Race and Home Ownership from the End of the Civil War to the Present Research

Race and Home Ownership from the End of the Civil War to the Present Abstract: We present race‐specific estimates of home ownership in the United States from 1870 to 2007. For the sample of households considered in this paper, the racial gap in ownership declined by 26 percentage points. Remarkably, 25 points of the long run decline in the gap occurred before World War One. We present estimates of home ownership for African‐American and white households from 1870 to 2007. These estimates, which pertain to a core sample of households headed by adult men (see below), update and extend our earlier paper’s analysis (Collins and Margo 2001) with figures for the pre‐ 1900 and post‐1990 periods, and for 1950. Long‐run patterns of racial differences and trends in home ownership provide a window on wealth accumulation and quality of life, on the economic nature and implications of discrimination, and on the influence of long‐term economic trends, specific economic events, and changes in government policy on housing market outcomes. Our goal in this short paper is to measure and succinctly explain long‐term changes in home ownership by race. Over the entire period African‐American households in the core sample increased their home ownership rate by 46 percentage points, whereas the rate for white households increased by 20 percentage points. Thus, in the long run, the racial gap declined by 26 percentage points. Remarkably, 25 of the 26 point long‐run narrowing in the racial gap occurred between 1870 and 1910, a novel finding that emerges from our new estimates for the pre‐1900 period. Since 1910, both white and black households have increased their rates of homeownership, but the long‐run growth in levels has been similar for both races, and therefore the racial gap measured in percentage points was, to a first approximation, constant over the twentieth century. The gap widened and narrowed appreciably over shorter periods, but the cumulative effects were not sufficient to result in substantial racial convergence after 1910. For example, from 1960 to 1980 the racial gap declined by 9 percentage points, but it widened by 6 points from 1940 to 1960, and so the net change from 1940 to 1980 was a narrowing of only 3 percentage points, which then was wiped out by a widening in the gap between 1980 and 2007 of 4 percentage points. Below, we document several factors that tended to narrow (or widen) the ownership gap at various historical junctures. Data and Basic Trends The federal census provides the necessary basis to construct long‐run time series of owner‐ occupancy rates for private households. We use data from the Integrated Public Use Micro‐data Series (IPUMS) and published census volumes to estimate rates of ownership for census years from 1870 to 2000 (Ruggles et al. 2008). For 2007, we rely on the American Community Survey (ACS), a large random sample that is designed to provide information similar to the decennial census. The 1890 census was the first in American history to inquire specifically about home ownership status. Each subsequent census included a direct question on home ownership, a tradition continued by 1 the ACS. Prior to 1890, the census asked questions about real estate ownership in 1850, 1860, and 1870. Although real estate ownership covers more than just home ownership, we believe it is possible to use these data to construct plausible estimates of owner‐occupancy for the pre‐1890 period. Our estimates of ownership begin in 1870, the first census year after the Civil War in which African‐ Americans were enumerated as free persons, and end in 2007, just before the onset of the recent housing and financial market crises.1 The census question on home ownership has remained essentially the same since it was first asked in 1890. The most important feature of the census question is that it pertains to the tenure status of the persons living in a housing unit at a specific street address or rural location.2 A family is “owner‐ occupying” the unit if someone in the family is the legal owner of the unit.3 If no one in the family is the legal owner, the family is considered to be renting.4 We aggregate household level outcomes into a rate per 100 households, or percent. To be included in the core sample used in this paper, the household head must be male, in the labor force, not currently enrolled in school, and between the ages of 25 and 64.5 We focus on this 1 It would be possible to use the 1850 and 1860 censuses to estimate owner‐occupancy using our procedures for white households and households headed by free blacks but not for slaves, who formed the bulk of the African‐ American population before the Civil War. 2 Prior to 1960 the census was taken by trained enumerators who visited households. The census takers went from dwelling to dwelling generally in order of street address or from one unit to the next in multi‐family housing. From 1960 to the present the census is taken by mail with follow‐ups by enumerators if forms are not returned. 3 It is occupancy that determines census ownership. A property that is mortgaged is not considered to be part‐ owned by the entity that extends the mortgage even if the entity happens to be another person (for example, if the mortgage was provided privately by the seller). More than one person in the family could be an owner (for example, the head’s spouse) but only one person can be designated as the household head. 4 Beginning in 1980, the Census Bureau adopted the term “householder” (rather than “head of household”) under which the person designated as the householder was supposed to be a family member who legally owned the dwelling unit or in whose name the dwelling unit was leased. Strictly speaking, therefore, our ownership estimates prior to 1980 pertain to households whose heads satisfy certain conditions (discussed below), whether or not the designated head actually owned or held the lease to the property. As a practical matter, we believe the distinction to be minor, but it should be kept in mind when interpreting the estimates. In 1890 the census recorded the identity of the person (or persons) in whose name the dwelling was owned (if owned). The published census volume on ownership in 1890 contains, for example, tables giving the number of male owners and male renters; a male owner might not be a household head, but if the household was renting the gender of the household head determined its classification. Unfortunately, the 1890 manuscripts were destroyed in a fire and so it is not possible to use them to determine the fraction of households living in owner occupied housing in which the owner was someone other than the household head. 5 The screening of the sample is accomplished using the IPUMS variables RELATE or RELATED (relationship to head of household), SCHOOL (school attendance during the census year), LABFORCE (labor force status), and age. There are inconsistencies across the census years in the meaning and construction of these variables (particularly, LABFORCE and to a lesser extent, RELATE) but we believe these inconsistencies have only a minor effect, if any, on our estimates. We use age 25 as the early cutoff age because this is the age for more recent decades by which 2 particular slice of the population because it has been intensively studied by scholars interested in the long‐run evolution of racial differences in income and educational attainment (see, for example, Smith 1984; Smith and Welch 1989; Donohue and Heckman 1991; Collins and Margo 2006). Although there are some differences in magnitudes, the fundamental time‐series patterns revealed by the core sample are not changed appreciably by broadening the sample to include all household heads (see Collins and Margo 2004; Collins and Margo, in progress; and Appendix Table 2). Figure 1 displays our estimates of owner‐occupancy by race. For 1900 to 1940, 1960 to 2000, and 2007, we derive the estimates directly from the IPUMS or ACS. For 1870, the estimates also derive from the IPUMS, but they are made indirectly by adjusting the data on real estate ownership. For 1890 and 1950, the estimates are based on adjustments made to published census data. The 1880 estimates are constructed in an indirect manner and should be viewed as less secure than for the other years. Details of the estimation and the data points underlying the figures in this paper are presented in the data appendix. In 1870, the rate of owner occupancy for African‐Americans in the sample was a scant 7.7 percent.6 In 2007, the most recent year in the series, the rate was 54 percent. Thus, over the approximately 140 years covered by our series, African‐American households represented in the core sample increased their rate of owner‐occupancy by 46.3 percentage points. The long‐run increase in African‐American owner‐occupancy, however, did not occur at a continuous pace. Broadly speaking, the data reveal two periods of rising owner‐occupancy for black households. The first period is from 1870 to 1910. During the first period, the ownership rate increased by 16 percentage points, from about 8 to 24 percentage points. The second is from 1940 to 1980, during which the black ownership rate increased by 37 percentage points. Interspersed between the periods of growth were periods of stasis or cyclical movement. There was a slight decline in black owner‐occupancy between 1910 and 1920, followed by an increase in the 1920s and a decline in the 1930s, leaving the 1940 rate below the 1910 rate. A second period of stasis is from 1980 to 2007. The black rate fluctuated during the second period, ending up in 2007 just below the 1980 level. most adult men have completed their schooling. The substantive patterns would not be affected if we used age 20 instead. 6 The 1870 census is known to have been deficient in coverage, particularly among black households in the South. Non‐enumerated households those that were more difficult for the census taker to find in the first place and these would be households without a fixed address. The odds that such households would be owner‐occupants are probably close to zero. In all likelihood, therefore, our estimate for black households overstates the true rate of owner‐occupancy. 3 While the long‐run dynamic for blacks was generally upward, homeownership for whites followed a U‐shaped pattern. The extent of decline during the downward portion of the U was fairly substantial, about 10 percentage points over the half century from 1870 to 1920. The rate remained constant in the 1920s before falling another five points in the 1930s. In 1940, the eve of American entry into World War II, the white rate of owner occupancy stood at 43.2 percent, 13.3 percentage points below the rate in 1870 (56.5 percent). Between 1940 and 1980 the white owner‐occupancy rate rose by about 34 percentage points. Considerably more of this increase occurred between 1940 and 1960 (about 26 percentage points) than between 1960 and 1980 (8 percentage points). From 1980 to 2007, the white rate in the core sample remained steady at approximately 77 percent. The racial gap in percentage points is also labeled in Figure 1 for select years (see also appendix table 1). From 1870 to 2007, the white ownership rate increased by about 20 percentage points and the black rate by 46 percentage points. Consequently, the racial gap fell from 49 to 23 percentage points, a sizable decline. The pace of racial convergence, however, was not constant over time. Convergence was much more rapid between 1870 and 1910 (about 25 percentage points) than after 1900. Approximately two‐thirds (62 percent) of the narrowing of the racial gap before 1910 can be attributed to the increase in the black ownership rate with the remainder due to the fall in the white rate. Between 1910 and 1940, the racial gap fell slightly, with all of the reduction occurring during the Great Depression. The decline in the racial gap occurred because the white rate fell more than the black rate in the 1930s. The narrowing of the gap during the 1930s was relatively modest measured against the long‐term trend, and the overall impression of the period between the two world wars is one of relative stasis. Both white and black households experienced rapidly rising owner‐occupancy rates from 1940 to 1960, but the increases were larger for white households, and so the racial gap widened to 28 percentage points. From 1960 to 1980, the rate of African‐American ownership increased by 17 percentage points, with the absolute gains split more or less equally between the 1960s and 1970s. The white rate also increased over this period but to a lesser extent. Consequently, the racial gap was at its lowest level, about 19 percentage points, in 1980. After four decades of impressive growth, home ownership rates stagnated after 1980. In 2007, the white ownership rate was about the same as in 1980, whereas the black ownership rate had declined by 4 percentage points. Table 1 reports regressions of the racial gap on a linear time trend to summarize the salient convergence patterns discussed above. In column 1 the racial gap is measured by the difference in percentage points (black – white). In the first row the regression is estimated over the entire period 4 (1870‐2007). In the other row, we allow the coefficient of the time trend to differ before and after 1910. When estimated over the entire period, the coefficient is positive and statistically significant. The time trend is measured so that a unit change corresponds to a decade; measuring the gap in percentage points, average (estimated) convergence is about 1.42 percentage points per decade. When the trend has a break point at 1910, convergence is very rapid before 1910 (nearly 7 percentage points per decade), but much more limited between 1910 and the present. Indeed, for the post‐1910 period, the coefficient on the trend variable in column 1 is not statistically significant. An alternative to measuring the racial gap in percentage points is to first transform the ownership rates by the logistic function, in which case the dependent variable in the regression is the logarithm of black‐to‐white odds ratio. As column 2 of table 1 shows, use of the logistic transformation does yield a statistically significant coefficient on the trend variable for the post‐1910 period but does not alter our principal finding: The majority of the long‐run racial convergence in home ownership took place before 1910. Explaining the Timing of Racial Convergence in Home Ownership According to the conventional economics of home ownership, households choose housing characteristics and housing tenure (owner‐occupancy versus renting) to maximize life‐cycle utility, subject to the household’s inter‐temporal budget constraint. The decision to owner‐occupy versus rent will vary with the type of housing demanded. Broadly speaking, single‐family detached housing units will be owner‐occupied more frequently than units in multi‐family housing (Glaeser and Shapiro 2002). The demand for single‐family detached housing is positively related to the household’s permanent income and, in most treatments of the issue, also influenced by state and federal tax policy and institutional features of mortgage markets. This simple framework suggests that narrowing racial gaps in household income and educational attainment—observable characteristics that are strongly correlated with permanent income—should be correlated with racial convergence in owner‐occupancy. At first glance, however, the framework does not seem very promising given the historical trends discussed above. While there is some evidence that racial differences in incomes and education narrowed between 1870 and 1910, the extent of convergence in owner‐occupancy over this period is far out of proportion; moreover, both racial groups appear to have increased their incomes over this period, which does not square with the apparent decline in white owner‐occupancy (Higgs 1982; Margo 1983; Collins and Margo 2004). Most economic 5 historians and labor economists believe that there has been substantial racial convergence in incomes and schooling from the early twentieth century to the present, but our data show, at best, limited racial convergence in owner‐occupancy since World War One. Movements over shorter periods in the twentieth century are also at odds with the framework. The black‐white income ratio is known to have risen significantly between 1940 and 1960, for example, yet our data show a widening of the racial ownership gap over these two decades. The conventional economics of home ownership was developed with post‐World War Two non‐ farm housing markets in mind, but this framework, we argue, is inadequate to fully understand the dynamics of long‐term racial change for two reasons. First, the conventional framework neglects the role played by the long‐term shift of households out of agriculture which, as we describe below, was racially distinct in its implications for home ownership. Second, institutional changes in housing markets, among other factors, dramatically increased the racial ownership gap at younger ages between 1940 and 1960. Home ownership, however, is part of a life‐cycle process, and as described below, when we focus on across‐cohort racial differences in owner‐occupancy at older ages, there is stronger evidence of persistent racial convergence in the long run, as well as clear links to income and schooling convergence. As a point of departure for discussion of the role of agriculture, Panels A and B of Table 2 show race‐specific owner‐occupancy by “farm home” status for 1870, 1890, and 1910.7 Panel A reveals an enormous farm‐nonfarm gap in owner‐occupancy among white households in the late nineteenth century. Over the 1870 to 1910 period, approximately 72 percent of white household heads residing in farm homes were owner‐occupants. This rate is comparable in magnitude to the aggregate white ownership in the late twentieth century when the proportion of households living in farm homes was negligible by comparison. 7 Farm home status is indicated by the nature of the dwelling—was it a “farm home” or not? Farm homes were those for which a companion farm schedule was issued. That is, the home was located on a farm that independently produced a quantity of crops or other agricultural products sufficient to be included in the agricultural census. When measured by the functional status of the home, the percent farm is lower than when measured by the occupation or industry of employment of the household head because some farm workers would have been living in non‐farm homes, as the census defined these to be. The difference for whites was small, as can be seen in table 2. In 1870 41 percent of homes occupied by white households were designated as farms, whereas 45 percent of household heads were employed in agriculture. The table refers to all household heads in order to make use of the published figures by farm home status in the 1890 census. The substantive results would not differ if, instead, we used just the core sample for 1870 and 1910. Farm home status is indicated by the nature of the dwelling—was it a “farm home” or not? Farm homes were those for which a companion farm schedule was issued. 6 Although the United States was one of a handful of countries to begin industrializing prior to the twentieth century, the American economy was dominated by agriculture well into the late nineteenth century. Relatively small, “family” farms were the norm, and for such households, owner‐operator status was central to life‐cycle wealth accumulation. For whites, owner‐operator status could be achieved through inheritance (or inter‐vivos transfer) or by moving up the so‐called “agricultural ladder,” possibly through the use of credit. Over the course of the nineteenth century, a well‐ functioning market for agricultural credit emerged, including farm mortgages, which facilitated high rates of farm ownership (Bogue 1955). Before the automobile era, it was generally too costly, in time and money, for farmers to live “in town” and commute to their farms. De facto, ownership of the farm meant owner‐occupancy of the farm home, which was part of the farm’s capital stock. High rates of farm ownership directly imply high rates of home ownership among nineteenth‐century white farm households. The magnitude of the farm‐nonfarm ownership gap among whites was so large in the late nineteenth century that the shift of whites out of agriculture resulted in a quantitatively significant decline in the aggregate home ownership rate. For instance, taking the sector‐specific (farm and nonfarm) ownership rates shown in Table 2 for 1870 and weighting by the shares of farm and nonfarm households in 1910 yields a predicted aggregate ownership rate of 48.6, compared to the actual 1870 rate of 56.4.8 Some portion of the farm‐nonfarm gap in white owner‐occupancy reflects the economics of urban areas at the time. Urban jobs, particularly in manufacturing, were concentrated in central business districts (CBD). Urban commuting costs were high, and workers preferred to live close to their place of employment. In dense urban neighborhoods, multi‐family housing economized on scarce land, but in such housing the majority of occupants were renters. Eventually, improvements in transportation—chiefly, but not exclusively, the automobile—led to a flattening of urban density gradients and to the concomitant development of neighborhoods on the periphery of the CBD, made up primarily of owner‐occupied single‐family dwellings. 8 To be sure, the shift of labor out of agriculture was not the only factor responsible for the downward trend in aggregate white ownership. Table 2 also shows that farm ownership rates were falling after 1870. Some of this decline, however, is really part of the same story because it reflects a lower rate of farm ownership in the South coupled with regional differences in the pace of economic development. The shift of labor out of agriculture was slower to take hold in the South than elsewhere. By 1910, a larger fraction of farm homes occupied by whites were located in the South than in 1870, thereby lowering the white farm ownership rate in the national average. In 1910, 40 percent of white farm homes occupied by our core sample were located in the South, compared with 28 percent in 1870. 7 However, it is unlikely that all of farm‐nonfarm ownership gap can be explained by intra‐ metropolitan commuting costs. Although the farm credit market was institutionally well‐developed by the late nineteenth century, the same cannot be said for the “home” (i.e., non‐farm) mortgage market. In some communities a local abundance of credit might make home ownership relatively easy to achieve whereas in other areas credit would be scarce and difficult to obtain. Virtually everywhere lenders required significant down‐payments, as much as 50 percent of the sale price. Term lengths were relatively short and the loans were non‐amortizing—so‐called “balloon” mortgages—that would need to be refinanced at the end of the loan unless the borrower had accumulated sufficient capital in the interim to purchase the property outright. Beginning in the late nineteenth century, a variety of institutional developments led to a gradual “thickening” of the home mortgage market, which left its imprint in the form of a secular narrowing farm‐nonfarm gap (Snowden 1997, 2003, 2006).9 On the cusp of the Great Depression, a majority of non‐farm white households still lived in rental housing, but their tenure choice, by and large, no longer reflected fundamental inefficiencies in the home mortgage market. For African‐American households the post‐bellum story is different in crucial ways, as illustrated by Panel B of Table 2. In 1870, 63 percent of black household heads in our sample claimed to be employed in agriculture, yet only 19 percent lived in a farm home, meaning that they were not sufficiently independent as farmers to warrant filling out a farm schedule in 1870. By 1890, however, 39 percent of African‐American households were resident in farm homes; the corresponding figure was slightly higher in 1910. The movement into farm homes was a step towards economic independence and, eventually, land ownership. The vast majority of blacks in the post‐bellum south could never count on inheriting land because either they or their fathers had been slaves. By necessity, the only way to become a farm owner was moving up the agricultural ladder, and most blacks started on the lowest rung as wage workers or in some type of tenancy contract with a white landlord. The climb to owner‐ operator status at the top of the ladder was long and arduous. Most southern blacks were dirt poor, and their meager incomes went primarily towards consumption rather than saving. Credit was generally scarce in the post‐bellum south. The racial climate was hostile at best and deadly at worst. Black farm 9 We estimated linear probability regressions of ownership using white households in the core sample; the key independent variable in these regressions is a dummy for farm home status. The regressions also include dummy variables for age, literacy, place of birth, marital status, household size, and county of residence. In the 1900 regression, the coefficient on farm home status is 0.209 (s.e. = 0.004). By 1920, the same coefficient had declined in magnitude to 0.126 (s.e. = 0.004) and then to 0.051 (s.e. = 0.004) in 1930. 8 owners, in particular, were a visible target for the Klan and others bent on enforcing the racial status quo. Despite the impediments placed in their way, the census data reveal that many black farmers did manage to become owner‐operators and thus home owners. In 1870, about 13 percent of blacks residing in farm homes were owner‐operators but, because just 19 percent of black homes were designated as farms, a scant 2.5 percent of the core sample of black men owned and occupied a farm home. By 1910, 40 percent of black homes were designated as farms, and the ownership rate had risen to about 26 percent. Thus, in 1910, about 10 percent of the black sample consisted of men who owned and occupied farm homes, a 400 percent increase over a 40 year period. The census data of the era are not especially rich in information, but they do offer some clues regarding the causal determinants of black achievement of home ownership in the late nineteenth century. A linear probability regression of black owner‐occupancy in 1900 reveals significant positive effects of literacy and occupational status, even with extensive controls for household structure, family size, and county fixed effects. The literacy connection is important because after the Civil War black children were more likely to receive some schooling and thus become literate than those who came of age just before the conflict.10 It is likely that the accumulation of human capital contributed to the accumulation of real property and higher rates of home ownership. After four decades of increase, black owner‐occupancy stagnated after 1910. It was about this time that the “Great Migration” of African‐Americans to urban areas began in earnest (Collins 1997). Black migrants usually settled in neighborhoods that were in close proximity to central business districts, where low‐skilled employment in manufacturing and related sectors could be obtained. Urban jobs, particularly in the North, paid far better than agricultural employment in the South, but the gains in income did not translate immediately into higher rates of black home ownership. This was partly for the reasons already described for urban whites, but also because neighborhoods where higher‐income black 10 Using the core sample of male household heads in 1900, and restricting the sample to individuals who were 31 to 40 years of age, we estimated an IV regression of ownership on literacy in which the instrument for literacy was a dummy variable for Southern birth after 1865 (the regression also includes dummies for age, race, and state of birth). The coefficient of the instrument in the first stage is positive (0.046) and highly significant (s.e. = 0.009) indicating that black men born just after the Civil War were more likely to be literate than those born just before. The 2SLS coefficient from the regression of ownership on (instrumented) literacy is 0.489 (s.e. = 0.376); multiplying this coefficient by the change in black literacy rates between 1870 and 1900 for the core sample (30 percentage points) accounts for nearly all of the gains in black ownership over time period. However, the same calculation applied to ages 55‐64 under‐predicts the gains in ownership, which suggests some pure “catch‐up” in wealth accumulation was also present. 9 households attempted to settle became contested terrain as whites resisted the black influx. The end result was an increase in the extent of residential segregation (Cutler, Glaeser, and Vigdor 1999).11 Was residential segregation an impediment to black home ownership in metropolitan areas during the first wave of the Great Migration? We investigated this question econometrically by estimating census year ownership regressions using pooled samples of black and white household heads from the IPUMS. The key independent variable is a standard measure of residential segregation, the dissimilarity index, interacted with a race dummy (Black = 1). The regressions include metropolitan area fixed effects, so while the level effect of segregation is not identified, the interaction effect is identified. For 1920 and 1930, the relevant census years during the initial phase of the black migration, the coefficient of the interaction effect is negative and significant, indicating that increasing segregation was associated with a widening racial ownership gap.12 The 1930s were a disaster for home ownership. During the Great Depression millions of Americans lost their homes as a consequence of economic downturn. The Depression story is well‐ known, so we do not recount it here, but in terms of owner‐occupancy the decline in levels was modestly larger for whites, and thus the racial gap in percentage points narrowed slightly between 1930 and 1940. The important point, however, is that in no other census year during the twentieth century were home ownership rates lower than in 1940. On the eve of World War Two, 43 percent of white households in our core sample were owner‐occupiers, compared with 21 percent of black households. In the two decades after 1940 the “American Dream” took on its modern form. Whites increased ownership from 43 percent to 69 percent and blacks from 21 to 41 percent. In percentage point terms, the gains for white were greater than for blacks, and thus the racial gap in percentage points increased from 22 to 28 points. The widening of the gap was split equally between the two decades.13 The widening of the racial ownership gap between 1940 and 1960 was unusually pronounced at younger ages (see Figure 2). In 1940, 24 percent of white men in our core samples ages 25‐34 were 11 Various cities in the US passed so‐called “segregation ordinances” restricting the residential locations of black households but these ordinances were declared unconstitutional by the US Supreme Court in 1917 in Buchanan v. Warley; see Rice (1968). 12 The core sample is used for the regressions restricted to residents of metropolitan areas for which estimates of the dissimilarity index is available from Cutler, Glaeser, and Vigdor (1999). The specification includes a fourth degree polynomial in age, race, dummy variables for literacy (1900‐1930) or single years of schooling (1940‐1990), marital status, citizenship status, and linear terms in family size and occupational status (as coded by IPUMS), SMSA fixed effects, and the interaction effect between the race dummy and the dissimilarity index. The coefficient of the interaction term for both 1920 and 1930 is ‐0.12, and is statistically significant at the 5 percent level. 13 The equal splitting result is a consequence of the new estimate for 1950. 10 owner‐occupants; the corresponding figure for African‐American men was a meager 9.8 percent. In 1960, the ownership rate for white men ages 25‐34 had more than doubled to 55.4 percent; the corresponding rate for black men had jumped to 26 percent. Thus, among younger men, the racial gap approximately doubled in size, from 14 percentage points to 29 percentage points, between 1940 and 1960. In 1960 the racial ownership gap (ages 25‐64) overall stood at 28 percentage points, a larger gap than in any earlier census year during the twentieth century. Over the next two decades, however, the gap began to close, declining to 23 percentage points in 1970 and then to 19 percentage points in 1980. From 1980 to the present, however, the gap has drifted upwards. The racial difference in 1980 was the smallest of any census year in the twentieth century. Some portion of the gains in ownership after World War Two can be attributed to institutional reforms in the mortgage market in the 1930s and 1940s. These reforms included the establishment of the FHA and associated changes in lending norms that led to lower down payments, the replacement of “balloon” mortgages with self‐amortizing loans, and longer term lengths. Recent work by Fetter (2010) demonstrates that subsidies to veterans also played an important role in the ownership run‐up among whites. The relaxing of credit constraints, along with veterans’ subsidies, would likely have their largest impact at younger ages, which is consistent with the pronounced widening of the racial gap between the ages of 25 and 34. It is also part of the conventional wisdom that the institutional reforms of the 1930s were biased against black home ownership. For example, the implementation of FHA loan guidelines included “red‐lining” of black neighborhoods, which made it more difficult for black families to obtain mortgages which, again, would tend to have larger effects at younger ages. Suburbanization per se may have played an important role in the post‐WW2 run‐up in white home ownership. Although suburbanization was underway before the war, improvements in inter‐ and intra‐urban transportation, such as the Interstate Highway System, facilitated the decentralization of jobs and people within metropolitan areas (Baum‐Snow 2007). Because land was cheaper on the urban periphery, among other reasons, suburban housing consisted disproportionately of detached, single family homes, among which ownership was the norm. Urban blacks, however, faced considerable difficulties in moving to the suburbs. As whites left the central city, urban ghettos expanded their geographic boundaries, and residential segregation increased further, reaching a peak about 1970. However, while our regressions for 1920 and 1930 suggest that residential segregation was an impediment to home ownership among urban blacks before World War Two, an analogous regression 11 for 1970 finds no evidence that residential segregation was positively associated with the racial ownership gap. The changing relationship between residential segregation and the ownership gap may be explained by the change in the nature of segregation before and after World War Two. Before World War Two, when blacks moved into central cities, their housing options were limited, and they faced high housing prices, relative to whites. After World War Two, white movement to the suburbs raised the level of segregation in metropolitan areas, but suburbanization also meant that blacks gained access to middle‐class urban neighborhoods that whites had previously occupied. White households fled the city, but the single‐family homes they had occupied remained behind, potentially available for black owner‐ occupants. Using census data, Boustan and Margo (2010) show that, other factors held constant, black homeownership in central cities increased disproportionately in cities that experienced above‐average white suburbanization between 1940 and 1980. Initially, therefore, as African‐Americans moved to metropolitan areas, there were significant impediments to home ownership, but these impediments began to erode over time. As they eroded, urban blacks who would have, except for the impediments, been more likely to owner‐occupy (i.e., those with higher incomes and education) could begin to achieve the American Dream, albeit later in the life‐cycle than otherwise similar whites. If this argument is correct we should see more continuous racial convergence over time in owner‐occupancy at older ages, as well as strong correlations between convergence in ownership and convergence in income and schooling. Figure 3 shows owner‐occupancy rates by race for core sample households in which the head’s age is between 55 and 64; also shown is the racial gap in percentage points. In 1870, the racial gap in this age group was enormous, a full 62 percentage points. Over time the gap closed dramatically, and in striking contrast to the overall series, it did so monotonically.14 Figure 4 contrasts the racial gap ownership for the age group 55‐64 with racial gaps in household income (panel A) or educational attainment (panel B) over the 1940 to 1980 period. The relationship between the ownership and education (or income) gaps is nearly linear. A narrowing in the education or income gap translates directly into a reduction in the ownership gap.15 14 See Appendix Table 3 for the data points underlying Figure 2. We also estimated an IV regression using core sample households for 1980 restricted to heads between the ages of 30 and 39 who were born in the United States during the first or fourth quarter. The dependent variable is owner‐occupancy and the regression includes dummy variables for age, race, and state of birth. We instrument for years of schooling using a dummy variable for birth in the fourth quarter, whose coefficient is expected (and is) positive. The second stage coefficient of years of schooling is 0.117 (s.e. = 0.074). The coefficient of the race dummy in this regression is negative and statistically significant (‐0.116, s.e. = 0.063) but is much smaller in 15 12 The importance of racial convergence in permanent income is also illustrated by the absence of significant convergence in ownership after 1980, even in the oldest age group in our sample (see Figure 3). The passage of the Fair Housing Act of 1968 ushered in a new era of federal regulation of housing markets aimed at mitigating the effects of racial discrimination inherited from the era of high levels of residential segregation as well as overt barriers such as redlining. Broadly speaking, these regulatory efforts focused on enhancing the ability of African‐Americans to become home owners by “opening up” the suburbs and by easing access to mortgage finance. In light of the scope of the regulatory effort, it is astonishing how little evidence there is of treatment effects. For example, the Fair Housing Act of 1968 had as its goal to equalize racial access to housing by, among other effects, imposing sanctions on real estate agents who discriminated. If the act had its intended effects we might expect to see a jump in the fraction of African‐Americans who lived in the suburbs because it was in the suburbs where black access had been most restricted. However, regression analysis of CPS micro‐data reveals no evidence of a structural break around 1968 in blacks’ likelihood of a suburban residence (Collins and Margo, in progress). The upshot is that, during the period of greatest regulatory effort at enhancing black access to housing finance and the suburbs, the racial gap in home ownership has stubbornly resisted further narrowing. Indeed, if anything, the gap has widened slightly, according to our time series. However, as has been well‐established by labor economists, racial convergence in incomes and education has largely ceased in the United States since 1980. Housing regulation aimed at combating racial discrimination in housing markets may be justified on equity grounds, but such policies cannot be responsible for any of the long‐run convergence in black‐white owner‐occupancy rates because such convergence predates the policies. Concluding Remarks New estimates of home ownership reveal that the majority of the long‐run narrowing of the racial gap in home ownership took place before World War One. The proper interpretation of this finding requires understanding how the shift of labor out of agriculture figures into secular trends in owner‐occupancy. As whites left the farm in the late nineteenth century, the aggregate rate of white owner‐occupancy declined because non‐farm ownership was far lower than farm ownership among magnitude than the difference in sample means or the coefficient of the race dummy if the regression is estimated by OLS. 13 whites. Blacks inherited little or nothing in the way of landownership from slavery but were able to substantially increase their ownership rates during the post‐bellum period. The end result was a substantial narrowing of the racial ownership gap from 1870 to 1910. As African‐Americans left the rural South for urban areas during the Great Migration various impediments limited access to owner‐occupied housing. White flight in the aftermath of World War Two opened up new neighborhoods in which middle‐class blacks could finally achieve the American dream, albeit at a later stage in the life‐cycle than whites. Despite a massive effort to monitor and regulate housing markets for racial discrimination, racial convergence in owner‐occupancy stalled after 1980, in our view because racial convergence in the key determinant of housing demand – permanent income – largely ceased. 14