The estimation of male earnings under panel attrition. A cross country comparison based on the European Community Household Panel

Abstract

The aim of the paper is analyzing the effect of attrition in the European Community Household Panel (ECHP) on earnings equations. By splitting the completely observed sample according to the response behavior of the following wave, we assess empirically the bias of an un-weighted as well as an inverse probability weighted estimator. Our findings lead us to conclude that the problem of attrition is no matter of great concern when estimating income equations of the Mincerian type based on the ECHP data. While we find in some cases smaller differences in the regression parameters due to attrition, the main findings seem rather unaffected by attrition. Concerning the question of correcting for possible attrition biases through inverse probability weighting, we conclude that the additional variance of the estimated response-probabilities outweighs possible gains from this correcting procedure and does not lead to an overall improvement compared to the un-weighted estimator.