If gathering salary data was high cost they could still address the question of having a wage gap with random sampling. And that's what they'd have to do anywa

If gathering salary data was high cost they could still address the question of having a wage gap with random sampling. And that's what they'd have to do anyway -- salary data wouldn't tell you if they had a discriminatory wage gap.