Differential validity of the SAT
In an earlier comment thread someone asked whether Asian-American college performance is commensurate with their SAT scores. If A-A SAT scores are artificially elevated by cramming then one might expect it to under-predict college GPA. (On the other hand, if Asians are more conscientious and hard working overall, one might* expect both SAT scores and college GPA to be elevated relative to other groups.) This data from the College Board shows that the validity of SAT as a predictor of college GPA is about the same for whites and Asians.
*Regarding cramming, I have yet to see any data which shows that large groups of people can significantly elevate their SAT scores through preparation. Test prep companies will claim this is possible, but detailed studies by ETS suggest otherwise. In our U Oregon data set (covering all students at the university over a 5 year period) it is quite rare to see a change of 1 population SD between max and average score for individual students who take the SAT multiple times.
(Click for larger version. FYGPA = Freshman Year GPA.)
