Structural Racism and Pedestrian Safety: Measuring the Association Between Historical Redlining and Contemporary Pedestrian Fatalities Across the United States, 2010‒2019

Abstract

Objectives. To examine the association between historical redlining and contemporary pedestrian fatalities across the United States.

Methods. We analyzed 2010–2019 traffic fatality data, obtained from the Fatality Analysis Reporting System, for all US pedestrian fatalities linked by location of crash to 1930s Home Owners’ Loan Corporation (HOLC) grades and current sociodemographic factors at the census tract level. We applied generalized estimating equation models to assess the relationship between the count of pedestrian fatalities and redlining.

Results. In an adjusted multivariable analysis, tracts graded D (“Hazardous”) had a 2.60 (95% confidence interval = 2.26, 2.99) incidence rate ratio (per residential population) of pedestrian fatalities compared with tracts graded A (“Best”). We found a significant dose‒response relationship: as grades worsened from A to D, rates of pedestrian fatalities increased.

Conclusions. Historical redlining policy, initiated in the 1930s, has an impact on present-day transportation inequities in the United States.

Public Health Implications. To reduce transportation inequities, understanding how structurally racist policies, past and present, have an impact on community-level investments in transportation and health is crucial. (Am J Public Health. 2023;113(4):420–428. https://doi.org/10.2105/AJPH.2022.307192)

Transportation is an important social determinant of health that affects the ability of people to move efficiently and safely through public and private spaces. Active transportation—specifically walking, cycling, and rolling—has direct and indirect impacts on health at both individual and community levels.¹ Injuries and deaths among road users, especially those walking, continue to be a significant public health problem. In the past decade, pedestrian deaths have risen by 54% while all other traffic deaths have increased by 13%.² Low-income communities and communities of color bear a disproportionate burden of pedestrian injuries and fatalities, with Native/Indigenous and Black pedestrians being especially overrepresented.^3–8 Moreover, roadway designs that enable speeding and discourage walking are more likely to be in areas that experience high rates of pedestrian fatalities, which are often lower-income, Black, or Hispanic/Latinx communities.^9,10 However, research focuses on identifying factors that increase or decrease risk rather than characterizing the policies that created and facilitated these unsafe built environments.

There is a current shift in public health, both research and fields of practice, that seeks to understand the ways in which structural racism fundamentally causes the health inequities we see today in the United States. Bailey et al. define structural racism as

the totality of ways in which societies foster racial discrimination, through mutually reinforcing inequitable systems (in housing, education, employment, earnings, benefits, credit, media, health care, criminal justice, and so on) that in turn reinforce discriminatory beliefs, values, and distribution of resources, which together affect the risk of adverse health outcomes.^11(p1454)

Inequities in residential housing practices measured by residential racial segregation or historical redlining are a common indicator of structural racism.¹² “Redlining” is a term that refers to a federally sponsored policy that was initiated in the United States in the 1930s. The government-sponsored Home Owners’ Loan Corporation (HOLC), created as part of President Franklin D. Roosevelt’s New Deal, made loans to new homeowners by refinancing mortgages at low-interest rates. HOLC used color-coded and letter-graded maps to group neighborhoods into financial risk and lending categories.¹³ Areas color-coded green (“A” or “Best”) and blue (“B” or “Still Desirable”) were predominantly White and were systematically approved for privately and publicly guaranteed home loans. However, neighborhoods color-coded yellow (“C” or “Definitely Declining”) and red (“D” or “Hazardous”), which were populated by Black people and immigrants, were denied homeownership loans, ultimately impacting generational wealth and limiting community-level investments. Redlining legalized discrimination in housing and systematized structural racism on a national scale.¹⁴ For an example of the language and maps created by HOLC, see Figure 1 and Figure A (available as a supplement to the online version of this article at https://ajph.org).¹⁵

FIGURE 1—

1930s Home Owners’ Loan Corporation Map for Durham, North Carolina, and Grade-Related Descriptions: University of Richmond Mapping Inequality

Source. Nelson et al.¹⁵

Redlining is associated with a wide variety of contemporary adverse health outcomes, both at the individual and community level. These health outcomes include (but are not limited to) smoking, infant mortality, life expectancy, and firearm violence.^16–24 Krieger et al. found that census tracts assigned worse HOLC grades in New York City had an elevated risk of adverse maternal health outcomes.¹⁹ Research on violence and historical redlining also found that neighborhoods that were redlined (graded D) in Louisville, Kentucky, had a greater incidence of gun violence compared with areas graded A, even after adjusting for census-level demographic factors.¹⁶ Moreover, redlining continues to be associated with racial segregation, poverty, and income inequality.^23,25 Mitchell et al. found that areas graded “Hazardous” by redlining maps remain areas with lower household incomes.²⁵ By tying the presence of Black people to low property values, negating the generational wealth of Black households, cementing the racial wealth gap, and perpetuating disinvestment in segregated Black neighborhoods, redlining continues to adversely affect community health throughout the United States.

To date, potential connections between redlining and transportation-related health outcomes have not been explored. Previous research has focused on how specific demographic, environmental, and behavioral factors are associated with pedestrian injuries and fatalities and may contribute to observed disparities in safety outcomes.^7,8,26 However, there is a gap in knowledge surrounding the impacts of inequitable neighborhood-level investments created by historical structurally racist policies, such as redlining and transportation-related disinvestments. As the field of traffic safety shifts from focusing primarily on changing individual-level factors to transforming structures that reinforce inequities in the transportation systems, there will be a need to understand how policies have created these inequitable systems. To hypothesize these relationships, we used a conceptual model to identify potential pathways between redlining and contemporary, neighborhood-level inequities in pedestrian fatalities.

We sought to address a gap in transportation safety research by assessing the impact of historical policies on contemporary pedestrian safety outcomes. In this study, we aimed to assess the impact of historical redlining on pedestrian fatalities within the United States between 2010 and 2019. Given that redlining is a leading contributing factor to economic disinvestment in neighborhoods, which may lead to a lack of pedestrian infrastructure in redlined neighborhoods, we hypothesized that historical redlining is associated with pedestrian fatalities throughout the United States. Specifically, we hypothesized that areas impacted by redlining or classified as “Definitely Declining” or “Hazardous” (graded C or D, respectively) will have higher contemporary rates of pedestrian fatalities. To our knowledge, this is the first study that seeks to describe the relationship between historical redlining and present-day transportation-related health outcomes on a national scale.

METHODS

We obtained the geocoded locations of all US traffic-related pedestrian fatalities from 2010 to 2019 from the Fatality Analysis Reporting System from the National Highway Traffic Safety Administration.²⁷ Of the 53 407 pedestrian fatalities that occurred between 2010 and 2019, we omitted 412 from the analysis because of unusable latitude and longitude values that were unreported, reported as unknown, or not available. We mapped the remaining 52 995 usable geographic coordinates by using ArcMap version 10.8 (ESRI, Redlands, CA). We then aggregated the counts of the pedestrian fatalities at the census tract level by using the 2019 US TIGER/Line Shapefiles.

Determining Redlined Areas

Our main exposure of interest was historical redlining, measured by the color-coded, A–D grades illustrated in the 1930s HOLC maps (Figure 1). For our analyses, we obtained shapefiles of all original HOLC maps from the University of Richmond Mapping Inequality project.¹⁵ The original HOLC boundaries do not align spatially with current census tract boundaries. Like other studies, we used the area of overlap technique to assign HOLC grades to census tracts.^16,17,28 We overlayed redlining shapefiles with 2019 US census tracts to determine the number of intersections and areas of HOLC grades that fell within each census tract. We selected the HOLC grade with the largest area within the census tract boundary as the HOLC color-coded categorization for that census tract. We dropped tracts with less than 10% area overlap with HOLC grades or coded as “E” (uncharacterized) in the data set from the analysis. We completed all spatial processes in ArcMap version 10.8.

Covariates

We obtained covariates of interest including census tract‒level self-reported race, ethnicity, age, gender, poverty, education, and population density from the 2010–2015 (5-year) American Community Survey (ACS), based on categories and descriptions developed by the US Census.²⁹ These variables included the percentage of non-Hispanic Black/African American, Hispanic/Latinx, male, those older than 18 years, those older than 65 years, those in poverty aged 18 years or older, and those older than 25 years with at least a high school degree or completion of general education development. We regrouped all variables, except for population density, into discrete “high” and “low” categories using mean and median distributions. We included population density as a continuous variable.

Statistical Analysis

We used descriptive statistics to assess the relationship between historical redlining, pedestrian fatalities, and sociodemographic factors at the census tract level. Our dependent variable was the count of pedestrian fatalities per census tract; therefore, we used models that account for count distributions and clustering at the census tract. We assessed generalized estimating equation regression, with log function and exchangeable correlation, to model the relationship between historical redlining and counts of pedestrian fatalities. We calculated incidence rate ratios (IRRs) with 95% confidence intervals (CIs) using population as an offset. Because of the potential clustering of pedestrian fatalities within our study area, we assessed residual spatial dependence among census tracts by using Moran’s I statistic. We found that our model indicated a weak but significant spatial dependence (Moran’s I statistic = 0.06; P < .001).

First, we modeled the unadjusted relationship between historical redlining and pedestrian fatalities using grade A as the referent. Second, we assessed multivariable models, using the goodness-of-fit statistic, by adjusting for census-tract level age, gender, and population density. We did not include all sociodemographic factors because of concerns that these factors may be mediators and could lead to overadjustment in our models.³⁰ To determine if redlining exhibited a dose‒response effect for pedestrian fatalities across HOLC grades A to D, we performed the Kruskal‒Wallis test for trends. We performed all statistical analyses with SAS version 9.4 (SAS Institute, Cary, NC).

RESULTS

We included a total of 15 289 census tracts in our analysis. We excluded census tracts with less than a 10% overlay with 1930s HOLC areas (n = 2412), with HOLC grade E (n = 10), and with zero population according to the 2015 ACS (n = 96). Our final sample included 13 377 census tracts across 38 states and 202 cities. Table 1 shows the characteristics of our study population. Within areas impacted by redlining, a total of 9631 pedestrian fatalities occurred between 2010 and 2019 with a maximum count of pedestrian fatalities per census tract of 17 (mean = 0.72; SD = 1.2). Roughly 45.1% of our 13 377 census tracts were graded C (n = 6037) followed by D (28.5%; n = 3816), B (20.0%; n = 2671), and A (6.3%; n = 853).

TABLE 1—

Census Tract‒Level Sociodemographic Characteristics and Count of Pedestrian Fatalities (n = 13 377): United States, 2010–2019

Characteristics	Mean	Median	Min	Max	SD	IQR
Outcome: pedestrian fatalities, no.	0.7	0.0	0.0	17	1.7	1.0
Covariates—sociodemographic factorsa
Non-Hispanic Black, %	26.2	10.7	0.0	100	31.2	38.8
Hispanic/Latinx, %	21.5	10.2	0.0	100	25.3	26.5
Male, %	48.5	48.4	0.0	100	4.7	5.1
Aged > 18 y, %	78.1	77.9	37.1	100	7.9	9.7
Aged > 65 y, %	12.3	11.5	0.0	100	5.9	7.2
Poverty, %	21.0	18.5	0.0	100	13.4	16.5
Education for those aged ≥ 25 y, %	26.2	27.2	0.0	100	11.4	18.9
Population density per 1000	6.5	3.1	0.0	93.3	9.0	5.6

Note. IQR = interquartile range. Education includes those aged ≥ 25 years with at least a high school degree or completion of general education development. Poverty includes those aged ≥ 18 years; federal poverty level according to the US Census.

^a 2010–2015 American Community Survey.²⁹

The overall rate of pedestrian fatalities was 2.0 per 100 000 person-years, with tracts assigned worse HOLC grades having higher pedestrian fatality rates per 100 000 person-years: A = 1.1; B = 1.6; C = 1.9; and D = 2.6. In addition, sociodemographic factors also varied by HOLC grades. Tracts graded D had a higher percentage of people of color (non-Hispanic Black and Hispanic/Latinx) and poverty compared with other tracts (Figure 2).

FIGURE 2—

Census Tract‒Level Characteristics by Home Owners’ Loan Corporation (HOLC) Grade for (a) Non-Hispanic Black Persons, (b) Hispanic/Latinx Persons, and (c) Poverty: United States, 2010–2019

Note. Federal poverty level according to the US Census.

Source. 2010‒2015 American Community Survey.²⁵

Comparison of Models and Multivariable Analysis

Table 2 illustrates the relationship between HOLC grades, demographic factors, and pedestrian fatalities. In our multivariable analysis, we found a significant relationship between historical redlining and contemporary pedestrian fatalities. In model 1, the unadjusted IRRs for HOLC grade D was 2.33 (95% CI = 2.03, 2.60) times the rate of pedestrian fatalities compared with grade A (Table 2, model 1). Tracts graded C (IRR = 1.71; 95% CI = 1.50, 1.96) or B (IRR = 1.38; 95% CI = 1.20, 1.60) were also associated with a higher IRR for pedestrian fatalities compared with tracts graded A. The estimated IRR from tracts graded D to those graded A showed a significant dose‒response relationship (Kruskal‒Wallis test: P < .001). After adjusting for census tract‒level demographic factors, the relationship between HOLC grade designation and pedestrian fatalities remained. When compared with tracts graded A, census tracts graded D (IRR = 2.60; 95% CI = 2.26, 2.99), C (IRR = 1.84; 95% CI = 1.61, 2.11), or B (IRR = 1.49; 95% CI = 1.29, 1.72) were associated with higher incidence rates of pedestrian fatalities (Table 2, model 2).

TABLE 2—

Multivariable Associations of Pedestrian Fatalities With Historical Redlining, Census Tract‒Level Demographic Factors: Fatality Analysis Reporting System, United States, 2010–2019

	Model 1,a IRR (95% CI)	Model 2,b IRR (95% CI)	Model 3,c IRR (95% CI)
HOLC grade
A (green)—“Best”	1 (Ref)	1 (Ref)	1 (Ref)
B (blue)—“Still Desirable”	1.38 (1.20, 1.60)	1.49 (1.29, 1.72)	1.49 (1.29, 1.72)
C (yellow)—“Definitely Declining”	1.71 (1.50, 1.96)	1.84 (1.61, 2.11)	1.84 (1.61, 2.11)
D (red)—“Hazardous”	2.33 (2.03, 2.68)	2.59 (2.25, 2.99)	2.60 (2.26, 2.99)
Model goodness of fit QICd	2705.22	3077.49	3046.04

Note. CI = confidence interval; HOLC = Home Owners’ Loan Corporation; IRR = incidence rate ratio; QIC = quasi likelihood under independence model criterion. Full descriptions of HOLC grades are in Figure A, available as a supplement to the online version of this article at https://ajph.org.

^a Unadjusted, HOLC grade only.

^b Fully adjusted for gender, age > 18 years, age > 65 years, and population density per 1000.

^c Final model adjusted for age > 65 years and population density per 1000.

^d The lowest goodness-of-fit statistic determines the best model.

DISCUSSION

In this novel study, we found a significant relationship between structural racism via historical redlining and contemporary, neighborhood-level inequities in pedestrian fatalities across the United States. More specifically, we found that census tracts graded D (“Hazardous”) in the 1930s had an increased rate of present-day pedestrian fatalities compared with tracts graded A or “Best.” In addition, we found a significant dose‒response relationship from grades A to D, meaning that as grade color categorization worsened, pedestrian fatality rates increased. This finding adds to the current literature describing the impact of historical redlining on present-day neighborhood-level health inequities.^{11,16–18,21,22} Similar to other studies, after adjustment for present-day demographic factors, the effect of redlining remained significant in tracts graded D, C, and B as compared with tracts graded A or “Best.”^16,21,22

While redlining is not a perfect or sole measure of structural racism, the long-lasting and intergenerational effects of this structurally racist policy are undeniable. Similar to other research, our study showed that census tracts adversely impacted by historical redlining (graded D) continue to be areas with a higher percentage of populations that are impoverished and belong to communities of color.²⁵ Furthermore, research continues to show that areas with high poverty rates and reliance on public transit or active transportation (e.g., walking, rolling, and cycling) have an increased risk of pedestrian crashes and are often characterized by limited, unsafe, high-speed roadway infrastructure.^10,26

Our findings indicate that redlining, a policy first implemented in the 1930s, which led to inequitable investments in communities, continues to adversely affect neighborhood-level transportation outcomes today. This is particularly noteworthy, given that most modern-day transportation safety programs, such as Vision Zero and the Safe System approach, do not typically acknowledge the impacts of structurally racist policies and racial inequities related to neighborhood-level transportation investments. They also do not undertake concrete efforts to rectify structural inequities in transportation and land use plans and investments. Rather than using “colorblind” approaches to transportation safety (e.g., allocating funds equally to communities for pedestrian infrastructure, regardless of the history of structural racism), these programs should aim to intentionally and directly address the underlying drivers of inequities in transportation safety and forge cross-agency partnerships with professionals in housing, community development, or public health to develop multidisciplinary approaches to rectify transportation inequities.

This study adds to a growing body of literature that confirms that historical policies that led to intergenerational neighborhood disinvestment must be redressed to improve public health, including the reduction of pedestrian fatalities throughout the nation. Departments of transportation at the local, state, and national level play key roles in the distribution and utilization of funding for roadway infrastructure, especially for large-scale highway projects. However, pedestrian infrastructure, such as sidewalks, is often left to the responsibility of private developers and property owners.³¹ Therefore, lower-income neighborhoods often lack sidewalks or have poorly maintained sidewalks with limited connectivity. Structural racism has governed the trajectories of communities across the United States, creating multidecade place-based effects. These effects are often not acknowledged as a fundamental cause of transportation inequities. Individually focused behavioral countermeasures and siloed infrastructure projects cannot sufficiently address present-day inequities. Transportation researchers must understand and address upstream factors—like redlining—that continue to undermine positive population-level transportation outcomes.

To our knowledge, this is the first study that (1) connects redlining to transportation, a key social determinant of health, and (2) connects redlining to a transportation-related health outcome by collectively examining all known redlined cities across the United States. Contrasting with other studies that have focused on the identification of specific risk factors or countermeasures that might pertain to individual road users or roadway locations, this study focuses on root causes of inequities and quantifies the multigenerational effects on population-level health outcomes. Research focused on the effects of historical redlining beyond a single or multicity level has been limited; however, our study examines the national-level impacts of historical redlining and neighborhood health. Our study underscores the ubiquity of this harmful policy and its longstanding effects on the health and safety of communities affected. Currently, the transportation and health communities suffer from an overreliance on individual-based research and behavior-based approaches to injury prevention, known to have limited effectiveness on complex, population-level challenges. More population-level research is needed to illuminate the effects of structural racism; assess historical, political, and social contexts; and highlight opportunities for systemic interventions that offer redress and high-impact, population-level health benefits.

Limitations

In our analysis, our outcome only included fatalities and did not include nonfatal injuries, which are also a significant public health problem. Although pedestrian fatalities are relatively rare events, census tracts with few or no fatalities may have had numerous nonfatal events. Moreover, residential population may not be the ideal denominator for examining exposure for all pedestrian behaviors (e.g., commuting or recreation); however, nationwide walking exposure information is limited. In addition, given that our data were aggregated to the census tract level, our findings are impacted by ecological fallacy. However, our findings support a well-established understanding that historical redlining impacts neighborhood-level health and can ultimately be used to explain place-based exposures and transportation inequities. Our results may also be impacted by both the uncertain geographic context problem and the modifiable areal unit problem. Census tracts are arbitrary boundaries, and pedestrian fatalities and roadway networks do not follow tract borders. However, less than 1% of crash locations fell within a 0.1-mile radius of borders that cross tracts that were graded D and A.

Our data are also subject to potential misclassification because of the assignment of 1930s HOLC grades to current-day census tracts when using overlay techniques. Roughly 16% of census tracts had less than 10% overlay with HOLC polygons, and removal of these tracts had little impact on our associations. While smaller geographic units may offer better resolution for interpretation of neighborhoods, studies using alternative geographic boundaries (e.g., HOLC polygons) have found similar results to our findings; therefore, we predict our relationship would remain.^17,24

Furthermore, our analysis did not account for potential temporal changes in HOLC grades from the 1930s to the present day and did not measure changes to neighborhood-level trajectories in health and investment attributable to gentrification or displacement. However, like other studies, we found that redlined census tracts remain areas with higher percentages of populations that are impoverished and belong to communities of color.^16,19,23,25

Despite controlling for census tract‒level factors, there may still be unmeasured confounding effects, such as factors related to civic engagement, political power, or other processes that have an impact on policy decisions (e.g., voting rights). In addition, our analysis did not completely account for spatial dependence; therefore, to further address the relationship between historical redlining and transportation outcomes, studies should consider spatial analysis methods. Finally, our analysis focused on the national-level impacts of HOLC policies on present-day neighborhood-level pedestrian fatalities; therefore, findings cannot be used to determine the impacts of redlining on transportation safety within a specific city or state. Future studies may focus on smaller geographic areas to calibrate estimates and offer more context-specific interpretations.

Public Health Implications

Our study adds to the current literature by examining the legacy of structurally racist historical policies that perpetuated transportation inequities that can be observed today. Our study has a variety of implications for public health and transportation safety. Public health efforts that address traffic safety—such as Vision Zero and the Safe System approach—must focus on actions that modify the political, social, and built environments that result in inequitable transportation systems driven by structural racism. The US Department of Transportation Equity Action Plan identified equitable actions focused on empowering communities, expanding access, increasing resources, and wealth creation; however, more should be done to provide state and local guidance on best practices to measure and reduce inequities created by past and present policies and transportation investments including redlining.³² Findings from this study underscore the importance of historical context and the availability of novel data sources, such as redlining maps, to support more nuanced and equity-focused decisions in land use, planning, policymaking, and transportation engineering.

Lessons learned from past interventions are also essential to public health approaches. Place-based funding interventions, often implemented by housing or revenue-focused governmental agencies, have increased property values and investment in disinvested communities; however, they have also led to gentrification and the residential displacement of low-income populations.^14,33 Some federal place-based funding programs that have attempted to equitably distribute funding to redlined areas have been associated with increases in property value.¹⁴ However, these gains in neighborhood wealth were also associated with reductions in Black resident homeownership in areas formerly graded D.¹⁴ These types of interventions show the possible benefits of place-based investments and reveal the consequences of displacement. Transportation interventions that focus on place-based funding should seek to reduce injuries and fatalities while also minimizing the consequences of displacement by centering community governance in transportation decision-making.

Future research should focus on 3 areas to address injury prevention inequities at the population level for pedestrian fatalities. First, more research is needed that focuses on measuring structural racism as the main exposure that drives health inequities. Racism is a public health crisis and has significant impacts on health across generations; however, investigations into how structural racism is a leading cause of transportation inequities have been limited and should be a focal point for future research. Analyzing other indicators of structural racism, such as school and residential racial segregation, racially discriminatory enforcement policies and practices, and the construction of high-speed roadways through communities of color, is important to understand and address present-day transportation inequities.^11,34

Second, epidemiological research should focus on analyzing modifiable exposures, multidimensional pathways, and potential mediators that result in disparate outcomes for communities of color within the United States. Differences in specific built environment and roadway features, as well as impacts of gentrification and displacement, could be assessed in future studies to determine if relationships exist between the presence of these features, disinvestments tied to structurally racist policies like redlining, and present-day transportation outcomes.

Finally, research should focus on the intergenerational impacts of historical policies on neighborhood development that ultimately affect public health. Several pathways, at the community level—which include social, policy, and built environment factors—should be further explored to better understand how redlining is associated with pedestrian fatalities and to develop sustained policy solutions that focus on redress and make progress toward health and transportation equity.