Using Call-Back Data to Adjust for Nonignorable Nonresponse: Results of an Empirical Study

• Activity Number: 685
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Paul Biemer*+ and Kevin Wang and Patrick Chen
• Companies: RTI International and RTI International and RTI International
• Keywords: selectivity bias ; drug use ; EM algorithm ; ML estimation ; paradata

Abstract

The number of call attempts can be used in response propensity models to reduce nonignorable nonresponse bias in sample surveys. In applying these models, a number of issues arise from the definition of process variables like "call attempt," "contact/noncontact,", and "nonresponse" to callback censoring, unrecorded call attempts, and other data errors. As will be demonstrated, these problems can reduce the fit and utility of the models. Despite these problems, callback models can substantially reduce the nonignorable nonresponse bias, particularly when data on nonrespondents are quite limited. This presentation will review some approaches for employing callback models for nonresponse bias adjustment and analysis. Some of the problems in using call-back data in nonresponse adjustments will be described and the potential of the approach will be demonstrated for a large field survey.