This is not necessarily the case. This is a vast oversimplification of a tech company's hiring process, but for the sake of simplicity let's say that phone interviews have a 50% rate of generating a false negative and on-sites have a 0% error rate.
Say a company does one phone-interview, and if passed, moves on to on-sites. But for diverse candidates they do two phone interviews and if either passes they move on to on-sites.
* Unqualified candidates always fail, since neither phone interviews or on-sites have a false-positive rate.
* Qualified non-diverse candidates have a 50% chance of getting the job.
* Qualified diverse candidates have a 75% of getting the job.
At no point do any unqualified candidates receive offers. But at the same time, qualified non-diverse candidates are still rejected at twice the rate as diverse candidates.
The "bar" so to speak isn't always something that people have a higher chance of passing the better their skills are. A working knowledge of data structures and concurrency is all that's necessary to pass all of my company's coding interviews. Practice and luck is probably a greater asset than skill in many interviews. But getting into the efficacy of technical interviews is opening a whole 'nother bag of worms.