For students who performed equally well on an achievement test given to a sample of 3796 high school seniors, being black or Hispanic will add a little more than .3 years of schooling. (For those of you who care, this result is statistically significant at the 1% level using a 2-sided t-test.) If you ignore controlling this variable, being Hispanic subtracts .2 years and being black subtracts .5 years of schooling. In addition, a regression on students who performed above the 96th percentile shows that race no longer becomes a (statistically) significant factor and expected years of schooling increases to 15.5 years.
Controlling for end-of-high-school academic achievement being the primary corrector the mainstream bias about race and education may not be a surprise for some, myself included. The more interesting phenomenon is the positive correlation between race and schooling after controlling for achievement. I don't have any good idea why this is the case. Feel free to share any hypotheses you may have.
Note: I used STATA to run regressions on this data set supplied by Professor Cecilia Rouse of Princeton University.