Measuring the true returns of education is impossible, both on the individual level (we can't replay history with alternative choices) and the population level (clean randomisation on education is socially unacceptable), so all we have are approximations and methodological hacks (technical summary in linked summary paper). But such approximations typically found that "ability bias" was not all that large.
Someone ordered that every household should have books at home while failing to realize that correlation != causation…
https://economix.blogs.nytimes.com/2009/08/27/sat-scores-and...
And last I checked, there is a strong correlation between maternal educational achievement and child educational achievement even after controlling for various things including wealth. The correlation with paternal achievement when controlling for other factors is much lower.
How do you know this possibility has been left unconsidered?
But what was not considered was that the causality here could be incorrect, instead showing us that people with traits optimized for success in life chose to enroll in higher education to use the additional knowledge and experience as a catalyst for their success (or they just otherwise loved academia or learning).
Due to the extent of which this first hypothesis was treated as a fact, we are left with a system that attempts to reinforce it at many levels of society, ignoring the large cost that it takes.