1. The piece confuses two different points: (1) it is not optimal for some students to go to a four-year college, including some of those who presently do and (2) it is not optimal for anyone to go to college. The first point is clearly correct and worth highlighting. The second is clearly, given the large literature on the effects of higher education on life outcomes, incorrect.
2. The optimal number of students who start but do not finish four-year college is not zero. College, particularly four-year college, is what economists call an experience good, which means that you cannot fully evaluate it without actually consuming some of it. I would agree that current dropout rates are probably too high, but that implies a call for research on how better identify students unlikely to finish based on their applications, something that colleges would actually like to do better. Those high dropout rates are not good publicity; for colleges that depend on the goodwill of the local state legislature, good publicity is very important.
3. Contra Richard Vedder, the literature that estimates the impact of college actually does take account of ability differences between those who do and do not attend college. Even an hour of research by Stossel's underlings would have made that quite clear. And Vedder surely knows better and should be spanked. The correct criticism to make is that estimates of the average impact of college on those who attend are often naively generalized to those who do not presently attend.
4. The problem with Hilary Clinton's "extra million dollars" argument or, more precisely, with similar numbers actually based on the literature, is its failure to discount, not its failure to take account of ability differences. Again, a bit of research, or perhaps talking to someone less ideological than Vedder and Riley, would have given Stossel a good criticism to use in place of all the stupid stuff.
5. There is plenty of scientific evidence that going to a higher quality university increases later earnings. You don't just have to take Harvard's word for it, as the Stossel piece suggests.
6. The research that pays tends to be research that matters. Most people who write pieces on "obscure topics for journals that nobody reads" are not getting grants to do it.
Shame on Stossel for putting this rubbish out and on my friends at reason for putting it on their website.