The Geography of Exclusion: Explaining state-level racial disparities in FHA and VA lending

• In Event: Beyond the Red Lines: New Directions in the Study of Housing Policy and Racial Inequality

Abstract

An expansive literature has drawn links between the contemporary racial geography in the U.S. and New Deal-era federal mortgage insurance programs. However, two of the largest programs—the Federal Housing Administration (FHA) mortgage insurance and Veterans Administration (VA) Home Loan Guaranty—have been seldom studied due to data limitations. The existing sources of loan-level data on these programs generally have small sample sizes and are limited to a few municipalities (e.g., Fishback et al. 2022), limiting our understanding of how the implementation and beneficiaries of these programs varied geographically. This variation matters because qualitative historical research shows that the implementation of these programs differed widely, and often in ways that reinforced inequality (Taylor 2019, Katznelson 2005).

In this paper, we examine geographic variation in racial and ethnic disparities in receipt of FHA-insured and VA-guaranteed loans. To do so, we construct the first quasi-national dataset of FHA and VA mortgage recipients by digitizing 120,000 records from the National Archives using Optical Character Recognition (OCR) models and machine learning classification models. We then link these records to the 1940 and 1950 Full Count Censuses to identify borrowers’ race and nativity. Our analysis shows that program beneficiaries were disproportionately white and U.S.-born. We also find significant state-level variation in race- and nativity-based disparities, especially in VA lending. 

In forthcoming analyses, we focus on explaining geographic variation in racially disparate lending, which we operationalize as the ratio of Black FHA/VA borrowers to the share of Black adults in each state. We predict the extent of racially disparate lending by using demographic (e.g., percent Black, percent foreign-born, decadal change in percent Black), spatial (e.g., Census region), political (e.g., state Democratic voting share, Black turnout rate), and economic variables (e.g., white homeownership rate, proxies of household income). Using similar models, we also analyze variation in the share of foreign-born FHA/VA borrowers. Where data permits, we replicate these analyses at the county-level. 

New Deal-era homeownership policies redistributed wealth disproportionately to white and U.S. born households. Our analysis focuses on the geographic dimensions of these racial disparities, situating them within social, political, and economic context. These results have important implications for studies of the long-term consequences of New Deal housing programs on racial and spatial inequality. 

Author

• Katherine Anne Thomas, New York University
  Presenting Author