21166 Viewpoint'6 Note No. 212 June T Infrastructure Reform, Better Subsidies, and the Information Deficit Andres G6mez- In developing countries the provision of water and sanitation services is often subsidized. These Lobo, Vivien Foster, and subsidies take the form of a general underpricing of water, numerous cross-subsidies, and inefficient Jonatban billing and collection. An essential part of infrastructure reform is the redesign of subsidies. In the Halpern design of an optimal subsidy scheme the key decisions are the choice of eligibility criteria, the level of the subsidy, and the budgetary requirements. However, the lack of consistent and reliable data sets which combine socioeconomic and water consumption information may be an important obstacle to making good decisions, undermining efforts to provide affordable water services for the poor. This Note discusses the type of information required, where it can be found, and ways to deal with shortcomings in the data. To illustrate, the Note draws on data from World Bank work in Panama. There are three types of information required to durable goods ownership and the property make informed choices on subsidy and other value of the dwelling. This information is use- policies in the water sector. These are: ful for determining the target population group * Willingness-to-pay data. Willingness to pay is for the subsidy, studying the targeting proper- the maximum amount that a household would ties of different eligibility criteria (such as geo- be prepared to spend to secure access to a given graphic location and wealth indicators) and quantity of the service. Thus, in economic terms, determining the proportion of household it represents the limit of affordability of the ser- income that is spent on water and sanitation. vice. A reasonable rule would be to set subsi- dies to cover the shortfall between a vulnerable The ideal data source would combine all three household's willingness to pay for a basic level types of information for the same households. of consumption and the associated bill. However, this ideal data set is rarely available, so * Water consumption data. These describe the the required information often has to be collated pattern of demand for different types of house- from different sources. holds. They can be used to establish the basic consumption level that will be subsidized. Water What sources? consumption data are also essential for estab- lishing the distributional incidence of cross-sub- One source of information is a water company's sidy schemes such as rising block tariffs. client database. This will typically contain * Socioeconomic data. These should ideally monthly information on the number of cus- include household income or expenditure lev- tomers, the value of their bills, the level of con- els, poverty lines, and receipt of other welfare sumption for customers with water meters, the benefits, as well as general indicators of basic location of customers, and the tariff structure needs, such as the quality of housing and its applicable to each customer. Client databases are associated facilities, and family wealth, such as generally the best source of information on The World Bank Group * Private Sector and Infrastructure Network F'LFflOPY *, r tp; % Infrastructure Reform, Better Subsidies, and the Information Deficit consumption levels, wlhich can be used to cre- Of the three types of data required for subsidy ate a frequency distribution of consumption pat- design, willingness-to-pay information is proba- terns for metered customers, and to determine bly the most difficult to obtain. One possibility is seasonal trends in consumption. to use a surrogate market approach. In this case, willingness to pay is inferred from the behavior However, client databases have several draw- of houselholds in a market for a goocd that is a backs as a soutrce of information. They do not complement or substitute for public water sup- provide information on the consumption patterns ply. The obvious example is demand for non- of informal connections. Neither are client data- piped sources of water suchi as bottled water or bases very informative as to the consumption truck-based vendors. This method was used in level of clients without meters. However, their the Dominican Republic to provide a crude first- most important drawback is the lack of data on cut estimate of the demand function for piped the socioeconomic characteristics of each house- water services. Households without a connection hold. This limits their usefulness in studying the to the public network were typically buying targeting properties of different subsidy designs. water for US$6.33 per cubic meter from water tankers, and were consuming approximately 7 Another very useftil source of information is the cubic meters per householdi per month, an data from household expenditure or socioeco- expenditure of US$44.3 per mionth. Households nomic surveys. These surveys are routinely under- with a connection to the public network paid taken by the statistical offices of most countries in around US$0.13 per cubic meter for, on average, order to update consumer price indices and to 37 cubic meters per household per month, an measure poverty levels or obtain other socioeco- expenditure of US$4.93 per month. These tAwo nomic information. Household surveys are prob- situLations essentially describe two points on the ably the best source of information on poverty demand curve for water, which can be used to levels, demographic characteristics, and general make inferences about the overall demand curve socioeconomic information. Many coujntries use and the corresponding willingness to pay. some form of the Living Standards Measurement Study (LSiMS) survey, which was developed by the Another alternative is to undertake a specific WVorlci Bank in 1980 and has subsequently been willingness-to-pay survey. These sun'Ceys use a adopted (with minor variations) in more than 20 'contingent valuation" approach to deduce developing countries, including Ilanama. But users' back-of-the-envelope preferences for the household surveys also have drawbacks: they are service. Many such surveys have been done in often outclated and the geographic divisions used Latin America. may not match the population for whom the su>b sidy is being designed. If a willingness-to-pay survey is not undertaken, the final option is to use parameters taken from A third source is national censuses. These usually other regional or international studies of willing- include a wide range of demographic, durable ness to pay, or to use a benchmark parameter. It goocds ownership and other socioeconomic infor- is striking that the results from a range of studies mation. and have the advantage over household in different countries are comparatively consis- surveys that they are representative at all levels of tent, with the willingness of poorer households disaggregation, including very fine geographic to pay for piped water supplies invariably falling divisions. However, census data have drawbacks between 3 and 5 percent of household income. too. They can be quite out of date (censuses are However, there is no guarantee that such rules of done every ten years or so). They usually measure thumb will be applicable to the area under study. income and expenditure poorly if at all. Finally, they provide no better information on water con- The characteristics of these data sources are sumption and bills than household sur-,eys. summarized in table 1. Most often, the ideal clata TABLE 1 DATA SOURCES AND THEIR RELATIVE ADVANTAGES i ii i 11 1i 1 1 r11r11 ii 1i |lklwtr_ Water company Consumption Detailed and usually reliable No information on informal connections client database No data on socioeconomic characteristics Absence o1 information on unmetered customers Household expenditure Socioeconomic variables Very detailed data to make Information may be outdated surveys Access to water supply judgment of household's May not be representative at the Expenditure on water economic well-being required geographical segmentation No data on physical water consumption Expenditure data may be unreliable Census Socioeconomic variables Representative of all zones May be ouldated Access to water supply and geographical levels No income or expenditure data No data on physical water consumption Willingness-to-pay surveys Willingness to pay Detailed control of Expensive to conduct Consumption data information gathered Methodological problems Some socioeconomic data Very flexible International parameters Willingness to pay Quick and inexpensive Parameters may not be correct or relevant Surrogate market studies Willingness to pay Quick and inexpensive May not be very reliable Required primary data may not be available base will not be available, so the required infor- Inferring unmeasured consumption. The con- mation wvill have to be collated from different sumption of househo]ds without meters cannot sources. The strategy adopted in Panama, for be directly observed, yet it is often relevant for example, wras to use multiple sources of infor- subsidy design, especially when metering cover- mation and integrate them to the greatest degree age is expected to increase in the near future. It possible. The 1997 LSMS survey was used as the cannot be assumed that consumption in house- basis of the analysis because it was the only holds without meters is the same as in house- source that combined data on socioeconomic holds with meters. One approach is to estimate conditions with data on water use. The client unmeasured consumption by examining the records of the water utility, Instituto de behavior of a client in the months after a meter Acueductos y Alcantarillados Nacional (IDAAN). is installed. It is reasonable to conjecture that the which contained information on the location of household will change its consumption gradually clients. the corresponding tariff structure, and after the installation of the meter. The more ran- the consumption each month (whether mica- domly new meters are installed, the more repre- sured, estimated or imputed) was used to check sentative of the market as a Nwhole such figures the accuracy of the consumption levels that were will be. Data from ]Panarna showed that the inferred from the expenditure data in the LSMS average monthly consumption for households survey. Finally, since neither of these sources facing the residential tariff declined by about 25 had information on willingness to pay, a special percent over the first four months of metering. contingent valuation survey was undertaken. Inferring consumption from expenditure data. Dealing with incomplete data sets The LSMS survey used in the Panama case study provided direct information on how much house- Data sets are often incomplete. Creative data holds spent on water, not how much water they analysis can overcome this problem. used. Moreover, the information on expenditure Infrastructure Reform, Better Subsidies, and the Information Deficit was deficient in a number of respects. It did not sumption levels from that part of the sample say whether customers had a meter or not, nor which reported water expenditure, a consump- detail the tariff structure under which they were tion function was estimated on the basis of the paying. As a result, the estimate of household numerous socioeconomic variables contained in water consumption inferred from the LSMS survey the survey. The estimated equation was then diverged substantially from the consumption used to predict the expenditure level of house recorded by the water utility. Notwithstanding holds with missing values for this variable. these deficiencies, an attempt was made to retrieve consumption information from the Conclusion expenditure data contained in the LSMS survey by making some reasonable assumptions about the This Note has summarized some methods of specific tariff that each household paid from acquiring, or producing, the information needed among the many different tariff structures charged to design a water subsidy scheme. Without such by the company. On this basis, water consump- information, and a simulation of the probable tion could be estimated for each household (albeit effect of different subsidies, there is little guaran- imperfectly) by applying the corresponding tariff tee that a subsidy system, however well inten- structure to the water expenditure figure. tioned, will have the desired effect. Once a suitable data set has been constructed, a simula- Inferring consumption when data were missing. tion model can be created using simple spread- Viewpoint is an open Socioeconomic data and the water consumption sheet software, providing figures on which a forum intended to data often come from different sources and a sta- design can be based. Armed with this informa- encourage tistically valid method for combining them is tion, the policymaker should be in a position to debateon ideasn required. A useful method has been developed design a subsidy program that reaches the innovations, and best by Rajah and Smith (1993) in their work on the intended beneficiaries, provides them with the practices for expanding distributional impacts of introdu-cing water level of financial support which is strictly neces- views published are metering in the United Kingdom. These authors sary, meets the overall budgetary restrictions, and those ofthe authors and had access to two data sets: does not have excessive administrative costs. should not be attributed * A large-scale nationallv representative house- to the WorldaBunkaor hold survey with a very rich set of informa- References organizations. Nor do tion on income and other socioeconomic Foster, \Vivien, Andr6s G6rnez-Loho, aind jlonailiast H-alpern. any of the conclusions indicators, but scant information about water Fortlcoming. I)esigning Direct Subsidies for Water and Sanitation represent officiul policy 5 rmPnm oiyRsac okn ae of the World Bank or of consumption. services Lessons from Panama." Policy Research Working Paper. its Executive Directors * A small-scale detailed local study of water con- World 13ank. 'ashington. D.C. G6nnez Loho. Andres, V ivien Poster, and Jonathan Halpern. or the countries they sumption withi a very limited range of socio- Forthcoming InformTation and Modeling IssLes in Designing represent. economic variables. Water and Sanitation Subsidv Schemnes.' Police Research Working To order additional Paper. World Bank. Washington, D C. copies please call Rajah. N., and S. Snith. 1993 Distrihotional Aspects Of addoicinal copies please call With the aim of incorporating a consumption Water Charges'" Fiscal Stities 14(3):86-108. 202 458 1111 or contact.. .* Suzanne Smith, editor, variable into the household survey, Rajah and Room Fl I K 208, Smith estimate a water consumption function The World Bank, from the second data set based on socioeco- Vivien Foster (vfoster@worldbank.org) and 1818 H Street, NW,bohdt Washington, D.C. 20433, nomic variables that are common to both data Jonathan Halpern (jhalpern@worldbank.org), or Internet address sets. On the basis of this equation, it was then Latin America and the Caribbean Region, and ssmith7@worldbank.org. possible to predict consumption levels in the Andrns G6mez-Lobo (agomezlo@ The series is also available on line large-scale household survey. This method was deconfiacea.uchile.cl), Departmnent of (www.worldbank.org/ used in Panama. About 17 percent of households Economics, U7niversity of Cbile html/fpd/notes/). in the LSMS survey stated that they did not pay ® Printed on recycled their water bills and, hence, their reported expen- paper. diture was equal to zero. Taking the inferred con studl ir Kr ishllamulr ti i 1I:NCF MC6- L73