Detecting and correcting attrition bias in longitudinal family research

Bias due to the attrition of respondents poses a threat To the internal and external validity of findings in longitudinal family research. Differences in characteristics between two samples at initial and subsequent waves is a threat to external validity, and differences in the relationships between variables in the samples is a threat to internal validity. Methods of detecting attrition bins are discussed, and Heckman's procedure to correct attrition bias is presented. Data from the University of Southern California Longitudinal Study of Generations (N = 2,044) are used to illustrate the detection and correction of attrition bias.