Panel: Using Administrative Data to Improve Survey Data Quality
(Tools of Analysis: Methods, Data, Informatics and Research Design)

Saturday, November 4, 2017: 1:45 PM-3:15 PM
McCormick (Hyatt Regency Chicago)

*Names in bold indicate Presenter

Panel Organizers:  Katherine Giefer, U.S. Census Bureau
Panel Chairs:  Joanna Motro, U.S. Census Bureau
Discussants:  Bruce Meyer, University of Chicago

Program Confusion in the 2014 SIPP: Can Detailed Data be Retained?
Katherine Giefer and Joanna Motro, U.S. Census Bureau

Sampling with Administrative Records in the National Survey of Children’s Health
Scott Albrecht, Jason Fields and Keith Finlay, U.S. Census Bureau

Ignorable Nonresponse? Improved Imputation and Administrative Data in the CPS Asec
Charles Hokayem1, Trivellore Raghunathan2 and Jonathan L Rothbaum1, (1)U.S. Census Bureau, (2)University of Michigan

Using Administrative Records and Parametric Models in 2014 SIPP Imputations
Veronica Roth and Joanna Motro, U.S. Census Bureau

Nationally representative public use survey data is widely used by researchers and policy makers to identify problems and recommend policies across numerous arenas. However, nonresponse and misreporting is a growing problem in survey data. The papers in this panel focus on how surveys are using administrative records to improve their research design and edited variables on public use data sets to help reduce the impact of these growing issues. The combination of survey and administrative data leverages available information leading to better quality data, which in turn allows researchers and policy makers to make decisions that are more informed.

The papers in this panel combine survey and administrative data in multiple ways. One paper shows how comparing administrative records to survey responses in the Survey of Income and Program Participation (SIPP) can be used to correct and retain respondent data when respondents confuse two government assistance programs, Social Security and Supplemental Security Income (SSI). The second paper shows how the National Survey of Children’s Health (NSCH) uses administrative records and the American Community Survey (ACS) to increase the NSCH’s sampling efficiency. The last two papers provide evidence that using sequential regression multivariate imputation (SRMI) and administrative records in imputations improve survey data.  One of these papers shows how these methods and data can reduce bias when imputing income values in the Current Population Survey Annual Social and Economic Supplement (CPS ASEC). The other paper provides evidence of how using SRMI and administrative records in the 2014 SIPP helps preserve the relationship between income and program receipt and mitigate the problem of data not missing at random.