Historical Neighborhood Redlining and Contemporary Environmental Racism

Abstract

To stabilize the housing market during the great depression, the government-sanctioned Home Owners’ Loan Corporation (HOLC) created color coded maps of nearly 200 United States cities according to lending risk. These maps were largely driven by racial segregation, with the worst graded neighborhoods colored in red, later termed redlined neighborhoods. We sought to investigate the association between historical redlining, and trends in environmental disparities across the US over the past few decades. We characterized environmental exposures including air pollutants (e.g., NO2 and fine particulate matter), vegetation, noise, and light at night, proximity hazardous emission sources (e.g., hazardous water facilities, wastewater discharge indicator) and other environmental and social indicators harnessed from various sources across HOLC graded neighborhoods and extrapolated census tracts (A [lowest risk neighborhoods] to D [highest risk neighborhoods]). Lower graded areas (C and D) had consistently higher exposures to worse environmental factors. Additionally, there were consistent relative disparities in the exposures to PM2.5 (1981–2018) and NO2 (2005–2019), without significant improvement in the gap compared with HOLC grade A neighborhoods. Our findings illustrate that historical redlining, a form of residential segregation largely based on racial discrimination is associated with environmental injustice over the past 2–4 decades.

Introduction

A substantial body of literature has documented links between racial segregation and various health outcomes in the United States.(Kramer and Hogue 2009) Although exact mechanisms are unclear, some studies have attributed this to socio-economic disadvantage, decreased access to care, chronic stress, and to environmental exposures.(Gaskin et al. 2012; Simons et al. 2018; Iceland and Wilkes 2006; Lee et al. 2022) Historical racial segregation practices, although outlawed, continue to shape current conditions.

In the 1930s, the government sanctioned Home Owners’ Loan Corporation (HOLC) was created to stabilize the housing market during the Great Depression. As a means to limit foreclosure risk, the HOLC created color-coded security maps of more than 200 United States cities according to the risk of investing in certain neighborhoods.(Bell 1986) Neighborhood grading was largely based on racial/ethnic composition, built environment, and housing conditions. The HOLC maps color-coded neighborhoods as follows: A (“best” or green), B (“still desirable” or blue), C (“definitely declining” or yellow) and D (“hazardous” or red). This approach resulted in neighborhood “redlining” and can serve as an indicator of intentional and protracted racially directed inequities in economic and social investments.

Redlining is not only associated with persistent segregation and economic inequality, but also disparities in public health indices, life expectancy and social vulnerability.(Lee et al. 2022) Given the known links between neighborhood level environmental exposures and health, we sought to investigate the association between redlining with historic and contemporary environmental exposures.

Methods

We characterized various environmental exposures across HOLC graded neighborhoods and their extrapolated census tract transformations. We obtained HOLC map shapefiles from the University of Richmond’s Inequality project.(Nelson et al 2022) The data provides a digitalized mapping of the historical HOLC neighborhood boundaries color coded with their respective grades. Each color-coded neighborhood was transformed into a colored polygon totalizing of 8,871 analyzable polygons falling into one of the four HOLC grades A, B, C, and D.

Fine particulate matter smaller than 2.5 mm in diameter (PM2.5) and Nitric Dioxide (NO2):

We compiled 38 years of annual PM2.5 data (1981 to 2018) and 15 years of annual NO2 data estimates (2005–2019) from the Atmospheric Composition Analysis Group.(Hammer et al. 2020; Meng et al. 2019; Cooper et al. 2022) This data combines information from satellite remote sensing, ground-based monitors, and chemical transport modeling to generate concentration estimates. PM2.5 and NO2 data was overlaid with HOLC polygons to generate HOLC graded neighborhood exposure estimates by using zonal statistics within QGIS v. 3.16.

Vegetation/Greenness, Noise, and light at night:

Vegetation was estimated from the normalized vegetation difference index (NDVI), an index calculating the difference in reflective value between red and infrared light to estimate green cover density.(Pettorelli 2013) It ranges from −1 (least green) to 1 (greenest). The estimates were generated using EROS Moderate Resolution Imaging Spectroradiometer (eMODIS) V6 data(Earth Resources Observation And Science (EROS) Center 2002) gathered from July 2018 at a 250m grid resolutions.

The 2018 National transportation noise map data files were downloaded from the Bureau of Transportation Statistics, which showcases annual average potential exposures to roads, aviation, and rail transportation noise.(“National Transportation Noise Map” n.d.) Light at night pollution data was obtained from Light Pollution Science and Technology Institute and reported in radiance.(Falchi et al., n.d.) PM2.5, NO2, vegetation, noise, and light pollution data were overlaid with HOLC polygons to generate exposure estimates for each of the HOLC graded neighborhoods by using zonal statistics within QGIS v. 3.16.

Census tract analysis:

Census tract scores were extrapolated from neighborhood scores through weighting by the percentage of land that a census tract overlaps with a graded neighborhood, generating a score as previously described (Richardson et al. 2020; Motairek et al. 2022). Census tracts with less than 20% land graded were excluded. The continuous score was then transformed back into one of the four HOLC grades according to an equal interval method of division: A (>1 and <1.75), B (>1.75 and <2.5), C (>2.5 and <3.25), and D (>3.25 and <4).

We categorized census tract-weighted exposures of variables collected from the Environmental Protection Agency’s Environmental Justice screening tool (“EJSCREEN: Environmental Justice Screening and Mapping Tool | US EPA”). We used the 2021 version of EJSCREEN that incorporates the most recently available environmental metrics: diesel particulate matter, air toxics respiratory hazard index, air toxics cancer risk, living in house built before 1960 (lead paint indicator), proximity to risk management plan facilities (RMP), proximity to national priority list sites (NPL), proximity to hazardous waste facilities, traffic proximity and volume, and proximity to underground storage tanks. We also included 4 socioeconomic variables available on EJSCREEN and compiled from the American Community Survey: Less than high school education, proportion with low income, unemployment rate, and median household income.

Statistical analysis:

After generating annual PM2.5 and NO2 concentrations of each HOLC graded neighborhood, we then calculated the average neighborhood exposure concentration within each HOLC grade for each year that the data was available. This was followed by the calculating the percent difference of mean level exposure of neighborhoods in HOLC grades B, C, and D from HOLC grade A for each year the data was available. Vegetation, noise, and light at night median and interquartile ranges (IQR) were generated for HOLC neighborhoods across the four HOLC grade (A, B, C, and D).

We also generated median and in IQR across the graded census tracts for the EPA-EJSCREEN variables, and the socioeconomic variables. Kruskal–Wallis one-way analysis of variance (ANOVA) was used to compare neighborhood and census tract-level exposure values across HOLC grades. Statistical analyses were performed using R version 4.1.3 open statistical software, and figures were generated via GraphPad Prism version 8.0.

Results:

A total of 8,871 HOLC-graded neighborhoods (polygons) were analyzed. (Figure 1) Out of these neighborhoods A-graded areas covered 1,040 (11.7%) neighborhoods, B-graded areas covered 2,332 (26.3%) neighborhoods, C-graded areas 3,381 (38.1%), and D-graded areas covered 2,118 (23.9%) neighborhoods.

Figure 1:

Continental United States maps showing the location of HOLC neighborhoods included in this analysis

PM2.5 and NO2 decreased in concentration across all four HOLC grades. HOLC grades C and D had consistently higher exposures of PM2.5 and NO2 than grades A and B over the years.

For PM2.5, the percent difference in exposure (representing relative disparity) of HOLC grades C and D decreased with time from around 6% for grades C and D to around 4%. That of grade B did not have a significant change which decreased the gap with grades C and D. (Figure 2 B)

Figure 2:

Percent differences in exposure to NO2 (A) and PM2.5 (B) in HOLC grades B, C, and D from grade A

For NO2, the percent difference was highest for HOLC grade C, followed by HOLC grades and D, and then B. The percent difference in NO2 remained constant over the years showing the persistence of relative disparities with HOLC grade A. (Figure 2 A)

There was a stepwise increase in exposure to transportation noise and light at night pollution and a stepwise decrease in NDVI across HOLC graded neighborhoods (A to D). (Figure 3)

Figure 3:

Median and interquartile ranges of three environmental exposures across HOLC graded neighborhoods. NDVI: Normalized Difference Vegetation Index; Noise: Transportation noise from roads, avation, and roads reported in Decibels (dBA). Light pollution: Light at night radiance reported in millicandela per square meter.

A total of 11,864 census tracts were included representing 842, 2,318, 4,976, and 3,728 census tracts falling in HOLC grades A, B, C, and D, respectively. HOLC grading was associated with a stepwise increment in census tract exposures for environmental indicators: diesel particulate matter, air toxics respiratory hazard index, air toxics cancer risk, proximity to risk management plan facilities (RMP), proximity to national priority list sites (NPL), proximity to hazardous waste facilities, traffic proximity and volume, and proximity to underground storage tanks. There was a stepwise decrement for pre-1960 housing proportion (Figure 4). For the socioeconomic variables, HOLC grading was associated with stepwise increment of proportion of individuals in the census tracts with less than high school education, with low income, and with unemployment. There was a stepwise decrease in median household income, P < 0.05 for all comparisons (Figure 5).

Figure 4:

Median and interquartile ranges of 10 environmental indicators in Unites States census tracts by HOLC grade. Abbreviations: HOLC, Home Owners’ Loan Corporation; PM, particulate matter; RMP, Risk Management Plan; NPL, National Priorities List.

Air toxics respiratory hazard index: ratio of exposure concentration to health-based reference concentration; Air toxics cancer risk: Lifetime cancer risk from inhalation of air toxics; Diesel PM: Diesel particulate matter level in air, μg/m3; Proximity to RMP Sites: Count of RMP (potential chemical accident management plan) facilities within 5 km (or nearest one beyond 5 km), each divided by distance in kilometers; Proximity to NPL sites: Count of proposed or listed NPL - also known as superfund - sites within 5 km (or nearest one beyond 5 km), each divided by distance in kilometers; Wastewater Discharge Indicator: Modeled Toxic Concentrations at stream segments within 500 meters, divided by distance in kilometers (km); Traffic Proximity and Volume: Average annual daily count of vehicles at major roads within 500 meters, divided by distance in meters; Proximity to Hazardous Waste Facilities: Count of hazardous waste facilities within 5 km (or nearest beyond 5 km), each divided by distance in kilometers. Pre-1960 Housing: Fraction of housing units built pre-1960, as indicator of potential lead paint exposure.

Figure 5:

Median and interquartile ranges of 4 socioeconomic indicators indicators in United States census tracts by HOLC grade. Less than highschool: Fraction of people age 25 or older in a census tract whose education is short of a high school diploma. Low income fraction: Fraction of a census tract’s population in households where the household income is less than or equal to twice the federal “poverty level”. Unemployment rate: Fraction of census tract’s population that did not have a job at all during the reporting period, made at least one specific active effort to find a job during the prior 4 weeks, and were available for work (unless temporarily ill). Median household income: median annual income of households living in a census tracts reported in United States Dollars.

Discussion

The statistical analysis reveals a significant association between historical redlining with an array of environmental exposures in the United States. Our data highlight that historical patterns of residential segregation, as indicated by redlining practices still have consequences to this day, contributing to persistent environmental injustice.

Few studies have investigated the association between redlining practices and environmental disparities. Lane et al investigated the effects of redlining on air pollution and showed an association between HOLC grading with PM2.5 and NO2 concentrations in 2010.(Lane et al. 2022) Our findings support and extend these findings elucidating trends of relative disparities of PM2.5 and NO2 over many years. Our study is unique in examining other vulnerabilities and in demonstrating how environmental injustice spans a multitude of socioenvironmental domains. With an estimated 6–9 million air pollution-related deaths globally, PM2.5 is the main air pollutant responsible for morbidity and mortality.(Murray et al. 2020) Despite reductions of PM2.5 nationally, relative disparities still exist. An analysis of over 65,000 census tracts found that census tracts that were most polluted in 1981 were still the most polluted in 2016.(Colmer et al. 2020) We highlight a similar trend for redlined areas across the nation, in which lower graded neighborhoods (HOLC grades C and D) continue to consistently have higher concentrations of PM2.5 and NO2 than their counterparts (HOLC A and B).

Today, many railroads and highways pass through historically redlined areas. A big proportion of highways were constructed well after 1930s or after the HOLC grading took place, highlighting a potential temporal trend between residential segregation and urban inequality.(Lane et al. 2022; Ananat 2011) The passage of the National Interstate and Defense Highways Act of 1956 resulted in the construction of highways and disruption of predominantly Black communities. Current transportation infrastructure preferentially targets predominantly Black communities displacing hundreds of thousands to provide access to white people while walling off minorities.(Lane et al. 2022; Wiese 2005; Archer, n.d.) Further, the erection of the highway system in black communities contributed to decades of poisonous environmental exposures closely related to vehicular emissions as well as adverse health sequelae. We reinforce these underpinnings in our analysis as we show that people living in redlined neighborhoods are more likely to experience proximity to traffic and higher exposures to transportation noise and transportation-related pollutants (NO2, Diesel PM, and PM2.5).

Road infrastructure is not the only constituent of the built environment compromised by redlining. Previous literature has shown that redlining is associated with various components of the built environment. One study showed that neighborhoods assigned worse HOLC grades in 1930s were still associated with reduced greenspace in 2010.(Nardone et al. 2010) Another study showed that redlined areas (HOLC grade D) had higher mean land surface temperatures than areas with more favorable ratings.(Wilson 2020) We build on previous literature and show that residents of redlined neighborhoods experience various adverse exposures inherent to residential areas compounded by the numerous hazardous facilities (hazardous waste facilities, wastewater discharge, NPL and RMP sites, and underground storage tanks) situated in C- and D-grade neighborhoods. In fact, redlined areas often adjoin areas coded as commercial and industrial with the HOLC Examiner’s notes sometimes referencing industrial nuisances as a factor in downgrading certain areas.(Nelson et al. 2022) This left an enduring impact on the health of inhabitants of these segregated areas that is akin to decades of sustained adverse exposures. We found a downward trend on pre-1960 housing, which may act a surrogate for lead exposure, with HOLC grading. This is probably due to poor housing conditions in redlined areas that made reconstruction necessary after the Fair Housing Act was enacted in 1968.

Causality between HOLC maps and lending practices may be hard to assert. Even though Michney and Winling found that segregation and race played an important role in impacting lending practices, HOLC gave loans to both White and Black borrowers during its “rescue phase” (Michney and Winling 2020). Moreover, Fishback et al. found that HOLC mapping did not have a strong influence on lending practices, and that the Federal Housing Administration (FHA) had their own methodology of preferential lending predating the emergence of these maps (Fishback et al. 2022). However, HOLC maps do indeed allow for the best available analysis of historical residential segregation practices immediately prior to World War II. Moreover, they present an archive of “hazardous” designated neighborhoods that would have been “redlined” by banks and the FHA in the post-war era. Finding links between redlined areas and current-day environmental racism sheds the light on how residential segregation impacted social and environmental inequality.

Aaronson et al found a causal correlation between growing up in historically redlined neighborhoods and socio-economic status including labor market outcomes, family structure, and incarceration.(Aaronson et al. 2021) Another study showed how firearm violence extends beyond the individual and may be linked to consequences that redlining had on the socioeconomic and demographic makeup of the redlined communities.(Poulson et al. 2021) In our study we show a trend of less employment, education, and income associated with HOLC grading. This goes with other studies showing that disinvestments in redlined neighborhoods may have led to significant place-based disadvantages with life-course altering consequences for children growing up generations later.

A significant body of literature has linked built environment with health outcomes.(Renalds, Smith, and Hale 2010; Motairek et al. 2022) In fact, disparities in adverse environmental exposures may lead in turn to widening disparities in various health outcomes. A few prior studies have investigated the association between redlining and health inequities.(Lee et al. 2022) However, our work may lay a foundations to explain some mechanisms responsible for health disparities that might exist between redlined and non-redlined neighborhoods.

Our study has a few limitations. First, most environmental exposure estimates in the study are modeled, and thus may be prone to estimation mischaracterization. Despite their modeled nature, these environmental outcomes have been widely used in the literature and are at many times the best available analysis of environmental exposures. Also, satellite estimates (i.e. air pollutants and NDVI) have moderate resolutions which may lead to some uncertainty around estimates. However, HOLC graded areas are relatively large compared to the resolution of grid scale minimizing the level of uncertainty. Further, there is no gold standard for defining “redlining”, and our methods may introduce some misclassification bias. Finally, in this national study we aggregate findings from various HOLC graded areas across different cities and regions in the United States. Findings may differ according to the region and the historic patterns of urban development. “Sunbelt” and western cities that had relatively smaller populations at the time HOLC maps followed by a rapid population increase in the decades after. The social and environmental impact of segregation in these cities are likely to be different than the mature “rust” belt” cities which have lost a considerable fraction of their populations.

Conclusions:

The legacy of historical redlining policies is still associated with various contemporary disparities. Our work expands the literature by highlighting the multifaceted associations between historical residential segregation and socioenvironmental attributes of society, many of which not previously investigated. Although causality between HOLC maps and lending practices may be hard to assert, HOLC maps provide a powerful tool to analyze a snapshot of the historical residential segregation directly prior to World War II. Our findings call for increased awareness of historical segregation policies that have sustained environmental injustice, advocating for a community-based approach targeting the most vulnerable areas assuring equity in the protection from environmental and health hazards. Further research on the mechanistic impact the environmental injustice in redlined communities on the sustained health disparities.

Table 1.

Characteristics of neighborhoods by HOLC grade

	HOLC grade A Median (IQR)	HOLC grade B Median (IQR)	HOLC grade C Median (IQR)	HOLC grade D Median (IQR)
Median household Income	89,616 (60,192–135,509)	61,576 (44,059–86,536)	50,670 (36,032–69,776)	40,814 (27,748–60,859)
Low income	0.16 (0.09–0.28)	0.30 (0.17–0.46)	0.41 (0.26–0.57)	0.51 (0.33–0.65)
Unemployment rate	0.037 (0.023–0.058)	0.05 (0.032–0.081)	0.06 (0.037–0.096)	0.07 (0.04–0.12)
Underground Storage Tanks	3.18 (1.36–6.57)	5.46 (2.5–11.93)	7.18 (3.37–14.95)	8.51 (3.78–17.87)
Less than high school education	0.04 (0.02–0.08)	0.09 (0.04–0.16)	0.15 (0.08–0.24)	0.19 (0.11–0.29)
NDVI	0.62 (0.51–0.7)	0.55 (0.44–0.64)	0.50 (0.37–0.60)	0.45 (0.32–0.56)
Noise	10.4 (6.03–24.93)	15.31 (8.11–40.99)	18.84 (9.66–47.68)	21.9 (10.03–48.75)
Light at night	3.73 (2.58–5.15)	4.11 (2.89–5.66)	4.53 (2.96–6.51)	4.85 (3.03–7.04)
Traffic Proximity and Volume	637 (319–1213)	737 (372–1418)	812 (402–1723)	1071 (502–2369)
Wastewater Discharge Indicator	2.0*10−3 (6.33*10−5 −3.96*10−2)	3.53*10−3 (4.6*10−4 −5.08*10−2)	6.39*10−3 (1.97*10−4 −1.04*10−1)	1.53*10−2 (5.46*10−4 −1.51*10−1)
Air Toxics Respiratory Hazard Index	0.4 (0.3–0.4)	0.4 (0.3–0.5)	0.4 (0.3–0.5)	0.4 (0.4–0.5)
Air Toxics Cancer Risk	30 (30–30)	30 (30–40)	30 (30–40)	30 (30–40)
Diesel PM	0.36 (0.29–0.45)	0.41 (0.32–0.58)	0.48 (0.34–0.69)	0.54 (0.38–0.81)
Proximity to NPL Sites	0.1 (0.06–0.16)	0.12 (0.07–0.19)	0.11 (0.06–0.19)	0.13 (0.06–0.27)
Pre-1960 Housing	0.74 (0.6–0.85)	0.76 (0.61–0.85)	0.7 (0.55–0.81)	0.62 (0.46–0.75)
Proximity to RMP Sites	0.52 (0.24–1.07)	0.635 (0.265–1.31)	0.78 (0.3–1.6)	1.28 (0.64–2.44)
Proximity to Hazardous Waste Facilities	2.26 (0.89–4.29)	3.33 (1.53–6.27)	4.18 (2.21–7.26)	5.3 (2.56–10.74)