![Sample our Geography journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5937/TOC-Subject-Sample-2021-Geography.jpg"})
Abstract
In this paper, I build on previous historical studies that have analyzed the lending activities of the Homeowners’ Loan Corporation (HOLC), a New Deal agency established in 1933 to mitigate the national mortgage foreclosure crisis caused by the economic fallout of the Great Depression. I investigate whether discriminatory lending practices took place in Greater Cleveland, Ohio. I also examine a second question related to the contemporaneous impact of HOLC lending: did lending actually decrease Depression-era foreclosure rates in areas of significant lending activity? Census tract-level ordinary least squares, two-stage least squares, and spatial regression models are estimated to answer these questions. While studies of the HOLC in other cities (including Philadelphia, Pittsburgh, St. Louis, and Chicago) have attempted to assess the possibility of discriminatory lending outcomes, their findings have remained inconclusive. In addition, previous studies have not assessed the HOLC’s practical success, especially in terms of foreclosure amelioration. I use extensive historical evidence, in conjunction with appropriate tools of empirical analysis, to explore the evidence that the HOLC’s lending practices were discriminatory in Cleveland and that the practical effect of the HOLC lending was nonexistent. Additionally, in areas of high HOLC lending concentrations for the years 1933–1935 (the years of its operation), there was no significant reduction in relative foreclosure rates in subsequent years (1936 and 1937).
Notes
1. C. Lowell Harriss’s institutional history of the HOLC for the National Bureau of Economic Research in 1951, History and Policies of the Home Owners’ Loan Corporation, contained a chapter on the lending activities and detailed that in some cities, “wholesale operations” took place in which persons who had loans with troubled banks were given priority in lending due to the bankruptcy or imminent insolvency of their lending institution. Harriss notes that policies varied by state, but the HOLC would often purchase the loans of an institution in crisis and provide financing on a case by case basis, after the fact of purchase. At the end of HOLC lending activities in 1936, 13 percent of all HOLC loans were originated through this method. See Harriss (Citation1951, p. 38–39).
2. The decline for 1933 and 1934 is due to the institution of a foreclosure moratorium in Ohio during these years.
3. The location quotient is modified due to the use of population in the final denominators instead of total federal spending, which did not exist in the Fishback data. The actual formulae for the state and county computations are:
States: ((State HOLC Funding ÷ United States HOLC Funding) ÷ (State Population ÷ US Population))
Counties: ((County HOLC Funding ÷ United States HOLC Funding) ÷ (County Population ÷ US Population)).
4. See Green (Citation1934) for information on the creation of the Cleveland Real Property Inventory. Green was also a consultant to the 1940 Census and was formative in the implementation of a standardized census tract adoption process that was extended to 61 cities for that Census. Also, see Krieger (Citation2006).
5. The index that was accessed is located at the Special Collections Division of the Cleveland State University Michael Schwartz Library.
6. These maps were obtained from the National Archives in College Park, Maryland and were digitized.
7. Crossney and Bartelt’s data come largely from 1940 Census of Housing and estimated the HOLC lending impacts from data compiled four years after lending ceased. Hillier did not estimate regression models assessing the influence of race on HOLC lending.
8. In a 1939 RPI publication titled Relative Desirability of Census Tracts in Greater Cleveland stated in part that “the most desirable places to live are assumed to be census tracts: With large percentages of family heads native white or born in the British Empire and small percentages born in Italy or in Russia, or Negro.” This was one desirability characteristic from a list of eight. See Green (Citation1939, p. 1).
9. For each model, previous instability ratios are instrumented for, but do not technically replace, the endogenous independent variables of interest. For the discriminatory lending model, this included the average of the 1933 and 1934 instability ratio instruments for the previous foreclosure rate variable and for the foreclosure model, this included the average of the 1934 and 1935 instability ratio instruments for the HOLC lending concentration variable. For both sets of models—through residual analysis—it was determined that the instrumental variable did in fact reduce heteraskedasticty through the 2SLS process by eliminating the correlation between the endogenous variable in each model and the error term in each model.
10. Foreclosure models were also estimated for the years 1938 and 1939 using a similar 2SLS and spatial 2SLS approach. Results were not at great variance from the first two foreclosure models. One interesting difference from the previous models was that, by 1939, the HOLC concentration variable was no longer significant. This may reflect a reversion toward a normal real estate lending environment by this time.
11. A description of the interventions, including HAMP, and approved “servicers” can be found at www.makinghomeaffordable.gov.
12. Several websites provide access to these maps. Nearly 50 city survey maps are now freely available. See http://library.osu.edu/find/collections/maps/redlining-maps-ohio and http://www.urbanoasis.org/2012/07/29/holc-maps/ for the best online collections.