Correcting for Underreporting: Using Administrative Records to Develop a Model for Improving Reported SNAP Receipt in the CPS ASEC

Liana Fox , U.S. Census Bureau
Kathryn Shantz, U.S. Census Bureau

Underreporting of program participation in household surveys has become an increasing concern among policy leaders who use these data to help inform and guide programmatic decisions. One possible solution is to use a microsimulation model, such as TRIM3, to correct for underreporting. However, recent work by Stevens, Fox and Heggeness (2018) finds evidence that TRIM3 may over-allocate benefits to the bottom of the income distribution, potentially overstating the impact of underreporting on the poverty rate. This paper provides an alternative option to correct for underreporting. Building on direct administrative data linkages, we use out-of-sample estimation to predict the likelihood of household SNAP receipiency and estimate predicted SNAP values. We will examine how replacing values from the CPS ASEC with either regression-based corrections or TRIM3 estimates changes poverty rate estimates using the Supplemental Poverty Measure (SPM).

See extended abstract

 Presented in Session 193. Measuring Poverty