Fair enough. However, there are statistical deviations that are significant enough that it's fair to consider them.
In the US, 90% of men are incarcerated. Sure, maybe it's just because men are suspect more women, therefore are caught with higher likelihood. But, given a random citizen, is the probability that they will commit a crime not significantly higher if they are a men? (Which is of course near 0 anyway, don't get me wrong.)
There are your ratios.
See the problem?