Description: Labor-Public Economics Workshop: Katherine Harris-Lagoudakis
Location: 368A Heady Hall
Abstract: This paper utilizes variation in the timing of benefit receipt that results from the semi-random assignment of SNAP distribution dates to evaluate the impact of SNAP issuance on household level spending patterns in retail scanner data. This approach is in contrast to recent research that utilizes variation in the timing of benefit receipt that stems from the proportion of SNAP benefits being issued over a given time period in a given state (i.e., variation in the likelihood of benefit receipt). We find that the likelihood of benefit receipt estimates are 2.0 to 2.9 times larger than the benefit receipt estimates. Benefit receipt estimates indicate a 13 percent increase in household level expenditure on the calendar week of benefit receipt, relative to calendar weeks of benefit non-receipt. We decompose the differences between these two sets of estimates and find that all of the difference is attributable to differences in endogeneity bias; explicitly, we find that the likelihood of benefit receipt estimates suffer endogeneity bias attributable to endogenous measurement error. Our findings illustrate that the utilization of group or time averages as proxies for individual stimuli may be subject to endogeneity concerns when the effect of treatment has both contemporaneous and lagged effects on the outcome of interest.
Contact person: John Winters