Key Points
Question
Is exposure to residential redlining practices by the Home Owners’ Loan Corporation (HOLC) in 1940 associated with increased risk of death later in life?
Findings
A total of 377 000 individuals who resided within neighborhoods graded as D (hazardous [ie, redlined]) by HOLC in 1940 had an estimated 1.44 years’ lower life expectancy at the age of 65 years than individuals who resided within A (best) graded areas.
Meaning
Exposure to residential redlining in 1940 had sizable and statistically significant associations with residents’ risk of death later in life.
Abstract
Importance
Historic redlining, the practice by the Home Owners’ Loan Corporation (HOLC) of systematically denying credit to borrowers in neighborhoods that were inhabited by primarily African American individuals, has been associated with poor community outcomes, but the association with individual risk of death is not clear.
Objective
To examine if exposure to residential redlining practices by HOLC in 1940 is associated with increased risk of death later in life.
Design, Setting, and Participants
The study linked individuals who resided within HOLC-graded neighborhoods (defined as Census Enumeration Districts) in 1940 with administrative death records data. The study estimated hazard ratios as well as age-specific life expectancy gaps (at age 55, 65, and 75 years) for HOLC grading exposure. This was done using methods that adapted standard parametric survival analysis to data with limited mortality coverage windows and incomplete observations of survivors. The analysis sample consisted of 961 719 individual-level observations across 13 912 enumeration districts within 30 of the largest US cities (based on 1940 population counts) across 23 states. Data were analyzed between December 1, 2023, and September 4, 2024.
Main Outcome and Measures
The exposure was HOLC grade based on historic HOLC maps, with A representing “best” or creditworthy areas; B, “still desirable”; C, “definitely declining”; and D, “hazardous” areas not worthy of credit (ie, redlined), and the main outcome was age at death from the Social Security Numident file.
Results
The 961 719-person individual sample had a mean (SD) age of 19.26 (9.26) years in 1940 and a mean (SD) age at death of 76.83 (9.22) years. In a model adjusted for sex (52.48% female; 47.52% male), race and ethnicity (7.36% African American; 92.64% White), and latent place effects, a 1-unit lower HOLC grade was associated with an 8% (hazard ratio, 1.08 [95% CI, 1.07-1.09]) increased risk of death. At age 65 years, these hazard differentials translated into an estimated life expectancy gap of −0.49 (95% CI, −0.56 to −0.43) years for each 1-unit decrease of the HOLC grade.
Conclusion
This study found that individuals who resided within redlined neighborhoods in 1940 had lower life expectancy later in life than individuals who resided within other HOLC-graded areas.
This study explores whether individuals who resided in US Home Owners’ Loan Corporation redlined neighborhoods in 1940 had lower life expectancy later in life compared with individuals who resided in neighborhoods that were not redlined.
Introduction
Large health and mortality disparities exist across race in the US.1 Structural racism, defined as the ways in which societies foster discrimination through mutually reinforcing inequitable systems, has been identified as a potential contributor to observed racial health and mortality disparities.2,3,4,5,6,7,8,9 To empirically examine the association between structural racism, health, and mortality, several recent studies have used the historic practice of residential redlining as a manifestation of explicit structural racism.10,11,12,13 Historic redlining refers to the practice by the Home Owners’ Loan Corporation (HOLC), an agency within the Federal Home Loan Bank Board that was created in 1932, of systematically denying credit to borrowers in neighborhoods that were inhabited by primarily African American individuals and which they marked as “hazardous” (colored red) within their residential security maps.14,15,16 Although the Fair Housing Act of 1968 prohibited discriminatory practices by HOLC, a growing number of studies have identified associations between neighborhood-level exposure to historic redlining and present-day neighborhood-level health and mortality outcomes.11,12
This study built on prior work that has examined the association between redlining and mortality disparities, using primarily community-level data within the confines of limited geographic areas.10,11,13 This study leveraged large administrative death records data from 30 cities (across 23 US states) to examine later-in-life mortality outcomes of individuals who lived in areas exposed to residential redlining by HOLC in 1940.
Methods
Study Sample
We combined data from several sources to construct our analytic sample. First, we obtained digitized historic HOLC maps from the Mapping Inequality project.17 Second, we obtained city enumeration district maps corresponding to the 1940 census from the Urban Transition Historical GIS Project.18,19 Enumeration districts designate historic areas that are comparable in population size to present-day census tracts. Third, we obtained the complete 1940 decennial census sample from Integrated Public Use Microdata Series, which allowed us to identify the residents within each enumeration district.20 Fourth, we used census-linked mortality data from the CenSoc project. We used the Numident dataset, which is based on linkages between the 1940 census and the National Archives’ public release of the Social Security Numident file.21 This dataset contains information on men and women and has coverage of deaths occurring between 1988 and 2005. For additional sensitivity analyses, we also drew on alternative mortality data from the CenSoc Death Master File, which has coverage of male (only) deaths for a wider set of years (1975-2005).22
Our primary analytic sample was constructed by identifying 1940 enumeration districts with exposure to HOLC grading, identifying those individuals residing within these areas in the 1940 census, and then linking these records with the individual-level Numident mortality data. Our inclusion criteria beyond residence within a HOLC-graded area were: (1) born in the US, with both parents being born in the US and (2) race reported as African American or White. These criteria ameliorate concerns of unobserved confounders related to direct individual (or indirect parental) exposure to a non-US environment. The resulting sample consisted of 961 719 individual-level observations nested within 13 912 enumeration districts within 30 of the largest cities (based on 1940 population counts) across 23 US states.
The institutional review board at Texas A&M University determined that this study was exempt from human subject review. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
Study Variables
Outcome Measure
Our outcome measure was the individual’s age at death. Age at death was computed as the number of years between the year of birth and the year of death based on records from the Social Security Numident files.
Exposure Measure
We used a constructed HOLC grade score as our exposure measure (this measure has been used in prior work).12,23 This measure was constructed at the enumeration district level. In summary, the measure was constructed as follows: first, each HOLC-graded area was given a numeric score, with A = 1, B = 2, C = 3, and D = 4. These letters corresponded to historic HOLC grades used on HOLC maps to indicate the credit quality of said area, with A representing “best”; B, “still desirable”; C, “definitely declining”; and D, “hazardous” (ie, redlined) areas. Second, we computed the proportion of each enumeration district that was contained within a given HOLC-graded area. Third, for enumeration districts that mapped directly into 1 unique HOLC-graded area, they were given the HOLC grade score of that area; for enumeration districts that mapped into multiple different HOLC-graded areas, they were given a score based on a weighted average across the overlapping HOLC area scores. These steps yielded a HOLC grade score that was a continuous measure between 1 (best) and 4 (redlined) for each enumeration district.12
Additional Variables
Individual characteristics related to sex (female or male) and race (African American or White) were included from the 1940 decennial census. City of residence was identified based on data from the Urban Transition Historical GIS Project.18 Lastly, individual birth year information was identified based on the Numident death record data.
Additional Sensitivity Analyses Variables
Additional sensitivity analyses variables included a labor force participation indicator, occupation income score (score reflecting the median income of people in that occupation in 1950), and home ownership status (owner, renter, or not applicable) indicator. These variables were sourced from the 1940 decennial census and they were part of additional sensitivity analyses.
Statistical Analysis
Two common features of large administrative death records data are limited mortality coverage windows (ie, censoring from above and below) and incomplete observations of survivors. To account for this, we used recently developed methods that adapt standard parametric survival analysis to data with these features.24,25 Our adjusted model used the (continuous) HOLC grade score exposure measure, along with covariates consisting of sex, race and ethnicity, and city of residence indicators. We used maximum likelihood methods to estimate the hazard rate and age-specific mortality differentials (at age 55, 65, and 75 years), assuming the age at death distribution of birth cohorts followed a parametric Gompertz model.24,25 Reported estimates referred to the increased mortality hazard (or reduced age-specific mortality differential) due to a 1-unit decrease in the HOLC grade score exposure, which corresponded to a HOLC letter grade decrease from grade A to grade B (where grade A serves as the reference grade). Data were analyzed between December 1, 2023, and September 4, 2024. Additional method details are provided in the eMethods in Supplement 1.
Results
Summary Statistics
Our main analytic sample consisted of 961 719 individuals. We observed a mean (SD) age at death of 76.83 (9.22) years, a mean (SD) age of 19.26 (9.26) years in 1940, 52% of the individuals were female, and 7% were African American. The mean (SD) age at death for individuals who resided within a HOLC A (best) graded area in 1940 was 78.52 (9.47) years compared with 75.92 (9.11) years for those who resided within D (hazardous) areas (Table 1). Additional age at death distribution plots stratified by HOLC grade are provided in eAppendix 1 and the eFigure in Supplement 1.
Table 1. Summary Statistics for Full Sample and Home Owners’ Loan Corporation (HOLC) Grade Subsamplesa.
| Characteristic | Mean (SD) | ||||
|---|---|---|---|---|---|
| Full sample (N = 961 719) | HOLC grade = A (n = 26 352; 2.74%) | HOLC grade = B (n = 147 054; 15.29%) | HOLC grade = C (n = 411 313; 42.77%) | HOLC grade = D (n = 377 000; 39.20%) | |
| Age at death, y | 76.83 (9.22) | 78.52 (9.47) | 77.95 (9.37) | 77.15 (9.16) | 75.94 (9.11) |
| Sex, % | |||||
| Female | 52.48 (49.93) | 57.47 (49.44) | 54.49 (49.80) | 52.46 (49.94) | 51.36 (49.98) |
| Male | 47.52 (49.93) | 42.53 (49.44) | 45.51 (49.80) | 47.54 (49.94) | 48.64 (49.98) |
| Race and ethnicity, % | |||||
| African American | 7.36 (26.11) | 1.04 (10.16) | 0.48 (6.91) | 1.10 (10.41) | 17.32 (37.84) |
| White | 92.64 (26.11) | 98.96 (10.16) | 99.52 (6.91) | 98.90 (10.41) | 82.68 (37.84) |
| Age in 1940 census, y | 19.26 (9.26) | 21.03 (9.82) | 20.4 (9.57) | 19.57 (9.19) | 18.34 (9.06) |
Subsamples were grouped based on their inferred HOLC grade letters. These grade letters were assigned based on the HOLC score of the enumeration district that the individual resided within in 1940. An area was marked as A if the HOLC score was less than 1.5; B if the HOLC score was greater than 1.5, but less than 2.5; C if the HOLC score was greater than 2.5, but less than 3.5; and D if the HOLC score was greater than 3.5.
Hazard Ratios Due to HOLC Grading Exposure
Unadjusted model results indicated that for a 1-unit decrease of our HOLC score (which corresponded to a HOLC letter grade decrease [eg, A to B]) there was a 7% (hazard ratio [HR], 1.07 [95% CI, 1.07-1.08]) increased risk (across all ages) of death. Adjusted model results, controlling for individual characteristics (sex, race and ethnicity) and for latent place effects (using city fixed effects) indicated that a 1-unit lower HOLC score was associated with an 8% (HR, 1.08 [95% CI, 1.07-1.09]) increased risk (across all ages) of death.
Life Expectancy Gap Estimates Due to HOLC Grading Exposure
Life expectancy gap estimates for the unadjusted model showed that at the age of 55 years, estimated hazard differentials translated into a life expectancy gap of −0.66 (95% CI, −0.73 to −0.59) years (Table 2). That is, a HOLC score decreased by 1 unit (eg, by changing an area’s HOLC grade from A to B) corresponded to a −0.66-year decrease in life expectancy. Similarly, changing an area’s HOLC grade from A to D corresponded to a −1.98-year decrease in life expectancy. At age 65 years, the life expectancy gap estimate was −0.59 (95% CI, −0.66 to −0.52) years per HOLC grade decrease. At the age of 75 years, the life expectancy gap was −0.48 (95% CI, −0.54 to −0.43) years per HOLC grade decrease.
Table 2. Difference in Life Expectancy at Ages 55, 65, and 75 Years Based on 1-Unit Difference in Home Owners’ Loan Corporation (HOLC) Gradea.
| Difference in life expectancy | Unadjusted (95% CI), y | Adjustedb (95% CI), y |
|---|---|---|
| HOLC grade (life expectancy gap at age 55 y) | −0.66 (−0.73 to −0.59) | −0.57 (−0.64 to −0.49) |
| HOLC grade (life expectancy gap at age 65 y) | −0.59 (−0.66 to −0.52) | −0.49 (−0.56 to −0.43) |
| HOLC grade (life expectancy gap at age 75 y) | −0.48 (−0.54 to −0.43) | −0.39 (−0.44 to −0.34) |
In the Adjusted column, the −0.57 estimate indicates that at age 55 years, those who resided within HOLC B (still desirable) graded areas in 1940 could expect to (on average) live −0.57 years less than those who resided within A (best) graded areas; those who resided within HOLC C (definitely declining) graded areas could expect to (on average) live 2 × −0.57 = −1.14 years less than those who resided within A (best) graded areas; and those residing within D (hazardous) graded areas could expect to (on average) live 3 × −0.57 = −1.71 years less than those who resided within A (best) graded areas.
Adjusted results control for sex, race and ethnicity, and city (fixed effects) indicators.
Life expectancy gap estimates for the adjusted model, controlling for individual characteristics (sex, race and ethnicity) and for latent place effects (using city fixed effects), showed that at the age of 55 years, the estimated life expectancy gap was −0.57 (95% CI, −0.64 to −0.49) years (Table 2). At the age of 65 years, the estimated life expectancy gap was −0.49 (95% CI, −0.56 to −0.43) years per HOLC grade decrease. Lastly, at the age of 75 years, the inferred life expectancy gap was −0.39 (95% CI, −0.44 to −0.34) years per HOLC grade decrease.
Sensitivity Analyses
Within subgroup sensitivity analyses, we found that significant associations persisted even across subsamples of White individuals only, labor force participants, homeowners, and renters. These results also persisted when adjusting for individual-level occupational income score data (eAppendix 2, eTables 1-4 in Supplement 1).
To further assess the robustness of our findings, we used an alternative administrative death records dataset, the CenSoc Death Master File, whose mortality coverage window was broader (ie, covered deaths between 1975-2005) than that within our main dataset.22 Results from this alternative dataset were qualitatively similar to our main findings (eAppendix 3, eTables 5-7 in Supplement 1).
Lastly, we also assessed the sensitivity of our choice of exposure measure using 3 alternative measures. These all resulted in qualitatively similar results (eAppendix 4, eTable 8 in Supplement 1).
Discussion
Using a unique dataset with data on individual death records, the study documented deaths of individuals who resided within HOLC-graded areas in 1940. Adjusting for sex, race and ethnicity, and latent place effects, the study found that a 1-unit decrease of the HOLC score (eg, changing the grade of a neighborhood from A to B) was associated with an 8% increased later-in-life risk of death and a corresponding −0.49 years of reduced life expectancy at the age of 65 years.
The findings contribute to existing literature that has documented associations between historic HOLC maps and present-day health and mortality outcomes.10,11,12,13 However, this study differs from these prior studies in several ways. First, individual- rather than aggregate-level data were used within the analyses. Second, outcomes were based on mortality among individuals who resided within HOLC-graded areas in 1940 and not on the community-level outcomes of all individuals residing within these areas currently. Third, the results were not confined to a single city or region, but instead drew on data from across 30 of the largest US cities based on 1940 population counts.
Although the discriminatory practices of HOLC were made illegal with the passage of the Fair Housing Act of 1968, there are several reasons why historic HOLC maps are associated with present-day health and mortality outcomes. First, as has been noted previously,22 HOLC maps are best conceptualized as reflections of historic beliefs about race, place, and value (during the time of the maps’ creation as well as before) that drove investment decisions by a broad set of both public and private agents.13 Second, the significance of the HOLC maps may have been further solidified by other mutually reinforcing practices, such as the Federal Housing Administration’s reluctance to provide mortgage insurance within majority racial and ethnic minority communities, zoning laws that aimed to exclude affordable housing within suburban communities, and other types of restrictions on new developments within suburban areas.13,26,27 Lastly, the resulting broad divestment decisions based on these beliefs and practices may in turn have affected the longer-term exposure of community residents to increased social risk factors (including housing instability, food insecurity, transport needs, economic needs, and safety), lower human capital (via reduced educational resources and/or opportunities), and lower health care resources (via limited access to health care facilities and providers). These are all mechanisms that have been linked to adverse health outcomes and premature mortality.28,29,30,31,32
To address long-standing health and mortality disparities, policymakers may wish to identify policies that directly target the potential mechanisms via which historical structural inequities may continue to affect present-day outcomes. Although there are several potential policies, such as minimum wage regulations, place-based investment initiatives, and health insurance expansion, further research is needed to help assess the potential relative efficacy of these (and other) policies.
Study Limitations
There are several study limitations. First, this study used an observational study design. As such, results should be interpreted as associations rather than as causal effects. With this noted, the study design took several steps to ameliorate bias concerns: (1) within the main analyses, the study adjusted for covariates and place-specific latent confounders (using city fixed effects); (2) the study used recent methods that were appropriate for mortality modeling, using censored administrative death records data, and that overcame the attenuation bias that occurs within standard regression modeling approaches; and (3) within additional subsample sensitivity analyses, the study found that the main results persisted even when analyses were limited to subpopulations that a priori might be more resistant to HOLC grading practices.
Second, administrative death records data, such as those used in this study, have technical limitations pertaining to truncated mortality coverage windows and incomplete observation of survivors. With regard to these limitations, the methods used within this study were designed to accommodate these features of the data and additional sensitivity analyses also found qualitatively similar results when alternative data with a broader mortality coverage window were used.
Third, since enumeration districts in 1940 do not perfectly map into historic HOLC maps, this presents a challenge concerning how to best construct a HOLC exposure measure at the enumeration district level. Although imperfect, additional sensitivity checks found that models using alternative (categorical) exposure measure definitions also resulted in qualitatively similar results.
Fourth, study data captured individuals residing within 30 of the most populous cities in 1940. As such, results may not be generalizable to areas beyond larger urban cities.
Conclusions
This study found that a 1-unit decrease of the HOLC score (eg, changing the grade of a neighborhood from A to B) was associated with an 8% increased later-in-life risk of death and a corresponding −0.49 years of reduced life expectancy at the age of 65 years.
Although these findings suggest that structural inequitable policies (such as historic redlining) may reduce individuals’ life expectancy, recent work has warned against fully attributing these associations to the actions of HOLC. In line with these recommendations, it appears that study associations may best be interpreted as the product of a much broader set of sociopolitical forces (including the actions and practices of both public and private stakeholders) that were all effectively encoded within HOLC maps.
eMethods. Additional Approach Details
eAppendix 1. Additional Descriptives and Results for Main Analytic Sample
eFigure 1. Age at Death Distributions by HOLC Grade
eAppendix 2. Robustness Check: Alternative Sub-Samples
eTable 1. Summary Statistics for Full Sample and HOLC Grade Subsamples with the Addition of Stratifying Variables
eTable 2. Summary Statistics for Full Sample and Robustness Check Subsamples
eTable 3. Additional Sensitivity Analyses Using Alternative Subsamples Based on Race, Labor Force Participation and Ownership of Property. Results Report Hazard Ratio Estimates for Each Subsample
eTable 4. Additional Sensitivity Analyses Using Alternative Subsamples Based on Race, Labor Force Participation and Ownership of Property. Results Report Life Expectancy Gap Estimates at 55, 65, and 75 Years of Age for Each Subsample
eAppendix 3. Robustness Check: Alternative Administrative Records Data
eTable 5. Summary Statistics for DMF Data
eTable 6. Hazard Ratio Estimates for DMF Data
eTable 7. Life Expectancy Gap Estimates at 55, 65, and 75 Years of Age for DMF Data
eAppendix 4. Robustness Check: Alternative Exposure Measures
eTable 8. Life Expectancy Gap Estimates – Alternative Exposure Measures
Data Sharing Statement
References
- 1.Luo J, Hendryx M, Wang F. Mortality disparities between Black and White Americans mediated by income and health behaviors. SSM Popul Health. 2021;17:101019. doi: 10.1016/j.ssmph.2021.101019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Agarwal S, Wade AN, Mbanya JC, et al. The role of structural racism and geographical inequity in diabetes outcomes. Lancet. 2023;402(10397):235-249. doi: 10.1016/S0140-6736(23)00909-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Devakumar D, Selvarajah S, Abubakar I, et al. Racism, xenophobia, discrimination, and the determination of health. Lancet. 2022;400(10368):2097-2108. doi: 10.1016/S0140-6736(22)01972-9 [DOI] [PubMed] [Google Scholar]
- 4.Egede LE, Walker RJ, Campbell JA, Linde S. Historic redlining and impact of structural racism on diabetes prevalence in a nationally representative sample of U.S. adults. Diabetes Care. 2024;47(6):964-969. doi: 10.2337/dc23-2184 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Egede LE, Walker RJ. Structural racism, social risk factors, and Covid-19—a dangerous convergence for Black Americans. N Engl J Med. 2020;383(12):e77. doi: 10.1056/NEJMp2023616 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Bailey ZD, Krieger N, Agénor M, Graves J, Linos N, Bassett MT. Structural racism and health inequities in the USA: evidence and interventions. Lancet. 2017;389(10077):1453-1463. doi: 10.1016/S0140-6736(17)30569-X [DOI] [PubMed] [Google Scholar]
- 7.Paradies Y, Ben J, Denson N, et al. Racism as a determinant of health: a systematic review and meta-analysis. PLoS One. 2015;10(9):e0138511. doi: 10.1371/journal.pone.0138511 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Phelan JC, Link BG. Is racism a fundamental cause of inequalities in health? Annu Rev Sociol. 2015;41. doi: 10.1146/annurev-soc-073014-112305 [DOI] [Google Scholar]
- 9.Williams DR, Mohammed SA. Racism and health I: pathways and scientific evidence. Am Behav Sci. 2013;57(8). doi: 10.1177/0002764213487340 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Egede LE, Campbell JA, Walker RJ, Linde S. Structural racism as an upstream social determinant of diabetes outcomes: a scoping review. Diabetes Care. 2023;46(4):667-677. doi: 10.2337/dci22-0044 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lee EK, Donley G, Ciesielski TH, et al. Health outcomes in redlined versus non-redlined neighborhoods: a systematic review and meta-analysis. Soc Sci Med. 2022;294:114696. doi: 10.1016/j.socscimed.2021.114696 [DOI] [PubMed] [Google Scholar]
- 12.Linde S, Walker RJ, Campbell JA, Egede LE. Historic residential redlining and present-day diabetes mortality and years of life lost: the persistence of structural racism. Diabetes Care. 2022;45(8):1772-1778. doi: 10.2337/dc21-2563 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Swope CB, Hernández D, Cushing LJ. The relationship of historical redlining with present-day neighborhood environmental and health outcomes: a scoping review and conceptual model. J Urban Health. 2022;99(6):959-983. doi: 10.1007/s11524-022-00665-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Aaronson D, Hartley D, Mazumder B. The effects of the 1930s HOLC “redlining” maps. Am Econ J Econ Policy. 2021;13(4):355-392. doi: 10.1257/pol.20190414 [DOI] [Google Scholar]
- 15.Richardson J, Mitchell B, Edlebi J, Meier H, Lynch E. The lasting impact of historic “redlining” on neighborhood health: higher prevalence of COVID-19 risk factors. Published online 2020. Accessed August 13, 2024. https://pacdc.org/2017/wp-content/uploads/2020/09/HOLC-and-public-health-affects-FINAL3.pdf
- 16.Rothstein R. The Color of Law: A Forgotten History of How Our Government Segregated America. Liveright; 2017. [Google Scholar]
- 17.Nelson RK, Winling L, Marciano R, Connolly N. Mapping inequality. Published online 2021. Accessed December 9, 2023. https://dsl.richmond.edu/panorama/redlining/#loc=5/39.1/-94.58
- 18.Logan JR, Zhang W. Developing GIS maps for US cities in 1930 and 1940. In: The Routledge Companion to Spatial History. 2018. doi: 10.4324/9781315099781 [DOI] [Google Scholar]
- 19.Logan JR, Minca E, Bellman B, Kisch A. From side street to ghetto: understanding the rising levels and changing spatial pattern of segregation, 1900-1940. City Community. 2023. doi: 10.1177/15356841231188968 [DOI] [Google Scholar]
- 20.Ruggles S, Fitch CA, Goeken R, et al. IPUMS Ancestry Full Count Data: version 3.0. Published online 2021. Accessed August 13, 2024. https://www.ipums.org/projects/ipums-usa/d014.v3.0
- 21.Goldstein RJ, Alexander M, Breen C, et al. CenSoc-Numident. Harvard Dataverse.
- 22.Goldstein JR, Alexander M, Breen C, et al. CenSoc-DMF. Harvard Dataverse.
- 23.Linde S, Walker RJ, Campbell JA, Egede LE. Historic residential redlining and present-day social determinants of health, home evictions, and food insecurity within US neighborhoods. J Gen Intern Med. 2023;38(15):3321-3328. doi: 10.1007/s11606-023-08258-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Goldstein JR, Osborne M, Atherwood S, Breen CF. Mortality modeling of partially observed cohorts using administrative death records. Popul Res Policy Rev. 2023;42(3). doi: 10.1007/s11113-023-09785-z [DOI] [Google Scholar]
- 25.Breen CF, Goldstein JR. Berkeley Unified Numident Mortality Database: public administrative records for individual-level mortality research. Demogr Res. 2022;47. doi: 10.4054/DemRes.2022.47.5 [DOI] [Google Scholar]
- 26.Gioielli R. The tyranny of the map: rethinking redlining. Published online 2022. Accessed April 26, 2024. https://themetropole.blog/2022/11/03/the-tyranny-of-the-map-rethinking-redlining
- 27.Fishback P, Rose J, Snowden KA, Storrs T. New evidence on redlining by federal housing programs in the 1930s. J Urban Econ. 2024;141. doi: 10.1016/j.jue.2022.103462 [DOI] [Google Scholar]
- 28.Chetty R, Stepner M, Abraham S, et al. The association between income and life expectancy in the United States, 2001-2014. JAMA. 2016;315(16):1750-1766. doi: 10.1001/jama.2016.4226 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Beatty K, Egen O, Dreyzehner J, Wykoff R. Poverty and health in Tennessee. South Med J. 2020;113(1):1-7. doi: 10.14423/SMJ.0000000000001055 [DOI] [PubMed] [Google Scholar]
- 30.Walker RJ, Garacci E, Dawson AZ, Williams JS, Ozieh M, Egede LE. Trends in food insecurity in the United States from 2011-2017: disparities by age, sex, race/ethnicity, and income. Popul Health Manag. 2021;24(4):496-501. doi: 10.1089/pop.2020.0123 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Borgonovi F, Pokropek A. Education and self-reported health: evidence from 23 countries on the role of years of schooling, cognitive skills and social capital. PLoS One. 2016;11(2):e0149716. doi: 10.1371/journal.pone.0149716 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Carroll C. Impeding access or promoting efficiency? effects of rural hospital closure on the cost and quality of care. Published online June 25, 2019. Accessed August 13, 2024. https://www.semanticscholar.org/paper/Impeding-Access-or-Promoting-Efficiency-Effects-of-Carroll/27800f7217e237f8b13828527bff73562212760c
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Citations
- Goldstein RJ, Alexander M, Breen C, et al. CenSoc-Numident. Harvard Dataverse.
- Goldstein JR, Alexander M, Breen C, et al. CenSoc-DMF. Harvard Dataverse.
Supplementary Materials
eMethods. Additional Approach Details
eAppendix 1. Additional Descriptives and Results for Main Analytic Sample
eFigure 1. Age at Death Distributions by HOLC Grade
eAppendix 2. Robustness Check: Alternative Sub-Samples
eTable 1. Summary Statistics for Full Sample and HOLC Grade Subsamples with the Addition of Stratifying Variables
eTable 2. Summary Statistics for Full Sample and Robustness Check Subsamples
eTable 3. Additional Sensitivity Analyses Using Alternative Subsamples Based on Race, Labor Force Participation and Ownership of Property. Results Report Hazard Ratio Estimates for Each Subsample
eTable 4. Additional Sensitivity Analyses Using Alternative Subsamples Based on Race, Labor Force Participation and Ownership of Property. Results Report Life Expectancy Gap Estimates at 55, 65, and 75 Years of Age for Each Subsample
eAppendix 3. Robustness Check: Alternative Administrative Records Data
eTable 5. Summary Statistics for DMF Data
eTable 6. Hazard Ratio Estimates for DMF Data
eTable 7. Life Expectancy Gap Estimates at 55, 65, and 75 Years of Age for DMF Data
eAppendix 4. Robustness Check: Alternative Exposure Measures
eTable 8. Life Expectancy Gap Estimates – Alternative Exposure Measures
Data Sharing Statement