Sunteți pe pagina 1din 13

Pergamon

WoddDevrlopment, Vol. 22, No. 1, pp. 71-83,1994 Copyright 0 1994 Pergamon Press Ltd Printed in Great Britain. All rights reserved 0305-7501094 $6.00 + 0.00

Water Markets, Market Reform and the Urban Poor: Results from Jakarta, Indonesia
RANDALL CRANE* University of California, Irvine Summary. - Roughly 20% of the eight million persons ia Jakarta receive water from municipal water
counectiens to their homes. Remaining supplies are essentially private and include street vendors, staudpipes, and wells - the last of these an inexpensive but inueasingly contaminated source in many areas. An expansion of the netwcrk of in-house connections is in progress, but will take several decades to complete. In the interim, various strategies have been advanced to deal with persistent problems of access to good water by the poor. This study examines the impacts of one such plan put to action, the April 1990 deregulation measure permitting private homes with water connections to resell municipal water. We present preliminary evidence that the primary social benefits of deregulation are the money savings and increased consumption by former vendor and standpipe customers. In most respects the aggregate effect is equivalent to a costless expansion in the standpipe system, with the difference on the supply side that significant transfers from vendors and standpipe operators to household resellers may

1. INTRODUCTION
Limited access by the poor to clean water is among the most basic issues faced by authorities in the

cities of the developing world. This and related concerns explain a good part of the billions of dollars planned for urban water infrastructure investment over the next few years. The funds are mainly intended to improve water treatment capabilities and expand the population served by in-house connections, but administrative and regulatory reforms are also increasingly advocated as either cost-saving strategies or equity-promoting measures, or both.* The value of these efforts has been difficult to estimate, however, owing to the limited availability of data on actual water consumption patterns, especially in squatter settlements. Information regarding the demand for water by the urban poor, including the amount they pay, where they find it, and how they would react to a change in either prices or supply structures, remains relatively scarce. As a consequence, the behavioral response to any given policy intervention has only infrequently been incorporated into infrastructure investment and administrative reform plans.* This paper addresses the gap by analyzing the effect of a competition-enhancing market reform in Jakarta, Indonesia. Most of the residents of Jakarta do not have direct access to piped water and increasingly find traditional low-cost sources, such as rivers and wells, contaminated or depleted. In some parts of the metropolitan area, people obtain most of their water from alternative sources, including street
71

pedlars who deliver water door-to-door and neighborhood standpipes. These supplies are somewhat less healthy and dramatically more expensive than water distributed through house connections. In some locations, inadequate access to reliable and low-cost sources of clean water has led consumption to fall to health-threatening levels. An aggressive standpipe investment program has reduced the distances between neighborhood taps over the past few years, thus lowering water transport costs and promoting competition among tap operators. As another component of the effort to increase the number of suppliers and thereby lower water costs to the urban poor, the municipal water enterprise also permitted all households with a metered connection to resell water starting in April 1990. This study provides preliminary evidence on the market impacts of the deregulation measure after its

*I am grateful to Dr. Jun Hang, formerly of The World Bank and now of the International Finance Corporation,
and to several senior staff of the Jakarta municipal water enterprise, Perusahaan Air Minum Jaya, for comments, access to data and other essential help. Gunnar Eskeland and Emmanuel Jimenez provided valuable advice on the design of the questionnaire. I also thank Marlon Boarnet, Amrita Daniere, Eric Heikkila, and especially three anonymous referees for many suggested improvements to earlier versions. The data for this study were obtained as part of a project completed for The World Bank, but the views expressed are mine alone and should not be attributed to The World Bank, its management, or any of its member countries. Final revision accepted: August 20,1993.

72

WORLD DEVRLOPMENT

The first is to generate new information regarding the linkages between water markets, market reform, and urban poverty. While the results support the conventional wisdom that market liberalization can only help consumers, as well as increase aggregate efficiency, they also underscore the pivotal role of private water markets in the larger urban service distribution system. Water is better described as a private market commodity than a public service for millions of people in Jakarta, and efforts to improve local circumstances must account for local market conditions to be most effective. The second aim of the study is to calculate the consumption value of deregulation, as in a benefitcost assessment. Within the limits of the sampling technique, the study considers the value of two categories of consumer benefits, which are labeled direct and indirect. Direct benefits are defined as those restricted to households choosing to buy from the new water source, and are reflected in the time and money saved over alternative supplies and the value of increased consumption. The so-called indirect benefits equal the money savings by current vendor and standpipe customers owing to price reductions induced by increased competition. The latter category of benefits do not accrue to the consumers of the deregulated water source, although they are a direct consequence of the new policy. In addition to these, the change in the welfare of sellers is a third, unmeasured, category of potential benefits. While sizable transfers from vendors and standpipe operators to household resellers might well occur, the primary social benefit of deregulation appears to be the money savings and increased consumption by former vendor and standpipe customers. Public sector resource costs are negligible, so the effect on consumers is similar to that of a zero-cost expansion in the standpipe system. The next section provides context by reviewing the highlights of the literature on urban water markets, and the structure of these markets in Jakarta. The results of a household expenditure survey are presented, including a description of water consumption patterns in the sample. The paper then examines the benefits of deregulation, as an element of future water sector reform strategies. Finally, the concluding section offers some further thoughts on water policy toward the poor in Jakarta.

first year. There are two objectives.

2. URBAN WATER MARKETS A number of studies have documented vending and standpipe practices in developing countries. Many have also analyzed the respective roles of public policy and private markets in the overall water distribution system. Most of this work has concen-

trated on strategies for improving the adequacy of water supply in rural areas, where the greater number of poor live and where the share of the population with reasonable access to good water is typically much lower than in cities.3 This focus, reflected in the rapid increases in rural infrastructure investment throughout the past three decades, was also driven by the argument that services to rural residents cordd be brought to acceptable levels at a lower per capita cost than in urban areas. Somewhat less space has been devoted to the study of water delivery systems servicing the urban poor in recent years. While individual cities have been examined in some detail for broad-based case studies of poverty- and development-related issues, systematic assessments of how urban water policies affect vending and other water sources for the poor are rare. Accelerated urban population growth over the last 20 years, coupled with diminished access to urban services in many instances, is sparking renewed interest in urban water delivery issues, however. A benchmark for this literature is the comparative survey of vending practices in Asia, Africa and South America by Zaroff and Okun (1984). The recurring themes in their data were the high relative user cost of vended water, on both a per unit and monthly basis, and its uneven quality due to poor handling and source pollution. They nonetheless conclude that vending is the most appropriate means of delivering water in many instances, though most vending systems could be improved through public sector intervention. The latter would include equipment upgrades, using house connections as vendor suppliers, and route and price regulation to increase water quality and lower its price. In a simulated application to Surabaya, Indonesia, these policies reduced the consumer price of vended water from about $3.90 to $1.04 per cubic meters Based on evidence from several locations, Whittington, el al. (1989) argue that vending is often sufficiently competitive to keep vendor profits below rent-earning levels. In Whittington. Lauria and Mu (1991). however, an analysis of the private water supply for the city of Onitsha, Nigeria, illustrates how sensitive this result is to market conditions. Only 8,000 households in Onitsha. a city of 700,000, have formal piped connections. The remainder obtain most water horn a system of tanker trucks and truck-supplied vendors. Trucks buy water at about $0.20 per cubic meter, which they in turn sell at $0.95 to $2.12 per cubic meter, at a markup of roughly 50&1000%. Small vendors resell this water at $2.7@-$7.90 per cubic meter, a further mark-up of about 300% over their cost. The significant labor costs of the latter group keep excess profits to a minimum. In the case of tanker trucks, however, substantial economic rents appear to be generated. The authors were unable to provide a definitive explana-

WATER MARKETS AND MARKET REFORM

13

tion the existence continued monopoly ings by owners, although suspect that truck owners enforce entry tions or the prospect a new Bankfinanced of the water system lowered profit and thus In an study of Haiti, Fass presented the of a of 2,600 and vendors in 1975. water enterprise estimated its population 400,000 people, half reached inhome and half 40 standpipes but the determined the to be 640,000 people, nearly 60% than thought. the populations by connections standpipes were lower than Approximately 435,000 used other, rect, sources their water none of were accounted in official sector plans. two most sources were obtained at cost from with in-house tions, used 15% of population, and water, purchased nearly 50%. water was in three by 25 at $7.50 $15.00 per meter, by households with connections at $2.50 to as high as $50.00 per cubic meter in the dry season, and by 14,000 traditional water sellers who bought their water from the second group and resold it at $1.50-.$2.75 per cubic meter. As the author remarks, it thus became clear that water was much less a public service than a large private industry, generating several million dollars in revenues per year. The city water board earned less than onefifth the amount people spent on vended water. Notable consequences include (a) the transfer of wealth from poor consumers to the households selling water from their homes, (b) the meager earnings of water carriers compared to sellers with piped connections, and (c) the wildly uneven level of access to good water across income groups and neighborhoods. Fass observed that these patterns were not unique to urban Haiti, and that indeed the situation in Jakarta, Indonesia, in the mid 1970s was similar in several respects. As shown in Berry and Sierra (1978) and Berry (1982), a system of vendors and private standpipes supplied as much as 10% of the city at $1.70~$2.10 per cubic meter in 1977. Vendors bought their supplies from private water connections at about $0.09 per cubic meter. These water connections differ somewhat from those in urban Haiti by their official status and restrictions on competition. Classified as standpipes. or hydrants, their location and operator were selected by the water authority. In addition, the prices charged were subject to controls. These controls appear to have been largely ineffective, but entry into the market was tightly regulated. Private households were pm-

hibited from selling to their neighbors, for example. Berry and Sierra (1978) evaluated the extent of monopoly power by standpipe operators, and its implication for consumer pricing. They also calculated the distances between standpipes that would retard excess profits while ensuring reasonable access by the poor. The main recommendation was an aggressive standpipe investment system to reduce existing distances accordingly. Two lessons of critical importance to our study emerge from this brief review. In many cities where access to good water is restricted, including Jakarta, the private sector plays a central role by reselling water obtained from private taps. In particular, indirect access to public sector water may be more important than direct access, especially among the urban poor. The other lesson is that the consumer price of water depends on market structure, i.e. on the degree of competition among resellers. This is influenced by both the regulatory environment and private market characteristics. The next section updates water market data in Jakarta to 1991, following more than a decade of considerable investment in both house connections and hydrant systems - and, in 1990, the legalization of household resales.

3. WATER MARKETS IN JAKARTA TODAY In many respects, access by the poor to good water in Jakarta has both improved and deteriorated over the past 15 years. As the commercial and administrative hub of the fourth most populous country in the world, a disproportionate share of the nations population growth over the past three decades has fallen on metropolitan Jakarta. By 1990, 20% of the urban population of Indonesia, or over eight million people, lived and worked in the city. The surrounding fringe areas contained an additional 8.4 million. The rapid urban growth and associated industrialization in the metropolitan Jakarta area are placing increasing stress on the natural resources of the region, with domestic and industrial wastes causing severe degradation of ground, surface and coastal waters. Fully 40% of the population still depend on groundwater, but the quality of this source is already poor in many areas and threatened in others. Incidence of infant mortality and of gastroenteric. pulmonary and viral diseases caused, in part, by polluted and congested conditions is high, especially among the poor and those living in northern, floodprone areas of the city. Moreover, the ground water table is saline from the coast to as far as five kilometers inland, leaving well water all but useless at some sites. The other main water source is the municipal water authority (PAM Jaya). This is distributed by several means, the most direct, convenient and least

14

WORLD DEZVELOPMFNT in the households barrel or reservoir. Customers buy water by the cart or by the piRul (two cans). Vendors in Jakarta commonly purchase their water from hydrants, although the vendors are often employees of the hydrant operators and work for a salary or at a piece-rate. In other cases, the vendor may rent the cart from a hydrant owner. Vendors tend to have regular routes, and other studies of Jakarta have reported that vendors stray from their normal route only when requested to do so by a consumer. Lingering concerns remain over the market power of both hydrant operators and vendors. In 1990 and 1991, the official consumer price of water delivered by an in-house piped connection was 350 Indonesian Rupiah (Rp) for each of the first 15 cubic meters per month (about $0.18 in mid-1991). An escalating block tariff applies to larger quantities, peaking at Rp 900 per cubic meter for amounts above 50 cubic meters per month.9 In addition, installation of the hook-up requires a fee of Rp 81,000. Each Hidran Umum and Hidran Contoh operator paid Rp 150 per cubic meter for water from PAM Jaya, at least in those cases where meters were functioning correctly. The official selling price, prominently displayed in every hydrant facility, was Rp 500 per cubic meter in 1991. Actual selling prices tend to be from Rp 1,000 to Rp 4,000 per cubic meter, including the rental price of the cart. The end-user price of vended water varies substantially from one area to another, but can be as high as Rp 7,000 per cubic meter, some 20 times the price for connected homes. These vast differentials have led to both equity and health concerns by water planners. While more hydrants are added each month, some observers have expressed the concern that hydrant competition is kept to a minimum by spacing hydrants sufficiently far apart. The vendor system is entirely private in nature, and unregulated. Apart from price controls, already in place to some extent for hydrants, the clearest means for improving consumer welfare is to introduce more competition. Lovei and Whittington (1993) and Shugart (1991) have reexamined these arguments, based on new evidence for Jakarta.lO In principle, the ability of hydrants and vendors to earn excessive profits is limited only by the availability of close substitutes for the services they offer, and in many areas these are few indeed. Shugart (1991) provides an especially clear picture of vending and hydrant practices in the northern and western parts of the city, where these markets are most active. He discusses how limited competition among vendors and the spatial pattern of hydrants interact, and suggests that the degree of competition among vendors varies by neighborhood - depending on. among other things, whether competing vendors are members of the same ethnic group or have migrated from the same village.rl Hydrants are another story, with the extent of price

expensive of which are in-house connections. The water enterprise listed 225,000 household water connections in June 1991, or about 20% of the households in the service area, with expansion plans calling for an additional 25,000 hi each of the next few years. In the southern, central and eastern parts of the city, those who lack official connections continue to rely on well water for most purposes. Residents of the northern and western areas, where groundwater supplies are of especially poor quality, generally buy their drinking and bathing water from either hydrants, a kind of private standpipe. or street vendors. The latter in turn obtain their supplies from hydrants, which are connected by pipe to the municipal distribution system. Although somewhat public in origin, private individuals operate most hydrants as for-profit enterprises. The two main types are (a) Hidran (Imum. the franchise for which was purchased outright by an individual, and (b) Hidran Contoh, which are ostensibly free to the community and operated by an individual awarded that privilege with the support of neighborhood leaders. A third hydrant type is the water terminal, or Terminal Air. These are freestanding water tanks supplied by water trucks in remote areas beyond the reach of the water mains. To purchase hydrant water, a household must transport buckets of water from the tap back to their home by hand, or by cart rented from the hydrant operator. In a few cases, Hidran Contoh use long hoses to deliver water to each house so that no water need be carried. The traditional Jakarta hydrant is the Hi&an Umum. with 1,241 in use in May 1991. The number of Hidran Contoh has risen dramatically in the past four years due to au aggressive investment program of the national planning ministry (BAPPENAS), and there were 633 by May 1991. While the Hidran Contoh are meant to be more community oriented with respect to both location and operation, whatever distinctions exist between the two hydrant types tend to be immaterial to consumers in most instances. Service quality differences from one hydrant to another appear to depend far more on the pressure in the distribution system at that location, and the size of the hydrant reservoir, than how the hydrant was financed or how the franchise was awarded. In particular, market forces largely determine selling prices rather than official price guidelines. A recent estimate by PAM Jaya of the number of vendors servicing Jakarta is around 8,000. An extensive housing survey conducted in 1988 indicated that 32% of the municipal population bought vended water (Struyk. Hoffman and Katsura, 1990). Vendors transport water in plastic or tin jerrycans. varying in size from 18 to 100 liters each, by means of single axle handcarts carrying four to 18 cans. To sell water, the vendor hand-carries jerrycans to the house from the street or alley and deposits the water

WATER MARKET.5 AND MARKET REFORM

75

competition depending almost entirely on the distances involved. Lovei and Whittington (1993) emphasize the potential for sizable rent earnings by both vendors and hydrant operators, and conclude that existing arrangements result in too few water taps, and too many resources hauling water by cart They argue that deregulation of water sales coupled with the easing of supply constraints could substantially lower both hauling costs and water prices, and thus reduce the ability of utility staff and neighborhood officials to capture rents. The municipal water authority implemented a policy with precisely this intent in mid-1990: The legalization of the resale of water from household connections. The remainder of the paper presents data on its effect and the corresponding implications for water sector policy in the metropolitan area.

4. A WATER EXPENDITURE SURVEY FOR NORTH JAKARTA To encourage a stable and competitive water sup ply for low-income families, the municipal water enterprise announced in April 1990 that all households with a metered connection would be permitted to sell water to their neighbors and to vendorsi A prohibition on this practice was strictly enforced up to that time. To assess the impacts of supply deregulation, a household expenditure survey was administered in Jakarta in mid-1991. A complementary purpose was to collect generic information on water consumption by low income households, including the sources used, the quantities bought and prices paid. Due to time and funding constraints, the study confined itself to the area of Jakarta with the worst ground water conditions: North Jakarta, a jurisdiction covering roughly 155 square kilometers, or 23% of the citys total land area, and home to just over one million people in 1989. Water availability is a particularly serious problem in this area and both vending and public hydrants are relatively widespread. It is also the area where the largest concentration of houses reselling water is located. Extensive interviews were conducted in July and August 1991 with 291 residents. The sampling procedure resembled a modified cluster approach, although severe constraints on time and resources made proper sampling techniques somewhat impractical. Address listings were unavailable for the neighborhoods of interest, for example. Low-income areas were chosen through site visits and identified on street maps, and we interviewed clusters of at least 30 households in each of the major subjurisdictions (kelurahan) of North Jakarta. Fewer than two weeks were available for the sample selection and data collection phases of the study, and as with other rapid appraisal type studies in developing coun-

tries, every attempt was made to collect data as rep resentative and thus useful as time and funds allowed. Although our results are consistent with other household data collected from North Jakarta in recent years, caution should be exercised when generalizing the results beyond the sample. The median monthly expenses of the sample households are just over Rp 250,000.r3 While the sample households are quite poor and live in areas lacking most public services, 48% own their homes and half have lived in their current home at least five years. The 291 sample households averaged 5.5 people per household, and represent a total of 1,341 people altogether. Table 1 summarizes water consumption behavior for the sample. These descriptive data are revealing of both fundamental trends as well as the impacts of the market reform. Over 73% of the sample households report buying water from water vendors, 27.5% from public hydrants, and 10.0% from household resellers; i.e. neighbors with in-house connections.*4 Some 13.4% of the sample obtained water from the PAM Jaya piped system. About 12% of these reported reselling water to their neighbors. The number of families buying some water from neighbors is approximately one.-third that buying water from hydrants, and one-sixth the number buying water from vendors. Not surprisingly, the average volume of water consumed is greatest for those households with water connections. Sample households with connections report buying nearly 62.2 liters per person each day. The second highest volume is for resold water, at 26.7 liters per person each day. The average amount purchased from hydrants is virtually the same. Vendor customers purchase much less than those of other suppliers, an average of only 14.6 liters per day per person. For North Jakarta overall this behavior is consistent with the reported cost of water, which is highest for vendor-supplied water, at Rp 5.040 per cubic meter. This is over I4 rimes the average cost of PAM Jaya water through a piped water connection. A larger share of vended water is reserved for drinking and cooking than for the other sources. however. Water from hydrants is the next most expensive after vended water, at about seven times the piped water cost. Water resold by households is only slightly less than hydrant water, at Rp 2.090 per cubic meter. Sample households report the largest expenditure share for vended water (7.5%). slightly less for hydrant water (4.8%), and less still for resold water (3.2%). Families rarely depend on resold water as their sole source, however, so this last term underestimates the average water cost for those households. The smallest budget share is for piped water (1.8%). Because piped water is considerably more inexpensive on a per unit basis, low-income households with connections choose to consume three to four times

16

WORLD DEVELOPMENT Table 1. Water cauumption profie (Sample means, except when? n&-d)
Vendor Hydrant customers Household resales customers 29 connected
hOU&OldS

Number of households Percentage of sample* Cubic meters per month Liters per day per person Price per cubic meter Minutes per week obtaining water Meters to nearest hydrant Meters to house connection Meters to household reseller Monthly household water cost Monthly water cost per person Share of household budget for water Median monthly expenditures

214 13.5% 2.4 14.6 $267 lb 131 91 113 Rp 13,590 ($7.06)

10.0%
8.5 25.9 Rp 2,430 ($1.26) 83 60 42 9-I 5.7 26.7 ?ZOY 87 104 20 Rpy521 ($4.b5) Rp 1,629 3.2% Rp 260,m

38 13.4% 13.2 62.2 %OY;; -

Rp4,854 ($2.52) & 650 1.8% Rp335,OOO

Rp 2,2io

Rp9,468 ($4.92)
Rp 1,663 4.8% Rp 295,ooo

7.5% Rp 232,500

YThe sample shares in the first four columns add up to more than 100% as 17% of the surveyed households purchased from moTe than one source. Source: The source for the data in all the tables of this paper is the household survey administered by the author and described in the text.

the volume of water consumed by vendor and hydrant customers. Median monthly expenses were highest for co~ected households, at about $174. and next highest for those buying hydrant water. Households purchasing neighbor supplied water were next lowest and, significantly, the lowest median monthly expenses were those of households buying vended water, at about $135. To measure the sensitivity of consumption to household and market variables, particularly prices and water transport requirements, a simple behavioral model of water demand was estimated for the households buying water from vendors or from hydrants. (See the Appendix.) Table Al in the Appendix shows that the demand for vended water is significantly influenced in our sample by the Price of water, the time per purchase, and the capacity of the home water reservoir, all with the expected signs.ls The point estimate of the price elasticity of water demand is -0.48, which is consistent with the estimates reported in other studies of water demand in developing countries. I6 The interpretation of this measure is that a 1% increase in the price per cubic meter will, on average, reduce the quantity bought by 0.48%. Controlling for differences in income, family size, and other family and community characteristics, the demand for vended water is thus moderately responsive to price variation. Purchases will fall if prices rise, and will increase if prices fall, all things considered. The expenditure elasticity of demand is estimated to be highly inelastic, at 0.03, but not significantly different from either zero or one. There may not be enough income variation in

the survey, controlling for all other factors, to demonstrate a separate role for household resources. It may also be that expenditures are a poor or biased meaSure of income, or not as complete a measure of the resources available to the household as would be useful. As shown in Table A2 in the Appendix, the price elasticity for hydrant water is estimated to be -0.60, in roughly the same range as that for vended water. In this case, the time per purchase and the capacity of the storage tank are not significant. This is certainly not what we would expect of the time variable, where the effort and time necessary to collect hydrant water is substantially above that necessary for vended water. Again, there may be a problem in the way the variable was measured. It will be useful, for example, to account for who it is that usually collects the water. A childs time may have relatively little value, while an adults time has higher opportunity costs. The absence of an effect of reservoir capacity is less surprising, given the likelihood that less water can be carried per trip than can be stored at the home.

5. THE VALUE OF DEREGULATION This paper considers two categories of consumer impacts of the change in regulatory policy, which we call direct and indirect for convenience. Direct impacts refer to those realized by new household water customers, including both vendors and direct purchases by other households. Ibe last of these is

WATER MARKEXS AND MARKET REFORM

77

measured by calculating the savings in water expenses and time for customers buying resold water. Indirect impacts include the effects of increased competition on prices charged by other suppliers, and are reflected in price reductions to customers of both hydrants and vendors. Without good longitudinal data, it is difficult to attribute changes iu enduser water prices to any one factor such as the introduction of new supply structures. Even where prices rise or remain stable, new supplies and other competition-enhancing reforms may well keep downward pressure on prices. Gur survey, based on the memories of the respondents, indicates steady prices over the first year of the new policy. This is consistent with the interviews with vendors and hydrant operators by Shugart (1991). who attributes the flat prices to expansions in the hydrant system. While consumers obtaining water exclusively from either hydrants or vendors have not benefited directly from household resales, the indirect benefits of entry by new suppliers are thus tangible. Whether new supplies are in the form of private households with connections, community controlled hydrants, or some other fixed location source would not appear to matter in this respect. The impact of this competition is difficult to estimate given the available data, but preliminary field visits suggest it is relatively minimal. The potentially sizable transfers among water suppliers are another matter, and represent a third category of impacts. To the extent household resellers capture market share from both vendors and hydrants, they will gain and the last two will lose. In some areas, household resellers have clearly captured a share of the revenues formally going to vendors - who, on average, may be poorer that the new resellers. Diminished market share for vendors is most likely unavoidable, in any case. These gains and losses am difficult to calculate in our case, however, as we have very few observations on water resellers. The issue is probably whether consumption benefits outweigh whatever negative impacts these transfers may have, and this is largely answered by the willingness of consumers to change suppliers. The discussion below thus emphasizes direct consumption benefits. Based on the sample data, there do appear to have been such benefits. In comparing sources from one year to the next, and for drinking and cooking versus bathing and cleaning. Table 2 shows that while vendors provide most of the respondent drinking and cooking water needs, the largest proportional growth has been in rainwater and water resold by households. The households represented in this sample used vendors and piped water (from new connections) more in mid-1991 than one year earlier, well water somewhat less, and hydrant and household supplied water about the same. One might have expected to see a larger gain from hydrant use, and

Table 2. Proportion of sample using sach water sowce in


July 1991, and the change in each proportion over the pmolu year

Use fa driakiag Use fa bathing and cleaaing sadcockiag Dry g seuml* 61.0% 5.4% 26.4% 0 11.0% 19.6%
1.7% 4.8% <.3% 21.4% 12.7% 57.1%

&?a 47.3% 15.8% 9.6%


2.7% 0.7% 32.3% 1::::

Vendor

Hydrsat Househdd $pe


Househddrcade Neighborwell Own well
RiVCl RlillW~

L$ 5.3% -1.9% 7.9%


12.5% 0 -0.9% -1:.6%

~numbersdduptomorcthrn1oO%rrmanyofthc surveyed householdspurchasefrom moccthan one same.

from household resales, given the recent expansion of the system but in many areas hydrants primarily serve vendors. At the same time, resales appear to play a supplemental rather than primary supply role. Few households buy water exclusively from neighbors, for example. Moreover, in areas where household connections are scarce, alleyways are narrow, and hydrants relatively plentiful, there was little evidence of household resales. Put another way, the impacts of water sales deregulation are not geographically uniform. Spatial access to sources supplying water which must be transported by households appear to be a factor. One would expect that the purchase of resold water is a decreasing function of distance to both reselling households and the potential suppliers represented by households with piped connections. For families living over 100 meters from any house. with a municipal connection, for example, only 1.4% bought water from neighbors with connections. Ihe number buying from households, however, rose to over 19% among all households within 50 meters of a house connection. Moreover, 43% of the households living within 50 meters of a household who sold water in turn bought some resold water (Figure 1). The spatial competition represented by the degree of access to hydrants may also be important, as the share of households buying water from neighbors falls with proximity to hydrants. The latter relationship does not hold up in the regression analysis reported in the Appendix, where distance to the nearest hydrant helps explain neither the consumption of vended or hydrant water, but it would be worth exploring carefully in future work. The household value of these changes can be estimated, at least approximately, to obtain a quantitative measure of the consumer benefits of deregulation. Generally speaking, we are interested in the

78

WORLD DEVELOPMENT

Hydrant

q Vended

water

n Hydrant water

q Resold

water

Figure 1. Percenrage of sample by source and by aWmce of potential sources.

size of these benefits, in the equity of their diatribution. and how they compare with any associated sup plier costs. Again, however, the last two of these are difficult to pin down with any real accuracy given the nature of our data. We therefore concentrate on the first. As defined in this study, the direct benefits of water sales deregulation come in two forms: the money and time savings from a change in suppliers, and the additional benefits of increased consumption. These must be positive overall, or the household would not have changed suppliers. but it is useful to examine how the various components compare. If a household had been consuming 100 liters per day of vended water at Rp 5 per liter, for example, and switched to resold water at Rp 2 per liter, the most obvious savings is the difference in the cost of 100 liters, or Rp 300 per day. That is the out-ofpocket savings. There may also be a time savings, reflecting the difference in the time and effort cost of obtaining water. A switch from vended to resold water would probably involve increased time, so this would enter the benefit side negatively. If the household had previously spent ten minutes per day obtaining water, and now must spend 30 minutes per day at that task, the net cost of the change is the money value of the additional 20 minutes. This in turn depends on which person in the household obtains the water, and the opportunity value of their time in other activities. If that person values that time at Rp 300 per hour, for example, the additional time required to obtain resold water effectively costs the household Rp 100 per day. There is also the additional direct benefit repre-

sented by any associated increase in water use, and by extension the health and other indirect benefits of increased consumption. It is likely that the household will choose to buy more than 100 liters per day now that a lower cost alternative is available. Say the household increased its consumption to 120 liters, for example. While the family must pay Rp 2 x 20 liters = Rp 40 per day for this water, it is also generally true that the household would be willing to pay even more. Thii surplus* can be considered a measure of how much better off the household is by purchasing more water than before, at a cost below what it is worth to them.* The conventional approximation of the household value of this surplus is just half the increased consumption times the change in price; i.e. (20 liters x Rp 3)/2 = Rp 30 in our example. Thus, a measure of the total direct benefits gained by this hypothetical family per day is Rp 300 money savings, less Rp 100 time cost, plus Rp 30 consumers surplus, or Rp 230. The fast step in this procedure is straightforward and can be estimated from observed data; i.e. the average price paid by source, and the average quantity consumed. The second step is more problematic, as it requires an accurate accounting of the time devoted to water collection and the value of that time.rg The third step, the calculation of the consumers surplus gained by changing suppliers, involves estimating the expected change in water use. We do not have data on this change, but we can approximate it by using the estimated elasticities of demand from the previous section to calculate how consumption will increase with a change in suppliers. The price elasticity of demand is a measure of

WATER MARKIXS AND MARKET REFORM Table 3. Welfare gains by customers of household ru&rs vahe

79

of price difference Switched from hydraat supplies Rp440 Rp 2,470

Per cubic meter Per month per family

Switched from vended supplies Rp 3,020 Rp 8,070

the propensity to vary water coosumptioo as the price of water varies. 10 the Appendix, we estimate that hydrant customers have a price elasticity of -0.60, which implies that if prices fell by 18%. the quantity of water purchased would rise by 60% of that amount, or roughly 11%. Table 3 summarixes these calculations for the hypothetical situation of the average household facing average prices. When a household changes suppliers, the calculation of consumption effects must also account for the change in the time-cost of waiting for and transporting water. While the average price reduction in going from vended to resold water is 60%. from Table 1, the household time devoted to bringing water into the house will increase - the real savings of switching from a vendor is therefore not so great as it might appear from the financial accounting. This component of the cost of water depends both on the time spent transporting resold water and the value of that time, which in turn depends on the alternative uses of the time of the person collecting the water. As in the price savings, there are two welfare effects: the value of the time savings 00 the initial quantity bought, and the value of increased coosumptioo attributable to the time difference. These time savings gained by new customers of household resellers can also be illustrated, though without any statistical measure of accuracy. prom Table 1, the average monthly difference in time spent buying water from resold water versus vended water is about 4.5 hours per month. The average number of purchases per week was five. We can then calculate an estimate of the increased time required per purchase when switching from vended to resold supplies of 15 minutes. This will reduce coosumptioo, holding all other factors constant, about 30 liters per month on average. The value of that decrease depends on the value of time in this use. The time savings of a switch from hydrant to resold supplies depends most directly on the relative proximity of the two supply sources. There may also be. differences in waiting times in the queue. Table A2 provides no evidence. however, of a significant relationship between the demand for hydrant water and the time required to purchase and transport that water. This counter-intuitive result may be due to the form of the data, the specification of the regression

equation. or other features of the local water market the model does not capture. 00 the surface, the effect of deregulation on prices paid and quantities consumed by low-income families thus appear both positive and significant. Many households have benefited from the availability of closer and in some cases less expensive water. Customers who have switched to resold water from vended water now face a lower unit price and no doubt increased their water consumption significantly as a result. Those who switched from hydrant water have realized less of a price decrease, on average, but in many instances likely gained some savings in time and effort costs associated with water transport The public costs of the new regulatory policy appear miniial. The effect of the new policy is thus equivalent to an inexpensive expansion in the standpipe system.

6. CONCLUDING REMARKS CONSIDERATIONS

AND

On balance, it is difficult to argue with a policy stimulating competition for an essential commodity where, in addition, the poor are burdened by prices well above those faced by others. In the case at hand, the costs of market deregulation are almost exclusively the rents lost by suppliers facing a diminished market share. These may or may not be a social coocem. The direct benefits of the legalization measure are reflected in prices below the market prices of other suppliers, in the additional consumption the lower prices permit, and in the reduced time necessary to transport resold water versus standpipe water. In our sample, consumer benefits are about 59% of the average household water bii for ex-vendor coosumers (less additional time costs), and about 26% of the average water bill for ex-hydrant consumers (plus any time savings). Vendors buying water from household resellers pay more than from their normal suppliers (standpipes), but do so to avoid waiting time in the queue and to reduce transport distance. The indirect benefits of deregulation, in the form of increased supply competition. are likely minimal in the short mo and are difficult to measure with our data in any case. The overall extent of reselling is limited, for example, by the comparatively high sup ply costs of household resellers due to rising block tariffs, and the reality that water which must be transported by the customer is not a perfect substitute for water delivered into the house. The latter characteristic also limits the value of an expanded standpipe system. Indeed, neither household resales nor the hydrant program are long-term substitutes for an expanded piped water system. These policies have eased supply problems for vendors and households alike in the interim but, especially considering

80

WORLD DEVELOPMENT five-fold after obtaining a connection, and the monthly cost of the water and the connection installment would still sum to less than the average vendor bill. Hydrant customers face similar savings. The credit plans are little used, however. These calculations also ignore both the Unofficial fees one must pay in many instances to facilitate the connection process and the necessary improvements to the interior of the structure, for which no formal installment plans are available. In addition, several households mentioned the difficulties associated with budgeting water expenses on a monthly rather than a daily basis. The policy implication is that fmancing a lump sum water investment is a complicated matter for the poor, and simply having credit available is inadequate to the task. Alternative strategies should be explored, including folding the connection fee into the water bill somehow and considering more flexible-meaning more frequent-payment plans. The financial constraints keeping households from obtaining piped connections that lower their costs, and increasing their water use. should be studied further. A fmal caveat: While the importance of accounting for private market activity in water sector planning is an unmistakable implication of this work, care should be taken when generalizing beyond the sample. The low-income areas of this city and others differ markedly in their physical and economic circumstances, including the pressure available in the piped water system, the land tenure of residents, the level of organization among vendors, the size of water cans vendors use, the way hydrants distribute water, the extent of in-house water connections, and the size of the alleys and pathways between homes. The basic lessons of market reform no doubt apply equally well to other cities in other countries, but the magnitude, character, and distribution of change will depend on local market structure. Future research should review and assess the particular circumstances of each area before advocating one kind of reform over another.

the value of time and effort, the consumer cost of water is substantially higher in either instance. Until connections become more widespread, a fundamental characteristic of the water sector in the areas of Jakarta with poor groundwater is that it is essentially private. Thus, public policies attempting to work within the framework of supply and demand at least avoid many of the unintended consequences of those which do not. Other obstacles remain, however. One of the more serious problems limiting the effectiveness of the deregulation is the widespread ignorance of the legality of resales. Among those families who were aware such sales are legal, over 28% bought some resold water. The purchase rate was below 7% for households who did not know, which may be partly explained by households who were told it was legal but were not sure. We would expect increased demand for resold water if village and other community leaders become more involved in promoting the legality and availability of sales from household connections. These efforts might take the form of neighborhood meetings, neighborhood directories, and even community facilities to encourage the practice of sales from house connections. This form of community participation will ensure that families have community support for the practice of buying from their neighbors, and that the activity is clearly legal. It will also improve the quality of information regarding sources of water, and thereby of prices paid. Surprisingly perhaps, not all households who have the opportunity to connect do so, even when it would save them money over time - as it would in most of the examples given here. The obstacle is partly the connection fee of Rp 81,000 (S42), which is a large lump sum payment for highly marginalixed families. Credit schemes exist that, on paper at least, should all but eliminate this burden. PAM Jaya allows the fee to be paid directly to them in three installments. The municipal bank offers a 12-month financing plan, requiring monthly payments of Rp 7,000 or so. From Table 1, the average vendor customer could increase his familys water consumption

NOTES 1. See, e.g., Munashinghe (1990), Okun and Lauria (1991) and the overviews by The World Bank (1992) aad The World Bank Water Demand Research Team (1993). 2. The value of such infamation is d&ussed in some detail by Whittington and Choe (1992). which contains a systematic review of the literature on the vahration of nonhealth benefits of water sector improvements. 3. E.g., Saunders and Warford (1976), Briscoe and deFerranti (1988), Whittington, e? al. (1989). Briscoe, et al. (1990). Wood and Palmer-Jones (1990), Okun (1991),

Whittiagtoa, Lauria aad Mu (1991), and The World Bank


Water Demsnd Research Team (1993). 4. Recent examples include Linn (1983) and Rasarda and Parnell (1993) on large cities, Schteingart (1989) on Latin Amesice, and Marcussen (1990) and Jelhnek (1991) on Jakarta. Exceptions include Trava Manxanilla, R. Calleros and B. Rodriguez (1991) for border cities in Mexico, Fass (1988) for Haiti, as discussed below, and Hardoy, Cairncross and Satterthwaite (1990).

WATER MARKETS AND MARKBTREFGRM 5. All dollar figures have been expressed in constant 1991 US$ terms, using the Consumer Price Index. 6. 7. Buro Pusat Statistik (1990). Yayasan Dian Desa (1990) and Shugart (1991).

81

and some curection. l&se data are considered somewhat more reliable than the coIleded income data for that reason, which in any event were not reported for nearly one-third of the sample. 14. Hydrants, vendors and co~ected households are all potential resellers of water. For clarity, we will tend to refer to connected households that resell as household resellers. 15. The decisioltmaking unit is the household, and the demand equations are thus estimated at the household level rather than on a per-person basis. Household sixe is included as a control variable. Still, the composition of the family clearly matters, with respect to both the demands for water by people of different ages and the availability of person at different ages to collect water. Alternative specifications of the family configuration did not significantly affect the price and time coefficients, however. 16. See the studies cited by Briscoe, et al. (1990), Bahl and Linn (1992), and Whittington and Choe (1992). 17. The survey took place during the dry season. Respondents were also asked about their consumption by scurce during the rainy season, during which time the proportion of households using rain water for drinking and cooking was about 10 times higher and the proportion using rain water for bathing wss about twice as high, on average, as during the dry season. Use of the other scurces did not vary much from one season to another. 18. Pindyck and Rubinfeld (1992) contains a basic discussion of this apl~oach to the measurement of consumer welfare. 19. Whittington. Mu and Roche (1990) is a recent example of an effort to calculate the value of time devoted to water collection.

8. The exchange rate in July 1991 was approximately 1,925 Rupiah to each US dollar. The rate structure was revised upward in late 1991. 9. The intent of an increasing block tariff is to both encourage conservation and provide a cross-subsidy to lowconsumption users. It also severely restricts the role private homes can play as indirect suppliers of water to the pocr, however, as the marginal supply price of water will rise significantly above that of other supplies with the amcunt sold. In this case, purchasers of household supplied water would cross-subsidize connected households that do not sell water, which in most situations are wealthier. Whittington (1992) discusses these issues in detail. 10. A case study of the Nortb Jakarta hydrant develop ment program in late 1989 by Yayasan Dian Desa (1990) also addressed many of these issues. 11. The average monthly net revenue of North Jakarta vendors in the survey results of Crane (1992) was Rp 230,000 in 1991, less any cart rental costs. lhis represents an average hourly wage of Rp 900 for the sample, which Lovei and Whittington (1993) indicate is two to three times the relevant average wage. 12. PAM Jaya Director General 030/INSTRIPAM/IV/l990, April 21990. Announcement

13. The questionnaire requested expenditure data on individual categories of goods (e.g., food, electricity, housing,) and the monthly total, which allowed for cross-checking

REFERENCES Altaf, Mir Anjum, Haroon Jamal, Jin Long Liu, V. Kerry Smith and Dale Whittington, Prices and connection decisions for public water systems in developing countries: A case study of the Punjab, Pakistan, unpublished (June 1990). Bahl, Roy W. and Johannes F. Linn, Urban Public Finance in Developing Countries (New York: Gxford University Press fcr The World Bank, 1992). Berry, Brian J.L. Clean water fa all: Equity-based urban water supply alternatives for Indonesias cities, Urbmr Geography, Vol. 3 (1982), pp. 281-299. Berry, Brian J. L., and Katherine Sierra, Public works investment strategy in a developing country: Urban water supply in Indonesia, Urban Planning, Policy Analysis and Administration Reseerch Report R78-3, Department of City and Regional Planning(Cambridge, MA: Harvard University, 1978). Briscoe, J., P. Furtado de Castro, C. Griffin, J. North and 0. Olsen, Toward equitable and mstab&le rural water sup plies: A contingent valuation study in Brazil. The World EenkEccnomic Review4 (1990), pp. 115-134. Briscoe, John P., and Daniel deFerranti, Water for Ruml Helping People Help Themselves conrnrwliti~: (Washington, DC: The World Bank, 1988). Buro Pusat Statist& (Indonesian Central Bureau of Statistics) Jakarta D&m Angka(Jakarta, Indonesia: 1990). Crane, Randall, lhe impact of water supply deregulation in Jakarta, Indonesia: Results from household and water vendor surveys in North Jakarta, Final Report, Infrastructure Division (Washington, DC The World Bank, April 1992). Fass, Simon M., Politkal Economy in Haiti: the Drama of Survival (New Brunswick, NJ: Transaction Books, 1988). Gertler, Paul, and Jacques van der Gaag, Measuring the willingness to pay for social savices in developing countries, Living Standards Measurement Study Working Paper No. 45 (Washington, DC: The World Bank, 1988). and David Hardoy, Jorge E., Sandy Cairncross Sattherthwaite (Eds.), lke Poor Die Young: Housing and Health in i%ird World Countries (London: Earthscan Ress, 1990).

82

WORLD DE7IELOFWENT

Jellinek, Lea, The Wheel ofFortune: The History #a Poor Community in Jaharka (Honolulu: University of Hawaii Press, 1991). Kasarda, John D., and Allan M. Pamell (Eds.), Third World Cities: Problems, Policies and Prospects (Newbury Park, CA: Sage Publications, 1993). L.inn, Johannes, Cities in the Developing World (New Yak: Oxford University Press, 1983). Lovei, Lasxlo, and Dale Whittington, Rent-extracting behavia by multiple agents in the provision of municipal water supply: A-study of Jakarta, Indonesia, W&r Resources Research 29 (July 1993). w. 1965-1974. Marcussen, Lars, i7rird WorldHousing~~ Social and Qnztial Development: The Case of Jakarta (Aldershot, UK: Avcbury, 1990). Mu, Kinming, Dale Whittington and John Briscoe, Modeling village water demand behavior: A discrete choice approach, Water Resources Rcpurrch 26 (1990), pp. 521-529. Munashinghe, Mohan, Water supply policies and issues in developing countries, Natural Resources Forum (February 1990), pp. 3348. Okun, Daniel A., A water and sanitation strategy for the developing world, Environment 33, No. 8, (October 1991), pp. 16-43. Okun, Daniel A., and Donald T. Lauris, Capacity building for water resources management: An international initiative for the 199oS, Division for Global and Interregional Programmes (New York: United Nations Development prOgramme, August 1991). Perusahaan Air Minum Jaya, Various unpublished data, Jakarta, Indonesia (1991). Pindyck, Robert, and Daniel Rubinfeld, Microeconomics, Second E&ion (New York: Macmillan, 1992). Saunders, Robert J., and Jeremy J. Wsrford, Vi&age Water Supply (Baltimore: Johns Hopkins University Press, for the World Bank, 1976). Schteingart, Martha (Ed.), Lar Cimbaaks La&oamericanas en las Crisis: Probkmas y Desajios (Mexico City: Trillas Press, 1989). Shugart, Chris, An exploratory study of the water standpipe-vendor system in Jakarta, Working paper (Cambridge, MA: Harvard Institute for International Development, 1991).

Struyk, Raymond J., Michael L. Hoffman and Harold M. Katsura, The Ma&r for Shelter in Indonesian Cities (Washington, DC The Urban Institute and Hasfarm Dian Komulta.n, 1990). Trava Manrani lla, I., J. Roman Calleros and F. Bemal RodrIguex, Manejo AmbiuatuJmente Adecuack de1 Agua en la Frontem MLxico - Eskados Vnidos: Situation Actuul y Pempectivus (Tijuana, Mexico: El Colegio de la Frontera Norm, 1991). Whittington, Dale, Possible adverse effects of increasing block water tariffs in developing countries, Economic Development and Cultural Change, Vol. 41 (October 1992), pp. 75-87. Whittington, Dale, and Kyeongac Choe, Economic benefits available from the p&is&n of improved potable water supplies, WASH Technical Report No. 77 (1992). Whittington, Dale, Donald T. Lauria and Kinming Mu, A study of water vending and willingness to pay for water in Onitsha, Nine& WoridDevelomnent. Vol. 19. No. 2l3 (1991),pp. i79-i98. . Whittington, Dale, Kinming Mu, and R. Roche, Calculating the value of time spent collecting water: Some estimates for Uktmda, Kenya, World Development, Vol. 18, No. 2 @?bNWJ lgi)cl), pp. 269-280. Whiaington, Dale, Donald T. Lauria, Daniel A. Okun and Kinming Mu, Water vending activities in developing countries: A case study of Ukunda, Kenya, Intemati~~& Journal of Water Resource Develomnent, Vol. 5 (1989). . _. pp. 158-l-68. Wood, Geoffrey D., and Richard Palmer-Jones, The Water Sellers (West Hartford, CN: Kumarian Press, 1990). World Bank, l7re World Bank Development Report 1992 (New York: Oxford Universitv Press for the World Bank. i992). World Bank Water Demand Research Team, Ibe demand for water in rural areas: Determinants and policy implications, The World Barth Research Observer, Vol. 8 (1993). pp. 47-70. Yayasan Dian Desa, Madtoting cmd Evaluation of Public Hydrants & Water Terminals in North Jakarta: Final Report (Yogyakarta, Indonesia: 1990). Zaroff, Barbara, and Daniel A. Okun, Water vending in developing countries, Aqua, Vol. 5 (1984), pp. 289-295.

APPENDIX: THE DETERMINANTS

OF HYDR ANT AND VENDOR SUPPLIED WATER USE errors. (Crane (1992) discusses related estimation and speciflcation issues applied to this model and these data.) The Bs are the parameters to be c&nated. Of psrt.ia~lar interest are 6, and & which may be interpreted as the price and income elasticities of demand, respectively. This model was edmated by ordinary least squares methods separately for customers of vended water and of hydrant water, and the results sre reported in Tables Al and A2. Note that several statistical issues are left unresolved in this approach. It would be useful to include the price of water from other sources in the estimated equation, but these are not available except in those relatively few cases where households purchase from more than one source. We do not observe the vendor price available to a household who chooses not to buy vended water, for example.

To measure the sensitivity of consumption to household and market variables, particularly prices and water transport requirements, a simple behavioral model of water demand was employed. The general specification has the form: log q@= $ + 6,log pe + s, log y,+ l3&+ B& gszj + Q.

where qti is the quantity of water purchased from source i by household j, pu is the per unit price of water from source i faced by household j, yj is the income of household j, rt+is the time spent per week on collecting water from source i, dp is the distance to source i, 4 is a vector of household and community characteristics expected to influence demand, and E is the errot term capturing data reporting and other

WATER MARKEZS AND MARRET REFGRM Table Al. DeterminrmtJ ofvoldcdwater use Dependentvariable: Log of cubic meters of water plrchased from vendors per month Independent variables E&mated coefficient -0.48 0.03 -0.03 -O.ooOl -0.012 -0.004 0.06 0.01 0.01 -0.02 0.0004 -0.19 0.001 -0.37 standard error 0.141+ 0.190 0.014t 0.001 0.193 0.003 0.038 0.010 0.007 0.172 0.001 0.148 0.003t 0.226 Table A2. Detenninmrts of hydrant water use Dependent variable: Log of cubic meters of water purchased from hydrants per month Independent variables Price per ms Oog) Household monthly income (log) Tie necessary for each purchase Distance to nearest hydrant (log) Purchase from vendor Time to commute (minutes) Household size Age cf head of household Years in house Tenant Floor area of house (in meters) Share toilet? (yes = 1) Reservoir capacity Rural location 0.763 R* 43 N *significant at 1% ~significant at 5% Ewimamd coefficient -0.60 0.12 -0.002 0.001 -1.44 -0.01 0.08 -0.04 -0.005 0.42 o.ooo2 -0.1 -O.).oooool -0.482

83

Standard crrac 0.265t 0.362 0.01 0.002 0.439* 0.011 0.065 0.016t 0.013 0.341 0.005 0.386 o.OOo&? 0.530

Price per m3 (log) Household monthly income 008) Time spent coRectin8 W&X (log) Distance to nearest hydrant (log) Purchase from hydrant Time to commute to work (minmes) Household sire Age of head of household Years in house Tenant Floa area of house (in meters) Share toilet? (yes = 1) Reservoir capacity Located in rural area R2 0.572 N 101 *sign&ant at 1% tsigniticant at 5%

Simultaneity issues may also be important, as in the endogeneity of reservoir sire. Given the length of tenure in this sample, this may not be a substantial problem The reservoir investment decision was certainly codetermined with the expected need for water, but in many instances that wss some years ago. Fcr all practical puposes. the size of the water storage facility can be considered exogenous to any

one weeks water purchase decision. This is less true of households who have recently acquired their storage tank, and the issue should be carefully addressed in future research. lbe codetermination of source and quantity purchased is a separate and somewhat complex issue not addressed here. See Mu, Whittington and Briscce (1990) and Altaf, et al. (1990) for empirical strategies.

S-ar putea să vă placă și