safe to drink from the tap, and so on. Suppose that the price of water for this improved system were [p1, p2, p3, p4, p5] per cubic meter. Because you would pay for the water you actually used, your water bill could vary from month to month. But suppose that, in a typical month, an average household would use about 15 cubic meters of water and receive a bill for [15p1, 15p2, 15p3, 15p4, 15p5].1,2 The price of a water connection would remain the same as it is today, or about $x. Suppose that a vote were held in [NAME OF TOWN] regarding a project to build a modern water system here. If the price of the new system were [p1, p2, p3, p4, p5]3 per cubic meter, would you vote for the new water supply project or against it? So, the first question will require a simple "yes/no/do not know" answer. Now suppose that the improved water supply system were installed in [NAME OF TOWN]. Would your household want to be connected to this system if the price of water from the new system were one of [p1, p2, p3, p4, p5] per cubic meter? Or would you make other arrangements for obtaining your water supply? This requires another "yes/no/do not know" answer. "Yes" means connect or stay connected to the new system, and "no" means do not want to be connected, will make other arrangements. 1The household would only be asked about one price in this question, the price determined by the part of the split sample to which the household was assigned. The assignments are random, so that each split of the sample is a random subset of the whole. 2These questions depend on the head of household or spouse knowing the number of cubic meters the household uses. In fact, in many developing countries, households with private metered connections know more about their water usage than do households in industrialized countries. Nonetheless, the survey designer must ensure that this assumption is generally valid in the country of the survey. 3A respondent would receive one of these prices, which would be randomly assigned. 14 Willingness to Pay for Water and Energy nothing inherent in the question that may influence predictions should be confirmed by the responses (and it has no starting point bias). estimated coefficients. For example, economic However, respondents may find it difficult to set a theory suggests that the percentage of the value, and they therefore report a zero amount. This respondents willing to pay a particular price difficulty could result in a high no-response rate, as should decrease as the price they are asked well as protest bids (Mitchell and Carson 1989). to pay increases. This is one of the Moreover, the open-ended format provides greater fundamental postulates of economic theory, opportunity for respondents to act strategically. The that is, the law of demand. literature on the CV method strongly recommends b. Correlation test (convergent validity). using a dichotomous format in order to minimize, This refers to the correlation between the WTP even if not completely eliminate, the biases (Arrow measures resulting from the CV method and and others 1993). The use of person-to-person the AE method. Using the revealed preference interviews to decrease sample selection bias and method to verify the results of the stated minimize the non-response rate is also generally preference method and compare the recommended (Arrow and others 1993). As a magnitude and correlation between the two result, the dichotomous format has been the methods is highly recommended and is preferred method in empirical work. employed in empirical studies. Carson and others (1996) review 83 studies and conclude There is also evidence that estimates of the WTP based that CV estimates are, on average, smaller on open-ended questions tend to be systematically than those based on revealed preference lower than those based on dichotomous choice techniques. They find that correlations range formats. This conclusion, however, mainly holds true from 0.78 to 0.92. for environmental and public goods rather than for private goods. When people are familiar with a This correlation test was used by Korman (2002) in particular service, as in the case of water supply, a study of the WTP for water improvements. The Korman (2002) suggests that the two methods are households analyzed in Famagusta (Cyprus) were likely to produce similar WTP estimates. adversely affected by municipal services, and, because alternative (private) water markets exist, 3.4.3 Validation of the CV results averting expenditures were used to meet the water requirement. Korman could then compare the The CV method remains subject to criticism, and results of the averting expenditure method to those good CV practice therefore includes a number of of the contingent valuation method and check the tests to validate results. validity and feasibility of using contingent valuation for estimating the benefits of improved potable a. Tests of the theoretical validity of the model. water services in Famagusta (see section 0). "Theoretical validity" is measured by examining the relationship between the WTP Griffin and others (1995) conducted an unusual and the explanatory variables that are experiment in the Indian State of Kerala in which theoretically the determinants of the WTP for they compared the stated responses to a CV the improved public services. Theoretical questionnaire with the actual connections to a piped Energy Working Notes 15 water supply by the same households a few years later Another important issue regarding the elicitation of (also a test of convergent validity). They found that information is the choice of the initial bid format. The stated responses are accurate in predicting the actual most commonly used format is a variable initial bid connections both on average and for particular whereby respondents are divided into different households in the survey. However, stated responses subsamples and given different starting bid amounts. provide a much poorer prediction of connections by Variable initial bids are useful in testing for starting similar households in a similar location. They point bias, and they make the calculation easier of concluded that the benefit transfer approach, that the mean and median WTP from binary choice data. is, when estimates from one location are applied to Whittington and others (1990) point out that variable a different location, may be more successful when initial bids may cause problems in eliciting information omitted factors that are different across the from respondents who may readily notice that different locations are taken into account more accurately. households have been told different costs for the This calls for rigorous qualitative work during the proposed plan. This may create uncertainty about the stages of questionnaire design and follow-up. cost of the program and result in bias in the reported WTP. While studying the WTP for water services in 3.4.4 Other recommendations southern Haiti, Whittington and others (1990) conducted tests to determine whether the starting Whittington (1998) reviews the issues that relate to point bias would be a major problem. Three versions the implementation of the CV method in developing of their questionnaire were randomly distributed, each countries. He reports that CV surveys are now starting with a different initial bid. They found that the becoming relatively easy and more straightforward. mean WTP did not vary much with a different starting The survey response rates are typically high in point bid. developing countries, and respondents are receptive and give consideration to the questions asked. 3.4.5 Theoretical underpinnings of the WTP Moreover, he indicates that it is much cheaper to carry measurement for improved services out in-person surveys in developing countries and therefore to have larger sample sizes than are usual What is the welfare impact on a household if the quality in industrial countries. He concludes that it seems of a public service increases? If the existing quality level feasible and desirable to use CV studies in developing of the service (say, municipal water) is given by Q0, its countries to help evaluate a wide range of projects. price for the household is given by pW0, and the price of Whittington also suggests that, with regard to cultural the other "composite" good is pZ0, then the improvement and ethical settings, CV researchers should be in water quality from Q0 to Q1, with no concomitant cautious in the design and implementation of the price changes, produces a welfare gain of: survey. This includes treating respondents with respect as citizens rather than as experimental subjects, CV = E(pW0, pZ0, Q0, U0) ­ E(pW0, pZ0, Q1, U0) > 0 giving them time to think about their responses, and making sure that referendum prices do not spread where CV is the compensating variation that provides confusion and misinformation about the cost of the true measure of the welfare effect of the water addressing possible problems of public concern. quality improvement that is being evaluated, E(.) represents the consumer's expenditure function, and 16 Willingness to Pay for Water and Energy U0 is the household's initial level of welfare. Note and standard ordinary least square methods that E(pW0, pZ0, Q0, U0) is simply the household's (and quintile regressions) can be employed. initial income, I0. The CV is the amount of money However, in CV surveys with an open-ended that the consumer is willing to pay to see the service's elicitation question, researchers often face a quality improve from Q0 to Q1 and is exactly the large proportion of zero WTP responses, some measure that the contingent valuation method aims of which reflect true negative WTP, which is to elicit from respondents to the household survey. then censored at 0; others are protest zeros, and still others are true, valid zeros. In such a In equation (2), U0 is the maximum utility (indirect situation, censored regression models (Tobit, utility) achieved by the consumer when facing prices for example) should be used. pW0 and pZ0, quality Q0, and income I0, that is, b. In the case of a simple binary (dichotomous U0 = V(pW0, pZ0, Q0, I0). After substituting, (3) can choice) elicitation question to a given cost of therefore be rewritten as a (generic) function (.) of quality improvement, a random utility-based the prices and qualities in the status-quo and in the method provides the underpinnings of the hypothesized improved scenario: estimation approach. The probability that the respondent agrees to pay $R for the quality CV = WTP = (pW0, pZ0, Q0, Q1, I0). improvement is written in terms of a utility comparison: 3.4.6 Econometric modeling of the WTP using CV methods Pr(yes to $R) = Pr(V(pW0, pZ0, Q1, I0 ­ R; Hi) + i > V(pW0, pZ0, Q0, I0 ; Hi) + i) Once an error term is added to (3), one obtains the empirical counterpart, the foundation for the WTP Assuming that the error term ei has a distribution estimation based on contingent valuation methods: function F(.), one can easily derive the familiar expression for dichotomous choice models: CVi = WTPi = (pW0, pZ0, Q0, Q1, I0 ; Hi) + i Pr(household i says "yes" to $R) = F(Xi'). Logistic or normality assumptions for F(.) give rise, respectively, where a vector Hi of household characteristics to logit and probit models. Today, these are easily (household size, the education and occupation of implemented in most statistical software. the household head and the spouse, and so on), as well as other variables reflecting preferences, tastes, c. If double bounded-dichotomous choice formats and local circumstances, has also been included. are used to elicit household WTP, the econometric models become slightly more complicated, but The econometric methods to be used depend in part are still easy to implement. Suppose a two-bid upon the elicitation format adopted by the survey format is used, where the first dichotomous investigator. A number of cases can be distinguished. choice question is an accept/not-accept $R offer, with a further offer of $(R+d) following a. If open-ended questions were used when acceptance, or a counteroffer of $(R­) establishing the WTP for the proposed quality following initial decline. The four possible change, WTPi, in (4) is a continuous variable, patterns of answers are summarized below. Energy Working Notes 17 Initial question ($R) lives (Arrow and others 1993). Furthermore, the Yes No method minimizes strategic bias and non-response Follow up yes $(R+d) $(R­) rates. Since respondents are familiar with the nature (+$d or -$) no $R 0 of the water supply service, it was assumed that hypothetical bias would be minimized. The four possible choices can be analyzed through an ordered probit (or logit) model, where the true In particular, Korman used a double bounded- latent (unobserved) WTP is WTP*, and the observed dichotomous choice format to increase the choices are seen as the outcomes of a discrete statistical efficiency of her WTP results. The open- random variable: ended question followed by the double bounded- dichotomous choice format was used to check for Pr(no, no) = Pr(R­ > WTP*), "yea-saying behavior" among respondents to the Pr(no, yes) = Pr(R > WTP* > R­), dichotomous choice format and to observe whether Pr(yes, no) = Pr(R < WTP* < R+d), and households were consistent in their answers to the Pr(yes, yes) = Pr( R+d < WTP*). WTP questions. By doing this, Korman expected to minimize the starting point bias. Furthermore, The coefficient estimates from the double bounded person-to-person interviews were employed to model are assumed to be more efficient (smaller decrease sample selection bias and minimize the variance of the estimated coefficients) than those non-response rate (Arrow and others 1993). from the single bounded estimation model. Because the CV method is still open to criticism, 3.4.7 Example: Estimating an average various tests were employed to confirm the validity willingness to pay for water service of the method (Carson, Flores, and Meade 2001; improvements in Famagusta (Cyprus) Carson and others 1996; Mitchell and Carson 1989). Theoretical validity is measured by examining the Korman (2002) used the AE and the CV methods to relationship between the WTP and the explanatory estimate the willingness of consumers to pay for variables, which are theoretically the determinants of improved water supply in Famagusta. She stresses the WTP for the improved water services. Econometric that, in the CV survey, consumer WTP for improved analysis is used to test the theoretical validity of the municipal water depends on the information widely CV methodology and to estimate the responses of available about the improvements consumers should households about any water improvements that expect. Although, in the CV method, estimates of the might be undertaken by the Cyprus government. WTP may be subject to potential biases (hypothetical, Furthermore, a convergent validity test was applied strategic behavior, and starting point bias), the to compare the CV results and the AE method. This guideline recommendations for minimizing these test compares the estimated mean and the correlation biases are incorporated in the design of her survey. of the WTP in two methodologies. The data As recommended by the National Oceanic and collected during the field survey permitted this Atmospheric Administration, Korman implemented comparative analysis. the dichotomous choice format because people are familiar with this type of arrangement in their daily 18 Willingness to Pay for Water and Energy 3.5 Stated choice experiments preceding methods, although the model specification is different. In each choice situation, respondents Over the past decade, many economists have face several options that differ in terms of their turned to the methods that are known as stated service or program attributes, such as the frequency preference, conjoint analysis, attribute-based and timing of outages, the level of voltage fluctuations, methods, or stated choice experiments. These the existence or non-existence of prior notification evaluation approaches have their origin in of outages, and the price of services. In each choice marketing and transportation research studies. situation, respondents choose the option that provides Stated preference methods are useful when the the highest level of utility. This choice can be objective is to value different attributes of a formulated as a standard or a mixed logit model. program or service (Louviere, Hensher, and Swait The latter model incorporates the variation in 2000). In a stated choice experiment, a survey preferences for different aspects of service quality respondent is asked to choose from several options, among customers (Revelt and Train 1998). Hensher, each of which is associated with a number of Shore, and Train (2003) offer an accessible characteristics and a price. A series of experiments discussion of the model's specification and an is presented to each respondent, varying attributes application to the valuation of drinking water and across respondents to provide the necessary the service quality of water disposal. variation so as to enable econometric estimation (Hensher, Shore, and Train 2003). 4 WILLINGNESS TO PAY FOR The stated choice provides the basis for estimating ENERGY SERVICES the marginal WTP of respondents for each attribute and each service level of the proposed program or 4.1 Introduction service. This approach is useful when the program in question is characterized by a multitude of traits Projects and policies that improve the access to and the policy objective is to choose the optimal reliable, modern energy services can make an combination of traits. In addition, choice important difference in the welfare of the poor. experiments can be helpful in predicting behavior. What is the starting point for improving access? An and others (2002) applied this framework to What kinds of improvements will poor households estimate not only the WTP of respondents for an and communities value? Answering these questions electricity supply improvement, but also the impact requires an understanding of how these households that fuel switching would have on the use of obtain and use energy services today both for fuelwood and the preservation of wildlife habitat. consumption and productive activities. It is also Carefully designed, the choice experiments can essential to understand the demand of poor also help alleviate the hypothetical bias inherent households for better energy services, as well in the CV approach. as the willingness of these households to pay for the services. The framework for analyzing responses to stated choice experiment questions follows the same logic Traditionally, data collection on these issues was as the analysis of the WTP relying on the three weak because state monopoly providers had limited Energy Working Notes 19 incentives and capacity to learn about current and information available is sufficiently rich, models of potential customers. More recently, policy advisers household demand for specific fuel types can be and donor agencies have attempted to understand developed. Households consume energy for lighting, the demand of the poor for services and to tailor heating, water heating, cooking, and so on, often projects to the preferences of the poor. Nonetheless, employing different types of fuel (electricity, wood, there is still a wide gap in the available data. central gas, liquid propane gas, kerosene, dung, and so forth) for each of these activities. Household It is commonly estimated that two billion people demand models can potentially explain household lack access to electricity. These people rely on varied fuel choices, as well as the quantities of each fuel type energy sources and often incur real costs far higher applied, as a function of fuel prices, household than those for equivalent energy supplied by electricity income, and other socioeconomic factors. Panel data networks. Improving the energy sector is, however, from repeated surveys are particularly helpful for not simply a matter of reaching 100 percent evaluating the effects of policy interventions over time. electrification. It means providing better options so that the change can involve cleaner, safer, cheaper The econometric work associated with using surveys on energy sources and energy markets that are more prices and quantities, as well as socioeconomic responsive to the needs and demands of households information, to estimate household demand functions and communities. Evidence suggests that the poor can be challenging and will vary from country to are, indeed, often willing to pay for better services. country. Infrastructure policy analysts need to ensure The ESMAP Energy and Development Report that there is variation in the prices of fuels (electricity) (ESMAP 2000) thus remarks that a major challenge that results from changing cost conditions in order to is to open markets so that this demand can be identify the demand function. Nationally uniform tariffs identified and met. may limit (spatial) price variation. Also, expanding the block tariff structure poses difficulties because the 4.2 Consumer choices and the demand marginal and average price a household pays for energy depends on the household's own decision about how much to use. At a minimum, the estimation of demand Sound policy analysis entails an understanding of functions will necessitate the gathering of data on the the ways policy interventions (for example, projects, tariff structures applied to different households in the investments, regulations) will affect development survey sample. The complete modeling of household outcomes. Causal models are needed to link the fuel choices and quantities often requires information improvements in energy infrastructure services with that is unavailable, such as information on the prices desired outcomes, such as improved household of all types of fuels (including those not currently income, health, or well-being. Analysts need to adopted by the household), the quantity of the chosen know how policy interventions will affect household fuel type actually consumed, the access to or decisions and, in turn, energy markets. constraints on each type of fuel, the socioeconomic characteristics at the household and local (village) Data generated through specialized surveys, a levels.8 Even time-series econometric analysis relying general-purpose LSMS, or the records of utilities on aggregate variables needs at least 20 years of data can be used to create these causal models. If the on electricity sales, the marginal prices of the electricity 20 Willingness to Pay for Water and Energy sold, and the prices of alternate fuels, as well as weather electrifying a zone with no electricity. In this case, think and economic data (for example, income). Sufficient of the second data point, (p1, q1), measured at time data are often not available; electricity supply may be 0, but for a village where electricity is not available. constrained, and other statistical issues arise, such as the The price and quantity refer now to an alternative identification problem. It may also not be possible to source of energy consumed (converted to electrical relate the resultant parameters of the econometric units), kerosene, for example. The consumer surplus in analysis based on a relatively large group of existing this case provides a rough measure of the WTP for consumers, say, on the country level, to the consumer electrification (ESMAP 2002). Once again, the group under investigation. Therefore, the econometric observation of household (aggregate) behavior with approach to demand estimation is usually employed respect to an alternative energy source supplies the sparingly. However, there are various alternative missing (non-market) information for a simple strategies that the analyst can pursue in order to estimate of the electricity demand. estimate the costs and benefits of infrastructure and policy interventions. These clearly depend not only on Clearly, the measurement of WTP here rests the nature of the policy action to be assessed, but also critically on a reliable estimate of the electricity on the type and richness of the data on hand. demand function. A major weakness of the approach in (5) lies in its reliance on a linear The approach that is frequently taken in practice for demand function; the use of the function has no policy or project evaluations is simple. It consists of theoretical basis and is mainly a matter of calculating the consumer surplus (CS in the equation) convenience. on the basis of a linear electricity demand function.9 Say that (p0, q0) represents the average price for electricity Choynowski (2002) presents an alternative approach and the quantity consumed by a representative to deriving the demand for electricity10 and proposes household at time 0 in a particular geographic area. a plausible semilog functional form, which is easy to The second average data point (p1, q1) can be obtained, estimate, is consistent with the underlying for the same area, at a later date (time 1), say, after a microeconomic theory, and readily lends itself to change in the tariff structure. This information should be a calculation of the economic benefit of electricity readily available from the utility supplying the electricity. provision.11 Note, however, that, even with this Consumer surplus of this change is then estimated as: generalization, the present approach is quite different from the econometric estimation of an CS = 1/2( p1- p0)(q1- q0), electricity demand function, which, when employed for predictive purposes (ex-ante analysis), is generally and is simply the area beneath the linear demand based on more than only two data points. Moreover, function, between q1 and q0. The consumer surplus in while the two-point approach is useful in ex-post this case provides an approximation of the change in assessment exercises at the aggregate level and the welfare of the representative household that results when the more demanding approaches are not from the change in the electricity price from p0 to p1. viable, it remains limited in its utility for a social and poverty impact analysis of the policy change in terms Note, though, that the same approach can, in principle, of various subgroups in the population. also be adopted to assess the value to the consumer of Energy Working Notes 21 A number of studies have attempted to evaluate the preference questionnaire for inclusion in the specialized welfare effects of utility privatization and the resulting or LSMS-type survey and to question respondents directly price and access changes. For example, a recent about their demand (willingness to pay) for improved WIDER project sponsored a series of studies to evaluate services. For example, respondents could be asked if the welfare effects of utility privatization in Latin America. they would agree to pay a specified monthly increase in For simplicity, some studies assumed linear or perfectly their electricity bill in exchange for improved service inelastic demand (Waddams Price and Hancock, 1998; reliability. Alternatively, choice modeling techniques might Waddams Price and Young, 2001; Barja and Urquiola, be employed to explore the conditions under which a 2001), while others estimated consumer surplus for a household would switch its fuels for different uses. range of demand elasticities (Freund and Wallich, 1995; Delfino and Casarin, 2001; Torrero and Pasco- 4.3 Willingness to pay for improved Font, 2001). McKenzie and Mookherjee (2003) energy services obtained electricity demand elasticity estimates from a complete demand system and calculated the equivalent The estimation of the value of service quality and compensating variation of the price changes and improvement in the electricity sector goes back to the access provision. While the approach that relies on late 1970s and the 1980s. The early studies relied on assumed or estimated elasticities is helpful for several approaches to value a reduction in outages, evaluating the welfare effect of a price change or the voltage fluctuations, and improvements in system provision of access to a new service, it is of limited reliability (Caves, Herriges, and Windle 1990). The benefit in the evaluation of service quality changes. methodological approaches ranged from examining the cost of backup generators and the valuation of Two other general approaches make use of survey data lost commercial or home production by firms or to estimate the costs and benefits of infrastructure households to stated preference methods. Recent investments and policy interventions, including the studies have adopted a stated choice experiment interventions that affect service quality. The first and framework to provide a marginal valuation of a range simplest approach is to use data on the household of service attributes (Goett, Hudson, and Train 2000). costs of coping with unreliable services as a measure of Nearly all the early and recent studies have valued the benefits of service quality improvement (see section the costs of unreliable services and occasional 2, for an example involving water). The argument is outages in developed countries, where the nature of that, if service quality is improved, households the outages tends to be different from those in the experience economic benefits in the form of cost developing countries.12 savings because they no longer have to spend financial resources trying to cope with unreliable services. In Outages have short-run and long-run effects on general, such avoided coping costs will be a lower household electricity consumption. In the short run, bound on the magnitude of the economic benefits of during an outage, households incur losses from not service improvements. being able to turn on their electric appliances to produce lighting, heating, or cooking services. If they The second approach relies on stated preference data. rely on backup equipment during outages, they will In special circumstances, it may be possible for an incur an associated cost. The same is true for firms. energy or infrastructure analyst to consider a stated While they are likely to incur higher outage costs in the 22 Willingness to Pay for Water and Energy short run, they adjust production methods over the welfare evaluation in the energy sector. A recent World long run to minimize the losses borne during outages. Bank (2001) study uses a revealed preference approach The optimal electricity reliability literature and energy to evaluate the impact of policy reforms in the power demand modeling studies have long recognized the sector on agriculture in India. The analysis is existence of these costs and have applied them as conducted at the farm microeconomic level, within a proxies for the WTP to avoid outages (Pasha, partial equilibrium framework, and focuses on the Ghaus, and Malik 1989; Matsukawa and Fujii impact of policy reforms on production costs, yields, 1994; Beenstock, Goldin, and Haitovsky 1997). cropping patterns, the demand for electricity, and Sanghvi (1983) and Westley (1984) have shown farm incomes for various categories of farmers that studies that apply this approach in a context in (classified as marginal, small, medium, and large). which adaptive response is possible substantially To evaluate the impact of policy reform quantitatively underestimate the WTP to avoid blackouts. at the microeconomic level, an econometric model based on data on the observed choices of farmers is Outages are associated with short-run costs, which employed. This predicts what farmers are likely to do are reflected in the outage costs, as well as with the when the policy changes occur. The study does not long-run costs of coping with unreliable electricity rely on CV methods to elicit responses and the WTP supply. The coping methods may include a search for the changes in policy. The report acknowledges for electricity substitutes, the purchase of non- the potential advantage of the CV approach in electric appliances, fuel switching, or other eliciting such information on a given hypothetical behavioral adjustments. The total shortage costs situation without the need to extrapolate from past consist of the expected outage costs and the behavior. However, it discards such an approach on adaptive response costs, which is the way Sanghvi the grounds that "earlier studies have pointed out (1983) refers to the coping costs. Due to the that responses are very sensitive to the manner in presence of the shortage costs, the WTP for each which the questions are phrased and upon the timing unit of unreliable electricity supply is lower than it of such a survey vis-à-vis the household's recent would be if the supply were more reliable, even if outage experience" (World Bank 2001, page 9). fuel switching does not take place. This implies an outward shift of the electricity demand curve Another recent study by the Energy and Resources associated with an improvement in supply.13 Sanghvi Institute (TERI) in New Delhi relies on a variety of (1983), Dias-Bandaranaike and Munasinghe stated and revealed preference techniques, and (1983), and Westley (1984) contend that ignoring estimates the WTP of firms for improved energy the demand curve shift can bias the results by supply in Haryana and Karnatala (TERI 2001). TERI grossly underestimating the benefits of a reliability surveyed about 500 manufacturing industries and improvement and of the optimal level of reliability. 900 farmers in each of the two Indian states using pre-tested schedules on a stratified sampling basis to 4.3.1 WTP of firms for electricity service collect primary data for the calculation of the cost of improvement unserved energy (CUE). Two recent studies provide an illustration of the use The starting point of the study is the recognition that of stated and revealed preference techniques in the power sector in many developing countries is Energy Working Notes 23 characterized by administered prices, a high level of The production loss method gives the maximum amount cross subsidization, low recovery of revenues, and that a consumer can pay, which may be interpreted as strong political influence. In such a scenario, it is the upper bound on electricity tariffs. A problem with the difficult to arrive at an acceptable restructuring of method is that adjustment processes are not accounted tariffs because the market gives only limited indications for in the responses, with the result that the CUE is often of the burdens that can be borne by different overstated. For example, industrial enterprises that suffer segments. In these circumstances, estimates of the from power cuts will seek to minimize the outage effects value of electricity and the CUE are important for by rescheduling production to other periods. Similarly, both investment planning and tariff-setting. agricultural users can often reschedule the use of pump sets. These rescheduling possibilities are not This study was undertaken in the two Indian states in incorporated into the estimates respondents make of order to provide guidance to decisionmakers regarding their lost production. A further disadvantage of the consumer perceptions of the quality and availability of method is its reliance on the recall ability of respondents. power supply, as well as their willingness to pay different The value of losses may also often be overstated because or higher tariffs for improved power supply. The study the entire loss of production is attributed to the power implemented three methods to estimate the CUE. cut. Moreover, the industries surveyed may be able to offer estimates of the loss in revenue, but may not be · The value of the production lost for each unit of able to estimate the loss in value added. power outage (production loss method); · The cost of alternative or backup power generation Despite these limitations, the production loss method (captive generation method); and provides an estimate of the upper limit on the electricity · The WTP for reliable and uninterrupted electricity tariffs that the firms would be willing to pay. supply (WTP method), that is, a CV analysis. The captive generation method These methods each have advantages and limitations (see below). Obtaining relevant information by Deriving the CUE by the captive generation method is surveying a number of farms and industries is crucial based on gauging the cost incurred by the consumer due for the implementation of the methods. to the use of standby energy generation. In other words, the average economic costs of backup power generation The production loss method by captive units in industries provide an estimate of the CUE. In agriculture, the cost of supplying an equivalent With the production loss method, one looks at the value amount of water using diesel pumps gives estimates of of the production loss that can be attributed to the the CUE by way of the captive generation model. interruption (or other poor quality indicators) of power supply to estimate the CUE. In the agricultural sector, The CUE obtained through this method supplies an the production loss method derives the CUE based on indirect estimate of the tariff a consumer will be willing the incremental crop output not realized (opportunity to pay for electricity. The assumption is that, if a loss) due to the non-availability of power for irrigation. consumer is willing to undertake self-generation, s/he would be willing to pay at least the cost of self- generation as the price for grid power. This is similar to 24 Willingness to Pay for Water and Energy the assumption underlying the revealed preference This approach is also subject to limitations. It is approaches discussed for water (see elsewhere above). highly dependent on questionnaire design and the understanding respondents possess of the hypothetical This approach too, has limitations. On the one hand, scenario (see Box 2 for an example).14 it may yield an overestimate in that consumers may be willing to pay the estimated price only for a The value of the CUE estimated by the CV method fraction of the units (those generated during the is usually above the estimates derived through the power cut), but not for all the electricity units in the captive generation method and below the estimates required production process. In other words, the derived through the production loss method. However, captive generation method only estimates the CUE in developing countries, due to the significant share of at the margin. On the other hand, the price may be poor people, the value is also influenced by consumer underestimated, as some of the costs cannot be ability to pay. Notwithstanding the limitations, this monetized, for example, the additional investment method helps to narrow the range between the upper required, the inconvenience of the hot air and noise and lower limits on electricity tariffs and provides from the generator, and the additional responsibility insights for a more practical policy on electricity tariffs. of manipulating the generator. Moreover, the price is generally only applicable to consumers with an TERI survey results available backup supply. Despite these limitations, the captive generation method provides a lower bound In industry, the three analytical methods applied by on the tariff a consumer is willing to pay at the margin. TERI (2001) gave different estimates of the CUE. For example, in Haryana state, the estimates were Contingent valuation methods rupees (Rs) 7.15 per kWh with the production loss method, Rs 3.38/kWh with the captive generation The consumer is willing to pay an amount that lies method, and Rs 5.16/kWh with the WTP method. As between the upper bound (production loss method) the actual tariff per kWh is only Rs 3.6, there is room and the lower bound (the cost of alternative for a tariff increase according to the methods as long generation). The WTP approach measures the price as this is accompanied by improved service quality. the consumers are willing to pay. The WTP is established through CV methods (estimating As expected, the CUE estimates generated through consumer WTP in a hypothetically improved power the WTP method are higher than those based on supply scenario) based on a bidding game approach. the cost of captive generation for industries, but less than those based on the production loss method. The Relative to the other two methods, the main advantage CUE estimated by the production loss method was of the CV method is that it generates a comprehensive also quite different in the two states. This reflects: measure of the total value of electricity supply to consumers. The success of the method in estimating a. Differences in the percentage share of industries the WTP depends on the extent to which respondents that had installed captive generation capacity are well informed and are able to assess the total (and were thus less reliant on grid supply); value of the electricity and the services provided. b. Differences in the value added per unit of output; and Energy Working Notes 25 c. The diversity in product manufactured and The estimates of the CUE by way of the captive in production processes (which also had generation method were similar in the two states; an effect on the confidence intervals of the the cost in low tension industries was found to be mean CUE). greater than the cost in high tension industries. Box 2. A Sample Questionnaire on WTP for Energy Supply Improvements for Farms (Teri 2001) Scenario Description (Interviewer to Read Out) "As you are aware, businesses in this state face problems of power supply. Some companies make do with their grid supply, while others rely partly or wholly on captive generation. As things stand, you pay a certain amount for grid supply, as well as paying for any damage due to voltage fluctuations and interruptions to supply. You may also pay for standby or captive generation, which can be far more expensive per kWh than grid supply. "An improved electricity supply may increase your company's productivity. However, if output is to be increased, you will incur costs besides the cost of electricity. Looking to the future, it is possible to fix the problems of restricted and poor quality grid supply. To do this will require significant new investment in generation, transmission, and distribution. This new investment is expensive, and the cost would have to be recovered through the electricity tariff." [The question numbering below is the same as in TERI 2001] 24. "We do not yet know exactly how much this investment would cost, but I would like to know if you agree in principle with paying for an improved grid supply?" 24.1 Yes 24.2 No If No, Go to Question 29 Bidding Game 25. "Now I am going to read you some costs per kWh that a stable and uninterrupted electricity supply might cost you. There would not be any other charge for supply. Please tell me if you would be willing to pay this amount for the improved service. Please do not agree to pay if your company cannot afford it or if you feel there are other things that are more important for you to spend money on." Rs/kWh [4 or x] 8.00 2.50 7.00 3.00 6.00 3.50 5.50 4.00 5.00 4.50 26 Willingness to Pay for Water and Energy Box 2. A Sample Questionnaire on WTP for Energy Supply Improvements for Farms (Teri 2001)-continued 26. "Do you have any preference for the power supplier?" 26.1 SEB 26.2 Private agency 26.3 Any other (specify) 26.4 Indifferent 27. "Is the answer to 11.2 [that is, the current tariff] equal to or higher than the highest number ticked in question 25?" 27.1 Yes 27.2 No If Yes, continue. Otherwise, Go to Question 30. Read Out: "From the figures you have given us, it appears that you are currently paying [from 11.2] per kWh for grid supply. You have also said that you are not willing to pay more than this for the improved electricity supply we have described. Is this really the case, or would you like to revise the amount you say you would be willing to pay for improved supply?" 28.1 No change 28.2 Revise bid. If No Change, Go to 29. If Revise Bid, Ask: 28.3 "So what is the maximum you are willing to pay per kWh for a stable and uninterrupted electricity supply?" Rs/kWh________ If 28.3 Completed, Go to 30 29. "Please, could you give the reasons why you are not willing to pay more for the improved electricity supply?" (Interviewer to code various reasons.) Code Mark a tick Reason 01 Cannot afford to pay (threat of bankruptcy) 02 Do not believe they would get improved supply 03 Demand is already met from self-generation 04 Can meet additional demand from self-generation at lower cost 05 Believe the current tariff is already above the cost of grid supply to industry 06 Believe the current tariff is above the industrial tariff in other states 07 Other (please specify) 30. "Which of the following statements best describes the way you felt when we described the option of an improved supply to you?" (Tick One Only] 30.1 The introduction of new investment in the electricity sector in this state has a good chance of improving the electricity supply. 30.2 This new investment may improve the electricity supply, but it will take so long to produce benefits that it is not worth much to me. 30.3 I just did not believe that the system for improved supply you described could be introduced or made to work in our state. Energy Working Notes 27 Some of the known difficulties were encountered in 4.3.2 Willingness of households to pay for the WTP method. Thus, a percentage of the energy service improvements industries surveyed did not believe that a scenario of improved power supply (reliable and good The production loss method and the captive quality; see the questionnaire in Box 2) from the grid generation method cannot easily be applied to was possible. Of those who accepted the scenario, households. The first method requires an estimation only about half were willing to pay more than they of the value of production in the home and the currently paid. (The estimated WTP is about Rs contribution of energy as an input to this production 5.2/kWh in both states.) The industries unwilling to pay process. Residential outage costs are the market value more expressed concerns that the current cost was of services lost due to an interruption. However, unaffordable--the industrial tariff was already above obtaining the market value of these services is difficult. the cost of supply--and stated that they would rather Since households are able to shift household rely on generation. Probit regressions were used to production to a time when electricity supply is restored, analyze the factors correlated to the willingness of the the production loss calculation is complicated further industries to pay more (yes/no answer). The variables (Caves, Herriges, and Windle 1990). The second that tended to be associated with the decision to bid method is based on the premise that a household will for improved electricity supply were: respond to poor service reliability by using backup generation equipment.15 The purchase and a. The small share of electricity to total cost; operating costs of this equipment provide a b. The low tariffs/prices compared to other states; measure of the losses due to poor service reliability. c. The type of industry and its sensitivity to While this is a reasonable approach to estimate the continuous electricity supply; losses of firms, many households do not own d. The experience of high levels of unscheduled backup generation equipment. It may thus be more outages; and useful to obtain information on the amounts e. The unofficial payments to state electricity households spend on electricity substitutes so as to boards for power restoration. adjust to poor supply quality or lack of supply. The results highlight the factors to be addressed so CV methods can be applied to households, as well that consumers can be persuaded to pay higher tariffs. as to commercial and industrial sectors in order to These factors include institutional issues (such as estimate the value of the improved reliability of unofficial payments made to a utility's staff) and factors energy services. However, the application of these related to the quality of electricity supply. In particular, methods in the case of energy is less straightforward the study underlines the urgent need to address than it is in the case of water. People are not used unscheduled outages, followed by scheduled outages to thinking of electricity consumption in terms of and voltage fluctuations; consumers should be kilowatt hours, whereas they are familiar with the provided with better information about supply so measurement of the consumption of water in terms they can plan their industrial activities accordingly of, for example, buckets. As a result, the questions and limit their production losses. about consumption levels at different prices in the TERI (2001) survey had to be phrased in terms of capacity rather than kWh. An and others (2002) 28 Willingness to Pay for Water and Energy were faced with a similar problem when some of the properly the attributes of a service or program that is survey respondents in China did not understand how being evaluated, as well as the so-called levels of each a range of prices per kWh would translate on a attribute. (Levels are the ranges that characterize each monthly electricity bill. If respondents do not have a attribute.) For example, the attributes might include the clear understanding of the unit of measurement, this outage levels, voltage fluctuations, the provision or can become a serious weakness of a CV study. In a non-provision of prior notification of an outage, and CV survey of households in Azerbaijan, Lampietti the price levels that are used in the various versions and others (2004) posed questions about of choice experiments. consumption levels at different prices in terms of the hours of use of each electric appliance rather than in terms of total monthly kilowatt hours.16 5 CONCLUDING REMARKS A range of service attributes may be important in To estimate the value of improved water or energy electricity sector reforms, and a service quality supply, analysts need to elicit the preferences of improvement cannot always be captured by one beneficiaries (or the consumer demand for the characteristic. A survey of Azeri households good). This is more difficult in cases where the (Lampietti and others 2004) revealed that low voltage analysis cannot rely on price and quantity data on is nearly as important an obstacle to the use of the goods and services involved. In these cases, electric appliances as are outages. Stated choice analysts have to resort to alternative, non-market experiments are particularly useful for eliciting the strategies to elicit the necessary information for a marginal WTP for different service attributes. An and cost-benefit analysis. others (2002) applied this framework to estimate electricity demand and to predict the likely impact of This paper has reviewed two main approaches a reduction in voltage fluctuations and in outage available when the standard estimation of the frequency on fuelwood use. Box 3 briefly presents this demand system is not viable: the contingent study of willingness to switch between energy sources valuation (or stated preference) approach and the (from fuelwood to electricity), with implications for giant coping cost method (or revealed preference in a panda habitat conservation in China. The marginal surrogate market). Examples of how to conduct each WTP for a reduction in voltage fluctuations and type of analysis have been provided with reference outages can easily be calculated once the demand to policy interventions aimed at improving the function and the parameters on service quality variables supply of water and energy in developing countries. are known. Thus, stated choice experiments, if carefully designed, permit the analysis of a range of The paper suggests that contingent valuation and policy issues and the tradeoffs faced by households. coping costs methods are useful tools for the evaluation of water and energy projects. This is so Since stated choice questions tend to be more particularly when analysts are aware of the technical complicated than the WTP questions, it is critically and practical issues to be confronted before these important to conduct focus group discussions and non-market preference-elicitation techniques may other qualitative work prior to finalizing the survey be trusted. instrument. This preparatory work helps to specify Energy Working Notes 29 Box 3. Modeling the Choice of Households to Switch between Energy Sources Despite its status as a nature reserve, Wolong Nature Reserve (China) has experienced continued loss of giant panda habitat due to human activities such as fuelwood collection. Though available throughout Wolong, electricity has not replaced fuelwood as an energy source. A recent paper by An and others (2002) relied on stated preference data obtained from in-person interviews to estimate a random utility model of the choice of adopting electricity for cooking and heating. The goal of the study was to estimate the demand for electricity under different conditions of price and quality, as well as to relate the demand to demographic characteristics. Although critical for the approach undertaken, adequate market data spanning these conditions were unavailable. For example, the amount of electricity employed by each household was not accurately recorded or not available in some areas, and the variation in electricity prices was small. Stated preference techniques were applied to overcome this problem. Willingness to switch to electricity was explained by demographic and electricity factors such as price, voltage, and outage frequency. In particular, the differences between the current value of these energy variables (that is, in the status-quo fuelwood-use scenario) and the hypothesized scenario (switching to electricity) were adopted as crucial explanatory determinants of the willingness to switch. The results suggested that lowering electricity prices and increasing electricity quality (mainly the voltage levels) would encourage local residents to switch from fuelwood to electricity and should be considered in the mix of policies aimed at promoting panda habitat conservation. Contingent valuation surveys are now becoming Coping cost methods also supply reasonably accurate relatively easy and more straightforward. The survey estimates of current time and cost expenditures. The response rates are typically high in developing coping costs avoided will generally represent a lower countries where respondents are receptive to listening bound on the magnitude of the economic benefits to and considering the questions asked. Moreover, it of service improvements and, hence, of willingness is much cheaper to carry out in-person surveys and to pay, reflecting the fact that often only a fraction of therefore to assemble larger sample sizes in the actual coping costs are observed by the analyst. developing countries than it is in industrial ones. If a lower bound estimate of benefits is greater than Assuming a thorough survey is undertaken, CV the costs of the intervention, this benefit estimate can methods can provide analysts with good data on the prove especially helpful. On the other hand, the data willingness to pay and on the potential revenues for and analysis requirements tend to be relatively modest. various levels of service. They can thus guide tariff subsidy and cost recovery policy. Another advantage The importance of cross-validating the WTP results lies in the similarity of the CV to public opinion polls, with the two approaches has also been highlighted which means that results are conceptually easy for in the paper as an example of best-practice strategy non-specialists and politicians to understand. Overall, whenever this is a feasible option for the analyst. it seems increasingly feasible and desirable to use CV studies in developing countries to help evaluate While the usefulness of revealed and stated a wide range of projects. However, CV researchers preference techniques for valuing non-market goods should apply caution in the design and implementation has been the main focus of the paper, this should not of the survey and take into account the specifics be taken as an attempt to dismiss more traditional and relating to a particular cultural and ethical setting. relatively well-known techniques for estimating a 30 Willingness to Pay for Water and Energy demand curve for water and energy services based on BIBLIOGRAPHY market price-quantity data. On the contrary, if it is practical for analysts to employ them, for example, The word processed describes informally when a rich and detailed dataset is available, then reproduced works that may not be commonly these techniques are likely to offer a more consistent, available through libraries. more reliable, and probably less controversial framework for analysis. However, the conditions Abelson, Peter W. 1979. "Property Prices and the necessary for the appropriate application of these Value of Amenities." Journal of Environmental methods (mainly, but not only, the required information Economics and Management 6: 11-28. base) are unlikely to be met in practice in most developing countries. Abrahams, Nil Adote, Bryan J. Hubbell, and Jeffrey L. Jordan. 2000. "Joint Production and Averting Expenditure Measures of Willingness to Pay: Do Water Expenditures Really Measure Avoidance Costs?" American Journal of Agricultural Economics 82, no. 2: 427­37. Ajzen, Icek, and B. L. Driver. 1992. "Contingent Value Measurement: On the Nature and Meaning of Willingness to Pay." Journal of Consumer Psychology 1, no.4: 297­316. Alberini, Anna, and James Cooper. 2000. "Applications of the Contingent Valuation Method in Developing Countries: A Survey." Economic and Social Development Paper 146. Food and Agriculture Organization of the United Nations, Rome. An, Li, Frank Lupi, Jianguo Liu, Marc A. Linderman, and Jinyan Huang. 2002. "Modeling the Choice to Switch from Fuelwood to Electricity: Implications for Giant Panda Habitat Conservation." Ecological Economics 42, no. 3: 445­57. Arimah, Ben C. 1992. "Hedonic Prices and the Demand for Housing Attributes in a Third World City: The Case of Ibadan, Nigeria." Urban Studies 29, no. 5: 639­51. Energy Working Notes 31 Arrow, Kenneth, Robert Solow, Paul R. Portney, Edward Bergstrom, John. 1990. "Concepts and Measures of E. Leamer, Roy Radner, and Howard Schuman. the Economic Value of Environmental Quality: A 1993. "Report of the NOAA Panel on Contingent Review." Journal of Environmental Management 31: Valuation." Federal Register 58, no. 10: 4601­14. 215-28. Asian Development Bank. 1999. Handbook for the Bishop, Richard C., and Thomas A. Heberlein. 1990. Economic Analysis of Water Supply Projects. Asian "The Contingent Valuation Method." In Rebecca L. Development Bank, Manila. Johnson and Gary V. Johnson, eds., Economic Valuation of Natural Resources. Boulder, CO: Bamberger, Michael, ed. 2000. "Integrating Westview Press. Quantitative and Qualitative Research in Development Projects." Directions in Development. World Bank, Blamey, R. K., J. W. Benett, and M. D. Morisson. Washington, D. C. 1999. "Yea-Saying in Contingent Valuation Surveys." Land Economics 75, no. 1: 126­41. Barja, Gover, and Miguel Urquiola. 2001. "Capitalization, Regulation and the Poor: Access to Blomquist, Glenn, and Lawrence Worley. 1981. Basic Services in Bolivia." WIDER Discussion Paper "Hedonic Prices, Demands for Urban Amenities, 2001/34. United Nations University­World Institute and Benefit Estimates." Journal of Urban Economics for Development Economics Research, Helsinki. 9, no. 2: 212­21. Bartik, Timothy J. 1988. "Evaluating the Benefits of Boardman, Anthony E., David H. Greenburg, Aidan Non-Marginal Reductions in Pollution Using R. Vining, and David L. Weimer. 2001. Cost-Benefit Information on Defensive Expenditures." Journal of Analysis: Concepts and Practice. Upper Saddle River, Environmental Economics and Management 15: NJ: Prentice Hall. 111-27. Bohara, Alok K., Michael McKee, Robert P. Berrens, Bateman, Ian J., Ian H. Langford, and Jon Rasbash. Hank Jenkins-Smith, Carol L. Silva and David S. 1999. "Willingness to Pay Question Format Effects Brookshire. 1998. "Effects of Total Cost and Group- in Contingent Valuation Studies." In Ian J. Bateman Size Information on Willingness to Pay Responses: and Kenneth G. Willis, eds., Valuing Environmental Open-Ended vs. Dichotomous Choice." Journal of Preferences: Theory and Practice of the Contingent Environmental Economics and Management 35, no. 2: Valuation Method in the US, EU, and Developing 142­63. Countries. New York: Oxford University Press. Boyle, Kevin J., and Richard C. Bishop. 1988. Beenstock, Michael, Ephraim Goldin, and Yoel "Welfare Measurements Using Contingent Haitovsky. 1997. "The Cost of Power Outages in Valuation: A Comparison of Techniques." American the Business and Public Sectors in Israel: Revealed Journal of Agricultural Economics 70, no. 1: 20-28. Preference vs. Subjective Valuation." The Energy Journal 18. no. 2: 39­61. 32 Willingness to Pay for Water and Energy Boyle, Kevin J., Richard C. Bishop, and Michael P. Carson, Richard T., Nicholas E. Flores, Kerry M. Welsh. 1985. "Starting Point Bias in Contingent Martin, and Jennifer L. Wright. 1996. "Contingent Valuation Bidding Games." Land Economics 61, no. 2: Valuation and Revealed Preference Methodologies: 188-94. Comparing the Estimates for Quasi-Public Goods." Land Economics 72, no. 1: 80­99. Brookshire, David S., Mark A. Thayer, William Schulze, and Ralph C. d'Arge. 1982. "Valuing Public Goods: Carson, Richard T., W. Michael Hanemann, Raymond A Comparison of Survey and Hedonic Approaches." J. Kopp, Jon A. Krosnick, Robert C. Mitchell, Stanley American Economic Review 72, no. 1: 165­77. Presser, Paul A. Rudd, and V. Kerry Smith. 1998. "Referendum Design and Contingent Valuation: The Brown, T., A. Champ, Richard C. Bishop, and D. W. NOAA Panel's No-Vote Recommendation." Review McCollum. 1996. "Which Response Format Reveals of Economics and Statistics 80, no. 3: 484­87. the Truth about Donations to a Public Good?" Land Economics 72, no. 1: 152­66. Caves, Douglas, Joseph A. Herriges, and Robert Windle. 1990. "Customer Demand for Service Calia, P., and E. Strazzera. 2000. "Bias and Efficiency Reliability in the Electric Power Industry: A Synthesis of Single Versus Double Bounded Models for of the Outage Cost Literature." Bulletin of Economic Contingent Valuation Studies: A Monte Carlo Analysis." Research 42, no. 2: 79­119. Applied Economics 32, no. 10: 1329­36. Choynowski, Peter. 2002. "Measuring Willingness to Carson, Richard T. 1997. "Contingent Valuation: Pay for Electricity." ERD Technical Note Series no. 3. Theoretical Advances since the NOAA Panel." Asian Development Bank, Economics and Research American Journal of Agricultural Economics 79, no. 5: Department, Manila. 1501­07. Courant, Paul N., and Richard C. Porter. 1981. ------. 2000. "Contingent Valuation: A User's "Averting Expenditure and the Cost of Pollution." Guide." Environmental Science and Technology 34, Journal of Environmental Economics and no. 8: 1413­28. Management 8, no. 4: 321-29. Carson, Richard T., and Robert C. Mitchell. 1993. "The Crane, Randall, and Amrita Daniere, with Stacey Value of Clean Water: The Public's Willingness to Harwood. 1997. "The Contributions of Pay for Boatable, Fishable, and Swimmable Quality Environmental Amenities to Low-Income Housing: A Water." Water Resources Research 29, no. 7: Comparative Study of Bangkok and Jakarta." Urban 2445­54. Studies 24, no. 9: 1495­512. Carson, Richard T., Nicholas E. Flores, and Norman F. Cummings, Ronald G., David S. Brookshire, and Meade. 2001. "Contingent Valuation: Controversies William D. Schulze, eds. 1986. Valuing and Evidence." Environmental and Resource Environmental Goods: A State of the Arts Economics 19, no. 2: 173­210. Assessment of the Contingent Valuation Method. Totowa, NJ: Rowman and Allanheld. Energy Working Notes 33 Daniere, Amrita G. 1994. "Estimating Willingness Follain, James R., and Emmanual Jimenez. 1985. to Pay for Housing Attributes: An Application to "The Demand for Housing Characteristics in Cairo and Manila." Regional Science and Urban Developing Countries." Urban Studies 22: 421­32. Economics 24, no. 5: 577­99. Freeman, A. Myrick, III. 1993. The Measurement of Delfino, Jose, and Ariel Casarin. 2001. "The Environmental and Resource Values: Theory and Reform of the Utilities Sector in Argentina." WIDER Methods. Washington, D. C.: Resource for the Discussion Paper 2001/74. United Nations Future Press. University­World Institute for Development Economics Research, Helsinki. Freund, Caroline, and Christine Wallich. 1995. "Raising Household Energy Prices in Poland: Who DfID (Department for International Development). Gains? Who Loses?" Policy Research Working Paper 1998. Guidance Manual on Water and Sanitation 1495. World Bank, Country Department II, Europe Projects. Department for International Development, and Central Asia, Office of the Director, London. Washington, D. C. Diamond, Peter, and Jerry A. Hausman. 1994. Goett, Andrew, Kathleen Hudson, and Kenneth "Contingent Valuation: Is Some Number Better Than Train. 2000. "Customers' Choice among Retail No Number." Journal of Economic Perspectives 8, Energy Suppliers: The Willingness-to-Pay for Service no. 4: 45­64. Attributes." The Energy Journal 21, no. 4: 1­28. Dias-Bandaranaike, R., and M. Munasinghe. 1983. Green, C. H., and S. M. Tunstall. 1991. "The "The Demand for Electricity and the Quality of Evaluation of River Quality Improvements by the Supply." The Energy Journal 4, no. 2: 49­71. Contingent Valuation Method." Applied Economics 23, 1135­46. Edwards, Steven. 1988. "Option Prices for Groundwater Protection." Journal of Environmental Griffin, Charles C., John Briscoe, Bhanwar Singh, Economics and Management 15, no. 4: 475-87. Radhika Ramasubban, and Ramesh Bhatia. 1995. "Contingent Valuation and Actual Behavior: Predicting ESMAP (Joint UNDP, World Bank Energy Sector Connections to New Water Systems in the State of Management Assistance Programme). 2000. Kerala, India." World Bank Economic Review 9, no. 3: "Energy Services for the World's Poor." ESMAP 373­95. Energy and Development Report. World Bank, Washington, D. C. Gross, D. J. 1988. "Estimating Willingness to Pay for Housing Characteristics: An Application of the ------. 2002. "Rural Electrification and Ellickson Bid-Rent Model." Journal of Urban Development in the Philippines: Measuring the Economics 24, no. 1: 95­112. Social and Economic Benefits." ESMAP Report 255/02. World Bank, Washington, D. C. 34 Willingness to Pay for Water and Energy Haab, Timothy C., and Kenneth E. McConnell. Johansson, Robert C. 2000. "Pricing Irrigation 2002. Valuing Environmental and Natural Water: A Literature Survey." Policy Research Working Resources: The Econometrics of Non-Market Paper 2449. World Bank, Development Research Valuation. Cheltanham, U. K.: Edward Elgar. Group, Rural Development, Washington, D. C. Hanemann, W. Michael. 1984. "Discrete- Kaufman, Daniel, and John M. Quigley. 1987. Continuous Models of Consumer Demand." "The Consumption Benefits of Investment in Econometrica 52, no. 3: 541-61. Infrastructure: The Evaluation of Sites-and-Services Programs in Underdeveloped Countries." Journal of ------. 1994. "Valuing the Environment through Development Economics 25, April: 263­84. Contingent Valuation." Journal of Economic Perspectives 8, no. 4: 19­43. Klugman, J., ed. 2002. A Sourcebook for Poverty Reduction Strategies, 2 vols. World Bank, Hanemann, W. Michael, John B. Loomis, and Washington, D. C. Barbara J. Kanninen. 1991. "Statistical Efficiency of Double-Bounded Dichotomous Choice Contingent Korman, Vijdan. 2002. "Willingness to Pay for Valuation." American Journal of Agricultural Water Improvements in Famagusta, North Cyprus." Economics 73, no. 4: 1255-63. Ph. D. Dissertation. Northeastern University, Boston. Hensher, David, Nina Shore, and Kenneth Train. Lampietti, Julian, Anthony Kolb, Gulyani Sumila, 2003. "Households' Willingness to Pay for Water and Avenesyan Vahram. 2001. "Utility Pricing and Service Attributes." University of Sydney, Faculty of the Poor: Lessons from Armenia." World Bank Economics and Business, School of Business, Technical Paper 497. Washington, D. C. Institute of Transport Studies, Sydney. Processed. Lampietti, Julian, and others. 2004. "Azerbaijan Hentschel, Jesko, and Peter Lanjouw. 1997. Household Energy Survey: Survey Documentation." "Household Welfare Measurement and the Pricing World Bank, Washington, D. C. Processed. of Basic Services." World Bank, Poverty Reduction and Economic Management Network, Poverty Lauria, Donald T., Dale Whittington, Kyeongae Group, and Development Research Group, Choe, Cynthia Turingan, and Virginia Abiad. 1999. Washington, D. C. Processed. "Household Demand for Improved Sanitation Services: A Case Study of Calamba, the Herriges, Joseph A., and Jason F. Shogren. 1996. Philippines." In Ian J. Bateman and Kenneth G. "Starting Point Bias in Dichotomous Choice Valuation Willis, eds., Valuing Environmental Preferences: with Follow-Up Questioning." Journal of Environmental Theory and Practice of the Contingent Valuation Economics and Management 30, no. 1: 112­31. Method in the US, EU, and Developing Countries. New York: Oxford University Press. Jimenez, Emmanuel. 1982. "The Value of Squatter Dwellings in Developing Countries." Economic Development and Cultural Change 30, no. 4: 739­52. Energy Working Notes 35 Lee, Linda K., and L. Joe Moffitt. 1993. "Defensive Mitchell, Robert C., and Richard T. Carson. 1989. Technology and Welfare Analysis of Environmental Using Surveys to Value Public Goods: The Quality Change with Uncertain Consumer Health Contingent Valuation Method. Washington D. C.: Impacts." American Journal of Agricultural Resources for the Future Press. Economics 75, no. 2: 361­66. Mu, Xinming, Dale Whittington, and John Briscoe. Leggett, Christopher G., and Nancy E. Bockstael. 1990. "Modeling Village Water Demand Behavior: 2000. "Evidence of the Effects of Water Quality on A Discrete Choice Approach." Water Resources Residential Land Prices." Journal of Environmental Research 26, no. 4: 521­29. Economics and Management 39, no. 2: 121­44. North, James H., and Charles C. Griffin. 1993. Linneman, Peter. 1981. "The Demand for Residence "Water Source as a Housing Characteristic: Site Characteristics." Journal of Urban Economics 9, Hedonic Property Valuation and Willingness to Pay." no. 2: 129­48. Water Resources Research 29, no. 7: 1923­29. Louviere, J., D. Hensher, and J. Swait. 2000. Stated O'Doherty, Richard. 1998. "The Theory of the Choice Methods: Analysis and Applications. New York: Contingent Valuation Method." Hume Papers on Cambridge University Press. Public Policy 6, no. 3: 67­87. Matsukawa, Isamu, and Yoshifumi Fujii. 1994. Pasha, Hafiz A., Aisha Ghaus, and Salman Malik. "Customer Preferences for Reliable Power Supply: 1989. "The Economic Cost of Power Outages in the Using Data on Actual Choices of Back-Up Industrial Sector of Pakistan." Energy Economics 11, Equipment." Review of Economics and Statistics 76, no. 4: 301­18. no. 3: 434­46. Perez-Pineda, Felipe. 1999. "Estimating Willingness McKenzie, David, and Dilip Mookherjee. 2003. to Pay for Water Services in Developing Countries: "Distributive Impact of Privatization in Latin America: A Case Study of the Use of a Contingent Valuation An Overview of Evidence from Four Countries." Survey in El Salvador, Central America." Ph.D. Economia 3, no. 2: 161­218. dissertation. Purdue University, West Lafayette, IN. McPhail, Alexander A. 1993. "The `Five Percent Rule' Portney, Paul R. 1994. "The Contingent Valuation for Improved Water Service: Can Households Afford Debate: Why Economists Should Care." Journal of More?" World Development 21, no. 6: 963­73. Economic Perspectives 8, no. 4: 3­17. ------. 1994. "Why Don't Households Connect to Ready, Richard, Janis Malzubris, and Silva Senkane. the Piped Water System?: Observations from Tunis, 1997. "A Contingent Valuation Study Estimating Tunisia." Land Economics 70, no. 2: 189­96. Willingness to Pay for Improved Water Quality in Sigulda, Latvia." Report to the C4EP Project, Harvard University, Harvard Institute for International Development, Cambridge, MA. 36 Willingness to Pay for Water and Energy Revelt, David, and Kenneth Train. 1998. "Mixed Logit Waddams Price, Catherine, and Ruth Hancock. with Repeated Choices: Households' Choices of 1998. "Distributional Effects of Liberalizing UK Appliance Efficiency Level." Review of Economics and Residential Utility Markets." Fiscal Studies 19, no. 3: Statistics 80, no. 4: 647­57. 295­319. Rosen, Sherwin. 1974. "Hedonic Prices and Implicit Waddams Price, Catherine, and Alison Young. Markets: Product Differentiation in Pure Competition." 2001. "UK Utility Reforms: Distributional Journal of Political Economy 82, no. 1: 34­55. Implications and Government Response." WIDER Discussion Paper 2001/10. United Nations Rubey, Lawrence, and Frank Lupi. "Predicting the University­World Institute for Development Effects of Market Reform in Zimbabwe: A Stated Economics Research, Helsinki. Preference Approach." American Journal of Agricultural Economics 79, no. 1: 89­99. Welsh, Michael P., and Gregory L. Poe. 1998. "Elicitation Effects in Contingent Valuation: Sanghvi, Arun. 1983. "Optimal Electricity Supply Comparisons to a Multiple Bounded Discrete Reliability Using Customer Shortage Costs." Energy Choice Approach." Journal of Environmental Economics 5, no. 2: 129­36. Economics and Management 36, no. 2: 170­85. Schultz, Steven D., and Bruce E. Lindsay. 1990. "The Westley, Glenn. 1984. "Electricity Demand in a Willingness to Pay for Groundwater Protection." Developing Country." Review of Economics and Water Resources Research 26, no. 9: 1869-75. Statistics 66, no. 3: 459­67. Smith, V. Kerry. 2000. "Fifty Years of Contingent Whitehead, John C., and George Van Houtven. Valuation." Paper Presented at the Kobe Conference on 1997. "Methods for Valuing the Benefits of the Safe the Theory and Application of Environmental Valuation, Drinking Water Act: Review and Assessment." Japan Forum of Environmental Valuation, Kobe Department of Economics Working Paper 9705, University, Rokkoudai Campus, Kobe, Japan, January. East Carolina University, Department of Economics, Greenville, NC. TERI (The Energy and Resources Institute). 2001. "Cost of Unserved Energy." Project Report 98PG42. Whittington, Dale. 1998. "Administering Contingent Submitted to the World Bank and Department for Valuation Studies in Developing Countries." World International Development (U. K.). The Energy and Development 26, no. 1: 21­30. Resources Institute, New Delhi. ------. 2000. "Environmental Issues." In Margaret Torrero, Máximo, and Alberto Pascó-Font. 2001. E. Grosh and Paul Glewwe, eds., Designing "The Social Impact of Privatization and the Household Survey Questionnaires for Developing Regulation of Utilities in Peru." WIDER Discussion Countries: Lessons from 15 Years of the Living Paper 2001/19. United Nations University­World Standards Measurement Study, 3 vols. World Bank, Institute for Development Economics Research, Helsinki. Washington, D. C. Energy Working Notes 37 Whittington, Dale, Donald T. Lauria, and Xinming NOTES Mu. 1991. "A Study of Water Vending and Willingness to Pay for Water in Onitsha, Nigeria." 1 For instance, a study by Lampietti and others (2001) World Development 19, no. 2­3: 179­98. has analyzed the scope and consequences of recent utility pricing reforms in both the water and energy Whittington, Dale, Xinming Mu, and Robert Roche. sector in Armenia. The motivation of the analysis, which 1990. "Calculating the Value of Time Spent was undertaken using some of the techniques discussed Collecting Water: Some Estimates for Ukunda, in this paper, was to understand how cost recovery Kenya." World Development 18, no. 2: 269­80. efforts in water and electricity affect the poor, the challenge for the government of Armenia being to Whittington, Dale, K. O'Sullivan, and D. Barnes. strengthen the financial stability of the utilities, while 2004. "Guidelines for Designing Energy Modules in improving service quality and minimizing the potential Living Standard Measurement Studies." World Bank, welfare burden of these efforts on the poor. Washington, D. C. Processed. 2 Here "costs" refer to the costs incurred by the state (or Whittington, Dale, John Briscoe, Xinming Mu, and private agency) that will undertake the project or policy William Barron. 1990. "Estimating the Willingness intervention, while "benefits" are the benefits accruing to Pay for Water Services in Developing Countries: to the final service users (for example, consumers) and A Case Study of the Use of Contingent Valuation should actually be thought of as net benefits (that is, Surveys in Southern Haiti." Economic Development net of the costs incurred by the final users themselves). and Cultural Change 38, no. 2: 293­311. 3 The assessment of a more comprehensive notion of Whittington, Dale, Jennifer Davis, Harry Miarsono, costs that also includes, for example, externalities, or and Richard Pollard. 2000. "Designing a non-use value in the case of environmental goods, `Neighborhood Deal' for Urban Sewers: A Case remains a more difficult issue. In the case of public goods Study of Semarang, Indonesia." Journal of Planning (for instance, turning a natural area into a public park), Education and Research 19, no. 3: 297­308. some of the infrastructure costs (for example, building costs) might still be easily computed; others, however, World Bank. 2001. "India: Power Supply to such as those generally referred to as "negative Agriculture." Report 22171-IN. South Asia Region, externalities," are much more difficult to evaluate. The Energy Sector Unit, Washington, D. C. park might, for instance, subtract pasture opportunities from a nearby farm or have a negative impact on wildlife Wu, Pei-Ing, and Chu-Li Huang. 2001. "Actual conservation. These costs, too, might be computed in Averting Expenditure Versus Stated Willingness to principle, but arguably more controversially and with a Pay." Applied Economics 33, no. 2: 277­83. smaller degree of confidence than in the case of standard types of costs. Similarly, some of the benefits might be easily estimated (in the park example, the travel costs saved by local households that now find a closer recreational area or tourist-related earnings for the local community), but other types of benefits, such as those 38 Willingness to Pay for Water and Energy reflecting the "intrinsic" value of the park for the into the production of services from a stock of community of beneficiaries involved, as well as the electricity-consuming equipment in the household. As "positive externalities" of less air pollution and better air such, it should not enter directly into the household's quality, are of more difficult assessment in money terms. utility function, but indirectly, through the user cost associated with the services produced by the 4 For a basic discussion of such issues, see Boardman electricity-consuming equipment. and others 2001. 11In particular, it features a negative relationship 5 For example, a quasi-public good such as public between price and quantity and, unlike many single water provision as opposed to pure-public goods such equation and demand system models, also allows for as a remote wilderness area or the overall level of the possibility that the demand may be finite even at air quality. CV methods are usually the only feasible zero price. Moreover, it does not assume a constant methods for including passive use considerations in price elasticity. The functional form is written ln(q) = the economic analysis of environmental goods or for a+bp, where a and b are parameters to be estimated. assessing pure public goods. Economic benefit (EB in the equation) is simply the area beneath the (nonlinear) demand function. It can 6 A specialized survey as opposed to an LSMS-type be reckoned as EB = q1(p1- 1/b)- q0(p0-1/b). survey, non-survey information, data from the records of the utilities, and so on. 12Two exceptions are an early study by Dias- Bandaranaike and Munasinghe (1983) that valued an 7 However, often the price paid does not reflect improvement in service reliability in Costa Rica and a consumer benefit or the willingness of consumers to recent study by An and others (2002) that conducted pay on account of various market distortions (for stated choice experiments to value a service reliability example, taxes and subsidies) that further complicate, improvement and the implications of the resulting both conceptually and econometrically, the demand fuel switching for a giant panda habitat in China. estimation based on observed price and quantities. See, for example, Hentschel and Lanjouw 1997. 13The area between the two demand curves provides an approximation of the true welfare 8 For an example of a detailed state-of-the-art change when the weak complementarity assumption questionnaire on household energy usage that is satisfied. In this setting, this assumption means assembles potentially all the information required for that the welfare level of those households that do a complete household microeconometric model not consume electricity is not affected by a change (including CV modules), see the Azerbaijan in the quality of supply. If this assumption is not Household Energy Survey (Lampietti and others 2004). satisfied, then the area between the demand curves for electricity at the original and improved quality 9 By definition, consumer surplus, plus revenue, levels does not capture all of the welfare gains from equals gross economic benefits (or willingness to pay). a quality improvement, and this area has no welfare significance. It is implicit in the analysis of Sanghvi 10As Choynowski (2002) stresses, the demand for (1983) and Westley (1984) that the weak energy is a derived demand and is essentially an input complementarity assumption is satisfied. Energy Working Notes 39 14TERI (2001) notes that, in the agriculture sector, this turned out to be too hypothetical for those difficulty was faced because most of the farmers were households that have become accustomed to the not used to kWh billing. As a result, the questionnaire poor quality of electricity supply in some rural areas in had to be designed for capacity (hp) based tariffs. The the last 10 years. Many households could not answer responses received from respondents showed that, while the questions because they did not believe the a significant number of farmers were willing to pay more scenario of a service quality improvement, because on a per hp basis, the same did not hold true on a per they were angry at the possibility of a price increase, kWh basis. As the additional hours of power demand or simply because they were no longer used to were factored into the analysis with the bid levels, only operating their electric stoves and heaters. an insignificant number of farmers were found willing to pay more on a kWh basis. As a result, no statistically significant conclusions could be drawn in the agriculture Working Notes sector through this method. The Working Notes series of the Energy and Mining Sector Board are intended to complement the 15 Matsukawa and Fujii (1994) found that few residential Energy and Mining Sector Board Discussion Papers. customers in Japan owned backup generation Working Notes are lightly edited notes prepared by World Bank staff on topical issues in the energy equipment. They had to narrow their study of outage sector. Working Notes are only available costs to industrial and commercial customers. electronically at www.worldbank.org/energy. Comments should be emailed to the authors(s). 16The 2003 Azerbaijan Household Energy Survey (Lampietti and others 2004) was conducted as part of a Disclaimer World Bank poverty and social impact analysis of The findings, interpretations, and conclusions expressed in this paper are entirely those of the proposed reforms in the electricity sector. The first authors and should not be attributed in any manner objective of the survey was to provide a measure of the to the World Bank, to its affiliated organizations, or WTP for a reduction in outages. The second objective to members of its Board of Executive Directors or was to supply information about hypothetical quantity the countries they represent. demand at a range of prices so as to enable the estimation of electricity demand. It was important to ask questions about the quantities respondents would consume at the new prices after the implementation of the future reform, assuming a perfect reliability level. Questions about the hours of use of appliances were clearer to respondents than were questions in terms of monthly electricity consumption in kilowatt hours. Since access to electricity is universal in Azerbaijan (though service quality has deteriorated in many rural areas in the last decade), Azeri households are familiar with electricity as a service and are accustomed to a range of electric appliances. However, even the questions in terms of hours of use of each appliance