While your point is theoretically valid, the evidence in the academic literature suggests that the causal effect of education on earnings is roughly the same as

While your point is theoretically valid, the evidence in the academic literature suggests that the causal effect of education on earnings is roughly the same as one would get by doing a naive regression of earnings on education at the individual level. There's a bias with respect to ability, but it's on the order of 10 percent, and measurement error in education variables attenuates the naive estimates, balancing things out. There are a number of natural experiments that caused specific groups to get more education than others. This evidence has led to a pretty broad consensus that the returns to one additional year of education are about 10 percent. See: http://www.demog.berkeley.edu/~ebenstei/litreview/Handbooks/Card_HB.pdf http://www.demog.berkeley.edu/~ebenstei/litreview/Handbooks/...