SWP560 Progressive Development and Affordability in the Design of Urban Shelter Projects Douglas H. Keare Emmanuel Jimenez WORLD BANK STAFF WORKING PAPERS Number 560 WORLD BANK STAFF WORKING PAPERS K Number 560 - ^ & 0 Progressive Development and Affordability t in the Design of Urban Shelter Projects Douglas H. Keare Emmanuel Jimenez ,~~~- 2r°1 The World Bank Washington, D.C., U.S.A. Copyright © 1983 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, N.W. Washington, D.C 20433, U.S.A. First printing May 1983 All rights reserved Manufactured in the United States of America This is a working document published informally by the World Bank. To present the results of research with the least possible delay, the typescript has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. The publication is supplied at a token charge to defray part of the cost of manufacture and distribution. The views and interpretations in this document are those of the author(s) and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting on their behalf. Any maps used have been prepared solely for the convenience of the readers; the denominations used and the boundaries shown do not imply, on the part of the World Bank and its affiliates, any judgment on the legal status of any territory or any endorsement or acceptance of such boundaries. The full range of World Bank publications is described in the Catalog of World Bank Publications; the continuing research program of the Bank is outlined in World Bank Research Program: Abstracts of Current Studies. Both booklets are updated annually, the most recent edition of each is available without charge from the Publications Distribution Unit of the Bank in Washington or from the European Office of the Bank, 66, avenue d'Iena, 75116 Pans, France. Douglas H. Keare is urban adviser in the Urban Development Department of the World Bank, Emmanuel Jimenez is assistant professor of economics at the University of Western Ontario and a Consultant to the World Bank. Library of Congress Cataloging in PublEcation Data Keare, Douglas H., 1935- Progressive development and affordability in the design of urban shelter projects. (World Bank staff working papers ; 560) lo Underdeveloped areas--Public housing--Design and construction--Economic aspects. 2. Underdeveloped areas-- Rent subsidies. 3. Underdeveloped areas--Self-help housing. 4. Underdeveloped areas--Poor. 5. Under- developed areas--Income. I. Jimenez, Emmanuel, 1952- . II. Title. III, Series. HD7288.77.K4 1983 363.5'8 83-3561 ISBN 0-8213-0166-7 Abstract The progressive development approach to house construction by and for the urban poor in sites and services and area upgrading schemes relies on affordable designs, tailored to meet individual families' needs. This feature was one of the main areas of investigation in a pilot program of evaluation of four of the early World Bank-supported projects in El Salvador, the Philippines, Senegal, and Zambia. The primary objective of this paper is to examine the concept and determination of affordability as a design factor. It begins by a discussion of the conceptual and measurement considerations which underlie affordability calculations; and then uses data generated by the evaluation program in El Salvador to conduct an empirical exercise to measure the determinants of willingness to pay. A principal finding of the investigation is that it matters greatly what definition of income one uses in making affordability calculations. Total family income, including "unearned" income, appears to be the most reliable indicator of families' willingness to pay for housing; and permanent income measures, or proxies, should be used where possible. Appropriately calculated measures of the propensity to consume housing indicate that owners spend a considerably greater proportion of their incomes on housing than do renters. This proportion also varies with other factors, such as certain household characteristics and city size. One standard feature, however, is that the proportion declines with income, with an income elasticity that remains within quite a narrow range across countries. Acknowledgements The authors are most grateful to their colleagues Michael Bamberger, Dani Kaufmann and David Lindauer for their very considerable contributions to the overall program of which this paper is but a small part; to participants in the November 1980 Conference on Evaluation held at Bank headquarters, at which the first draft of this paper was presented and discussed; and to Michael Cohen, Harold Dunkerley and Alberto Harth for particularly helpful comments on the subsequent draft. The authors alone, however, are responsible for any reraining errors. TABLE OF CONTENTS Page Nos. I. Introduction .............................. . I II. Conceptual Considerations . ............................. 6 III. Measurement Considerations..................... ... .12 A. The Measurement of Costs .........................13 B. The Measurement of "Willingness to Pay ...........15 Income in the Measurement of "a .............. ...-19 Housing Expenditures in the Measurement of "a ....27 IV. Housing Consumption in El Savador: An Empirical Exercise ............... ......----------29 A. Estimates of the Propensity to Consume Housing... .31 B. The Determinants of Housing Consumption . . 45 V. Principal Conclusions And Recommendations ............. 55 PROGRESSIVE DEVELOPMENT AND AFFORDABILITY IN THE DESIGN OF URBAN SHELTER PROJECTS by Douglas H. Keare and Emmanuel Jimenez I. INTRODUCTION The Progressive Development Model: When, in the early 1970s, the World Bank began to commit considerable resources to the development of urban housing in Third World countries, the resulting lending program was based on the proposition that sites and services and squatter area upgrading projects, -/ providing secure tenure and a range of basic services, 2/ would enable and encourage low-income households to improve their housing through self-help financing and/or construction. Improvements in living environments were expected to lead to growth in productivity and incomes over time. Both project types were characterized as "progressive development," implying that improvements would be made at a pace dependent on the incomes and preferences of project households. 1/ Sites and services development consists of the provision of new serviced sites which selected applicants buy or lease. Upgrading involves improvements to existing areas, although some families may have to be resettled to neighboring serviced sites to make room for essential infrastructure. 2/ Including, for example, water supply, sewerage, storm water drainage, footpaths, streets (with various surfaces), optional electricity and some core construction. Progressive development can be further defined formally as a method of housing construction or upgrading achieved through: (a) staged development, in which the i'nfrastructure cid occasionally part of the house are built by a contrractor and the rest of the shelter is completed by the household; (b) flexibility in housing design, construction time and materials used; and (c) self-help components, which can be organized in the following ways: (i) mutual help, in which families work together in groups, often with supervision from project management; (ii) self-help construction in wihich the household hires a contractor to build the shelter; (iii) self-help construction in which the household hires and supervises individual laborers; (iv) self-help construction in which the household uses its own labor to build its house. The progressive development approach assumes that it ls better to take advantage of existing housing ctock within a city, whatever its condition, than to demolish and replece it;, and better also to provlde new housing which is more nearly tailored to the needs o£ poor families than that provided through most othez publfc sector programs. Upgrading schemes provide essential infrastructure and secure tenure to dwellers in existing urban slums, as well as extending credit, materials and - 3 - technical assistance for the improvement of their own dwellings. Sites and services schemes provide contractor-built utility infrastructure and in most cases core units as well, leaving project participants in a position to complete their own new dwellings, again with provisions for materials loans and technical assistance. Both progressive development strategies attempt to make project designs as flexible as possible, giving participant families as broad a range of choice as possible of the timing and methods of construction, as well as of materials and sources of labor. A key design feature of both upgrading and sites-and-services projects has been provision of opportunities for participants to use family labor in self-help and mutual-help construction processes. The flexibility built into such programs was intended to enable families to adjust the pace of house construction to their frequently fluctuating incomes, and to use their own labor (or exchange labor with others) to reduce "out-of- pocket" expenditures. All these features of progressive development practice were intended to make improved housing accessible to poor families, and even to affect the quality of social existence within low income areas. It was expected that the encouragement of self-help and mutual-help house construction would also promote a sense of civic responsibility and community solidarity previously lacking in many impoverished neighborhoods. Some planners and policy makers remain to be convinced, and the inevitable problems of project implementation and the occasional "problem projects" fuel these doubts; nevertheless, the results of these efforts to develop means of improving the quality of urban life on a mass scale have so far been very encouraging0 The successes have been both numerous and dramatic enough to cause project administrators and designers to believe that the progressive development model has proved a valid strategy for coping with ever-burgeoning urban populations in the Third World. Designing Affordable Projects: The principles followed in designing urban projects financed by the World Bank center on the practice of delivering "packages" of tenure, services and assistance in house construction, either in new subdivisions (sites and services) or in existing squatter areas (upgrading)0 The design of these projects seeks to ensure that beneficiaries are neither prevented from spending what they want and are able to afford on housing, nor induced to spend beyond their capacity and willingness0 These considerations define the concern for affordability. Closely related to the issue of afford- ability are those of accessibility and replicability. Accessibility is achieved when families within the prescribed range of incomes actually gain entrance to the projecto Replicability refers to the goal of having project costs paid for by the beneficiaries, with little or no subsidy0 This is necessary for such programs to be extended on a much wider scale, consonant with needs0 without causing unsustainable burdens on municipal or national budgets0 The aim of improving project design is to have these three objectives converge0 even though there is no assurance that they will0 If a project is not affordable0 it can either be replicable (by the entry of higher-income families) or accessible (through subsidy of the targeted families), but not both0 On the other hand, an affordable - 5 - project may not be replicable (if the cost recovery mechanism fails) or accessible (if, for example, selection procedures favor higher-income families over members of the target population). In any case, afford- ability remains a central issue in project design. Objectives of the Paper: The primary objective of this paper is to examine the concept and determination of affordability as a design factor in urban housing projects utilizing progressive development approaches. It begins by a discussion of the conceptual and measurement considerations which underlie many affordability calculations. It then uses data generated by the evaluation program in El Salvador -L to conduct an empirical exercise to measure the determinants of willingness to pay. It aims, thereby, to provide improved guidelines on factors to consider in making affordability calculations, given existing field conditions. The paper's message is intended for two audiences: for project and program managers seeking to make the best use of available data; and for national statistical offices (which should be) seeking to improve the utility of data bases for such purposes over time. 1/ This data base is one of several generated during a five-year pilot effort to rigorously evaluate the execution and effects of early urban development projects in El Salvador, the Philippines, Senegal, and Zambia. See Keare, Douglas H. and Scott Parris, "Evaluation of Shelter Programs for the Urban Poor: Principal Find'ngs", World Bank Staff Working Paper No. 547, 1982, for a more detailed discussion of the methods and particularly the results of the evaluation. This program was jointly financed by the International Development Research Centre (Canada) and the Bank. II. CONCEPTUAL CONSIDERATIONS For more than a decade, the World Bank has been trying '.o assist agencies in its member countries in designing and implementing wiorkable solutions to urban shelter problems. The worilng defini2ion of affordability used in appraisal reports has been as follows: a certain level of urban services is affordable to a low-income beneficiary household if the amount the household is uilling and able to pay for shelter-related expenses is sufficient to cover the monthly cost of services. The definition may be written symbolically0 If "a" represents the average propensity to consume housing (the proport.on of monthly income Y a household is wilLing to spend on it), then a unit of project housing with service level "j" and monthly costs "C " is estimated to be affordable down the "i"th percentile of the income distribution if the following is true: aYi > C c As an example, consider a project whose costs are such that, if it were to pay for itself, beneficiaries must be charged $100 per month. If all households are assumed to be willing and able to devote 20 percent of their incomes to housing (a 20), then a household must 1/ In the estimation of "a" it is typically necessary to rely on measures of monthly rent, whether actual or implicit. When using the resulting estimates of "a" to assess affordability, one is concerned with design costs0 To fill the bill, the cost concept used must be (fully amortized) occupancy costs0 Occupancy costs are defined as the total costs (capital charges, taxes, operational and maintenance costs, utilities charges, etc) of occupylng the housing unit0 "Fully amortized" means that, using credit tc the extent necessary, costs are conceived as a steady monthly charge over a very extended period (2030 years). earn at least $500 per month to afford project housing. If the income distribution is as follows: Percentile of Population Y = Monthly Income 14 Y1L $490 15 Y15 $500 16 Y16 $508 then the project is affordable down to the 15th (income) percentile of the population. This example, however, is complicated by the fact that the value of "a" is unlikely to be constant. The proportion of their incomes that households are willing and able to spend may vary in a distribuition with mean ai and standard deviation ai for the "i"th income percentile. This is because "a" is a function not only of income, but of other variables as well, such as relative prices, family size, sex of household head, assets and certain characteristics of the extended family. In addition "a" may be different for renters (aR) versus owners (a0). This type of relationship is depicted in Figure 1, where the solid lines represent loci of means, and the broken ones correspond to a general notion of the proportion of the population that housing policy should try to reach. Thus, in this illustration, the "a" that is chosen (representing either one of the dotted lines) would be set such that no more than say 5 to 15 percent of families would be found in the shaded areas. Conceptually this means that 85 to 95 AVERAGE PROPENSITY TO CONSUME HOUSING OWNERS -o -- -…- - - - - - -- - - - - . RENTERS -R 0 Yo INCOME Wold Ban!-24785 - 9 - percent of families would choose to join and stay in the projects, their own "a"s being greater than those assumed in the design. The Bank's interest in affordability has hinged upon avoiding projects that are too expensive for target groups, thus emphasizing keeping standards and costs down. This explains why the affordability criterion says that: cost must be equal to or less than the affordable proportion of income, and why programs of the Bank have stressed the harmful effects of designing projects that target groups cannot afford. In essence, housing programs and projects with distributional objectives such as the above should reach as far down as possible in the income distribution. They should also be replicable, which requires good cost recovery performance. For this reason there are admissions criteria in projects financed by the Bank, the objectives being to admit "successes" and exclude "failures." However, this is not a simple task. There are possibilities of erring in several ways in these projects, for which lower and upper income bounds usually are set: If the lower bound is set too low, one may admit households that fail -- at high costs to individual projects and the perceived credibility of the approach. If the lower bound is set too high, one will exclude poor households that would have succeeded, and this will compromise distributional objectives. Project designers may sometimes be tempted to set the upper bound too high, so as to ensure adequate demand for project participation. On the other hand, it would probably not be desirable to set this ceiling so low as to totally forego diversity within the project population, or the attendant opportunities for cross- subsidy and market interactions. Although the principles are reasonably - 10 - subsidy and market interactions. Although the principles are reasonably clear, as a practical matter, information on which to base efficient decisions about these cut-off levels is thus fag in short supply. At both ends of the income spectrum, but particularly at the lower end, it is necessary to constantly improve the affordability criteria on which judgments are based. Findings to date indicate that to reach the lowest income groups, total income must be carefully estimated and the costs of a project mandated by the public sector kept to a minimum. For this reason progressive development projects are designed to provide large numbers of relatively lolf quality sites, possibly including core units, rather than a few high quality houses. These steps insure against screening out large numbers of families who1, though poor, can afford to improve their housing at least to the modest degree implied by the lower standards. At the same time1, project designers must be careful not to constrain unduly the behavior of households that have both the desire and capacity to do more than the minimum. As we wIllR document in later sections of this paper, the evidence indicates that participantc' behavior varies greatly, due to the interaction of a number of forces that sometimes conflict with and sometimes reinforce each other0 These include variable incomes, variable investment plans and variable timing of house construction and investment0 The range of outcomes is probably accentuated in newly settled sites-and-services areas1, where one can see shacks juxtaposed with completed family dwellings1, and large hcuses containing rental accommodations for several people0 Even though the extent of this diversity is likely to be reduced in the long run, it is - 11 - almost certainly desirable to encourage diversity it in the short run; for it can be a means of moving projects toward the fulfillment of their major objectives. This approach calls for considerable flexibility in credit programs and other project components so that they can address a wide range of the population's needs and capacities. - 12 III. MEASUREMENT CONSIDERATIONS Calculations on affordability depend upon the way in twhich each of the components of the inequality is measuredo The equation can also be written as a > C /Yi An ideal data base would enable an analyst seeking to develop an estimate of "a" for a particular target group in a given society to simply calculate "a" using the present incomes and housing costs actually incurred by the target population, or some sub-sample oL it. However, data bases even approaching the ideal are very rarely encountered in reality. For the sub-populations in questions, it is frequently the case that, if suitable cost estimates exist, income estimates either do not exist or are for the population as a whole, or for a different sub-sample. On the other hand, if the income data exist, suitable cost estimates may be missing. Not infrequently both kinds of information are missingo In addition, the target population could be a squatter community, or squatters generally; in which case estimating affordability based on what they spend currently might be misleading.l/ In contemplating a first project in such circumstances, one is typically forced to assume an "a" based on a wider sample, or to use an "a" calculated from another sample judged to be similar or 1/ In most cases it would be misleading. Because of the risk (of eviction without compensation) involved in squatting, ratioral squatters will invest less in their housing than if they had absolutely secure tenure to the same parcel of land0 - 13 - comparable. Then the task of designing the project is one of applying this figure judiciously to the incomes obtained or estimated for the target population. This paper seeks to take advantage of a data base generated by the aforementioned evaluation program to derive more reliable estimates of "a" than are generally possible in field circumstances. - While it focuses primarily on the left-hand side of this inequality, a few comments first need to be made about the measurement of Cj. A. The Measurement of Costs The crucial question that needs to be addressed is what definition of costs to use. A number of alternatives are possible. As outlined in Figure 2, the direct project costs (Cl) provide only a partial view. Adding the costs (C2) which households voluntarily incur for materials and labor to complete their houses, is conceptually straightforward but difficult to execute because costs shift both with varying levels of expenditure and a broad range of time periods over which they may be extended. In designing a project, one must estimate or guess what these patterns will be. In the earliest projects, and to some extent still in first projects anywhere, there is little information on which even to base a guess. As experience increases, however, particularly if it is systematically exploited through 1/ Being designed specifically to this purpose (the assessment of affordability factors) among others, the data generated by the evaluation coordinate information on income and costs for the same sample, are more carefully generated (and therefore of higher quality or reliability) than most data and are generated for more than one point in time. 14 - FIGURE 2: THE DESIGN OF PROJECT COSTS c4 C5 - ~~~~~~C1,C3 where C1 - Project Costs (payment for land, cervicee, core houSe, etc., not including credit components)O C2 = Added Materials and Labor Costs (still no credit components). C3 Total Costs (Cl + C2) C4 = Amortized Costs ("True" C, with "ideal" credit system fully amortizing total costs to a constant monthly payment). C5 m Project Costs (Actual costs , with actual credit scheme that does not fully amortize all costs). 1/ 1/- The difference between C5 and CA may be thought of as due to "inadequacies" in design. - 15 - evaluation programs, the basis for sound estimates grows. Ideally the concept used should be that of fully amortized occupancy costs, or C4 in Figure 2. Project designers may, however, be fully excused for using the "blemished" C5, as this again reflects the real world in which they must operate; what matters is how close the project -- with its imperfect credit mechanism(s) -- comes to rendering its offering affordable. Unnecessary confusion is, however, caused by using more limited (and misleading) concepts such as C1, C2 or some mixture of them. The other principal consideration, when one is attempting to calculate the ratio of costs to income, is choosing what income definition to use. The crucial point here is that whoever is doing the affordability calculation should be aware what components of income (as well as costs) went into the estimate of "a" which is being compared with the estimates of C/Y. This is examined in further detail below, in the discussion of the definition of Y. B. The Measurement of "Willingness to Pay" We now turn to the left-hand side of the affordability relationship. Until the evaluation and research programs designed to inform the World Bank's policy decisions, 1/ as well as those of some of I/ The Bank's formal research program began to achieve significant dimensions by the late 1960s. Research on urban and regional economics has constituted a part of the larger program since 1971. Research on affordability in housing has been a part of this program since 1972/3. It has been carried out both under formal research projects and under the evaluation program which gives rise to this paper. There is a variety of publications available on urban development. For details see The World Bank Catalog of Publications 1983 and The World Bank Research News both available free of charge. - 16 - its client agencies,D / had been in operation long enough to genergte appropriate data bases, rigorous estimates for "a" virtually did not exist. Policy formulation and project design necessarily relied upon a very few, opportunistically located data sets, and arbitrary judgments concerning the stability of relevant parameters across geographical boundaries and income levels initially had to be made in the interest of getting the learning by doing process underway0 Grimes was one of the first to collect information for devel- oping countries. -/ Using aggregate data, he concluded that: the average percentage of household expenditure devoted to housing 000 falls into a fairly narrow range (across cities), from 11.7 percent in Kingston to about 20 percent in Mexico City and Seoul03/ However, in looking at some rough and strictly non-comparable government micro-level data for five countries, Grimes also found that: The fragmentary data available on expenditure by various income groups 00. indicate a general tendency for housing expenditure by low-income households to claim a higher than average share 0o0 As income rises and other demands are met, the share of income devoted to housing may remain constant or fallo 4/ By the second year of the evaluation program it had been determined not only that the progressive development model was viable (in the sense that housing was being completed and occupied, or up- graded, according to the case), but also that the "first generation" 1/ The Fundacion Salvadorena de Desarzollo y Vivienda Minima, or FSDVM, on whose project and data this paper is based, initiated its own evaluation program before the Bank did0 2/ Grimes, 00 F0 (1976)o HousLna for Lo^y-Income Urban Families: Economics and Policy in the Developing World0 Baltimore: Johns Hopkins, p° 65o 3/ Ibid., p° 65. 4/ Ibid., pp. 65-67. - 17 - projects being supported by the Bank were also affordable and consistent with Grimes' findings. Based on initial, detailed evaluation of one sub-project in the city of Santa Ana, the larger El Salvador project appeared to be affordable down to about the 20th percentile of the income distribution. -L Despite measurement difficulties, this result seemed to be borne out as well by the evaluations in Senegal and Zambia, and to be broadly confirmed by other project experiences up to that time. These generally encouraging results were subject to more than one interpretation among the project managers, researchers, and institutions involved. For example, the precise results in the case of El Salvador showed that, assuming an average propensity to consume housing of 20 percent, direct project costs (C1 of Figure 2) were affordable to those with incomes down to the 17th percentile. However, if crude estimates of additional private costs were included, it appeared that total costs (C1 + C2, partly "amortized to approximate C4) might be affordable only down to the 25th percentile of the income distribution. Such calculations suggested that assessing affordability was a matter of greater complexity than had been assumed in the original project design and appraisal. More important, they illustrated the importance, that has been stressed here, of being clear about the cost and income definitions used and of obtaining direct estimates for "a". 1/ Bamberger, Michael, Edgardo Gonzalez-Polio, and Umnuay Sae-Hau, "Evaluation of Sites and Services Projects: The Evidence from El Salvador," World Bank Staff Working Paper No. 549, 1982. - 18 - By the time an adequate data base hnad been accumulated, it was found that the implied values of a" assumed in the first generat:.on of urban projects financed by the Bank had been distributed about the upper end of the narrow range described by Grimes. An internal analye:^. of 36 projects approved from 1972 through 1979 by the Executive Directors of the World Bank recorded extreme values for "a" of 8 and 50 percent. However, the bulk of observations wazs approximately normally distributed about a mean of 20 percent, within a range cf 12 to 33 percent. This reflected a number of differing circumstances encountered in the field -- ranging from differences in available data to differences in what it was hoped projects would achieve. What is most striking about these figures, however, is the large variation in a parameter previously considered by most to be quite stable. And, although obtaining more accurate estimates of "a" was only one of the several complementary goals of the analysis of affordability under the evaluation program, several questions regarding this parameter must be answered in order to gauge its value as an operational tool. Is the value of "a" constant (or close to it) across various income groups of the target population? In most of these early projects, Bank staff assumed that "a" was fixed across all income levels. In some projects (Ivory Coast, El Salvador, Thailand, Botswana, and Brazil) it was tacitly assumed that "a" rises with income. The first project Li Bolivia apparently assumed the opposite tendency. The divergent. assumptions naturally lead to the next two questions. If "a" is not constant, how does it vary with income? How sensitive is "a" to the definitions of housing expenditures and income? While we w^ill attempt - 19 - to answer these questions in Section IV, in the rest of this section we will focus our attention on measurement issues. In the estimation of "a" we must typically rely on measures of income (Y) and actual housing expenditures or rents (R), so that a = R/Y . This can be viewed as the average propensity to consume housing services over a given period. The problems in estimating this value are complicated by the non-homogenous nature of housing, by the uncertainty and variability of personal income in poor countries, and by the heterogeneity of the population itself. 1/ The problems dealt with here concern definition and measurement of, first, the income (Y) and, then, housing expenditures (R) of a prescribed segment of a market or population. Income in the Measurement of "a": To minimize the risk of measurement errors, ambiguities in definition must be removed; the ambiguities which are most pronounced are those involved in determining whose incomes and what sources to include. In early project design as well as research efforts, a developed-country bias crept in, resulting in a tendency to think only, or primarily, of the household head's wage income, since that was judged to be the most stable component of total household income. A number of studies have since led unambiguously to the conclusion that this assumption brings about serious underestimations and distortions of hcusehold incomes. 1/ This is important, because in project design it is frequently necessary to employ estimates of "a" which come from sub- populations different from those to be "treated" in the project. - 20 - Programs too have admitted a wide variety of defiuiitions of income. In some, such as the Zambia project, income comprises solely the earned (labor) income of the household head; in others, the earned (labor) incomes of all family members have been included; and still others have counted the family's total income from all sources. The distinction is important, because total income can be twice as great on average as the labor income of the household head. Y Income, moreover, fluctuates significantly during the year and over the family life cycle, and these fluctuations influence the perceptions families hold regarding their access to resources for housing investment. There is also the problem of measurement accuracyo The evaluation program has established even firmer evidence than had been collected before that, in addition to the sources of error typically associated with income estimates from socioeconomic surveys, some families purposely distort reported incomeo Among the many potential reasons for this are some directly associated with projects; for example, some families may claim that their incomes are less than they really are so as not to be considered too well off to join the project, 1/ This variation in definitions implies that estimates of "a" can differ by at least a factor of 2, based on chis factor alone. On this basis alone, there would seem to be a strong case for consistency in the way incomes are measured or assessed, as in defining and measuring occupancy costs and in most other aopects of this work0 This, of course is not always possible, given variations in data and other circumstances. Still, it Ind:icates that -- in the absence of data we want -it may be better to estimate data according to consistent principles rather tPan to accept actual data gathered according to inconsistent ones0 - 21 - while others may claim a higher income in order not to be considered financial risks. Obviously, program affordability definitions, based as they have been in the past on particular sources and definitions of income, have affected the degree to which low income households can participate in projects. With respect to income, exclusion can occur because projects' regulations may require participants to have certain verifiable kinds of employment, as in the formal sector alone; may consider only the earnings of certain household members, such as the head; or, indeed, may limit the income definition to earnings. As an example of an important type of exclusion in many projects up to now, let us consider transfer income. Evidence from the evaluations in El Salvador, the Philippines, and Senegal, as well as from a related investigation in Cartagena, Colombia, sh_ws tna. Inter- household transfer payments within the extendŽd family constitute an important source of income, and this source is most important for the lowest income families. In Santa Ana, El Salvador, for example, recurrent transfers have been found to augment the incomes of roughly one-third of the families living within the "informal" housing sector, including 58 percent of those in the lowest and 48 percent of those in the next lowest deciles; 1/ and transfers in that city represent a stable source of approximately 40 percent of household income for these 1/ The average monthly amount transferred is 86 colones (US$34.40) per recipient. - 22 - these two deciles, as opposed to 10 prTeint Cor t-he t:oftal oamp2lo Similar patterns appear to hold among ,he proj.t.,. arza S in the Tondo area of Manila, the Philipp neo, an.d fSn the South Orient zone in Cartagena, Colombia. In Tondo, 43 perens or ! Zs a&.ieo in the lowest third of the income distribution, as 0..cc" to 21 percent of the . receive transfers accounting for over haP? thea: ..-comes. In the Cartagena project area the correspond-.ir. f2ges are 50 percent, '2 percent and 41 percent. Total income, e'oFrD rther than the w-ages of primary or secondary earEeners, D to 'he a z morL appropriate measure of the sum of a household's di0.ponable reoourceo aerd inhereent purchasing power. There id still considerable rcsitzncs to th-1s last idea, however, and most affordabiRity and demLas 3t:½:ate to date have been made on the basis of earned income sou.ces onLyo Thiae disputed pTectice persists partly as a matter of defPnii.l o, ane partly due to ishe added difficulty of obtaining reliable infors:tio r. oz gif to ,nd traa;fers. One argument has been that unearned incoF.2 ohould` not be considered when estimating affordability, because such inco:ie ls sub4ect to the whims and tribulations of others, and is there27ore unreliable. Recent studies have indicated, however, that transfero,E o; t;- y within the 1/ See, for example, Kaufmann, DanLelD, `S3cial I nteraceion as a Strategy of Economic Survival Amiong the U.rbarn Poor: A Theory and Evidence," Unpublished Ph.D. Dissertation, I-nzvard University, 1982. The sample studied consists of 181 pa ticapant and 32(0 control group households in Sanza Ana0 The inforrnatlon on thece families is also consistent with that: fre an earlier survey in Belo Horizonte, Brazil0 (See, for examp.e, Anna Mzr!a Sant°Anna, Thomas W. Merrick and Dipak Mazurfda7, ""ncome Distribution ^and the Economy of the Urban Household: The Case of ielo Rorizonte", World Bank Staff Working Paper, No. 237, J;r:e J lune ) - 23 - extended family, are regularly received and thus not only an important but also a stable component of household income among low-income urban families in developing countries. Moreover, it appears that transfers may be allocated in a compensatory manner, to help the most impoverished families meet their basic needs. -/ Transfers are sensitive to household composition and employment status, as well as to shifts in individual incomes. In studies conducted to date (in Colombia, El Salvador and the Philippines), female-headed households receive more and larger transfers than male-headed ones, 2/ and households headed by unemployed workers are more apt to receive transfers than those headed by working members. These transfers are progressive: most of the net transfers are to the lowest 65 percent of the income distribution. The relatively few net givers falling in this lower segment of the population are concentrated between the 40th and 65th percentiles from the bottom, with transfers again flowing toward poorer families. The transferred funds, moreover, are often "earmarked" for basic needs expenditures, a further 1/ Hypotheses about the motivation for transfers include the proposition that an underlying "contract" exists by which the better-off households provide resources to less well-off households until basic needs criteria are met. These "exchangers" might even reverse roles, moreover: transfer givers may become transfer receivers if income or wealth among extended family members is shifted. in general, the motivaLion for transfers does not seem to be only utilitarian, whLch would posit a more continuous transfer relationship across the income spectrum, nor does it appear to be based on altruistic motives, since the income differential between givers and receivers is at times very smlall. Ibid., Ch. 2,3,4. 2/ In the Santa Ana sample, 60 percent of female-headed households, as opposed to but 25 percent of male-headed ones, received trans- fers. These transfers constituted more than 40 percent of income for more than a quarter of fem.ale-headed hotuseholds. Ibid, p.12- - 24 -- reflection of the "social contract" attached to the transfers between kin-linked households. For many reasons, then, the very existence of the t:ransfers and the characteristi,cs of receiving hoU.3eholds demonas trate that social equi.ty objectives wL!l not be well s-erved by shelter programs if eligibility is based on earnings alone rat'aer than on total income. It is evident that the precise definition of the components of Incoze is crucial in 6aking the affoxdabil.ty calculation, We supgest, based on the research described above, that the total nousehold income -- including the incomes of all family members. from all sources is the appropriate definition to use. There is, however, a more subtle definitional issue that also needs to be settled0 It stems from an hypothesis popularized by economists like M. Friedman and A. Modigliani that households do not base their consumption decisions on their current intake of ircome. I/ If they dic. sc, they v7ould cut back consumption, not only of housing b1.tt of al.l other coniicditties, wdhen ti.nes are unexpect.edly lear arl i :ncz4.!.ase it, when there is winiifall income0 r'he permanent inicoine hypothesis postulates that, in a world where households ,-an borrow and lend freely (eogo, in the form of savings accounts), they can protect themselves from fluctuations in consumption even though income is varying. 0 Households thus calculate what they believe to be that component of their income which is stable or permanent and base 1/ See, for example, Mo Friedman (l957), A Theory of Consumption, Princeton: University Press0 2/ Based on the level of income/consumption and access to formal and/or informal credit sources, this ability to maintain consumption levels is of variable longevity0 - 25 - their consumption decisions on that. In the life cycle variant of this hypothesis, households smooth out their entire lifetime consumption, so that there are rot: discrete jumps in consumption upon entering the labor force or upon retLremeni. In its present form, this theory predicts that a household would imake an implicit calculation of the present: value of its stream of future lifetime r'arnings and form a consumption plan in accordance w-i th that stream. A practica:L way to estimate "permanent income" is t.,-l as.sume that householde take past income, adjusted for a trend and expec$tations, as a good predictor of permanent income. Thus, an average of incomes in past years (which may be weighted, depending on the discount factor) would seem to be a good proxy for permanent income. Nevertheless, formidable measurement problems persist, so we must again distinguish between what is desirable in establishing standards in a research or evaluation program, and what is feasible or practical in an operational setting. Available data sets are typically one-time household surveys which provide bases neither for averaging nor for adjustment, let alone for purposes of estimating permanent income. Thus, strictly from the standpoint of measurement accuracy, the World Bank evaluation program enjoys two tremendous planned advantages in this respect. First, surveys are repeated to give at least three readings at one to two year intervals. Second, an unprecedented survey in the Philippines followed incomes and expenditures of a small sample of households over a three-year period, enabling detailed observations concerning variations in income over time, changes in the composition of income and relationships with other variables. Such observations are very useful in the formulation and refinement of hypotheses. - 26 There is, however, competition here between measurement accuracy and practicality, which is analogous to that regarding cost concepts discussed earlier. Policymakers and project designers seldom have more than one good data set, if thatD at their disposal. They are thus faced with two practical alternatives: (a) to estimate "a" using current income and apply it to current income measures or estimates; or (b) to estimate "a" using permanent income and a?ply it to consumption expendi.ture, as a proxy for permanent income, or even to estimate "a" using this proxy, as has been done by some analysts0 j/ Which approach will produce the best or most cost effective resuits is a questlcn to be answered only as empirical evidence accumulates. 2! Remaining measurement problems relate to capturing accurately all income sources for all persons within a household. These problems include failure or refusal (by the "counters") to record all the sources; difficulties attendant on collecting unearned income information, such as lack of recall aad concea:.ment; and complications caused by such social arrangements as polygamy. There are several likely causes of data unreliability0 One is the impossibility of accounting for fluctuations in incomes in one-time surveys0 Another is the informants possible ignorance of total income sources, whether due 1/ See, for example, Follain, J., B. Renaud and, G. C. Lim, "The Demand for Housing in Developing Counties: The Case of Korea, Journal of Urban Economics", 7:313--336 (1980)O 2/ Recent analysis using the El Salvador data base indicates that the permanent income hypothesis may be weaker in developing countries where capital markets are far from perfect and where windfall income may thus have an impact on consumption (see E. Jimeeaz and D0H0 Keare, "Housing Consumption and Income in the Low Inccme Urban Setting: Evidence From Panel Data in El Salvador", Journal of Urban Economics, (forthcoming)0 - 27 - to poor record-keeping, secrecy of other family members, or inability to reckon the effects of such activities as self-help construction or the family's raising of its own food. Then there is the possibility of outright fabrication of figures, perhaps to meet project criteria, to avoid purveying information that might reach the tax collectors, or simply to get rid of the interviewers. Such problems have consistently plagued analysts attempting to obtain accurate estimates of income. Though the difficulties are real, they are often exaggerated by practitioners and researchers uncomfortable with quantification. Much is now understood about all aspects of sampling, surveying and cross- checking; and experience has shown that acceptable results can be obtained with careful, cost-effective application of this knowledge. Housing Expenditures in the Measurement of "a": In gauging or estimating housing expenditures, or the value of housing services consumed during a year, the first problem is to be clear about what is being measured. This is a relatively straightforward matter where renters are concerned. The rental charge is usually unambiguous and consistent, although there are practices (particularly where rent controls are in vogue) such as "key money", surreptitious payments, and excess charges for certain services, that make the calculation more elaborate. 1/ The task is considerably more difficult for owners, since there is seldom any summary measure of housing expenditures over a given period, such as a month. For low-income households, self-help in house 1/ A colleague has also advised us of instances -- presumably depend- ing on particular market conditions -- of landlords behaving, at least in a limited sense, as "discriminating monopolists" and relating rental charges to tenants' incomes. - 28 - construction complicates the issue further, as one must also place a value on non-purchased inputs such as family labor and scavenged materials. There are two ways of approaching this problem0 One is f o bring all costs down to an estimate of monthly occupancy costo For a host of reasons, this figure is difficult to estimate accuratelyo.-i/ Thus a second, easier method is more often relied upon. This approach entails asking the owner two questions: What would you sell your house for? and what would you rent it for? It is assumed that, if inflation exists, the respondent is equally aware of it in answering both questions, and that money illusion therefore does not affect the results. Recent evidence shows that the answers to these questions, when provided by low-income househoLds, are accurate representations of value. Research performed using the Philippines data base, for example, showed that, on average, squatter dwellers had the same estimate of value of their dwellings as a professional appraiser0 2/ 1/ The appropriate conceptual approach, however, is clear and has been presented above, ppo11-13 2/ See Jimenez, Emmanuel, "The Value of Squatter Dwellings in Developing Countries," Economic Development and Cultural Change, July 1982. - 29 - IV. HOUSING CONSUMPTION IN EL SALVADOR AN EMPIRICAL EXERCISE We hope to illustrate in this section the effects of these measurement considerations using data from El Salvador. -/ We will show how sensitive measures of "a" are to the definitions used. We will then attempt to answer some of the questions raised earlier about the factors which determine "a". The data for this analysis have been collected in El Salvador as part of the evaluation program. They consist of responses to extensive socio-economic questionnaires applied in 1977, 1979, and 1980 1/ The El Salvador sites and services project was one of the first shelter projects to be appraised by the World Bank. The Appraisal Report, published in 1974, stressed the existence of a large and increasing housing deficit which the small and fairly costly government housing programs were not able to resolve either numerically or in terms of providing units which were economically accessible to the urban poor. Given this increasing deficit and the inability of existing programs to generate sufficient new units or to produce them at a price affordable to the low-income population, the objectives of the project were to demonstrate that a sites and services program with partially built units is a practical alternative to conventional fully-built Government housing programs; ease the severe shortage of low-cost urban shelter; demonstrate the potential role of the private sector in providing self-financing, low-income housing; encourage provision of adequate community facilities; and generate employment. The project, to be implemented through the Fundacion Salvadorena de Desarrollo y Vivienda Minima (FSDVM) would consist of the following components: approximately 7000 lots serviced with water, sewerage, storm water drainage, unpaved streets, footpaths and optional electricity; approximately 7000 sanitary units and 3500 basic dwellings; off-site infrastructure, consisting of water distribution and sewer mains and the upgrading of access roads; financing for a materials fund designed to induce self-help extension of core units; community facilities; small industries loans; and training and technical assistance. - 30 - to samples of approximately 260 households in each of two cities, Santa Ana and Sonsonate (The number of families varied with attrition over time0)0 These comprised the control groups for the evaluation; thc,t is, families continuing to live in one of fhree types of housing- rental rooms in mesones, quasi-legal dwellings in cclonias ilegales, and illegal dwellings in tugurios. Mesones are rentrt. tenements of from five to fifty rooms, typically clustered arodnd a central patio, where, among other things, cooking, bathing and sanitary facilities are locatedo In colonias ilegales, legal owners sell plots to legal purchasers who then build houses on them. These subdivisions, however, do not meet subdivision codes, and che homes bult there may not meet building codes0 Tugurios consist of groups of shacks and hovels thrown together in rights of way, arroyos, and other undesirable locations0 1! It appears that estimates derived from data on these families will prove superior to those previously available to guide project design0 Measurement of income and housing expenditures has been more complete and accurate in this case than in most other comparable surveys0 The richness of the survey also enables researchers to control 1/ As of 1975, when this program got underway, these three types of housing (termed vivienda popular) sheltered from 56 to 75 percent of the populations of El Salvaoor's f-ive principal citieso Mesones accounted for the bulk of the population living in such hoasing -- ranging from 53 percent in San Salvac.or up to 77 percent In Santa Ana and 85 percent in Sonsonatea Colonlas ilegales twere tihe most variable component, accounting 'or almost 40 percent of t&e population living in vivienda popular in Usultutaa and a tiM-d in San Salvador, but only 13 and 20 \ / ~ ~ ~~~\ / Total - Mesones i 0 J*l 20 30 40 50 60 70 80 90 100 Income Decile1 - 36 - Tables 1 and 2 provide results for all low-income dwellers in the Santa Ana and Sonsonate surveys, incorporating owners by imputing a rental value to their homes. This imputed rental value is then added to the total income of the owner households (YTT)D since this is impLicit rental income for owner occupiers. The general result is that owners exhibit a greater average propensity to consume housing than do renters. Further study will be needed before confident statements can be made concerning either the magnitude of, or the reasons for, this differenceo - Generally speaking, however, the difference is thought to result partly from definitional ambiguity, 2/ in part from data problems (the imputation of rental values to owners) and in part from real behavioral differences0./ Future studies will be most concerned with these behavioral differences0 1/ Such study is underway in a World Bank Research Project on "Housing Demand and Finance in Developing Countries"0 Among other things, the study, which was begun in mid-1981, will estimate demand for housing by owners and renters in about a dozen countries0 In most countries, at least two cities, and in three countries, several cities, will be covered0 2/ This refers principally to the extent to which owners' housing outlays are, as with renters, for their own housing consumption, as opposed to the creation and mairtenance of rental or business accomodations, or other investment purposeso 3/ Theory tells us, that, if all households were alike (which, of course, they aren't) and face identical unit prices, no such differences would existo In the real world, however, we would expect differences to exist for at least the following three reasons: (1) Part of owners' (apparent) consumption of housing services is to lease space to renters. (2) Owners probably also view housing, in part as an asset to hold0 (3) Households differ in other respects -- e.g. family size and age -- than household income0 - 37 - Another result which is evident in Tables 1 and 2 is that the value of willingness to pay alters drastically, depending upon what type of income is considered. For both cities, and for renters as well as owners, simply considering the wage of the household head will lead to a value of "a"(al) which is almost double that which is calculated using total income (a3). Again, the lesson is that definitions often matter greatly. It has often been assumed that the average propensity to consume varies systematically with income. Figures 3 and 4 generally confirm the downward trend hypothesized by Grimes five years ago.1/ This is so despite certain shortcomings of the data: the relatively small size of the sample generates extreme values for a few cells, particularly for the stratifications by residential type; and, as with all other studies of this type, this phenomenon is most marked--and the results correspondingly least certain--for the lowest 2 to 3 deciles. 2/ Notwithstanding these shortcomings, the following results happen to be unambiguous: First, for both Sonsonate and Santa Ana, the total sample values decline, from .21-.22 for the bottom decile, quite abruptly through the first three deciles, and then fluctuate in a narrow range between about .10 and .08. Second, the stratified sub-samples adopt a similar pattern. 1/ Gritmes, op. cit. 2/ We know very little about housing market behavior among those in the lowest two or three deciles. It is one of the aims of the aforementioned Research Project to improve our understanding in this area. - 38 - These trends appear clearly in the figures0l/ It should be noted, though, that they reflect only this particular stratification of data according to incomes, the observations having been grouped by tenure type. A rough stratification of the data according to housing qu&aity alone reveals that the trend could look quite different: Average Total Income a3 Tugurio ¢ 320 per month .06 Meson 6 470 per month 07 Colonia ¢ 724 per month .19 Failing to note that this result is produced by stratifying according to tenure type, one might be led to hypothesize in error -- that the propensity to consume housing increases, rather than decreases, with income. Thus, it is crucially important to consider micro-level data which are not stratified--or, alternatively, w-here the implications of 1/ These findings from El Salvador confirm those of similar studies in the U0S., Korea and Colombia (See Mayo, S. K., "Theory and Estimation of the Economics of Housing Demand", Journal of Urban Economics, 10:95-116 (1981); Follain, J., B. Renaud and G. D. Lim, op0cit0; and Ingram, G. K., "Analysis of Housing Demand in Bogota and Cali", paper presented to the Meetings of the Eastern Economic Association (Philadelphia)0 While the treatments are not strictly comparable, because the studies of Colombia and Korea had as their samples the entire urban populations of Bogota and Cali, Colombia, and of Korea, respectively, they have all found that the average propensity to consume housing declines with income0 We have found that this declining trend is most pronounced for the louyeot three deciles of income, and is imperceptible for approximately the highest three deciles. The latter phenomenon is of little interest in the planning of housing projects for the urban poor. HoTiever, as indicated, a more precise appreciation of the lotwest deciles' behavior is of considerable importance and will be the subject of further study. - 39 - Table 3: SANTA ANA RENT/INCOME RATIO (APC) FOR HOUSEHOLDS BY INCOME CLASS 1976 1979 Renters 1980 Owners O - 10% .3843 .2281 .1723 .1964 10 - 20 .1442 .1213 .1110 .2236 20 - 30 .1338 .1009 .0833 .1194 30 - 40 .1296 .1293 .0970 .2169 40 - 50 .1005 .0640 .0862 .1114 50 - 60 .1062 .1055 .0816 .1554 60 - 70 .0815 .0593 .0821 .22]1 70 - 80 .0811 .0739 .0699 .1310 80 - 90 .0547 .0642 .0529 .1249 90 - 100 .0339 .0490 .0524 .1206 Total .1140 .0894 .0862 .1561 (.114) (.061) .0967 (.073) - 40 - Table 4: SONSONATE RENT/INCOME RATIO (APC) FOR HOUSEHOLDS BY INCOME CLASS 1976 1979 Renters 1980 Owners O - 10% o2794 o2033 o2432 .2341 10 - 20 o1496 o1080 .1698 20 - 30 o1269 o1120 l409 .1745 30 - 40 o1461 oO712 .O930 .1705 40 - 50 o1022 o0893 oO777 .2112 50 60 o O758 .986 .9786 o O902 60 - 70 .1051 .0616 o0690 .J710 70 - 80 oO707 oO557 oO583 ol233 80 - 90 oO588 oO714 oO669 .3979 90 -100 o0316 .0382 .0601 .1830 Total .1118 .0816 .0978 .1479 (.097) (.052) .0963 (.073) APC SANTAANA OWNERS 1980 1 * ooo o .750 U_ (Y) tl) 3WOONI iN38Ufl OO?O 0091 OOZI 008 00+ 0 + +44. + + + + -44+ + .~~~~~~~~++ + w 4- + ++ + tk+t + ++ + 4+ + GZ +~~ + + OSZ + - TSi C1 +~~ r-4 ~~~~~ 0 a C) 1 ooo ln m --4 C3 0 0 .* I 0861 S~I3IN3 3iVNOSNOS OdV - 44 - the strafications chosen have been very carefully considered-in order to discover the true underlying relationships. _/ Thus, for most purposes, simple stratification into owners and renters, as in Tables 3 and 4, will suffice. it is however, also important to note the differences among households, as is most evident in the scatter diagrams of Figures 5 - 7. Although the general trenc is somewhat obscured, these diagrams highlight the tremendous differences in consumption (at given income levels), particularly at low levels of income and most notably among ownerso Because of these differences, families should never be consigned to live permanently with their relative priorities of the moment, nor forced to adopt new priorities (different "aa"s) precipitously. For example, in a project designed for a target population consisting of a large number of renting families with representative incomes but low (ex-ante) "a"s, the use of standards based on owners' behavior (i.e., higher implied "a"s) will result in a large percentage of these families being designed out of the project. There are basically two ways of avoidina this outcome0 One is to attempt to design the potentially excluded families into the project as owners by setting base standards very low0 The other is to attempt to design them in, as continuing renters, by setting higher standards but providing specific rental components or incentives. Usually a mix of strategies should be employed0 1/ For a more complete treatment of this subject, see Ingram, o tp0it - 45 - Given that our measure of affordability varies with so many factors, an average value of "a" is at best a very rough guide to the other factors in project design. An average figure, rigidly adhered to, will pose problems both for households with a low "a" whose income is adequate, and for those whose high "a" would offset a low income, in an unrestricted situation. Since what the data show, particularly at low levels of income, tends to be obscured by the use of averages (some families spending next to nothing on housing, and other spending a great deal), the most desirable approach, if it can be implemented, will be to generate an expected rent profile for a given population's income distribution, as in Figure 1. Projects should then be designed to eliminate only the bottom (say, 5 to 15 percent) of this profile, where all factors are taken into consideration. This premise argues in favor of projects with options, and with the lowest-cost options providing very low entry costs, but also the opportunity to add substantial amounts of investment. This is the opposite of what most public designs of low income housing I/ offer, but it is precisely what the private markets provide, albeit usually in an illegal or quasi-legal fashion. B. The Determinants of Housing Consumption: This section will attempt to explain what determines housing consumption by controlling for the many factors which seem to affect it, since it is apparent that there is no simple relationship between housing consumption and income. The basic methodology used is multivariate analysis. In particular, the objective is to estimate the 1/ Though not those financed in part by the World Bank. Table 5: International Comparison of i4ajor Housing Demand Parameters Follain 2/ Lim end 1/ __I_E_ Renaud Log:Log Linear Income Elasticity of Demand o6 o78 o35 to 73 Rises with income Rises with city Ouners .6 Renters ° Price Elasticity of Demand -o2 to -.3 3/ -o86 --o63 to 085 R4ses with income Elasticity of Houoing Consumption/Expenditure with Respect to: Family Size o13 to o19 4/ o13 ol2 to o28 Falls with income Age of Read n/a ol4 .07 to o23 Falls with income Sex of Head n/a -.6 =07 to -.16 Falls with income Mean Income ($) $1872 Year 1972 1/ Permanent income (consumption) equations only; national end roups; log-lo 2/ Household income 3/ -o,60 to -o67 for cities other than Seoul /7 Higher for current income and renters - 47 - role of various income measures on housing consumption, while holding constant such parameters as household size and other demogFaphlc characteristics, as well as whether the household is a renter or owner. This analysis is related to the empirical literature on housing demand, under the assumption that prices are held fixed. The housing demand literature has been reviewed, for developed countries, by de Leeuw, -/ and more recently by Mayo. The general conclusion is that "for a wide range of analyses employing different data bases and methodologies, the permanent income elasticity of demand for housing is estimated to be well below one on average. Studies that have used data based on individual observations universally produce estimates 'below unity for both renters and- owners.. 2/ The previously cited papers by Ingram (1980) and Follain, Renaud, and Lim (1980) provide the first rigorous attempts to conduct housing demand estimation in developing countries. The latter analyzed data collected in 1976 in Korea for the regular Family Income and Expenditure Survey, and a Special Housing Survey. Ingram has used data collected in Bogota, Colombia in 1972, aa part of the Bogota Phase II planning exercise, and in 1978 as part of a subsequent major research project conducted by the World Bank and Colombian collaborators. Both studies are free from the aggregation and specification biases that had plagued many earlier studies in the United States. The approaches taken in this paper are generally very similar. These three studies produce 1/ de Leeuw, F. (1971), "The Demand for Housing: A Review of the Cross-Section Evidence," Rev. Econ. Statistics, 53:1-10. 2/ Mayo, op. cit. 48 - strikingly similar results which, though they deal with samples with widely varying incomes, are consistent both with one another and with comparable studies carried out using U.S. data. These results are presented in Table 5 They indicate that the income elasticity of demand, as well as the average propensity to consume are quite stable across these countries. Although the results are not reproduced here, the cited studies also find that the average propensity to consume housing starts out high at low incomes, drops off rather sharply (since the minimal consumption of housing services varies little at low levels of income) and then appears to tend asymptotically to some lower value. According to the Korean data, this lover value is on the order of 8 to 12 percent, the typical range is from 12 to 25 percent, and the intercept may be inferred to be on the order of 30 percent or more. The Bogota data reveal that the highest income groups' average propensities to consume housing fall within the range of 12 to 16 percent, and the intercept may again be inferred to be above 30 percent0 The last stage of our analysis has been to trace the determinants of housing consumption in order to try to understand what lies behind the observed trends0 The analysis is similar to that of the growing body of work on housing demand0 However, there are a few important differences0 One is that we have been able to obtain a better measure of permanent income, which is estimated as a simple average over a three year period0 (All income variables prefixed by P are to be considered as permanent income measures0) This is a result of the panel nature of our data0 The other is that we have postulated that different - 49 - forms of income may contribute in varying proportions to the decision to consume housing. We wished to investigate the hypothesis that the wage income of the household head (YWH) was the primary determinant of consumption behavior. So, in contrast to other studies, we have included other measures of income such as wage income of other household members (YWF), other income of the household (YoF), and total household income (YTT) in the sample. It is clear that the two concerns -- obtaining an accurate measure of permanent income and including other measures of income -- are related to one another. The underlying assumption, which seems to have pervaded most affordability calculations in the past, is that the wage income of the household head is more stable than income from other sources, and thus is a better proxy for permanent income. Since this does not appear to be borne out by the data, we claim that this is at best a testable hypothesis. We therefore include "permanent" notions of other income sources as well in our final regression. The housing equation estimated is of the linear form, and for the "i"th household is written: Ei = Zib + ui where Ei = PHHi = housing expenditures in terms of actual or imputed rent; Zi is a row vector of explanatory variables, Zi = (Yi Xi), where Y is a vector of measures of income and Xi is a vector of demographic variables; b is a column vector of coefficients; and ui is a random error term. The results of the estimation are presented in Tables 6 and 7, the former using current income and the latter permanent income measures for Santa Ana, in 1980. (Tugurios are excluded because, although they constitute Table 6: HOUSING CONSUMPTION IN SANTA ANA 1980--CURRENT INCOME OWNERS RENTERS ALL CONSTANT 131.933 122.130 -40.780 -20.828+ -15.612 2.240 (2.846) (2.894) (-2.235) (-1.036) (-.7110) (.1050) YTT .0912 .1875 .1570 (3.451) (9.142) (8.900) YWH .0732 .2420 .1932 (1.826) (11.864) (8.541) YWF .0972 .0720 .1039 (2.593) (2.659) (4.121) YOF .1076 .114 5 .160O (2.4 85) (2.441) (4.834 ) AGEn -.0318 .1494 .2694 -.1619 .4785 .2159 (-.0430) (.2240) (.9631) (-.5386 (1.375) (.6458) SEXHI -19.432 -25.383 -23.206 -.6786 -22.832 -9.590 (-.836) (1.249) (-2.334) (-.0615) (-1.890) (.8496) UHBSIZE -7.2 64 -6. 854 .3589 -4.477 -.6579 -3.660 (-1.551) (-1.610) (.4613) (1.979) (-.2544) (-1.5213) R2 .16 .16 .54 .40 .33 .30 N 72 72 136 136 208 208 * significant at 95% ** significant at 90% + coefficient exceeds standard error Table 7: HOUSING CONSUMPTION IN SANTA ANA 1980--PERMANENT INCOME OWNERS RENTERS ALL CONSTANT 113.462 100.524 16.965 9.990 .4133 -6.364 (2.343) (2.204) (1.730) (1.043) (.0161) (-.263) PYTT .0895** .0295** .119** (2.795) (2.566) (5.617) PYWH .0681 .0197+ .1059** (1.324) (1.193) (3.370) PYWF .1113** .0111+ .1142** (2.439) (.950) (3.963) PYOF .1012** .1048** .1522** (2.019) (5.618) (4.263) Utn AGEH -.0802+ .1620 -.0743 .1304 .279 .4119+ (-1.966) (.234) (-.538) (.964) (.704) (1.121) SEXH -7.683 -18.803 7.021+ 5.471k -6.437 -8.906 (-.298) (-.851) (1.524) (1.219) (-.474) (-.738) HHSIZE -4.713+ -3.576 .8850+ -.1317 .1358 -.140 (-1.012) (-.829) (1.003) (-.1424) (.0526) (-.058) R2 .13 .12 .29 .10 .22 .21 N 63 63 93 93 156 156 * significant at 95% ** significant at 90% + coefficient exceeds standard error - 52 - only a small proportion of the population, they make up a very large proportion of the sample.) The first observation is that each of the "other" income variables, representing sources other than the wage income of the household head, contributes significantly to explaining the variability of housing consumption. According to the current income measures in Table 6, the marginal propensity to consume housing out of "other" income exceeds that for all other income variables0 This finding is consistent with observations for the Philippines and Senegal that house construction for a majority of families is financed by savings and transfers from the extended family network, rather than out of current wage income0 These results are confirmed when permanent income measures are used in Table 7. The fact that the marginal propensity to consume out of other incomes is larger than that from wages leads us to believe that, under the permanent income hypothesis, there are large transitory components at work0 Further revealing information is contained in Table 8, which presents estimates of the income elasticities of housing demand for the various sub-groups at their respective mean incomes and rents0 These results support the conclusion of the previously cited studies that demand for housing is relatively income inelastic0 This is true for renters and owners in Santa Ana, across all four categories of income considered0 The elasticity of demand with respect to "permanent" wage income of other household members and with respect to other "permanent" non-wage income is stable across income groups0 The elasticity of demand with respect to the three-year average of wage income of the Table 8: INCOME ELASTICITIES OF DEMAND FOR HOUSING (EVALUATED AT THE MEAN) Santa Ana Bogota Call Korea Renters Owners Renters Owners Renters Owners Renters Owners Permanent Income Wage of Head .1599 .1589 Wage of Others .0600 .2316 Other Income .1285 .0647 Total Income .4338 .4 530 .78 .75 .47 .73 .42 .62 * Current income estimates. = 54 household head appears to be quite similar for renters and ounerso A similar result is apparent with regard to total permanent incomes of the entire household. Household size does not seem to contribute significantly to housing demand. However, the sign of the coefficients reflects angram-s finding that increased family size leads to increased demand among renters, but not among owners. This w^jould be especially true fox meson rental units which tend to be one-room dwellings0 The only other significant demographic variable is sex of household head. Our estimated coefficients indicate that female-headed households tend to demand more housing, and are thus again consistent with Ingrams findings for Colombia. - 55 - V. PRINCIPAL CONCLUSIONS AND RECOMMENDATIONS This exercise has comprised an evaluation of experience with respect to the affordability of projects, and of the usefulness of affordability measures used in their design. Taken together, these findings suggest how affordability considerations should be brought to bear in the design of future projects. The Affordability Record: Evaluation to date shows that the urban housing projects so far financed by the World Bank have generally been affordable to their target populations. This has been so even though actual costs have turned out to be considerably higher than originally estimated. There are two reasons behind the higher costs. First, materials prices have risen at unforeseen rates, and the impacts of this and other inflationary forces have been intensified by substantial delays in project execution. Second, the additional amounts that families have voluntarily invested in their plots and dwellings have greatly exceeded initial estimates. This trend has resulted partly from inflation and partly from the fact that families have generally allocated more resources to house-building than had been anticipated. Notwithstanding these factors, it is heartening to find that projects have been affordable even to families within the bottom 20 percent of the income distribution. 4/ This appears to have been due in large measure to the validity of the progressive development approach 1/ This is not the same as saying that distributional impacts of the projects have been optimal, which subject is currently under investigation. 56 - and to conscientious, and largely successful, efforts to keep project standards and costs down to (or near to) what were presumed to be affordable levels. It was also due in part to luck, in that the target populations' income levels in these first projects were underestimated because little was known initially about the importance of transfers. The unanticipated availability of transfers within the extended family, of course, links back to the unexpected degree of investment in the home. In El Salvador, in particular, there were some additional reasons for the affordability of the product. The Fundacion's program was of long standing, and considerable experience had been built up0 Furthermore, great emphasis had been placed on affordable design from the outset, and modifications -- in standards, in plot layouts and in site layouts -- were continually being made to improve performance in this regard0 Great care was taken in the selection of participant families0 In addition, the Fundacion utilized the mutual help process to ensure that all families' downpayments could be made, and utilized both rewards and sanctions to encourage good payment performance0 The positive findings on affordability are supported by indirect indicators as well0 One such indicator (of lack of afford- ability or "affordability problems") would have been uncommonly high turnover of project populationso This is not an unambiguous indicator, as people might move out of projects at high rates for reasons quite unrelated to affordability0 For example, they could be realizing substantial capital gains0 On the other hand, they could be staying on in spite of affordability problems, in which case the effects would likely show up eventually in one or more of the indicators discussed - 57 - below. But it is worth stressing that in none of the projects studied has turnover among project populations been higher than that among controls. In El Salvador, the turnover rate has been much lower than that among the controls or the population at large, once allowance is made for some early departures among families that found participation in the mandatory mutual help program onerous. This is unlikely, how- ever, to be attributable solely to affordable design. It appears that the most important additional explanation in the El Salvador case is simply the overall meticulousness of the entire process -- in which the careful selection of participants, their organization into and involvement in mutual help teams and the careful explanation of project (and community) benefits and obligations play an important part. It is also likely that, where subsidization of (a portion of) project costs _ coexists with barriers to transfer of the subsidy in the event of resale, 2/ the participant may be, or at least feel, "locked in" for a time. Another possible indicator would have been an unexpectedly slow rate of house consolidation. This phenomenon has been observed in the Senegal project, and initially affordability was involved. Government spokesmen raised people's expectations of what would be done for them, while the project initially had no loan component. When one 1/ Something which occurred only to a very modest extent (with respect to interest rates) in the El Salvador project, but has been of some significance in some other projects. 2/ This is proscribed, in El Salvador, by the fact that the tenure arrangement is leasehold for the first five years. 58 was introduced, it was inadequate, in that the ceiling amount for individual loans constituted only a very small vproportion of the costs of the homes the participants were attempting to build. Even after these problems were resolved, project progress has been sloW, for reasons having to do with more general design failures and implementation problems. Individual house construction in all other projects has proceeded at least as rapidly as expected. Still another possibility is that families might have attempted to adapt to affordability problems by altering income and/or expenditure patterns. There was particular concern in the early phases of program development that participants might be forced to reduce consumption of other necessities, such as food or health services, to meet housing payments0 In fact, there have been adjustments on the income side: participating families have a high propensity to expand their houses and rent out part of the additional space, and there have been some indications of possible positive effects on labor force participation. These, however, can hardly be considered indicators of affordability problems so long as the families are successful. On the other hand, there have been no indications that investments in housing have led to reduced consumption of fcod, medicine, or other necessities. A final indicator would have been a relatively high rate of default, which is linked to the first indicator, as, presumably, households whose payment lags persist will eventually be evicted0 Here the evidence is mixed0 In El Salvador, default rates were virtually nil up to the outbreak of the major civil conflict in 1979. In the Philippines, collections have barely started. In the two African - 59 - countries, however, default rates on plot payments (and in Zambia on loan repayments) have been distressingly high. In the case of Senegal, it is difficult to separate this phenomenon from the complex of problems that have plagued the project; however, affordability does not appear to have been at the heart of the problem. In Zambia, it was initially thought that affordability might be the key factor, particularly given the economic reversals faced by the country during the period of project execution. Further research, however, has shown this not to be the case. Defaulting has not been related to income, but caused by other factors, notably poor financial administration. Affordability Measures: This paper has also attempted to evaluate affordability measures currently in use for progressive development projects. It is essential that the measure ("a") of willingness and ability to pay is constructed from appropriate income estimates, combined with reliable figures on rents or occupancy costs. In conducting research on affordability, the measure of income chosen is crucial. The disaggregated results of early World Bank- financed experimentation with urban housing show that housing expenditures are least responsive to the measure of income originally favored--wages of household head--and considerably more responsive to other income components. As we have emphasized, most important for project design purposes is total household income, including "unearned" as well as earned income sources. Total household income comprises a more encompassing definition of family purchasing power, which reduces the risk of screening out some of the people who should and can be helped by housing programs. Also contrary to original expectations, - 60 - total household income is less subject to variation than is wage income taken alone. This measure of income must still be used with care, and in constructing such measures attention must be paid wherever feasible to the distinction between transfers that are stable components of income and those that are not. Data collected as part of the Evaluation Program, initiated and partly funded by the World Bank, make it possible to provide reliable guidance on this subject. The other main lesson is that, if available, measures of permanent income should be used in calculating "a. While the World Bank's project staff are unlikely to have access to the type of panel data necessary for such calculations, there is available the alternative discussed above of using consumption expenditures as a proxy for permanent income, and such figures are frequently available. It appears that total household income, including transfers, may in most months closely approximate "permanent" income in that a major function of transfers is to smooth out income fluctuations. Plotting rents (or occupancy costs) against total income for the selected distribution will give the locus of values for "a". World Bank evaluation studies and comparable recent research have found that a has the following characteristics: It basically declines as income rises, because housing is a necessity. The elasticities estimated for the various cited studies of housing consumption are strikingly similar, falling within quite a narrow range. This does not imply identical values of "a" among countries and cities. In fact, the values of a" appear to vary considerably among countries and also according to city size. In addition, "a" varies with other factors, including household - 61 - characteristics, and nature of tenure. Owners tend to consume considerably more housing than do renters, and, as income rises, the ratio of owners to renters consumption rises. Even if we adjust the calculation for owners to ensure strict comparability by adding implicit rent to their incomes so that, for owners, a =Y , only part of the discrepancy disappears. This furnishes some evidence that homeowners own their homes not just for the annual housing services provided, but for some combination of other reasons, including use of (part of) the premises to earn income, and expected capital gains. Affordability Considerations in Project Design: Practitioners in the field need reliable estimates of "a" to work with. 1/ Just at the moment these are only available for a very few countries 2/ and cities. As mentioned, research currently underway will add substan- tially to this list during the coming 18 months. Once this larger selection is available, practitioners should in most instances be able to transfer existing parameter estimates from comparable environments or to select appropriate values of "a" through a process of interpolation. 3 Until that time, and in some instances thereafter, it may be necessary to estimate the parameter ("a") using data from the site. In 1/ Presented, as in this paper, as plots of a against Y. 2/ Colombia, El Salvador, Korea and perhaps one or two others. 3/ When outside estimates from other data sources are used in the comparison of project costs to the income disribution of the target population, the use of consistent definitions is crucial. If outside estimates of "a" are to be used, how that value of "a" was determined should be ascertained. Specifically, it is necessary to know whether the calculation relies upon the same components of costs and income as the C and Y to be used in the affordability calculation. - 62 - such instances1, it is advisable to apply insofar as possible the approach outlined in this paper. This means1, again insofar as possible, combining decent data on rent (or occupancy cost) and income for the same sample population, which should be "comparable" 1/ to the target population for the project. Rather than using "inappropriate" data,92 practitioners should (preferably) return to the list of available parameter estimates for other countries/cities and make informed guesses as to the value(s) of "a" to use; alternatively they should swallow (or induce the borrower to swallow) the costs of conducting appropriate small surveys and estimating "a" directlyo Sometimes it may be particularly worthwhile to do this, because the survey is required, or may serve a number of additional purposes -- and the delay entailed can be endured. -/ Occasionally, projects should be used to generate new samples in any case. The ultimate objective should be to persuade natiLonal sta- tistical agencies systematically to generate a set of micro-data, including periodic household sample surveys, with geographical 1/ The following should suffice to meet the definition for comparable: for upgrading or sites and services projects, a city- wide sample, or frequently a national urban sample w-- where the practitioner/analyst would select that part of the overall distribution which pertains to the sample population; for upgrading projects, samples of the same area or similar areas; and, for sites and services projects, a sample of some sub-sample of the entire distribution which covers more-or-less the target income range. 2/ Data of poor quality or data for a sample population that is not reasonably comparable to the target population (actual o.: assumed)0 3/ In such cases, research advice on designing and conducting the survey should be obtained0 This advice should be extended to the creation and clearing of the data tapes as wiell as to their storage for future use. - 63 - breakdowns, such as urban/rural, as well as by major cities. 1/ Appropriately planned and executed such a program will greatly reduce the need to generate new survey data for specific purposes and populations. 1/ Colombia and Korea currently do this; Kenya and the Philippines are on the verge of doing so; and a number of other countries have begun undertaking such surveys on ad hoc bases. World Bnk Publications of Related Interesf NEW The Bertaud Model: A Model Designing the Site and Ser- Cities In the Developing for the Analysis of Altema- vice Plot Allocation Process: World: Policies for Their tives for Low-income Shelter Lessons from Project Equitable and Efficient In the Developing World Experience Growth (Prepared by PADCO, Inc., in Lauren E. Cooper Johannes Fi Linn collaboration with the staff of This analysis provides a framework in Delineates the major policy issues the World Bank.) which to identify and trace the steps that arise in the efforts to adapt to This model is a working tool for required to implement the site and the growth of cities in developing technicians and policymakers who are service plot allocation process. It is countries and discusses policies responsible for low-income settle- based on a survey of twelve urban designed to increase the efficiency development projects and two in- and equity of urban development. ment projects. It Identifies tradeoffs depth project studies. A detailed pic- Particular areas covered include among land use, physical design, and ture emerges of the factors associ- urban employment income financial parameters. ated with allocating plots and the redistribution through the fiscal Urban Development Technical Paper relationships and interdependencies system, transport housing, and Number 2. December 191. 153 pages of one factor to another. Potential social services. The policy instru- (including 3 statistical annexes). problem areas and recommendations ments considered include public on how to avoid them are identified. investment, pricing, taxation, I5SSN 0253-3324. Stock No. UD-0002.anreuti. $5.00 Urban Development Technical Paper an regulation. Number 3. March 1982. 30 pages Oxford Uniuersity Press. 1983. 240 (including annex). pages (including bibliography, index) ISSN 0253-3324. Stock No. UD-0003. ISBN 0-19-520382-8 $2750 hardcover: $3.00. ISBN 0-19-520383-6. $12 50 paper- back. NEW NEW Housing for Low-income Urban Families: Economics Environmental Management Evaluation of Sites and and Policy in the of Urban Solid Wastes In Senrices Projects: The Developing World Developing Countries: A Experience from Lusaka, Orville r Grimes, Jr. Project Guide Zambia Analyzes the operation of urban Sandra J. Cointreau Michael Bamberger, housing markets in developing coun- This guide provides Information and Bishwapriya Sanyal, and tries to determine the kinds of dwell- procedures for planning and Imple- Nelson Valverde Ings affordable by the urban poor mentation of solid-waste manage- This report presents the findings of a The Johns flopkins University Plss, ment Improvements It Is designed to five-year evaluation of the First 1976. 190 pages (Including statistical facilltate project preparatlon. Lusaka Upgrading and Sites and Ser- appendix, select bibliography. index) appraisaL and Implementation of vices Project made possible through a LC 76-4934 ISBN 0.8018-1853-2. Dank-assisted sold-waste projects In cooperative research project sup- $18 50 (59-50) hardcover, urban aeas Current Bank objectives, ported jointly by the World Bank and ISBN 0-8018 854-0. $895 (55.50) policies. and project requiremeints are the international Development 1B -0815-,$ 5U.0 summarized Research Centre of Canada. paperback Urban Development Technical Paper World Bank 5taff Working Paper Spanish Vivienda para familias urbanas NumberS August 1982 214 pages NIo 54.8 1982 201 pages,y de baol icngs5 aspmundos ecn dsromcos (including S annexes)ydepftcenemuoenesrlo ISBN 0-8213-0063-6 ISS 0253-3324 1SBN 0-8213-015-2. $10.00 Editorial Tecnas. 1978 $500 ISBN 84-309-0778-5. 480 pesetas NEW NEW National Urbanization Policy House Registration i In Developing Countries Evaluation of Shelter Handbook: A Model for Bertrand M. Renaud Programs for the Urban Registering Houses and tlational urbanization pollcies in loon Principal Findings Plots In Unplanned developing countries often attempt Douglas 11. Keare and Settlements without a full understanding of the Scott Parris Saad Yahya forces at work to block the growth of the largest cities and to induce This report provides an evaluation of A manual for professionals and decentralization This book takes a four sites and services and area administrators working in urban critical look at such policies and their upgrading projects in El Salvador, the uncontrolled settlements In the third weak conceptual foundations and Philippines Senegal. and Zambia and world Examines the problem of describes problems inherent in confirms that Bank-supported urban registering of houses and plots in Implementation. The coverage is shelter projects have been remarkably unplanned settilements in the large comprehensivc and both global and successful and recommendations are urban centers of the developing national trends are analyzed rnade for future projects countries. World Bank Staff Working Paper Urban Development Technical Paper pages U ncluding indexs append1es No 547 1982 109 pages Number 4 June 1982 75 pages LC 81-3999 ISBN 0-19-520264-3. ISBN 0-8213-0114-4 $5 00 (including 3 appendixes) $1695 hardcover. ISBr1 0-19-520265-1, ISBN 0-8213-0053-9 ISSN 0253-3324. $7 95 paperback __ _ 3 __ $ 00 NEW Also available in lXorean from the forea Research Inst5bute for Human Evaluation of Sites and Ser. Housing Settlements vices ProJects: The Evidence Reviews the importance of housing from El Salvador within the framework of urban Ownerhip and Efficiency Michael Bamberger, development problems, and makes in Urban Buses recommendations for World DankiiUra Bue Edgardo Gonzalez-Polio, and assistance In integrated urban Charles Feibel and IJmnuay Sae-Hau planning A A. Walters This report presents the findings of Sector Policy Paper May 1975 World Bank Staff Working Paper No a fve-year evaluation of the first 75 pages lincluding 13 annexes) 371 February 1980 19 pages (includ- El Salvador sites and services project Engilsh, French, Spanish, and Arabic ing selected bibliography) made possible through a cooperative research project supportedjointly by Stock Nos. PP-7506-E, PP-7506-F. Stock Nto WP-0371 $3 00 the World Bank and the International PP- 7506-S. PP- 7506-A $5 00 Development Research Centre of Canada World Bank 5taff Working Paper No 549 1982 233 pages ISBN 0-8213-01U6-0 $1000 rhe People of Bogota: Who Urban and Spatial NEW rhey Are, What They Earn, Development in Mexico Where They Live Ian Scott Urban Land Policy: Issues Rakesh Mohan Examines urbanization in a country and Opportunities World Bank Staff Working Paper No. in which that process has been Harold B. Dunkerley, 390. May 1980.153 pages (including 2 particularly rapid and in which such coordinating editor, with the Maynixs b980.i15r3pagsyinl). n issues as provision of jobs, shelter,.sitneo appendixes, bibliography), public services, and mass transit are assistance of Stock No. WP-0390. $5.00. urgent. Also considers issues that Christine N.E. Whitehead arise because of the size and form of Various authors with experience in the system of large cities and the their respective fields discuss major Pirate Subdivisions and the linkages between them: centraliza- problems of urban land: the rising Market for Residential Lots tion, rural-urban integration, and pat- cost, the relation of different types of in Bogota terns of interregional development. land tenure to equity and efficiency, Alan Carroll The study is relevant to other coun- the rationale for government interven- Alan Carroil tries in which similar problems will tion and the forms it may take-taxa- World Bank Staff Working Paper No. undoubtedly become increasingly tion, acquisition and development of 135. October 1980. 116 pages (includ- urgent. land, regulation of land use, and ing 2 appendixes, tables, references). The Johns Hopkins University Press, other forms of control. The analyses 5tock No WP-0435 $5.00.1982. .344 pages (including maps, refine and illuminate many of the Stock Nio. WP-0435. $5.00. 1982.344 pages bibluding urban problems that confront appendixes, bibliography, index), developing countries and provide Sites and Services Projects LC 80-8023. ISBN 0-8018-2499-0, practical guidelines for dealing with $32.50 (£20.75) hardcouer; them. Chapter authors are John M. Urban Development Technical Paper ISBN 0-8018-2498-2, $9.50 (i6.75) Courtney, William A. Doebele, Harold 'Yumber 1. April 1974. 47 pages (includ- paperback. B. Dunkerley, Malcolm D. Rivkin, ~ng annex). English, French, and Donald C. Shoup, Alan A. Walters, and Spanish. Christine M.E. Whitehead. 5toc Nos PP-401E, P-740-F,Urban Economic and Plan- Oxford University Press. 1985. About Pt 7401ock $5.001 ning Models: Assessing the 208 pages. PP-7401 -S. $5.00. Potential for Cities in LC 82-20247. ISBN 0-19-520403-4, Developing Countries $22.50 hardcover. rhe Transformation of Rakesh Mohan Urban Housing: The Applies current urban modeling REPRINTS Experience of Upgrading techniques to cities in developing The Costs of Urbanization In in Cartagena countries. Problems intrinsic to the Developing Countries N. Paul Strassmann modeling of all urban areas are Johannes Linn considered along with those particu- World Bank Reprint Series Number 227 ln urban development, it is now lar to cities in poor countries. Reprnted from Economic Development and 'fficient to upgrade the existing The Johns Hopkins University Press, Cultural Change, vol J0. no 3 (April lousing stock than to replace it with 1979. 192 pages (including bibliogra- 1982)S625.48 iew construction. Housing transfor- phy, author index). Stock No. RP-0227 Free of charge ,nation should not only be expected LC 78-8437. ISBN 0-8018-2141-X, The Economis of Self-Help Housing: Fostered as a means to increase $7.50 (.5.25) paperback. Theory and Some Evidence from a production in an important field, to Developing Country 3enerate employment, and to Emmanuel Jimenez mprove equity in the distribution of NEW World Bank Reprint Series. IYumber 223 iousing. This study shows how the Reprinted from Journal of Urban Economics, lousing stock in Cartagena was The Urban Edge uol 1 (1982):205-28 .ipgraded during the 1970s by Herbert H. Werlin (Editor) Stock No. RP-0223. Free of charge. zwner-occupants and, to a lesser and George G. Wynne zxtent, by landlords. The principal a G . The Spatial Structure of Latin :onclusions are supported by exten- (Publications Director) American Cities sive econometric tests and by This six-page newsletter is published Gregory K. Ingram and Alan Carroll interviews with local experts and for the World Bank by the Council for World Bank Reprint Series' Number 211. suthorities. International Urban Liaison. It is con- Reprinted from Journal of Urban Economics. rhe JohnsHopkins University Press, cerned with practical approaches to vol. 9. no. 2 (March 1981): 257-73. rhe Johns lopkins urban problems in developing Stock Nlo. RP-0211. Free of charge. !982. 239 pages. countries. 'C 81 -48176. ISBIY 0-8018-2805 -8, Ten times a year. English, french, 522.50 hardcover. Spanish, Subscriptions at $20 a year (plus $4 for overseas postage) can be obtained from CIUL, 818 18th Street, N.W., Washington, D.C. 20006, U.SA. WORLD BANK PUBLICATIONS ORDER FORM SEND TO: WORLD BANK PUBLICATIONS WORLD BANK PUBLICATIONS RO. BOX 37525 or 66, AVENUE D'IENA WASHINGTON, D.C. 20013 75116 PARIS, FRANCE U.S.A. Name: Address: Stock or ISBN # Author, Title Qty Price Total _ _ _ _ _ _= l__ _ _ _ _ _ I _ Sub-Total Cost -_ Postage & handling fee for more than two free items ($1.00 each). Total copies _ _ Air mail surcharge ($2 00 each) TOTAL PAYMENT ENCLOSED_ Make checks payable: WORLD BANK PUBLICATIONS Prepayment on orders from individuals is requested Purchase orders are accepted from booksellers. library suppliers. libraries. and institutions All prices include cost of postage by the least expensive means. The prices and publication dates quoted in this Catalog are sublect to change without notice No refunds will be given for items that cannot be filled Credit will be appiied towards future orders No more than two free publications will be provided without charge Requests for additional copies will be tilled at a charge of US $1 00 per copy to cover handling and postage costs Airmail delivery will require a prepayment of US $2.00 per copy Mail-order payment to the World Bank need not be in U S. dollars, but the amount remitted must be at the rate of exchange on the day the order is placed The World Bank will also accept Unesco coupons. The World Bank Headquarters: 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. Telephone: (202) 477-1234 Telex: WUI 64145 WORLDBANK RCA 248423 WORLDBK Cable address: INTBAFRAD WASHINGTONDC European Office: 66, avenue d'1ena 75116 Paris, France Telephone: (1) 723-54.21 Telex: 842-620628 Tokyo Office: Kokusai Building 1-1, Marunouchi 3-chome Chiyoda-ku, Tokyo 100, Japan Telephone: (03) 214-5001 Telex: 781-26838 ISSN 0253-2115/ISBN 0-8213-0166-7