We estimate the public wage gap in France for the period 1990-2002, both at the mean and at different quantiles of the wage distribution, for men and women separately. We account for unobserved heterogeneity by using fixed effects estimations on panel data and, departing from usual practice, allow the public wage markup to vary over time. We also provide one of the very first applications of fixed effects quantile regressions. Contrary to common belief, results convey that monetary returns are not fundamentally different in the public sector. Firstly, public wage ‘premia’ (for women) or ‘penalties’ (for men) are essentially the result of selection. After controlling for unobserved heterogeneity, only small pay differences between sectors remain over time, reflecting fluctuations due to specific public policies and to the pro-cyclicality of private sector wages. The long-term difference is essentially zero. Secondly, the relative compression of the wage distribution by the public sector is also partly due to unobserved characteristics. The most natural explanation for these results is that the civil sector manages to attract better workers in the lower part of the distribution, in part because of non-monetary gains (including job protection), but fails to retain the most productive ones at the top.