An Attempt to Reduce Survey Costs via Logistic Regression and Paradata

• Activity Number: 383
• Type: Invited
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Frost A. Hubbard*+ and James R. Wagner
• Companies: University of Michigan and University of Michigan
• Keywords: Paradata ; Health and Retirement Study ; Responsive Design ; Institute for Social Research

Abstract

Creating new methods for reducing survey costs and errors is an increasingly important issue as survey response rates continue to decline. To reduce survey costs on the Health and Retirement Study, we developed a logistic regression model that predicts the likelihood of a sampled address completing the screening interview.

The screening completion propensity model will be used in two ways. First, the propensity scores for all cases will be calculated each day and matched with the cases the interviewer attempted the previous night. With this we will offer suggestions to make their contact attempts more efficient. Second, when selecting a two-phase sample, non finalized cases with higher propensities to complete the screening interview will be oversampled to increase the efficiency of the second phase sample.

To determine if the screening propensity model helped reduce costs, we will compare cost metrics from before the use of the model and after. Finally, we will discuss how to further this process so that it also helps reduce the potential for nonresponse bias by using predicted values of key statistics of interest.