Univariate analyses are ridiculously reductionist. For complex social interactions, like occupational pay scales, you simply cannot reduce it to "well these people have penises and these people have vaginas, and that's why they get paid differently". It's silly. There are any number of intervening variables that exert influence on compensation. Any time you see someone complaining about the 'gender wage gap', just ignore them as silly, unscientific ideologues, because that's the only reason someone would put forward such a stupid argument.