Emil O W Kirkegaard

The beta for IQ is somewhat stronger here, but p>a due the model having too little free variation to estimate precisely (large SEs).

Ok, fair. But Im not sure this settles the question of direction of causality. That will perhaps require an additional study.

It could be done using an experimental setup where subjects are given ambivalent info and asked whether they were discriminated against.

Prediction here being that the less bright will be more likely to ascribe it to discrimination holding the ambiguity constant.

I like this but I think at best it reveals what the first reg more or less confirms (association btwn IQ and perceived discrimination).

But if we want to know whether lower IQ causes discrimination I think it has to be probed a diff way. Rough example: give respondents a

passage "written by" a representative of each SIRE/skin_tone category and vary this passage in terms of quality (keeps content constant)

then ask respondents to rate IQ of passage authors and likelihood of engagement in stereotypical behavior by authors. Check for correlation.