Evaluating the Survival Analysis Method for Estimating Residency Rates

• Activity Number: 105
• Type: Topic Contributed
• Date/Time: Monday, August 9, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Barbara L. Carlson*+ and Daniel Kasprzyk
• Companies: Mathematica Policy Research, Inc. and Mathematica Policy Research, Inc.
• Keywords: RDD ; CASRO ; response rate ; AAPOR ; Community Tracking Study

Abstract

To calculate a response rate for a survey, each sample member must first be classified according to its eligibility status. For telephone surveys with random-digit-dial samples, we try to determine for each sampled telephone number whether it is in service and belongs to a residence. The residency status of a certain portion of the sample, however, is never resolved even after making many call attempts. Several methods have been proposed to estimate residency among unresolved telephone numbers. The most conservative of these assumes the same residency rate among the unresolved cases as was found for the resolved cases. This inevitably overestimates the number of eligible telephone numbers and unduly reduces the response rate. A newer approach proposed in the literature tries to estimate a residency rate by modeling the "time to resolution" using survival analysis. The numbers designated as residential, nonresidential, and unresolved at each attempt are incorporated into the model. We apply a variation of this methodology to the HSC Community Tracking Study Household Survey and report on how it compared to the more conservative method of estimating residency.