The Effects of SNAP on Non-Food Expenditure: An Instrumental Variables Approach

Background: A vast body of literature has documented SNAP’s effectiveness in increasing food expenditures, reducing food insecurity and poverty, and to a lesser extent, improving the nutritional quality of diets and overall health of participants (Gundersen & Oliveira, 2001; Hoynes & Schanzenbach, 2009). A handful of studies have considered non-food related outcomes including SNAP’s effects on reducing material hardships (Schaefer & Gutierrez, 2013), medical spending (Meyerhoefer & Pylypchuk, 2008; Nicholas, 2011), labor supply (Hoynes & Schanzenbach, 2012), and changes in time allocation (Roy, Millimet, & Tchernis, 2012). However, relatively little is known on the effects of SNAP participation on other direct measures of non-food expenditures. Economic theory predicts that SNAP participation has the potential to free up income for households to spend on other goods and services – yet the empirical evidence of the effect of SNAP participation on non-food spending is seriously lacking. This paper fills this gap in the literature by investigating the causal effects of SNAP participation on household spending, with a central focus on non-food related expenditures. We apply an instrumental variable (IV) approach that exploits changes in state-level SNAP administration policies to address participation endogeneity.

Methods: We investigate the causal effects of SNAP participation on household non-food related expenditures. Our analysis focuses on 4 broad categories of non-food expenditures that are associated with child and family wellbeing using multiple waves of data from the Consumer Expenditure Survey. To address selection, we use instrumental variables approach. The main instrumental variables are rates of overpayments, the implementation of fingerprinting, and the operation of Combined Application Projects (CAP). To address potential measurement error in SNAP participation status we adopt an approach based on parametric methods for misclassified binary dependent variables that produces consistent estimates when using instrumental variables.

Results: The results from OLS are consistent with previous studies reporting that SNAP households spend significantly less on non-food goods and services compared to low-income non-SNAP households, and that the expenditure share of non-food items is also smaller for SNAP households. However, the model does not take into account the endogeniety problems of program participation. Using the IV’s described above, we show that the state-level SNAP policies are strong predictors of SNAP participation among all low-income households, conditional on a rich set of covariates. The IV models confirm there is non-random selection into SNAP. The findings suggest that SNAP participation causes an increase in non-food expenditure levels as well as an increase in the share of non-food expenditures. The findings are driven primarily by increases in spending on enrichment goods and services.

Conclusions and Implications: The findings from our study will inform policy makers on how SNAP affects household expenditures on certain non-food related goods and services that have the potential to improve family wellbeing and alleviate non-food related hardships. Such a finding support continued and expanded efforts to improve the well-being of recipients.