After first controlling for the effects of age, race, gender, union status, and location on ...

Cognitive ability and years of schooling are likely positively correlated.  Higher ability may translate into greater labor market productivity and therefore above average earnings. By not holding constant for cognitive ability in the analysis, it's entirely possible that at least some of the observed 8% rate of return is, in reality, a return to cognitive ability rather than to additional education.  If this is the case, then the rate of return estimate exaggerates  (there is upward bias) the effect of schooling on earnings.