Gender pay gap widens at lower end after selection correction

The paper introduces a novel approach to estimate selection-corrected quantiles of the gender wage gap, addressing the bias caused by non-random selection into full-time employment.

Many studies compare wages only among those observed working full-time, but employment itself is not random; individuals with higher earning potential are more likely to work full-time.

Ignoring this selection can lead to misleading wage gaps that vary across different earning segments.

The new method employs instrumental variables that explain variation in latent wages without directly affecting selection, conditional on the latent process.

This allows for semiparametric identification of quantile parameters without imposing parametric restrictions on selection probability.

Estimation is based on inverse probability weighting combined with quantile regression, with selection probabilities estimated flexibly.

Applying the method to German administrative data, the study uses early-career wages as an instrument, capturing persistent individual characteristics influencing long-run earning potential.

Results show positive selection into full-time employment for both women and men, with observed full-time workers having higher potential wages.

Selection is strong for women at the lower wage distribution end and for highly educated men at the top.

After correction, estimated wages are lower, especially for lower-paid women.

This increases the gender wage gap at lower quantiles but slightly reduces it at higher quantiles, particularly among highly educated workers.

These findings highlight the importance of accounting for selection heterogeneity across groups and wage distributions.