Racial Segregation in Housing Markets and the Erosion of Black Wealth1 Prottoy A. Akbar (University of Pittsburgh) Sijie Li (University of Pittsburgh) Allison Shertzer (University of Pittsburgh and NBER) Randall P. Walsh (University of Pittsburgh and NBER) November 2018 Abstract: Housing is the most important asset for the vast majority of American households and a key driver of racial disparities in wealth. This paper studies how residential segregation by race eroded black household wealth in U.S. cities. Using a novel sample of matched addresses from prewar American cities, we find that rental prices and occupancy soared by about 40 percent in blocks that transitioned from all white to majority black. However, home values fell on average by 10 percent over the first decade of racial transition and by a staggering 50 percent in major African American destinations such as Chicago, Philadelphia, and Detroit. These findings suggest that, because of the segregated housing market, black families faced dual barriers to wealth accumulation: they paid more in rent for similar housing while the homes they were able to purchase rapidly declined in value. 1 Support for this research was provided by the National Science Foundation (SES-1459847). We thank Werner Troesken, Arie Beresteneau, William Even, Thomas Davidoff, and seminar participants at UC-Irvine, Miami of Ohio, Pittsburgh, and Tennessee for helpful comments. We thank John Logan for assistance with enumeration district mapping and Carlos Villarreal and the Union Army Project (www.uadata.org) for the 1930 street files. Antonio Diaz-Guy, Phil Wetzel, Jeremy Brown, Andrew O’Rourke, Aly Caito, Loleta Lee, Zach Gozlan, and Sai Konduru provided outstanding research assistance. Corresponding author’s email: [email protected] (A. Shertzer). 1 “Daisy and Bill Myers, the first black family to move into Levittown, Pennsylvania, were greeted with protests and a burning cross. A neighbor who opposed the family said that Bill Myers was ‘probably a nice guy, but every time I look at him I see $2,000 drop off the value of my house.’ ” - Ta-Nehisi Coates, We Were Eight Years in Power: An American Tragedy (2017) “During the early nineteen twenties it is estimated that more than 200,000 Negroes migrated to Harlem… It was a typical slum and tenement area little different from many others in New York except for the fact that in Harlem rents were higher… Before Negroes inhabited them, they could be let for virtually a song. Afterwards, however, they brought handsome incomes.” - Frank Boyd, American Life Histories Manuscripts (WPA Federal Writers' Project, 1938) I. Introduction Housing is the most important asset for the vast majority of American households and a key driver of racial disparities in wealth (Blau and Graham 1990, Wolff 2014, Albouy and Zabek 2016). Social scientists have long hypothesized that racial income inequality reproduces itself in housing wealth, with minority groups who face discrimination in the labor market less able to build equity in their homes. The result is impoverished neighborhoods that impede the health, educational attainment, and upward mobility of the next generation of black children (Ananat 2011; Wilson 2012; Chetty, Hendren, Kline, and Saez 2014). In this narrative, segregated neighborhoods harm the socioeconomic standing of their residents through a complex system of interrelated disadvantages. Recent scholarship has focused on how discrimination in credit markets served to exacerbate forces operating in labor markets and educational systems to limit black accumulation of housing wealth.2 Far less attention has been paid to the direct role that segregated housing markets played in sapping black households of resources and inhibiting their ability to accumulate assets. Using a novel dataset of rents, home values, and the racial composition of city 2 A key focus of this work has been the role of redlining by both private lenders and federal agencies, namely the Federal Housing Authority. For instance, see Aaronson, Hartley, and Mazumder (2017). 2 blocks in prewar American cities, we examine the distinct disadvantages faced by black households due to the establishment and expansion of segregated neighborhoods. We find that rental prices and occupancy rates soared when city blocks transitioned from all white to majority black. Further, when black families were able to escape these escalated rents through the purchase of their own home, these same market dynamics led to the erosion of their homes’ value. The history of racial transition in twentieth century American neighborhoods contains many references to “blockbusting,” the process by which the black ghetto expanded into formerly white neighborhoods (Massey and Denton, 1993). Black families, who were desperate for quality housing, would outbid whites for apartments and homes outside (but typically adjacent to) the existing ghetto. Furthermore, because blacks were willing to pay a premium relative to whites for even substandard housing, formerly single-family homes could be subdivided to accommodate more residents. As highlighted by the two passages that begin this paper, that blacks would pay higher rents stands contrast to an expectation that housing prices would fall as the block transitioned. Popular accounts of the day focus extensively on the concerns of white homeowners who expected their property to lose substantial value once black families began settling nearby. 3 These concerns fueled their exodus to neighborhoods safe from the threat of black encroachment (Mehlhorn 1998, Boustan 2010). The notion that racial transition could be associated with both increases in aggregate rental prices and decreases in property values poses a puzzle at first glance. Indeed, to our knowledge, no existing empirical papers argue that the arrival of black residents caused the 3 A recent theory paper (Ouazad, 2015) models blockbusting as a problem in which fee-motivated real estate agents trade off a decline in property values associated with black entry with an increase in commissions due to higher turnover. We cannot observe the actions of real estate agents in our setting and thus restrict our attention to prices in the housing market. 3 prices of owned and rented housing to diverge. To understand the mechanisms that could generate such a finding, we consider the capitalization rate that is implied by a no arbitrage condition between rents and home values (Kearl 1979; Poterba 1992). Here, the capitalization rate is the percent of a property’s value that must be received as rent each year to make an investor indifferent between holding the asset and receiving rent and selling the property at its market value. Under the no arbitrage condition, all else equal, a real estate investor must be compensated by higher rents today if she is to purchase a property whose price is expected to fall in future periods. Similarly, expectations of high maintenance costs would require higher rents. The causal effect of racial transition on rental and home prices is difficult to identify because the ghetto tended to expand endogenously into areas populated by older residents and with lower quality housing. Accordingly, our empirical approach addresses the concern that black families may have been allowed to move into blocks where home prices would have fallen even in the absence of racial transition. To facilitate the identification of the causal impact of racial transition on prices, we match the universe of household addresses from ten major northern cities across the 1930 and 1940 federal censuses to create a panel dataset of single- family homes and apartment buildings. The 1930 and 1940 censuses were the first to ask about home values and rents, and the expiration of census confidentiality rules enables us to observe the same address in both years, along with a reported rent or valuation and the race of the occupants. This panel dataset provides several avenues for causal identification of the racial transition effect on home prices. First, to minimize the potential for omitted variable bias related to where black families already lived, our baseline sample consists of city blocks that were all white in 1930. We further restrict our attention to addresses that were single-family, owner- 4 occupied homes at the start of the decade. Importantly, with our linked sample, we can use 1930 price as a proxy for all time-consistent unobserved address-level characteristics. Linking also allows us to control for the 1930 occupancy-rate and a set of city block-level characteristics such as rental rate. In our preferred specification, we also include fixed effects for the enumeration district (a geographic area typically comprising less than four city blocks). Identification thus relies on variation in block-level racial transition within these narrowly defined neighborhoods that cannot be predicted by 1930 housing value, occupancy rate or city block-level rental share. Impacts of racial transition are large. We find that rental prices and occupancy soared by 40 percent in blocks that transitioned from all white to majority black. In contrast, home values fell by 10 percent relative to blocks that remained all white. Further analysis suggests that increases in occupancy in homes on transitioning blocks were a key mechanism underlying these declines in value. When we focus our analysis solely on homes that saw both increases in occupancy and block-level racial transition, the estimated decline in price increases to 28 percent. In contrast, we observe sharp increases in rental prices for all houses on transitioning blocks, whether occupancy increased or not. Summarizing the impact of racial transition on the relationship between rents and prices, we estimate that the capitalization rate on blocks that became majority black was about 17 percent, compared with 11 percent on blocks that remained white. Black families bore the brunt of both shifts in prices. When we decompose the fall in home prices by the racial composition of the block and race of the homeowner, we find that the prices fell the most during the beginning of the transition process, when black arrivals tended to be better off and were more likely to buy their home. On the other hand, rental prices increased the most after the block attained majority black status and black arrivals were typically poorer 5 renters. Furthermore, rents only increased for black families, not white families who remained during the transition process. Finally, we document significant heterogeneity across cities with those cities that saw the largest inflows of black migrants (e.g. Chicago, Philadelphia and Detroit) experiencing the largest rent and price impacts from racial transition. In these cities, capitalization rates in blocks that transitioned to majority black exceeded 20 percent and home values on these blocks dropped by a staggering 50 percent over the first decade of transition. Our findings have important implications for understanding racial disparities in wealth. Because of the segregated housing market in American cities, black families faced dual barriers to wealth accumulation: they paid more in rent for similar housing while simultaneously experiencing declining values in the homes they were able to purchase. Meanwhile, real estate investors faced such large demand from black renter households desperate for housing outside of the already underserved ghetto that they were able to charge sufficiently high rental prices to overcome the expected losses due to future declines in the value of their capital. The large gaps in capitalization rates that we document in this paper provide strong evidence that as of 1940, the market was already anticipating that homeowners in black neighborhoods would see their wealth severely eroded by housing market forces over the coming years. II. Background and Related Work A. Historical Background The Great Migration saw millions of African Americans leave the poverty and oppression of the Jim Crow South for better lives in northern cities. However, they soon discovered that the North maintained its own system of racial segregation, particularly in housing markets. Black families found themselves largely restricted to homes in the existing black ghetto through a 6 mixture of threats, actual violence, and discriminatory real estate practices. The narrative history emphasizes collective action taken by whites to maintain the color line, which shifted over time from angry mobs in the early days of the Great Migration to the later establishment of genteel neighborhood “improvement” associations (Massey and Denton, 1993). Such associations were created to lower the costs of adopting of restrictive covenants, which were deed provisions prohibiting the sale of a house to a black family. Such covenants were enforceable until the 1948 Supreme Court case Buchanan v. Warley. Yet the color line did not reliably hold. The 1920s and 1930s saw significant expansions of the ghetto in most northern cities. Urban historians underscore the desperation of black families for better housing and their tendency to outbid whites for homes near the ghetto. On the other hand, real estate professionals and academics were united in their belief that black entry would harm home values.4 Such expectations made banks reluctant to underwrite a mortgage for a “pioneer” black family entering a white neighborhood where the lending institution already held loans. One historian summarized the dichotomy thusly: “One of the most interesting points made in the [real estate] broker comments is the recurring theme that while sellers may not get their price from whites (who are reluctant to consider an area undergoing racial transition), they probably can from nonwhites. This is quite different from the unqualified prediction that all prices in an ‘invaded’ area fall” (Laurenti 1960, p. 20). The fact that the ghetto expanded even though black families tended to have fewer assets to use for a down payment suggests that some banks did in fact underwrite mortgages for them. While banks were typically reluctant to initiate racial transition on a block, they appear willing to have made loans in neighborhoods “destined” to turn. Surveys of real estate brokers from the period suggest that the first family to enter a white neighborhood often sought a mortgage from a 4 Some social scientists had a more nuanced view of the process. For instance, Gunnar Myrdal argued in An American Dilemma that white racism was the primary cause of drops in home values as a block began transitioning and that prices should recover once the neighborhood was majority black (p. 623). 7 distant bank that did not have exposure to the area in question (Schietinger 1953, p. 172). The narrative history on the issue of mortgage terms is mixed, with some surveys finding blacks and whites received similar terms (Rapkin and Grigsby 1960, p. 77) and other scholars arguing that African American borrowers were steered towards installment contracts where they could be lose possession of their home if they were late on a single payment (Satter 2009, p. 4). Of course, not all black families bought their own home. As we discuss below, we find that the proportion of renters increased throughout the transition process. The question of who owned properties rented out to black families is thus important for interpreting our results. The census does not allow us to observe the identity of property owners in the case where the occupants are renters. We thus turn again to the narrative history, which suggests white investors purchased properties in the black ghetto with the perhaps self-fulfilling expectation that their investment would sharply depreciate over time.5 Real estate brokers believed that houses that were converted to multi-family rentals would lose value over time and were generally unwilling to make loans for the purchase of such properties (McEntire 1960, ch. xiii). It would thus be necessary to buy these properties with cash. It is also likely the case that some landlords were simply household heads who decided to rent out the former family home instead of selling. Both considerations underscore the likelihood that property owners in our setting were typically white. B. Related Literature in Economics A large body of work in economics and related fields seeks to understand the causes and consequences of segregated housing markets. Of particular interest is the question of how preferences for racial residential segregation is manifested in housing prices. The consensus in 5 See for instance United States Congress House Committee on the District of Columbia, 1935, Rent Commission: Hearings before the subcommittee on Fiscal Affairs on H.R. 3809, p. 7. 8 the literature is that segregation that arises from constraints on black housing supply will result in black families paying higher prices for similar housing relative to whites. Indeed, most papers that examine racial housing price disparities between 1940 and 1970 have argued that blacks paid such a premium (King and Mieszkowski 1973; Yinger 1978; Schafter 1979). The passage of the Fair Housing Act in 1968 reduced the tools available to white families to maintain the color line, and most papers working with data from after 1970 argue that segregation was maintained by whites paying a premium to avoid black neighbors (Follain and Malpezzi 1981; Chambers 1992). Yet establishing that black and white families paid different amounts for the same quality of housing is extraordinarily difficult. Much of the research on such differentials in the years after the Fair Housing Act was passed necessarily compares housing in very different neighborhoods because so many whites had already moved to the suburbs. The seminal paper on this topic is Cutler, Glaeser, and Vidgor (1999), which proposes an indirect empirical test of the hypothesis that segregation generates price premia. They note that the black main effect on rental price is negative in every period they study, from 1940 to 1990, likely due to unobserved differences in housing quality. The authors thus draw inference from the interaction between black household and measured racial segregation in a particular city: a positive interaction term is than interpreted as evidence that blacks paid more for housing in segregated cities, hinting at the existence of supply constraints, while a negative interaction suggests that whites pay a premium. Interpreting estimates of these interaction terms, Cutler Glaeser and Vigdor conclude that blacks paid a premium in the 1940s and whites a premium in the 1990s. In any case, the finding that the typical black family paid a premium for housing circa 1940 is difficult to square with the anecdotal literature on the impact of racial transition on 9 property values in the early to mid-twentieth century. The history of the Great Migration makes many references to the deleterious impact of black arrivals on home values in northern cities. The FHA underwriting manual emphasized maintaining the racial composition of neighborhoods for this reason (FHA, 1936). In any case, it remains necessary to reconcile the potential drop in property values associated with pre-Fair Housing Act black in-migration in the with the black rental premium found in other work. Economists have recently dedicated a great deal of attention to government involvement in housing markets that may have had a discriminatory impact, particularly “redlining” in mortgage insurance (for instance, see Aaronson, Hartley, and Mazumder 2017). Beginning in 1934, the Federal Housing Authority initiated underwriting mortgages under policies that would disadvantage black neighborhoods in central cities. However, by 1940, the FHA had underwritten only 60,339 mortgages on existing homes in the metropolitan areas of the cities we study in this paper.6 Federal urban renewal policies did not begin until the 1949 Housing Act (Collins and Shester, 2013). It is thus unlikely that federal government policies can explain the findings of this paper. Instead, the FHA and subsequent federal policies likely served to institutionalize and reinforce the private market dynamics that we document. Our paper is also related to the literature on white flight from central cities. A number of researchers have investigated the population dynamics associated with neighborhood racial turnover starting with the Schelling (1971) model of neighborhood tipping. Card, Mas, and Rothstein (2008) explored the existence of neighborhood tipping in the late twentieth century. 6 These numbers come from the FHA’s Annual report for 1940 (FHA, 1941). We have been unable to identify exactly how large the metropolitan areas were for this reported data. However, as an example, the FHA reported more homes insured in the New York City Metropolitan area than it reported for the entire state of New York, suggesting that they used broad metropolitan area definitions. Thus, this number should likely be viewed as a conservative upper bound. In which case, FHA penetration into our sample would still have been quite limited as of 1940 (likely on the order of 2 to 4 percent). 10