. 2 1468 Viewpoint , *.. . X, g Better Household Surveys for Better Design of Infrastructure Subsidies Andres G6mez- Reform of the water, electricity, and telecommunications sectors is gathering momentum in Lobo, Vivien nearly all developing countries. Reform should include an assessment of whether subsidies are Foster, atnd Jonathan necessary and if so, how to design subsidies that reach their intended beneficiaries accurately Halpern and do not distort the market. A major challenge for governments is to build the capability to do this fast enough for subsidy redesign to be incorporated in sector reform. Clearly, it would save time to use existing sources of information. Potentially, one of the most useful sources is the Living Standards Measurement Study (LSMS) survey. However, the LSMS questionnaires do not generate all the information needed for subsidy design. Fortunately, with a few simple and inexpensive changes, these surveys could be made much more useful for the design of subsidies and for devising policies that would give the poor better access to infrastructure services. More than 20 developing countries use some form types of subsidies on the target population. An of the LSMS survey developed by the World Bank LSMS survey is an invaluable source of the infor- in 1980. These household surveys are a useful tool mation needed. for measuring living standards and poverty, and provide crucial information for the design of gov- However, the use of figures from an LSMS sur- ernment policies and evaluating social programs. vey for a study of the water sector in Panama As well as socioeconomic inquiries, most surveys revealed that, as in most countries, the data on include questions on utility services such as water, water expenditure produced by such surveys electricity and telecommunications. This makes have significant limitations. This Note describes them particularly valuable because they are the difficulties encountered with the Panama potentially the only source combining informa- LSMS survey of 1997 and suggests changes that tion on households' usage of utility services would make the survey much more useful for the together with socioeconomic characteristics. formulation of water sector policy and the design of policies to promote access of the poor to infra- Interest is growing in the rational design of structure services in general. subsidies for utility services as an integral part of tariff reforms. Explicit subsidies, with well- The challenge defined objectives, budgets and instruments, are intended to replace the implicit subsidy arrange- When designing a water tariff and subsidy struc- ments that are ubiquitous in developing coun- ture it is important to simulate the social and dis- tries. However, the process of designing a tributive impacts of different designs. For subsicdy schleme should be supported by empir- example, rising block tariffs might be proposed ical analysis to simulate the impact of alternative as a way to make water affordable for the poor UN The World Bank Group * Private Sector and Infrastructure Network Better Household Surveys for Better Design of Infrastructure Subsidies without impairing the economic efficiency and the water sector. More typical are general aggre- financial sustainabilitv of the service provider. gations, such as urban or riral. whici clo not The unit price of water would be lower for the allow the analysis of the impact of policies first units of consumption. up to the level con- I)ased on finer geographical (livisions. sidered sufficient for the basic needs of poor * ½nuey design. The surveys ask householdls householcls Water in the subsequent blocks 1-town much they spend on water; not how much could then be priced at the true economic cost water they consume. Consumption has to be or higher. The efficiency and effectiveness of inferred by applying the prevailing tariff stric- these types of social tariffs will depend on the ture to the expenditure figures. This has proven correlation between household water consump- very unreliable as there are often multiple tar- tion and household poverty levels. To evaluate ilf strictures in force and billing is often based this correlation it is necessary to have a database on unmetered consumiption. Consequently. that records for each household both socioeco- there are likely to be some inaccuracies in the nomic variables-required to determine the inci- estimates of household consumption infelTed dence and degree of poverty-and physical fiom LSMS expenditure data. water consumption. a IJterview protocols. Although interviewers were instructed to ask householdls for their last Suchi a clatahase would he useful in setting the cor- water bill, the analysis of the survey results rect prices and the size of the blocks. It would also presented below seems to suggest either that help with other policy questions. For example, the this was not done, or that households were targeting properties of different subsidy design unable to produce the bill. It appears that most options and eligibility criteria could be studied responses were based on recall, not read from empirically to choose the most appropriate. the hill. Besides the measuring errors intro- duced by poor recall, figures given from mem- LSMS surveys record a large number of socio- owy tend to overstate the amount clharged for- economic variables that can be used to ascer- Water, because the bills include items not tain the poverty level of the sampled related to water services (for example, refuse households. For water use all surveys Ask how collection). much the household spent on water services during the last month or the last payment These shortcomings imply that information on the period. They also ask the source of water sup- water sector provided by an LSMS survey may not ply. the average number of hours a dlay in which be capable of providing adequate answers to the a dwelling receives water, and whetlher there is key questions raised by subsidy design. As there a sewerage connection. They sometimes ask are fexv other data sources, the choice is either not how far the dwelling is from the water supply to use these data for water policy (lesign, or to source, where the tap is withiin the perimeter of risk poody designed policies. Either wax' the the dwelling's premises, and other characteris- result is less than optimal policies to aid the access tics of the water and sexverage services. of poorer houselholdls to basic water services. Experience with the Panama LSMS survey in the The problem of unsuitable sample dlesign may design stage of a new water subsidy scheme be expensive to overcome. Finer geographical revealed the shortcomings of such surv,eys for divisions require more intervSiews if they are to the design of sectoral policy. These relate to: be statistically reliable. They might also be unac- Sample design. Sampling frames are not ceptable if thev made the samples unsuitable for cdesigned to ansvwer usefLl policy questions at the other purposes of the LSMS survey. the sectoral level. Rarely will the sample be rep- However, the survey design and interview pro- resentative of specific urban localities, a level of tocols would be less difficult to adapt. This Note dissaggregation neecled for policy analysis in will recommend inexpensive changes to them FIGURE 1 MONTHLY WATER BILL DISTRIBUTIONS FROM DIFFERENT INFORMATION SOURCES, STANDARD RESIDENTIAL CUSTOMERS, PANAMA, 1997 Frequency (percent) 25 LSMS 20 _~ ~ ~ 1 N 15 10 5 0 0 0.8 1.6 3.2 4.0 4.8 5.6 7.2 8.0 8.8 9.6 11.2 12.0 12.8 13.6 14.4 16.0 16.8 17.6 18.4 20.0 20.8 21.6 22.4 24.0 24+ Monthly water bill (U.S. dollars) Source: Instituto de Acueductos y Aicantarillados Nacional and Encuesta de Niveles de Vida, Panama, 1997. that would immediately make the LSMS survey the frequency distribution of the actual water more useftul for water policy formulation. Before expenditure as shown in IDAAN's client (lata set making these recommendations, the Note con- and that reported in the LSMS survey (figure 1). siders the evidence from the Panama LSMS sur- Part of the explanation for the poor quality of vey that illustrates the shortcomings of such the survey expenditure data may be that house- surveys for dcsign of water sector policy. holds based their answer on rccall, not the actual bill, Particularly revealing was the spike in the A case of divergent numbers LSMS distribution at US$12, exactly the amount of the minimum residential bill including refuse Ideally, the designers of tariff and subsidy collection charges. schemes for infrastructure services should have access to a household-level data set that contains Calculating the volume of water consumed intro- both socioeconomic dlata and data on physical duces further error. The bills must be transformed consumption for each observation. The LSMS into volumes using the tariff structure. However, survey comes close to this ideal data source, multiple tariff structures are applied to residential because it records the amount each household customers and the survey contained no infor- spent oIn water. The volunme consumed can be imation on which tariff applied to wllich house- calculated using the prevailing tariff structure. hold. Nor did it indicate whether a given household had a water meter. Therefore it is Unfortunately, a detailed analysis of the data in impossible to know whether the expenditure Panama's 1997 LSMS survey reveals some incon- transformation gives actual or imputed water con- sistencies. There was wide divergence between sumption. This a a flaw common to LSMS surveys Better Household Surveys for Better Design of Infrastructure Subsidies in most countries and it limits their usefulness as presented above suggests that interviewers were a tool for policy design. not following the established procedure. Better training and supervision may be required. Better surveys and interviews When the bill is not available, the interviewer Several changes to the survey and interview pro- should: tocols could be made to improve the precision * Ask, based on the household's recollection, of LSMS water expenditure data and facilitate the the same information as above. estimation of the volume consumed by each * Remind the interviewee to exclude from the household. Although the recommendations have estimate the costs of irrelevant services such as been derived for the water sector, many of them refuse collection or other services charged on will be relevant to other utilities such as elec- the utility bill. tricity and telecommunications. Overall these * Record that the expenditure figure is based on recommendations illustrate the importance of the household's recollection, not the bill. maintaining a dialogue between LSMS practi- * Be alert to question the household if a stated tioners and clownstream infrastructure analysts expenditure figure (based on recollection) is to ensure that the supply of information meets implausible (for example, if it is below the the policy needs of the sector. minimum charge of the tariff structure). * Record, particularly for water, whether the In particular, if the interviewer is able to see the household has a meter. Viewpoint is an open bill, it would be desirable to record the follow- forum intended to ing information in addition to the amount spent These recommendations are relatively simple encourage on the service: and would not increase the burden or costs of dissemination of and debate on ideas, * Whether the bill was based on measured, esti- administering a survey. They could bring sub- innovations, and best mated (based on past meter readings), or im- stantial benefits by providing the needed data to practices for expanding puted consumption. A more ambitious make informed policy choices. The marginal the private sector The views published in this strategy would be to record the customer num- costs would be small compared with the poten- series are thosse of the her of Lhe client. This variable could then be tial benefits to tlie poor frorn better dlesigned and authors and shouid not used to cross the household survey data with targeted subsidy and tariff schemes for infra- be attribstad to the World Bank or any of its the wx ater company's client data base. structure services. affiliated crganizations. However, unless the statistical office did the Nor do any of the con- crossing of data sets itself, before the informa- References clusions represent official policy of the tion is made public, this altemative would World Bank or of ts probably violate the confidentiality rules that eain)n o Anguts. 19Ap. The Artaoyse of Household SunBati, A .the,-oeconomnetr-c Approach rv Dez.eloapment PoliCy' Baltimore, Executive Directors or apply to such surveys. Std.: Johns Hopkins tniversity Press. the countries they represent. * The tariff that applied to the household (if this Foster, Vivien, Andres Gcmez-1Lobo, and Jonathan Halpern. information is clearly specified in the bill). orthco ing rDesigning DirPm P licy Research Working PSpeo To order adcitional * The quantity of service consumed (if this infor- World fBank, Washington, D.C. 2024581111 or contact mation is recorded in the bill) as well as the o6niez-Lofo, Andres, Vivien Foster, and Jonathan Halpern. Suzanne Smith, editor, amount of iiiorie- chiarged for it. Forthcoming. "Intormatnon and Modeling Issues in Designing Rosin F11K-208, The Wloe ater and Sanitation Subhsidy Schemnes.' Policy Research Working World Bank, 1818 H * The identity of the setv ice supplier. This may be Paper. \orld Bank, WSashington, D.C. Street, NW, Washington, important when there are alternative suppliers. DC. 20433, or Internet Some surv-eys already include such a question. I itien Foster (4'bsterfworldbank.org), and address ssmith7@ worldbank.org. The Jonathan Halpern (jhalpernogworldbank. org), series is also available The interviewer training should cmphasize the Latin America and the Caribbean Region, and on-line (www.worldbank. org/html/fpd/notes/). importance of asking to see the bill. Although Androes Gomez-Lobo (agomezlo@ @ Printed on recycled interniew guidelines-at least in Panama-give deconfacea.uchile.c,), Department of paper. clear instrnctions in this respect, the evidence Economics, University p/Chile