IQ can be measured in middle school (before the student has a job or income), and the resulting score can be used to predict future income and job performance. In that sense, it is a predictor rather than a mere association.

Ok, but there are still myriad confounding factors, even at that age. Do students from higher income schools in aggregate have higher IQs than those from lower income schools?

All of these things are controlled for.

These are professors whose job it is to make sure you don't have such confounding variables.

You even have twin studies where one twin was adopted by rich and one by poor parents and it still holds.

What about Stoolmiller in Psychol. Bull. 125:392-409 who found families deciding to adopt forming a pointedly restricted range. Being for example exceedingly intelligent with IQ over 120 on average, which is obviously different than general average.

Source?