Residential location decisions of multiple worker households in Bogota, Colombia

Urban and. Regional Report No. 81 10 RESIDENTIAL LOCATION DECISIONS OF MULTIPLE WORKER HOUSEHOLDS IN BOGOTA, COLOMBIA By Jose Fernando Pineda July, 1981 This report was prepared under the auspices of the City Study Research Project (RPO 671-47) as City Study Project Paper No. 22. 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 Deing circulated to stimulate discussion and comment. Urban and Regional Economics Division Development Economics Department Development Policy Staff The World Bank Washington, D.C.  ABSTRACT In Bogota, as in many other cities of the world, the number of workers per household is increasing,as women and other household members joing the labor force in greater numbers. The number of workers per household in Bogota has increased,from 1.42 in 1972 to 1.70 in 1978, for example. Preliminary analyses in other countries has suggested that increasing the number of workers per household has virtually no effect on residential location on housing consumption of households, and this results is confirmed in Bogota when all workers other than the household head are treated as a single group of secondary workers. However, when secondary workers are stratified by their level of qualification (adult and educated vs. others) or by their workplace (central business district vs. other) significant and offsetting effects are observed that tend to cancel each other out when all secondary workers are pooled. In terms-of residential location effects, qualified secondary workers tend to move the residential location of the household toward the center of the city and to shorten the head's commuting distance while other secondary workers move the residential location of the house- hold away from the center of the city and lengthen the head's commuting distance. These-effects seem to reflect some joint optimisation across workers on the part of the household because qualified secondary workers are likely to work in the city center while other secondary workers tend to work at more peripheral worksites. The presesnce of secondary workers also has some effect on housing consumption: secondary workers decrease housing consumption by their presence but increase it through their added income. Taking these two effects together, secondary workers increase housing consumption slightly. The empirical results in the paper are based on a sample of renter households in Bogota collected in 1978.  PREFACE This paper forms part of a large program of research grouped under the rubric of the "City Study" of Bogota, Colombia, being conducted at the World Bank in collaboration with Corporacion Centro Regional de Poblacion. The goal of the City Study is to increase our understanding of the workings of five major urban sectors -- housing, transport, employment location, labor markets, and the public sector -- in order that the impact of policies and projects can be assessed more accurately. The author thanks Gregory K. Ingram and Alvaro Pachon for helpful comments and discussions. He also thanks Sonia Rodriguez and Elsa de Crane for valuable research assistance, and Leticia de Noriega and Maria Elena Edwards for manuscript preparation. Other City Study Papers dealing with housing and residential location include: 1. Rafael Stevenson, "Housing Programs and Policies in Bogota: An Historical/Descriptive Analysis," Washington, D.C., The World Bank, Urban and Regional Report No. 79-8, June 1978 (City Study Project Paper No. 3). 2. Alan Carroll, "Pirate Subdivisions and the Market for Residential Lots in Bogota," Washington, D.C., World Bank Staff Working Paper No. 435, October, 1980. 3. Gregory K. Ingram, "Housing Demand in the Developing Metropolis: Estimates from Bogota and Cali, Colombia" Washington, D.C., World Bank, Urban and Regional Report No. 81-11, June, 1981 (City Study Project Paper No. 20). L ..... .  I. INTRODUCTION The changing labor force composition of households in developing countries has stirred some interest on the impact of additional family workers on housing consumption and residential location. For Bogota in 1972 the average number of workers per family was 1.42. In 1978 this numwber had risen to a comparable figure of 1.70, equivalent to an annual increment of 3%. The distribution of households by the number of workers can be obserVed in Table 1. Households with more than one worker represented half of the total number of households interviewed in 1978, and 20% of the interviewed families had three or more family memberF who declared some form of gainful occupation. From 1972 to 1978 the proportion of households with jus't one worker had been"reduced"by 1'0%. One of the major obstacles to the measurement of the consequences of this phenomenon is the multiplicity of effects generated by the presence of additional workers in the household. This difficulty is further aggravated by the untenable nature of some of the traditional assumptions of residential location models a la Alonso, such as the monocentricity assumption. In moving from the monocentric to the multicentric city, space loses its one dimensional quality, and makes the analytical treatment of the subject somewhat more complex. In this brief essay, part of a larger work on residential location patterns, we present some estimates of the impact of additional workers on housing consumption and residential location in Bogota. The first part of the  -2- essay reviews briefly the residential location models and some previous findings. The second part provides a brief summary of Bogota and a description of the data base. The third section presents some of the results obtained, and the last part is a summary of conclusions. II. RESIDENTIAL LOCATION MODELS In the traditional residential location models households have two elements in their utility function: housing, Q, and other goods, Z. The price of housing is made up of a series of attributes like age of the dwelling unit, quality, amenities in the neighborhoodl and distance to the workzone. One of the major contributions of these models is their demonstration that a differential accessibility rent is reflected in the price of housing when the other housing attributes are controlled. As shown by Montesano, having an additive utility function and transport costs that are positively associated with distance to the workplace results in a housing price that declines with distance from the 1/ workzone-- In addition the household incurs some transport expenditures that decline with proximity to the employment center. In an equilibrium condition the marginal costs of transport with regard to distance should be equal to the absolute value of the marginal decrement in the expenditure on housing while holding its quantity and quality constant. For each quantity of housing consumed we can then find a minimum expenditure location as shown in figure 1. The actual amount of expenditure on housing and the remaining portion of income destined  -3- TABLE 1 BOGOTA 1972- 1978 % DISTRIBUTION OF HOUSEHOLDS ACCORDING TO NUMBER OF WORKERS PER HOUSEHOLD. 1/ YEAR Number of. 1972 1978 worker in the household 1 60 50 2 23 30 3 9 12 4 and more 8 8 TOTAL 100.0 100.0. 1/ Sources 1972: Phase II household survey 1978 : World Bank- DANE household survey (EH-21)  to other goods consumption at the i-Lnimum expenditure point indicates not only the housing expenditure but also the optimal location for that given amount of housing and the expenditures on Z. This is graphically shown in Figure 2. This very simple description allows us to understand the impact of having additional workers. Still holding the monocentricity assumption, an extra worker implies an incremental expend .:ure on transport and, ceteris paribus, a more central location. However, it also means an increase in the household income which might be represented by an increase in housing consumption, forcing a more peripheral location. But in addition it also represents a decrease in the leisure time available to the household. If there is some degree of complementary between housing and leisure time, an extra worker implies, ceteris paribus, a decrease in the quantity of housing services purchased. Again this effect would point to a more central location. Michelle J. White seeked to derive a bid-price function for two earner households where the husband works at the center and the wife at a 2/ peripheral location.- The housing price offer curve depends on the value of the wife's leisure time (relative to her husbands) in the household's utility function. Since the slopes of the bid-rent functions is known but not their intercepts, White attempted to show by trial and error, that households with two workers would seek to locate closer to the wife's work place. However, this depends OI the assumption that the bid rent function of two worker households is flatter between  THE RESIDENTIAL EQUILIBRIUM OF THE HOUSEHOLD - 5 - Total Residential \ | Costs (Q1 Total Residen Costs (QO ) (Housing + Trans. Cost' ,Transport Costs \Housing Costs Housing Costs (Q ) C.B.D. t t Distance to CBD. FIGURE 1 z II (Q,Z) I QO Q, (quantity of housing) I0 Distance to CBD t t Q Q, (quantity of housing) FIGURE 2  -6- the two job centers than the one worker households. This is only true if the diffzrence between the husband's and wife's wage rates are smaller than the commuting outlays, a fact that does not necessarily hold. So, two earner households might well stay closer to t:ie husband's job then one worker households. Partially based on White's reasoning, Madden developed a model of three simultaneous equations to analyze differences in housing consumption between one and two earner households.-/ The first equation explaina the separation between place of work and residence for a worker in a multi-worker household as a function of each worker's wages, housing prices, number of hours worked by each household worker, and the household's unearned income. Wages are in turn a function of workplace location and worker's attributes, and housing prices depend both on the workplace-residence separation, and on a vector of market attributes. The second equation is for housing size, in turn a function of the separation between work and residence, each worker's income, household unearned income, and household demographic characteristics. The third equation is for quality of housing, also a function of the variables included in the previous equation. These equations are estimated across several cities simultaneously using Survey Research Center Panel Data on income Dynamics. The prodedure followed is to estimate wages first as a function of distance of job from residence (instead of distance from job to city center) and then include  7 it to the estimate the value of the workplace to residence distance with all the additional variables. Number of rooms in the househol's housing unit represents housing size or quantity and price per room indicates quality. The three equations are estimate separately for men and women workers since sex is correlated with demographic labor force characteristics. Madden concludes that differences in housing consumption between one earner and two earner households are fully explained by their differences in money income and fertility. In contrast to White's conclusions, Madden finds that men from two earner households live closer to their jobs than men with either non-employed wives or employed wives and employed children. However none of these locational differences are statistically significant. In short, one and two earner households behave just the same when other relevant variables are controlled. The methodology we use is somewhat different. First we estimate housing prices by using the workzone stratification approach employed by Ingram.4/ Briefly described, the theoretical approach derives from the well known facts that workers commute down the rent gradient from their workplace and do so in the steepest direction and that workers facing the steepest gradient have the longest commute. Hence different workplaces imply different residential areas and these differences in location are derived From the trade offs between housing prices and travel costs. The different workplace opportunity sets available to workers provide the basis for workplace-based price variations when using cross-section data. The hedonic price equations are obtained by regressing che rent of a unit on its variables include d4.stance to  -8- the workplace, distance from workplace to the city's CBD, type of building structure and quality of the dwelling unit. Having defined an "standard" housing unit we estimate workzone specific housing prices by multiplying the estimated coefficients of each workzonie by the respective attribute of the "standard":ihousing unit. If we consider rent expenditures as price times quantity we obtain quantity of housing by dividing monthly rent by price. Once quantity is obtained we proceed to estimate the housing demand equation introducing income, household size, sex of the household head, price, and number of workers as independent variables. We also seek to explain workplace-residence separation as a function of quantity of housing consumed and household characteristics. The quality of housing is taken care of in the hedonic price index. Since all workers in a household live in the same residence, their impact on location is estimated by looking at the length of the journey to work of just one of them. We have chosen the cmmmuting distance of the household head as our dependent variable in the equation that explains workplace-residence separation. III. BOGOTA AND THE DATA BASE Bogota is a city of roughly 4.000.000 people located in the highlands of the Colombian Andes. During the early 1960's its population expanded at'annual geometric rates close to 7%. In the middle of the last decade its population growth rate has declined to less than 4% as a result  -9- of a decline in its crude birth rates (from 4.2% in 1960 to 2.6% in 1978) and a reduction in the inflow of migrants. From 1972 to 1978 its real per capita income grew at something close to 3% per year and auto owner- ship expanded from 14% to 18% of all households. The city has also become more decentralized. Estimates of the 5/ populatiorn density gradient are -0.177 and -0.112 for 1972 and 1978 respectively, The estimated dernsity at the center ,or che two yea7:s went down from 356.7 inhabitants per hectare to 262.6. Employment has also become more decentralized. In 1972 the Central Business District had 22% of the total number of jobs existing in the city and this percentage went down to 14% six years later. Anpther indication of the same phenomenon is the change in employmei%t density gradients. Using a negative expenential function, the values of the gradient are -0.240 and -0.204 and the intercepts 135 and 116 jobs per hectare for 1972 and 1978 respectively. Bogota is located in a plateau against the eastern range of mountains. The city resembles a semicircle with the mountains tracing the division (see Maps 1 and 2). The CBD is more or less at the center, and the wealthiest neighborhoods are located along a corridor that goes from the city center toward its northernmost portion. Low income hou3seholds are located almost everywhere. One could say it is the upper income groups that are- seggregated and not low income households, Growth has occured in the periphery of the city as can be observed on map 3, which shows population growth rates by distance from the canter for the 1924-1972 period. In  - 10 - MAP No. 1 BOGOTA 1910 aw,~ - ) 7/  IAP 2 - 21 BOGOTA - 1975 IV (Shadowed area corresponds to city limits in 1910) +%|{. e ..,,t tg,t ^jlSt................Figlure 2. Bogota and vicinity MCLCINTRO oi-N X i, al- BABR CHIC op.ado ; |%LdzAOT CENTE7 n O - ~CANDELAR4IA IvonIa 2OdeJUL'- -- | SN FE"NArp E, II ZEtEPO TO A.9 ^ CSK ,- / /~~AEROPUEIT(v flsi }t +- / t /2 /8 ELDORADO { >h> \ X/ . D \ jih-up ar. -3 .- 4 1 fMadr/id y c IA I t,. HG r-M LA A  BOGOTA 1964e-1973 -12 -A 30C-OTA INTERCENSAL GROSITH RATE OF POPULATION BY RING 1.29r 0\5\7&9 3 .43 ,' <X 5.65 2. . - /,  - 13 - short Bogota shows a marked change in its spatial structure and all the signs of the emergence of a sub-urbanization process. However, given the growth of the city administrative boundaries most of the population living and working in the city is covered by only one local administration. This differentiates the city from large urban agglomerations in North- America where there is much local government fragmentation and different utility companies within the same metropolitan area. Our data comes from two households surveys. The first one was carried out in 1972 by a United Nations study on transportation. It includes roughly 4.000 households who were living within the city administrative boundaries. The survey contains information on transportation, housing, labor force, and employment location. In 1978 the World Bank City Study project sponsored a household survey to update the 1972 information. This survey covered 3.056 households and represented some marked improvement on the quality of the information obtained, in particular with regard to income estimates. Every worker in the household was interviewed about his or her inco:,e from all sources including fringe benefits, something that was not included in the 1972 household survey. The city was divided into 38 zones, called communas by the National Statistical Agency. Expansion factors were calculated at the comuna level (see Map 4 for the comunas division). In the file created to carry out this analysis we excluded households where some of the income information was missing. In addition, households with no workzone data or with workers commuting outside tf the  MAP No. 4 BOGOTA - BCIJNDARIES OF COMUNAS CFAA AN f\ ,<1' K..,, \,J( ,- t -,- .'-f "f --- - *J '-\ ,-I (7, (v, )  - 15 - city boundaries were also excluded. We also eliminated workers with no fixed place of work: taxi-drivers, door to door, salesmen, etc. In the analysis, maids or domestic servants were not considered to be part of the household and are not included in the number of persons or number of workers in a household. Finally, households with monthly earnings inferior to 2.000 pesos (U.S.$50.00) were also excluded. IV. RESULTS Our analysis started with some very simple measures of residential location. For each household with two or more workers we derived three distance measures: D1, or distance between the residence and the place of work of the household head; D2, or distance between the place of work of the secondary worker and his or her residence' and D3, distance between the two workzones. In Tables 2, 3, and 4 we present some cross tabulations of the three distance measures. The entries of each table represent rounded distance in kilometers. Entry 0, for exmaple, means a distance between 0 kilometers andJl kilometers; entry 1, between 1 kilometer and 2 kilometers, etc. Looking at these tables we can observe the existance of five commuting patterns. One is what we call the short commute case where both the household head and the secondary worker labor close to home. The second one is called household head commutes, where the household head travels long distances while the secondary worker stays close to home. The third pattern is the complement of the second one: the head commutes less than one kilometer and the additional worker travels longer distances. The four.h case is where both household heads  I ti iN I INOD) (0001 Z,r 1213 013 9tL 0 9'L 915 0,9 r'L 13'9z ivio OKIC 1291 III. LOI 001 t' 6 ?01 LZI 013 L6 391N2l10' I 01l I L0 I 01l 1 Z' I 120 I 1'*0 I IIZO 120 I 90, I 01 I I 911 1 0'6 I V'Zl 1 0*& 6I 121' 1 6-9 I I'' 1 0I 1 r'o I 612 1 UJL I 12121 1 9'6 I 1'zl I 9'LZ 1 1312 I 179 1 61? I 131 I 9'L I 12121 I