Layering income, environment, and health outcomes across census-tract neighborhoods to locate — and target — geographic health inequity
Every health-equity map starts with a boundary choice. Analysts typically use U.S. Census tracts — geographic units of roughly 1,200–8,000 residents drawn to be relatively homogeneous — as the "neighborhood" unit, because that is the finest resolution at which the American Community Survey (ACS) and CDC PLACES release reliable estimates. Before any data layer is applied, the grid is simply population: where do people live, and how densely?
Zip codes were designed for mail routing, not demography — they vary wildly in population (200 to 100,000+) and often straddle wealthy and poor blocks. Named "neighborhoods" have no consistent legal boundary and change over time.
Census tracts solve both problems: the Census Bureau draws them explicitly to contain a statistically stable, relatively homogeneous population (typically 1,200–8,000 people), redrawing boundaries only once per decade. This stability is what lets researchers track the same geography across the 1930s Home Owners' Loan Corporation (HOLC) maps, the 2000 Census, and 2024 CDC PLACES estimates — enabling multi-decade disparity studies.
Smaller than tracts, census block groups (600–3,000 people) offer even finer resolution but carry wider margins of error in survey-based estimates, so most disparity dashboards default to the tract level as the sweet spot between resolution and statistical reliability.
Before layering any socioeconomic or health variable, density itself is informative. Dense low-rise tracts near a central business district often signal older, historically disinvested housing stock; low-density peripheral tracts often signal newer, higher-income suburban development — a pattern with roots in mid-20th-century zoning and highway construction that physically separated the two.
Density also matters methodologically: sparser tracts produce noisier survey-based estimates (larger confidence intervals), which is why CDC PLACES uses small-area model-based estimation — borrowing statistical strength from demographically similar tracts — rather than raw survey averages, to stabilize numbers in thinly-populated areas.
It is tempting to jump straight to a disparity map, but analysts who skip the neutral baseline risk anchoring their read of the data to whichever layer they see first. Establishing the plain population grid up front — the same 104 tracts that will carry every subsequent layer — lets a viewer notice, on their own, how tightly the socioeconomic, environmental, and health layers end up correlating with each other once they are added.
This simulator keeps the same underlying 104-tract city across all five stages (only the coloring changes) so the geographic clustering you see later is not an artifact of re-randomizing the map — it is the same neighborhoods, viewed through five different data lenses.
When income and educational attainment are painted onto the grid, the pattern rarely looks random — it looks clustered, often in shapes that trace back to explicit mid-20th-century housing policy. The 1930s Home Owners' Loan Corporation (HOLC) graded neighborhoods A ("best") through D ("hazardous," colored red on official maps — the origin of the term "redlining"), systematically denying federally-backed mortgages to Black and immigrant communities graded D. Those boundaries are still visible in wealth, housing quality, and health data nine decades later.
HOLC "residential security" maps were never intended as health documents — they were mortgage-risk assessments used by banks and the Federal Housing Administration to decide who could get a home loan. Grade-D neighborhoods were starved of mortgage credit for decades, which suppressed home-ownership rates, property investment, and generational wealth accumulation precisely in the areas that needed it most.
The compounding effect took decades to fully register: disinvested housing stock aged without renovation capital, commercial corridors lost investment, and public infrastructure (parks, transit, schools) was built preferentially in graded-A and B areas. By the time CDC PLACES and ACS data exist to measure current outcomes, the historical credit decision has become a present-day socioeconomic geography.
Household income and educational attainment are the two socioeconomic variables most tightly correlated with downstream health outcomes in the literature, but they rarely act alone. Low-income tracts disproportionately overlap with: fewer employer-sponsored insurance plans, higher housing-cost burden (rent >30% of income), lower homeownership rates, and — critically for this simulator — worse environmental and food-access conditions, which is why the next layer tends to mirror this one so closely.
The built and natural environment is where socioeconomic disadvantage becomes physically embodied. Tree canopy and park access reduce heat-island exposure and support physical activity; grocery-store proximity determines whether "healthy food" is a five-minute walk or a bus transfer away; and industrial zoning, highway proximity, and diesel truck routes concentrate air-pollution exposure — and all three tend to sort along the same historic boundaries as income.
Formerly redlined neighborhoods measure, on average, several degrees Celsius hotter than grade-A neighborhoods in the same city during summer heat waves — a documented "urban heat island" effect driven directly by lower tree canopy, more impervious surface (asphalt, unshaded rooftops), and fewer parks. Heat is not a cosmetic difference: it drives excess cardiovascular and heat-stroke mortality, disproportionately among elderly residents without reliable home cooling.
USDA and municipal urban-forestry data consistently show 15–30 percentage-point canopy-cover gaps between the highest- and lowest-income quartile tracts within the same city — a gap traceable to decades of differential public and private tree-planting investment.
The USDA Food Access Research Atlas classifies a tract as "low access" if a significant share of residents live more than one mile (urban) or ten miles (rural) from a supermarket. Low-access, low-income tracts are colloquially called "food deserts" — though many researchers prefer "food apartheid" to emphasize that the pattern reflects policy and disinvestment decisions rather than a natural or accidental gap.
Corner stores and fast-food outlets often fill the retail vacuum, offering calorie-dense, nutrient-poor options at higher relative prices than a full-service grocery store — a structural contributor to the chronic-disease layer mapped next.
The EPA's EJScreen tool combines 13 environmental indicators — PM2.5, ozone, diesel particulate matter, proximity to Superfund and hazardous-waste sites, traffic proximity, and more — into a single environmental-justice index, explicitly designed to surface tracts facing both high pollution burden and high socioeconomic vulnerability simultaneously. Highway placement in the 1950s–60s Interstate era frequently ran directly through — and physically bisected — the same neighborhoods graded "hazardous" by HOLC two decades earlier, layering fresh pollution exposure onto pre-existing disinvestment.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Tree canopy / green space | Municipal LiDAR, NLCD land cover | % tree/park cover per tract, heat-island proxy | Cools tracts, supports activity, lowers stress |
| Food access | USDA Food Access Research Atlas | Distance to nearest supermarket vs. income | Enables affordable, nutrient-dense diets |
| Air pollution burden | EPA EJScreen, PM2.5 monitors | Particulate matter, diesel exposure, traffic density | Lower burden = fewer respiratory/cardiac events |
| Industrial / hazard proximity | EPA Superfund, TRI facility registry | Distance-weighted count of hazardous sites | Distance reduces toxic exposure risk |
CDC PLACES (Population Level Analysis and Community Estimates) produces small-area model-based estimates of chronic disease prevalence, health behaviors, and clinical preventive-service use for every U.S. census tract — even where no local survey exists — by combining Behavioral Risk Factor Surveillance System (BRFSS) responses with tract-level demographic and socioeconomic predictors. Overlaying these outcomes on the grid typically produces a near mirror image of the socioeconomic and environmental layers already mapped.
PLACES uses multilevel regression and poststratification (MRP): a statistical technique that models the relationship between individual BRFSS survey responses and demographic/geographic predictors, then applies that model to the full tract-level population using Census demographic composition. This lets PLACES publish credible diabetes, hypertension, obesity, and COPD prevalence estimates for tracts that were never directly surveyed — the small-area equivalent of the same model-based approach used for state and county estimates, just at far finer resolution.
Outputs are cross-validated against direct survey estimates where available and against clinical claims data, and are published as open, downloadable tract-level datasets — the primary raw material for most public-facing health-equity mapping tools built since 2020.
One of the most cited illustrations of neighborhood health disparity is the "subway/metro life expectancy gradient": researchers at Virginia Commonwealth University's Center on Society and Health mapped life expectancy by census tract along transit lines in Chicago, Washington D.C., New Orleans, Richmond, and other cities, finding that life expectancy could fall by roughly a year for every mile — sometimes as much as 20–30 years across a metro area — traveling from a high-advantage tract to a low-advantage tract just a few transit stops away.
The finding reframes health disparity from an abstract statistic into something legible on a literal subway map: "your zip code matters more than your genetic code" became the project's widely repeated tagline.
Mapping disparity is only useful if it changes where resources go. The final stage computes a disparity score for every tract — how much worse its health outcome is than its socioeconomic and environmental profile alone would predict — flags the tracts furthest below expectation, and simulates how additional intervention investment (community health workers, mobile clinics, clinic siting, green infrastructure grants) can be routed to the worst-off areas first.
A disparity score asks a specific counterfactual question: given this tract's socioeconomic and environmental profile, how much better should its health outcomes be than they actually are? A tract with low income AND poor outcomes is "expected" — disadvantage predicting disadvantage. A tract with low income and disproportionately worse outcomes than similar tracts, or a tract where outcomes lag well behind an otherwise average socioeconomic profile, is flagged as a priority: something beyond the modeled predictors — a missing clinic, a contamination source, an access barrier — may be driving excess burden.
The Socioeconomic Weight slider in this simulator controls how much the flagging algorithm leans on the socioeconomic/environmental profile (predicted disadvantage) versus raw health-outcome deficit alone — mirroring a real methodological choice health departments make when designing equity dashboards.
Several production tools apply this same layered-map logic at scale: the CDC/ATSDR Social Vulnerability Index (SVI) combines 16 census variables across four domains (socioeconomic status, household composition, minority/language status, housing/transportation) into a single percentile score per tract, widely used for disaster preparedness and public-health resource allocation. PolicyMap and the Robert Wood Johnson Foundation's County Health Rankings & Roadmaps layer similar indicators for county- and tract-level planning. The EPA's EJScreen and the White House Climate and Economic Justice Screening Tool (CEJST) apply the same method to environmental and climate burden for federal grant-targeting decisions (e.g., the Justice40 Initiative, which committed 40% of certain federal climate/infrastructure investment benefits to disadvantaged communities).
Once flagged, tracts become the geographic unit for targeted intervention: nonprofit hospitals' IRS-mandated Community Health Needs Assessments (CHNAs) use tract-level disparity data to direct required community-benefit spending; Medicaid Accountable Care Organizations increasingly use similar geo-targeting to deploy community health workers and mobile clinics; municipal capital-improvement budgets use canopy and heat-island maps to prioritize tree planting; and federal place-based programs (Promise Zones, Opportunity Zones, Justice40) use composite disparity/vulnerability indices to direct billions in infrastructure and economic-development funding toward the tracts modeled here as "flagged."
The intervention-investment coverage metric in this simulator is a simplified stand-in for that real allocation process: as investment increases, coverage extends first to the tracts with the highest computed disparity score — a "worst-first" targeting rule common to most public equity-funding formulas.