Modeling how grocery access geography shapes diet quality, obesity, and diabetes risk across a neighborhood
The USDA Economic Research Service defines the food environment as the mix of retail food outlets available to a community — full-service supermarkets and grocery stores that stock fresh produce, meat, and dairy, versus fast-food restaurants, convenience stores, and dollar stores that stock predominantly shelf-stable, energy-dense, low-nutrient products. Retailer-density audits like the USDA Food Environment Atlas and the Modified Retail Food Environment Index (mRFEI) quantify this balance census tract by census tract, and the imbalance itself is the first measurable predictor of downstream diet quality.
A census tract qualifies as a food desert when it is simultaneously:
• Low-income: poverty rate ≥20%, or median family income ≤80% of the surrounding area median • Low-access: at least 500 people or 33% of the population live more than 1 mile (urban) or 10 miles (rural) from the nearest supermarket, supercenter, or large grocery store
The definition deliberately combines an economic constraint with a geographic one — distance alone does not create a food desert if households have reliable vehicle access and disposable income; poverty alone does not create one if a supermarket is a short walk away. It is the intersection of the two that removes the practical option of routine fresh-food shopping.
The USDA Food Access Research Atlas (updated periodically since 2011) geocodes every US census tract against this definition, producing the standard reference dataset used in nearly all subsequent food-access research.
A growing body of research argues that raw supermarket distance understates the problem. Cooksey-Stowers, Schwartz & Brownell (2017, Int. J. Environ. Res. Public Health) introduced the "food swamp" concept: areas where unhealthy food outlets (fast food, convenience, dollar stores) so heavily outnumber healthy ones that the ratio itself — not distance to any single store — better predicts obesity prevalence than the desert classification alone.
The Modified Retail Food Environment Index (mRFEI) operationalizes this by scoring each tract on the proportion of "healthy" versus "less healthy" retailers within a defined radius, producing a 0–100 index. Tracts scoring below 10 are considered severely imbalanced. This is the metric visualized in Stage 1 of the simulator: two supermarkets against seven fast-food and convenience outlets is a food-swamp pattern layered directly on top of a food-desert distance problem.
Food deserts and food swamps are related but distinct: a household can live within 1 mile of a supermarket and still be surrounded by ten fast-food outlets that make the supermarket the least convenient, least habitual option. Intervention design increasingly targets the ratio, not just the nearest-store distance.
Distance to a supermarket is not an abstract number — it is a recurring weekly cost measured in time, transportation, and opportunity. For a household without a reliable vehicle, a 1.2-mile trip to the nearest supermarket can mean a 40-minute bus transfer each way, carrying groceries on foot, or paying for a rideshare that erodes any savings from buying in bulk. The access radius drawn around each store in this simulator is the geographic core of the USDA low-access definition, and it is what separates households into two structurally different shopping realities.
Distance interacts multiplicatively with two other constraints common in low-income households: time poverty and vehicle access. Ver Ploeg et al. (USDA ERS, 2009) — the foundational federal report "Access to Affordable and Nutritious Food" — found that households without vehicles in low-access areas shop less frequently, buy smaller quantities per trip, and substitute toward shelf-stable, longer-lasting products that tolerate infrequent restocking. Fresh produce, with its short shelf life, is disproportionately squeezed out of this shopping pattern regardless of a household's underlying nutritional preferences.
The simulator models this by classifying each household as "in-access" or "in-desert" based on straight-line distance to the nearest full-service grocery store within the 1-mile radius — a simplification of the network-distance and transit-time calculations used in real GIS-based food-access research, but one that preserves the same structural divide.
Food access geography changes over time as retailers open and close. A single supermarket closure can push an entire tract across the low-access threshold overnight, while a new store opening can pull hundreds of households back within range. Longitudinal USDA tract reclassifications show meaningful churn between survey years — roughly 10–15% of low-access tracts change status between Food Access Research Atlas updates, driven mostly by retail entry and exit rather than population change.
This volatility is why access radius is treated as a dynamic, slider-controllable variable in Stage 5 of this simulator rather than a fixed baseline fact: neighborhood food access is a policy-responsive geography, not a permanent condition.
GIS-based access studies increasingly use network distance (actual walkable/drivable route length) rather than straight-line distance, which can understate real travel burden by 30–50% in areas with highways, rivers, or irregular street grids acting as barriers.
Reduced access to fresh produce does not immediately show up as a health outcome — it first shows up as a substitution pattern in the weekly shopping basket. Repeated national dietary surveys (NHANES) consistently find that adults in low-access, low-income tracts consume fewer daily servings of fruits and vegetables than adults in high-access tracts, even after adjusting for income alone. Distance and retail imbalance operate as an independent drag on diet quality, layered on top of — not merely a proxy for — economic constraint.
When fresh produce requires a special, costly trip while shelf-stable and fast food are available at every corner, households rationally shift the marginal purchase toward what is convenient — not because preferences change, but because the effective price of fresh food (time + transport + spoilage risk) rises relative to the effective price of processed alternatives. This is modeled in the simulator as each household's diet-quality gauge drifting toward a lower equilibrium when it falls outside the access radius, and holding near a higher equilibrium when inside it.
The drift is gradual by design: dietary substitution accumulates over weeks and months of routine shopping decisions, not in a single trip. This mirrors longitudinal survey data, where measurable declines in produce intake associated with a supermarket closure typically take 2–6 months to fully register.
Diet-quality decline is not uniform. Households with a vehicle, flexible work schedules, or strong social shopping networks (carpooling to a distant supermarket) can partly buffer the access gap. Households working multiple jobs, caring for children alone, or without reliable transportation show the steepest declines. The simulator represents this heterogeneity with per-household random variation around the access-driven target, so that the population-level average masks a real spread of individual outcomes — exactly as it does in observed cohort data.
A commonly cited RAND/NIH cohort finding: each additional mile to the nearest supermarket is associated with roughly a 0.03–0.05 fewer daily cup-equivalents of fruit and vegetable intake, a small per-mile effect that compounds across a population unevenly distributed across access distances.
Sustained low fruit and vegetable intake is a well-established upstream risk factor for excess weight gain and impaired glucose metabolism, operating through higher net caloric density, lower fiber intake, and displacement of nutrient-dense calories by processed alternatives. Population-level studies consistently find higher obesity and type 2 diabetes prevalence in food-desert and food-swamp tracts than in demographically matched high-access tracts — an association that strengthens the longer a household remains in a low-access environment.
Metabolic risk does not respond instantly to a bad month of shopping; it accumulates from sustained exposure, similar to other chronic-disease risk factors like smoking pack-years. The simulator models each household's obesity and diabetes risk as a slow integral of "diet deficit" over simulated time — the further and longer a household's diet quality sits below a healthy threshold, the faster its risk indicator climbs, visualized as an expanding red ring around at-risk households.
This mirrors the epidemiological reality: cross-sectional studies (a single snapshot) understate the food-desert effect, because the households most affected are the ones who have lived in the low-access environment longest. Longitudinal cohorts that track the same households over 5–10 years find substantially larger diet-attributable risk gaps than one-time surveys.
Food-desert residence correlates strongly with poverty, and poverty independently predicts obesity and diabetes through stress, food insecurity, and reduced healthcare access — so isolating the causal contribution of retail geography alone is genuinely difficult. The strongest causal evidence comes from natural experiments: supermarket openings, closures, and relocations that shift access for an otherwise stable population without those households choosing to move.
The consensus reading of this literature (Larson, Story & Nelson 2009; Cooksey-Stowers et al. 2017; Dubowitz et al. 2015 RAND) is that food environment access explains a meaningful but partial share of the obesity and diabetes disparity — large enough to justify intervention, but not large enough that access alone would close the gap.
CDC PLACES tract-level small-area estimates show food-desert census tracts averaging roughly 31–35% adult obesity prevalence versus 24–27% in matched high-access tracts within the same metro area — a gap that persists after adjusting for median household income.
Three policy levers dominate the food-access intervention literature: bringing food to people (mobile markets, food pantries with fresh produce), making healthy food cheaper (SNAP incentive programs like the Gus Schumacher Nutrition Incentive Program), and bringing people to food (siting new grocery stores via financing programs like the Healthy Food Financing Initiative). Evaluation evidence shows each works — but not equally, and not alone.
Mobile markets and produce-delivery programs directly shrink effective distance without waiting for a permanent retailer to commit capital — they show measurable short-term upticks in fresh produce purchases among participating households, though effects fade quickly once the program stops visiting a location.
SNAP healthy-food incentive programs (GusNIP, Double Up Food Bucks) lower the effective price of fruits and vegetables at the point of sale. Multi-site evaluations find consistent, if modest, increases in produce purchasing among participants — typically an added 0.2–0.5 cup-equivalents per day — with larger effects among households with the lowest baseline intake.
New grocery store siting (financed via the Healthy Food Financing Initiative and similar state programs) is the most visible intervention but has the most mixed evidence base for changing individual health outcomes on its own.
Cummins, Flint & Matthews (2014, Health Affairs) studied a new supermarket opening in a Philadelphia food desert using a matched comparison-neighborhood design. Awareness and self-reported use of the new store were high, and residents' perceptions of food access improved substantially — but measured fruit and vegetable intake and BMI did not significantly change relative to the comparison neighborhood after one year.
The interpretation that has shaped subsequent program design: proximity removes one barrier, but habitual purchasing behavior, price sensitivity, and marketing exposure to competing food environments (the food-swamp effect) can offset the access gain unless siting is paired with affordability and behavioral programs.
The strongest documented outcomes combine levers: new or improved grocery access paired with SNAP incentive programs and nutrition education outperforms any single intervention in the evaluation literature — which is why this simulator lets you combine grocery density and subsidy strength simultaneously rather than testing them in isolation.