I think what GP is getting at is that you have to separate selection effects of university admissions from actual effects of the university. The people who don

I think what GP is getting at is that you have to separate selection effects of university admissions from actual effects of the university. The people who don't attend university are not in aggregate equivalent to the people who do - I think I remember seeing a study once that people who were admitted to MIT but chose not to attend had similar career outcomes (and incomes) to people who did end up attending MIT (as one example). In that case the signal was the selection effect of being admitted to MIT, not anything that MIT specifically did to help people make more money over the long term. So attending University might look good on average and lead to higher career earnings, but it could just be a selection effect of people who would have made more money either way.