The kicker is "similarly employed". I've seen people with the same job title earn a differential of 3x, but that's simply because one negotiated better and was more economically valuable (had about 1.5 more decades' experience) than the other, yet they had the same job title. I guess we could argue to control for years' experience, but I've seen people who were better at a job after 6 months than those in the same job were after 10 years. There are even jobs where people get worse over time, which isn't intuitive but easy to find examples (e.g. you forget documentation if life gets busy for a year or two, and you're less effective because of that).

I suspect "similarly employed" is indefinable in any realistic sense. I'm not against trying, just very sober about the probability of crafting a policy that outperforms the imperfect but free-ish market.