It would be surprising and novel if the DoL statisticians had produced an unambiguous correct analysis. They're working in a complex field, and under pressure a

It would be surprising and novel if the DoL statisticians had produced an unambiguous correct analysis. They're working in a complex field, and under pressure and regulated to use the wrong tools in many cases---so I'm sure plenty of the people there are excellent and could do great work---DoL statistics in general are not a good basis for action. Example: DoL is required by law to track non-white applicants and hiring to jobs. Alice is not white, and applies to ten open jobs at Example.com, a hot new startup. She's hired for one. DoL reports this as 90% of non-white applicants being rejected. If 10 non-white applicants each apply for all those ten jobs and are all hired, one job each, that's 90% rejection. A mix of law and regulation requires counting this way. Near one edge case (plausibly including SV today!) the error doesn't matter much. But as you get anywhere near justice, this isn't helpful. So without seeing the details of this analysis, I'm skeptical that DoL's way of counting is helpful here.