For decades, public health researchers have documented the geography of food deserts — low-income, low-access areas where residents lack reasonable proximity to affordable, nutritious food. The USDA estimates 54 million Americans live in food deserts, and the health consequences are well-established: higher rates of obesity, diabetes, and cardiovascular disease.
Then came e-commerce grocery delivery, and with it the hope that digital infrastructure could solve what physical infrastructure had failed. If a low-income household in rural Mississippi or inner-city Detroit could order fresh produce from Amazon Fresh or Instacart and have it delivered the next day, the food desert would cease to exist.
Except that is not what is happening.
The Digital Food Desert Problem
E-commerce grocery delivery platforms do not serve all communities equally. Delivery availability, pricing, and service quality are algorithmically determined by fulfillment cost models that systematically deprioritize low-density, low-income, and minority communities. I call these areas digital food deserts — places where not only is physical food access limited, but digital grocery delivery has also failed to provide an alternative.
My research maps e-commerce grocery delivery coverage across all 42,522 U.S. ZIP codes for four major platforms — Amazon Fresh, Instacart, Walmart Grocery Delivery, and Kroger Delivery — and tests whether coverage gaps systematically correspond to food desert geography and chronic disease prevalence.
"Amazon's fulfillment algorithm is one of the most sophisticated optimization systems in human history. It is optimized for cost efficiency. It is not optimized for health equity. Those are different objectives, and the difference shows up in which ZIP codes get next-day delivery."
Why This Research Matters at the Policy Level
The USDA Food Environment Atlas documents physical food access barriers. The FTC's 2022 study of food retail markets examined competitive dynamics in grocery. But neither federal agency has systematically documented the e-commerce delivery coverage gap and its health consequences. My research proposes to fill that gap — producing what would be the first national census-level analysis of digital food desert distribution and its compounding effects on chronic disease outcomes.
The policy implications are direct. Federal programs like SNAP and WIC are increasingly usable for online grocery purchases. If the delivery platforms that accept these benefits are not actually available in the communities with the highest need, the policy has failed its own beneficiaries.
Research Design
Paper 4 of the health analytics program uses automated Python scripts to test delivery availability across all U.S. ZIP codes, constructing a Digital Food Desert Index (DFI). The study then tests whether DFI compounds physical food desert status in predicting chronic disease rates at the county level. Target journals: JAMA Network Open, Health Affairs.
The Amazon Perspective
I want to be precise about something. Working at Amazon has given me respect for the operational complexity of large-scale fulfillment. The coverage gaps I am studying are not the result of malicious intent. They are the result of cost optimization algorithms making rational economic decisions — decisions that, in aggregate, reproduce and amplify existing geographic inequality.
Understanding that mechanism is the first step to designing interventions that work with, rather than against, the economics of delivery logistics. That is the practical value of having a researcher with operational experience inside these systems.