A Study of Survey Nonresponse Bias Using Propensity Score Adjustment
Author: Su-hao Tu, Wan-yun Lo, Yung-tai Hung
Abstract / Chinese PDF Download
It is common to use socio-demographic variables for weighting, but this doesnot mean that other inferences, especially on attitudes and behavior variables, willbe free of bias. This article takes voting behavior in a presidential election as anexample to examine the effects of adjusting the telephone survey results using subclassificationson the propensity score as suggested by Lee (2006). We first used theTaiwan Social Change Survey (TSCS, 2004) wave 4 module for a citizenship studyto produce a pseudo-population dataset based on bootstrapping sampling. A randomsample of 200 cases was drawn from this dataset. In addition, a stratified randomsample of 800 cases was drawn from a telephone survey. The two samples werecombined into a sample of 1,000 cases. Propensity Score Adjustment (PSA) allowsus to adjust and evaluate the proper estimation of voting behavior based on the subclassificationof propensity score. The procedure was repeated 2,000 times. Theresults showed that the PSA method does effectively reduce telephone surveyestimate bias.