Handling Attrition in Longitudinal Studies: The Case for Refreshment\n Samples

Abstract

Panel studies typically suffer from attrition, which reduces sample size and\ncan result in biased inferences. It is impossible to know whether or not the\nattrition causes bias from the observed panel data alone. Refreshment samples -\nnew, randomly sampled respondents given the questionnaire at the same time as a\nsubsequent wave of the panel - offer information that can be used to diagnose\nand adjust for bias due to attrition. We review and bolster the case for the\nuse of refreshment samples in panel studies. We include examples of both a\nfully Bayesian approach for analyzing the concatenated panel and refreshment\ndata, and a multiple imputation approach for analyzing only the original panel.\nFor the latter, we document a positive bias in the usual multiple imputation\nvariance estimator. We present models appropriate for three waves and two\nrefreshment samples, including nonterminal attrition. We illustrate the\nthree-wave analysis using the 2007-2008 Associated Press-Yahoo! News Election\nPoll.\n