In case of discrimination alegations (race, disability and veteran-related matters in the US), the burden is on the accused to provide proof of non-discrimination.
In this case, the usual practice for proving innocence is to show that the statistical distribution of hires is similar to the statistical distribution of applicants. For example, if I had 50% latino applicants, 50% of my hires should be latinos.
That being said, the sort of publicity these issues get mean that companies try really hard to steer clear of any discrimination accusations. It's a delicate problem to handle.
I dislike employment anti-discrimination laws for this and other reasons. I suggest a compromise of limiting it to manual labor jobs and the like for which the anti-discrimination laws were actually designed for.
At a well-known Java school in the Southern US the student body is ~ 2/3 Indian and 1/3 Chinese, the secretarial staff is 1/2 white, 1/2 Latino, and the tech support staff is 100 % white. It's an odd distribution, no matter how much you dislike anti-discrimination laws.
In places as rotten as that one you'd ideally fire everyone and rehire from 4 states away. Antidiscrimination laws are weak tea, but the only tool we have. You sit and hope that the University President takes note after the first suit and fires management.
In this case, the usual practice for proving innocence is to show that the statistical distribution of hires is similar to the statistical distribution of applicants. For example, if I had 50% latino applicants, 50% of my hires should be latinos.
That being said, the sort of publicity these issues get mean that companies try really hard to steer clear of any discrimination accusations. It's a delicate problem to handle.