Redesign of the Income Questions in the Current Population Survey Annual Social and Economic Supplement: Further Analysis of the 2014 Split-Sample Test

Publisher: Washington, DC: Mathematica Policy Research
Sep 27, 2015
Authors
John L. Czajka and Randy Rosso

Key Findings:

  • Except for wage and salary income and Social Security, the reporting of income receipt by source shows considerable volatility between 2013 and 2014, which is only partially explained by imputation.
  • The new questions were associated with significant increases in the reporting of Social Security, disability benefits, retirement income, interest, and financial assistance from others.
  • The increases in total income and unearned income were significant as well.
  • Household poverty was not reduced by the new questions, but the fraction of households above 5 times the poverty level was increased.
The Census Bureau introduced a new battery of income questions to the CPS ASEC through a split-sample test in which part of the 2014 sample received the new questions and the remainder received the traditional questions. The Census Bureau assessed the impact of the new questions by comparing estimates from the two subsamples. This report uses matched 2013 and 2014 CPS ASEC samples to conduct a difference-in-differences analysis between households responding to the traditional questions in 2013 and the new questions in 2014 and households responding to the traditional questions in both years. We find evidence of increased reporting of both total income and unearned income as well as specific sources of unearned income (earned income questions were not changed) with the new questions. We also find, unexpectedly, a significant reduction in the poverty rate among children in households receiving the traditional questions in both years but no significant change in households receiving the new questions in 2014. The difference-in-differences is not significant, but the findings lend support to the Census Bureau’s concerns that the sample split in 2014 did not yield strictly comparable samples. Repeating the matched sample analysis with the 2015 CPS ASEC could yield further insight into the impact of the new income questions. 
Senior Staff

John Czajka
Read More