Households and housing : residential mobility, tenure choice, and space consumption in the developing metropolis

Urban and Regional Report No. 81-20 - HOUSEHOLDS AND HOUSING: RESIDENTIAL MOBILITY, TENURE CHOICE, AND SPACE CONSUMPTION IN THE DEVELOPING METROPOLIS by Andrew Mfarshall Hamer December 1981 This report was prepared under the auspices of the City Study Research Project (RPO 671-47) as City Study Project Paper No. 26. The views reported here are those of the author, and they should not be interpreted as reflecting the views of the World Bank or its affiliated organizations. This report is being circulated to stimulate discussion and comment. Urban and Regional Economics Division Development Economics Department Development Policy Staff The World Bank Washington, D.C.  ABSTRACT Households and Rotusing: Residential Mobility, Tenure Choice, and Space Consumption in the Developing Metropolis Bogota and Cali are cities of considerable size populated by households which, in the main, have fairly high incomes by the standards of the developing world. The household heads tend to be relatively young, overwhelmingly male, and migrants by birth. The latter fact does rot seem to be associated with widespread "marginalization". The households typically include 5 members and that size is declining over time. They live in low density environments, and roughly half awn their dwellings. Residences are more concentrated in particular macro-zones than jobs and in neither case is there a pronounced tendency to live or work near the core. Using various measures, one can identify particular areas of each city which are "affluent" or "disadvantaged". The data reviewed uncovered significant and similar rates of residential relocation over time in both cities. Roughly 23 percent of all households moved within the year prior to the surveys. The one longitudinal study examined suggests that hypermobility is confined to only a small fraction of households. Mobility is held down by rising income, the aging of the household head, and increasing household size. Also depressing mobility are those factors which point to "settled" behavior, such as a lengthy prior tenure and a long commitment to the present job by the household head. Independent of other characteristics, household heads with a recent history of residence in the city are likely to have higher mobility rates thani their counterparts. When patterns of relocation are examined by concentric ring, decentralization appears to be slowing down in Bogota and is fairly modest in Cali. Reverse migration, from farther out to closer rings is rare. Furthermore most moves take place within the ring of origin. The rings are very different in size, and so heterogeneous in their resident household income characteristics that it proves difficult to examine the validity of hypotheses that individuals living in certain zones are "locationally disadvantaged" and face barriers in attempts to move in search of greater opportunities. Using more homogeneous radial sectors as units of observation suggests that a strong pattern of upper- income self-segregation is at work in Cali and Bogota. It is, however,true that sectors characterized as lower income in character do have higher "retention" rates when only one move is considered. An attempt was made to model the determinants of move distances. While city-specific factors play an important role, in both cities an increased length of tenure in the prior residences increases distance moved. Households living in higher density environments tend to move shorter distances; only in Cali, however, is there evidence that  prior residence in a "dis'dvantaged" sector has a negative impact on distance moved. Other variables also prove significanL in only one or the other city. Though some variables of interest are not available (such as- relative prices), it is possible Lo model the tenure choice behavior of owners and renters. The probability of ownership is enhanced by rising household income, the aging of the household head, growing household size and the two indices of household "stability". Other variables such as the number of workers contributing to the household incomes, the recent arrival of the household head to the city, and a location in a lower-income sector prior to the most recent move, all tend to reduce the probability that the household will own. The coefficients obtained in the multivariate estimations have similar values but opposite signs from those found in the mobility equations, suggesting that factors which are associated with greater mobility depress homeownership and vice versa. Owners occupy more space than renters, whether viewed as a whole or within categories like "low income". Low-income households are smaller in size than the average for each over-all category, and their space utilization is also smaller. Among owners, space consumption is enhanced ,by household income and household size. The need to rely on self-help housing methods, to depend on secondary workers, and to house tenants, tends to depress consumption. There is evidence of an independent "sector of origin" effect. The magnitude of these effects is similar in Bogota and Cali. Among renters, space consumption is enhanced by the self-same variables cited in the homeowners case, plus the age of the household head and a history of recent migration into the city. Among the continuous variables tested, there is a broad similarity in elasticities obtained at the mean between owners and renters; one clear exception is the far larger family size effect found among renter households. Origination in a disadvantaged sector depresses space consumption among renters and does so far more than among owners. As expected, pooling the data for owners and renters confirms the role of income, household size, and ownership in boosting space use. Factors such as the reliance on tenants and secondary workers, and previous residence in a disadvantaged sector, all depress space utilization. 6'.  SECTION I: AN INTRODUCTION The Issues There are many reasons for wanting to undertake a detailed analysis of household behavior in making choices related to residential behavior. Those who influence household behavior through public sector investments, taxation policies, and regulatory practices must determine, at a minimum, whether behavior in this area is random or whether, instead, it follows certain predictable patterns which might be taken into account. More specifically the degree and nature of residential mobility is important to anyone concerned with changes in the spatial patterns of location in a rapidly growing metropolis. Do households change residences rapidly over time? Are there certain variables that help identify the degree to which households change residences? Are these movements scattered across the urban landscape or do they cluster within well-defined perimeters? If households do cluster their moves why do they do so? Can such a phenomenon be reconciled with a process of outward movement of jobs and residences over time? Finally, can such a phenomenon be reconciled with the freedom to move laterally, across radial sectors, or are these urban areas mired in a process of ghettoization, where certain sectors act as "black holes", effectively barring its residents from taking advantage of opportunities elsewhere?  -2- The answers to these questions could help clarify the likely effectiveness of policies such as the promotion of transportation invest- ments focused radially upon the traditional center versus those targeted at the promotion of lateral movement in the periphery. Alternatively the location of publicly subsidized housing schemes could be a critical ingredient to their success; there may well be reasons why the organized removal of low income households from certain locations to others are more or less successful. Policy makers often treat shelter as a "basic need" whose provision must be shaped by direct or indirect intervention in the housing market. They must decide whether or not to provide public housing. They must consider whether or not these facilities arc to be owned by a government authority overseeing tenants or by the occupants themselves. They must choose whether or not to construct owner-occupant structures in such a way that they are modifiable over time and open to multiple uses. There are, to name three examples, very different implications to building finished high-rise apartments, providing a single-story shell which can be expanded, and subsidizing nothing but the acquisition of building materials and basic public services. To each one of these concerns the driving forces behind tenancy choice and space consumption are crucial. There are other important issues, as well, upon which tenure and space consumption determinants have a bearing. Governments may regulate land use or financial institutions in ways that have very profound impacts on the shelter options of the poorer households. Would a restructuring  3- of residential subdivision standards or of mortgage institutions be more likely to achieve "basic shelter" objectives than existing insti- tutional arrangements permit? Such questions are crucial to any effective public policy in this field. There are many ways to tackle issues of the type taken up here. The one used is to probe individual household behavior through the use of several surveys undertaken in Colombia. 'These micro-level data sets are examined in often painstaking fashion to develop models of behavior from which, ultimately, policy-relevant conc.lusions may be derived. The Data Base The Bogota Urban Development Study Phase II was undertaken to establish a 1980 plan for the metropolis, dealing with urban infrastructure investment, land use controls, taxation policies and administrative reorganization. The study began in August 1972 and concluded in March 1974. In 1972, a household survey was undertaken, using questionnaires designed to produce detailed information on household characteristics, including those related to housing conditions and the location of household residences and the workplace of labor force participants. The survey was based on a sample of 4,675 households, representing the most complete data collection effort following the Census of 1964. In terms of detail, the Phase II household survey is comparable only to the 1978 survey performed by the national census bureau, the Departamento Administrativo Nacional de Estadistica, in conjunction with the World Bank. The 1978 study was conducted in Bogota and Cali among 3066 and 983 households, respectively. Like the 1972 study it contained  -4- very detailed questions on the residence and work place of the members of households in the labor force. This was supplemented by information on dwelling and structure characteristics, and relevant demographic and socio-econom:Lc data, paralleling, once again, the Phase II survey. Some use was also made of a retrospective history of 730 adults 20 to 45 years of age living in Bogota. This was derived from a survey conducced in 1975 by Robert Corno. The choice of the 20 to 45 age group was meant to focus attention on a large number of migrants. As in the other two surveys, the Corno study was based on a detailed sampling frame (drawn from Phase II, uncorrected for changes thzough 1975) on the basis of which the responses were "expanded" to represent the population of the city as a whole in the relevant age category. The most valuable data collected during this survey were the longitudinal records of the origin and destination of residential moves made by adults throughout their life in Bogota. A General Profile of Households Any explanation of residential behavior in Bogota and Cali can begin profitably with a brief introduction to the conditions in those cities at the time of the surveys. In 1972, Bogota recorded approximately 503,000 households averaging 5.7 persons in size, for a total population of 2,850,000.A 1/ Household responses vary according to the different sets of questions asked. Therefore totals in tables presented below may not correspond to the cited numbers. The 1973 Census reports a total of 495,259 households, averaging 5.2 persons in size, and a total population of 2,571,548. The reasons for the discrepancy have not yet been ex?lored. See XIV Censo Nacional de Poblacion y III de Vivienda (Octobre 24 de 1973), Bogota, D.E. Bogota, Departamento Administrativo Nacional de Estadistica, September 1980, p. 29.  -5- By 1978, the total household count had risen to roughly 722,000, averaging a lower 4.8 persons per unit, for a population of 3,439,000. Cali, by contrast,is a smaller city, with 219,000 households in 1978, averaging 4.8 persons in size, for a population of 1,055,000. 2 Being only 30% the size of the population of Bogota, Cali provides the basis for interesting cross-sectional tests of relationships uncovered in the capital area. For purposes of descriptive analysis, many household variables were aggregated into more manageable categories. One of these was the distribution of income. An attempt was made to define a group of poverty households, proxied by a reported income level (labor plus non-labor income J/) equal to one-half the median i:or the city as a whole. An intermediate income category was defined by the interval between the poverty level and the median, and a final category of "better off" households was allocated to the above-median income group. Because 1972 income was itself reported in interval form, the median had to be estimated using a piecewise Pareto approximation procedure. - That yielded a median income of 1731 pesos 2/ The 1973 Census reports 175,000 households, averaging 5.2 persons in size, for a total population of 918,057. See XIV Censo Nacional de Poblacion y III de Vivienda, Octobre 24 de 1973: Departamento del Valle, Bogota, Departamento Administrativo Nacional de Estadistica, December 1980, p. 63. 3/ A decision was made to omit the imputed income from homeownership in the discussion that follows. 4/ Alvaro Pachon, Urban Structure, Modal Choice, and Auto Ownership in Bogota, 1972, Washington, D.C., World Bank City Study Intermediate Paper No. 31, June 1979, Table 3.  -6- per month. For the sake of simplicity, in this study the level of 2000 pesos per month was used as a proxy for the median, with 1000 pesos per month defining the cut-off below which households were labelled as poor. No effort was made to create a more elaborate set of definitions, where total, adult, or adult-equivalent household size would be considered irn allocating a unit to one or another of the intervals defining the poor, the intermediate-income households and the higher-income households. For 1978, where household income data was reported in absolute terms, the Bogota median was set at 8000 pesos per month, and the poverty cut-off at 4000 pesos per month. In Cali, the levels proved to be lower: the median totalled 7016 pesos per month (rounded off to 7000 pesos) and the poverty level was defined as 3500 pesos per month. Since the cost of living in both cities was approximately the same in 1978, these nominal differences probably represent real differences as well. It is difficult to translate these units into a standard easily identifiable by those unfamiliar with Colombia. Since the dollar is one 5/ This conclusion was originally based on an updated version, by Alvaro Pachon, of work carried out in 1969 by Alberto Hernandez Garcia ("Comparaciones de Precios y Costo de Vida en Cuatro Ciudades Colombianas", Bogota, Universidad de los Andes, Economics Department, Masters Thesis). Hernandez used city-specific market baskets derived from a 1968 survey, bypassing the Census Bureau's outdated 1953 market basket, which remained in use until the late 1970s. The original work suggested Cali's market basket cost only 92% that of Bogota's. The update, using 1978 prices suggests an elimination of any differential. This finds corroboration in the recent work of the Census Bureau. It now uses bundles of goods and services based on 1971 consumption surveys, providing estimates for blue collar and for white collar workers. Taking Bogota values as a base, the December 1978 Cali index for white collar workers is equal to .89 and for blue collar workers is equal to 1.01. (Data supplied by the Census Bureau). AnP assumption of rough parity therefore seems reasonable.  -7- standard unit of comparison, the above cited levels can be converted in dollars of the year in qutestion using purchasing power parity rates rather than foreign exchange rates. The former take into consideration those goods and services that do not enter into international trade,the so-called non-traded goods. The 1972 dollar totals were inflated to 1978 dollar equivalents by simply applying the conversion factor suggested by the U.S. 6/ consumer price index. - The results, which are merely indicative given the manipulations involved, suggest that a poverty level of 1000 pesos per month in 1972 translates into an annual incomne of 2124 dollars (1978), while the proxied median was twice that, or 4248 dollars. For 1978, the poverty level for Bogota was equivalent to 3000 dollars per year, and the median, to 6000 dollars per year. In Cali, the equivalent levels for the poverty cut-off and the median translated into 2628 dollars and 5256 dollars, respectively. The high implicit growth in dollar incomes for Bogota between 1972 and 1978 is an artifact of' differences in income coverage. Based on 6/ Following I. Kravis, A. Heston, R. Summers, International Comparisons of Real Product and Purchasing Power, Baltimore, Johns Hopkins University Press, 1978, the average foreign exchange rate prevailing in 1972 and 1978 was adjusted downward to reflect the purchasing power parity rate, which stood at approximately 40% (pesos per dollar) of the former in the two years (1970, 1973) for which Kravis, et al made detailed compari- sons (1970, 1973). The rates used were thus 8.8 pesos per dollar for 1972 and 16 pesos per doll.ar for 1978. The U.S. consumer price index, on an average annual basis, was 56% higher in 1978 than 1972 (see U.S. Department of Commerce, Bureau of the Census, Statistical Abstract of the United States, 1979 (100th Edition), Washington D.C., 1979). Thus the Colombian data for 1972, once converted to 1972 dollar equivalent,was inflated by 1.56 to reflect a common 1978 dollar base.  -8- estimates of total personal income derived from regional income accounts, for example, the 1972 survey accounts for 67% of the actual personal income C; 7/ in Bogota, while the 1978 survey covers 92% of the relevant total. - The poverty households, as defined, account for between one- fifth and one-quarter of all units in the surveys. As already noted, care should be taken in interpreting these figures, which include only reported labor and non-labor income and which provide for no adjustmnnt by household size. Turning to other indicators, one finds that between one-quarter and one-fifth of the households in the surveys were headed by young individuals who were 30 years of age or less. Another quarter to three- tenths was concentrated in the age grouping, 31-40, while the remaining half were over 40 years of age. These households were headed by males in the overwhelming majority of cases, though the percentage of female- headed households,appears on the rise for Bogota, increasing from below to above one-fifth, and is above one-quarter in Cali. There is a tendency for household size to decrease over time. The data suggest fairly dramatic changes in Bogota, especially for small 7/ See A. Pachon, Urban Structure ...., op. cit., Chapter 2, and R. Mohan, The Determinants of Labor Earnings in Developing Metropoli: Estimates from Bogota and Cali, Colombia, Washington D.C., Urban and Regional Report No. 80-14, The World Bank, November 1980, p. 119.  -9- (1-3 persons) and large households (7 or more persons). Large families fell from one-third to one-fifth of the total over the period, while small ones increased their share in mirror-image fashion. Cali, though a smaller city, parallels Bogota in 1978 in the distribution of household size. In all 3 cases, the bulk of households, with 45% to 50% of the total, include 4 to 6 members. These trends could have two components: declining birth rates and reduction in the number of relatives and friends sharing meals with a nuclear family, thus qualifying as household members in these surveys. / Bogota and Cali are cities with a large number of migrants, defined as non-natives by birth. Fully three-quarters of the household heads are non-native in these cities, and approximately 12-13% are relatively young household heads who arrived in Bogota over the 5 years preceding the survey year. This heavy representation of non-natives among household heads is not matched to the same extent for the po-1lation at large. Apparently migrants arrive in these cities at an early enough age so that substantial numbers of their off-spring are native born. Thus, while three-quarters of the households heads are migrants, somewhat less than one-half of the population of Bogota is non-native; in Cali migrants form a bare majority of the population. 8/ One study of total fertility rates (children ever born per female) in urban Colombia suggests that between 1973 and 1978 the rate fell by 19% from 3.6 to 2.9 children. Seee Nancy Birdsall, "Colombia: Declining Fertility and Human Welfare," Washington, Population and Human Resources Division, Development Economics Department, World Bank-, 1979, p. 15. Similar conclusions for Colombia are reached in S. Quick, Coutitry Demographic Profiles: Colombia, Washington D.C., Bureau of the Census, U.S. Department of Commerce, October 1979, Table A-8. Note should again be taken of the discrepancies uncovered earlier between Phase II and the 1973 Census.  - 10 - Nor is the heavy influx of migrants a source of "marginalization" for vast sections of these cities' populations. Native households and non-native households have roughly simil'ar income profiles, and it is only among the relatively young households of the "recent arrivals" -/that the wreight of poverty is noticeably higher than among the native households. 10/ ' Since age and income are linked by such factors as education and experience, - this disadvantage among recent migrant households is apparently temporary, for it hardly looms in the comparison of household income between units headed by natives and those headed by all non-natives, (I-1). Thus among 1972 Bogotanos native and migrant households (the latter of which include a recent migrant component equal to approximately 26%), poverty units equal just under one-fifth amd one-quarter of their respective group; for recent migrants the poverty households account for over 30% of the total. In the intermediate income categories, all three types of households are represented by roughly one-third of their total number. The native households maintain an advantage over migrant households in the above-median income category (48% to 41%), with recent migrants having the smallest proportion, at 36%. 9/ Roughly 40% of those household heads who arrived in the previous 5 years are under 30 years of age, a rate twice that for all households. 10/ Data from all three surveys suggest that the percentage of households earning less than the poverty level standard falls dramatically for households headed by individuals past the age of 30. This prevails until the household head reaches the age of retirement, when the proportion of households in poverty rises to levels approaching or surpassing that of the youngest households. See also R. Tiohan, The People of Bogota: Who They Are, What They-Earn, Where They Live, Washington, D.C., World Bank Staff Working Paper No. 390, May 1980, pp. 82-91, and especially Table 6.12.  Table I-1 Monthly Income Comparisons Between Households with Native, Migratit, and Recent Migrant Heads A. Bogota, 1972 Household Head tatus Migrant, Arrived Native Migrant 1967-1972 Income Category 1000 pesos or less 18% 25% 31% 1001 - 2000 pesos 33% 34% 34% Over 2000 pesos 48% 41% 36% B. Bogota, 1978 Household Head Status Native Migrant Migrant, Arrived Income Within Last Category Five Years 4000 pesos or less 20% 25% 27% 4001 - 8000 pesos 23% 27% 31% Over 8000 pesos 54% 48% 41% C. Cali, 1978 Household Head Status Native Migrant Migrant, Arrived Income Within Last Category Five Years 3500 pesos or less 18% 20% 29% 3501 - 7000 pesos 27% 29% 29% Over 7000 pesos 56% 50% 41%  - 12 - In 1978, the gap in the proportions of each Bogota household group falling into the below-poverty category is narrower (20%, 25%, 27%, respectively), and the equiproportional representation of each group in the intermediate category no longer prevails. But the key finding is, as before, that migrant households as a whole have income profiles not very dissimilar from that of native households, in spite of the "drag" imposed by recent arrivals.. Furthermore, as in 1972, substantial proportions of the.house- holds headed by recent migrants are found in the above-median income category. l/ The evidence from Cali is broadly consistent with the information already presented for Bogota. Among the poorest income categories (up to 3500 pesos), the diffevences between native and migrant households, taken as a whole, is negligible, slightly favoring the native population. Recent migrants, almost two-fifths of whose heads are under 30 years of age, are more heavily represented in this category. Roughly three-tenths of all such households are classified as poor as opposed to one-fifth of the other two groups. Paralleling the 1972 Bogota results, all three types of households are found in the intermediate income category in roughly equal proportions. In the better-off category the difference between natives and 11/ R. Mohan, in his study, The People of Bogota ...., op. cit., pp. 77-80, reports on data from the 1973 Census and the 1977 Household Survey which suggest in broad terms, that recent migrant workers earn less than long-term migrants and that migrants who have been in Bogota for many years earn more than native-born workers. The conclusions make no reference to household income.  - 13 - all migrants is again small, with both groups represented by approximately half of their households. Recent migrants trail, with only about two- fifths of their numbers in this category. Whatever one's prior expectations, these are cities where homeownership plays a major role. Roughly half of the households are homeowners and the remaining half rent or otherwise occupy premises as usufructuaries (Table I-2). These households opt for a relatively low density environment, showing a strong preference for single and two household buildings. These account for about three-quarters of all units in Bogota in 1972, and 90% of the units in 1978. In Cali these types of buildings account for 92% of the housing stock. The building data is reinforced by the information of dwelling types. In all three cases,' independent houses prevail over all other options, representing 62% of the household choices in 1972 Bogota and over 80% of the choices in Bogota and Cali in 1978 (Table 1-3) Some of the shifts between the 1972 and 1978 Bogota totals are the result of differences in the administration of the surveys and may not represent true shifts in the direction of indepen6ent houses and their corresponding building types. A comparison of the 1972 and 1978 survey instruments, for example, points to the removal of the building type "rooming house" (inquilinato) in the latter survey. One assumes that these types of units, barring their unlikely disappearance (they represented 11% of the building types occupied in 1972) were redefined so as to fit other categories. It is  -14- Table I-2 Present Tenure Choice of all Households A. Bogota, 1972 Count - Own 260,719 52 Rent 225,922 45 Other 16,160 5 Total 502,801 100.0 B. Bogota, 1978 Count % Own 330,543 46 Rent 3.39,587 48 Other 41,862 6 Total 711,992 100.0 C. Cali, 1978 Count % Own 110,105 51 Rent 94,171 43 Other 13,792 6 Total 218,068 100.0  Table I-3 A. BOGOTA, 1972 type of Dwelling Compared with Type of Building \Building D Single Two Multi-- Rooming Family Family Family House Total Type\ Independent House 84 17 4 1 100 %ROW 100 32 14 7 62 %COL Apartment or Floor 0 58 37 5 100 %ROW O 43 41 8 17 %COL Ln Room (s) 0 26 31 43 100 %ROW 0 25 44 85 21 %COL Total 52 22 15 11 100 mRow 100 100 100 100 100 %COL  Table 1-3 B. BOGOTA, 1978 Type of Dwelling Compared'-with Type of Building Building Type Apart- Detached Attached Bi-Family House Under w i> ment House House House Casalote Construction Tugurio Total DwelMilg'-, Type, Apartment 73 19 2 6 0 0 0 lOO %ROW 95 4 2 32 0 0 0 13 ZCOL hlouse 0 67 18 2 11 0 loo XROW 3 93 96 65 79 84 12 81 XCOL Room(s) 6 39 7 0 40 5 3 loo R0W 2 3 2 0 15 8 13 4 %COL House of Scrap 0 . 0 0 2 42 12 44 100 ZRDW O 0 0 2 6 8 68 2 ZCOL Otlher 0 0 0 0 0 0 100 100 %ROW 0 0 0 0 0 8 0 ZCOL Total 10 58 15 2 11 2 1 lOO ZROW 100 100 .100 100 100 100 100 1OO ZCOL 7~~~~~ I,.i~ ...'. . -^,,vo,rtS.3'f z 5j^'h''.S -S ti4 f>'5fESe';t XS>h--i4t;;-ABS<>