Attrition and Income in a Panel Survey: A Comparative Analysis with Cross-sectional Surveys

Abstract

This paper examines whether attrition in a panel survey causes bias in the estimation of individual income. The Japanese Life Course Panel Survey (JLPS), which has been conducted every year since 2007, is analyzed along with the 2006 and 2012 Japanese General Social Surveys and the 2015 SocialStratification and Mobility Survey. Regression results for income are used to compare the JLPS andthese major cross-sectional surveys. The magnitudes of coefficients differ at the initial stage of JLPS, but they become closer as the survey progresses. Even though JLPS has experienced an approximately 40% reduction in the original sample after 8 years, the results do not provide strong evidence that it has become unrepresentative.