Measuring The Enduring Imprint Of Structural Racism On American Neighborhoods

A long history of discriminatory policies in the United States has created disparities in neighborhood resources that shape ethnoracial health inequities today. To quantify these differences, we organized publicly available data on forty-two variables at the census tract level within nine domains affected by structural racism: built environment, criminal justice, education, employment, housing, income and poverty, social cohesion, transportation, and wealth. Using data from multiple sources at several levels of geography, we developed scores in each domain, as well as a summary score that we call the Structural Racism Effect Index. We examined correlations with life expectancy and other measures of health for this index and other commonly used area-based indices. The Structural Racism Effect Index was more strongly associated with each health outcome than were the other indices. Its domain and summary scores can be used to describe differences in social risk factors, and they provide powerful new tools to guide policies and investments to advance health equity.


SREI data sources
We relied on several additional data sources, in each case using the most recent data available, such as 2014 data from the U.S. Environmental Protection Agency's (EPA) National Air Toxin Assessment to measure total air toxins and cancer risk due to air pollutant exposure.(18)Variables about food availability based on distance to grocers for many populations were available from the Economic Research Service of the U.S. Department (25) Local eviction data from 2016 was downloaded through the Princeton Eviction Lab.Eviction data for some census tracts are imputed based on data available at larger geographies.(26) The Healthcare Delivery Research Program National Cancer Institute provided a 2012 dataset on racial segregation including location quotients (LQ) and indices of concentration at the extremes (ICE) using measures described in Bemanian, et al. (27,28)

Comparison Indices, additional information
The ADI is available at the census block group level from the University of Wisconsin School of Medicine and Public Health Center for Health Disparities Research Neighborhood Atlas and is constructed from 17 ACS variables.(30,31) Organized into domains, the SVI is a composite of 16 ACS variables including demographic data such as ethnoracial group and disability status.(32) The SDI is a composite of seven ACS derived variables.(33)The COI creates scores from 29 variables organized into three domains.The variables include data from the ACS, the Department of Education, Environmental Protection Agency, and other Federal Agencies.

Variable selection
To select variables for inclusion, first, we used exploratory factor analysis to examine variables in each domain, clarifying some underlying patterns and leading to some changes to ensure that we measured what we intended.For example, the transportation domain originally seemed to be driven largely by urbanicity.Since transportation is important across all population densities, we included other variables that were less associated with urbanicity, such as carpooling.Second, we examined correlations and scatterplots within each domain to look for surprising directional, non-linear, or markedly high correlations between variables.
While examining many candidate variables, we sought fewer variables for the final index, valuing parsimony for several reasons.First, collinearity can obscure the impact of closely related variables in the same domain.Second, we sought to avoid creating a false sense of sophistication by including variables that would add little additional information (for example, "job availability" and "retail job availability" were 96% correlated, so only one was included).Finally, given our goal of developing a measure to be used in policy and advocacy, simplicity was a virtue.
The best approach to selecting an appropriate time frame for data was not initially clear.We opted for using the most recent data available at the time, both for the sake of simplicity and to best reflect current realities with shifting demographics, post-Covid emigration, urban revitalization efforts, etc.Though the ACS data used is 2015-2019 data, conceptually, the index captures more than just the status quo.For example, the education domain includes both current resources in the education system (per pupil spending) and the long-term effects of past resources in the education system (% of residents with a bachelor's degree).
Where data were unavailable at the census tract level, we used one of two methods to determine a tract level value.For non-intersecting boundaries (county, state, etc.), the census tract adopted its "parent" geography's value.For intersecting geographies (school district, ZCTA, etc.), we used weighted averages based on the fractions of the total land area in each census tract.Due to the irregular nature of some of the boundaries, such as school district, a population-weighted method is not immediately apparent as even census blocks are sometimes crossed.Notes: People of color is defined as Black, Latine, and Indigenous.ADI is Area Deprivation Index, SVI is Social Vulnerability Index, COI is Child Opportunity Index, SDI is Social Deprivation index, SREI is Structural Racism Effect Index.Unit of analysis is census tract.The top row of data shows the mean of each outcome by decile (i.e. the mean life expectancy in the decile of neighborhoods with the lowest life expectancy).The following rows show the mean value of the outcome per decile of neighborhoods ranked by the SREI and its individual domains (i.e. the average life expectancy in the decile of neighborhoods scoring the lowest on the index is 82.6 years).

Supplement 4 .
Mean Life Expectancy, Diabetes Prevalence, and Ethnoracial Makeup by Decile of the Structural Racism Effect Index and Domains Sources: Mean life expectancy from National Center for Health Statistics' US Small Area Life Expectancy Estimates Project (2010-2015).Diabetes Prevalence from Centers for Disease Control and Prevention PLACES Project (2020).Ethnoracial makeup from U.S. Census Bureau American Community Survey 2015-2019 estimates.
of Agriculture Food Access 2019 Research Atlas.(19)We accessed 2018 county-level data on the total jail population, pre-trial jail population, and prison population per 100,000 from the Vera Institute of Justice which aggregates local data and data from the U.S. Department of Justice Bureau of Justice Statistics Annual Survey of Jails and Census of Jails.(20)State and municipal law enforcement officer data from 2019 was provided by the Federal Bureau of Investigation Uniform Crime Reporting Program Police Employee Data for census places (municipalities) and states.(21)Per pupil spending data by school district from 2019 was obtained from the U.S. Census Bureau Annual Survey of School System Finances.(22)The U.S. Department of Housing and Urban Development and Department of Transportation Location Affordability Index Model Version 3 which uses 2012-2016 data was the source for job availability, retail job availability, and housing and transportation cost burdens for different household configurations.(23)Information on vacancy and foreclosure risk was obtained from U.S. Department of Housing and Urban Development Office of Policy Development and Research Neighborhood Stabilization program which uses 2010 data.(24)U.S. Census Bureau ACS Supplemental Poverty Measure resources provided an alternate poverty measure using 2019 data.