Is There Nonrandom Selection in Matched Administrative Earnings Data? Evidence from the Health and Retirement Study

• Activity Number: 314
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Jesse Bricker*+
• Companies: Federal Reserve Board
• Keywords: sample selection ; survey quality ; non-random sample ; administrative data ; survey data ; earnings

Abstract

An innovation in the study of earnings is the availability of administrative records matched to individual responses in household surveys. These data are frequently used in empirical studies. A strong assumption about ignorability of sample selection is needed, though, as the subset of respondents who consented to have their earnings matched may be non-random; a counterfactual (administrative earnings of non-consenters) is generally not observed. We circumvent this obstacle by exploiting the differential timing of consent for respondents of the Health and Retirement Study; those who refused to consent in 1992, but later consented in 2004 form the counterfactual group for those who consented in 1992. Earnings, as well as their determinants, are similar between the two groups. Overall, there is little evidence of non-random selection from using the administrative data to study earnings.