The USDA's food desert framework assumes that distance to grocery stores is a primary barrier to food access. [1] Federal policy has directed resources toward bringing stores closer to underserved communities, from the Healthy Food Financing Initiative to state-level incentive programs. [2]
Our earlier analysis of 7 Bay Area and major metro counties suggested this framing may be backwards: the most vulnerable communities already had closer stores, not farther ones. But that finding was limited to ~2,000 census tracts in urban California. Does the pattern hold statewide? And does it hold for specific demographic groups?
The article reports a test across 9,033 residential census tracts using racial-composition and housing-tenure cuts. Related pages report incompatible populations, measures, and directions, so the results below are preserved as published claims under reconciliation.
The food access vulnerability paradox is the hypothesis that communities can have shorter grocery distances while experiencing worse food-security outcomes. This article's statewide estimates are under reconciliation and should not be used to establish that pattern yet.
The Statewide Test
We can apply the same methodology used in the 7-county pilot: grocery store distances from population-weighted centroids, mobility desert classification (grocery within 1 mile but inadequate transit), and a composite vulnerability index incorporating food access, poverty, renter status, minority percentage, and sprawl.
We examine two demographic lenses. First, we compare majority-minority tracts (where non-white residents exceed 50% of the population) to majority-white tracts. [3] Second, we compare renter-dominated tracts (where renters exceed 50% of occupied housing units) to owner-dominated tracts.
The article reports a common pattern across both comparisons. The related pages do not currently agree, so that synthesis is provisional.
Race and Food Access
The article reports a 9,033-tract population split into 6,098 majority-minority and 2,935 majority-white tracts. That population conflicts with the Transit Equity and Race page and is under reconciliation.
| Metric | Majority-Minority | Majority-White |
|---|---|---|
| Tracts | 6,098 | 2,935 |
| Mean distance to grocery | 1.03 mi | 1.78 mi |
| Food desert rate | 21.9% | 39.9% |
| Mobility desert rate | 9.7% | 16.9% |
| Vulnerability index | 0.337 | 0.279 |
The article reports distances of 1.03 versus 1.78 miles, food-desert rates of 21.9% versus 39.9%, and mobility-desert rates of 9.7% versus 16.9%. The last comparison runs opposite to the Transit Equity and Race result for similar stated vintages, so the full set is under reconciliation.
The article also reports vulnerability scores of 0.337 versus 0.279. The component inputs and interpretation have not been reproduced from a matching public package.
Controlling for Income and Density
The raw comparison conflates racial composition with urban location. Majority-minority tracts concentrate in urban cores where stores are dense. To isolate the racial composition effect, we estimated a linear probability model predicting mobility desert status, controlling for median household income and population density:
| Variable | Coefficient | Clustered SE | p-value |
|---|---|---|---|
| Majority-Minority (dummy) | -0.082 | 0.013 | <0.001 |
| Median Income (per $10K) | -0.000 | 0.000 | 0.43 |
| Population Density (per 1000) | +0.015 | 0.003 | <0.001 |
Linear probability model predicting mobility desert status (binary). N = 9,033 California census tracts. Standard errors clustered at county level (58 clusters). R² = 0.017.
The article reports an 8.2 percentage point lower probability after controlling for income and population density. That estimate is under reconciliation with the opposite direction reported on the Transit Equity and Race page and should not be described as robust yet.
Note on clustered standard errors: We cluster standard errors at the county level rather than treating each tract as independent. [4] Tracts within the same county share transit systems, policy environments, and geographic constraints. Ignoring this clustering would understate uncertainty.
Housing Tenure and Food Access
The same pattern emerges when we compare renter-dominated to owner-dominated tracts. Research documents that renters have lower car ownership rates than homeowners, making transit access particularly consequential for this group. [5]
| Metric | Renter-Dominated | Owner-Dominated |
|---|---|---|
| Tracts | 3,473 | 5,560 |
| Mean distance to grocery | 0.74 mi | 1.60 mi |
| Food desert rate | 10.3% | 38.7% |
| Mobility desert rate | 6.0% | 15.8% |
| Vulnerability index | 0.357 | 0.294 |
The article reports distances of 0.74 versus 1.60 miles, food-desert rates of 10.3% versus 38.7%, and mobility-desert rates of 6.0% versus 15.8%. These values conflict with the Housing Tenure article and are under reconciliation.
The reported vulnerability scores of 0.357 versus 0.294 are also provisional because they use the unresolved tract population.
The Income Quintile Test
One might argue that the renter advantage simply reflects income composition—renters have lower incomes on average, and lower-income neighborhoods happen to be in denser urban areas. To test this, we stratified by tract median household income quintile:
| Income Quintile | Renter Rate | Owner Rate | Difference |
|---|---|---|---|
| Q1 (Lowest) | 4.5% | 14.1% | -9.7 pp |
| Q2 | 6.7% | 13.9% | -7.2 pp |
| Q3 | 7.8% | 12.9% | -5.1 pp |
| Q4 | 7.9% | 16.7% | -8.7 pp |
| Q5 (Highest) | 5.4% | 18.7% | -13.3 pp |
The article reports a renter advantage at every income level and a 13.3 percentage point gap in the highest quintile. Those comparisons have not been reconciled with the separate Housing Tenure article.
No conclusion about income composition, housing type, or location choice should be drawn from this table until the populations and measures match.
The Vulnerability Paradox
The article reports a shared pattern across two demographic cuts. Conflicts with the related race and housing-tenure pages mean that synthesis is provisional.
Urban geography is one possible explanation, but the current public record cannot distinguish explanation from measurement conflict.
Urban-core concentration is one article-proposed explanation to test. The page has not reproduced whether its tract groups cluster in the named cities or whether development form explains the disputed comparisons. [6]
External research documents links among historical redlining, residential patterns, and present-day food environments, but it does not validate this page's unresolved tract comparison. [7]
The article proposes that geographic and economic components offset one another in its vulnerability index. That decomposition remains under reconciliation with the input populations and related pages.
Regional Variation
The article reports the same direction across 9 regions. Because the statewide comparison conflicts with a related page, this table is preserved for reconciliation and is not robustness evidence:
| Region | MM Rate | MW Rate | Difference | MM N | MW N |
|---|---|---|---|---|---|
| Los Angeles Metro | 7.9% | 15.3% | -7.4 pp | 3,180 | 1,059 |
| Bay Area | 5.7% | 11.2% | -5.5 pp | 1,226 | 526 |
| Central Valley | 11.8% | 18.7% | -6.9 pp | 787 | 187 |
| San Diego | 25.3% | 31.0% | -5.7 pp | 431 | 345 |
| Sacramento | 19.4% | 20.9% | -1.5 pp | 284 | 316 |
| Central Coast | 6.8% | 17.6% | -10.8 pp | 161 | 193 |
| North State | 10.0% | 14.2% | -4.2 pp | 20 | 141 |
| North Coast | 0.0% | 7.1% | -7.1 pp | 7 | 84 |
| Mountain Sierra | 0.0% | 7.7% | -7.7 pp | 1 | 91 |
The article reports regional gaps from 1.5 to 10.8 percentage points in one direction. Because the statewide direction conflicts with a related page, the regional pattern is also under reconciliation.
What This Means for Policy
The USDA food desert framework and related policy interventions assume that distance to stores is a primary barrier to food access. Federal programs have directed resources toward building grocery stores in "underserved" areas and improving transit to reach distant stores. [2]
The article's published estimates would complicate this framing if a matched rerun reproduces them.
The current public record does not establish whether the affected communities have better or worse mobility access because related pages report opposite directions.
Affordability can remain a barrier even when stores are nearby, but this article's unresolved composite analysis cannot establish that it is the primary barrier for the reported tract groups. Research on transit-dependent populations provides external context. [8]
The article reports lower minority-tract mobility-desert rates inside and outside major metros, but a related page reports the opposite statewide direction. Geographic generalization is postponed until those records are reconciled.
Policy choice should therefore use independently verified local evidence rather than the disputed statewide comparison. Economic and geographic interventions remain hypotheses to test against a matched population.
Limitations
This analysis has important constraints.
Tract-level aggregation. Census tract data captures neighborhood averages, not individual experience. A renter near a transit stop and a renter far from one both count toward the tract's renter percentage. Within-tract variation may be substantial.
Cross-sectional design. We observe each tract at a single point in time. Demographic composition changes through gentrification, migration, and displacement. The patterns observed in 2018-2022 data may not persist.
Binary classifications. Treating >50% as the threshold for "majority-minority" or "renter-dominated" loses information. The article reports a continuous-measure check, but no matching public output reproduces it. It cannot establish robustness while the main direction remains unresolved.
Geographic access only. We measure distance and transit availability, not affordability or store quality. A tract with a nearby store charging premium prices may have worse effective access than a tract with a more distant discount store.
Correlation, not causation. Geography is one possible explanation for the article-reported renter comparison. The conflicting housing-tenure page and missing matched output prevent treating mediation by geography as established.
Data and Methods
Material status: Article only. No matching public package currently derives the tract populations, tables, models, or policy conclusions on this page.
Data sources:
- Race/ethnicity: ACS 2018-2022, Table B03002
- Housing tenure: ACS 2018-2022, Table B25003
- Transit stops: Cal-ITP statewide GTFS (2024)
- Grocery distances: Calculated from population-weighted centroids using SafeGraph (2023)
- Vulnerability index: Composite of food access, poverty, renter status, minority percentage, sprawl
Classification:
- Majority-minority: > 50% non-Hispanic non-white
- Renter-dominated: > 50% renter-occupied units
- Mobility desert: Grocery within 1 mile, transit stop > 0.5 miles or < 2 stops within 0.5 miles
Statistical analysis:
- Descriptive comparison of means
- Linear probability model with income and density controls
- Standard errors clustered at county level (58 clusters)
Notes
[1] Ver Ploeg, M., et al. (2009). Access to Affordable and Nutritious Food: Measuring and Understanding Food Deserts and Their Consequences. USDA Economic Research Service. The USDA defines food deserts as low-income census tracts where a substantial share of residents has limited access to supermarkets or large grocery stores. ↩
[2] The Healthy Food Financing Initiative (HFFI), launched in 2010, provides grants and loans to bring grocery stores to underserved communities. State programs include the California FreshWorks Fund. See PolicyLink (2019), "Access to Healthy Food." ↩
[3] Majority-minority defined using ACS Table B03002 (Hispanic or Latino Origin by Race). Non-Hispanic white percentage calculated as B03002_003E / B03002_001E. Tracts where this ratio ≤ 0.50 are classified as majority-minority. ↩
[4] Cameron, A. C., & Miller, D. L. (2015). A practitioner's guide to cluster-robust inference. Journal of Human Resources, 50(2), 317-372. https://doi.org/10.3368/jhr.50.2.317 ↩
[5] Blumenberg, E., & Pierce, G. (2012). Automobile ownership and travel by the poor: Evidence from the 2009 National Household Travel Survey. Transportation Research Record, 2320(1), 28-36. ↩
[6] Levine, J., & Inam, A. (2004). The market for transportation-land use integration: Do developers want smarter growth than regulations allow? Transportation, 31(4), 409-427. ↩
[7] Li, M., & Yuan, F. (2022). Historical redlining and food environments: A study of 102 urban areas in the United States. Health & Place, 75, 102775. https://doi.org/10.1016/j.healthplace.2022.102775 ↩
[8] Bayly, R., Pustz, J., Stopka, T. J., Metzger, J., & Waters, M. C. (2025). Transit bus access to healthy, affordable food: A novel geographic information system (GIS) and community-informed analysis. SSM - Population Health, 29, 101753. https://doi.org/10.1016/j.ssmph.2025.101753 ↩
Frequently Asked Questions
Do minority neighborhoods have worse food access in California?
This article reports shorter grocery distances and lower food-desert rates in majority-minority tracts. A related article reports the opposite direction for mobility-desert rates using ostensibly similar inputs, so the result is under reconciliation.
How much lower are mobility desert rates in minority neighborhoods?
The article reports an 8.2 percentage point lower probability after controls. That estimate is under reconciliation with a related article that reports the opposite substantive direction.
How many California census tracts were analyzed?
The article reports 9,033 tracts split into 6,098 majority-minority and 2,935 majority-white tracts. Those populations conflict with related pages and are under reconciliation.
Does better geographic access mean better food security?
The article reports higher vulnerability scores despite shorter grocery distances. Its component inputs and policy interpretation are under reconciliation with related statewide pages.
How to Cite This Research
Cholette, V. (2025, December 24). Food access comparison under reconciliation. Too Early To Say. https://tooearlytosay.com/research/food-security/food-access-vulnerability-paradox/Copy citation