The value of squatter dwellings in developing countries

THE WORLD BANK DEVELOPMENT ECONOMICS DEPARTMENT URBAN AND REGIONAL ECONOMICS DIVISION URBAN AND REGIONAL REPORT No. 80-17 THE VALUE OF SQUATTER DWELLINGS IN DEVELOPING COUNTRIES EMMANUEL JIMENEZ University of Western Ontario, Canada November 1980 The author is a Consultant to DEDRB. This report was prepared as part of a program of Monitoring and Evaluation of Urban Shelter Programs, which is being conducted by DEDRB. The views expressed here are those of the author and should not be interpreted as reflecting the views of the World Bank or its affiliated organizations. The Value of Squatter Dwellings in Developing Countries Emmanuel Jimenez'. The University of Western Ontario and the World Bank I. Introduction A great proportion of urban dwellers in developing countries live in illegal squatter communities.. Recent studies. estimate that some 20-30% of the- population of the largest Asian cities inhabit so-called "uncontrolled settlements". Similar figures obtain for African and Latin American cities, .1 although over a wider range. Despite the fact that they are frequently made up of makeshift and dilapidated materials, do not have access to basic ser- vices, and are under constant threat of being razed by the government authorities, squatter dwellings, .1ike those in the established legal (or formal) sector, are economic goods and can be characterized by a.market value. Thus, given the relative size of the population that they shelter, any analysis of urban housing markets in developing countries which relied solely on statistics- on the formal sector would most likely be misleading. Social scientists have begun to dispel the traditional notion that squatter housing units are more than temporary, valueless structures which 2 are shoddily built on vacant land. However, figures have not been documented and relatively little is known about this "informal" market. How much are these .squatter homes worth? What determines the value of these dwellings? Does this market differ signiZicantly from that found in the formal sector? 1* 2 These questions have important implications for policy measures, such as, urban development projects, which are intended to improve the lot of the low-income urban population. This paper is intended as a first step in filling this gap in the literature. Aside from presenting estimates on the value of squatter dwell- ings for a sample in the Philippines, the statistical relationship between . value and housing characteristics (such as the availability of certain ser- vices) will also be derived. These resulting coefficients can be inter- preted as the marginal prices of these characteristics. This exercise has some direct policy implications. In particular, the valuations of the dwelling units can be used to measure the implicit costs of housing projects which raze existing settlements to make room for dwellings exhibiting a certain set of characteristics. Also, if these valuations are indeed signi- ficant and are found to be consistent with physical measures of housing quality, then, it is evidence that the official valuation of housing stock, based primarily on. "formal" housing units, may be seriously underestimated. Further, the derivation of the determinants of housing value provides valu- able price information about the implicit market for housing character- . 3 istics. As such, the resulting coefficients can be used: to control for price changes when value is used as a predictor of how housing quality changes over time; to gain some empirical insights into the benefits of providing certain housing characteristics, such as sounder building materials, running 4 water and sanitary facilities, in urban development projects; and as an -initial step in estimating the parameters of the demand for housing char- 5 acteristics. * .3 Prior to any estimation, we must first confront the primary problem . of all studies which require the value of oWner-occupied dwellings--that of obtaining adequate data. The most accurate estimate- of value would be .the sale price of the house at the time of purchase. However, the frequency . of such transactions in studying the characteristics of a housing market in a neighborhood limit the use of such information. Moreover, because no official records are kept of any of these largely illegal transactions, it is difficult to obtain information of the sale price of. dwelling units in squatter areas. One easily accessible estimate is that of the "owner" of the structure '(but not the land). Indeed, even for the United States, many studies use. the owner's estimate of housing value (see Kain. and Quigley 6 for a list of such studies ). However, this procedure raises additional questions regarding the accuracy and possible biases of the estimates. A secondary goal of this paper, then, is to evaluate the reliability of the owner s estimate of the value of his/her dwelling. In particular, it will compare owner 's estimates with appraisal estimates, and evaluate the results in light of two similar studies for U. S. data, by Kish and Lansing and by Kain and Quigley which concluded that, while errors of estimate (i.e., the discrepancy between owner and appraiser valuations) are quite large for . individual properties, they are largely offsetting for reasonably sized . 7 samples. II. The Data In 1978, the Tondo Foreshore area of Manila contained the Phillipine's . largest concentration of squatter dwellings--over 200,000 low-income indi- viduals out of a total city-wide population of 1.2 million in a landfill area of 137 hectares. It was also the oldest and most established squatter community with a long and well-documented political history. The struc- tures exhibited a wide variance in terms of quality--makeshift homes of scrap .metal could be found not' far. from those made of concrete and brick. However, they all shared a common legacy of illegality and lack of services. Since then a World Bank financed urban development project has .legalized the residents' status. This project is miant primarily to provide basic services and tenure to the community. A monitoring and evaluation effort by the Phillipine National Housing Authority 's Research and Analysis Division (RAD) is gathering .detailed housing information on 96 households (randomly selected from the Tondo through stratified samples) as subjects of a house consolidation study. Designed to eventually deter- mine the extent and the speed of house upgrading activities, this data set provides infdrmation on the initial. (i.e., before project implementation) housing conditions of the households. This includes estimates of the value of each house by varipus appraisers, as well as measures of its characteristics. It is the primary data source.for this study. *As stated in the introduction, the informal housing market, by defi- nition, operates in a world which is beyond the scrutiny of conventional government activity. Since data and tax collection agencies do not enter this world, legislated restrictions such as minimum servicing requirements for housing are irrelevant. It is thus not a trivial task to study the market for squatter .dwellings when selling prices are not recorded. In the absence of these recorded prices, the National Housing Authority's RAD was very re- sourceful in obtaining informiation on alternative measures of housing value. Household heads were asked by interviewers "How much do you think you could sell your house for?" For the 96 heads who responded, this figure is inter- preted to be the owner's own valuation of the structure. A local professional .5 appraiser-engineer was then hired to valuate the same dwelling units. Only one; appraiser was used. A third potentially useful measure had-to be dis- regarded for this study. The five most proximate neighbors were asked how much they 'thought the dwelling unit in question was worth. Unfortunately, the only figure available for this study was an average in which the upper extreme was dropped for each dwelling unit, thus seriously underestimating the results. III. The Value of Squatter Dwellings The housing value estimates are summnarized in Table 1. The owneras assessment is 14,145.83-pesos (U.S. $1886) on average for the 96 households, which is not significantly different from the appraiser's average estimate of-14,092 (U.-S. $1879). The .first observation is, that, contrary to what one might expect, these households do not live in hovels which have little value. .Since average annual income in Tondo is about 7400 pesos (US $982), the average appraised value of the house is about 90% greater than the aver- age annual income per capita from regular sources, a signficant amount which 9 is not much different from the U.S. figure of 85%. This is partly attri- butable to the fact that Tondo is an established community, even though it has only recently shed its illegality. Also, among squatter communities, the . Tondo dwellers appear to be slightly better off. Most of its workers are employed in the port area or in Manila's biggest market area. Still, the relatively high valuations tend to substantiate the notion that many squatter 10 dwellings are more than makeshift and temporary. This has important impli- cations for the cost side in the economic evaluation of a project which razes squatter communities. 6 The mean value owner and appraiser estimates are almost exactly the,same. In fact, they are not statistically distinguishable even at the strictest confidence levels. The Philippine sample does well when com- pared with Kain and Quigley's results for St. Louis differences in average valuations. Then .conclusion then, is that on average, owners valuate their dwellings consistently with valuations obtained from accepted appraising practices. For individual estimates, the results are rather different. The aver- age of the absolute value of the differences between owned and appraised val iations is just over one thousand dollars for the Phillipines sample. This is approximately 65% of the mean appraised value. The comparable figure for Kain and Quigley is approximately 20%. This discrepancy in individual estimates is highlighted in Table 2, which tabulates the distribution of differences for the Philippines sample and the two existing U.S. studies. Within the discrete categories of the table, the distribution appears to be bimodal in the Philippines, where three respondents out of four were unable to estimate the value of the home within 30% of an independent appraiser 's . valuation. On the other- hand, both Pearson and Spearman correlation coeffi- cients are relatively high which indicates that -the estimates tend to vary together, both in terms of magnitudes and of rankings. An interesting hypo- thesis is that the use of these estimates as dependent variables in a re- gression equation of the value of the characteristics of the dwelling units would not lead to very different results.11 IV. The Determinants of Hou3ing Value The preceding section established that owners' valuations of their own dwellings are, on average, the same as those of the appraisers' estimations; that these two magnitudes tend to vary together; and that these valuations 7 are relatively large. The next task of this paper is to derive, through hedonic price techniques, which characteristics of squatter dwellings con- tribute most to housing value and whether the use of owner, versus appraiser valuation changes this relationship. The basic premise. of hedonic price analysis'is that there exists a reasonably well-,fitting relationship between 12 the price of the good in question and the characteristics of that good. In the most general functional form, this relationship can be represented as: (1) V f(C,C2,. . .,CN) where V is the price (or value) of the house and the C's are the character- istics of the house (number of rooms, lot size,etc.). The exact.relationship between the characteristics of housing and the price is not known. The standard hedonic method is to assume that this relationship can be expressed in the linear form: (2) V p0 + p1C1 2C2 NCN+ error terms. This equation can then be estimated using linear regression analysis (assuming that- the errors are randomly distributed and-are not correlated with one another). The coefficients of the characteristics can be interpreted as the shadow price of that characteristic, reflecting the interaction of demrander's bids and the . 13. supplier ls offer. . Characteristics of Squatter Dwellings: Squatter communities have been the subject of many studies. However, there has been relatively little done in quantifying the stock of housing in this informal setting. The RAD surveys were able to obtain a fairly complete list of the characteristics of each house. These characteristics, which will be used as independent variables in estimating (2), are described in Table 3. The first observation is that the average dwelling unit's character- istics do not appear to describe a temporary, makeshift shack, which is the stylized notion of a squatter unit. This observation is consistent with the earlier findings regarding housing value and is most clearly illus- trated by the average values of the "quality variables" of Table 3. The average age of the structure is almost 12 years, with a range of 2 to'30 years, attesting to the Tondo's ability to survive numerous threats of- wholesale eviction. Whether or not structural age is expected to be a proxy for quality and, thus, to vary inversely with its value, as in housing markets of developed countries, needs further discussion. Normally, newly built houses would command the highest prices. -If the rate of maintenance does not keep pace with the rate of structural deterioration, all other variables (sith as- neighborhood quality) constant, depreciation will occur 14 over time. However, age in a squatter community may be an indicator of. '"staying power" and the durability of the structure. In fact, since many. dwellings are built by self-help methods, progressive development is prob- ably prevalent. Households would build their dwellings slow4ly over time, . as their resources allow them. Thus, age sight point to a larger and/or higher quality house simply because a'household (or succeeding occupants) has had the time to put more work and materials into it. Thus, it is hypo- thesized that a positive relationship between age and value could exist for the squatter dwellings in the sample. More direct quality measures were also obtained by the survey. The next three variables of Table 3 describe the materials used in housebuilding. The variables CMNTWALL, FINWALL and SOLIDF take on the value of unity when t)he dwelling is characterized, respectively, by cement walls, finished (painted) outside walls and a concrete foundation. Otherwise, the value of the relevant variable is zero. The average values of these dummies indicate that 30% of the sample have cement walls, which tend to be the highest quality of materials used in these communities, since wooden walls usu- ally mean scrap planks. About 80% do not have finished walls and 70% . of residents do not have dwellings -built on soli'd foundations. Housing value is expected to be directly related to all three variables. The surveys also obtained a large amount of info=ation on the juality (i'..e.., the present condition) of the houses. Interviewers were asked to rate the dwellings from zero to three, depending upon certain structural conditions. For example, one variable was called "wear on floors". If a dwelling unit had a dirt floor, it was assigned a value of zero.. If' there was very substantial amount of wear, the variable was assigned a value of .one. Moderate wear was worth a value of two and no wear, three. There. are 16 othe; such variables in the data base. They are listed under . Tab le 4. Because the inclusion of all these variables would severely restrict the estimating equation' degrees of freedom, certain summary indices of quality had to be constructed. There is no reason to believe why these value judgments can simply be added up and, following Kain. and Quigley, 15 factor analysis was also used to combine them into indices which can be easily interpreted. Table 4 summarizes the three factor solutions for the 17 vari- ables. Two, four and five factor solutions were also attempted but they did not result in variable combinations which explained the variation of the components better nor were these combinations easily interpretable. The factor loadings indicate the correlation between the factor and the component variables. In Table 4, there are three groups of variables. Factor 1 10 can be termed the outside structural quality of the unit. It measures ther quality of the walls, windows, doors, roofs and foundations. It is called. QSTRUCT. Factor 2 loads heavily on variables which have to do with the quality of the steps. It is coded QSTAIRS. Finally, Factor 3 loads heavily on the quality of the floors and it is coded as QFLOORS. All three are expected to .be positively related to housing value. The second group of variables described in Table 3 are those which describe the size of the dwellings. Once again, the numbers indicate that the stylized view of the squatter unit is misleading.. The average lot size is approximately 61 square meters, which is not an insignificant size. This quantity is, of course, the size perceived by the owner since, there are no titles. However, in Tondo, titles are not necessary since neighbors appear to be aware of the boundaries of the lots, althcagh there are probably many overlapping claims. Houses of more than one story are also very-much in'evidence in.Tondo. Approximately half of the sample units have more than one story. Size is obviously expected to vary directly with value. The third group of variables of Table 3 indicate the 'existence of sanitary and water-related facilities in the squatter household. TOILET and SINK equal one when these facilities are respectively present in the dwelling unit. The differences between formal housing and squatter communities are most evident when one examines these variables. Over half of the families have no toilet facilities at all. A similar percentage does not have a sink which is a proxy for individual water connections. Finally RICH is a rough measure of neighborhood effects. This is a dummy variable which takes oa the value of unity of a household 's house is on 16 a superblock with an average monthly household income which exceeds th mean. 11 Subjective information regarding neighborhood conditions was not available. No measures of location were available. This possible source of bias is minimal because all of the dwellings in the sample were located within the Tondo area, which is fairly confined. A great majority of the household heads do not have.to commute to the work place. Results: Table 5 presents the main results of the hedonic equation estimation with the variables' described earlier. Linear and nonlinear specifications (with respect to age,.AGE2, and lot size, LOT2) are used and Beta coefficients, which standardize for the unit' measures, are also shown. The two dependent *variables are. OWNRVAL, which is owner 's own valuation of the worth of the dwelling unit, and CONSVAL which is the appraiser 's valuation. The overall results appear to be similar when a comparison is made between the coefficients of equations with different dependent variables although no exact econometric text is possible. The signs and magnitudes are roughly of the' same order of magnitude. Also the relative ranking of the contributions of the various characteristics to housing value is not changed by large amounts. The rank-order correlation (Spearmants rho) between, the beta coefficients of the nonlinear specifications of the equations using CONSVAL and OWNRVAL is .66. The signs of the coefficients conform to expectations, except for AGE2 and TOILET in the nonlinear specification of the owner's valuation. Although its significance level is relatively low, the positive coefficient for AGE suggests that progressive development and/or the lessened degree of risk signalled by age are having some effect. The sign is maintained throughout the regressions. This can be contrasted to findings for more developed ..countries where the age coefficient is consistently negative and significant.17 12. Housing quality variables seem to be the more important determinants of value. A concrete wall and some sort of finish on that wall enhances the value considerably. The beta coefficients for CMNTWALL and FINWALL consistently rank highly in the-equations. (Many studies in the U.S. found the same result for- paint' or a brick. exterior: see studies -cited in footnote 17.) In flood- prone Tondo, having a solid foundation is, not surprisingly, strongly correlated with value. The variables which account for the maintenance of a dwelling unit (QSTRUCT and QFLOORS) perform poorly in the owner equations but not in the appraiser equations. It may be, for example, that owners do not value. maintenance or. at least tend to associate it with the durability of materials used in the house. Although the price of the land in this squatter community is presumably nil to the households, the extent of the lot seems to. have a great deal to do with housing value. Like homesteaders, squatter families are very quick. to stake out a claim to the land and the lot sizes vary. The boundaries appear to be respected in Tondo, as long as they are within reason and the owner can justify the.claim by force or moral suasion. The evidence indicates that homeowners feel that an additional square meter of space adds 4 to 11 pesos to the value of the house.- It is interesting to note that the government's urban development project eventually. sold the land to the squatters at a highly subsidized rate of 5 pesos per square meter, as agreed over 15 years ago. Surprisingly, the variables which measure service levels do not enter into the equation very significantly. In particular, sanitary facilities do not appear to add much to explaining the variance in either the owner's or the architect's valuation. This may simply be a reflection of tastes, as conditioned by the relatively little familiarity these residents have with 13 18 toilet fixtures and their importance for sanitation. This might have grave implications for projects which provide these facilities. Greater attention may have to be paid to education components to. teach their value and use. Simply providing them may not be sufficient to deliver intended health benefits. V. Conclusions Squatter communities have not been included in the housing statistics of many developing countries. Yet they constitute a large portion of urban dwellers and the dwelling units are obviously not without value. It is important to estimate the determinants as well as the magnitude of those values to use in the evaluation of housing and urban. development projects or for eventual use in housing demand studies. This paper uses data from the Philippines to determine 4hether or not the squatter owner's valuation of his own house compares with that of an independent appraiser's estimate. The results show .that, while .the discrepancies in the estimates are quite large for individual properties,. they are largely offsetting for. reasonably sized samples. This tends to confirm earlier findings for the U.S. by Kish and Lansing. and Kain . and Quigley for conventional housing. In addition it is shown that the two estimates are highly correlated with one another and yield similar results when they are used as.dependent variables in a hedonic price equation. The determinants of the value of squatter dwellings tend to be similar to those of conventional "formal sector" dwellings. The external appearance and quality of materials used in construction are among the most important variables. Water and sanitary facilities may not be valued as much as expected in the market. Finally, there is preliminary evidence that age may be positively correlated with value, because housing services are improved more gradually in squatter communities and because longevity in a particular area is a sign of reduced risk. Overall, it can be stated that squatter housing markets appear. to behave as economically rational entities which. valuate dwelling units similarly to conventional markets. They should be accounted for in any analysis regarding housing markets in developing countries and it appears that. simple household surveys would be reasonably effective in obtaining the requisite information for such work. -~- 15 Footnotes * The author acknowledges .J. M. Bamberger, G. K. Ingram, D. H. Keare, K. S. Lee, J. C. Leith, D. Lindauer, J. F. Linn, S. Margolis and J. Quigley for comments. dn an earlier draft. Special thanks are due to Professor Mila.A. Reforma of the University of the Philippines and the staff of the Research and Analysis Division of the Philippine National Housing Authority for designing and implementing the collection of the data. Some estimates are as follows: Manila (22%), Jakarta (25%), Seoul (30%), Dakar (30%), Lusaka (27%), Rio de Janeiro (27%), Mexico City (40%). It should be noted that the source of -these figures, 0. F. Grimes, Housing for Low-Income Families (Baltimore: Johns Hopkins University Press, 1972),makes it clear that the definition of what constitutes a squatter community varies from country to country. We define squatter settlements as spontaneous (unplanned) agglom- erations of dwellings whose residents do not hold title' to the. land. Iowever, the residents may "own" the structures since they built them. 2 See, for example, Grimes, A. A. Laquian, Slums are for People (Honolulu: East-West Center Press, 1969), and .J. Perlman, The Myth of Marginality (Berkeley-: University of California' Press, 1976). . 3 Z..Griliches (ed.), Price Indexes and Quality Change (Cambridge: Harvard University Press,l1971). 4 The view that an improvement in benefits.will be partially capitalized in higher land and housing prices and the use of the hedonic techniques in ob- taining benefit measures are complicated issues- and are well discussed in the literature--see A. M. Polinsky and D. L. Rubinfeld, "The Long-Run Effects of a Residential Property Tax and Local Public Services," Journal of Urban Economics 5 (1978): 241-62, for a partial bibliography. 16 5- As i'n* A. D. Witte, H. J. Sumka, and H. Erekson, "An Estimate of a * Structural Hedonic Price Model of a Housing Market," Econometrica 47 (1980): 1151-73. 6 J., Kain and J. Quigley, "Owner's Estimate of Housing Value," Journal of the American Statistical Association 67 (1972): 803-6. 7 Kain and Quigley and L. Kish and J. Lansing, "Response Errors in Estimating the Value of Homes," Journal of the American Statistical Association 49 (1954): 520-38. 8 Data from developing countries are notoriously unreliable.. However, the data gathered for this analysis iere collected scrupulously. The same house- holds were also the subjects for RAD's income and expenditure survey, in which households recorded daily their incomes and expenditures under the supervision of an interviewer. These interviewers then built up trust within the community and were also used to administer the house consolidation surveys. 9 Kain and Quigley. 10 It should be noted that, in most cities of the developing world, there are two types of squatter communities. One type has the following characteristics: a large number of dwellings; a long history and struggle for services; a favor- able location on large tracts of government or private land. Another type of community typically consists of a small number of dwellings perched precariously along railroad rights-of-way, leaning alongside the wall of a building, in deep gullies and have no real community sense. Tondo is of the former, variety (Laquian). 17 11 Note that it is not clear in this case which of the estimates is the "right" one. Although the professional appraiser was a highly skilled struc- * tural engineer, she may not have been fully cognizant of the market forces which valuate dwellings in a squatter community. Thus, this paper has diverged from the convention of the earlier studies such as Kain and Quigley by not denoting the difference between the owner and appraisal estimates as "response error". However, for comparative purposes, a regression using income, sex of.head, the age of the house, and housing quality as explanatory variables were used to explain the discrepancies between owner and appraiser estimates. The dependent variables were absolute and relative discrepancies between owner . and appriased value estimates. The results (available from the author) re- vealed that higher income households tend to "underestimate" with respect to the appraised value of the structure. Sex enters insignificantly in the equa- tions. The coefficient for'age indicates that owner valuations deviate more ftom appraised valuations for older homes. Owners- also tend to overestimate for higher quality homes, which is at variance with -Kain and Quigley 's findings. 12See Griliches. .13 .See S. Rosen, "Hedonic Prices and Implicit Markets," Journal of Political Economy 82 (1974): 34-55. 14 After a certain point, urban renewal might occur, and the dwelling may.be upgraded again. But this would tend to occur only in very old homes (over 35 years) and is a fairly recent phenomenon in the central core of older American cities. . 15 J. Kain and J. Quigley, "Measuring the Value of Housing Quality,"' Journal of the American Statistical Association 65 (1970): 532-48. 18 16 Tondo has been divided into twenty-four "superblocks" of approxi- mately 700-800 families each to facilitate project implementation. ; 17 See M. J. Ball, "Recent Empirical Work on the Determinants of; Relative House Price," trban Studies-10 (1973): 213-31; A. T. King, "The Demand for Housing," in Household Production and Consumption, ed. Terleckyj (New York: NBER Studies in Income and Wealth, 1976); Kain and Quigley, "Measuring the Value of Housiing Quality"; V. Lapham, "Do Blacks Pay More for Housing," Journal of Political Economy 79 (1971): 1244-57; A. B. Schnare and R. Struyk, "An Analysis of Ghetto Prices over time," in Residential Location and Urban Housing Markets, ed. G. Ingram (Cambridge: Ballinger Co., 1977); and M. Stegman and H. Sumka, Non-Metropolitan Urban Housing (Cambridge: Ballinger Co.., 1977). .18 Housing characteristics and their hedonic prices can show demand pro- perties of traditional goods and prices only under certain conditions,, M. P. Murray, "Hedonic Prices and Composite Commodities," Journal of Urban Economics 5 (1978): 188-97. Table 1.: COMPARISON OF INDEPENDENTLY APPRAISED HOUSING VALUES WITH OWNER ESTIMATES Philippines 1979: Kain and Quigley 1972 St. Louis Kain and Quigley- 1972 St. Louis ._____Study (All Owner Occupied Homes): (Single Detached Homes Only): N. (1) Average Appraised Values Z C]l $1,879 $14,431 $14,719 i=l N (2) Average Owner Estimated [ 0 1] 1,886 14,473 14,488 Values i=l (3) Difference [(1)-(2)] -7 -42 .+271 (4) % Difference [(3)*(l)] -0.4% -0.3% - +1.8% - N (5) Absolute Value of 1 N C 0,] 1,027 3,058 2,825 Difference i=1 (6) Absolute Value of Difference as a 65.4% 21.2% -19.2% % of Mean Approved Value [(5)j((1)] (7) Pearson Correlation Coefficient .77 .87 .84 (8) Spearman Correlation Coefficient .78 (9) Sample Size 96 113 83 a th Ci= CONSVAL = value of the i home as estimated by consulting professional appraiser. th th 0 = OWNRVAL = value of the i home as estimated by the owner of the i home. b i b.Table 1 figures converted to US dollars at $US1 = Y7.5 Philippines pesos. Table 2: FREQUENCY DISTRIBUTION OF OWNER ESTIMATES AS A % OF APPRAISED VALUE a Philippines 1979 St. Louis 1972 US National 1954 Owner Occupiers Owner Occupiers Under 70% . . 25%, 12%. 6% 70 -89. . 8 24 20 .90 -109 .. 8 .26 37 S110-129 .. 9 22 19 130-149 6 8. 9 150-200 17 8 9 200 and over 24. . a a * Kish and Lansing.,. . ... ... Table 3: HOUSING CHARACTERISTICS IN THE TONDO AREA (N =96 HOUSES) Quality Variables: Meana AGEC . Age of the structure in years 11.6 (8.57) CMNTWALL % of dwellings with solid (cement or brick). 30 walls. (.47) FINWALL % of dwellings with wall finish (e.g., 17 paint) (.37) SOLIDF %'of dwellings with concrete foundations 13 (.46) b QSTRUCT Index of quality of the structure. .013 (1.003) QSTAIRS Index of quality of the stairs . .004 (.996) b QFLOORS Index of quality of the floors .014 (.995) Size Variables: LOTC Average lot size in square meters 60.7 (48.44) STORY Number of floors . 1.5 (.50) Facilities Variables: TOILET % of dwe*llings with bucket-flushed or 44 other water-sealed toilet . (.50) WATER % of dwellings with sink (and water . 54 connection) (.50) Neighborhood Variable: RICH % of dwellings in neighborhoods 45 (superblocks) with monthly incomes above (.50) a . Standard deviations in parentheses. b See text for explanation of how these variables were formulated. c2 2 AGE2 = (AGE) and LOT2 = (LOT) Table 4: FACTOR LOADINGS ON INDIVIDUAL QUALITY VARIABLESa Variables Factor 1 Factor 2 Factor 3 VARO63 Wear on floors 0.94394 VAR064 Defects floors - 0.94613 VARO65 Sagging or .bulging floors - 0.91645 VAR066 Wear on steps - 0.93917 VARO67 Defects on steps - 0.95275 - VARO68, Unsafe steps or railing - 0.92148 - VAR069 Defects on. ceiling. * VARO70 Defects on in walls 0.69608 - VARO71 Broken, missing window panes 0.80596 VARO72. Rotted, loose window frame 0.79213 - VAR073 Deep wiar on doorsill, frames 0.76212 - VARO74 Sagging, bulging outwalls 0.80229 - VARO75 Defects on outwalls 0.80669 - VARO76 Sagging, bulging roof 0.61832 - VARO77 Defects on roof * 0.51990 - VARO78 Defects on foundations 0.64005 VARO79 Bad gutters and downspouts a "-"'indicates a standardized factor loading less than .5. Table .5 HEDONIC PRICE EQUATIONSa DEPENDENT VARIABLE = OWNRVAL DEPENDENT VARIABLE = CONSVAL A p.. p. p. CONSTANT -12505.18** -13800.29** -11760.84** -16006.81** (6074.47)- (8088.07) (5938.35) . (7922.52) AGE . 252.38* .13 132.20 . .07 11.13. .01 504.18 .22 (186.65) (717.89) . (182.47) (703.19) AGE2 2.74 .05 -15.55 -.23 (21.36) (20.93) CMNTWALL 9755.09** .27 10253.52** .29 7717.80** .18 8295.38** .19 (4033.39) (4131.88) (3943.01) (4047.30) FINWALL 7562.89** .17 -6102.44* .14 16998.20** .32 16760.39** .31 (4586.22). (4843.03) (4483.45) (4743.90) QSTRUCT 967.98 .06 1059.93 .06 1840.39* .09 1755.75* .09 (1543.07) (1561.48) (1508.49) (1529.52) SOLIDF 8820.92** .18 7928.54** .17 9294.42** .16 9088.35** .16 (4701.87) (4814.32) (4596.51) (4715.78) -QFLOORS 1406.43. .09 .1283.93 .08 3634.43** .18 3567.47** .18 (1582.58)- (1597.70) (1547.12) (1565.00) QSTAIRS 1018.30 .06 933.11 .06 927.39 . .05 826.87 .04 (1494.18) . (1509.33) (1460.70) (1478.43) LOT. 4.30* .13 10.67* .32 10.63** .26 13.10** .32 (2.97) (7.15) (2.91) (7.00) LOT2 -.001 -.20 . -.001 -.06 (.001) , (.002) STORY 5611.96** .17 6139.81** .19 7380.79** .18 7650.00** .19 (3326.67) (3393.49) (3252.12) (3324.03) TOILET 587.15 .02 -170.15 -.01 1242.75 .03 1087.00 .03 (3510.69) (3612.87) (3432.03) (3538.92) WATER 6696.91* .18 5647.51 .15 1167,66 .03 938.15 .02 (4254.75) (4408.05) (4159.40). (4317.82) RICH 5035.87** .15 4950.44*' .15 1172.20 .03 2171.05 .05 (2802.67) (3119.79) (2739.86) (3055.93) 2 R. .51 .52 .69 .69 N 96 96 96 96 F 7.21 6.18 15.21 12.88 .a Standard errors are in parentheses Significant at .10 confidence levels (two-tailed test) ** Coefficients larger than standard error P's.are coefficients of standardized variables.