Project:
Did the Rollout of the Food Stamp Affect Retailers?

Year: 2019

Research Center: Tufts University/University of Connecticut (UConn) RIDGE Program

Investigator: Timothy Beatty, Marianne Bitler, and Cynthia van der Werf

Institution: University of California at Davis

Project Contact:
Marianne Bitler
UC Davis
Department of Economics
1 Shields Avenue,
Davis, CA 95616
Phone: 415-305-5644
Email: bitler@ucdavis.edu

Summary:

The Supplemental Nutrition Assistance Program (SNAP) is a cornerstone of the food economy. In 2013, SNAP benefits were accepted in 252,962 firms. Around $76 billion in SNAP benefits were redeemed in 2013, with the majority being used at superstores or supermarkets (about $62 billion) on a total base of supermarket sales of around $649 billion. During the Great Recession, SNAP spending accounted for more than 10 percent of food-at-home spending and was the largest source of food-at-home spending for low-income Americans.

Although many studies have analyzed SNAP’s effect on food spending by low-income households, work on how SNAP shapes retailers, the retail food environment, and the food industry more broadly is virtually nonexistent. The size and scope of SNAP and the long history of the program are prima facie evidence that SNAP dollars may have shaped the retail food environment faced by the poor. Retailer location is an equilibrium outcome for firms; all else equal, retailers will choose to locate in areas of highest demand, and SNAP also likely helps shape this choice.

Ignoring the general equilibrium effect of food assistance on the retail food industry will lead to an incomplete accounting of its effects. First, to the extent that food assistance programs lead to a higher quality food environment, it benefits both participants and nonparticipants. Prior work evaluating food assistance programs has neither considered nor quantified these externalities. Thus, prior work looking at the effects of SNAP may have conflated the direct effect of increased food spending for participants with the indirect effect on the retail food environment, which could affect a broader group. In other words, increased SNAP participation and benefits in an area may induce stores to open or expand offerings, which in turn may increase the value of a given level of benefits to participants and nonparticipants through lower prices, travel times, and increased choice.

Our research uses the rollout of the Food Stamp program as a source of variation to identify effects of food assistance on the retail food environment at that time. Food Stamps were rolled out in a quasi-random fashion across counties during the 1960s and early 1970s. We build on previous work showing that the idiosyncratic rollout of the Food Stamp program across counties affected a wide range of outcomes, including food purchases. Our hypothesis is that locations which implement Food Stamps will have more stores selling food, more workers in those stores, and higher sales than they did before adopting the program.

We tackle our research question using three data sources, County Business Patterns (CBP) data from 1970-78, newly digitized data from the Census of Retail Trade/Census of Business on retail trade establishments from 1954 through 1977, and U.S. Decennial Census data from 1960/1970/1980 to study this question. Using CBP data, we study either the effects of the Food Stamp rollout on employment in food and grocery stores with paid employees in the county as well as on payroll. Using the Census of Retail Trade/Census of Business, we analyze the effects of the rollout on real sales, the number of establishments, and the share of establishments in various categories. Using census data, we examine the number of persons employed and decompose employment into various occupations. We use a difference-in-difference research design across all specifications. Finally, we control for time invariant characteristics of counties (CBP and Census of Retail) and commuting zones (Census) as well as national shocks and other controls for welfare spending and the business cycle.

Our findings suggest that rollout of the Food Stamp program in the early 1960s and 1970s led to about a 5-percent increase in employment in food and grocery stores. These findings hold using both 1970-78 County Business Patterns data and pooled 1960/1970/1980 Decennial Census data. This is consistent with evidence from newly digitized Census of Retail Trade data on sales, the number of establishments, and the share of establishments in these industry categories. Rollout of Food Stamps led to a 1.3–1.6-percent increase in real sales, robust to controlling for county trends. Effects on sales in eating and drinking establishments and all other establishments are smaller in magnitude and never statistically significant. Our findings suggest that the rollout of Food Stamps had substantive effects on sales and employment in food stores.