The impact of food geography on consumer and producer welfare

I study the interactions between food prices, food retail environment and food demand. Given the rising debate on how to improve the quality of food purchases and consumption by changing the food geography in the United States, the overall goal of my dissertation is to understand the nature, causes and impact of living in a food desert (low-income areas with limited access to supermarkets) in the United States. Low-income households have poorer diets and higher rates of obesity than high-income households in the United States. However, it is unclear why these disparities exist. Poor diets can be driven by three factors: limited availability of healthy foods, higher prices of healthy foods, or preferences for unhealthy foods. Depending on which of these causes hold, improving access to healthy foods may or may not have an effect on the foods purchased by households living in food deserts. The first chapter of my dissertation studies the first two potential drivers of poor diets, prices and availability. I compare food prices and availability between food deserts and non-food deserts to observe whether and to what extent consumers face higher prices for a complete diet in food deserts. If a nutritionally complete diet costs significantly more in food deserts, resident consumers may be constrained from consuming healthier foods. I am able to attribute a dollar value to the importance of food variety and incorporate food variety into the calculation of the price index to allow for the availability of food products to differ across food deserts and non-food deserts I use the 2012 IRI InfoScan that records weekly store-level sales in a nationally representative sample of counties at over 15,000 retailers to identify the products available and prices of these products offered in each census tract. In doing so, I calculate a census-tract level Exact Price Index (EPI) based on store sales in home and contiguous census tracts and a food basket defined by the Thrifty Food Plan (TFP). The EPI addresses potential biases from both product heterogeneity and product variety. I find that after controlling for observable demand, the stores charge slightly higher prices (0.9%) and have lower food availability (2.6%) in low access tracts than in their high access counterparts. In combination, the variety-adjusted price index is only 3.5% higher in low access areas. The small difference in EPI between food deserts and non-food deserts suggests both limited market power of existing stores in food deserts and a relatively small welfare impact of living in a food desert, at least for those who can travel to neighboring census tracts to shop. The second chapter of my dissertation studies whether better availability or lower prices of fruits and vegetables in food deserts improves consumers’ welfare; both policies are implemented in the real world. I use a structural estimation approach from the industrial organization literature and estimate a discrete choice demand for different types of food stores. I explicitly model consumer demand with spatial differentiation and heterogeneous preferences. If the existence of food deserts is caused by lower demand for access to healthy foods, then improving food access alone will not be effective in changing consumer behavior. Thus it is important to allow consumers in food deserts and non-food deserts to have differential preferences for store proximity and other store attributes. With specific knowledge of consumers’ demand and preferences, I can conduct counterfactual analyses and study the welfare implications of intervention policies on food deserts ex ante. I use a panel of quantities and prices for 174 food stores from 11 randomly chosen mid-sized counties over a period of 16 quarters (2009-2012), collected using scanning devices from Information Resources Inc. (IRI). Store characteristics come from Nielsen TDLinx store directory data and census-tract level socio-demographics from 2008-2012 American Community Survey (ACS). The census-tract level food deserts indicators are from 2010 USDA Food Access Research Atlas. I find that first, price is the most important factor when consumers decide where to shop and is much more important than availability of fruits and vegetables and store proximity. Second, consumer’s welfare does not greatly increase when the number of fruits and vegetables are increased by a small amount in each store nearest to a food desert. This result flows from the fact that consumers do not value availability of fruits and vegetables as much as prices, and most consumers have access to vehicles and are thus less constrained by store distances to shop in the nearest supermarkets. Third, fruits and vegetables price subsidies are more effective policies to improve consumer welfare and have better distributional effects that they benefit food deserts households more. Lastly, food deserts households have lower demand for better access to fruits and vegetables and may explain, at least in part, why food deserts exist. Building on the second chapter, in the third chapter I incorporate supply side price competition and producers’ surplus into the welfare analysis of supermarket entry. Specifically, I evaluate the welfare impact of a subsidy to a low-cost grocery store to enter food deserts. Thus I consider the full effect of the entry of a new grocery store, not just on increased food availability but also on prices in existing stores. As found in chapter 2, price is the most important factor in consumers’ choices of stores. Thus incorporating price effect and supply side allows us to more accurately assess the total policy impact. I use a multinomial mixed logit demand model combined with a Nash-Bertrand price competition model to analyze the food store retail industry. After obtaining consumer preferences and cost parameters, I conduct a counterfactual analysis to see what happens when a low-cost grocery store comes to the most severe food desert in the county. I use Walmart Neighborhood Market (WNM) as an example of a low-cost grocery store because WNM offers prices that are 10%-15% lower than an average grocery store, which may be particularly appealing to low-income consumers and is expanding fast in the U.S. recently. The most severe food desert is the low-access census block group with the highest poverty rate within a county. I use USDA’s new National Household Food Acquisition and Purchase Survey (FoodAPS) combined with 2012 IRI InfoScan to estimate consumers’ food store choices and food stores’ pricing behavior and distribution of marginal costs. The FoodAPS contains detailed information about the foods purchased or acquired by surveyed households for at-home consumption and away-from-home in 2012. The 2012 Information Resources Inc. (IRI) InfoScan records weekly barcode level sales and quantities sold in major food chain stores that cover over 80% of grocery purchases in the US. These data allow me to address limitations in the literature that are critical to understanding store choices and policy interventions to improve food access. There are three central findings from the analyses. First, price is the most important store attribute when consumers decide where to shop, especially for food deserts (FD) households. Second, after a low-cost grocery store comes to the most severe food desert, FD households benefit more than NFD households. On the supply side, larger grocery stores and stores that are close to the incoming new grocery store are more negatively affected. Third, the average annual county-level social welfare increases by $219,810, in which consumer welfare increases by $3,843,880 and producer welfare decreases by $3,624,070. In addition, the fixed costs of operating a low-cost grocery store in the most severe food deserts is more than the variable profit obtained, thus casting doubt on the effectiveness of the policy to subsidize low-cost grocery stores into the most severe food deserts. This result is robust across various low-cost grocery retail chains including WNM and Food 4 Less