Obesity is a significant public health concern that is socially patterned and influenced by a complex interplay of social, economic, and environmental factors, which are deeply intertwined with structural access to resources. As defined by a high body mass index, obesity is associated with an increased risk of all-cause mortality in the US [14]. In addition, it increases the risk of developing several chronic conditions, including kidney disease, hypertension, diabetes, cardiovascular disease, osteoarthritis, stroke, and certain preventable cancers [15–19]. Approximately 41.9% of adults have obesity, with 9.2% having severe obesity [20]. Obesity imposes a significant financial burden to US healthcare systems, with annual obesity-related medical expenditures estimated at $170 billion [21].

The risk of obesity or a high body mass index has previously been associated with various SDOH, including neighborhood socioeconomic disadvantages, income, and racialized status. Research has shown that the highest prevalence of obesity is among Black and Hispanic populations and those with lower socioeconomic status (SES) [22–25]. Disparities in obesity among racialized groups are marked; however, only limited literature has investigated the role of structural racism in creating these disparities [26–31]. Structural racism may contribute to obesity risk through a lack of investment in or disenfranchisement of racially segregated areas that have primarily Black or racially minoritized residents, by reinforcing neighborhoods with fewer health-promoting and more obesogenic factors [13, 32, 33]. Wealth inequality has also been linked to obesity, with individuals in lower-wealth quintiles having a higher risk of developing obesity than those in wealthier quintiles [34]. There is a high theoretical potential for redlining to explain disparities in obesity through social and economic pathways, including wealth accumulation and neighborhood economic deprivation.

Research investigating the interplay between structurally racist policies, SDOH, and obesity outcomes is scarce [27, 35]. Although some studies have shown that individuals with lower wealth have a higher prevalence of obesity [36–38], even fewer have measured wealth accumulation over multiple generations [39]. There is a gap in the literature that applies generational or life-course approaches to understanding the complex relationship between obesity and wealth, as well as the impact of structural racism on wealth and health [39, 40].

Research indicates a possible connection between structural racism and wealth, which may be linked to obesogenic environments and home values [26, 41]. Drewnowski et al. found that areas near crime, liquor stores, and fast-food stores were associated with lower property values [42]. Many families in redlined neighborhoods may have faced obstacles in building wealth through homeownership due to barriers created by redlining to obtain low-interest loans. Home equity is a significant source of wealth for many Americans, accounting for over a quarter of the portfolio assets of middle-class Americans [43]. Structural racism is also hypothesized to decrease household wealth for racialized minorities because of lower property values in segregated areas [44, 45], which may contribute to a lack of health-promoting amenities, such as safe walkable areas, green spaces, and healthy food access [13, 26, 46–50]. Greenspace, which includes tree canopy coverage, significantly contributes to physical activity, and thus the maintenance of a healthy weight, as this environmental resource commonly encourages residents to engage in outdoor activities [13, 46, 51]. Evidence supports the notion that structural racism and neighborhood segregation contribute to the racial wealth gap. However, no study has used redlining policies as a specific measure of structural racism to quantify their contribution to the gap in generational wealth accumulation [52, 53].

In the 1930s, President Roosevelt’s New Deal included the creation of the Federal Housing Administration (FHA). The FHA sponsored a federal agency called Home Owners’ Loan Corporation (HOLC) [54], which developed a discriminatory system to appraise homes and assess neighborhoods across 239 of America’s largest cities using maps to determine perceived credit risk level for home mortgages based on neighborhood characteristics, including demographic composition [55]. The grading scale ranged from the following: A (green — “best” deemed the lowest credit risk), B (blue — “still desirable”), C (yellow — “definitely declining”), and D (red — “hazardous” deemed the highest credit risk) [54–59]. The racial and ethnic composition of a neighborhood played a significant role in determining the grade an area received, as historical accounts revealed that HOLC graders remarked when “subversive racial elements” were present or increasing in a graded area [60]. Notably, neighborhoods with a higher proportion of racialized minorities, specifically “Negros” (i.e., Blacks), were often labeled as “hazardous” and given a D grade or colored red (hence the term redlining) [58]. The deployment of HOLC maps for assessing mortgage risk has been outlawed since the mid-1970s [54, 61], but the legacy of this practice is still evident in the persistent residential segregation and long-run decline of once redlined neighborhoods throughout the US [55, 58, 62].

We leverage the geographical nature of HOLC’s assigned credit security ratings (i.e., red, yellow, blue, and green) as a manifestation of neighborhood-level structural racism. These designations allowed us to evaluate the long-term effects of the redlining. Research has shown that HOLC grading is associated with lower home prices in neighborhoods that were previously redlined, indicating a potential disinvestment in these areas [55, 58, 62, 63]. Furthermore, redlining has been shown to strengthen segregation in many neighborhoods, thereby solidifying racial residential segregation as places previously redlined have become fairly immutable in terms of racial composition over time [55, 58, 62, 63]. The HOLC security maps, endorsed by the federal government through the Federal Housing Authority (FHA), provide a unique opportunity for a natural experiment.

Racial wealth inequality and health disparities highlight the need to assess the causal role of structural racism on both wealth accumulation over multiple generations, and health risk, such as BMI. Our study addresses a crucial gap in the literature by identifying the degree to which redlining is implicated in producing wealth and obesity disparities among racialized minorities over multiple generations. Therefore, this study adopts a temporal causal pathway framework wherein exposure is experienced by the first generation (hereafter called grandparents) and the outcome is measured in the third generation (hereafter called grandchildren). HOLC credit ratings, commonly known as redlining, had a profound impact on mortgage rates by systematically denying loans or offering them at higher (or rather predatory) rates to residents of neighborhoods deemed risky, primarily due to the race or ethnicity of a particular neighborhood. Government policies, originating at the local level, enforced racial zoning ordinances that isolated White families in all-white urban areas, aligning with the discriminatory objectives of redlining [64]. These ordinances aimed to block lower-income African Americans from middle-class White neighborhoods and impede middle-class African Americans from purchasing homes there [64]. Zoning practices, distinct from redlining but aligned with its discriminatory goals, further entrenched segregation by ensuring that many colored families were ineligible for FHA-insured mortgages, thus perpetuating racial disparities at both local and federal levels. There is evidence that this discriminatory practice led to suppressed property values in redlined areas, reinforced neighborhood segregation, and contributed to the divestment of Black neighborhoods [10, 61, 65, 66]. Consequently, the inability to access equitable and fair loan credit and diminished property values due to redlining appraisals potentially hindered wealth accumulation for marginalized communities and ultimately exacerbated racialized wealth disparities [67].

We posit that structural racism acts as an exposure that accumulates risk clustered across multiple generations through family links, which affects access to social and economic health resources and influences inequities in the risk of diseases, particularly BMI.

This study aimed to investigate the long-term effects of redlining on intergenerational outcomes related to wealth and health indicators, particularly BMI. To achieve this, a geographical regression discontinuity (GRD) design was employed to determine whether HOLC-defined discontinuities in perceived lending credit risk levels affect intergenerational wealth and health outcomes. To undertake this study, we utilized the genealogical design of the Panel Study of Income Dynamics (PSID). First, a sample of grandparents (first generation of PSID) for whom the location of their family home was recorded in 1968 was identified. The “treatment” status was assigned by overlaying digitized HOLC category boundaries onto these census blocks and classifying the families as “treated” or exposed to redlining if their census block fell within a D category (red) and as “comparison” if their census block fell within the C category (yellow). The distance from the census block to the nearest D-C HOLC boundary line was calculated and used as the running variable in the continuity-based regression discontinuity analysis.

Furthermore, we carried out falsification tests to evaluate the plausibility of our assumptions, the validity of our regression discontinuity design, and the robustness of our primary results. Specifically, we implemented three tests: (1) the balance of predetermined covariates, (2) the density of observations across the score, and (3) the sensitivity of observations near the cutoff (all described below). Additionally, we conducted a sensitivity analysis by examining our primary outcome(s) wealth adjusted for family size and BMI using the highest value observed. The details of these analyses are available in the online appendix (Tables S1 and S4).

We used plots to graphically depict discontinuity in our overall data [103, 107]. The horizontal x-axis corresponds to the score “distance(m),” which signifies treatment assignment, while the vertical or y-axis represents the outcome (average wealth or mean BMI). In an RD plot, a sudden and significant change in the outcome as the variable crosses the cutoff point (which in this case is equal to zero) signifies a pattern in the plot, or discontinuity, as to whether treatment has had an effect on the outcome [103, 105, 107]. The plots are composed of a global polynomial fit to show a smooth approximation of the regression functions and the local means are constructed with mimicking variance using quantile-spaced bins [103, 105, 122]. The quantile-spaced bins ensured consistent observation counts within each bin. The quantile-bin method adapts the bin length based on the data density along the score, resulting in more observations near and fewer away from the cutoff point [105].

Geographical and statistical analyses were performed using RStudio and the RD design was implemented using the rdrobust package [123–125].

Our primary analysis involved a sample of 237 grandchildren, who were descendants of 121 grandparents. As shown in Table 1, the sample statistics for both grandparent and grandchild participants were categorized by the HOLC graded classification. Among grandparents identified as heads of households in the PSID, similarities in characteristics exist between those living in HOLC grade C (yellowlined) and HOLC D (redlined) areas. These commonalities include mean age (C 39 years versus D 40 years), the proportion of females (C 48% versus D 43%), the percentage of those in marital or cohabitating relationships (C 50% vs. D 54%), individuals who grew up with parents of low socioeconomic status (C 60% versus D 56%), mean home value (C $1300 ± $59,800 versus D $1300 ± $9600), and mean labor income (C $4900 ± $2800 versus D $4600 ± $3100). However, a substantial difference emerged in the racial composition of the yellowlined and redlined areas among grandparent residents. Redlined areas have a higher concentration of Blacks (C 52% versus D 67%) and other racialized ethnic groups (C 2% versus D 8%), coupled with lower education levels (C 45% versus D 23%) than the yellowlined areas.

Compared to grandchildren whose grandparents resided in yellowlined regions, grandchildren whose grandparents lived in redlined areas exhibited a lower likelihood of attaining a college degree (C 35% vs. D 21%) and a higher likelihood of not completing high school (C 8% vs. D 23%) (Table 1). Furthermore, grandchildren with grandparents from redlined areas tend to have lower mean family income (C $71,000 ± $63,000 vs. D $48,000 ± $39,000) and wealth (C $82,000 ± $190,000 vs. D $53,000 ± $230,000), but a higher average BMI (C 26 kg/m² ± 4.2 kg/m² vs. D 28 kg/m² ± 6.1 kg/m²). However, the proportion of grandchildren with financially well-off parents was similar in both yellowlined (27%) and redlined areas (31%).

Figure 2 displays RD plots for the primary outcome, wealth. A noticeable discontinuity emerges at the threshold for wealth outcome. Figure 3 presents RD plots for mean BMI outcomes. A small discontinuity arises at the threshold when examining the sample variance along the score. In general, when comparing the outcomes of grandchildren in redlined and yellowlined areas, the RD plots reveal an abrupt change in both average wealth and mean BMI.

Table 2 details the findings on average household wealth for the study. The results from the models show a substantial, statistically significant reduction in household wealth across generations of grandchildren whose grandparents lived in redlined areas compared with those in yellowlined areas. The unadjusted model (1a) indicates that having a grandparent who lived in a redlined area versus a yellowlined area is associated with a lower family wealth of $96,100 (95% CI_rbc – $243,710, – $23,539). The adjusted models (2a and 3a), which include age, calendar year, and gender, yielded similar results with narrower confidence intervals (Table 2). The fully adjusted model(4a), which also included race and ethnicity, revealed a similar association, but lower in magnitude, indicating a persistent association between grandparent residence in a redlined area and their grandchildren having lower household wealth by – $35,419 (95% CI_rbc – $37,423, – $7,615) compared to yellowlined area grandchildren. Further wealth analyses, considering family size adjustments, revealed a similar pattern (Table S1). Additional race- and ethnic-stratified analyses are available in the online appendix (Table S2).

Results of the local linear regression models with the mean BMI as the continuous outcome are presented in Table 3. Overall, when examining the relationship between HOLC grade location for grandparents and the mean BMI of grandchildren, we observed that the mean BMI was notably elevated in our treatment group — grandchildren with a grandparent residing in a redlined region, in contrast to the comparison group — grandchildren with a grandparent living in a yellowlined region. The unadjusted model 1b reveals a higher mean, but not significant, BMI (β = 5.16 kg/m²; 95% CI_rbc – 2.54 kg/m², 9.81 kg/m²) for redlined grandchildren versus yellowlined grandchildren. The associations in models 2b and 3b are positive, yet not statistically significant. The fully adjusted model(4b) includes covariates age, year, gender, and race/ethnicity and reveals a substantial effect size that is much greater than the observed effect size of other models (β = 7.46 kg/m²; 95% CI_rbc – 4.00 kg/m², 16.60 kg/m²). This suggests a consistent association between grandchildren having a grandparent residing in a redlined neighborhood and an overall higher mean BMI than those with a grandparent residing in a yellowlined neighborhood. Race- and ethnicity-based stratified and highest BMI analyses are available in the online supplementary material.

Our study, which employs a quasi-experimental approach, aligns with existing research on the racialized wealth gap. Previous research has shown that the racial wealth gap cannot be attributed to differences in saving rates, educational attainment, income, or labor opportunities among racialized groups but is intricately linked to structural racism [127–129]. To shed light on this, we leveraged the HOLC mortgage loan security maps to investigate the effects of codified neighborhood-level structural racism on wealth. We capitalize on the longitudinal nature of the PSID to capture genealogical and cumulative wealth data, a robust approach supported by the literature for studying the racialized wealth gap [52]. While our analysis included individuals of all races whose grandparents had lived in redlined versus yellowlined neighborhoods, previous literature has demonstrated that Black families were substantially more likely to have lived in neighborhoods graded red by HOLC. Furthermore, the difference in wealth between redlined and yellowlined grandchildren demonstrates that redlining likely contributes to the racialized wealth gap [58, 62]. In our fully adjusted model (4a), which incorporated race as a precision variable, we observed a smaller disparity in average household wealth between redlined and yellowlined grandchildren. However, it is unclear from our data how including race as an additional adjustment affected our effect size. Nonetheless, our study establishes a plausible causal relationship between discriminatory policies in the US and the manifestation of the racialized intergenerational wealth gap, making this a significant contribution to the literature.

Killewald et al. (2017) highlight several methodological concerns when studying wealth inequality and accumulation, including the unexplained differences in wealth compared to income levels and the use of transformation when analyzing highly skewed measures such as wealth [52]. Our research addresses these concerns by utilizing a historical policy lever to implement a natural experiment that examines the effect of wealth accumulation over multiple generations, accounting for debt and zero net worth by averaging wealth over multiple years [130], and using an intergenerational framework to analyze the effects of historical redlining practices on the accumulation of household wealth. Research on intergenerational wealth indicates that wealth acquired by grandchildren through early life investments, such as access to advantageous neighborhoods, homeownership, and other forms of social and cultural capital, may account for the racial wealth gap as much as direct gifts and bequests [52, 131]. Our findings demonstrate the solidifying effects of historical disadvantages on contemporary inequities in intergenerational wealth accumulation and social mobility between Black and White populations, ultimately strengthening the argument that social origins and historical structural racism have lasting effects on wealth outcomes across multiple generations.

Previous research has found that wealth is inversely associated with obesity [34]. Our study design provides an opportunity to delve deeper into the relationship between wealth and BMI disparity. By examining the impact of generational exposure to structural racism on BMI outcomes, we discovered a greater disparity in BMI outcomes for grandchildren whose grandparents were exposed to redlining compared to those with grandparents in yellowlined areas. Additionally, studies have shown a positive relationship between housing segregation, socioeconomic status (SES), and BMI, particularly among Black women [29, 132–134]. These findings are consistent with the outcomes of our primary and racially stratified models, albeit our results fall short of achieving statistical significance. Our research suggests that race-based segregation of neighborhoods, operating over multiple generations, may shape the social and economic environment of communities, particularly those composed mostly of Blacks and other historically disadvantaged groups, leading to neighborhoods that are disproportionately exposed to obesogenic environments. We maintain that race-based bifurcation of neighborhoods in the US, perpetuated and cemented by discriminatory HOLC maps, has contributed to systematic disinvestment in health-promoting neighborhood amenities and the emergence of obesogenic environments. We speculate that these factors likely contributed to poorer health outcomes, as indicated by higher BMI in the current study and the documented worse physical and mental health outcomes for racially segregated minority neighborhood, as identified by Lynch et al. (2021) [10]. Our work advances research on the causal link between BMI and neighborhood-level structural racism. Since redlining was based on the credit risk assigned to neighborhoods, this policy likely had far-reaching impacts on the economic and social trajectories of neighborhood resources.

This study has several limitations that must be considered. First, the regression discontinuity design utilized in this study was limited to individuals who were closer to the border of the HOLC red and yellow thresholds, which restricted our ability to generalize our findings to those who were farther away. Additionally, the use of fine-scaled geocoded data, such as address or longitude/latitude, is ideally suited for regression discontinuity design, and we were unable to utilize such data because of technical and privacy limitations [98]. Furthermore, we did not have a precise location for the grandparents and used aggregated areal units that may not accurately reflect the spatial variation in the units being measured [98]. To address these geographical limitations, we utilized the most granular spatial census unit available, census blocks, to identify the proximal location of the first-generation PSID participants in the HOLC areas.

Another limitation is the limited sample for our racially stratified analyses; we were unable to obtain disaggregated data for all racialized groups, particularly those classified as “Other” within the PSID, which limited our ability to understand the impact of redlining on these groups. Additionally, the density of our data for participants on yellowlined side, particularly for the BMI analysis, was sparse.

Finally, we acknowledge the use of body mass index as a measure of obesity is suboptimal, although it is the only available measure of obesity currently captured by the PSID [135, 136]. Additionally, since we lacked adequate data on information that may influence BMI such as medication and pregnancy, we used an average of BMI for outcome. The use of body mass index (BMI) as a measure of adiposity and its association with morbidity and mortality has been extensively studied, particularly regarding its application across different racial and ethnic groups. Studies, such as those conducted by Seo and Torabi (2006) and Jackson et al. (2014), have shed light on racial disparities in BMI and its relationship with health outcomes, especially among Black individuals [137, 138]. While Seo and Torabi’s (2006) findings revealed a higher mean BMI among non-Hispanic Black women compared to non-Hispanic White women, Jackson et al. (2014) suggested a weaker association between BMI and mortality risk in Black individuals, particularly Black women, compared to their White counterparts. Several factors may contribute to this observed difference, including the possibility that BMI may be a weaker indicator of adiposity in Black populations [138]. Moreover, structural racism and disparities in access to healthcare and resources may exacerbate the impact of obesity-related conditions on minority populations [139]. The study by Park et al. (2012) further emphasizes the complexity of BMI-mortality associations across ethnic groups, indicating variations in the strength of these associations and underscoring the importance of considering early adulthood BMI and ethnic-specific pathways to obesity-related diseases [140]. Furthermore, researchers have highlighted the differential trends in obesity-related cardiovascular mortality by race, sex, and place of residence, underscoring the need for targeted structural interventions to address disparities [141].

Overall, while BMI remains a widely used measure of adiposity, its usefulness across diverse racial and ethnic populations warrants careful consideration, taking into account variations in body composition, healthcare access, and societal factors. Despite these variations, there is strong evidence that generally high BMI (rather than directly measured adiposity) is highly correlated with a multitude of chronic diseases and is highly correlated with directly measured adiposity among non-elderly adults [80, 81, 142, 143]. Future research should continue to explore the complex interplay between BMI, race, and health outcomes to inform more tailored structural approaches to obesity prevention and healthy weight management in minority communities.