Using Linked Survey and Administrative Data to Better Measure Income: Implications for Poverty, Program Effectiveness, and Holes in the Safety Net

Abstract

We examine the consequences of survey underreporting of transfer programs for prototypical analyses of low-income populations. We link administrative data for four transfer programs to the CPS to correct its severe understatement of transfer dollars received. Using survey data sharply understates the income of poor households, distorts our understanding of program targeting, and greatly understates the effects of anti-poverty programs. Using the combined data, the poverty-reducing effect of all programs together is nearly doubled. The effect of housing assistance is tripled. Correcting survey error often reduces the share of single mothers falling through the safety net by one-half or more. (JEL C83, I32, I38)

Access and enrollment in safety net programs in the wake of COVID-19: A national cross-sectional survey

“…Future research, with larger sample sizes, could beneficially compare differences in program participation based on residence in areas experiencing higher COVID-19 case rates [28]. Third, as with all surveys on public program participation, there is likely to be some under-identification of program participation [29]. Self-reported program participation may be particularly problematic for some programs such as SNAP [30], although we have no reason to believe that reporting bias will be differential between people who experienced COVID-19 employment loss versus those that did not.…”

The Dynamics of the Racial Wealth Gap

“…32 Whatever decade of data we look at in the SCF, we see a large gap between the Black and white CEFs. There is considerable uncertainty in the measurement of lifetime earnings in the NLSY79, generated by alternative approaches to imputing missing observations longitudinally (Nielsen (2015), Nielsen (2020)) along with the known underreporting of income in surveys in the left tail of the distribution (Meyer and Mittag (2019), Meyer and Sullivan (2003)). As we describe in Appendix A.2, our preferred approach is to impute missing observations of earnings longitudinally as the most recent previous observation.…”

Evaluating the Utility of Linked Administrative Data for Nonresponse Bias Adjustment in a Piggyback Longitudinal Survey

“…Typically, administrative data are generated longitudinally, which makes them a viable source for studying biographical changes, program evaluation, or simply as a complement to surveys to lower the burden of data collection (Olson 1999;Scholz et al 2006;Antoni and Bethmann 2019). Besides substantive research, linked-administrative data are also used for methodological purposes, such as the assessment of nonresponse and measurement errors (Kreuter et al 2010a;Meyer and Mittag 2019) and for improving survey data that are affected by these errors (Davern et al 2019).…”

“…If this is not the case, then the linked sample cases are only representative of the subpopulation that overlaps with the administrative population. For example, Meyer and Mittag (2019) report an imperfect overlap in a linkage of the New York State sample of the 2008-2013 Current Population Survey Annual Social and Economic Supplement (CPS-ASEC) to administrative records from the Office of Temporary and Disability Assistance and the Department of Housing and Urban Development. The authors correct for this shortcoming using an inverse probability weighting procedure.…”