Association for Public Policy Analysis & Management

Many scholars place blame on the federal government for segregating America through housing policies initially developed during the New Deal and expanded in subsequent decades. The Home Owners Loan Corporation (HOLC), Federal Housing Administration (FHA), and GI Bill largely excluded people and communities of color from affordable mortgage credit, thereby reifying racialized neighborhood boundaries. Despite broad consensus about the impact of these policies, there is little empirical evidence connecting their implementation to contemporary patterns of spatial inequality—or changes over time in those patterns. Recent efforts of the Mapping Inequality group (Nelson et al. 2019) to digitize HOLC maps has led to work showing that redlining carried significant consequences for racial isolation and housing values on the neighborhood-level (Aaronson et al. 2018). However, no work has thus far explored whether redlining practices affected racial geography measured on higher levels of aggregation (e.g. cities and towns).

This is the first paper to investigate the relationship between implementation of segregationist housing policies in the first half of the Twentieth Century and subsequent change over time in segregation patterns. Specifically, I combine Mapping Inequality data with nine decades of census data to draw a direct connection between federally-mandated redlining and segregation measured on the place-level. In doing so, I provide strong evidence in support of claims that the consequences of these federal interventions into the hierarchy of place can be seen today. I show that the cities and towns appraised by HOLC became far more segregated than cities and towns ignored by HOLC. Both groups of cities experienced dramatic increases in black-white segregation between 1940 and 1960, though the rise was significantly steeper for cities and towns HOLC evaluated. HOLC-graded places continued to be more segregated than their ungraded counterparts through the latter half of the Twentieth Century. In 2010, the black-white dissimilarity, black isolation, and white-black information theory indices were 10.9, 14.8, and 7.8 points higher, respectively, in places touched by HOLC than in other cities. Results are robust across multiple identification strategies, including difference-in-difference estimation with city fixed effects, subsample analyses among cities near the recommended (but not universally adhered to) population cutoff for HOLC appraisal (i.e. 40,000), and proximity analysis.