A conference on Using FoodAPS for Research in Diet, Health, Nutrition, and Food Security took place in Washington DC on December 7-8. Research Associate Marianne Bitler, the University of California at Davis and NBER and Program Director Janet Currie, Princeton University and NBER organized the meeting, sponsored by the Economic Research Service, the Food and Nutrition Service, and the U.S. Department of Agriculture. These researchers' papers were presented and discussed:
Bruce D. Meyer, the University of Chicago and NBER, and Nikolas Mittag, CERGE-EI
Misreporting of Government Transfers: How Important are Survey Design and Geography?
Recent studies linking household surveys to administrative records have revealed high rates of misreporting of program receipt. These studies use data from general household surveys for one or two states. Here Meyer and Mittag use the FoodAPS survey to examine whether these findings generalize to a survey with a narrow focus and across many states. First, they study how reporting errors differ from other surveys and to what extent the differences are likely due to survey design. The researchers find a lower rate of false negatives (failures to report true receipt) in FoodAPS than in other surveys. As Meyer and Mittag also find that a shorter time since last SNAP use reduces false negatives, the shorter recall period of FoodAPS likely contributes to its lower error rate. As in other surveys, they find household characteristics such as income, employment, and minority status to predict misreporting. Thus, the resulting measurement errors are neither classical nor independent of covariates conditional on true receipt. Second, the researchers examine variation in how administrative and reported SNAP receipt differ between U.S. states. Understanding geographic variation in misreporting is important to assess whether Meyer and Mittag can correct estimates of official statistics, such as program receipt or poverty rates, for the entire U.S. based on validation data from a few states. They document systematic differences between states in unconditional error rates, but cannot reject that misreporting differs between states conditional on common covariates. The researchers show that despite the geographic heterogeneity, extrapolating across states can yield substantially more accurate estimates than estimates based on survey reports.
David E. Frisvold, the University of Iowa and NBER, and Joseph Price, Brigham Young University and NBER
The Contribution of the School Environment to the Overall Food Environment Experienced by Children
With the expansion of various school meal programs, children receive an increasing fraction of their meals at school. Despite the important role of schools in the food environment faced by children, most research examining access to healthy food has focused on the retail sector. In this paper, Frisvold and Price examine the roles that both the retail- and school-based food environments have on the nutritional choices of children. Also, they determine whether the school-based food environment, as measured through the healthfulness of foods available to students through the school meal programs, compensates for the poor retail food environment in food deserts or exacerbates the problem. Further, the researchers examine how these relationships differ based on whether children receive free or reduced-price meals and whether children are in school. Their project brings together two important strands of research into one framework: the impact of school meal programs on what children eat and the impact of the retail food environment on children.
One important contribution of our project is the creation of a database of school menus covering about 90% of students in the US coupled with nutritional information about each item on the menu. Using this database, Frisvold and Price document variation in the healthfulness of school meals across the country. They then merge these data with records of food purchases and acquisitions of children and households and the retail food environment contained in the National Household Food Acquisition and Purchase Survey (FoodAPS). Descriptively, initial results show that the healthfulness of the school lunch offered to students is greater for children in neighborhoods with low access to healthy food. Comparing students within neighborhoods (Census tracts) who are exposed to the same retail environment, but who attend different school districts, the researchers find that the healthy eating index scores of students are lower in school districts which offer healthier school lunches.
Marianne Bitler; Timothy Beatty, Xinzhe Cheng, and Cynthia van der Werf, the University of California at Davis
The Pay Check Cycle
In this paper, Bitler, Beatty, Cheng, and van der Werf explore a form of cyclicality rarely studied. While a host of papers have looked at the SNAP and Social Security cycles in spending, hardly any work has focused on cyclicality associated with pay checks and spending on food (exceptions include papers by Stephens, Berniell, and Zaki). The researchers compare food acquisition in states where paychecks must be paid at least 2 times per month with food acquisition in states with no such laws, finding evidence that spending is less concentrated at the start of the month if the state required semi-monthly paychecks.
Timothy Beatty, Marianne Bitler, and Cynthia van der Werf
Do Food Assistance Programs Affect Retailers?
Food assistance is a large part of the food economy, with SNAP redemptions totaling $76 billion in 2013, or more than 10% of sales at supermarkets. Yet, Beatty, Bitler, and van der Werf know next to nothing about how food assistance shapes the retail food environment. They fill this gap, using validated causal research strategies from the literature. First, did the roll-out of Food Stamps and WIC during the 1960s and 1970s affect the retail environment at the time? The researchers find that locations with earlier Food Stamp or WIC programs have more food stores and more workers in those stores. Second, did the large increase, and subsequent decrease in the size of the SNAP program driven by the American Reinvestment and Recovery Act affect the retail food environment? Did increases and decreases have symmetric effects? Beatty, Bitler, and van der Werf anticipate that the ARRA increase should have led to more sales, with the decline leading to some decreases.
Erin T. Bronchetti, Swarthmore College; Garret S. Christensen, the University of California at Berkeley; and Benjamin Hansen, the University of Oregon and NBER
USDA Food Assistance Programs (SNAP, the National School Lunch Program, and the School Breakfast Program) and Healthy Food Choices: Quasi-Experimental Evidence from Geographic Variation in Food Prices
The Supplemental Nutrition Assistance Program (SNAP, formerly food stamps) is one of the most important elements of the social safety net. Unlike most other safety net programs, SNAP varies little across states and over time, which creates challenges for quasi-experimental evaluation. Notably, SNAP benefits are fixed across 48 states; but local food prices vary, leading to geographic variation in the real value of SNAP benefits. Bronchetti, Christensen, and Hansen provide estimates that leverage variation in local food prices across markets to examine effects of (real) SNAP benefit generosity on on nutrition. They link cross-sectional data on local food prices to food acquisition data, control for observables, and use the Oster (2016) method of bounding their estimates with assumptions about the relative degree of selection into treatment on observables and unobservables. The researchers find that children in market regions with higher SNAP purchasing power have small improvements in nutritional intake, while adults see higher levels of obesity.
Charles J. Courtemanche and Rusty Tchernis, Georgia State University and NBER, and Augustine Denteh, Georgia State University
The Impacts of SNAP on Food Insecurity, Obesity, and Food Purchases: Who Misreports and Does it Matter?
Administrative data are considered the "gold standard" when measuring program participation, but little evidence exists on the potential problems with administrative records or their implications for econometric estimates. Courtemanche, Denteh, and Tchernis explore measurement error in administrative data using data from the FoodAPS, a unique dataset that contains two different administrative measures of Supplemental Nutrition Assistance Program (SNAP) participation as well as a survey-based measure. They first document substantial missing data in the two administrative participation variables and show that they are only slightly more strongly correlated with each other than with self-reported participation. Next, the researchers find that estimated misreporting rates can vary considerably depending on assumptions used to consolidate the two administrative variables into a single "true" participation measure. They then show that instrumental variables estimates of the effects of SNAP on food insecurity, obesity, and the Healthy Eating Index are also quite sensitive to these assumptions. Using their preferred approach, which combines information from all three SNAP participation measures, SNAP is not statistically significantly associated with food security, BMI, or obesity, but increases severe obesity while worsening the healthfulness of food purchases.
Amy Ellen Schwartz, Syracuse University, and Augustina Laurito, New York University
Does School Lunch Fill the "SNAP Gap" at the End of the Month?
Schwartz and Laurito examine the relationship between the timing of SNAP benefit payments and participation in school lunch and breakfast using the National Household Survey of Food Acquisitions and Purchases (FoodAPS). An event study approach examines participation over the five-day window before and after the SNAP payment. They find that school lunch participation decreases 17 to 23 percentage points immediately after the SNAP payment among 11-18 year olds while breakfast drops 19 percentage points to 36 percentage points. The decline begins the day prior to payment. The researchers find no effects for 5-10 year olds. Models examining participation over the full SNAP month using individual random effects yield similar findings. Among teenagers, participation in school lunch and breakfast decline in the first two weeks of the SNAP month, increasing afterwards. Non-school meals show the opposite pattern. Overall, results indicate SNAP households rely more on school lunch and breakfast toward the end of the SNAP month. Older children substitute away from school meals to non-subsidized meal options earlier in the SNAP benefit cycle.
Robert A. Moffitt, Johns Hopkins University and NBER, and Kyungmin Kang, Johns Hopkins University
The Effect of SNAP and School Food Programs on Food Spending, Diet Quality, and Food Security: Sensitivity to Program and Income Reporting Error
There is an extensive research literature on the effects of the Supplemental Nutrition Assistance Program (SNAP) on food-related outcomes which has shown somewhat mixed results but generally favorable effects. However, most of the research has used data sets whose information on SNAP participation is gathered from responses on household surveys, and such responses are subject to reporting error. Moffitt and Kang use the FoodAPS data set to examine the effect of reporting error, for that data set contains information on SNAP participation gathered from government administrative records. Their analysis shows that the degree of reporting error is extremely small and has very little effect on the estimated impact of participation in the SNAP program on food security, diet quality, and food spending. A supplemental analysis of the effect of school food programs is also provided.
Helen H. Jensen, Brent Kreider, and Oleksandr Zhylyevskyy, Iowa State University
Investigating Causal Effects of SNAP and WIC on Food Insecurity Using FoodAPS
The U.S. Department of Agriculture (USDA) operates several food assistance programs aimed at alleviating food insecurity among low-income households. While many assistance recipients participate in more than one food program, little is known about how the programs interact. The National Household Food Acquisition and Purchase Survey (FoodAPS) provides self-reported information about a households participation in selected food programs and validates participation status in the Supplemental Nutrition Assistance Program (SNAP), which is the largest program by expenditures. Jensen, Kreider, and Zhylyevskyy leverage these key features of FoodAPS to focus on two programsSNAP and the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC)and address the following questions:
To what extent does joint participation in SNAP and WIC alleviate food insecurity compared with participation in SNAP or WIC alone?
How can the researchers combine self-reported program participation with administrative SNAP validation data to tighten inference on the causal effects of the programs?
Jacob S. Goldin, Stanford University; Tatiana Homonoff, New York University; and Katherine H. Meckel, Texas A&M University
Issuance and Incidence: SNAP Benefit Cycles and Grocery Prices
In-kind benefit transfer programs like the Supplemental Nutrition Assistance Program (SNAP) are an increasingly important part of the U.S. safety net. Because states issue SNAP benefits to each recipient once per month, retailers experience predictable cyclicality in consumer demand for food. In response to these fluctuations, retailers face incentives to vary food prices throughout the month, potentially shaping the incidence of the benefits transferred through the SNAP program. Using a large panel data set from households and retailers, Goldin, Homonoff, and Meckel document large intra-month cycles in food expenditures among that closely track state issuance policies. However, they find evidence that retailers do not vary their prices in response to such fluctuations by an economically significant magnitude. This finding is consistent with recent evidence showing that grocery retailers largely adopt a strategy of uniform pricing at the expense of substantial increases in profits.
Di Fang, Aaron M. Novotny, Rodolfo Nayga, and Michael Thomsen, the University of Arkansas
WIC Participation and Relative Quality of Household Food Purchases: Evidence from FoodAPS
Fang, Novotny, Nayga, and Thomsen examine the effect of the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) participation on the relative quality of household food purchases using the National Household Food Acquisition and Purchase Survey (FoodAPS) and propensity score matching. Quality of household food purchases was measured using a healthy purchasing index (HPI). Results suggest that households participating in WIC have a higher HPI value in comparison to eligible non-participating households. Importantly, this difference is driven entirely by WIC participating households who redeemed WIC foods during the interview week. There was no significant difference between WIC participating households who did not redeem WIC foods during the interview week and WIC-eligible but non-participating households. Robustness checks suggest that WIC foods explain the improvement in relative quality of household food purchases, not self-selection of more nutrition-conscious households into the program. As a secondary objective, the researchers examine whether geographic barriers limit WIC participation. Locations of WIC clinics were added to the already detailed FoodAPS information on food store locations. There is no evidence in this sample that access to clinics is adversely affecting WIC participation rates nor are there meaningful differences in the HPI outcomes among participants with and without supermarket access.