摘要：Ordinary least squares (OLS) estimates of the impact of gender on earnings are potentially biased owing to nonrandomness
in sample selection. In this note, OLS estimates are compared with the results of two methods that have
been proposed to allow for these selection effects – first Heckman's method and secondly a novel approach based on
quantile regression promulgated by D'Haultfoeuille et al. (2018). Estimates are provided for 18 countries over a recent
three year period. Differences between the results obtained using the alternative methods are highlighted and
explained, with lessons drawn for the application of these techniques in future exercises.
