Sample Loss and Survey Bias in Estimates of Social Security Beneficiaries: A Tale of Two Surveys

Data from the Census Bureau’s Survey of Income and Program Participation (SIPP) and the Current Population Survey (CPS) provide information on current and potential beneficiaries served by Social Security Administration (SSA) programs. SSA also links administrative records to the records of survey respondents who provide Social Security numbers. These matched data expand the content of the SIPP and CPS files to fields available only through SSA and Internal Revenue Service records—such as lifetime earnings histories and aspects of program participation not collected in the surveys. The continuing usefulness of these surveys is being undercut by growing sample loss resulting from survey nonresponse, including initial nonresponse and attrition, as well as respondents’ reluctance to provide their Social Security numbers. This report examines sample loss in the SIPP and CPS and finds no evidence that attrition bias or match bias in the linking of administrative records to survey data has increased in the SIPP since the 1996 panel.