Linking environmental injustices in Detroit, MI to institutional racial segregation through historical federal redlining

The study investigates whether and how historical federal redlining, as formalized in the 1930s HOLC maps, is linked to present-day environmental injustices in Detroit, Michigan. HOLC’s racially biased grading system (A–D) institutionalized segregation, devalued neighborhoods with Black and immigrant residents, and influenced siting of industrial and transportation infrastructure. The research questions are: (1) Are current environmental exposures higher in historically redlined neighborhoods compared to non-redlined areas? (2) Which environmental exposures most pervasively characterize historically redlined neighborhoods today? The purpose is to inform remedial policies by identifying priority environmental hazards tied to structural racism, with a particular focus on transportation-related air and noise pollution. The study’s importance lies in directly connecting a historical, institutional practice to contemporary environmental disparities in a highly racialized city context.

Prior work has demonstrated associations between historical redlining and present-day environmental and health disparities, including higher intra-urban heat, less greenspace, more brownfields, oil and gas well siting, and elevated air pollution. Health outcomes linked to redlining include higher risks of cardiovascular disease, asthma, preterm birth, and certain cancers, as well as poorer mental and physical health. In Detroit, environmental racism has been documented in relation to industrialization, discriminatory housing practices, and cumulative pollution burdens (air pollution, proximity to landfills and Superfund sites, and toxic releases). However, few studies focus specifically on Detroit or examine a broad suite of environmental exposures simultaneously. Existing findings for Detroit are mixed regarding hazardous facilities, and there has been limited exploration of noise as an environmental justice issue. This study addresses these gaps by assessing multiple EPA-relevant pollutants alongside transportation noise to identify the environmental markers most tied to historical redlining in Detroit.

Design: Ecological, place-based analysis using 1939 HOLC-defined neighborhood boundaries for Detroit. Data sources: (1) HOLC shapefile from the University of Richmond’s Mapping Inequality Project (238 neighborhoods; 233 included after boundary issues). (2) EPA EJScreen v1.0 environmental and demographic indicators at neighborhood level via overlay within HOLC boundaries. Environmental indicators included: air toxics cancer risk (2014), respiratory hazard index (2014), diesel PM (µg/m³; 2014), PM2.5 (µg/m³; 2016), ozone (ppb; 2016), traffic proximity and volume (2017), lead paint indicator (% pre-1960 housing; 2013–2017), proximity counts within 5 km for RMP sites (2019), hazardous waste facilities (2019), Superfund sites (2019), and a wastewater discharge indicator (2017). Demographic indicators (2013–2017 ACS): % low-income, % minority (non-white, non-Hispanic), % less than high school, % linguistically isolated, % under 5, % over 64; plus EPA’s Dindex (average % low-income and % minority). (3) DOT National Transportation Noise Map: 24-h road traffic noise Leq(24h) in dBA modeled via FHWA TNM 2.3. Using R v4.0.2, point estimates were assigned to HOLC neighborhoods to compute the percentage of neighborhood area with road noise >70 dBA (hazardous noise indicator). Exposure definition: HOLC grade D = historically redlined; grades A/B/C grouped as non-redlined for primary analyses. Statistical analysis: Descriptive summaries by HOLC grade. Simple linear regression to estimate percent differences in environmental indicators between redlined and non-redlined neighborhoods (log-transforming right-skewed indicators), with alpha=0.05. Boosted classification tree (BCT) analysis (Bernoulli distribution; 500 iterations; shrinkage 0.01; tree complexity 2) to identify the most important predictors of a neighborhood being historically redlined using current environmental and demographic indicators. Highly correlated predictors (r>0.7) were removed (traffic volume dropped due to correlation with % road noise >70 dBA; respiratory hazard index dropped due to correlation with diesel PM; Dindex substituted for highly correlated demographic measures). Lead paint indicator was excluded from BCT due to inverse directionality with redlining. Variable importance and partial dependence plots were used to interpret predictors. Sensitivity analyses compared D- vs C-graded neighborhoods only, repeating regressions and BCT.

• Sample: 233 neighborhoods; 61 (26.2%) were historically redlined (D-grade). Demographics: higher % without high school diploma (22.1% vs 13.7%), higher % low-income (61.6% vs 43.7%), and higher % minority (74.1% vs 55.3%) in redlined vs non-redlined neighborhoods; minimal differences for % under 5, % over 64, or linguistic isolation. - Environmental disparities (redlined vs non-redlined): • Air toxics cancer risk: +4.4% (95% CI: 2.9–6.6%). • Respiratory hazard index: +3.9% (95% CI: 2.1–5.6%). • Diesel PM: +12.1% (95% CI: 7.2–17.1%). • PM2.5: +0.5% (95% CI: 0–1%) (not statistically significant). • Ozone: no significant difference. • Traffic volume: +32.2% (95% CI: 3.3–69.3%). • Hazardous road noise (% area >70 dBA): +65.7% (95% CI: 8.6–152.8%). • Proximity within 5 km: hazardous waste sites 1.7× (95% CI: 1.4–2.1); RMP sites 2.0× (95% CI: 1.5–2.7); Superfund and wastewater discharge indicators: no significant differences. • Lead paint indicator (% pre-1960 housing): −13.6% (95% CI: −19.6 to −7.6%). - BCT (D vs A/B/C): Most important predictors of redlining status: proximity to RMP sites (17.9% importance), % road noise >70 dBA (16.1%), diesel PM (15.2%), % without high school diploma (9.2%), and air toxics cancer risk (7.5%). Partial dependence plots showed positive, sigmoidal associations: higher values corresponded to higher redlining probability. - Sensitivity (D vs C only): Disparities persisted and were directionally similar; proximity to hazardous waste and RMP sites, air toxics cancer risk, respiratory hazard index, % road noise >70 dBA, and diesel PM remained significantly higher; traffic volume difference attenuated and was no longer statistically significant. BCT top predictors: RMP site proximity (16.4%), diesel PM (15.9%), % road noise >70 dBA (14.2%); air toxics cancer risk fell below threshold; Dindex emerged as important (7.9%).

Findings indicate a sustained, disproportionate environmental burden in historically redlined Detroit neighborhoods, aligning with patterns of structural racism that concentrated pollution sources and transportation infrastructure in marginalized communities. Key exposures elevated in redlined areas—proximity to RMP sites, hazardous road noise, diesel PM, and air toxics cancer risk—suggest compounded risks, particularly for cancer and respiratory outcomes. The road noise and diesel findings reflect historical and ongoing highway construction and freight routes disproportionately traversing redlined neighborhoods, contributing to both noise and air pollution. The association with hazardous waste proximity echoes long-standing environmental justice patterns in facility siting. The inverse association with the lead paint indicator likely reflects Detroit’s housing demolition patterns and makes this metric a poor proxy for lead exposure in this context. Sensitivity analyses comparing D to C neighborhoods, which were often industrial by design, show disparities persist beyond general industrialization, supporting the role of targeted environmental racism. Comparisons to literature largely corroborate observed patterns (e.g., higher diesel PM, less greenspace, higher air toxics risk in D-grade areas) and add novel evidence on noise and RMP proximity as prominent markers of historical redlining impacts. These results emphasize the need for targeted interventions in transportation and industrial risk management to remediate legacy harms.

Historical redlining in Detroit is associated with elevated present-day environmental hazards, especially proximity to RMP facilities, hazardous road noise, diesel particulate matter, and increased air toxics cancer risk. Priority interventions should focus on transportation-related air and noise pollution, particularly diesel sources, through policy (e.g., anti-idling rules, greenspace requirements) and engineering controls (e.g., diesel PM filters, roadway barrier walls, fleet electrification, building air filtration). Future research should investigate individual-level health outcomes among residents of historically redlined neighborhoods and longitudinal mobility effects to elucidate causal pathways. However, policy action to mitigate identified environmental injustices should proceed without delay.

• Exposure scope limited to EPA EJScreen indicators and DOT NTNM road noise; other pollutants (e.g., heavy metals, pesticides) were not assessed and could affect rankings. - EJScreen provides modeled point estimates subject to measurement error; temporal mismatches exist across indicators (2013–2019). - Ecological design limits causal inference; individual-level data were not analyzed. - Findings may not generalize beyond Detroit due to city-specific histories and appraisal biases across HOLC maps. - Potential residual confounding by demographic and land-use factors despite inclusion of demographic covariates and correlation screening in BCT. - Grouping A/B/C as non-redlined could mask heterogeneity; addressed via D vs C sensitivity analysis but residual differences may remain.