A Bootstrap Strategy for the Detection of a Panel Attrition Bias in a Household Panel with an Application to the German Socio-Economic Panel (GSOEP)

Abstract

A bootstrap strategy for detecting non-ignorable panel attrition is proposed. The strategy is based on the diflerence of the original estimate and an estimate that is obtained by reducing the original sample by a second attrition experiment. The attrition propensities are estimated from previous wave information and field work information of the current panel wave. The routine may be used to estimate the bias due to panel attrition. The bootstrap routine is applied to two income estimates with data from the first 8 waves of the GSOEP.