Race and Home Ownership, 1900-1990

William J. Collins, Vanderbilt University and NBER, and

Robert A. Margo, Vanderbilt University and NBER

The historical evolution of racial differences in labor incomes has received considerable attention from economists. In contrast to labor incomes, far less attention has been paid to the historical evolution of racial differences in wealth. This scholarly neglect is unfortunate because racial differences in wealth were B and still are B much larger than racial differences in labor incomes and also because wealth per se is an important determinant of living standards, independent of earnings. This paper uses IPUMS data to examine the historical evolution of racial differences in one aspect of wealth B home ownership B from 1900 to 1990.

Information about home ownership is available in every micro sample of the census since 1900, with the exception of 1950. We extracted from each census micro sample observations of male household heads between 20 and 64 years of age who were not in school. Implicitly, we assume that if the house was owner-occupied, then it was the household head who owned the house, though it is possible that someone else in the household (e.g., a grandparent) was the true owner.

We first examined male home ownership rates by race from 1900 to 1990. The white rate was at all times considerably higher than the black rate, and for the most part, the black and white rates moved in the same direction from decade to decade. Remarkably, not until 1970 did the black home ownership rate reach the level of the white rate at the turn of the century (46 percent). The slight declines in both the black and white rates between 1920 and 1940 were followed by sharp rises from 1940 to 1960 (24.2 points for whites and 18.6 for blacks) and continuing increases until 1980 when the rates leveled off. Although the white level of ownership was always higher than the black level, the size of the gap varied over time. The gap jumped by 5.5 points between 1940 and 1960 and then collapsed from 1960 to 1980, falling 7.8 points.

We used linear probability models to estimate the relationship between various factors and the probability of living in an owner-occupied home. The regressions reveal significant changes in the nature of home ownership over the century's course. For example, the age profile steepened sharply between 1940 and 1960 as home ownership became the norm for white men in their 30s. Over the same period, the marriage indicator became strongly positive and significant. The coefficient for the farm status variable went from strongly positive at the turn of the century to essentially zero by 1960, including a big drop between 1920 and 1940, presumably as small, family-owned farms became increasingly rare. Finally, the substantial inter-regional differences which were apparent in 1910 had dissipated by 1990 as the high propensity for home-ownership in the West and Midwest converged on that of the Northeast (omitted category). Central city residents and inter-regional migrants (natives residing in a region different from their region of birth) were always less likely to own than non-migrants, and literate men and those in relatively higher paying occupations were always more likely to own their homes.

Controlling for other factors, race exerted a negative effect on home ownership, but the magnitude of the coefficient was uniformly smaller than the sample mean gaps. Importantly, controlling for factors other than race eliminated nearly all of the increase in the gap between 1940 and 1960. It is also apparent that the pure effect of race (the coefficient of the race dummy) began to decline significantly after 1960. Controlling for other factors the racial gap was about 15.5 percentage points in 1900, compared with 14.9 points in 1960. By 1980, however, the pure effect of race had fallen to 9.0 percentage points.

Next, we decomposed the change in the ownership gap over the 1940-60 and 1960-80 periods into three components: (1) changes in the characteristics of blacks and whites (2) with changes in how characteristics were correlated with home ownership (3) changes in the distribution of residuals. Over the 1940-60 period changes in the racial gap in characteristics accounted for a significant amount of the increase in the racial home ownership gap, particularly the movement of blacks out of the South and into the central cities. Approximately 31 percent (0.016/0.050) of the change in the gap can be accounted for by changes in the coefficients relating the various characteristics to the propensity for home ownership. Most importantly, the sizable increase in the coefficient on education tended to increase the size of the racial ownership gap because whites had significantly higher levels of education than blacks. Over the 1960-80 period, changes in characteristics and changes in coefficients account for only 39 percent of the overall decline in the home ownership gap. Most of the gap's decline was accounted for by changes in the residuals, and in particular, by the improvement of the average black man's position within the residual distribution.

The 1900, 1910, 1920, 1980, and 1990 census micro data samples report the mortgage status of owner-occupied homes. The importance of the mortgage in American home ownership grew enormously over the century for both blacks and whites and in all regions. In 1900, only 16 percent of all white male household heads held a mortgage and only 6 percent of blacks did, but by 1990, 57 percent of whites held a mortgage compared to 43 percent of blacks. Moreover, among home owners, nearly 80 percent of blacks and whites held mortgages in 1990 compared to about one-third of home owners in 1900. A linear probability regression of mortgage status in 1990 on the same set of characteristics used in the home ownership decompositions suggests that blacks were about 6.5 percentage points less likely than whites to hold a mortgage. An analogously specified regression for 1920 (using literacy rather than education level) returned similar results: a sample including all male household heads suggests that blacks were 5.1 percentage points less likely to hold mortgage. Because the overall incidence of mortgages was much greater in 1990 than in 1920, it follows that race per se had a larger negative impact in proportionate terms on the probability of holding a mortgage in 1920 than in 1990.

Since 1940, the census has reported a measure of house and property value of owner-occupied units. Among home owners, the West has consistently had a high ratio of black-to-white home values relative to the rest of the country. The South had by far the lowest value in 1940, though it caught up to the Midwest and Northeast by 1980. The nationwide ratio was very low in 1940 because blacks were relatively concentrated in the South where home values were quite low compared to the rest of the country. In the years between 1940 and 1980, however, the ratio of black-owned to white-owned home values increased substantially, from 35 percent to 62 percent.

For 1960 and 1980 we can observe several dimensions of housing quality in the IPUMS: the number of rooms, the number of bathrooms, the type of heating system, and the age of the building. The log value of housing (V) can be conceived as the product of a vector of house characteristics (Q) and the market prices of those characteristics (P) (Rosen 1974). That is, V = PQ, and differences in V across races and changes in the size of that racial gap can then be decomposed in the same manner after estimating hedonic regressions of V on Q for home owners. The regressions' explanatory variables (Q) include the house quality variables already mentioned, regional indicators, and center city and suburban indicators. The dependent variable is the log of house value, and the P coefficients are estimated by OLS on the 1960 and 1980 samples of home owners.

The relative improvement in black-owned house characteristics was the driving force behind the increase of black-owned house values relative to white values. In fact, the changes in the characteristics over-predict the magnitude of the change in relative values, implying that either changes in the "prices" or changes in the residual distribution offset, to some extent, the effect of changes in observed house characteristics. Using the "prices" estimated from the samples of white home owners, it appears that the widening of the residual distribution and a decline in blacks' average position in the standardized residual distribution both tended to offset the convergence of housing values.

Our findings bear on several important aspects of the evolution of black-white differences in economic status in the twentieth century. First, it is clear that black households have shared in an important aspect of the "American Dream" B ownership of a home. The proportion of black male household heads who were homeowners rose by 34.4 percentage points between 1900 and 1990. The corresponding increase among white male household heads was smaller B 28.1 percentage points B and, thus, the racial gap in ownership fell over the course of the 20th century. The decline in the racial gap in ownership is consistent with a variety of other economic and social indicators showing declining racial gaps, such as income, mortality, and schooling.

The racial gap in ownership did not, however, decline consistently across census years. The gap widened between 1940 and 1960, when increasing numbers of blacks left the rural South in search of a better life in the urban North. Thus, a consequence of the Great Migration B one that is largely unrecognized in the literature B was a widening racial gap in home ownership. The gap widened because rates of ownership, for both races, were higher in the South than, in particular, in the Northeast, and because residents of central cities were far less likely to be home owners.

The timing of changes in the racial gap also suggest an important role for institutional factors that dramatically altered mortgage markets and led the government to intervene in housing markets to combat racial discrimination. In 1911-1914, the average down payment for (new and existing) single-family houses in 22 cities was almost 68 percent of the purchase price, and that 46 percent of homes were acquired debt free. Mortgage contracts were typically of short duration (6 to 11 years) and those with mortgages had to refinance their debt with some frequency. The financial collapse in the 1930s led the government to redesign the mortgage lending industry in a way that made home ownership substantially more attractive and affordable than ever before.

The first (successful) intervention was the establishment of Home Owner's Loan Corporation (HOLC) in 1933. The chief mission of the HOLC was to reduce the rate of foreclosure which, in the early years of the Great Depression, had risen to astonishing levels. The HOLC also systematized the mortgage appraisal process, developing elaborate, uniform procedures to assess whether a particular loan should be granted. The rationalization began with the development of a series of so-called "Residential Security Maps", by which cities were divided up into neighborhoods to be rated according to their desirability. The best areas, typically occupied by high-income professional whites, were rated "A" and shaded green on the maps. They were followed by somewhat less desirable "B" neighborhoods (shaded blue); "C" neighborhoods, deemed to be "declining" and shaded yellow; and, finally, the lowest rank B fourth grade B already well past the deteriorating condition of "C" neighborhoods, and shaded red. A central aspect of the rating system was that it was based not only on the quality of the housing stock but also explicitly on the socioeconomic, ethnic, and B especially B racial characteristics of the neighborhood residents. Use of non-housing criteria in the appraisal process did not originate with the HOLC, but both in scope and in detail, the agency's revisions were unprecedented. Private financial institutions incorporated the new rating system in their own appraisals, thereby beginning the widespread institutionalization of the practice known as "red-lining."

Also established in the 1930s was the Federal Housing Administration (FHA). The FHA, along with a very similar program run by Veteran=s Administration (VA) after the passage of the GI Bill, did not issue mortgages, but rather insured them. FHA insurance reduced the risk to lenders, thereby enabling much lower down-payments (as little as 3 percent in the case of the FHA, and even zero in the case of the VA), interest rates, and closing costs. However,

While there is no doubt that the FHA and VA greatly facilitated the extension of home ownership to millions of American families, the agencies have also been accused of racial bias. For example, the FHA's 1939 Underwriting Manual openly recommended the use of racial restrictive covenants B that is, clauses in housing deeds prohibiting the occupancy of the structure in question by a black family. Although the Supreme Court ruled that such covenants were unconstitutional in 1948 (Shelley v. Kraemer) the FHA did not implement the ruling until 1950. The FHA and VA monitored and predicted the residential patterns of blacks, using the information to refuse insurance to neighborhoods on a wholesale basis B redlining. Not until 1966 did the agencies back off from their use of redlining by making funds increasingly available in low-rated neighborhoods, as well as lessening its credit-worthy requirements, thereby improving the odds that black applicants would obtain mortgages. The indirect reasons are more subtle. Even though the majority of FHA-insured loans were made for existing structures, and the majority of existing homes at the time were located within city boundaries, the FHA nonetheless greatly favored new construction in the suburbs while simultaneously discouraging loans for repairing and upgrading existing units, prior to the 1960s.

With the advent of the Civil Rights Movement and widespread urban unrest in the mid-to-late-1960s, public awareness of the federal government's culpability in fostering urban decay and racial segregation in housing was heightened. One important outcome was the eventual passage of the fair housing legislation in 1968 at the federal level, which outlawed racial discrimination in the purchase or rental of housing, and several related acts that were passed subsequently. In the 1980s and early 1990s various studies, the results of several of which were widely-reported and commented on in the press, asserted that the likelihood that minority applicants would be denied a mortgage exceeded the likelihood among otherwise identical white applicants, and that prospective black home owners, even those with relatively high incomes, continued to encounter discriminatory treatment from real estate agents. The Federal Reserve clamped down, in one case refusing to allow the merger of two banks because one of the banks (Shawmut, in Boston) was accused of discriminating against black mortgage applicants.

Although the data we have examined cannot be used directly to assess the effects of FHA-VA policies on black accumulation of housing equity, several pieces of evidence suggest that the indictment of these policies may be somewhat overblown. The black home ownership rate did not begin to rise until after 1940 B that is, after the establishment of the FHA. Although the raw racial gap in home ownership rose between 1940 and 1960, the regressions demonstrate that the racial gap was stable once we controlled for factors affecting the probability of home ownership. Chief among these was the migration of blacks from the rural South, were home ownership rates were relatively high, to the inner cities of the urban South and urban Northeast, where home ownership rates were relatively low. But the negative impact of central city residence cannot be attributed to FHA-VA policies per se because a similar negative impact existed prior to 1940 B that is well before the federal government began to intervene in housing markets. Blacks did fall behind a bit in the distribution of standardized residuals, but the magnitude of this slip (1 percentage point) is rather small. Moreover, the "ghettoization" of blacks in the urban Midwest and Northeast tended to increase the aggregate black-to-white ratio of housing values, because this ratio was higher in the Midwest and Northeast to begin with than in the South. And, whatever the role of redlining, the census data on mortgages clearly demonstrate that the percentage of black homeowners who held mortgages was vastly higher after 1940 than before; that is, even if black households lacked access to FHA-insured loans, they must have had access to other types of mortgage finance. An alternative, complementary explanation is that by accelerating the pace of suburbanization of white households, the FHA enabled larger numbers of black middle-class households than ever before to become home owners via a "filtering" process. In sum, while FHA-VA policies coupled with racial discrimination in access to suburban housing no doubt lowered the pace at which black households left the inner city, the policies did not per se prevent blacks from becoming home owners and accumulating some housing equity.

Our evidence on the efficacy of fair housing laws is mixed. The increase in black home ownership rates between 1960 and 1980 exceeded the pace that would have been predicted by the observed changes in household characteristics, consistent with the view that fair housing legislation has had a positive effect on the ability of black households to become home owners. Effective fair housing laws should also have improved the position of black home owners in the value decomposition's standardized residual distribution by allowing movement to neighborhoods which were "better" along unobserved dimensions. However, we did not find evidence of such an improvement.

The analysis in this paper could be extended in two directions. First, we have ignored local housing market characteristics in our analysis (such as the relative price of rental and owner-occupied housing, and levels of segregation), yet there is evidence, at least for the post-1970 period, that such characteristics may have influenced the likelihood of home ownership for blacks. Second, previous studies have noted that family structure per se has a significant impact on home ownership; in particular, female-headed households have much lower rates of home ownership than other types of families. Because racial differences in the incidence of female headship are both large and growing, including such households in the analysis may significantly affect both the level and trend in the racial gap in home ownership and housing values.