Quite often those differences exist because of systemic or cultural bias that affects the test design. Tests are often validated based off of other tests that showed a difference, but those tests often had a severe sampling bias that showed a group difference where non-existed. It then became an established theory that if you design a test that measures e.g. “emotional intelligence” (whatever that means) and it didn’t show a group difference, it was invalid and had to be adjusted until it did.