Heterogeneous impacts of the Supplemental Nutrition Assistance Program on food insecurity

We study the effects of SNAP participation on food insecurity allowing for a priori unspecified heterogeneous treatment effects. Using finite mixture models, we identify a low food security class comprising almost 60% of the samples for whom SNAP participation increases the probability of no food insecurity by 14-37 percentage points across specifications and decreases the probability of very high insecurity by 14-35 percentage points. We find that SNAP participation has a small and statistically insignificant effect on food insecurity for the remaining 40% of the population. By examining posterior probabilities of class membership, we discover that individuals in the latter class are less likely to report unmet food needs and live in larger households where more consumption smoothing may be possible.