Multiple Imputation for Missing Data in the KLoSA Study

Most survey data include missing values due to nonresponse. Especially, sensitive questions such as income or assets tend to show higher percentage of missing values. When missing values occur, complete-case analysis may lead to biased estimates in parameters. Korean Longitudinal Study of Aging(KLoSA) is a longitudinal study to evaluate aging trends in the Korean population and apply the results to the social welfare and labor policy. In 2006, KLoSA collected baseline data. We conduct multiple imputation based on hotdeck to handle missing values in the KLoSA baseline data. In this study, we explain the imputation method for filling in missing values and discuss the results of imputation.