> If talent is not correlated to race/sex, then we can expect that the number of people of various races/sexes that are represented in the industry should rough

> If talent is not correlated to race/sex, then we can expect that the number of people of various races/sexes that are represented in the industry should roughly match the general makeup of the population (ie it's a mostly random distribution). Employer hiring should match the candidate population, not the general population. This is a subtle but very important difference when we also see, for instance, gender bias in graduation rates of STEM fields. (I'm not claiming that this accounts for all bias.) Really, when you say "there is a bias somewhere in the process", it must be remembered that the "process" being spoken of is the entire process that leads to a person being a valid candidate, from schooling to prior experience. (It's hard to get hired as a senior if you can't get hired as a junior first.)