What they're counting matters. It seems if Google has the goal of hiring more females than industry average, then the stats are stacked against them when you look at their worker pay distribution, even if Google isn't doing anything wrong. Suppose the population is distributed in their experience and productivity, and that you can measure these to some extent. Suppose your field has way more males than females. This…
If you are fairly paying both genders, it's fair to assume that women would be paid less statistically.
They usually retire from work mid-career for pregnancy and possibly for a much longer time to raise the child (when it's not the man doing it, and when the couple don't have a nursery/grandparent, etc...).
Retiring from work at the peak time for career growth is a disastrous move on one's income. (Be it a man or a woman).