Association for Public Policy Analysis & Management

The Supplemental Nutrition Assistance Program (SNAP), formerly the Food Stamp Program, provides in-kind food benefits to approximately 40 million people each month. Nearly 70% of the current SNAP caseload is composed of people in families with children. Federal statistics on the characteristics of SNAP/food stamps-participating households display a striking trend over the past two decades—an increase in the proportion of households with no source of cash income. In 2015, approximately 22% of SNAP households had zero gross income compared to only 9% in 1997. This trend coincides with increases in $2.00 per person per day poverty observed in other studies (Edin & Shaefer, 2015).

Understanding of zero-income SNAP households is otherwise fairly limited, yet given their meager economic resources and their prevalence in the contemporary population of SNAP beneficiaries they warrant attention. In particular, the extent and nature of this condition in households with children requires investigation. Families with children compose a large portion of the SNAP caseload, and economic deprivation has potential consequences for this group in areas such as health, emotional well-being, and educational success.

Research Questions

In this study, I address three questions related to SNAP-receiving households with children reporting no cash income. First, has the proportion of households with zero income changed over time among households with children specifically? Second, does the prevalence of these households on the SNAP caseload vary by state? Finally, how—if at all—do reports of zero cash income relate to changes in access to cash benefits under Temporary Assistance for Needy Families?

Methods

I construct a dataset of SNAP units in all fifty states from 2001 to 2015 using the SNAP Quality Control database, a monthly sample of SNAP units primarily intended to estimate state error rates in SNAP administration. These data also contain information on benefit unit composition (e.g., number of members, relationships between members and unit head, key demographic information) and sources of income. I restrict the sample to households with children (n=346,631).

I examine patterns across time and across states using descriptive and graphical analyses of the proportion of SNAP units with children reporting zero income. Next, I estimate a logistic regression model of the probability of a SNAP case reporting zero income. The model includes state and year effects, household compositional characteristics (e.g., number and age of children, presence of people with disabilities), and a set of state-year variables operationalizing relevant economic (e.g., unemployment rate) and policy factors. The key state-year variable is the ratio of TANF-receiving families to families in poverty, a measure of cash assistance accessibility (Center on Budget and Policy Priorities, 2018). The marginal effect of the coefficient on this variable is the main object of interpretation. I repeat these procedures for a subsample composed of only units headed by a single female (n=205,052), ostensibly the households at greatest risk for economic hardship and the most likely to be sensitive to changes in cash welfare policy.