Food Security Research

Understanding Food Deserts: What the Data Actually Shows

The term "food desert" has dominated policy discussions about hunger in America for two decades. The concept is intuitive: if people live far from grocery stores, they cannot access healthy food. Build more stores, solve the problem. Yet when we analyzed 6,613 grocery stores across California and matched them to 9,033 census tracts, we found something unexpected.

Zero food deserts exist in Santa Clara County by the federal definition. Every resident lives within one mile of a supermarket. Yet SNAP (food stamp) participation varies 4.7-fold across demographically similar communities. Some neighborhoods with identical income levels and grocery access show enrollment rates of 3%, while others exceed 14%.

This finding challenges the core assumption of food desert policy: that geographic proximity determines food security. Our research suggests the relationship is more complex.

The Proximity Paradox

When we mapped grocery store density against food insecurity indicators, we discovered what we call the vulnerability paradox: the most economically vulnerable communities often have the best geographic food access. Dense urban neighborhoods with high poverty rates typically have multiple grocery options within walking distance. Wealthier suburban areas, by contrast, may require a car to reach the nearest store.

This pattern makes sense when we consider how grocery stores choose locations. They follow population density and purchasing power. Low-income urban neighborhoods offer both: concentrated populations and consistent demand for staple goods. The result is adequate geographic access but persistent food insecurity.

If proximity were sufficient, these neighborhoods would show the lowest rates of food insecurity. They do not.

What Actually Predicts Food Insecurity

Our research identifies four factors that better predict food insecurity than distance to the nearest grocery store:

1. Transit accessibility — Can residents actually reach stores? A grocery store one mile away means little if the bus route takes 45 minutes with two transfers. We calculated 2.7 million transit routes to measure real-world accessibility, finding that "mobility deserts" better predict food insecurity than traditional food deserts.
2. Housing cost burden — Households spending more than 50% of income on housing have little remaining for food, regardless of store proximity. This "residual income" approach explains more variation in food security than geographic measures.
3. Employment stability — Irregular work hours make it difficult to shop during store hours, even when stores are nearby. Shift workers and gig economy participants face barriers invisible to geographic analysis.
4. Enrollment friction — Eligible families often do not enroll in SNAP due to administrative barriers, stigma, or lack of awareness. In Silicon Valley, we found that half of income-eligible households do not participate in available food assistance programs.

Implications for Policy

These findings suggest that building more grocery stores—the primary policy response to food deserts—addresses only part of the problem. Stores already exist in most high-need areas. The barriers are economic and logistical, not geographic.

More promising interventions might include:

• Improving transit routes to existing grocery stores
• Reducing SNAP enrollment friction through simplified applications
• Addressing housing cost burden that leaves no budget for food
• Supporting flexible shopping options for shift workers

The food desert framework is not wrong—geographic access matters. But it has become a convenient shorthand that obscures the economic and structural factors that actually determine whether families can put food on the table.

Research Methodology

Our food security research uses multiple data sources and methods:

• Grocery store data: Google Places API collection of 6,613 stores, validated with a custom classifier achieving 94% accuracy
• Census data: American Community Survey 5-year estimates at the tract level for demographic and economic indicators
• Transit data: GTFS feeds from Bay Area transit agencies, processed with r5py routing engine
• SNAP data: California Department of Social Services enrollment records matched to census geography

All analysis code and processed data are available in our public replication repository.

Frequently Asked Questions