46881 WATER 5 T A R I . . S & S U B S I D I E S I N S O U T H A S I A APERP Can subsidies be better targeted? RE.ORM O. THE WATER AND SANITATION SECTOR is occurring in many countries, and offers the potential to improve services to all. Of particular concern, however, is the situation of the poor, and reform must be designed so that they receive increased access to affordable services. A key issue in this regard is water pricing, which is one of the main variables affecting the distribution of benefits between Project different stakeholders. However, experience .IRE-D shows that water pricing, and the subsidies which are often delivered through water tariffs, can be a source of major inefficiencies in the sector. Indo-USAID by While affordability has been one of the prime concerns of those setting tariffs and designing Picture A subsidies, there may be significant flaws in many common pricing strategies and subsidy delivery n earlier paper in this series examined the extent to which mechanisms. Rather than providing affordable current water subsidies in two south Asian cities water to the poor, these may in fact be leading Bangalore and Kathmandu succeeded in reaching poor to financial unsustainability of utilities, lack of households. The paper concluded that subsidies to private taps access to services, and inequity. The reform were very poorly targeted; since barely 30% of the beneficiaries process provides the opportunity to rationalize are poor, and only 25% of the subsidy resources are captured by and reconsider the design of tariff and subsidy poor households. A key underlying reason for this is that more structures, and seek new ones which may provide than half of poor households in these cities do not have private better results. water connections at all. Subsidies to public taps perform considerably better in equity terms, in that around 70% of the beneficiaries are poor and they in turn capture around 70% of the This series of papers is designed to examine these issues in South Asia. It is designed to present the basics of tariff subsidy resources. However, even in the case of public taps, a and subsidy issues, to disseminate recent research findings, significant proportion of the poor in the two cities fail to benefit and to stimulate debate on the subject. from the subsidies. The preparation of these papers was funded by the Public- The poor targeting of subsidies for water consumption from private Private Infrastructure Advisory .acility (PPIA.). Additional financing was provided by the World Bank, the World taps raises the question of whether anything can be done to improve Bank Institute and the Water and Sanitation Program. the way subsidies are designed. Building on the same information base for the cities of Bangalore and Kathmandu, it is possible to Water and Sanitation Program CAN SUBSIDIES BE BETTER TARGETED? simulate alternative subsidy systems and examine whether subsidies essentially prove easier to target because there they perform any better than the status quo. In particular, is already a much higher concentration of poor people it is interesting to investigate whether use of geographic among the unconnected population than among those or individual targeting mechanisms that select subsidy who already enjoy access to piped water. beneficiaries on the basis of poverty criteria perform any better in delivering subsidies than the traditional Increasing How were the simulations done? Block Tariff (IBT) structure currently used in both cities. This discussion is based on the results of two city-level Another important issue is whether it is easier to reach household surveys conducted in Kathmandu (Nepal) in the poor by subsidizing private water connections rather April 2001, and Bangalore (India) in August 2001. Both than water consumption. surveys collected a wide range of information including Targeting of subsidies can substantially improve the household water expenditure, type of water supply, financial position of utilities, with revenues rising between physical characteristics of the dwelling, and socioeconomic characteristics of the household (including overall income or expenditure). A more Targeting of subsidies can substantially improve complete description of the data, and the water supply the financial position of utilities, with revenues situation in the two cities can be found in an earlier rising between three and fivefold in the two paper in this series1. Using the survey data, it was possible to estimate cities in the cases considered. the amount of water consumed by each household in the two cities. Different methods were used depending on the type of customer. .or those relying on public three and fivefold in the two cities in the cases considered. taps, consumption was estimated from reports about the Moreover, the various forms of targeting considered prove number of containers that they filled on average each to be extremely successful in reducing the leakage of day, and the volume of those containers. .or those with subsidies towards households that do not really need them. metered private taps, consumption was inferred from Indeed, targeting on the basis of geographical location or reported water expenditure by applying the current tariff housing characteristics can double the share of subsidy structure. .or those with unmetered private taps, expenditure that reaches the poor. consumption was imputed based on other household However, unfortunately, targeting criteria also have characteristics (family size, type of dwelling, etc), using the effect of mistakenly excluding households that are a statistical water consumption model developed with genuinely poor, and this may be a major drawback if data from metered households. the ultimate policy objective is to ensure that all poor It is important to note that, due to the intermittent households can afford to meet their subsistence needs nature of water supply in South Asia, meters can often for water. Targeted connection subsidies on the other become damaged, either under-recording true hand do an equally good job of avoiding leakage to consumption or breaking down altogether. This clearly 2 undeserving households, while at the same time reaching makes it difficult to achieve reliable estimates of water a much higher proportion of the poor. Connection consumption from water expenditure, particularly since 1Paper #4 Do current subsidies reach the poor? W A T E R T A R I . . S A N D S U B S I D I E S I N S O U T H A S I A there are no reliable estimates of the extent to which overall volume of subsidies remains within a fixed meters malfunction or under-record. The likely effect of subsidy budget. .rom a methodological perspective, it this phenomenon is to dampen the variation in water is important to ensure that the total amount of subsidy consumption estimates between households. However, given is held constant across each of the different tariff as long as there is no systematic relationship between structures that are simulated for each city. Since it is meter malfunctioning and income level, the much easier to reach the poor with a larger subsidy phenomenon should not necessarily affect measurements budget, it would be unfair to make comparisons across of the distribution of water consumption across income different targeting criteria unless the subsidy budget were groups, which is the primary focus of this analysis. the same in each case. These consumption estimates provide the basis for a Customers that do not meet the poverty criteria are simulation process that explores the distributional charged full cost recovery tariffs. This is in contrast to impacts of targeting subsidies in a variety of different the status quo, where the use of Increasing Block Tariffs ways. The key steps in the process are summarized (IBTs) with large initial blocks and only small increases graphically in .igure 1. in the upper blocks (never reaching cost recovery levels) .igure 1: Schematic representation of simulation process means that subsidies are effectively being allocated indiscriminately to all customers. .or the purposes of these simulations, cost recovery tariffs are taken from NON-POOR Apply cost recent engineering studies that quantify operating and recovery tariffs maintenance costs as well as debt service charges. .IX APPLY COMPUTE Information on capital depreciation was not available, SUBSIDY POVERTY EQUITY BUDGET CRITERIA OUTCOMES and hence the values of NPR 13/m3 (US$ 0.17/m3) in POOR Kathmandu, and Rs. 17/m3 (US$ 0.34/m3) in Bangalore Determine subsidized traiff still fall significantly short of full cost recovery. iteratively Nonetheless, they represent increases of 300 to 400% over the tariff currently charged for the first consumption All the simulations involve dramatically reducing the block, and increases of 33 to 50% over the marginal total volume of subsidies relative to the status quo. tariff currently faced by most customers. An earlier paper Specifically, they are reduced to 36% of current levels in this series looked at the extent to which cost recovery in Kathmandu, and 18% in Bangalore. As a result, utility tariffs were affordable for Indian households2. The results revenues increase by threefold in Kathmandu and showed that at the proposed tariff level, about 95% of fivefold in Bangalore. This reduction in subsidies is Indian households could afford a basic subsistence achieved by applying tight eligibility criteria for consumption of five cubic meters per month, but only subsidies. Customers who meet these criteria face prices 60% could afford to buy 10 cubic meters. This is that are only a little higher than those currently charged consistent with data from the Kathmandu study, which for the first consumption block in each city. suggests that the poor could afford to pay more than 7 The exact level of subsidized tariffs for these times the price of the current lifeline block of water3. 3 customers is determined iteratively to ensure that the Indeed, the willingness to pay by the poor, reported in 2Paper #2 A Scorecard for India 3Based on a median income for the poor of NPR 6000, a lifeline block priced at NPR 40 , and an affordability threshold of 5% of income. S.K. Pattanayak, J.C. Yang, D. Whittington, and B. Kumar K.C., 2002, Willingness to Pay for Improved Water Supply in Kathmandu Valley, Nepal, RTI, North Carolina W A T E R T A R I . . S A N D S U B S I D I E S I N S O U T H A S I A CAN SUBSIDIES BE BETTER TARGETED? the study, was as high as NPR 620 per month (about US$ 8.00) for a reliable piped water supply4. Since the simulations involve changing prices for just about all customers, it was essential to make assumptions as to how peoples demand for water would change in the light of rising prices. Based on the Project economic literature, a price elasticity5 of 0.5 was used, meaning that consumers reduce water consumption by .IRE-D 5% for every 10% increase in prices. In some cases, price increases were large enough that demand would Indo-USAID theoretically drop to zero if this assumption were by applied. To avoid this problem, a maximum demand Picture reduction of 33% was arbitrarily imposed for large price changes. current IBT tariffs as a way of targeting subsidies to the .inally, for the purpose of evaluating whether poor. Evidently, the ideal would be to allocate subsidies subsidies reach the poor, the people in the bottom 40% to households on the basis of their actual income, also of the income distribution for each city are considered known as means-testing. However, in practice it is very to be living in poverty. The simulations deliberately difficult to estimate a households income with any avoided the use of official poverty lines, preferring to degree of accuracy, particularly in developing countries apply a relative rather than absolute concept of poverty. where there is a high degree of informality. .or this This has the advantage of ensuring that the definition of reason, targeted subsidy schemes typically rely on more poverty is consistent across the two city cases. Using readily observable indicators, or proxies for poverty. this definition of poverty, the equity outcome of each In this case, three types of poverty proxies are subsidy simulation is evaluated on the basis of a number considered: the volume of water consumption, the of standard indicators. These include the leakage rate characteristics of the neighborhood, and the (or proportion of total subsidy resources captured by characteristics of the dwelling. the non-poor), the errors of inclusion (or proportion of The first approach is to modify the IBT structure so total subsidy beneficiaries who are not poor), and the that the size of the first block is reduced to a level more errors of exclusion (or proportion of the poor who are consistent with the idea that this is an amount needed not subsidy beneficiaries). A full set of results are for subsistence consumption. The initial subsidized block provided in the Annex, and the main highlights are is thus reduced to six cubic meters per month for both summarized as follows. cities, with all consumption above this threshold charged at cost recovery levels. Two problems still remain: given Does it help to target consumption subsidies? that the difference in average water consumption Akey question is whether it is possible to improve upon between rich and poor was only around 20% for both 4 4In India, willingness to pay studies have shown that households in several cities are willing to pay more than the current tariff for reliable water services: for instance 1995 studies in Dehradun and Baroda showed a willingness to pay of 2 times and 3.4 times the current tariff, respectively. Water and Sanitation Program, Willing to Pay but Unwilling to Charge, .ieldnote, June 1999 5 Price elasticity is defined as the responsiveness of quantity demanded to a change in price, with all other factors held constant, defined as proportionate change in quantity demanded divided by proportionate change in price W A T E R T A R I . . S A N D S U B S I D I E S I N S O U T H A S I A cities, there is no reason to assume that the consumption non-poor households than poor households. In fact, only of the poor falls entirely into the subsistence consumption a third of poor households meet these criteria. block, particularly if they share a connection between The results of applying these different targeting several families. schemes can be compared in a number of different ways The second approach is to introduce geographical (.igure 2). An important criterion is the share of total targeting, by limiting subsidies to residents of subsidy resources that leak away to non-poor households neighborhoods that satisfied certain poverty criteria. In in each case; shown as the light yellow columns in Bangalore, this meant limiting subsidies to residents of .igure 2. These are closely related to the errors of officially designated slums. Since such officially inclusion or the percentage of total subsidy beneficiaries designated slums do not exist in Kathmandu, the subsidy who are not poor; shown as the darker yellow line in was applied to neighborhoods where at least 60% of .igure 2, which follows the light yellow columns quite residents lived under the poverty line. Initial explorations closely. .inally, it is also important to look at errors of showed that such geographical criteria are able to exclusion, which give the percentage of the poor who correctly classify around 62% of the population as either do not benefit from the subsidy; these are represented poor or non-poor. However, the proportion of the poor who live in geographically defined slum areas is less than 20%. The modified IBT barely performs any The third approach is to introduce individual targeting better than the original one, indicating that it by identifying a number of variables that are good is not possible to improve targeting simply predictors of poverty at the household level but at the by playing around with the design of the same time are relatively easy to observe and difficult to misrepresent. In both cases, a formula based on housing IBT structure. characteristics was developed. .or Bangalore, a statistical model was developed to by maroon lines in .igure 2. Since all of these indicators predict the probability that a particular household was capture failings in the subsidy system, the lower their poor based on the nature of floor and roof materials, value the better the performance of the subsidy. the size of the dwelling, the type of dwelling, the The conclusions that emerge from this analysis are presence of a separate kitchen, and the presence of a clear, and also very consistent across the two cities. flush toilet. Prior tests showed that this formula correctly On the one hand, the modified IBT tariff barely identified 64% of households as either poor or not poor; performs any better than the original IBT, indicating that however only a quarter of the poor live in deficient it is not possible to improve targeting simply by playing housing of this type. around with the design of the IBT structure. Nonetheless, .or Kathmandu, subsidies were applied to any it is important to note that the modified IBT generates household that met at least two of the following three three to five times more revenue for the utility than the characteristics: lacking a telephone; living in a dwelling status quo, without worsening the targeting performance that does not incorporate cement in either floor, walls, or of the tariffs. roof; using traditional fuels rather than gas or electricity On the other hand, use of explicit targeting 5 for cooking. This formula correctly identified 82% of the schemes be they geographic or individual leads households it was applied to as either poor or not poor. to major improvements in performance. The However, once again, it does much better at identifying proportion of subsidies leaking away to non-poor W A T E R T A R I . . S A N D S U B S I D I E S I N S O U T H A S I A CAN SUBSIDIES BE BETTER TARGETED? households, as well as the errors of inclusion, fall to Unfortunately, there is a downside. The price of approximately half of their original levels (see the lowering the errors of inclusion is to substantially raise yellow lines and columns in .igure 2). The difference the errors of exclusion, to a point where as many as 80 in results between geographic and individual to 90% of the poor are excluded from the subsidy (see targeting is not that great; although geographic the maroon lines in .igure 2). There is thus a trade-off targeting performs slightly better in Kathmandu and between errors of inclusion and errors of exclusion individual targeting slightly better in Bangalore. The this is visually evident from the X shape that appears fact that the accuracy of geographic targeting is just in the graph. However, since only about 50% of the as good as individual targeting is an important finding poor are connected, the errors of exclusion cannot, by given that the administrative costs of geographic definition, be any lower than 50%. targeting are substantially lower all that would be The intuition behind this trade-off is straightforward. required is for the utility to include in its customer In order to avoid the major problems associated with database an indicator of whether a customer lived in direct means-testing of beneficiaries, poverty proxies a designated slum. have been used that are based either on neighborhood or housing quality. However, while there are very few rich people that live in slums or very basic housing, The fact that the accuracy of geographical there are plenty of poor people who live outside slums or have managed to secure slightly more decent housing, targeting is just as good as individual targeting perhaps as poor tenants in better-off neighborhoods. is an important finding given that the Hence, by applying targeting criteria of this kind, both administrative costs are substantially lower. rich and poor are being excluded from the subsidy, leading to the results that have been seen for errors of inclusion and exclusion. .igure 2: Comparison of alternative forms of targeting 6 (a) Kathmandu (b) Bangalore W A T E R T A R I . . S A N D S U B S I D I E S I N S O U T H A S I A Given these trade-offs, which is the best approach? to the Indian National Sample Survey (50th Round) The answer, of course, depends on the relative 1993/4, 68% of non-poor households in India have importance one attaches to errors of inclusion and private taps, whereas 55% of unconnected exclusion. If the policy objective is to deliver resources households are poor. Moreover, an earlier paper in as cost-effectively as possible to the poor, then the most this series6 found that connection subsidies were, in important consideration is to minimize the errors of practice, widely used in India, given that connections inclusion, and targeting subsidies would therefore make are typically charged at Rs. 1,000 (US$ 20), which sense. If, on the other hand, the policy objective is to is likely to be only about 10% of the full cost. ensure that all the poor have a basic safety net of In order to evaluate this hypothesis, an additional set affordable access to water, then errors of exclusion of simulations were performed. They involve eliminating become the overriding consideration, suggesting an all subsidies for water consumption, so that all consumers IBT approach. are charged full cost recovery tariffs on all units of It is important to put these results in perspective. consumption. The subsidy budget that is thereby saved is The problem of effectively targeting subsidies is neither then allocated in its entirety to subsidizing new unique to South Asia, nor unique to the water sector. A connections, at an estimated cost of US$ 75 per connection recent study of water subsidies in Chile and Colombia in Kathmandu and US$ 150 in Bangalore. This policy is found errors of inclusion in the 60 to 80% range, while pursued consistently over a number of years until universal a number of studies of electricity subsidies around the coverage is reached. Given the available subsidy world have found that barely 10 to 35% of these subsidy resources, it takes no more than a decade to reach universal resources reach poor households. Moreover, a study by coverage in both cities, although this assumes that parallel the Indian National Institute for Public .inance and investments in network expansion and densification could Policy (NIP.P) concluded that food subsidies allocated also be financed. via the Public Distribution System (PDS) suffered from .igure 3: Coverage of water services in India errors of inclusion between 34-52%, and errors of exclusion between 25-98%. Are connection subsidies any more equitable? Given the problems that have been seen with targeting of consumption subsidies, it is relevant to ask whether connection subsidies would perform any better in targeting terms. There are reasons to think that they might. In particular, it is noteworthy that almost all higher income households are already connected to the water network, whereas a high proportion of unconnected households tend to belong to lower 7 income groups (see .igure 3). .or example, according Source: National Sample Survey (50th Round) 1993/4 6Paper #2 A Scorecard for India W A T E R T A R I . . S A N D S U B S I D I E S I N S O U T H A S I A CAN SUBSIDIES BE BETTER TARGETED? Both targeted and untargeted connection subsidies rates, as well as errors of inclusion and exclusion. To are considered. In the absence of targeting, connection recap, the leakage rates (represented by light yellow subsidies are allocated at random to unconnected columns on the graph) indicate the percentage of subsidy households. Under the other scenario, the same resources that fail to reach the poor, and are related to individual targeting scheme developed for consumption the errors of inclusion (represented by the yellow lines subsidies (based on housing characteristics) is used, so on the graph), which capture the proportion of subsidy that connection subsidies are only granted to households beneficiaries who are not poor. .inally, it is also that comply with these criteria. important to evaluate the errors of exclusion (represented To facilitate comparisons, the results for targeted and by the maroon lines on the graph), which capture the untargeted connection subsidies are presented alongside proportion of the poor who fail to receive any subsidy. those for the status quo, as well as the best performing Once again, the conclusions are comparatively clear of the targeted consumption subsidies for each of the and consistent across the two cities. two cities (.igure 4). As before, the performance of the The most striking feature of the results is that the trade- different schemes can be evaluated in terms of leakage off between errors of inclusion and exclusion essentially disappears. That is to say that targeted connection subsidies Targeted connection subsidies have leakage have leakage rates and errors of inclusion that are barely a quarter of those associated with the status quo IBT, but rates and errors of exclusion that are barely a most importantly they also have lower errors of exclusion. quarter of those associated with the Targeted connection subsidies also perform significantly status quo IBT. better than targeted consumption subsidies, across all three performance indicators, even though the same targeting mechanism is being used in both cases. Even untargeted .igure 4: Comparison of consumption and connection subsidies 8 (a) Kathmandu (b) Bangalore W A T E R T A R I . . S A N D S U B S I D I E S I N S O U T H A S I A connection subsidies perform considerably better than the households without subsidies may not be able to afford to status quo. meet their subsistence needs, or if means testing is administratively complex or prone to corruption, then IBTs Conclusions may provide a second best solution in that at least all To summarize, the study finds that targeting of subsidies the connected poor receive some subsidized water. to poor consumers and introduction of cost recovery However, in order to improve the financial health of the tariffs has a major impact on utility revenues, raising utility by containing subsidy expenditure, IBTs need to be them by three to fivefold for the specific cases considered much better designed, with shorter initial blocks and here. At the same time, targeting on the basis of steeper gradients, ensuring that the upper blocks reach geographical location or housing characteristics can cost recovery levels. Nevertheless, IBTs do nothing to assist reduce the extent of subsidy leakage by about a half, the unconnected poor who must often make do with a thereby doubling the share of subsidy expenditure that much less adequate water service potentially at higher reaches the poor. However, unfortunately, targeting cost (and in fact a redesigned IBT with steeper gradients criteria also have the effect of mistakenly excluding puts poor households who share connections at households that are genuinely poor, so that errors of a disadvantage). exclusion rise from 50% to around 80%. The only way to ensure that these people benefit This creates a trade-off that can only be resolved with from government subsidies to water utilities is to increase reference to ultimate policy goals. Essentially, targeting coverage of private connections. The simulations show subsidies makes sense if the objective is to deliver resources that targeted connection subsidies perform much better to the poor as cost-effectively as possible. On the other than targeted consumption subsidies, even when the hand, if policy makers are particularly worried that poor same targeting mechanism is used. They are significantly Annex Summary of Results Subsidy scheme Quasi-Gini Poor get Errors of Errors percentage subsidy exclusion of inclusion Kath. Blore Kath. Blore Kath. Blore Kath. Blore Status quo Private taps 0.243 0.217 22% 27% 53% 51% 71% 71% Public taps -0.301 -0.588 61% 86% 72% 61% 38% 23% Improved targeting of consumption subsidies a. Modified IBT 0.116 0.190 29% 25% 52% 50% 70% 71% b. Geographic targeting -0.196 -0.516 53% 77% 82% 91% 37% 40% c. Individual targeting -0.183 -0.514 51% 80% 64% 76% 42% 34% 9 Connection subsidies a. Untargeted -0.277 -0.315 60% 63% 48% 50% 40% 37% b. Individual targeting -0.512 -0.479 84% 78% 48% 42% 16% 22% W A T E R T A R I . . S A N D S U B S I D I E S I N S O U T H A S I A CAN SUBSIDIES BE BETTER TARGETED? Prepared by: Vivien .oster, Subhrendu Pattanayak, and Linda Stalker Prokopy. The paper is based on the following consultants reports: Prokopy, L. 2002 Distributional Incidence of Current and Potential Water Tariffs and Subsidies in Bangalore, India, University of North Carolina; and Pattanayak, S.K. and Yang, J.C. 2002 Distributional Incidence of Water Tariffs and Subsidies in Kathmandu, Nepal, Research Triangle Institute, North Carolina more effective at avoiding leakage to undeserving households, while at the same time reaching a similar Series Editor: Clarissa Brocklehurst overall percentage of the poor. As much as 80% of This series of papers was prepared as part of the Capacity Building subsidy resources can be delivered to the poor by this and Learning initiative undertaken by the South Asia Energy and Infrastructure Unit of the World Bank in collaboration with the means and with a similar subsidy budget to that Government of India, the World Bank Institute and the Water used for consumption subsidies it may be possible and Sanitation Program, supported by a grant from the Public- to reach all of the unconnected poor within a decade. Private Infrastructure Advisory .acility (PPIA.), a multi-donor Connection subsidies essentially prove easier to target technical assistance facility aimed at helping developing countries improve the quality of their infrastructure through private sector because there is already a much higher concentration involvement (for more information see http://www.ppiaf.org). The of poor people among the unconnected population than initiative is designed to support the implementation of sector among those who already enjoy access to piped water. reforms and public-private partnerships in the provision and In practice, however, it may not always be possible to financing of water supply and sanitation services in India. choose between consumption and connection subsidies. April 2003 In the two South Asian cities studies, approximately half Task Manager: Midori Makino of the poor were connected and half unconnected, so Design and Production Coordinator: Vandana Mehra that any policy that aims to reach all of the poor will Thanks to: Lee Travers, Deepak Sanan and Kirsten Hommann need to consider both of these approaches. .inally, it is important to note that subsidizing prices Designed by: Roots Advertising Services Pvt. Ltd. Printed by: PS Press Services Pvt. Ltd. of basic goods such as water is essentially just an indirect way of redistributing income. Precisely for this reason, economists argue that it is far preferable to meet income distribution goals through a comprehensive, PPIAF government-administered social safety net, in the form 1818 H Street, N.W. of targeted income transfers. This avoids the side effect Washington, D.C. 20433 Phone: +1(202) 458-5588 of distorting the prices of goods and services. It also Fax: +1(202) 522-7466 gives the beneficiary household the freedom to Email: info@ppiaf.org determine how these resources should be spent. Website: www.ppiaf.org However, in many developing countries, the administrative obstacles to such a system are often too great, and subsidies to specific essential services may therefore become justified. While there is no easy Water and solution to the problem of targeting subsidies in the Sanitation Program urban water sector, the simulations presented here have 1818 H Street, N.W. shown that there are relatively straightforward ways of Washington, D.C. 20433 making significant improvements; these mechanisms Phone: +1(202) 473-9785 10 offer the potential for substantially reducing subsidy Fax: +1(202) 522-3313, 522-3228 Email: info@wsp.org budgets while still providing assistance to the poor. Website: www.wsp.org The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors and should not be attributed in any manner to the Public-Private Infrastructure Advisory .acility (PPIA.) or to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. Neither PPIA. nor the World Bank guarantees the accuracy of the data included in this publication or accepts responsibility for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this report do not imply on the part of PPIA. or the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. W A T E R T A R I . . S A N D S U B S I D I E S I N S O U T H A S I A