The effect of benefit underreporting on estimates of poverty in the United States

The household income data used most frequently to estimate poverty rates in the United States substantially underreports the value of means-tested transfers. This paper investigates how underreporting affects estimates of the incidence and composition of poverty in the U.S. from 2013 to 2015. Specifically, I apply benefit adjustments for the underreporting of three social transfers to the Current Population Survey (CPS ASEC) to provide more accurate estimates of poverty rates. Diagnostic checks indicate that the imputed benefit adjustments are imperfect, but do provide a more accurate representation of household income than the uncorrected CPS data. In 2015, the benefit adjustments add more than $30 billion of income transfers to the CPS ASEC, primarily concentrated among low-income households with children. I test the effects of the benefit corrections on two conceptualizations of poverty: the U.S. Supplemental Poverty Measure (SPM) and a relative measure of poverty set at 50% of federal median income. In 2015, the SPM poverty rate for the total population falls from 14.3 to 12.7%, a 1.6 percentage point (11%) decline, after adjusting for underreporting. Among children, the SPM poverty rate falls from 16.1 to 12.8%, a 3.3 percentage point (20%) decline. The percent-of-median poverty rate experiences similar declines after applying the benefit imputations. The findings suggest that the uncorrected CPS data meaningfully overestimates the incidence of poverty in the U.S., particularly among households with children. Documentation for applying the benefit adjustments to the CPS is provided for improved estimates in future poverty research.