Wrong.
In order to be in a position to do it, you need to start with a sample of many thousands of people being hired into a specific kind of job, all of whom have been given the test in question, and all of whom are then judged by standardized metrics. ONLY THEN can you collect enough data to make strong statistical inferences on this issue.
The military can meet this threshold. Very, very few private employers can. And claiming otherwise is wrong.
...the expected utility of general intelligence testing, even where any such benefit is rationally expected to be born out, is pretty low compared to more specific job-focused alternatives that are also easier to demonstrate are job related if they have a disparate impact.
I do not believe this to be the case for most entry-level jobs. Can you cite specific research indicating otherwise?
If it can be established as a fact, its actionable. It can't be assumed to be a fact for the specific jobs it is applied to, however -- and while Griggs may be a leading case, that's pretty clearly the intent of the burden established by the Civil Rights Act for the employer in disparate impact situations.
> But very few companies are in a position to do such research for themselves.
Any company is in the position to do it (or pay to have it done), for very few companies is the expected benefit justify the cost, because the expected utility of general intelligence testing, even where any such benefit is rationally expected to be born out, is pretty low compared to more specific job-focused alternatives that are also easier to demonstrate are job related if they have a disparate impact.