06-28-2015, 09:38 PM
The scariest thing about the SAT is that it measures something that actually seems to have no statistical relation to college success, and yet it is used to make a decision that alters lives. It is hard to find good quality (independent) statistical analysis of the utility of the SAT, but the several studies I have seen over the decades that follow up the success of students in college and compare it to their SAT scores before college all agree on one thing: the SAT score has no statistically significant correlation with any kind of college success: graduation rates, college GPAs, retention rates, etc.
(High school GPA, by the way, does turn out to have a fairly significant correlation--even when considering students from high schools with vastly different levels of rigor. Why? The best hypothesis is that GPA does not measure intelligence or learning so much as it measures self-discipline. A high GPA in high school shows a student who turns in assignments on time, does all the work, isn't truant, and knows how to cram for exams. It is a little sad, but those qualities seem to be more important than having a brilliant mind when it comes to college success--at least the way college is typically taught.)
But even high school GPA is eclipsed by the greatest predictor of college success: Social Economic Status (SES) of parents. It has a high correlation with SAT scores. Since the SAT indirectly measures SES, of course it predicts college success--but if you control for SES when running SAT scores through a regression you will get nearly random results. That is, if you take two white boys from Saratoga California whose fathers and mothers all have advanced degrees and whose families make more than $300k, the difference in SAT scores between them will have no predictive value over which of the two will be most successful in college.
Now, my question is whether all the SAT changes that some say are "dumbing it down" will actually make it better or worse at predicting college success. On the one hand, I am all for changes that remove the influence of SES from the scores. However, I think it is possible that if they really succeed in doing this, they will be left with a test that measures nothing at all, because the entire concept of a test that can fairly and accurately predict college success is spurious.Â
(High school GPA, by the way, does turn out to have a fairly significant correlation--even when considering students from high schools with vastly different levels of rigor. Why? The best hypothesis is that GPA does not measure intelligence or learning so much as it measures self-discipline. A high GPA in high school shows a student who turns in assignments on time, does all the work, isn't truant, and knows how to cram for exams. It is a little sad, but those qualities seem to be more important than having a brilliant mind when it comes to college success--at least the way college is typically taught.)
But even high school GPA is eclipsed by the greatest predictor of college success: Social Economic Status (SES) of parents. It has a high correlation with SAT scores. Since the SAT indirectly measures SES, of course it predicts college success--but if you control for SES when running SAT scores through a regression you will get nearly random results. That is, if you take two white boys from Saratoga California whose fathers and mothers all have advanced degrees and whose families make more than $300k, the difference in SAT scores between them will have no predictive value over which of the two will be most successful in college.
Now, my question is whether all the SAT changes that some say are "dumbing it down" will actually make it better or worse at predicting college success. On the one hand, I am all for changes that remove the influence of SES from the scores. However, I think it is possible that if they really succeed in doing this, they will be left with a test that measures nothing at all, because the entire concept of a test that can fairly and accurately predict college success is spurious.Â