Black-Swan Wage Premiums? Evidence from Expectile Regression

Abstract

We estimate wage premiums (male, union member and public sector) in West Germany using post penalized expectile regression. The estimation relies on data from the Socio-Economic Panel (SOEP) 2010-2017. We use a double robust LASSO as penalization technique, which allows us to deal with potential omitted variable bias and to conduct expectile-specific model selection. In addition, we enrich our study by comparing our estimates with the ones derived from a competing approach: the quantile regression method. We find that, depending on the underlying estimation method (expectile or quantile regression), the estimated wage premiums may differ substantially. Particularly, the premiums in the tails exhibit stronger sensitivity to the estimation method. The analysis detects `black-swan' wage premiums, i.e. significant premiums not uncovered otherwise.