Effects of sample attrition and data analysis in the retirement history study

Abstract

Bias can be created through the attrition of respondents in panel studies of older people. After a review of previous studies, a change in the method of examining attrition is suggested. Tests comparing those eligible to respond and those actually continuing in the panel are reported for five waves of data from the Retirement History Study. Results indicate that this data set is relatively free of bias caused by respondent attrition. Although caution is urged, attrition is suggested to be less of a factor in this data set and in panel studies in general than earlier studies involving samples of older individuals might indicate.