IigW/é/K/g/li/W/WW/W/l/W/fll This is to certify that the dissertation entitled AN ANALYSIS OF A LOW INCOME HOUSING MARKET IN URBAN ZAMBIA presented by Manenga Chilala Ndulo has been accepted towards fulfillment of the requirements for Ph.D. (mgflfin Economics Major pr essor Dme May 11, 1983 M5 U is an Affirmative Action / Equal Opportunity Institution 0-12771 Ll “ . 2% Y Miami-3m fimte ‘ L University _~_‘ W “'7 I' MSU LIBRARIES v RETURNING MATERIALS: Place in book drop to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. 74 AN ANALYSIS OF A LOW INCOME HOUSING MARKET IN URBAN ZAMBIA by Manenga Chilala Ndulo A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Economics 1983 /37~c97o’27 ABSTRACT AN ANALYSIS OF A LOW INCOME HOUSING MARKET IN URBAN ZAMBIA by Manenga Chilala Ndulo This dissertation studies economic behavior and resource allocation in a low income urban housing market in Lusaka, Zambia. It specifically addresses itself to how housing consumption is affected by income and household demographic characteristics, to the analysis 50 19.1% mean (years) 40.93 (0.90) Gender of household head: male 90.4% female 9.6% Schooling completed by household head (years): < 8 62.4% 8-12 33.8% > 12 3.8% mean (years) 5.97 Size of household: mean number of members 7.24 (0.24) Number of renters or lodgers: mean 0.26 (0.05) NOTE: Standard errors are in parentheses. SOURCE: M. Ndulo, 1979 Zambia Home Improvement Survey. 60 when they are stabilized, they are followed by female members of households. Ninety percent of the owner-occupant households surveyed for the present study were headed by men; 10 percent were headed by women. However, there is evidence that the proportion of female household heads is increasing. Studies by the LHPET in George found that the prOportion of household heads who were female increased from 5.4 percent in 1973 to 8.1 percent in 1976 (Singini, 1978, p. 7). Another study of George, one year later, found the pr0portion of female-headed house- holds to be 9 percent (LHPET, 1977a, p. 3). This trend might be due to the increasing independence of women and the increasing number of women born in urban areas, who look upon urban areas as their permanent home. Education The average head of an owner—occupied household had spent about 6 years at school. Sixty-two percent had spent less than 8 years at school; 34 percent, between 8 and 12 years, and 4 percent, more than 12 years. Twelve percent had not attended school; and 88 percent had had one or more years of schooling. The standard of schooling among the household heads is high compared with that in mbst African countries. This fact may indicate that those who have attended school are more likely to migrate to urban areas. 61 Size of the Household The size of a household is an important factor in how much housing it consumes. A larger household means more demand for space. It is likely to have a larger disposable income. There are likely to be more members working, or more older children who can help with income generating household activities. But at giygn incomes, demand for space usually falls as size rises. The average household's population in our sample of owner-occupants was 7. This is a slightly higher level than has been reported by other studies, which have estimated the mean household size to be 6.6 members in Chunga; 5.4 in Kaunda Square; 5.2 in John Howard; and 4.7 in George (LHPET, 1976a, p. 8). These figures are estimates for a sample of owner-occupant and tenant households; our estimate is for owner-occupant households only. It is possible that owner-occupants have larger households than do tenant occupants. What is the composition of the typical household? Eighty-three percent of the owner-occupant households in the present study's sample had more than four members, while 17 percent had fewer than five members. The typical household consisted primarily of members younger than fourteen years of age. An average family had four members younger than fourteen years and three members fourteen or more years old. Sixty percent of the households had four or more members younger than fourteen. Eighty-five 62 percent had one or more children under five, while fifteen percent had no children under five. The preponderance of children in low income housing areas was also evidenced by a study which showed that 48 percent of the population of George was children under fourteen years old. Individuals over 45 years old were only 5 percent of the population (Singini, 1978, p. 7). Some households take in lodgers, renters, or relatives. The survey conducted for the present study showed that 19 percent of the households had lodgers or renters. A study of George found that 60 percent of the households consisted of nuclear families, and 23 percent had relatives staying with them (Singini, 1978, p. l). A large household size implies overcrowding. A survey conducted by the National Housing Authority shows that there is less overcrowding in site-and-service areas than in public-housing areas, partly because owner—occupant households can enlarge their homes or move out in response to increased household size (Simoko, 1979, p. 2). Public housing cannot be enlarged because this is in the final analysis the responsibility of the landlord. Moving out is less likely because this is done through an administrative process which is not easily responsive to such needs. Household Mobility The average household head in our sample had migrated to Lusaka during the past decade, a fact reflecting the huge surge of rural-urban migration during the post-independence 63 era. Only 6 percent of the residents had been born in Lusaka; 94 percent had not. This finding is similar to those of other studies. A 1976 LHPET study of George found that 2 percent of the sampled household heads were born in Lusaka; 53 percent had come to Lusaka from rural areas and 43 percent had come either from other urban areas or from outside Zambia (Singini, 1978, p. 10). Data collected for the present study does not show which households had migrated from rural to urban areas. However, figures from the George survey (Singini, 1978) suggest that about 50 percent of urban poor households had recently migrated from the rural areas. Thus, migrants are equally likely to have migrated directly from the countryside to Lusaka or to have moved first to other urban areas before arriving in Lusaka. Studies of households in low income residential areas in Lusaka show that the most common pattern of .migration is for the migrant initially to stay with relatives until he or she finds a job. The person then starts looking for accommodations, and usually rents a room in the area for about $13 (UN/ECA—Bouwcentrum, 1973, p. 111-44; Mubanga, 1979b, p. 17). The pattern of migration just described is consistent with the migration literature. However, it is likely to change over time as the country grows and the rate of migration to urban areas moderates and then stabilizes. 64 The average household in the survey conducted for the present study had been living at the current site for the past six years (standard error: 0.4). Given the average age of the housing areas, this implies much inter- and intra-urban mobility, with households moving within their residential areas and outside of them to other residential areas. Within the urban areas, the predominant trend has been for low-income and middle-income households to move from council and public housing to private low income housing areas. This trend has been spurred by a variety of factors-~among them, high rents and a shortage of housing units. A study in 1973 found that Matero, a public housing area near George, was the most common area of previous residence for George households (Singini, 1978, p. 3). A LHPET study of Chawama found that 64 percent of the households with houses and plots in Chawama Overspill had come from outside Chawama, while 78 percent of those buying houses within the upgraded area of Chawama had come from within Chawama (Banda, 1979b, p. ii). The figures probably reflect two main kinds of moves: an influx of presumably higher-income households from outside, especially from public-housing areas, whose main motivation is ownership of a house; and the motion of a group of households which are mobile within the residential area. Table 3.8 shows the principal reasons why the 65 TABLE 3. 8 PERCENTAGE DISTRIBUTION OF THE MAJOR REASONS WHY HOUSEHOLDS MOVED TO THEIR PRESENT LOCATIONS Reason HNdlIinsbeehro 1°de PeHrofiiesnechgled 50f To be closer to work 8 5.13 To be in a better neighborhood 10 6.41 To pay less 3 1.92 To become a homeowner 76 48.72 fro obtain a bigger or 12 7 7 better-quality home ’ Othera 47 28. 21 aIncludes resettlement. SOURCE: M. Ndulo, 1979 Zambia Home Improvement Survey. 66 households surveyed for the present study moved to the present area. The main reason given was to become an owner: 49 percent of the households moved for this reason. The other important reasons were to obtain a bigger or better- quality dwelling (8 percent); to find a better neighborhood (6 percent); and to be closer to work (5 percent). A similar study of recent movers in Chawama suggested the same reasons. Asked for the main reason why they bought houses in Chawama, respondents named becoming an owner (51 percent); obtaining a bigger dwelling (26 percent); and being closer to work (2 percent) (Banda, 1979b, p. iii). Only about 10 percent of the owner—occupants surveyed for the present study said they were thinking of moving again. The main reason given by those household heads was to obtain a bigger and better-quality dwelling. Structure of Income and Employment A household may depend on various sources for its disposable income, and these sources may change from time to time, depending on the position in the life cycle of the household. Some households produce for their own consumption. In the urban areas, the major source of income is labor earnings, particularly from.work in the small scale production sector. Some households also receive returns from assets owned, or public or private transfer payments. The importance of these different sources of incone 67 is likely to depend on the level of development in the country and on the standard of living of a household. In this section, the structure of household incomes, and occupations of owner-occupant households, is explored. Occupation of the Household Head Table 3.9 shows the occupations reported by owner- occupant household heads in our survey. Forty-eight percent were employed as either unskilled or skilled manual workers. The third most important occupation was self- employment--ownership of a store or other business: 12 percent of the household heads ran a store or other business. Self-employment has been found to be common in low income residential areas. For example, a study of George and Chawama produced the estimate that 13.4 percent of the heads of households in George were self—employed, and 18.6 percent of the work force in Chawama was self-employed (LHPET, 1977d, p. 9). Domestic service is one of the most common occupational areas among poor households in Lusaka; in our sample, however, only 2.5 percent of the heads of households were employed in domestic service. This low proportion can be explained by the fact that domestic servants often receive housing at their places of work--usua11y behind the employer's house. Furthermore, if they reside away from work, they are likely to be renters because they have lower incomes than do people in most other urban occupations. 68 TABLE 3.9 PERCENTAGE DISTRIBUTION OF THE OCCUPATIONS OF HOUSEHOLD HEADS . Number of Percentage of Occupation Households Households Salesperson; vendor 7 4.5 Owner of store or other business 18 11’5 Police, military, or 14 8 9 other personal service ’ Domestic service 4 2.5 Unskilled worker 37 23.6 Skilled worker 38 24.2 Office worker 12 7.6 Professional, technician, l4 8 9 or foreman ' Other 13 8.3 TOTAL 157 100.0 SOURCE: M. Ndulo, 1979 Zambia Home Improvement Survey. 69 The impression generated by a review of the results of the present study's survey is that most heads of house- holds are employed. However, there may be a high rate of unemployment among other members of the household: 32.5 percent of the households had one or more unemployed workers, while 67.5 percent had none. Other studies carried out in low income residential areas generally show a high level of employment among the heads of households and a high level of unemployment among adults in general. A LHPET study showed that, among heads of households, an estimated 94 percent were working, 3 percent were seeking work, and 3 percent were neither working nor seeking work. Among the total population of adults, however, it was estimated that 45.8 percent were working, 3.6 percent were seeking work, and 50.6 percent were neither working nor seeking work (LHPET, 1976a, p. 10). Another author has reported an estimate that, among the unemployed, 64 percent had previously had paid employment in Lusaka. Periods of unemployment were varied; Among the unemployed, 52 percent had been unemployed for less than two years, 12 percent had been jobless for more than two years, and 36 percent had never had a job (Rakodi, 1978, P- 7). 70 Incomes The percentage distribution of the sources of income for owner-occupant households in our sample is shown in Table 3.10. Labor earnings were the dominant source of income: 85 percent of the households depended in part or entirely on wages and salaries. This figure reflects the importance of work in the formal sector for the urban poor. The second most prevalent category of income source reported in the survey was sales or work on one's account: 26 percent of the households earned this type of income. Rental income from lodgers in the building was also important: 15 percent of the households reported this type of income. Aid in cash from family members, pensions, and other aid were less commonly reported. Aid from other members of the family might be low because many urban residents, though poor, are nonetheless likely to be better off than are their rural relatives. It is the urban dwellers, by and large, who are expected to send remittances to rural family members. However, the urbanites might receive transfers from better-off residents in other urban areas. The pattern of the income sources shown in this sample is similar to that reported by other studies. For George residents, it has been found that, apart from wages and salaries, the most prevalent sources of income were self- employment (especially retail trade and the renting out of 71 TABLE 3.10 PERCENTAGE DISTRIBUTION OF THE VARIOUS SOURCES OF INCOME FOR OWNER-OCCUPANT HOUSEHOLDS Number of Percentage of Source Households Households Wages and salaries 134 85.4 Sales or work on one's account 40 25.5 Rent from lodgers 23 14.6 Aid in cash from family members not living in the same home Pensions, other aid 4 2.58 SOURCE: M. Ndulo, 1979 Zambia Home Improvement Survey. 72 rooms) and allowances from relatives (Rakodi, 1978, p. 4). The distribution of monthly disposable income levels in our sample of owner-occupant households is shown in Table 3.11. The mean monthly household income is $162.50 (standard error: $12.27). This is above the averages found in other studies: $128 for a survey of Chuunga, Kaunda Square, and John Howard (LHPET, 1976a, p. 13); and $116 for Chawama (Banda, 1979b, p. 10). It is higher because more care was taken to cover all income sources and because of concentration on owner-occupants. Home Improvement In this section, home improvement—-the process by which the household extends, improves, subdivides, and rebuilds the dwelling-~is examined. This is one of the main means of enlarging the stock of low-income housing in Zambia. Number of Home Improvements Data from our survey indicates that most low—income, urban, owner-occupant households improve their dwellings. The distribution of improvements carried out by the households is shown in Table 3.12. Ninety-five percent carried out one or more types of dwelling improvements; only 5 percent did not. The mean number of types of improvements implemented was 4.3 per household (standard error: 0.22). The most popular kinds of improvements were adding a room, improving 73 TABLE 3.11 PERCENTAGE DISTRIBUTION OF MONTHLY PERSONAL DISPOSABLE INCOME FOR OWNER-OCCUPANT HOUSEHOLDS I nc ome Number of Percentage of Households All Households < $51 9 5.7 $51-100 56 35.7 $101-150 26 16.6 $151-200 24 15.3 $201-300 24 15.3 $301-—400 11 7.0 > $400 7 4.5 SOURCE: M. Ndulo, 1979 Zambia Home Improvement Survey. 74 TABLE 3.12 DISTRIBUTION OF HOME IMPROVEMENTS CARRIED OUT BY OWNER-OCCUPANT HOUSEHOLDS Number of Types Number of Percentage of of Improvements Households All Households 0 8 5.1 l-2 40 25.1 3-»5 57 36.3 6+ 52 33.1 TOTAL 157 100.0 SOURCE: M. Ndulo, 1979 Zambia Home Improvement Survey. 75 floor and roof materials, adding plaster and paint, and improving the toilet (see Table 3.13). By the end of the average period of residence (about six years per house), these home improvements had helped increase the value of the house by about 95 percent over its original value, as confirmed by respondents' evaluations of their dwellings: 78 percent of the respondents said that their dwellings were now better than when they had acquired them, 17 percent believed they were Of the same quality, and only 5 percent believed the homes were worse than when they had acquired them. Forty-two percent of the respondents said they had added one or more rooms. The average cost of doing so was $334.00 per room (standard error: $58.00). Households added rooms for various reasons, among them the presence of additional children (72.3 percent) and of other additional relatives (13.8 percent), and investment purposes (4.6 percent). Findings Of other surveys also indicate that most low-income households, given Opportunities and resources, improve their dwellingsn//One such study found that 67 percent of the households in Kaunda Square, 50 percent in John Howard, and 20 percent in Chuunga were at the time making extensions to their dwellings (LHPET, 1976a, p. 21). Another survey, in George, found that 39.8 percent of the improving owner-occupants had made extensions to their dwellings; 29.1 percent had improved walls; 25.8 percent had improved 76 TABLE 3.13 MAJOR HOME IMPROVEMENTS MADE BY OWNER-OCCUPANT HOUSEHOLDS Type of Improvement fifigfifiinaf PeISSEOSEHgOf Households Additional rooms 66 42.0 Better kitchen 47 29.9 Better toilet 79 50.3 Plaster and paint 87 55.4 Wall materials 45 28.7 Roof materials 89 56.7 Flooring 89 56.7 Windows and doors 61 39.1 Earth fill 55 35.5 NOTE: Other improvements included adding a porch, ceiling work, fencing, installing a more conven- ient water source, etc. SOURCE: M. Ndulo, 1979 Zambia Home Improvement Survey. 77 roofs; 8.9 percent had rebuilt their dwellings; and 10.4 percent had made other kinds of improvements. Furthermore, 54.8 percent of those who had extended their dwellings had added more than one room (Singini, 1978, p. 19). Intention to Improve Even if households do not improve their dwellings in the end, the intention is nearly always there. Failure to improve the dwelling is attributable to limited resource levels--both time and money. Even households of the most modest means are likely to carry out inexpensive improvements with their limited resources. Studies confirm this analysis. Asked about their future intentions regarding improvements, 93.2 percent of the participants in the Lusaka Housing Project expressed intent to extent their homes (LHPET, 1977c, p. 10). Another study shows that households which had recently moved to Chawama Overspill bought. or acquired half—complete structures, and all intended to improve the structures in one way or another; indeed, within a few years, 93 percent had already made some sort of improvements, spending an average of $749.37 per household (Banda, 1979b, pp. ii, 7). In summary, then, the general desire for improving dwellings has led households, even in housing areas with low incomes, to undertake improvements to their dwellings, with cheaper materials when necessary. 78 Decision to Improve For households that intend to improve their dwellings, much depends on the resources available to the household-- both time and money--as well as the household's ability to reduce other expenses in order to save money for dwelling improvements. While households carry out improvements to keep dwellings from depreciating or to increase the flow of services, there are also "non-economic" flows. A quality dwelling provides status, pride, community respect, and other such intangibles, which by themselves are enough to motivate many households to improve their dwellings. However, households renting their dwellings have no incentive to improve. There is also little such incentive in an owner-occupant household if the flow of services from the dwelling is satisfactory, if the household intends to move soon, or if it has no room to extend the dwelling. (Singini, 1978, p. 5). How do institutional changes, such as the provision of basic services, affect the incentive to carry out home improvements? Respondents were asked about the influence of the household's access to piped water on dwelling improvements. Fifty-eight percent of the respondents with access to piped water said they had made additional improve— ments to the dwelling as a result of such access; 34 percent had not made any additional improvements; and 8 percent had made fewer improvements as a result. The same question was asked of households without 79 access to piped water. For 74 percent of the respondents, the lack Of piped water had no great effect on improvements; 24 percent made fewer additional improvements because of this lack; and 2.5 percent made more improvements as a consequence of the lack Of piped water. What is clear from this is that access to piped water does affect dwelling improvements. The strength of this relationship is not clear. Financing for Dwelling Improvements Since low-income households have limited incomes, and the Opportunities for increasing that income are limited, amassing resources for improving dwellings is a formidable task which requires ingenuity and foresight. Financing is needed to buy materials and pay labor for extending or improving dwellings. Sources of loans and other credit are scarce; households depend mainly on accumulated savings or on wages. The percentage distribution of the various sources of finances for dwelling improvements in our sample is shown in Table 3.14. In all, 47.6 percent of the respondents paid cash for materials and used self-help labor, while 42.9 percent used accumulated savings for materials and paid labor. Only one respondent, a landlord, said he used rental income to finance dwelling improvements; for other households, no clear distinction between wage income and rental income is made when it comes to spending decisions. Only 3.4 percent of the respondents used loans TABLE 3.14 PERCENTAGE DISTRIBUTION OF THE SOURCES OF FINANCING FOR DWELLING IMPROVEMENTS Number of Percentage of Source of Finances Households All Households Financed by owner (in rented dwellings) Cash paid for materials; self-help labor used 70 ' 47'62 Credit from materials supplier, or loan for materials from 7 5.44 others; self—help labor used Hired labor and materials, financed with savings or by 63 42.86 sale of prOperty Loans for everything obtained from formal loan sources SOURCE: M. Ndulo, 1979 Zambia Home Improvement Survey. 81 from formal sources to finance their dwelling improvements. The most important source of finance for dwelling improve- ments, therefore, is accumulated savings and wages. Most households were unwilling to take out mortgages to finance either buying houses or making improvements. Households refuse to risk loss of a house if loan payments cannot be made. When respondents were asked whether they would be willing to mortgage their houses in order to get a loan, 71 percent said they would not; only 29 percent would. Even those low-income households wishing to take out mortgages are generally closed off from formal sources of finance. The Zambia National Building Society has a minimum mortgage of $6,329.00 for people under age 36 who can qualify for the longest mortgage term, 30 years. The person taking such a mortgage would need a monthly income Of about $190; the loan would be 100 percent of the value of the dwelling. These conditions completely cut off a big segment of the low-income population. Furthermore, the administrative process of getting a loan is cumbersome and difficult (LHPET, 1977d, p. 17). Summary and Conclusion In this chapter, the sample survey data was examined, with a focus on the prominent features of the dwelling, the owner—occupant household, and the home—improvement process. The data from the survey conducted for the present study 82 was also compared with results of other sample surveys, which produced findings generally comparable to ours. However, since our pOpulation is the low income urban population in low income private housing markets, our sample is unlikely to be representative of the rest of the nation. The average household in our sample is large, with about seven members. Most of these family members are children. The head of the household is usually a man. The typical household head was not born in Lusaka; he migrated either from a rural area or from another town. He is usually middle aged, has had about six years of schooling, has wage employment, and walks to work. The average disposable income of the average household is about $162.50, most of which comes from wage earnings. The typical dwelling in our sample is worth about $2,793.00, has three rooms, and a floor space Of 97 square meters. The household either bought or built it. If the latter, the household most likely hired labor to construct the dwelling. The walls of the typical dwelling are made of concrete blocks or burnt bricks; the roof is made of asbestos or metal sheets. For their source Of water, most dwellings use a public standpipe, and for sanitary facilities, they have pit latrines. The average household did carry out improvements to its dwelling--for the most part, four different types per home. The major source of financing for this work was 83 accumulated savings and wage income. One conclusion emerges clearly from this analysis of the data--that time and income are constraints on household activities. In the next chapters, the data analyzed in this chapter is used to develop a model of the low income housing market. 84 CHAPTER FOUR PATTERNS OF DEMAND FOR HOUSING Introduction In this chapter, the housing market surveyed for the present study is analyzed for patterns of demand. The focus is on demand for the housing commodity as a whole. The objective of this analysis is the estimate the response of housing consumption to households' income and demographic characteristics. Also examined are differences in the consumption patterns of households with different preferences, achieved by disaggregating the population into a series of somewhat homogeneous cohorts and estimating demand relation— ships for each group. Methodological Framework A perfectly competitive housing market is assumed. The housing commodity consists of a bundle Of different attributes; however, in this chapter, the demand for housing as a homogeneous good is studied. In other words, the examination focuses on the demand for the composite housing good, given its heterogeneity. (Olsen, 1969; Quigley, 1979, p. 396). 85 We derive the expenditure demand function for housing from the theoretical model described in Chapter One. Each individual household maximizes its consumption subject to budget constraints. Following Muth (1969), we assume a general utility function, U = U (xi) (4.1) n Y = i Pixi . (4.2) where xi = individual commodity, pi = price of individual commodity, and Y = disposable income. For the point-of—utility maximization, we have the Lagrangian and the first-order conditions: n L = U (xi) - )(.{ pixi - Y) (4.3) i=1 dL _ dU _ _ HE? ‘ JET A Pi ' 0 (4'4) 1 1 dL __ _ _ ET - Xpixi Y — 0 (4.5) If we assume that both the first-order and the second- Order conditions of the Lagrangian are satisfied, we can derive a general expenditure demand function: a. = pixi = di (Y, pi) (4.6) 86 Specifically for our housing commodity, we have: H = plxl = H (Y, 0) ' (4.7) where x = housing commodity, p1 = price of house, 1 Y = disposable income, and a = demographic variable. ‘Equation 4.7 is the expenditure demand function for housing. Each household's housing expenditure (plxl) is functionally related to the household's disposable income (Y) and to a set of household demographic variables (a). Our estimated demand function relates to the demand for the composite housing (x1), given its heterogeneity. Empirical Specification To examine the patterns of demand for housing, the expenditure demand equation (4.7) is estimated using the logarithmic linear functional form. This functional form has been found to be convenient in empirical work because it permits elasticities of housing expenditures to be estimated by use of ordinary least squares estimation (Boyes and Gerking, 1980, p. 287). It also yielded a better fit with earlier experimentation with the data (Ndulo, 1981). The value of the house is regressed on income and demographic variables: household size, gender, age, and the location of the household. The value of the house is used as a proxy for the total housing expenditure of the household. The empirical formulation of the model to be estimated takes the form: 87 LnINCi + 8 LnHSi + B D 2 3 1i (4.8) where H is the value of the house. The continuous variables are INC, monthly income of the household; and HS, the equivalent household size. There are also five dummies for the discrete variables: the gender and age of the head of the household, and the neighborhood location of the household. 6 is the random error. Empirical Results The value of the house is regressed on income without demographic variables and, separately, on income with demographic variables. The best estimates of the two regression equations are shown in Table 4.1. The dependent variable is the value of the house. This is the asking price for which the house can be sold on the market, as estimated by the household. Income and Household Size Monthly family income is an estimate of the disposable income currently available to the household from various sources. Most households in our survey reported that wage income was the most important component of the family's disposable income. For those households which include people who work in the public sector, wage income is likely to be stable over a long period: Public-sector employment is 88 TABLE 4.1 REGRESSION OF PROPERTY VALUE ON INCOME AND DEMOGRAPHIC VARIABLES INDEPENDENT VARIABLE Estimated Coefficienta Without Demographic With Demographic Variables Variables Income 0.810b 0.56b (6.098) (5.47) Household size 0.60b (3.79) Gender of head of -0.05 household male (0.22) Age of household head 30-50 years -0.14 (0.69) 50 years 0.18 (0.71) Location b Chawama 1.09 (6.55 Kaunda Square 2.09 (10.50) b b Constant 3.241 2.39 (5.260) (4.56) Sample size 155 155 ‘82 0.19 0.57 F 37.18 29.62 SOURCE: Calculated from data in M. Ndulo, Improvement Survey. 1979 Zambia Home at-statistics of coefficients are in parentheses. bcoefficient significantly different from zero at the 0.01 level. 89 more or less permanent because of employment regulations. This implies that current disposable income is likely to be used as a basis for decisions with a long-term horizon, such as housing decisions. However, public-sector employment is likely to be unimportant for the majority of low-income households. It is theoretically accepted that permanent income is one of the major determinants of demand for housing (Quigley, 1979, p. 396), based on the assumption that households look beyond their current-planning-period income in making demand choices. With perfect capital markets, households can borrow against their future incomes and spread out consumption Of housing services over their horizons consistent with their permanent incomes. In making improvements to the house, however, transitory income can be a factor. A sudden windfall can contribute to the marginal decision to improve the house. For elderly households, such windfalls--for example, dowry--are likely to be substantial. Recently, Goodman and Kawai (1980) urged that transitory and permanent income components be included as separate independent variables. They argued that such a separation of measured income (into transitory and permanent components) in the housing demand regression equation should substantially improve estimation power. However, for the purposes of the present study, monthly 90 measured income, rather than permanent income, is used in the regression equation. Empirical studies employing some concept of permanent income in the estimation of the income elasticity of demand for housing have reported a higher elasticity than have studies using measured income, because Of the fixity and the high transaction costs Of housing demand choices, in the short run. Con- sequently, the estimation of income elasticities in the present study is expected to be biased downward (Quigley, 1979, p. 396). The number of equivalent adults per household is used as a measure of the size of the household. This number was found by giving one point to any member of the household who is Older than fourteen years, and half a point to any member of the household yOunger than four- teen years, under the hypothesis that the number of equivalent adults per household will have more influence on housing demand than would alternative household size formulations. This hypotheSis was tested in regressions with the sizeof the household and the ratio of household members over fourteen to those under fourteen. The formu- lation with equivalent household size performed better. Equivalent household size is looked at in order to es- timate in more detail the effect of new adult family mem- bers. In this way, the effect ofhaving relatives from rural areas joining urban hOuseholds can be discerned. The estimated income elasticity of demand for 91 housing is 0.6. This means that if income increases by 10 percent, the demand for housing will increase, over time, by six percent. The estimated coefficient is sig- nificantly different from zero at the 0.01 level. This result is similar to that found elsewhere, i.e., that the income elasticity is less than one. For example, in Colombia, the income elasticity of demand for owners has been estimated to be about 0.8 (Ingram, 1980, p. 23) and for South Korea, 0.21. (Follain, Lim and Renaud, 1980, p. 330). This compares favorably with estimates from similar studies in the United States. These range between 0.2 and 0.5. (Mayo, 1981, p. 97). The demographic coefficient for equivalent household size is 0.6. It is also significantly different from zero at the 0.01 level. The coefficient is positive, and corresponds to a priori expectation. The estimated coefficient means that for everytxuipercent increase in the equivalent household size, holding everything else constant, there is a six percent increase in the demand for housing. A ten percent increase in equivalent household size means either a ten percent increase in the number of older members or a twenty percent increase in the number of younger members in the household, or some combination thereof. (Recall that an older member is worth a point and a younger member is worth one-half point in calculating the equivalent household size.) Therefore, an increase in the number of adults in the household raises the demand for housing at a much higher rate 2 . T‘.‘ ’17 1‘12. '1)! .3101- t) (i ’_ Do. (1’ 'I‘ll. I II. .‘I I 1})! I! 2.)] $11115 I . 3(111‘ lil‘ II‘,‘(| (li‘ ‘ I I. . II 4. \‘I‘l‘I‘Jl‘lill 11%,}j ) 92 than the same increase in the number of children in the household. An increase in the number of adults in a household is likely to result primarily from children getting older and relatives from rural areas joining urban households. These factors have more serious implications in terms of the demand for housing than does the birth of additional children, which is less serious until the children pass the age of fourteen. Differences in the consumption patterns of households whose tastes differ are examined by use of separate regressions for three household size cohorts and two income cohorts. The results are shown in Table 4.2. For the household-size cohort, the behavior of the constant term and the income elasticity coefficient shows a bowl-shaped curve. The constant term for small households is large; it decreases for middle-sized households, and then increases for large- sized households. On the other hand, the income-elasticity coefficient is low for small households, increases for middle-sized households, and then increases slightly again for large-sized households. Analytically, it can be said this behavior pattern implies that small households make large base expenditures for housing. This leads them to devote larger portions of increases in income to non-housing consumption expenditures. With a ten percent increase in income, everything else constant, households will devote only five percent to housing. Middle-sized households make lower base expenditures for 93 TABLE 4.2 HOUSING DEMAND CHARACTERISTICS FOR OWNER-OCCUPANT HOUSEHOLD COHORTS Estimated Parametersa Household Cohorts 2 80 81 R N b b All households 3.24 0.81 0.20 155 (5.26) (6.10) Household size b 1-3 persons 4.08 0.45 0.11 15 (2.61) (1.27) 4-6 persons 2.86b 0.85b 0.22 48 (2.58) (3.62) > 6 persons 3.22b 0.87b 0.21 92 (3.93) (4.90) Income b Above median 1.98 1.04 0.12 76 (1.19) (3.24) Below median 1.81. 1.18b 0.14 79 (1.33) (3.51) Gender b b Male 3.03 0.85 0.21 140 (4.66) (6.14) Female 4.61b 0.52 0.07 15 (2.10) (1.03) Life cycle stage (household head) b < 30 years 2.52 0.92 0.23 21 (1.39) (2.36) 30-50 years 3.44b 0.76b 0.19 104 (4.80) (4.96) > 50 years 1.75 1.238 0.27 30 (1.05)' (3.24) SOURCE: Calculated from M. Ndulo, 1979 Zambia Home Im- provement Study data. at-statistics Of coefficients are in parentheses. b Coefficient significantly different from zero at the 0.01 level. 94 housing and devote larger portions of increases in income to housing expenditures. For a ten percent increase in income, they increase their consumption of housing by nine percent. Large households, like small households, make large base expenditures for housing; however, their income elasticity is higher than for small households, and about the same as middle-sized households'. Large-sized households make large base expenditures and also devote larger portions of increases in income to housing expenditures. Nine percent more is devoted to housing expenditures for every ten percent increase in income. The classification of households into higher- and lower-income classes according to a median income of $115 per month does not reveal a sharp discontinuity in consumption patterns. Base expenditures and the income-elasticity coefficient for the two income cohorts are nearly the same. Gender and Life Cycle The importance of the gender and the life-cycle stage of the head of the household in affecting housing choices is analyzed by use of dummies for gender (female or male) and for life-cycle categories (less than 30 years old, between 30 and 50, and Older than 50). The demographic coefficient with respect to male- headed households is negative and about 0.1 (see Table 4.1). This implies that households headed by females demand more 95 housing services than do those headed by males, a finding Opposite our a priori expectations. However, the coefficient is not significantly different from zero at the 0.01 level. The demographic coefficient with respect to life— cycle stage is negative for middle-aged heads of households and positive for Older heads of households (see Table 4.1). This implies that households headed by middle-aged persons demanded fewer housing services than do households with both younger and older heads. Households with older heads demand more housing services than do those with young heads; however, neither coefficient is significantly different from zero at the 0.01 level. Middle-aged households had been expected to show the strongest demand for housing; the results are therefore contrary to the a priori expectations. The results of separate regressions for two gender cohorts and three life—cycle cohorts are shown in Table 4.2. These reveal a sharp discontinuity in consumption patterns. Female-headed households show larger base housing expenditures (more than 50 percent higher) than do male-headed households. However, income elasticity is about 39 percent lower for female-headed households than for male-headed households. This implies that as income increases proportionately, male- headed households devote more income to housing expenditures than do female-headed households. It is possible that female—headed households have a larger base, but undertake less self-help improvement with increases in income. 96 The classification of the households into the three life-cycle cohorts also shows a sharp distinction Of consumption patterns between the three groups. While households are young (the head of the household is less than 30 years of age), they make low base expendi- tures on housing, with a high income elasticity. In the middle stages of life (head of household between 30 and 50 years old), base expenditures increase by about 37 percent, while income elasticity falls by about 27 percent. In the latter stages of life, base expenditures decrease, while income elasticity increases to a much higher level (1.23). Thus, although Older households make small base expenditures on housing, they spend more on housing for every proportion- ate increase in income than do the young and the middle-aged households. It may be that older households have fewer other commitments because of changes in the life cycle and can therefore spend more on housing as income increases. Location of the Household In this section, the effect of the location of the household on the demand for housing is examined. The house— holds are located in three housing areas--a squatter area (Bauleni), an improvement area (Chawama), and a site-and- service housing area (Kaunda Square). Taking the squatter housing area as the basis, we notice that households in site-and-service areas spend more for housing services than do those in squatter areas (see Table 4.1). Households in 97 improvement areas spend more for housing services than those in squatter areas. This result is as anticipated: It was expected that household in improvement areas would spend more for housing services than would those in squatter areas. The estimated coefficient for the location dummy for a site-and-service area is 2.09 and positive. For an improvement area, the estimated coefficient is 1.09 and also positive. Both coefficients are significantly different from zero at the 0.01 level. Services to the House and Housing Demand In this section, the effect of services to the house, such as better water and better sanitary facilities, is examined. The regression equation, with dummies for better water and better sanitary facilities, is shown in Table 4.3. The coefficient for the better water facilities dummy is about 1.1, and that for better sanitary facilities is about 0.3. Neither coefficient is significantly different from zero at the 0.01 level. The availability Of these two services is correlated positively with increases in expenditures for housing services. The effect of water is much stronger than is the effect of sanitary facilities. 98 TABLE 4.3 REGRESSION OF PROPERTY VALUE ON INCOME AND SERVICE DUMMIES Independent Estimated a Variable Coefficient Income 0.68b (5.76) Services Better water facilities 1.09 (1.53) Better sanitary facilities 0.29 (0.40) Constant 3.52b Sample size» 155 ‘82 0.40 F 34.66 SOURCE: Calculated from data in M. Ndulo, 1979 Zambia Home .Improvement Survpy. at-statistics Of coefficient are in parentheses. bcoefficient significantly different from zero at the 0.01 level. 99 Households With and Without Lodgers Low-income households are likely to start up with a one-room unit. Later, they may add another room, either because they have saved enough money to do so or they have had windfall income from a relative. The rental income from the additional room is likely to be used to improve the dwelling. The demand pattern for housing for such households is likely to be different from that of other households. For the present study, therefore, it was decided to distinguish between households with lodgers and those without. The results Of the regression equations for the two lodger cohorts are shown in Table 4.4. There is a sharp distinction between the demand patterns for households with lodgers and the patterns for those without lodgers. Both the income elasticity and the equivalent household size elasticity are higher for the households with lodgers. However, the constant term for households with lodgers is lower than for those without lodgers. Households without lodgers have larger base expenditures for housing than do households with lodgers, although both the income and the household size elasticities of demand for housing are greater for households with lodgers (in both cases they are greater than 50 percent). Furthermore, the income and household-size elasticities are similar for households without lodgers. The effect of the sex of the head of the household gives Opposite results for the two cohorts. Among households I100 TABLE 4.4 REGRESSION OF PROPERTY VALUE ON 'INCOME AND DEMOGRAPHIC VARIABLES BY LODGER CATEGORY Estimated Coefficientsa Independent Variable Households Households With Without Lodgers Lodgers Income 0.84b 0.53b (3.37) (4.54) Household Size 0.82c 0.53b (2.45) (2.85) Gender Male -0.49 0.31 (0.81) (1.11) Age of Household Head 30-50 years -0.50 -0.06 (1.21) (0.23) > 50 years 0.12 0.36 (0.24) (1.15) Location b Chawama 0.71 1.08 (1.52) (5.91) Kaunda Square 2.13b 2.03b (3.32) (9.50) Constant 1.89d 2.22b (1.79) (3.43) Sample Size 29 126 —2 R 0.58 0.57 F-statistic 6.43 23.43 SOURCE: Calculated from data in M. Ndulo, 1979 Zambia Home ImproVement Survey. at-statistics of coefficients are in parentheses. bcoefficient significantly different from zero at the 0.01 level. ccoefficient significantly different from zero at the 0.05 level. dcoefficient significantly different from zero at the 0.10 level. 101 with logers, female-headed households demanded more housing services than did male-headed households. However, among households without lodgers, male-headed households demanded demanded more housing services than did female-headed households. The coefficients for both lodger categories were insignificant, though. Middle-aged households demanded less housing services than did younger households regardless of whether the household had lodgers. The pattern for older households is the same for both lodger cohorts: Older households demanded more housing services than did younger households. On the whole, therefore, we can conclude that the patterns of demand differ between the two lodger cohorts. Households with lodgers made a lower level of base expen- ditures for housing than did households without lodgers, but demand for housing increased prOportionately more for households with lodgers than for those without for the same proportional increase in income. Households at Different LeVels of Education In this section, the effect of schooling on the demand for housing is examined. The total sample is divided into two cohorts. The first consists of households with primary education (those whose head has had seven or fewer years of schooling); the second is composed of households with secondary education (those whose head has had eight or more years of schooling). The hypothesis is that household heads 102 who have had secondary education will show a greater demand for housing than will those whose education has not passed the primary level. The results of the regression equations for the two schooling cohorts are shown in Table 4.5. The coefficients for the income elasticity of demand and the household-size elasticity are dissimilar. The income elasticity coefficients are significantly different from zero at the 0.01 level. The results suggest that income and household size elasticities differ between the two types of households. While it was expected that those households with more years of schooling would have higher elasticities than would those with fewer years, households with primary education instead had higher elasticities than did households with higher levels of education. A sharp discontinuity in consumption patterns with respect to the sex of the head of the household is seen. In the case of the households with primary education, female- headed households consumed more housing services than did male-headed households. .The situation is reversed for households with secondary education. Here, male-headed households consumed more housing services than did female- headed houses. There is also a sharp discontinuity in consumption patterns with respect to age and location of the household; the direction of the effect of the dummies is the same for both cohorts. To sum up, it seems that consumption patterns for both 103 TABLE 4.5 REGRESSION OF PROPERTY VALUE ON INCOME AND DEMOGRAPHIC VARIABLES BY EDUCATION CATEGORY Estimated Coefficientsa Households Households Independent Variable With With Primary Secondary Education Education Income 0.66b 0.48b (4.62) (2.67) Households size 0.79b 0.39 (3.57) (1.42) Gender -0.13 0.66 (0.50) (1.16) Age of household head 30-50 years -0.49 -0.21 (1.19) (0.72) > 50 years -0.26 _ (0.60) Location b b Chawama 1.13 1.25 (5.83) (3.29) Kaunda Square 2.03b 2.40b (8.47) (5.59) Constant 2.25b 2.360 (3.35) (2.29) Sample Size 98 57 82 0.58 0.52 F 20.10 10.92 SOURCE: Calculated from data in M. Ndulo, 1979 Zambia Home Improvement Survey. at-statistics of coefficients are in parentheses. bcoefficient significantly different from zero at the 0.01 level. ccoefficient significant at the 0.05 level. 104 income and household size were different for households with primary education than for those with secondary education. However, households with fewer years of schooling had higher elasticities than did households with more years of schooling. Conclusion This chapter has presented an analysis of the demand for housing. Some of the findings are similar to those of other studies of housing demand elsewhere. The estimated income elasticity confirms the expectation that the demand for housing for homeowners with respect to current income is less than one. The results imply that, as long as the household remains in the low-income settlements, a doubling of a household's current income, everything else also held constant, will be accompanied by a 60 percent increase in housing expenditures. If households were given transfers, they would only spend about 60 percent of the transfers on housing expenditures. If policymakers want households to spend more of these transfers on housing, they need to earmark them for housing expenses. (For example, transfers could be given in the form of building materials.) The income elasticity coefficient was seen to vary over different household sizes, but not over income groups (those earning less than the median income of $115 per month and those earning more than the median income). Furthermore, income and equivalent household elasticities are about 105 similar. There were three particularly interesting results in the analysis. It was found that female-headed households demand more housing services than do male-headed households ceteris paribus. This finding implies that current policies in the less-developed countries to make female-headed households in urban areas economically stronger will also improve housing. It was also found that households whose heads are in the older stage of the life cycle demand more housing services than do the young households, which in turn demand more housing services than do the middle-aged households. Households with access to piped water spend more on housing services than do those without piped water. This effect is stronger than the effect of access to better sanitary facilities. CHAPTER FIVE DETERMINANTS OF DWELLING VALUES Introduction In this chapter the market price of the houses in the housing market is studied from the point of the separate attributes. In a perfectly competitive market, each attribute is valued because of the flow of housing services it emits over its economic life; these housing services are capitalized into the property value of the house (Muth 1969). The amounts of services emitted by various houses, and the prices of such services, cannot be directly observed in the market. The method of hedonic indices is used in the present study to relate the market value of each property to the quantity of its housing services. The observable attributes of the house are used as surrogates for housing services. The purpose is to understand the relative importance and the relation of the attributes (and therefore of the housing services of a housing commodity) to the value of the house. Marginal weights are estimated to give this information, and are later used for estimating demand functions for each specific attribute. 106 107 Methodological Framework A perfectly competitive housing market is assumed. The housing commodity is composed of a bundle of different attributes. The housing market determines an equilibrium price, based on the interaction of demand and supply, which for each house is influenced by the amount and combination of the attributes in each house (Myrick Freeman II 1974; Lucas 1975; Ball and Kirwan 1975). When the housing market clears, at equilibrium, QD = 05 (5.1) where QD is the total market demand and QS is the total market supply of housing. This point yields an equilibrium price, from which a hedonic price equation can be deduced (Lucas 1975): Pi = P(vl, v2, . . . , vm) . (5.2) (l = l, . . . , n) where pi is the equilibrium price level and vi's are a set of attributes in each house. The observed relationship in the market between the price of the dwelling and the various attributes of the dwelling is a result of the interaction between demand and supply, and gives the relationship between the value of the dwelling and the emitted flow of housing services, for which the attributes are proxies. 108 Empirical Specification To analyze the structure of the attributes of a dwelling, the hedonic price equation (5.2) is estimated using the log-linear functional form. The price of the house is regressed against attributes of the house grouped in the categories space, access, structure, and related services to the house. It is generally agreed in the literature that the appropriate functional form for a hedonic price equation cannot be specified a priori. Rather, it is usually based on convenience in dealing with the problem at hand (Pollakowski 1982, p. 91). For the purposes Of the present study, the log-linear functional form shall be used because it Offers convenience, and also because earlier experimen— tation with the data has shown that it yields a better fit (Ndulo 1981). The empirical formulation of the model will take the form, Ln Pi = 80 + len FLSi + ean AGEi + B3Ln TOWKi + B4Ln YARDi + 85Ln NORMi + Blei + B7D21 + 88D3i + B9D4i + 8101351 + 8i (5'3) where P is the price of the dwelling: representing the property value of the house. Continuous attributes are FLS, floor-space area of the house; AGE, age of the house, TOWK, the time it takes the head of the household to travel 109 to work; YARD, the yard area of the site; and NORM, the number of rooms in the dwelling. For the discrete attributes of the dwelling, there are the five dummy variables--for better water facilities, D1; better wall materials, D2; better roof materials, D3; better sanitary facilities, D4; and whether or not the house has been plastered and painted, D5. 8 is the random error. Empirical Results A number of attributes of a house can be estimated, each representing a flow of household services for each specific attribute. In the present case, ten housing attributes were estimated, grouped into four major categories --structure, space, access, and related services to the dwelling. The best estimate of the regression equation is shown in Table 5.1; the coefficient correlation matrix of the variables in the estimated equation is given in Appendix B. The dependent variable is the value of the house. This is the expected price for which the dwelling would sell in the market under current market conditions; it is the asking price. The value of this variable depends on how well informed each household is about current market conditions. It was observed that most households tended to be well informed and were able to accurately predict the value of the house. 110 TABLE 5.1 REGRESSION OF PROPERTY VALUE ON ATTRIBUTES OF A DWELLING Independent Variable Estimated Coefficient a Structure Number of rooms 0.935b (5.309) Age of dwelling -0.104 (0.871) Wall materials 1.271b . (6.90) Roof materials 0.171 (1.040) Plaster and paint 0.491b (4.040) Space Floor-space area 0.089 (0.729) Yard area 0.078 (0.791) Access Travel time to work -0.086 (1.226) Services Water facilities 0.3600 (1.489) Sanitary facilities 0.232c (1.312) Constant 4 . 318b (5.619) Sample size 134 i2 0.729 F 36.86 SOURCE: Calculated from data in M. Ndulo, 1979 Zambia Home Improvement Survey. . at-statistics of coefficients are in.parentheses. bcoefficient significantly different from zero at the 0.05 level. ccoefficient significantly different from zero at the 0.20 level. 111 Structure In order to analyze the importance of the structure of the house in contributing to its value, several factors were examined: the number of rooms, the age of the dwelling, and dummies for the quality of wall and roof materials and whether or not the house had been plastered and painted. The number of rooms is the total excluding the kitchen, toilet, and bathroom. A house with better wall materials is made of permanent wall materials, defined as either concrete blocks or burnt bricks. Better roof materials were defined as asbestos-cement sheets. The plaster-and- paint variable is self-explanatory. Number of rooms A positive relationship is expected, a priori, between the number of rooms and the value of the house. The rapid rate of urbanization in Zambia (especially during the past two decades) and the associated massive influx of economically weak migrants from rural areas are expected to create over-crowding in low income housing markets, as households attempt to share what little accommodation is available. (Most migrants stay with relatives while looking for jobs, and later move out when they find employment.) Studies carried out by the Lusaka City Council's Lusaka Housing Project Unit show that over-crowding is high in all low income housing areas, but it is higher in the public housing run by the Lusaka City Council than in the private housing areas. This is probably because households in private 112 low income housing areas can more easily adjust their demand for rooms within the constraints of their budgets: They can always add on rooms or move to more-preferred bundles of housing services within the market. This is not possible for residents in public housing: They face both a rationing constraint and a budget constraint. Whenever they are dissatisfied with their allocation of rooms in the dwelling, they must go to the City Council's administrative allocation mechanism, and enter a queue to satisfy such demand. This might mean waiting for one, two, or more years. The empirical results of this analysis confirm the a priori expectations. A positive relationship was found between the value of the house and the number of rooms. The estimated coefficient is 0.94, meaning that for every percentage increase in the number of rooms, a nearly equal percentate increase in the value of the house was Obtained. Other things being equal, there are constant returns to the number of rooms. The estimated coefficient is significantly different from zero at the 0.05 level. Age Of the dwelling The value of the dwelling was expected to be negatively related to its age. Households view the dwelling as a capital asset, with an expected economic life. As time goes on, the dwelling is expected to depreciate in value, everything else constant. This will be reflected in a reduced market value as the economic lifespan of the dwelling is reduced. Furthermore, if the dwelling is of low quality 113 in terms of construction, the rate of depreciation (and thus of the reduction in the market value) will be faster. On the other hand, site value rises with time, and maintenance and improvement expenditures can offset or more than Offset depreciation. However, residents of Kaunda Square complained about the durability of their dwellings: They were worried about roof leaks and cracks in wall structures (Ndulo 1979). An inverse relationship was found between the value of the house and its age, but the estimated coefficient was quite small, 0.10, and was not significant. Wall materials Wall materials were apparently the most important aspect of the structure of the dwelling in terms of the market value. In comparing two average dwellings with the same attributes, except in terms of quality of wall materials, a big difference is found in the values of the dwellings. The dwelling with better wall materials-~concrete blocks or burnt bricks--is worth about $5,100.00, and the dwelling with poorer wall materials--such as adobe or sunrdried bricks--is worth only about $1,390.00. Thus, an otherwise- identical house with better wall materials is worth about four times the value of a dwelling with poorer wall .materials. The estimated coefficient is 1.3, significantly different from zero at the 0.05 level. This result is supported by available evidence from the participants of a house-upgrading scheme in George run jointly by the Lusaka City Council and the World Bank. 114 Participants generally preferred concrete blocks for their walls, as against other alternatives, such as sun-dried bricks, although the latter would cost less. Residents believed that the only durable house is one with concrete- block walls. One resident is reported to have said, "This is the only chance I have to build a permanent house, and I cannot build it with sun-dried bricks. I want a concrete one" (Mulenga 1978, p. 6). Plaster and paint The third most important aspect of the structure of the house in terms of determining its value is whether or not the dwelling is painted and plastered. On casual Observation it can be noticed that many dwellings in the low income housing areas not only are plastered and painted, but also are well decorated. Apparently, the general appearance of a dwelling has high marginal significance. It is possible that, because much of the day is spent outside the dwelling, there is a trade-off between the functional aspects and the general appearance of the house. This suspicion is bolstered by a study by the Zambia National Housing Authority, which showed that residents in site-and-service areas would rather spend their income on the surface appearance than on more space. The most common example of such surface embellishment was painting. Residents painted figures or flowers on the walls of the house. The empirical evidence gathered for the present study also supports this assertion. Dwellings which are plastered 115 and painted are worth more than others. An average dwelling with everything the same except for a lack of plaster and paint is valued at $3,093.00, while its plastered and painted counterpart is worth $5,100.00. This is an increase in value of about 65 percent. The estimated coefficient for the plaster-and-paint dummy is 0.49, significantly different from zero at the 0.05 level. There is only a low correlation between use of permanent materials and plaster and paint. The correlation coefficients between plaster and paint and better wall materials is 0.08; between plaster and paint and better roof materials, 0.11. These results imply that households tend to plaster and paint their dwellings regardless of the standard of structural materials. Roof materials Two houses, identical except that one has an asbestos roof and the other does not, will differ in market value. Most houses have either asbestos roofing or corrugated metal sheets; houses with metal sheets are worth less, perhaps because asbestos roofing affords protection from heat during the hot season and keeps the house warm during the cold season. The estimated coefficient is 0.17; however, it does not seem to bear significantly upon the market value of the house. 116 Space The measures of space used for the present study are the floor space of the dwelling and the yard area of the site on which the house is located. The floor space area is the total floor area of the dwelling, including the bathrooms, kitchen, and toilet facilities. The yard area of the site is the total site area minus the total area covered by the dwelling. Many household activities in low income housing areas take place outdoors during much of the year; the yard area is therefore an extension of the inside living space. This is especially true for the "swept area" of the yard, which has definite uses for household activities such as cooking, washing, and leisure (LHPET 1976, p. 27). The yard area is also used for such activities as vegetable gardening, storage (especially for charcoal or wood), children's playground activity, and diverse business activities such as small-scale manufacturing or commerce (LHPET 1976, p. 29). Sometimes the goods sold are those grown in the vegetable garden. Available studies by the Lusaka Housing Project Evaluation Team (1976) and Schlyter and Schlyter (Ward 1981) confirm the importance of the yard area of the site. Additionally, the household with a bigger yard area will more likely expand--i.e., add another room-- than decide to relocate as economic resources change. The regression equation gave a positive relationship between the value of the house and its floor space (0.09) and yard area (0.08). The coefficient for the yard area is 117 higher than the coefficient for the floor-space area; however, neither estimated coefficient is significant at the 0.05 level. Access The literature on location theory devotes much energy to explaining that accessibility should positively affect the value of the house: that is, that there is an inverse relationship between the value of a house and a measure of accessibility. Households are expected to be willing to pay more to be situated near areas of employment. As the measure of accessibility for the present study, the amount of time spent by the head of the household to get to his or her place of work was used. The empirical results are consistent with the theoretical notion. The estimated coefficient is 0.09. This implies that for every 10-percent increase in the time it takes to go to work, there is only a one-percent decrease in the value of the house. Services to the House Two major services to the house are access to piped water and sanitary facilities. In the regression equation for the present study dummies are used for each service. A dwelling either did or did not have piped water. It had better sanitary facilities if it had a septic tank or a flush toilet. A positive relationship between the value of the house and the dummies for piped water and better 118 sanitary facilities was expected. The estimated results were as postulated. The coefficients for better water and better sanitary facilities were 0.36 and 0.23, respectively. Looking again at the case of two identical houses, this time with everything the same except that one dwelling has a flush toilet and the other does not, the former is worth $5,100.00 and the latter is worth $4,176.00. The addition of a flush toilet increases the value of the house by 22 percent. An average house with no piped water is worth $3,419.00. Improving the house with the addition of indoor piped water increases its value by 49 percent. The estimated coefficients of both dummies are significantly different from zero at the 0.20 level, implying that the provision of both sanitary and water facilities does have significance for the value of the house. These findings are very important for policy consider- ations. Currently, a major question is whether public policy should encourage residents in low income private markets to erect indoor piped water and water-borne sanitation systems. The results of the present study indicate that having piped water or a flush toilet in the house is not highly signifi- cantly related to the value of the house. The survey (Ndulo 1979) showed that nearly all residents without better sanitary facilities were dissatisfied with their sanitary facilities. In Chawama, most respondents would have liked the Lusaka Housing Project 119 to intervene more directly in the erection of pit latrines than in anything else. One can argue that this indicates the importance residents attach to improving their sanitary facilities, but not necessarily in terms of flush toilets and septic tanks, as yet. The dwelling attributes are ranked by means of Beta coefficients in Table 5.2. Each Beta coefficient is a weight measuring the relative effect of each attribute on the value of the dwelling. The most important attributes in contributing to the value of the dwelling are those related to its structure. In this case, better wall materials, number of rooms, and plaster and paint. The least important attributes are related to the space of the dwelling. These are floor-space area and yard area. Conclusion A clearer understanding of the relationship between the value of a house and its attributes is an important contribution to the recent concern about low income housing markets. Clear understanding should help public policymakers to identify which attributes of the dwelling to focus on when trying to help residents in low income housing areas. The empirical analysis was an attempt to identify such attributes which can be affected by public policy. These attributes were ranked by means of Beta coefficients. The number of rooms, quality of wall materials, plaster and BETA 120 TABLE 5.2 COEFFICIENTS OF THE ATTRIBUTES OF A DWELLING Attribute Beta Coefficient Rank Number of rooms 0.324 2 Age of dwelling -0.050 8 Better wall materials 0.486 1 Better roof materials 0.069 6 Plaster and paint 0.199 3 Floor-space area 0.044 9(tie) Yard area 0.044 9(tie) Travel time to work -0.059 7 Better water facilities 0.081 4(tie) Better sanitary facilities 0.081 4(tie) SOURCE: Calculated from Table 5.1. 121 paint, water facilities, sanitary facilities, quality of roof materials, and travel time to work seem to be relatively important attributes. The other attributes do not seem to bear significantly upon the value of the house. The equal rating of access to piped water and better sanitary facilities is interesting. One can deduce from this that once residents in low income housing have acquired durable dwellings, they see the provision of basic services such as piped water and imporved sanitary facilities as equally important. Not until permanent dwellings have been acquired do residents attempt to get better sewage systems (such as flush toilets) or water piped into the dwelling. CHAPTER SIX DEMAND FUNCTIONS FOR DWELLING ATTRIBUTES Introduction In this chapter we wish to estimate the households' demand functions for the attributes of the dwelling. In Chapter Five we estimated the coefficients of the dwelling attributes from a hedonic price equation. The coefficients are the marginal weights of the various attributes in our housing bundle. We shall use the information from Chapter Five to derive the marginal prices and together with the attribute quantities estimate the households‘ demand functions for each attribute. We are interested in examining the relation between attribute quantities and their own prices, and household income. We expect to find an inverse relation between attribute quantities and their own prices and a direct relation between attribute quantities and income. We also examine the effect of households' taste variables, such as household size and gender, on the attribute demand structure. 122 123 Methodological Framework Following the work of Nelson (1978), Harrison and Rubinfeld (1978), Freeman (1979a, 1979b), and Quigley (1982), we have a housing market which will be in equilibrium when the quantity demanded (OD) and the quantity supplied of housing (OS) are equal. This situation will give us an equilibrium quantity of housing services (Q) and an equili- brium price (P). However, since each dwelling is a hetero- geneous commodity, it contains different combinations of attributes. Therefore, when the homogeneous commodity bundle is in equilibrium, QD = 08 (6.1) the various attributes of the housing bundle, the Vi's shall be so packaged as to also produce an arbitrary set of attributes in equilibrium, Qn‘Vi’ = 05(vi) (6.2) Households derive utility from consuming various combinations of the attributes in the housing bundle subject to the budget constraint. The attributes are provided by suppliers who strive to minimize their costs in their productive activities. The market demand functions for the various attributes will be determined by the attributes' own prices, income and tastes. The market supply for the attributes shall be determined by the technology and cost 124 conditions faced by suppliers, and shall be a function of the attributes' own prices and a set Of supplier characteristics. At equilibrium households will have packaged their combination of attributes such that quantities demanded of the various attributes is equal to quantities supplied, as expressed by equations (6.1) and (6.2). We therefore have the observed attribute quantities of the housing bundles, Vi's, and the unobserved marginal prices of the various attributes, MPi's. Equations (6.1) and (6.2) and the equilibrium prices shall give us the hedonic price equation, P = f(Vl, v2, . . . , Vm) . (6.3) We shall use the hedonic price equation to translate a vector of housing attributes into a marginal price, MP1. The estimated implicit marginal price influences the decisions of both suppliers and demanders of housing attributes. The marginal price of an attribute can be found by differentiating the hedonic price equation with respect to that attribute, 8P This gives us the increase in expenditure on the dwelling that is required to Obtain a dwelling with one more unit of that attribute, holding everything else constant. 125 For example, if a household desires more of a particular attribute in a dwelling, it has to repackage the existing bundle of attributes so as to have one unit of the desired attribute in the new bundle than in the previous bundle. The marginal price of the attribute is the extra amount that the household will be willing to pay for a dwelling unit with one more of that particular attribute than the previous dwelling unit had. We use the information on marginal prices, attribute quantities, income and household taste variables to derive the demand functions for the various attributes (Rubinfeld and Harrison, 1978; Freeman, 1979). Formally the demand functions will take the form, MPi = f(Vi, y, a) (6.5) where MPi is the marginal price of the attribute, Vi is the attribute quantity, y is income, and a represents the households' taste variables. On the supply side of the marginal price will be a function of the attribute quantities and other supplier characteristics. However, our basic assumption is that the supply Of attributes is perfectly inelastic. This assumption is reasonable in terms of short run cross section analysis. In the short run we expect the supply of individual attributes to be relatively fixed and prices to be therefore demand determined. 126 Empirical Specification We wish to generate quantitative estimates of the demand functions for attributes. In Chapter Five we analyzed the structure of the attributes of the dwelling. To do this, we estimated a hedonic price equation (5.2) using a logarithmic functional form. The nature of this functional form will give us enough information to enable us to identify the demand functions we are interested in (Rubinfeld and Harrison, 1978; Brown and Rosen, 1982). We can write the estimated form of equation (5.2) as, . 10 A 5 . Ln P = 80 + .21 silnxi + '26 B.D. (6.6) l: 1: where P is the value of the dwelling, xi is a continuous attribute of the dwelling, and Di is a discrete attribute of the dwelling. We can derive the marginal price of an attribute for each Observation in our sample by total differentiating the hedonic price equation (6.6), 1 E * dxi _ dP = B. —— (607) P i=1 1 xi . dP and $01Ving for dxi , dP ._ “ P ._ ‘83:.— 7 313a“ " MPi' (6‘8) 1 l which is the marginal price of attribute Xi' The marginal prices of the attributes represented by the dummy variables (Di i=6, . . . , 10) in the estimated equation (6.6) are 127 undefined because the dummies are discrete (Di = 1 or 0). For our analysis we therefore concentrate on the continuous variables for the structure, space and access aspects of the dwelling. These are the number of rooms, age of the dwelling, living area (floorspace area), yard area and travel time to work. Since for each attribute we have a demand and supply function, we are likely to be faced with the problem of identification. But if our assumption on the supply side is that of perfectly inelastic supply we can identify the demand function. Changes in the level of the attributes over the sample allow us to get information about the households' demand for the attributes. Therefore our general regression equation for the demand function for each attribute derived from equation (6.5) will take the form, MP. = + i “0 Xi + a a1 2 INCi + a3 HSi + a4 D1 + “5 D2 + 66 03 + 6i (6.9) where MPi is the marginal price of an attribute, X1 is the quantity of the attribute, INCi is household income, HSi is the size of the household, D is the dummy for the gender 1 of the household head (Dl=1 if male; D = 0 if female), D 1 2 and D3 are dummies for life cycle categories (D2=1 if between 30 and 50 years old; D2=0 if otherwise, and D =1 if greater 3 than 50 years old; D3=0 if otherwise), and Si is the error term. 128 The demand price of each attribute is a function of the attributes own quantity level, income and household size and other household taste variables. We exclude the quantities of other attributes in the regression equation on the assumption that the cross price effects are close enough to zero to be ignored. The marginal price of an attribute is the demand price for the household. It is the amount of money the household will be willing to pay for one additonal unit of the attribute, holding everything else constant. We expect that the marginal price of the attribute will decrease with an increase in the quantity of the attribute consumed, other things held constant. On the other hand, if income increases, while holding everything else constant, we expect the household to increase its consumption of that attribute. We therefore hypothesize an inverse relation between quantities of the attributes demanded and the attributes' own marginal prices and a direct relation with household income. We hypothesize a direct relation between the marginal prices of the attributes and household size, which in this case is the equivalent household size. This is so because we expect larger households to have a stronger demand for the various attributes than smaller households. The demand price of large households is therefore likely to be higher than that of smaller households. 129 We are not certain of the expected relation between demand price and the gender of the household dummy. For the life cycle categories, we expect a direct relation between the marginal price of the attributes and middle- aged heads of households (those between 30 and 50 years old), and an inverse relation between the marginal price of the attributes and the Old heads Of households (those Older than 50). Middle-aged households are in the most active stage of their life cycle and are therefore likely to have a stronger demand for the various housing attributes than other households. On the other hand, old households are past this most active stage of their life cycle and are less likely to have a strong demand for the housing attributes than most other households. Empirical Results We estimate the demand functions for the continuous attributes of the dwelling; structure, space, and access. The dependent variable is the marginal price of the relevant attribute in the demand function, as expressed in equation (6.9). The estimated regression equations for the continuous attributes of the dwelling structure are shown in Table 6.2 and those of the dwelling space and access in Table 6.3. The market wide average marginal prices, price and income elasticities of the dwelling attributes are shown in Table 6.1. 130. TABLE 6.1 AVERAGE MARGINAL PRICES, PRICE AND INCOME DEMAND , _ ELASTICITIES Average Marginal Price Income Prise Elasticity Elasticity (MP) ' (e ) (e ) P Y Number of $676.74 +2.63 -0.76 rooms (56.23) (room) Yard area 0.97 -1.65 0.71 (m ) (0.13) Age of the 76.85 -1.13 0.53 dwelling (10.10) . (year) Living area 2.64 -5.23 1.59 (m ) (0.24) Travel time 16.38 -0.84 0.67 to work (minute) (3.98) SOURCE: Calculated from data in M. Ndulo, 1979 Zambia Home Improvement Survey and Equation (6.9). NOTES: 1. '2. Standard errors are in parentheses. The average marginal prices are marketwide and are calculated at the mean value of each attribute for the total sample. Thus for n households, the average marginal price for an attribute is, nMP. MP .. {——1 i=1“ Price elasticities are calculated from the estimated equation as, i e 310'—. P /1 31 Income elasticities are calculated from the estimated equation as, “2 Y e =—,_. a _ Y 1 xi The negative income elasticit for number of rooms is because of positive quantity coefficient in the estimated regression. 131 TABLE 6.2 REGRESSION OF MARGINAL PRICES OF HOUSING ATTRIBUTES ON INCOME AND DEMOGRAPHIC VARIABLES: STRUCTURE DEPENDENT MARGINAL PRICE OF VARIABLE Rooms Age INDEPENDENT a VARIABLE ESTIMATED COEFFICIENTS Number of rooms Age of dwelling Income Household size Gender of head male Life cycle 30-50 years > 50 years Constant Sample size §2 F 63.243 (1.763) b -9.760 (4.308) 1.283b 0.232b (3.311) (3.794) 37.231 4.637 (1.523) (1.147) -111.614 -10.275 (0.664) (0.356) -26.777 -4.493 (0.196) (0.192) 115.121 16.911 (0.645) (0.563) 91.567 75.364 (0.367) (1.779) 134 134 0.13 0.20 4.44 6.42 SOURCE: Calculated from data in M. Ndulo, 1979 Zambia Home Improvement Survey. at-statistics of coefficient are in parentheses. bcoefficient significantly different from zero at the 0.01 level. 1532 TABLE 6.3 REGRESSION OF MARGINAL PRICES OF HOUSING ATTRIBUTES ON INCOME AND DEMOGRAPHIC VARIABLES: SPACE AND ACCESS MARGINAL PRICE OF DEPENDENT ”ITravel Time VARIABLE Liv1ng Area Yard Area to Work INDEPENDENT a VARIABLE ESTIMATED COEFFICIENTS Living area -0.426 x 10-2 (1.485) Yard area -0.107 x 10-20 (4.460) Travel time -0.370c (3.718) Income 0.520 x 10"2c 0.261 x 10'26 0.083c (3.060) (3.559) (3.568) Household size 0.072 0.072 1.198 (0.642) (1.477) (0.795) Gender -0.507 -0.443 -34.169c male (0.664) (1.298) (3.250) Life cycle 30-50 years -0.555 -0.123 -7.285 (0.894) (0.448) (0.848) 50 years 0.239 0.149 15.573 (0.298) (0.418) (1.403) Constant 2.415” 1.021” 46.331c (2.255) (2.044) (3.048) Sample size 134 134 134 §2 0.04 0.20 0.25 F 2.02 6.63 8.59 SOURCE: Calculated from data in M. Ndulo, Survey Data (1979). at-statistics of coefficient are in parentheses. b 1979 Zambia Home Improvement coefficient significantly different from zero at the 0.01 level. ccoefficient significantly different from zero at the 0.05 level. 133 Structure: Number of Rooms and Age of the Dwelling For the dwelling structure we have two continuous attributes--the number of rooms and age of the dwelling. In this section, we analyze the demand functions for these attributes. The estimated coefficients of the demand functions are shown in Table 6.2. Our first equation is a regression of the marginal price of rooms on number of rooms, income, household size, gender and life cycle dummies. We get the expected relation except for the number of rooms variable and the life cycle dummies. The quantity of rooms demanded varies directly with the demand price, holding everything else constant. That is, households are always willing to pay more for additional rooms, ceteris paribus. Our estimated coefficient is 63.2, giving us a positive price elasticity of 2.6. That is for every one unit increase in the number of rooms, households are willing to pay $63.2 more than they paid for the last room. From an average of $677, they will go to $740 for one more room. Alternatively for every one percent increase in the price of rooms, there is a 2.6 percent increase in the quantity of rooms demanded. This is different from our theoretical expectations. A priori, we expected an inverse relation between marginal price and number of rooms demanded. However, the estimated coefficient is not only positive but is not significantly different from zero at the 0.05 level. This result might be because the household in our 134 sample is both the main consumer and supplier of the attribute. In this case it is plausible that the fact that he can always rent out rooms makes his demand for rooms so strong that he is always willing to pay more for an additional room, holding everything else constant. In addition, we may have misspecified the issue, and marginal rooms might be harder to add on or be of higher quality in undetected ways. It may be further noted that these are extra rooms at giygg space, in effect subdivisions. Middle-aged households pay less and old-aged house- holds pay more for an additional room. On the other hand our estimated regression shows that both large and high income households are willing to pay more, while male headed households pay less for an additional room. However, only the income coefficient is significant at the 0.05 level. Our estimated demand function for the age of the dwelling shows that households pay less for the same dwelling with a marginal increase in age, but at the same time, holding everything else constant, with a marginal income or household size increase, they will pay more for the same dwelling. The estimated price and income elasticities for the age attribute of the dwelling are -l.13 and 0.53 respectively and are both significantly different from zero at the 0.01 level. 135 Space: Living and Yard Areas We have two variables measuring space. These are the living and yard areas of the dwelling. The estimated demand functions for living and yard area attributes are shown in Table 6.3. The estimated coefficients for the living area demand function gives us the expected signs except for the life cycle dummies. These show that middle-aged households demand less and old-aged households demand more living area space than other households. However'both coefficients are not significant at the 0.05 level. The demand for living area space is both income and price elastic. The estimated elasticities are 1.59 and -5.23 respectively. The estimated demand function for yard area gives us the same relationship as that of the living area attribute. However, in this case the demand function for yard area is income inelastic and price elastic. The estimated coefficients are 0.71 and -l.65 respectively. Access: TraVel Time to WOrk The estimated results of the demand function for access is shown in Table 6.3. We use the time taken by the head of the household to travel to work as a proxy for access. This function will therefore be a measure of the value of accessibility to the household. The estimated regression shows that the household 136 will be willing to pay less for the same dwelling with a marginal increase in the travel time to work--a decrease in accessibility. However, with marginal increases in income, family size or as households get older, ceteris paribus, households will pay more for the same dwelling with equal accessibility. The estimated price and income elasticities are -0.8 and 0.7 respectively. This gives us a price and income inelastic demand for access. Conclusion Accessibility, number of rooms, age, living and yard areas are some of the important attributes of a dwelling. This chapter gave us an idea of the importance of these attrib utes in terms of consumer demand. Several facts came out clearly. As expected, except for rooms, there is an inverse relation between the demand price of each attribute and its quantity. Households demand more of the attributes with marginal increases in income and family size. Female-headed and old-aged households have consistently a higher demand for attributes than male-headed, young and middle-aged households. The demand for living area is both price and income elastic, while the demand for yard area and age (durability), are price elastic and income inelastic. The demand for access is both price and income inelastic. CHAPTER SEVEN THE HOME IMPROVEMENT PROCESS Introduction This chapter looks at the supply of housing. One can study housing supply in terms of either the existing stock of housing or new housing or their jointly determined interacting changes. In terms of the existing stock of housing, one can examine sales of old houses, enlargement and improvement of existing dwellings, conversion of houses into non-residential use, and the subdivision and demolition of dwellings. In this chapter we are only interested in the enlargement and improvement of existing dwellings--the home improvement process. We are interested in the home improve- ment process because it is the major means through which households expand and improve the housing stock and quality on the supply side. We are interested in what sorts of households improve dwellings. We therefore wish to analyze the factors that are likely to determine the actions of the owner occupant household to make improvements to the dwelling. In the analysis we shall concentrate on two types of actions; making any improvement and adding an extra room or more to the 137 138 dwelling. Methodological Framework Our underlying conceptual framework of the home improvement process follows the work of Mendelsohn (1977). The house supplies to the household a flow of housing services and it is at the same time a component of the household's investment portfolio. The owner occupant household is therefore both a consumer and an investor. In his or her role as both a consumer and an investor, the household wishes to maximize its utility function; U(H.X.A,T) (7.1) given its discount rate and subject to the income and time constraint. Where H = quantity of housing, X = quantity of other goods, A = assets and T = leisure time. Income earned by the household is allocated to eXpenses for home improvements, to other goods and to other assets, such as financial assets. The household's time is allocated between leisure, work and home improvements. Total assets are allocated between other assets and the value of the house. We also have the housing production function H = H(H0, K, L) (7.2) where H0 = initial stock of housing, K = total expenditure on home improvements and L = household's input of labor into 139 home improvements. In his double role as consumer and investor, the owner occupant will work on home improvements until the marginal rewards from that work are equal to the marginal value of this leisure time. He will spread his income such that the marginal value of a dollar spent on improvements, assets or consumption of other goods is the same. He will also use the wage rate to determine whether to use hired labor or not. For any given stock of housing, H0, the housing production function becomes strictly the home improvement function, describing our home improvement process. Thus we now have a home improvement function where the quantity of home improvements is a function of the initial stock of housing, the total expenditures on home improvements, the household's input of labor into home improvements and other factors that affect the supply of home improvements. Mathematically we can write our home improvement function; HI = H (H0, K, L, F) (7.3) where HI = quantity of home improvements, and F = other supply factors. The other supply attributes of the owner occupant household that affects the supply of dwelling improvements are the household's age and family size; the length of his residence at the site; the presence of lodgers; and the 140 receipt of a one-time special income. The total expenditure on home improvements and the household's input of labor into home improvements is a function of the household's income. The analysis of the sample survey data showed that most households financed their home improvements from their wage earnings. In terms of labor used for home improvements, most households used hired labor for their home improvements, because of lack of skills and atime constraint. This is most likely to happen as wage income increases. The higher the wage income of the owner occupant, the less likely she or he will perform the home improvements by himself (herself). However, in the cases of those who make home improve- ments using their own labor, we do not have information about the time owner occupants spent improving their dwellings. We also do not have information on the total expenditures on home improvements by the households. However, since both total expenditures on home improvements and the households' input of labor into home improvements is a function of income, we substitute the household's disposable income for total home improvement expenditures and total input of labor into home improvements in the home improvement function 6.1. Thus we have a modified function; HI = H (H0, Y, F) (7.4) where HI = quantity of home improvements; Y = household disposable income; and F = other supply attributes of the 141 household. Our basic empirical problem is to determine how the household's supply attributes in the home improvement function affect the process of making improvements. It is difficult to observe or measure the probability that a household with given attributes will make an improvement. However, given a sample we can identify those households who have made improvements and those who have not made improvements. In this case, the dependent variable, the quantity of home improvements, becomes a binary variable. The Home Improvement function (7.4) can then be expressed in the form; HI (1, 0) = H (Ho, Y, F) (7.5) where HI = l, is the case where the household made an improvement, and HI = 0, is the case where the household did not make an improvement. We hypothesize a positive relationship between the probability of making an improvement and housing value, the household's monthly disposable income, the size of the household, the years of residence at the present site, the presence of lodgers and the one-time receipt of special income. We hypothesize a negative relationship between the probability of making an improvement and the age of the household head. If the head of the household is older, given years of residence, the less likely she or he will have.made improvements to the dwelling because of her or his stage in the life cycle. In the following sections we 142 empirically test these hypotheses. Empirical Specification Our basic assumption is that the probability of making an improvement to the dwelling, Pi' can be character- ized by the logistic form, -1 Pi = [1 + e ] (7.6) where 80 is a constant, 8 is the coefficient vector and Xi is a vector of explanatory variables in (7.5). (Pindyck and Rubinfeld, 1981; Watson, 1974). Alternatively we can write your logistic function in the form, p. log [1:57] = 80 + slei + 32 Yi + 33 115i + 34 AGEHi l + B YROCi + B D . + (7.7) 5 6 11 B7 ”21 ' where P = probability of making an improvement, HV = the value of the dwelling, Y = the household's monthly disposable income, HS = the equivalent household size, AGEH the age of the head of the household, and YROC = the number of years the household has occupied the present site. D1 and D2 are two dummies for the presence of lodgers and the one-time receipt of special income. We are interested in two kinds of actions by owner occupant households; the making of an improvement and the adding of a room or more to the dwelling. In our sample 143 survey, we have data on the types of improvements made by the households. There are fourteen types altogether. These are additional rooms, better kitchen, more convenient water, a better toilet, plaster and paint, better basic wall materials, better roofing, inside ceiling, improved flooring, better \ windows or doors, terrace or porch, fence around the property, earth fill and other unspecified improvements. However, in the collection of the data, only types of improvements were counted, and not volume or repeats of given types. For example, a roofing change counted only once. These improvements are therefore not quantifiable for equal weight. One cannot make types of improvements the dependent variable. On the other hand, we can identify from the data which households made any type of improvement and which did not. We can analyze the home improvement process in terms of binary choice. We use logit anlayze for this analysis because it is the most appropriate method of analysis for use in binary choice studies (Watson, 1974). However, because one or two types of improvements may be insignificant minor changes, we shall look at the making of three or more types of improvements instead of the making of any improvement to the dwelling. We look at the addition of rooms to the dwelling because it is one of the major improvements carried by owner occupant households. In our sample survey we found that 69.4 percent of the owner occupant households carried out three or more improvements to the dwelling and 42 percent of the improving 144 households built one or more additional rooms. An extra room can be a source of rental income for the household, but also expands the capacity of the housing stock. We shall use equation 7.7 to estimate the parameters associated with making three or more improvements and adding a room or more to the dwelling in the following section. The estimation was done with the maximum likelihood estima- tion procedure, using the program QUAIL, developed at the University of California (Berkman, et. al., 1979). Empirical Results The estimated logit equations are shown in Table 7.1. The dependent variables are the logarithm of the odds that a particular choice will be‘made: in this case, that the household will make three or more dwelling improvements and that it will add one or more rooms. We postulated that households with higher incomes are most likely to make additional improvements to the dwelling. Holding everything else constant, households with higher incomes are more likely to have more money put aside for dwelling improvements. However, the estimation results do not demonstrate that more income makes it likely for a household to carry out additional improvements. The coefficients for both types of improvements-~making three or more improvements and adding a room are insignificant, and the trend is even in the wrong direction. For our average owner occupant in Table 7.2 with a 145 TABLE 7.1 MAXIMUM LIKELIHOOD ESTIMATES OF HOME IMPROVEMENT EQUATIONS INDEPENDENT VARIABLES DEPENDENT VARIABLE Making Additional Making Additional Improvements Rooms Q3 Q1 L09 (IT—6;) L09 ("l-7‘51“) ESTIMATED LOGIT COEFFICIENTSa Household disposable -0.358 x 10-3 -0.166 x 10-2 income (-.167) (0.823) Housing value 0.561 x 10‘3b 0.611 x 10'4 (2.927) (0.774) Household size 0.119 0.064 (1.084) (0.688) Age of household head -0.028d -0.313 x 10‘2 (1.333) (0.168) Years of residence at 0.108C 0.115b the site (1.762) (2.161) Lodgers = 1 if yes, 0.587 -0.236 = 0 if no. (0.889) (0.485) Special income = 1 if yes, 0.468 0.260 = 0 if no. (0.787) (0.559) Constant -0.143 -1.2606 (0.142) (1.365) Likelihood ratio statistic 56.89 12.79 SOURCE: Calculated from data in M. Ndulo, 1979 Zambia Home Improvement Survey. D) Asymptotic t—statistic in parentheses. O" coefficient significant at the 0.05 level. coefficient significant at the 0.10 level. 940 coefficient significant at the 0.20 level. 03’ probability of making three or more improvements. 01’ probability of adding a room or more. 146 ~ TABLE 7. 2. PROBABILITIES OF MAKING IMPROVEMENTS FOR AN AVERAGE OWNER OCCUPANTa *3» Probabilities or more Selected Attributes Types of Additional Improvements Rooms Average owner occupant ' household 0.748 0.404 Average owner occupant household with lodgers 0.843 0.348 Average owner occupant household with lodgers and had special income 0.896 0.409 SOURCE: Calculated from Table 7.1. aAn.average owner occupant household occupying a dwelling with a housing value of $2792.89, with a disposable income of $162.50, 6 years of residence at the present site, an equivalent household size of 6 and with a household head, 41 years old. 147 monthly disposable income of $162.50, the probability of making three or more improvements is 0.748. When income falls to say $90.00, the probability is increased to 0.752, and when income rises to say $300.00, the probability is reduced to 0.741. Similarly, the probability of adding rooms is 0.404 for the average owner occupant. When income falls to $90.00, the probability is increased to 0.428. When income rises to $300.00, the probability is reduced to 0.361. It is plausible that owner households will opt to buy another dwelling, rather than stay on and improve the dwelling when income increases. Owner occupant households with higher incomes are also likely to start with a better quality dwelling which needs less improvement than otherwise. This is likely to be the case in Kaunda square where the households started off with a higher level of services to the dwelling than most other households in the sample. The estimated results for housing value are as we expected. They demonstrate that the higher the value of the dwelling, holding everything else constant, the more likely that households will have made additional improvements to the dwelling. The coefficient is significant at the 0.05 level for making three or more improvements. It is insignificant for making additional rooms. We included the household size variable in the logit regressions to allow for the possibility that the size of the household would have an important effect on making 148 improvements. In our sample survey, we found that house- holds added rooms mainly because of additional children and relatives. Furthermore, the larger the household, the more likely that there will be some disguised unemployed workers who can help in making the improvements. Our estimated results suggest that, holding everything else constant, the larger the household size, the more likely that the house- hold will make additional improvements. However, neither coefficient is significant. We also included the age of the head of the household. Our expectation is that as the age increases, the probability of making an additional improvement to the dwelling declines. Our estimated results demonstrate that this is the case. The coefficient for making three or more improvements is significant at the 0.20 level, while that for adding a room is not significant. The residence variable represents the number of years that the household has lived at the present site. It is expected that as the time of residency increases, the more likely that households feel attached to the community. Other things equal, the more that households feel attached to the community, the more likely that they carry out improvements to the dwelling. Besides, there will have been more time for improving. The estimated results show that the longer the time of residence at the site, the more likely that the household will make additional improvements. The results also show that the estimated coefficients are 149 significant. We have two dummies in our regressions; the presence of lodgers and the receipt of a one-time special income. We expect that households with lodgers are more likely to make improvements to the dwelling. We also expect that households who have received some special income, such as an inheritance or an unusually good loan, before deciding to buy or build the house are more likely to make additional improvements to the dwelling. Our estimated results confirm our prediction for making additional improvements (three or more different types). Households with lodgers are more likely to make additional improvements than those without. An average owner occupant household without lodgers has a probability 0.748 of making additional improvements. If the same household has lodgers, the probability increases to 0.843. If furthermore, the household had received some special income before building or buying the house, the probability increases to 0.896. However, contrary to our expectations, households with lodgers are less likely to add rooms to their present dwelling. Presumably, the fact that they have lodgers means that they have already achieved their optimum number of rooms in the house. Therefore any kind of improvement will be directed to improving the quality of the dwelling. Thus, for the average owner occupant household without lodgers, the probability of adding rooms is 0.404. For 150 those households with lodgers, the probability declines to 0.348. If the same household had received special income, the probability of adding a room increases to 0.409. However, both the dummy coefficients for lodgers and special income are insignificant. ConclusiOn In this chapter we attempted to estimate the probab- ility of an owner occupant household making additional improvements (three or more different types) to the dwelling. Our empirical results agreed with our theoretical expectations except for the changes in the household's monthly disposable income. The direction of the empirical relationship between adding rooms to the dwelling and the presence of lodgers is also contrary to Our theoretical expectations. Our estimated results seem to indicate that the value of the dwelling and the years of residence at the site are the most important predictors of additional improvements having been made to the dwelling. CHAPTER EIGHT CONCLUSIONS AND POLICY CONSIDERATIONS The purpose of this study was to analyze the operation of a low income housing market in urban Zambia, taking that of Lusaka as a case study. In so doing, we tried to answer certain questions in regard to the demand for housing and its attributes, the relative importance of the components of the housing bundle and we examined the process of home improvement, in terms of how much of it goes on and what kinds of households are likely to have carried out improvements to the dwelling. The analysis of these aspects of the study was achieved through the analysis of the cross section data collected from a sample of homeowners from the low income housing market in Lusaka. The major techniques used in the study are ordinary least squares, hedonic and logit regression analysis. Conclusions In the analysis of the data, several conclusions were derived in answer to our questions. These are outlined below. 151 152 1. Demand for Housipg. We have examined the response of housing consumption to income and taste variables of the household. Our results showed that the income elasticity of demand for housing is 0.6. This implies that the demand for housing is relatively insensitive to income. Our analysis is of course based on cross section micro observations and uses current income, rather than permanent income, for the income variable. However, our estimated result is similar to that of studies elsewhere, both in developing and developed countries, which have often shown the income elasticity to be less than one. For example we noted that the income elasticity of demand for owners has been estimated at about 0.8 for Colombia and 0.21 for South Korea. The estimates from similar studies in the United States vary between 0.2 and 0.5. The elasticity estimates appear especially low for studies limited to parts of the market, such as renters only, low income households only, residents in one neighbor- hood, etc. However, although less than one, the elasticity might be higher in LDC's because households rarely have other investment opportunities--except for education, small businesses, and work materials. Our analysis of the taste variables showed that the equivalent household size elasticity of demand for housing is 0.6. This is similar to that of income and implies that the demand for housing is relatively insensitive to the size 153 of the household. This suggests that urban demand for housing will not be much affected by future changes in the average household size. When households are located in either a site and service scheme or an improvement area, their demand for housing increases, compared with those households located in a squatter settlement. This implies that site and service and squatter upgrading schemes are beneficial, since they bring about changes in the quality of the neighborhood in such a way as to increase the demand for housing by households. Our analysis found a sharp distinction between the demand patterns for households with lodgers and those without lodgers, and those with more education (post primary education) and those with less education (primary education). On the whole, households with lodgers, and those with less education demanded more housing services than those without lodgers and with more education. 2. The Value of Attributes in the Housing Bundle. In terms of the relative importance of the attributes in the housing bundle, we found that the most important attributes in explaining the value of the house are the number of rooms in the dwelling, the quality of wall materials, plaster and paint, and access to better water and better sanitary facilities. Households attached more importance to space and the permanency of the dwelling implied by the quality of the wall materials than they did to the quality of water and 154 sanitary facilities. 2. Demand for Dwelling Attributes. Our analysis for the demand for dwelling attributes is preliminary and exploratory. We limited our analysis to five dwelling attributes, in other words, assuming that implicit markets for other attributes can be ignored. We also assumed that cross price elasticities are zero. Given these restrictions we were able to say something about the nature of the demand for dwelling attributes. The demand for housing attributes was price elastic and income inelastic with respect to yard area and age, and price and income inelastic with respect to access. It was price and income elastic with respect to living area. This implies that only the demand for living area is relatively elastic and will be much affected by changes in its own price and income. However, the income and price elasticities of demand vary for different attributes. The growth of income would therefore result in changing patterns of demand for attributes. 4. Home Improvements. Our analysis showed us that most owner occupants over time improve their dwellings. 95 percent of the households carried out one or more types of dwelling improvements with a mean of about four types of improvements per household. Most households used hired labor and financed their improvements with accumulated savings and wages. Furthermore, households who are older, who live in highly valued dwellings, 155 and those with long years of residence at the site are likely to have made more dwelling improvements to the dwelling. Policy Considerations The main contribution of this study is in increasing the knowledge about the low income housing market in Zambia. The results obtained are of course limited to Lusaka and specifically to low income owner occupants. There is therefore need for further research of the same kind both in Lusaka and other urban areas to reject or verify the conclusions of this study. Further research is also needed to analyze the rental market, and other aspects of the low income urban housing market, such as its relation to public housing. However, given the modest nature of our study and its limitations, we have now general and empirical evidence to use in the formulation of policy, which did not before exist. On the basis of this, several implications for policy can be discussed. At the outset, it is important to stress that if policy makers are to deal effectively with housing problems, specific housing policies should be underpinned by a properly conceived overall policy framework. The focus of urban low income housing policy is to improve the housing situation of low income households through squatter upgrading and site and service schemes. To be effective these specific policies 156 need to be coordinated with the overall policy framework. This should include the reorientation of such policies as monetary and fiscal policies to be beneficial to the housing market, inducing the development of grassroot financial institutions and the reorganization of the public housing institutions. Currently enormous public resources are spent on high income housing. This bias in resource use is an institutional characteristic of the urban housing sector. Policy makers should strive to change this situation. Policies should be directed to deemphasize the unlimited commitment of the government to provide housing for its workers. Resources saved with such policies should be used to stimulate low income housing through squatter upgrading and site and service schemes. It is in this context that specific policies addressed to low income housing issues will have a desirable impact on their solution. Our analysis implies that in the case of Lusaka, policies of squatter upgrading and site and service schemes are beneficial to low income housing. Such policies have in the final analysis led to improved housing conditions for the low income urban population. Governments faced with severe housing problems and a large squatter pOpulation are better off encouraging squatter upgrading, rather than squatter eradication. Sites should also be provided on which house- holds can acquire plots to build houses. The question which has often arisen is about the 157 initial level of services to be provided on such plots. Our analysis shows that households value space and the permanency of their dwellings more than any other component of the dwelling. Better wall materials, number of rooms and plaster and paint are highly ranked. On the other hand, the provision of better water and sanitary facilities such as piped water and flush toilets are not as highly ranked as the former. This implies that the current policy of providing standard sites rather than normal sites in the site and service schemes in Lusaka is justified. Policy makers now have empirical evidence to use in determining how housing demand responds to, for example, income growth and population changes--both in terms of its growth and migration. They should understand better what form of subsidy is best for increasing the consumption of housing by low income households, given an inelastic demand for housing. If policy.makers wished to increase housing consumption through transfer payments to low income households, they are better off earmarking them to specific housing expenses, otherwise only about 60 percent of the transfer payment will be spent on housing. The patterns of demand for dwelling attributes also change with income, price and population changes. Policy makers can now determine in what way they will change and how much households are willing to pay for some specific attributes. This kind of information is important for 158 project analysis. I The ranking of the various attributes indicates to the policy maker which attributes matter most to households. On this basis a selective policy to help low income house building activity can be focused on those attributes, such as subsidized loans towards the building of an additional room. The higher demand for housing by households with lodgers means that an enlightened rental policy--i.e. encouraging households to take in lodgers, is important. Such a policy will also, through this new source of income, increase the households' ability to improve and expand the house and thus increase rental housing. The present system of technical assistance to help households build a better house is directed at the households, on the assumption that they will use self help labor to improve the dwelling. However, most households use hired labor. There is, therefore, a need to reexamine this policy, and perhaps to direct technical assistance and advice to small scale builders. One other important issue exists for policymakers. Low income households are by definition poor households. If there are to be substantial house building activities, potential sources of finance must be generated to pay for both materials and labor. Therefore, institutions are needed to develOp these potential credit sources for home improvements. APPENDICES APPENDIX A RESEARCH METHODOLOGY The data for this study was collected in Lusaka by the author with the help of five University of Zambia students during the summer of 1979. The collection of the data covered a period of three months--June to September. The project was part of the international reserach on low cost housing and employment generation in LDCs underway at Michigan State University and included Sri Lanka, Columbia, Pakistan, Kenya, Tunisia, Peru and Zambia. The research was sponsored by the Bureau of Science and Technology, Agency for International Development. The focus of the research was the low income urban population living in private low income settlements. Information was gathered through a sample survey of the targeted population. The questionnaire for the survey was designed at Michigan State University by the principal investigator of the international project. Lusaka was selected as the site for the research in Zambia because the author is familiar with Lusaka. Further- more, it was believed that research assistants and transpor- tation would be readily available there. The questionnaires 159 160 for the survey were written in English and then were translated into the local languages during the interviews; it was easier to do this for Lusaka than it would have been elsewhere. It was decided to sample households in selected low income settlements. These represented a site and service scheme, a squatter settlement, an upgraded housing scheme and an expandable core housing scheme. Kaunda Square was chosen for a site-and-service scheme, and Chawama for an upgraded housing and an expandable core housing scheme because they were the oldest areas with such schemes. Bauleni was chosen for a squatter settlement on the basis of discussions with housing officials in Lusaka. The upgraded and expandable core housing sections of Chawama are together formally known as the Improvement Housing area. We lump these two schemes together in the study. Given the limited time and resources available, it was decided to interview 168 households with a quarter from each housing scheme. However, only 162 households were interviewed. Six households were either unavailable or were unacceptable respondents. The distribution of the respondents is shown in table A.l. The five University of Zambia students employed as research assistants for the survey all had previous field research experience. Initially, the research assistants studied the questionnaire for two days at the 161 TABLE A.1 DISTRIBUTION OF SAMPLE SURVEY RESPONDENTS Area of Survey Respondents Potential Actual Kaunda Square Site and Service 42 36 Bauleni Squatter Settlenent 42 41 Core Housing 42 41 Chawama Up-Graded 42 42 TOTAL 168 162 162 University of Zambia Main Campus. The study covered concepts, definitions and the structure used in the questionnaire. Then a pilot survey of twenty-five households was undertaken in Kalingalinga so that the research assistants could be further familiarized with the questions. The households interviewed in Kaunda Square were selected, using a table of random numbers, from a map of Kaunda Square provided by the Lusaka City Council. Kaunda Square Stage I has a total of 119 blocks, with an average of 18 houses in each block. We numbered the 119 blocks and, using a table of random numbers, we came up with a sample of 50 blocks. We then assigned numbers to the houses in the 50 blocks and came up with the 42 houses to be surveyed. In Chawama, samples were taken from both the expandable core-housing and the up-graded sections. In the expandable core-housing section, the sample was chosen from a map provided by the Lusaka Housing Project Unit. Houses were assigned numbers and then 42 houses were chosen using a table of random numbers. The numbering of houses in the up-graded part of Chawama was disorderly and there was no map available. However, the area was clearly divided into sections for political—party organization. We used these sections to select our sample of 42 households for the area. The same procedure was used in Bauleni. Each questionnaire was checked at the end of the day 163 by the author with the help of Dr. David Todd of the Urban Community Research Unit, University of Zambia. Questionnaires which had unsatisfactory answers were followed up the next day. The major problem experienced by the research assis- tants was that of finding the respondent at home. Initially, our surveys were done during the day. This produced very few completed questionnaires. After three days we realized that respondents were not usually at home in the day during the week, since most of them worked. We therefore changed our interview schedules to evenings and weekends. This was more successful in contacting the respondent at the first visit. After the survey was complete, the data was punched and put on tape for analysis and storage. Both the questionnaire for the survey and the data are available at Michigan State University. 164 .u:«m& was umummflm n mun “mowuwawomu umum3 u man “wamflumume Ham3 u qzn «mamaumuse moon a man kneauwawomm Suspends I mmn .wuwm on» no swam can» u Gad» «xuoz cu mafia He>muu u x309 umcHHHm3o may no wow u med umEDOH mo Hones: u zoomoz ”menu womanhood“ u mo I udtm «mmqdem<> WBNMUWHD “mmfldemdfi mDOZHBZOUm oo.H sH.o ea.o eo.o HH.e eH.o oa.e- eH.o- om.o on.o He.o one .HH oops em.o me.o se.o ao.o- em.o- oH.o- s~.o oa.o ~e.e one .oe oo.H e~.o e~.o No.o- o~.ou eH.o- e~.o ea.e em.e men .s eo.H ee.o Hm.ou ve.o- «H.o- s~.o ne.o em.e use .e oo.H se.eu em.o- mo.o- m~.o oH.o ~m.e axe .s oo.H ea.o so.o «H.o co.o m~.ou not .e oo.e ma.o eH.o- so.o- m~.o- one» .m eo.H oH.on ~o.o H~.ou egos .e oe.H ~m.o ~m.o zoomoz .m oo.a oe.o monsoon .~ oo.H mean .H as ea e o s e n e n N H flmmDA¢> OZHAAmzn ho mHZ¢ZH2¢mBmQ m ”HQZMAAG ”mmqdem¢> UZOZd mfiszUHhhmOO ZOH8¢AmmmOU APPENDIX C A NOTE ON MONETARY VALUES All monetary values in the study were converted to U.S. dollars at the prevailing exchange rate in 1979. The rate was 1.00 U.S. dollar = 0.79 Zambian kwacha. 165 BIBLIOGRAPHY BIBLIOGRAPHY Advisory Commission on the Review of the Constitution of the Federation of Rhodesia and Nyasaland. (1960). Survey of DevelOpments since 1953. London: Her Majesty's Stationery Office. Andrews, F.M. and Phillips, G.W. (1970). "The Squatters of Lima: Who They Are and What They Want". The Journal of Develpping Areas. January. Baldwin, R.E. (1966). Economic Deve10pment and Export Growth: A Study of Northern Rhodesia 1920-1960. Berkeley: University of California Press. Ball, M. and Kirwin, R. (1975). The Economics of an Urban Housing Market, Bristol Area Study. London: Center for Environmental Studies Research Report No. 15. Banda, M.C. (1978a). Observations of a Mutual Help Scheme in Desai: Mutual Help Case Study No. 2. Working Paper No. 13. Lusaka: Lusaka City Council Housing Project Unit. , (1979). People Buying Plots and Houses in Chawama Overspill and Chawama Existipg. Working Paper No. 29. Lusaka: Lusaka City Council Housing Project Unit. , (1978b). Garden Mutual Help Study. Working Paper No. 20. Lusaka: Lusaka City Council Housing Project Unit. Berkman, J., Brownstone, D., and Associates. (1979). QUAIL User's Manual. Department of Economics. University of California. Boswell, D.M. (1975). Buildings and Households in Roberts/ Chawama, A Comparison of the Situation in 1965 and 1974. University of Zambia. Unpublished Manuscript. Boyes, W.J. and Gerking, S.D. (1980). "The Role of the Functional Form in Estimating Elasticities of Housing Expenditures". Southern Economic Journal. Vol. 47, No. 2, October. 166 167 Burns, 3., and Grebler, L. (1976). "Resource Allocation to Housing Investment: A Comparative International Study". Economic Development and Cultural Change, October. , and Khing, T. (1967). "Does Good Housing Contribute to Sound Economic Development?". Journal of Housing. March. Carliner, G. (1973). "Income Elasticity of Housing Demand". Review of EconOmics and Statistics. November. Davies, D.H. (1969). Lusaka, Zambia: Some Town Planning Problems in an African City at Independence. Zambia Urban Studies No. 1. Institute of African Studies. University of Zambia. DeLeeuw, F. (1971). "The Demand for Housing-A Review of Cross Sectional Evidence". "Review of Economics and Statistics. February. Ereckson, H.O. and Witte, A.D. (1979). "The Demand for Housing: Comment". Southern Economic Journal. October. Follain, J., Lim, G., and Renaud, B. (1980). "The Demand for Housing in Developing Countries: The Case of Korea". Journal of Urban EdonomicS. May. Freeman III, A.M. (1974). "Air Pollution and Property Values: A Further Comment". Review of Economics ‘and Statistics. November. , (1979a). "The Hedonic Price Approach to Measuring Demand For Neighborhood Characteristics". In The Economics of Neighborhood, edited by David Segal. New York: Academic PreSs. , (1979b). "Hedonic Prices, Property Values and Measuring Environmental Benefits: A Survey of the Issues". Scandinavian Journal of Economics. Goodman, A.C. and Kawai, M. (1980). 'Permanent‘IncomeL Hedonic'Prices and Demand for Honsing: New Evidence. Working Papers in EconomIcs No. 61. The Johns Hopkins University. June. Grether, D. and Mieszekowski, P. (1974). "Determinants of Real Estate Values". ‘Journal of Urban Economics. April. Harrison, D. and Rubinfled, D.I. (1978). "Hedonic Housing Prices and the Demand for Urban Air Quality". JOurnal of Environmental Management, (5). 168 Ingram, G.K. (1980). Housing Demand in the Developing Metropolis. Economics Department. MIEhigan State University. Unpublished Economic Development Workshop Seminar Paper. Kain, J.F. (1970). "Measuring the Value of Housing Quality". Journal of American Statistical Association, (65). , and Quigley, J.M. (1975). (Housing Markets and Racial Discrimination. New York: NatiOnal Bureau of Economic Research. King, T.A. (1976). "The Demand for Housing: A Lancastrian Approach". Southern Economic Journal. October. Knauder, S. (1978). Integrated Housing in Zambia. Lusaka. Unpublished ManuscrIpt. Lancaster, K. (1966). "A New Approach to Consumer Theory". Journal of Political Economy. April. Lee, T.H. (1963). "Demand For Housing: Cross Section Analysis". Review of Economics and Statistics. May. . (1968). "Housing and Permanent Income: Texts Based on a Three Year Re-interview Survey". Review of Economics and Statistics. May. Linn, J.F. (1979). Policies for Efficient and Equitable Growth of CitIes in Developing Countries. World Bank Occasional Paper. Linneman, P. (1981). "Some Empirical Results on the Nature of the Hedonic Price Function for the Urban Housing Market". Journal of Urban Economics. March. Lucas, R.E.B. (1975). "Hedonic Price Functions". Economic Inquiry. June. Lusaka Housing Project Evaluation Team. (1976a). Study of Plot Use in Site and Service Schemes'and a quatter Area in Lusaka. Working Paper No. 2. August. Lusaka: Lusaka City Council Housing Project Unit. , (1976b). A Report on Observations of'a Pilot Mutual Help Scheme in NYerere in Lusaka. August. Lusaka: Lusaka City Council HousIng Project Unit. , (1977a). George EsSential Resettlement: The Drop Outs. Working Paper No. 7. November. Lusaka: Lusaka City Council Housing Project Unit. , (l977b). Community Development and Change: The Community Development Process in Chawama and George. Working Paper No. 10. September. Lusaka: Lusaka City Council Housing Project Unit. 169 , (1977c). Application, Allocation and Briefing Procedures for the Serviced Plot Programme. Working Paper No. 11. September. Lusaka: Lusaka City Council Housing Project Unit. , (l977d). George OverSpill House Consolidation: Preliminary Results. Working Paper No. 12. October. Lusaka: Lusaka City Council Housing Project Unit. , (l977e). Chaisa Complex: Some Population Statistics. Working Paper No. 5. May. Lusaka: Lusaka City Council Housing Project Unit. , (l977f). Employment Generation Study: First Re rt. Working Paper No. 14. November. Lusaka: LusaEa City Council Housing Project Unit. Maclennan, D. (1977). "Some Thoughts on the Nature and Purpose of House Price Studies". Urban Studies. Vol. 14. Maimbo, F.J. and Fry, J. (1971). "An Investigation into the Changes in the Terms of Trade between the Rural and Urban Sectors of Zambia". African Social Research. December. Maisel, J.S. (1971). "The Demand for Housing: A Comment". Review of Econdmids and Statistics. Vol. 53. Mangin, W. (1967). "Latin American Squatter Settlements: A Problem and A Solution". ‘Latin AmeriCan Research Review. Martin, R. (l977a). "Lusaka Residents are Licenced". Geographical.Magazine. May. , (1977b). "Housing Options, Lusaka, Zambia". Eskistics. August. Mayo, S.K. (1981). "Theory and Estimation in the Economics of Housing Demand". JOurnal of Urban Economics (10). McDougall, 6.5. (1976). "Hedonic Prices and the Demand for Local Public Goods". Public Finance. Vol. XXXI. No. 2. McLain, W.T. (1978). "Legal Aspects of Housing and Planning in Lusaka". In Kanyeihamba, G.W. and McAuslan, J.P.W.B. (ed.) Urban Legal Problems in Eastern'Africa. Uppsala: Scandinavian Institute of African Studies. Mendelsohn, R. (1977). "Empirical Evidence on Home Improvements". Journal of Urban Economics. (4). 170 Mohan, R. (1979). Urban Economic and Planning Models. World Bank Staff Occasional Papers. No. 5. Mubanga, A. (1979a). Survey of Applicants to Sites and Services Earning Less than K70.00 to Assess the Low Takeepp Of Plots. Working Paper No. 34. Lusaka: Lusaka City Council Housing Project Unit. , (1979b). Survey of Applicants to Sites and Services Earning Less than K85.00 to Assess the Low Take-up of Plots. June. Lusaka: Lusaka City CounEil Housing Project Unit. , and Turok, M. (1979). Progress of Sh0ps and Small Industny Plot Development. September. Lusaka: Lusaka City Council Housing Project Unit. Mulenga, A.D. (1978). Mutual Help BlOck-making in Garden Overspill. Workifig Paper No. 23. Lusaka: Lusaka City Council Housing Project Unit. Mulford, J. (1979). Income Elasticity of Housing Demand. Housing Assistance Supply Experiment R—2449-HUD. Santa Monica, California: Rand Corporation. Muth, R.F. (1960). "Demand for Non-Farm Housing". in A.C. Harberger (ed). The Demand for Durable Goods. Chicago: University of Chicago Press. , (1969). Cities and HOusing. Chicago: University of Chicago Press. v Mwiinga, H. (1979). "Housing for the Zambian Family". In Situ. August. Ndulo, M. (1979). l979'Zambia Home Improvement Survey: Report of a Project on Low Income Housing and Employment Generation. EconomiEs Department. Michigan State University. , (1981). An Analysis of Low Income Demand for Shelter: The Case of Lusaka, Zambia. EconomiEs Department. Michigan State University. Nelson, J.P. (1978). "Resdential Choice, Hedonic Prices, and the Demand for Urban Air Quality". Journal of ‘Urban Economics. (5). Northern Rhodesia Government. (1952). Northern Rhodesia Annual Report. Lusaka: Government Printers. Nziramasanga, M. and Obidegwu, C.F. (1981). Copper and Zambia. Lexington: Lexington Books. 171 Olsen, E.O. (1969). “A Competitive Theory of the Housing Market". ‘American’EconOmic Review. September. Pindyck, R. and Rubinfeld, D. (1981). Econometric Models and Economic Forecasts. New York: McGraw Hill. Polinsky, M.A. (1977). "The Demand for Housing: A Study in Specification and Grouping". Econometrica. March. , and Elwood, D.T. (1979). "An Empirical Reconciliation of Macro and Grouped Estimates of the Demand for Housing". ‘Review of Economics and Statistics. May. Pollakowski, H.O. (1982). Urban Housing Markets and Residential Location. Lexington: Lexington Books. Quugley, J.M. (1979). "What Have We Learned About Housing Markets?". In Current ISsues in Urban Economics, edited by Mieszkowski, Peter and Staszheim, Mahlon. Baltimore: The Johns Hopkins University Press. , (1982). "Nonlinear Budget Constraints and Consumer Demand: An Application to Public Programs for Residential Housing". JOurnal of Urban Economics. (12). Rakodi, C. (1978). ‘George 1976: ‘Initial Results of the First Primary Sample Survey. Operational and Poliey Implications. Supplement to Working Paper No. 4. April. Lusaka: Lusaka City Council Housing Project Unit. Reid, M.G. (1972). Housing and Income. Chicago: University of Chicago Press. Republic of Zambia. (1966). First National Development Plan 1966-1970. Office of National Development and Planning. Lusaka. , (1971). Second National Develppment Plan 1972-1976. Ministry of DevelOpment Planning and National Guidance. Lusaka. , (1979). Third National Development Plan 1979-1983. National Commission for Development Planning. Lusaka. , (1972). ‘Annual Report of the Department of Town and Country Planning for 1970. Lusaka: Government Printers. , (1974). Annual'Reporpvof the Department of Town and Country Planning for 1971. Lusaka: Government Printers. 172 , (1974). The Housingi(Statutory and Improvement Acts) Act No. 30 of 1974. Lusaka: Government Printers. , (1979). Monthly Digest of Statistics. April/June. Lusaka: Central Statistical Office. , (1979). 1974 Sample Census of Population. Second Reper . Lusaka: Central Statistical Office. Richards, P.J. (1979). "Housing and Employment". International Labor Review. January/February. Robertson, J.T. (1975). "The Urban Situation: Shelter". in Human Settlements in Zambia. Report of the Workshop held at the University of Zambia. 12-14 September. Rosen, S. (1974). "Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition". Journal of Political'Economy. January/February. , and Brown, J.N. (1982). "On the Estimation of Structural Hedonic Price Models". Econometrica. May. Sanyal, B. (1978). Cost Recovery in the Lusaka Urban Development Project. 'First Draft. Lusaka: Lusaka City Council Housing Project Unit. , (1981). "Who Gets What, Where, Why and How: A Critical Look at the Housing Subsidies in Zambia". Development and‘Change. ‘Vol. 12. Seymour, T. (1976). "The Causes of Squatter Settlements: The Case of Lusaka, Zambia, in an International Context". in Simons, H.J. et. a1. Slums or Self- Reliance?. Urban Growth in zambia. Institute 5? African Studies. University of Zambia. Lusaka. Simoko, P. (1977). It Will Be Better To—morr0w. Lusaka. Variety Printers. Simons, H.J. et. al. (1976). 'Slums or Self-Reliance? Urban Growth in zambia. Institute or African Studies. University of Zambia. Lusaka. Singini, R.E. (1978). George 1978 Primary Surveys I and III - Comparisons and their Operational and Policy Implications. Working Paper No. 21. September. Lusaka: Lusaka City Council Housing Project Unit. Strassmann, W.P. (1970). "The Construction Section in Economic Development". Scottish Journal of Political Economy. November. 173 Sunday Times of Zambia. 1981. February 22. Tipple, A.G. (1976). "Self-Help Housing Policies in a Zambian Mining Town". Urban Studies. June. Todd, D. (1979). Draft Proposal for a Joint Study of Small-Scale Enterprises in LusakaL Leading to Action-Research Projects. Lusaka: Urban Community Research Unit. Institute of African Studies. University of Zambia. Turner, J.F.C. (1967). "Barriers and Channels for Housing Development in Modernizing Countries". Journal of the American Institute of Planners. Vol. XXXIII. No. 3. May. , (1968). "Housing Priorities, Settlement Patterns and Urban Development in Modernizing Countries". Journal of the American Institute of Planners. Vol. XXXIV. No. 6. November. Turin, A. (1974). "The Role of the Construction Sector in Economic DevelOpment". International Development Review. UN/ECA/FAO. (1964). Report of the UN/ECA/FAO Economic Survey Mission 0n the Ec0nomic Development of Zambia. Ndola: Falcon Press. UN/ECA/Bouwcentrum. (1973). Third International Course on House BuildingfiCosts. 'East and West Africa. The Netherlands. August-November. Vaughan, G. (1976). "Sources of Downward Bias in Estimating the Demand Income Elasticity for Urban Housing". Journal of Urban Economics. January. Ward, P. (1981). "A Review of George-Development of a Squatter Settlement in Lusaka, Zambia". Third World Planning Review. Vol. 3, No. 3. August. Watson, P.H. (1974). "Choice of Estimated Procedures for Models of Binary Choices". Regional and Urban Economics. (4). Wilkinson, R.K. (1973). "The Income Elasticity of Demand for Housing". Oxford Economic Pepers. November. Winger, A.R. (1968). "Housing and Income". Western Economic Journal. Vol. VI, No. 3. June. Winpenny, J.T. (1975). Housing and Jobs for the Poor. Development Planning Unit Paper No. 2. University College, London. 174 Witte, A.D., Sumka, H.J., and Ereckson, H. (1979). "An Estimate of a Structural Hedonic Price Model of the Housing Market: An Application of Rosen's Theory of Imperfect Markets". EcOnometrica. September. "Iillillllllfilllllis