Eye color, unlike Race or Gender, is pretty evenly distributed over the obvious confounding variables like "Age" or "Preference of staying home with children". I'd expect it to be +/- 10%, though probably not "equal" enough to keep "disparate impact" folks from calling it out.