Gender inequality in academia is well documented yet persistently difficult to eradicate. Here, we adopt a new perspective by investigating the effect of institutional design on gender gap in research productivity on the example of habilitation - a centralized performance assessment procedure. Using a new, population-based dataset comprising 2,704,688 researchers from 45 countries we show that the presence of a habilitation has unexpected gendering counter-effect on research productivity, as it systematically lowers performance of female, but not male researchers. We also show that while international mobility improves research productivity of both genders in both systems, in systems without habilitation the selection effect for mobile women is particularly strong, and their returns to mobility in terms of increased productivity are the smallest of all mobile groups. We conclude that centralised, top-down regulation of `excellence in science' causes irreparable damage to careers of female researchers.