The estimation of union wage differentials and the impact of methodological choices

Abstract

This paper demonstrates that methodological differences can matter a lot in the estimation of union/non-union wage differentials. Using individual-level data from the 1991 Wave of the British Household Panel Survey and a model evolved from replicating six existing British studies, we find that the model specification adopted has an important impact on the estimated differential and that the choice of which group means to use when evaluating the mean differential in multi-equation models is of considerable importance. There are also important differences between membership and coverage differentials and the earnings measure used and sample selected also make a difference. However, apart from firm size, the contents of the control vector used is not found to be of great importance.