Printer Friendly

Racial differences in homeownership and housing wealth, 1970-1986.


Black couples are found to own a disproportionately low share of aggregate housing

wealth in the United States because they are less likely than whites to be homeowners

and because black-owned houses have lower market values than white-owned houses.

Probability of ownership and house value equations (corrected for selectivity bias) are

estimated with national data for 1970, 1980, and 1986. Trends in racial differences

in homeownership and house value are identified and reasons for their existence are



The wage gap between blacks and whites has been the subject of many labor market studies during the last two decades. These studies spawned similar analyses of racial differences in other markets such as housing. Concern over the black-white differential in economic well-being continues today, with recent attention focusing on the wealth gap between blacks and whites. Andrew Brimmer [1988] points out that blacks owned only 3.0 percent of accumulated wealth in the United States in 1984, even though blacks received 7.6 percent of total money income that year and they comprised 11 percent of all households.(1) The equity accumulated in homes constituted the most important form of wealth held by blacks, but even so the black share of home equity was only 4.4 percent of the total.

The value of black-owned housing will be disproportionately low if (a) blacks are not as likely as whites to own homes and (b) blacks' houses are relatively less valuable than whites' houses. In this paper we investigate both of these possibilities with national data on black and white households covering the period 1970-86.(2) Existing research has concentrated primarily on racial differences in the probability of homeownership, with an eye toward determining the extent of discrimination.(3) At least two prominent Congressmen (Joseph Kennedy of Massachusetts and Henry Gonzalez of Texas) have charged that government regulators routinely allow discrimination in home mortgage lending. Because most studies have relied on data from the 1960s and early 1970s, often covering only certain metropolitan areas, no reliable estimates of the national trend in black homeownership and housing market discrimination during more recent years exist. The question of how much housing blacks consume relative to whites, given the decision to own, has received only scant attention in the literature, yet both decisions influence wealth distribution. Our study compares the racial differentials in these two dimensions of the housing choice--ownership and value. A basic hypothesis of the paper is that house value differentials by race will be larger than homeownership differentials by race, since the factors that determine house prices are more numerous and probably more susceptible to discriminatory influence than the conditions that influence the decision to own rather than rent.


This section describes a model of the homeownership decision that can be applied to available data. The model is similar to those of previous studies in that it relates the probability of owning a home to several explanatory variables reflecting economic, demographic, and locational characteristics of the household. It differs in an important respect, however, in that it is estimated on a sample that contains only husband-and-wife households.(4) Most earlier studies that have used samples drawn from the entire population have controlled for different household types (e.g., single, female-headed, widowed) by means of dummy variables. This procedure generally constrains the impact of race on the housing choice to be equal for all household types. In addition, McDonald [1974] has pointed out that the current values of variables such as income and labor market activity are more apt to misrepresent their permanent levels for single-person or female-headed households. Consequently, housing market differentials attributed to race may actually stem from racially-biased phenomena such as high labor turnover or marital dissolution. Restricting the analysis to married couples provides more homogeneous samples of blacks and whites and thus should minimize bias in the "residual" estimates of racial discrimination in homeownership.

The dependent variable, homeownership, is measured by a dichotomous variable (1 = own, 0 = rent). The explanatory variables include dummy variables indicating the race of the household head (BLACK), veteran status (VETERAN) and age (using AGE20, AGE30, AGE50, AGE60, and AGE70 to represent ten-year intervals), and whether the household head was unemployed (UNEMP) or self-employed (SELFEMP) at the time of the survey in 1986. Binary variables are also used to control for the location of the household in a southern(5) state (SOUTH) and in the central city of a metropolitan area (CCITY). The income and wealth characteristics of the household are measured by an estimate of "permanent" family income in 1985 (PERMINC); transitory income (TRANINC); the capitalized value of net income from interest, dividends, royalties, and estates (ASSETS); and the fraction of income received from public assistance or welfare (WELFARE).(6) The number of persons in the household (PERSONS) completes the list of independent variables.

Virtually every previous analysis of the homeownership decision has found that blacks are less likely than whites to own, so we expect a negative coefficient for BLACK in the homeownership equation.(7) The probability of ownership is predicted to increase with income and wealth, perhaps nonlinearly so with age. Most studies have estimated a positive relationship between homeownership and family size, although in the Kain and Quigley [1972] equations the natural log of this variable was negatively related to ownership. On the basis of the Roistacher and Goodman [1976] and Silberman, Yochum, and Ihlanfeldt [1982] studies that have included location variables like ours, we anticipate a positive coefficient for SOUTH and a negative one for CCITY. Assuming that an occurrence of unemployment or the receipt of public assistance unfavorably affects a household's credit rating, the variables UNEMP and WELFARE may be expected to carry negative signs. In contrast, veterans may receive preferential treatment when securing mortgages, and therefore VETERAN is likely to be positively related to homeownership. Previous studies provide no guidance in predicting the effect of self-employment status on homeownership; Ladenson [1978] is the only study we are aware of which controlled for this variable and his results were inconclusive.


The homeownership equation was estimated with data from the 1986 Current Population Survey on black and white husband-and-wife households in which the household head was at least twenty years old. Because the dependent variable is dichotomous, the logistic regression procedure was utilized. The results are reported in equation (1) of Table I. All of the explanatory variables are highly significant, and those whose influence on the probability of homeownership could be confidently predicted on the basis of earlier studies generally exhibit the correct signs. One surprising result is the negative (but very small) influence of wealth on the probability of homeownership. However, this finding may not indicate a behavioral response as much as it might be a result of measurement error and a high correlation between ASSETS, income, and lifecycle variables in the model.(8)

The actual proportions of homeowners among black and white households in our sample are .632 and .785, respectively, which yields a black/white homeownership ratio of .805 for 1986. In the population as a whole, blacks are only about 64 percent as likely as whites to own a home,(9) which indicates that the racial gap in homeownership is reduced considerably when more homogeneous groups of blacks and whites (e.g., husband-wife families) are compared. Furthermore, the value of the race coefficient in equation (1) implies that once economic, demographic, and locational differences by race are accounted for, the probability of owning a home is only 6.3 percentage points lower for blacks than for whites.(10) Therefore, adjusting for the determinants of homeownership raises the estimated black/white ownership ratio to 0.92.

Several authors have questioned the conclusion initially accepted by Kain and Quigley [1972], and presumed in our equation (1), that the homeownership equation does not differ by race.(11) When we test for this possibility, we find that the black and white ownership equations reported in columns (2) and (3) are statistically different.(12) Some of the more obvious differences are seen in the signs and significance levels of the coefficients of the intercept, SOUTH, SELFEMP, ASSETS, and UNEMP. Also, whether it is measured in permanent or transitory terms, an increase in household income raises homeownership more for blacks than for whites. This last result would be surprising were it not consistent with previous studies and with the Birnbaum and Weston [1974] evidence that blacks have a greater tendency than whites to concentrate wealth in the form of home equity as income rises.

By estimating black and white homeownership equations, we can decompose the racial homeownership gap into two components: a portion resulting from racial differences in household characteristics, such as income, age, and so forth, and a residual component that stems from racial differences in the homeownership equation. The mean "propensities to own" predicted with the logistic homeownership models for whites ([bar] O.sub.w]) and blacks ([bar] O.sub.b]) can be expressed a (1) [Mathematical Expression Omitted] and (2) [Mathematical Expression Omitted] where f denotes the vector of logit coefficients and X [bar] is the vector of mean values of household characteristics. Accordingly, the racial difference in the propensity to own, [[bar] O.sub.w] - [[bar] O.sub.b] can be written as (3) [Mathematical Expression Omitted] The first difference above equals the sum of racial differences in mean values, each weighted by the corresponding logit coefficient for whites. If positive (negative), this measure indicates that overall the racial disparities in income and other household characteristics favor white (black) households. The second or residual difference in equation (3) equals the sum of racial differences in the logit coefficients of the homeownership equation, each weighted by the corresponding black mean value. A positive (negative) value indicates that racial differences in the homeownership equation parameters favor whites (blacks) in the tenure choice. These parameter differences may result from housing market discrimination, omitted variables, or misspecification of the tenure choice model.

It is equally correct to express the propensity-to-own differential as (4) [Mathematical Expression Omitted] in which case the first component uses black logit coefficients as weights and the residual component uses white household mean values. Consequently, there are two different estimates of the proportion of the white-black homeownership gap attributable to racial differences in household characteristics, and two separate estimates of the residual component of the homeownership differential. Because of space constraints, in this paper we report the weighted average value of the two estimates of each component, using the numbers of black and white households in the sample as weights.

Table II shows that the proportion of the white-black homeownership gap in 1986 explained by racial differences in household characteristics averaged 66.41 percent, whereas the residual portion averaged only 33.59 percent. In other words, race per se is only half as important in determining whether a household rents or owns as those characteristics that are merely coincident with race. Of all the potential differences between black and white households controlled for in the model, permanent-income and central-city-residence differentials are most important by far, each responsible for over 30 percent of the observed black-white homeownership gap. The remaining racial disparities are statistically significant but relatively unimportant as far as contributing to the homeownership gap between black and white couples.(13)

Table II also reveals that relative black homeownership is increased by racial differences in location, permanent income, and assets coefficients that are favorable to blacks. However, the black-white homeownership disparity is widened because (1) black couples in their twenties and thirties are much less likely than young white couples to own homes and (2) the intercept in the housing choice model is much more negative for blacks than for whites. As noted above, these "unexplained" differences by race in the homeownership equations might be the result of various forms of discrimination in housing and other markets. Kain and Quigley [1972] note that discrimination can take the form of supply restrictions on the tenure choice of blacks and on the kinds of housing available to blacks. McDonald [1974] and Ladenson [1978] speculate that blacks may have less access to mortgage credit than whites, and Yinger [1986] finds that blacks have inferior information about available housing. It is significant that our study attributes a smaller fraction of the housing differential between blacks and whites to discrimination (as measured by the residual) than the study by Silberman, Yochum, and Ihlanfeldt [1982]. Our use of black and white samples that are more homogeneous than those underlying previous studies would be expected to produce this result, as would a reduction in housing market discrimination over time. Furthermore, there are several reasons to believe that our discrimination estimate is biased upwards. First, the homeownership equation omits local housing market variables such as the availability of single-family housing and the cost of owning relative to renting.(14) Second, because the dependent variable measures homeownership rather than current home purchase, any lingering effects of previous homeownership barriers confronting blacks are included in the estimate of current (1986) housing discrimination.(15)


Previous studies have shown that the racial differential in homeownership has changed over time, but not consistently in the same direction. Ladenson [1978] found that the adjusted gap between black and white home purchase probabilities declined from 1969 to 1972 and then rose between 1973 and 1974. Silberman, Yochum, and Ihlanfeldt [1982] estimated that the residual (discrimination) portion of the purchase differential shrank by over 30 percent between 1974 and 1978. These studies are comparable in terms of their data source and the control variables included in the home purchase equation, so their results may be combined to suggest that the intensity of discrimination fluctuated during the 1969-78 period. In this section of the paper we investigate whether this conclusion applies when a longer time period is examined using substantially larger samples.

Husband-and-wife household records from the Bureau of the Census's 1-in-1000 Public Use Samples for 1970 and 1980 were used to estimate separate black and white homeownership equations analogous to those reported above for 1986. With these equations we computed the predicted black/white homeownership ratios reported in Table III. The proportion of all black husband-wife couples residing in their onw homes rose by 33 percent--from .475 to .632--between 1970 and 1986. The rise in homeownership among white households was considerably smaller, 14.9 percent. Consequently, the actual ratio of black to white homeownership rates increased by nearly 16 percent--from .695 to .805--over this period. A comparison of changes in the two predicted black/white ownership ratios indicates that the increase over time in relative black homeownership resulted more from a narrowing of racial differences in homeownership functions than from a convergence of black and white household characteristics. The black/white homeownership ratio that assumes equal ownership functions, but allows income, age, location, and other characteristics to vary by race, increased by 3.0 percent from 1970 to 1986. In contrast, the rise in the equal-characteristics homeownership ratio, which reflects racial differences in homeownership functions, was over twice as great (7.9 percent).

The years covered by our analyses of racial differences in homeownership, 1970 to 1986, were characterized by increased suburbanization in the United States. If an expanded supply of homes available to blacks resulted primarily from dwellings being deserted by whites fleeing from central cities, rather than from new construction in the inner city or from suburban neighborhoods opening up to blacks, then the progress in eliminating homeownership discrimination was more apparent than real. Furthermore, if whites were upgrading their houses by moving out to the suburbs, then the increase in black/white homeownership between 1970 and 1986 may have been accompanied by a decline in the relative value of blacks' homes.

In light of these possibilities, we reexamined the trends in homeownership using samples restricted to black and white husband-wife households residing outside central cities. Separate black and white homeownership equations (omitting the CCITY variable, of course) were estimated for 1970, 1980, and 1986 and the results used to prepare the right-hand-side estimates in Table III. Among households residing in rural and suburban areas, the actual ratio of black to white homeowner proportions increased by 10 percent between 1970 and 1986. Although this increase is less than the 16 percent gain recorded among all husband-wife households, it is still impressive, especially in light of the Carlson and Swartz [1988] finding of a relatively smaller earnings improvement among black males during the 1970s. Once again the movements in the predicted ownership ratios imply that the rise in relative black ownership resulted more from a reduction in discriminatory barriers to black homeownership than from a narrowing of racial differences in economic and demographic characteristics.

In summary, the homeownership analyses are consistent with the view that housing market discrimination which restricts the opportunities for blacks to own homes is relatively unimportant today, at least for black households whose structure matches that of most white homeowners (i.e., husband-and-wife households). However, the housing choice involves much more than the decision to rent or own; the location, age, size, and other characteristics of the residence must be selected. Housing market discrimination may limit the amount of mortgage credit available to blacks and restrict their choice of neighborhoods and quality of homes. As a result, blacks who own homes may have less valuable houses than comparable white homeowners. This issue is investigated in the remainder of the paper.


Racial differences in house values have not come under the same scrutiny as homeownership rates have, perhaps because of the more limited availability of house value data. Nonetheless, studies that have been based on national data report substantial differences in value between white and black-owned homes. For example, data from the 1967 Survey of Economic Opportunity, cited in Birnbaum and Weston [1974], reveal that black homeowners had net equity in their homes of $8,374 and white homeowners $12,113, which yields a black/white home equity ratio of only .691--substantially lower than even the unadjusted black/white homeownership ratio in recent years. Using data from the 1976 and 1978 National Longitudinal Surveys, Blau and Graham [1989] estimated that young black married couples have only 30 percent as much housing equity as their white counterparts, although much of the racial difference in home equity can be explained by socioeconomic differences between the races.

Several other studies have investigated the issue of racial differentials in house values, but only on a very limited geographical basis. Using a sample of Atlanta homeowners in the 1978 Annual Housing Survey, Ihlanfeldt and Martinez-Vazquez [1986] regressed owner-assessed house value on a dummy variable for blacks and various characteristics of the dwelling unit, such as number of rooms, location, and type of heating and cooling system. Blacks' houses were estimated to be 28 percent less valuable than comparable houses owned by whites, although the house value differential narrowed when racial differences in homeowner characteristics such as income were held constant. Using 1980 Census of Population data on house values in Los Angeles, Chicago, Houston, and New York, Boehm and Hofler [1987] found that homes owned by nonwhites ranged from $9,000 to $23,000 less in value on average than homes owned by whites.

In this section we specify a model of house value that will be estimated using a national sample of black and white households that own homes. Unlike previous multivariate studies of house value, our study controls for the possibility of selectivity bias in the house value equation by following Heckman's [1976] and Lee and Trost's [1978] two-step method. The first step involves estimating a model of the homeownership decision with probit analysis and using the vector of coefficient estimates ([Theta]) to construct a new variable, W, which equals f(Z [Theta])/F(Z [Theta]), where f and F are the density function and cumulative distribution function of the standard normal and Z is a matrix of household characteristics.(16) The second step involves estimating a house value equation with the new variable, W, included as an additional regressor. House value is assumed to be related to (1) socioeconomic characteristics of the household that are measured by the previously defined variables BLACK, PERMINC, TRANINC, ASSETS, and PERSONS, as well as the household head's age in years (AGE); (2) the geographical location of the house, which is proxied by dummy variables interacting region and metropolitan residence; and (3) physical characteristics of housing unit, such as the number of rooms (ROOMS), the number of bedrooms (BEDRMS), the number of baths (BATHS), the existence of central air conditioning (CENTAIR), and the number of years since the house was built (YRBUILT).


The 1986 Current Population Survey file does not contain information on house value, so the most recent data at our disposal for estimating the house value equation came from the 1980 Census Public Use Sample. To maintain consistency with our analysis of homeownership, we restricted the sample to black and white husband-and-wife households above age twenty who owned their homes (excluding mobile homes and trailers).(17) The dependent variable used in the model is the homeowner's estimate of how much the house would sell for on the current market.(18) House value should be expressed in real terms, but this is complicated by the unavailability of detailed cost-of-living estimates for most metropolitan areas covered by the 1980 census, least of all for rural or nonmetropolitan areas. The procedure followed in this paper was to deflate house value (and the income variables in the model) with a predicted index of relative living costs computed on the basis of a regression model that explains inter-area differences in annual household budget costs for medium income families.(19) Since this adjustment technique (and the inherent sample restrictions) may influence the results, Table IV reports both the nominal house value equations (the odd-numbered columns) and the real house value models (the even-numbered columns).

The nominal house value equation shown in column (1) is statistically significant and its explanatory power is relatively high considering that individual household data are utilized. The estimated variations in house value that are based on household and housing unit characteristics are consistent with previous studies, and the regional differences in house value are in accordance with well-known geographical differences in the cost of living. The coefficient of the sample selectivity variable is positive and highly significant, suggesting the presence of selectivity bias.(20) According to the BLACK coefficient, black-owned houses were $9,330 less valuable than white-owned houses in 1980, other things equal. Since the sample mean values of blacks' and whites' houses were $37,563 and $60,653, respectively, the variables in the model explain nearly $14,000 of the racial difference in nominal house values.

According to the estimates reported in column (2), black-owned houses were $11,352 less valuable in real terms than white-owned houses, after correction for sample selectivity and adjustment for racial differences in the explanatory variables. In percentage terms, roughly 40 percent of the racial differential in house value was unexplained by either the nominal or the real value equation. The adjustment for cost-of-living differences did not produce any major changes in most parameter estimates other than the region variable coefficients, whose magnitudes revealed that real house values in 1980 were substantially higher in the West, but fairly uniform across the rest of the country.

Using a single dummy variable for black households may be an appropriate technique for estimating racial differences in house value if blacks and whites have the same house value equation. However, this hypothesis can be rejected; therefore, the house value models were estimated separately for blacks and whites.(21) Comparing columns (3)-(6) in Table IV reveals important differences between the black and white equations: (1) the marginal effects of income on house value are more than twice as large for whites as for blacks, which is consistent with findings reported by Boehm and Hofler [1987]; (2) black houses tend to be less valuable than white houses having the same number of total rooms, bedrooms, and bathrooms; and (3) the selectivity variable coefficients are positive and statistically significant for whites alone.

When the racial difference in house value is decomposed (see Table V), we find that about 59 percent of the gap between white and black house values in 1980 can be explained on the basis of racial differences in the model's included variables, and this is true whether nominal or real values are utilized. The primary factors contributing to the relatively low black housing wealth are that black homeowners have relatively lower permanent income levels than white homeowners and their houses are physically smaller (in terms of numbers of rooms and baths), older, and less apt to be located in western metropolitan areas.(22)

The residual (unexplained) portion of the black-white housing value differential is over 40 percent, which is considerably larger than the unexplained gap in black and white homeownership. In a numerical sense, the residual house value gap results primarily because black house values do not increase with the homeowner's income and age and the house's size as much as white house values rise. Of course, we can only speculate about the sources of these phenomena.(23) They may stem from housing discrimination against blacks on the supply side. For example, financial institutions may restrict the amount of mortgage credit available to blacks and charge less favorable terms than whites confront, and Yinger [1986] suggests that realtors may steer blacks away from the more expensive white neighborhoods in order to avoid alienating potential white clients. Alternatively, black homeowners may indirectly select lower-valued houses by choosing racially segregated neighborhoods for cultural, ethnic, or purely personal reasons. If black neighborhoods have relatively higher crime rates, more smog, or other disamenities, then black-owned houses will be less valuable than physically similar white-owned houses because land values (a component of house value) will be lower in black areas. Over time the ratio of black to white housing wealth will diminish if land values rise less rapidly in black neighborhoods than white ones. However, if blacks' desire for racial clustering really amounts to an effort to avoid hostility toward blacks in integrated neighborhoods, then housing market discrimination may still exist but is occurring on the demand side, in the form of white homeowner preferences for white (or non-black) neighbors.


Changes in the relative value of black-owned houses during the 1970s were investigated by estimating the house value equations on samples of black and white homeowners obtained from the 1970 census Public Use Sample. The results were used to prepare Table VI, which shows that among all black and white husband-and-wife households, the actual nominal house value ratio declined by 4.2 percent between 1970 and 1980. This finding contrasts with the 11 percent increase in relative black homeownership recorded during the 1970s (see Table III). In real terms the actual black/white ratio declined by 8 percent (results not shown). A decrease in relative black housing wealth overall is consistent with a pattern of white flight to the suburbs, as noted above in section IV, but other explanations are possible. Therefore, to obtain additional understanding of the trend in relative house values, we again restricted the samples to households residing outside central cities and reestimated the house value equations for 1970 and 1980. Table VI shows that the actual black/white nominal house value ratio for homes located in rural and suburban areas rose by 7 percent during the 1970s. (For real house values the increase was 4 percent.)

The changes in the predicted black/white house value ratios reported in Table VI tell an interesting story. The equal-house-value-function ratio for homeowners outside central cities increased by over 15 percent during the 1970s. Consequently, racial differences in homeowners and housing unit characteristics narrowed considerably during the decade, by more than enough to offset a widening of disparities between black and white housing value equations. Among homeowners overall, blacks' incomes and house sizes did not increase enough (relative to whites') to raise the average value of blacks' homes compared to whites' homes. By implication, black homeowners residing in central cities must have experienced an erosion of their housing wealth relative to whites.(24) With our data we cannot determine whether this was caused by falling land prices in black neighborhoods, physical depreciation of black housing units, or other factors. What our data do reveal is that, although recent homeownership and housing wealth changes may have provided black households with more of the homeowner benefits (such as income tax savings and protection from inflation) long enjoyed by white couples, the gains have not been uniform among the black population.


This study finds that blacks in the United States own a disproportionately low share of aggregate housing wealth for two distinct reasons. First, blacks are less likely than whites to own their houses. Second, black-owned houses have lower market values than white-owned houses. As long as racial differences in income, marital status, and other factors that influence housing choices disfavor blacks, the rate of black homeownership and the value of blacks' homes can be expected to remain relatively low.

A disappearance of racial differences in economic and demographic characteristics will not automatically eliminate homeownership and house value disparities between blacks and whites if housing market discrimination exists. Our econometric analyses of the housing choices made by husband-and-wife households suggest that, if housing discrimination occurs, its major impact seems to be to limit the value of housing investment by blacks rather than to restrict homeownership among blacks. Homeownership among black households has increased over time so that today black couples have nearly achieved ownership parity with comparable white couples. Government monitoring of the credit practices of financial institutions may have contributed to the narrowing of the homeownership gap between blacks and whites.

However, as recently as 1980 the houses owned by black couples were substantially less valuable than those of similar white couples, and the house value differential overall had not shrunk significantly during the 1970s. The size and persistence of the white-black house value gap imply that it is difficult to eliminate the less overt and more subtle forms of discrimination that do not restrict homeownership as much as the sizes, locations, and types of houses available to blacks. If Simpson and Yinger [1985] are correct that white homeowners are more receptive to black neighbors with higher incomes, then this type of housing discrimination may be expected to dissipate over time as long as relative black wages and incomes continue to rise. In this case, new or intensified anti-discrimination policies are not warranted unless policy makers are concerned with the length of time it will take for black and white house values to converge.


Logistic Regression Estimates of the Homeownership Equations for 1986
Explanatory Total Sample Whites Blacks
 Variable (1) (2) (3)
Intercept -.15068 -.15690 -.75578
 (-1.61) (-1.59) (-2.60)
BLACK -.41097 -- --
SOUTH .15479 .14824 .40919
 (4.25) (3.84) (3.42)
CCITY -.92917 -.95314 -.73268
 (-25.98) (25.20) (-6.35)
PERSONS .09986 .09946 .08958
 (7.45) (6.91) (2.34)
VETERAN .27265 .27352 .36056
 (7.47) (7.14) (2.84)
SELFEMP .47258 .47689 .20630
 (8.18) (8.08) (.69)
PERMINC .03977 .03975 .04223
($1,000) (30.34) (28.68) (7.14)
TRANINC .03532 .03448 .04683
($1,000) (25.96) (24.96) (9.81)
ASSETS -.00110 -.00102 .01012
($1,000) (-2.83) (-2.59) (1.88)
UNEMP -.45420 -.48468 -.10351
 (-5.96) (-6.02) (-.44)
WELFARE -1.93022 -1.92220 -1.83056
 (-7.33) (-6.86) (-2.30)
AGE20 -1.41967 -1.39194 -1.76696
 (-25.79) (-24.12) (-8.84)
AGE30 -.43881 -.40547 -.78263
 (-9.46) (-8.28) (-5.23)
AGE50 .61907 .59229 .83298
 (10.09) (9.16) (4.28)
AGE60 1.17968 1.15926 1.29678
 (16.55) (15.39) (5.87)
AGE70 1.15724 1.13941 1.36870
 (14.97) (14.03) (5.35)
[[Chi].sub.2] 6,532(*) 5,869(*) 517(*)
n 31,252 29,238 2,014

Note: Asymptotic t-values is parenthesis.

(*)Indicates that the equation is statistically significant at the .01 level.


Decomposition of the White-Black Homeownership
 Differential in 1986
 Percentage of the Homeownership Differential
 Due to Racial Differences in
Explanatory Household Ownership
Variable Characteristics(a) Functions(a)
SOUTH -3.47 -11.55
 (3.81)(b) (2.08)
CCITY 30.60 -11.14
 (-24.00) (1.82)
PERSONS -4.33 3.86
 (6.62) (0.24)
VETERAN 2.37 -2.96
 (6.87) (0.66)
SELFEMP 3.02 1.26
 (7.61) (0.89)
PERMINC 34.04 -7.78
 (2.33) (2.06)
TRANINC .08 .04
 (27.52) (2.51)
ASSETS -.73 -7.30
 (23.86) (0.40)
UNEMP .95 -2.05
 (5.66) (1.53)
WELFARE 1.73 -.13
 (6.57) (0.11)
AGE20 -.27 5.20
 (23.15) (1.80)
AGE30 .01 10.59
 (8.09) (2.39)
AGE50 .40 -4.25
 (8.85) (1.17)
AGE60 .51 -2.05
 (14.77) (0.59)
AGE70 1.49 -2.20
 (13.47) (0.85)
TOTAL 66.41 33.59

(includes intercept)

(a)Percentages reported are weighted averages of the values obtained using white and black weights.

(b)Absolute value of t-ratios in parentheses. [Tabular Data III to VI Omitted]

(1)Racial differences in wealth and income are long-standing. In an earlier study using 1967 data, Birnbaum and Weston [1974] found that whites had 4.5 times as much wealth as blacks and 1.5 times as much income. (2)Households which do not identify their race as either "white" or "black" (Negro in the 1970 data) are excluded from the analyses. It is natural to focus on blacks since they are the largest racial minority in the United States. It would be interesting to see if the housing market experiences of other racial minorities (e.g., Japanese, Chinese, American Indian) are as varied in relation to whites as Carlson and Swartz [1988] find their labor market success to be, but this is a subject for future study. (3)The most frequently cited studies by economists in this area are those of Kain and Quigley [1972], McDonald [1974], Birnbaum and Weston [1974], Roistacher and Goodman [1976], Ladenson [1978], and Silberman, Yochum, and Ihlanfeldt [1982]. (4)The sample is further restricted to exclude households residing in mobile homes or trailers. It is unclear from previous studies whether this restriction is customary. Since the data sources we use to examine house value differences by race (see below) do not contain value estimates for mobile homes or trailers, we favor this restriction for consistency. Preliminary analysis indicated that the black/white homeownership ratio is slightly higher when occupants of mobile homes or trailers are included in the sample. (5)In this study the South consists of the states of Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia. (6)Cameron [1986], Goodman [1988], and others have suggested that permanent or long-run income is more appropriate for explaining housing market choices than current income. In the study permanent income (PERMINC) is predicted on the basis of a family income regression containing measures of the household members' age, education, veteran status, health status, and labor market activity. Transitory income (TRANINC) is the fitted residual. Separate income regressions were estimated for blacks and whites. Because PERMINC and TRANINC are estimated regressor, their reported t-ratios in the homeownership equation will be too high. (7)When the dependent variable measures home purchase rather than homeownership, studies such as Ladenson [1978] and Roistacher and Goodman [1976] have found that the coefficient of race is not statistically difference from zero. (8)The study by Birnbaum and Weston [1974] is noteworthy because it shows that the race coefficient is substantially reduced in magnitude and level of significance when a comprehensive measure of wealth is included in the homeownership model. However, the authors admit that their result may be a statistical artifact caused by the inclusion of home equity in the wealth variable. When the wealth variable was measured more narrowly (like our ASSETS measure) it apparently had little influence on homeownership. (9)The 68 percent figure refers to the black/white homeownership ratio for 1988, reported by U.S. Bureau of the Census [1989]. (10)The value of 6.3 percentage points is obtained when we compare the two predicted probabilities of homeownership that result from first evaluating equation (1) at sample means but ignoring the BLACK coefficient in the calculation, and then subtracting -- .41097 from the value obtained in the first step above. (11)For example, see BirnBaum and Weston [1974]. (12)The likelihood ratio test compared the pooled model to the fully interacted model. Minus two times the logarithm of the likelihood ratio equaled 93.44, which exceeds the critical value of the chi-square distribution with sixteen degree of freedom for [Alpha] = .005, which is 34.3. (13)The t-ratios reported in Table II are weighted averages of the t-ratios associated with the white-weights and the black-weights measure of the two components of the racial differential in the propensity to own. The t-ratios refer to the absolute size of the components of the homeownership differential, and not to the percentage of the total differential accounted for by each component. (14)Kain and Quigley [1972] and McDonald [1974] suggest that local market factors influence the tenure choice. We reestimated the homeownership equation with 1980 census data which allowed two additional variables to be included: (1) the percentage of year-round housing structures in an SMSA containing one unit and (2) the ratio of median gross rent in renter-occupied units to median owner cost in owner-occupied units. The results suggest that the intensity of homeownership discrimination is overstated if local housing market variables are omitted from the homeownership model. (15)Evidence of this result was obtained by reestimating the homeownership model on a sample of households contained in the 1970 census who had changed residence in the last five years and, presumably, had recently decided to purchase or rent a home. The black-white homeownership gap after adjusting for racial differences in the model's variables was substantially smaller for the "movers" sample. (16)We used essentially the same homeownership model described above except that age was measured by a continuous variable. Separate probit equations were estimated for blacks and whites. (17)Because of the aberrant housing markets in Alaska and Hawaii, households residing in these two states were also excluded from the analysis. (18)Other measures of house value employed in empirical research include sales price and the assessed value for property tax purposes. See Ihlanfeldt and Martinez-Vazquez [1986] for discussion and evidence relating to the differences in these measures. Research by Kain and Quigley [1975] and Kish and Lansing [1954] concluded that owner-occupant estimates of housing value are closely related to professional appraisers' estimates. (19)The budget estimates for thirty-eight metropolitan areas were reported in "Standards of Living for an Urban Family of Four Persons," U.S. Bureau of Labor Statistics, Supplement to Bulletin 1570-5. The cost-of-living regression included metropolitan population; median gross rent of renter-occupie residences; and dummy variables for northeast, north central, and western locations. While this model may be parsimonious, it explained 76 percent of the variation in relative living costs. Parameters from the model were used to predict living costs for all 272 metropolitan areas except Honolulu and Anchorage. Homeowners residing in non-metropolitan areas ere omitted from the analysis of real house values. (20)A correction for selectivity is employed when samples are not random but instead are based on individual choices. Individuals are presumed to choose between owning and renting housing in an optimizing manner, although analysis of this choice is complicated by the fact that homeownership provides both consumption and investment benefits. The only alternative to owning is renting, which represents a substitute for the consumption benefits of housing only. The positive sign of the selectivity variable indicates that if the sample of homeowners were random, the average house value would be lower, which no doubt reflects that fact that many couples not owning homes have lower incomes than homeowners. In other words, homeownership is a normal good. (21)A standard Chow test yielded a test statistic for the nominal house val ue model of F(19,28241) = 20.374, which exceeds the critical value of the .01 level. The test statistic for the real house value model, F(15,19605) = 22.56, is equally significant. (22)The importance of permanent income is understated when housing unit characteristics are included in the house value equation, since income affects the type of house purchased in terms of its number of rooms, cooling system, and so forth. Permanent income alone explains 46 percent of the house value gap when the model is reestimated without ROOMS, BEDRMS, BATHS, YRBUILT, and CENTAIR. (23)One possible reason not discussed below is that white homeowners overestimate actual house values more so than blacks. Ihlanfedlt and Martinez-Vazquez [1986] found that both blacks and whites overestimated the value of their homes, relative to the predicted value based on a sales price equation. However, whites overestimated by $1,324 more than blacks. This amount of upward bias in white house value seems far too small to wholly explain the racial differences in house value equation parameters. (24)Among households residing in central cities, the black/white actual (unadjusted) house value ratio declined by 21 percent between 1970 and 1980.


Birnbaum, Howard and Rafael Weston. "Home Ownership an the Wealth Position of Black and White Americans." Review of Income and Wealth, March 1974, 103-18. Blau, Francine D. and John W. Graham. "Black/white Differences in Wealth and Asset Composition." National Bureau of Economic Research Working Paper No. 2898, 1989. Boehm, Thomas P. and Richard A. Hofler." A Frontier Approach to Measuring the Effect of Market Discrimination: A Housing Illustration." Southern Economic Journal, October 1987, 301-15. Brimmer, Andrew F. "Income, Wealth, and Investment Behavior in the Black Community." American Economic Review Papers and Proceedings, May 1988, 151-5. Cameron, Trudy Ann. "Permanent and Transitory Income in Models of Housing Demand." Journal of Urban Economics, September 1986, 205-10. Carlson, Leonard A. and Caroline Swartz. "The Earnings of Women and Ethnic Minorities, 1959-1979." Industrial and Labor Relations Review, July 1988, 530-46. Goodman, Allen C. "An Econometric Model of Housing Price, Permanent Income, Tenure Choice, and Housing Demand." Journal of Urban Economics, May 1988, 327-53. Heckman, James J. "The Common Structure of Statistical Models of Truncation, Sample Selection, and Limite Dependent Variables and a Simple Estimator for Such Models." Annals of Economic and Social Measurement, 5(4), 1976, 475-92. Ihlanfeldt, Keith R. and Jorge Martinez-Vazquez. "Alternative Value Estimates of Owner-Occupied Housing: Evidence on Sample Selection Bias and Systematic Errors." Journal of Urban Economics, November 1986, 356-69. Kain, John F. and John M. Quigley. "Housing Market Discrimination, Homeownership, and Savings Behavior." American Economic Review, June 1972, 263-77. __. Housing Markets and Racial Discrimination: A Microeconomic Analysis. New York: National Bureau of Economic Research, 1975. Kish, Leslie and John B. Lansing. "Response Errors in Estimating the Value of Homes." Journal of the American Statistical Association, September 1954, 520-38. Ladenson, Mark L. "Race and Sex Discrimination in Housing: The Evidence from Probabilities of Homeownership." Southern Economic Journal, October 1978, 559-75. Lee, Lung-Fei and Robeert P. Trost. "Estimation of Some Limited Dependent Variable Models with Application to Housing Demand." Journal of Econometrics, December 1978, 357-82. McDonald, John F. "Housing Market Discrimination, Homeownership, and Savings Behavior: Comment." American Economic Review, March 1974, 225-31. Roistacher, Elizabeth A. and John L. Goodman. "Race and Homeownership: Is Discrimination Disappearing?" Economic Inquiry, March 1976, 59-69. Silberman, Jonathan, Gilbert Yochum, and Keith Ihlanfeldt. "Racial Differentials in Home Purchase: The Evidence from Newly-Formed Househols." Economic Inquiry, July 1982, 443-57. Simpson, George E. and J. Milton Yinger. Racial and Cultural Minorities: An Analysis of Prejudice and Discrimination, 5th ed. New York: Plenum Press, 1985. U.S. Bureau of the Census. Census of Population-1970, 1-in-1000 Public Use Sample, processed computer tape. U.S. Bureau of the Census. Census of Population-1980, 1-in-1000 Public Use Sample, processed computer tape. U.S. Bureau of the Census. Current Population Survey: Annual Demographic File, 1986, processed computer tape. U.S. Bureau of the Census. Statistical Abstract of the United States: 1989, Washington, D.C.: Government Printing Office, 1989. Yinger, John. "Measuring Racial Discrimination with Fair Housing Audits: Caught in the Act." American Economic Review, December 1986, 881-93.

JAMES E. LONG and STEVEN B. CAUDILL, Professor and Associate of Economics, Auburn University. We wish to thank Keith Ihlanfeldt, the editor, and two anonymous referees for helpful comments on a previous version of this article.
COPYRIGHT 1992 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1992 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Long, James E.; Caudill, Steven B.
Publication:Economic Inquiry
Date:Jan 1, 1992
Previous Article:The effect of gender-sorting on propensity to coauthor: implications for academic promotion.
Next Article:Structural unemployment in the United States: the effects of interindustry and interregional dispersion.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters