The Legacy of Redlining:
Residential Segregation in 147 American Cities

History of the HOLC "Redlining" Maps

The Home Owners' Loan Corportation (HOLC) was a New Deal era agency that helped provide emergency bailouts to banks and homeowners during the 1930s mortgage crisis that resulted from the Great Depression. The HOLC created credit worthiness maps for local banks and the Federal Housing Administration that indicated the lending risk by neighborhood. Since risk is based on property values as well as individual credit worthiness, the agency relied on local banks and realtors in 239 cities to appraise neighborhoods based for various levels of economic conditions, categorizing them into "Best", "Still Desirable", "Definitely Declining", and "Hazardous".

Guidelines for lenders are:

• "The First grade or A areas are 'hot spots'; they are not fully built up. In demand during good times and bad.
• "The Second grade or B areas, as a rule, are completely developed."
• "The Third grade or C areas are characterized by age, obsolescence, and change of style; expiring restrictions or lack of them; infiltration of a lower grade population."
• "The fourth grade or D areas...are characterized by detrimental influences in a pronounced degree, undesirable population or an infiltration of it...Some mortgage lenders may refuse to make loans in these neighborhoods and others will lend only on a conservative basis."

(FHLBB Division of Research and Statistics, 1937)

These appraisers evaluated neighborhoods based on criteria such as housing conditions, transportation access, distance to amenities and disamentiies, the economic class and status of the neighborhood's residents, and their ethnic and racial composition. (NRSC, 2018) The presence of minority (mostly black at the time) homeowners was considered to devalue both the property and the neighborhood; therefore, local appraisers also had an incentive to reinforce racial homogeneity.

Many consider the creation of these "residential security maps" to have faciliated residential redlining - the widespread practice of illegally denying home loans (and the opportunity to build wealth) based on race. The influence of these maps is debated: some argue that these maps directly bolstered governmental redlining (Jackson, 1987, Massey and Denton, 1993), while others say that, since these maps were made after 1936, when most of the HOLC bailout loans were provided, there is no direct impact (Hillier, 2003a, Hillier, 2003b). However, it is undeniable that there was (and continues to be) institutional discrimination against minority borrowers. These maps reflect the perceived lending risk held by banks and other real estate actors.

The 1968 Fair Housing Act made housing discriminination based on "race, color, religion, or national origin" a federal crime. By this point, however, the spatial structure of neighborhood segregation had already been established.

This map aims to show that the impact of discriminatory housing practices on the landscape of American cities and on the welfare of its residents is still seen today. Many of those neighborhoods deemed "Hazardous" or "Definitely Declining" were and remain minority areas with lower levels of income, education, and employment than their counterparts with higher HOLC grades.

Data

This map is indebted to the work done by the Digital Scholarship Lab, who geo-referenced the original HOLC maps as well as produced the shapefiles. They have a great interactive feature of all the HOLC maps. They also have a project mapping the impact of urban displacement due to "slum" clearance (a.k.a urban renewal) projects during the mid-20th century.

The historical decennial census and American Community Survey data comes from IPUMS's National Historical Geographic Information System (NHGIS) at the University of Minnesota. Because census areas, categories, and collection methods changed every decennial census, comparison between decades is not straightforward. Here are some notes about the data used:

• Before the 1940, the census only took place in metropolitan areas. Not all of our cities have coverage in 1930.
• Education information is not available until 1940.
• Income information is not available until 1950.
• Only starting in the 1960 census was Hispanic origin information collected. The census considers Hispanic origin not be a race but a "heritage, nationality, lineage, or country of birth". But only starting in 1990 did the Census start to separate race groups by Hispanic and non-Hispanic (for ex: Hispanic asian and non-Hispanic asian). Starting in the 1990s, I use non-Hispanic populations for White, Black, Asian, Native American, Pacific Islander, and other races.
• Unemployment information is not available until 1970.
• Until the 1970 census, the only required races were "White","Black", and "Other non-white".
• In 1960 and 1970s, median household income information only exists in ranges. I've created a random uniform sample of each income range based on the number of households in each range to estmate the median income value.
• All median household incomes have been inflation-adjusted from levels on the year before the census year (this is the year the information was actually collected) to 2016 levels. For instance, the 1960 census has incomes adjusted from 1959 to 2016.
• Staring in 2010, I use the American Community Survey - a monthly rolling survey published annually - as the longform census (containing additional questions such as education and income) was replaced by the ACS.
• The 2010 data uses the 2008-2012 5-year American Community Survey as an unbiased estimate of 2010 statistics.
• The 2016 data uses the 2016 1-year American Community Survey.

To learn more, the US Census Bureau provides a historical overview of each decade's census.

Method

The census data for HOLC zones is created through a population-weighted average of census data for each year, where the population is assumed to be proportional to the percentage overlap of the tract or block group with the HOLC zone. (This assumes that people are distributed uniformly across each census area, which of course, is not true.) As census units change size, each year's census needs to be calculated separately. Until 1980, I use the smallest geographic unit available, which is the census tract. Starting in 1990, I use census block groups.

Get in touch!

This project was created by Wenfei Xu. I am a PhD candidate in urban planning at Columbia GSAPP.

The original HOLC residential security map data was created by the Digital Scholarship Lab. Please reach out if you'd like a copy of my data: xu.wenfei[at]gmail.com