Various quantile regression approaches are implemented to analyze thecharacteristics of Italian data on earnings in the tails. A changing coefficientspattern across quantiles shows increasing returns to education along the wagedistribution. A quantile decomposition approach shows that higher educationgrants higher return at all quantiles, thus implying additional, non-linear returnsto higher education throughout the entire pattern of the earning distribution.Wage gender gap displays a decreasing pattern across quantiles, and it doesnot disappear at the higher quantiles. The southern workers penalty decreasesacross quantiles as well for highly educated workers.