Residential mobility and panel attrition: using the interview process as identifying instruments

This paper aims to assess the extent of attrition bias in the Keio Household Panel Survey (KHPS) 2004-2006, focusing on household residential mobility. We use the unique dataset of the interview process as identifying instruments for a sample selection model. The results show that sample attrition does lead to statistically significant bias in wave 2, but it seems to be attenuated in the later wave. The results of the bias analysis for each estimated coefficient are also presented. Reprinted by permission of Keio Economic Society