Food Insecurity Dynamics and the Food Stamp Program
Research Center: Southern Rural Development Center, Mississippi State University
Investigator: Mills, Bradford, Elton Mykerezi, and Yiran Li
Institution: Virginia Tech University
Food-insecure households have “limited or uncertain availability of nutritionally adequate and safe foods or limited or uncertain availability to acquire acceptable foods in socially acceptable ways.” The USDA Supplemental Nutrition Assistance Program (SNAP), formerly the Food Stamp Program (FSP), is the Nation’s largest domestic food and nutrition assistance program. SNAP/FSP is designed to provide food assistance via benefit payments to low-income households and, thereby, increase their levels of food security. This report covers a time period before SNAP came into being; thus, all references are to FSP.
This project examines the link between FSP participation and food insecurity (FI) using monthly data. This link has been surprisingly difficult to establish empirically, as households with higher levels of food insecurity are often more likely to participate in the FSP for unobserved reasons. Another potential reason for the weaker than expected link between FSP receipt and increased household food security is the relatively aggregate annual timeframe of FI and FSP participation employed in analyses to date. However, a household’s FI and FSP use often evolves within a year. Thus, intra-annual measures of FI and FSP receipt are crucial for accurate measurement of FSP impacts.
Two sets of models are employed in this report. The first set, Poisson models, examine how household socioeconomic characteristics affect the intensity of FI, participating the FSP, and being unemployed within a year. The second set, fixed effect Linear Probability Models (LPM), employ intra-annual measures of FI and FSP use to examine the underlying dynamics of FSP impacts on FI. Specifically, the study uses a monthly panel of FI status, FSP use, and employment status to examine how FI status varies within households over months before and after the start of a FSP spell. The FI status of participants in pre-FSP and during-FSP months is compared to the months that are not close to an FSP spell.
A panel dataset of households with children for the years 1998, 2000, and 2002 was drawn from the Panel Survey of Income Dynamics (PSID). The national sample contains 1,753 households in 1998, 2,760 households in 2000, and 3,120 households in 2002. The subsample of rural households consists of 509 households in 1998, 614 households in 2000, and 759 households in 2002. The Poisson and Linear Probability models are estimated using both the national and rural samples.
Poisson model estimates reveal that higher income significantly reduces the number of months households are unemployed, in the FSP, and food insecure. Other results are similar to those found in the existing literature. For instance, higher levels of education reduce the intensity of exposure to the FSP and to FI. Simple descriptive statistics show that household food insecurity first increases significantly in the months immediately prior to FSP participation and then starts to decline once the household participates in the FSP. Results from the fixed effect LPM confirm the trend displayed in descriptive statistics (Figure 1). Compared with the baseline when a household is more than 12 months removed from a FSP spell, households are somewhat better off in months 4-12 prior to participation. Then FI conditions are statistically similar (although worse off numerically) compared with the baseline in the period two and three months prior to participation. The probability of being food secure is 4 percentage points higher relative to the baseline in the month prior to FSP participation and reaches its highest deviation from the baseline in the month when the household enrolls in the FSP. After joining the FSP, FI declines steadily until it is only one percent higher than the baseline probability after 4-12 months in the program.
Descriptive statistics for rural food insecurity in relation to the timing of FSP participation are similar to those for the nation (i.e., food security deteriorates immediately prior to participation and then improves with time spent in the FSP). Statistical impacts of FSP on FI are, however, weaker in the fixed effect LPM for the rural households. Only 2 of the 8 indicator variables of FSP participation status are statistically significant with the smaller rural sample.
When a similar model with annual FSP participation and FI measures is estimated, no evidence of FSP impact is found. This result suggests that the use of annual measures may be an important reason for the lack of empirical evidence of FSP impacts on FI research in most studies to date. The results also imply a need to re-conceptualize the pathway through which the FSP impacts household food security. Specifically, the FSP does not improve food security per se. Rather, the FSP ameliorates the impact that negative household economic shocks have on FI.