Panel Paper: Ignorable Nonresponse? Improved Imputation and Administrative Data in the CPS Asec

Saturday, November 4, 2017
McCormick (Hyatt Regency Chicago)

*Names in bold indicate Presenter

Charles Hokayem1, Trivellore Raghunathan2 and Jonathan L Rothbaum1, (1)U.S. Census Bureau, (2)University of Michigan


We test an improved imputation technique for the Current Population Survey Annual Social and Economic Supplement to address match bias. Further, we augment the model with administrative data to test for nonignorable nonresponse. We find that the current practice biases distribution statistics, downward for poverty and inequaility and upward for median income. From the marginal distributions and OLS and quantile mincer regressions, we find that match bias is particularly relevant at the lower end of the unconditional and conditional distributions. However, nonresponse biases conditional and unconditional distribution statistics at high percentiles, particularly affecting inequality, the return to experience, and the association between education and earnings.

Full Paper: