A Response Propensity Modeling Navigator for Paradata

• Activity Number: 321
• Type: Invited
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Young Chun*+ and Mike Kwanisai and Linda Hamilton
• Companies: NORC and NORC and American Institutes for Research
• Keywords: paradata ; nonresponse ; propensity model ; NAEP ; complex surveys ; pandata

Abstract

The purpose of this paper is to identify factors affecting nonresponse of 12th graders in the National Assessment of Educational Progress, by using social isolation as a theoretical navigator. We also evaluate the statistical impact of nonresponse bias on NAEP estimates by taking advantage of response propensity models built on a social isolation framework. We use the 2000 NAEP science survey data and its contact history paradata, both of which are linked to the school administrative data from over 20,000 seniors in the 2000 High School Transcript Study whose sampling frame is identical to NAEP. We apply the final robust response propensity model to reweight NAEP estimates with additional covariates extracted from the HSTS administrative data. We introduce the concept of "pandata," the muiply-linked data including paradata and administrative data to improve nonresponse adjustment.