What's linked is an article (another from NPR) quoting a college official as claiming that, which, it's true, isn't research. However, the linked article itself links to the particular piece of research that both articles are referring to, which contains:
"Furthermore, our data indicated that high school GPA had a stronger correlation with college success for Non- Submitters than the ACT/SAT (for the 27% of Non-Submitters for whom we had test scores) -- both in terms of college cumulative GPA and graduation rate. While test scores had a generally stronger relationship with college GPAs for the Submitters, for the Non-Submitters they tended to show a weaker relationship, essentially under- predicting the college GPA. The test scores continued to most strongly correlate with family income."
So they may be right or may be wrong, and the research may be good or it may be crap, but NPR does not appear to be misrepresenting its own sources.
> for the Non-Submitters they tended to show a weaker relationship, essentially under- predicting the college GPA
I wonder if that's at least partially due to non-submitters choosing not to submit because they did worse than expected on the test, compared to how they did on practice tests. Even assuming no exogenous events like a bad night's sleep, test anxiety, etc, your SAT score is still an estimate of your "true" score, with some variance. If you have an expected score due to taking practice tests, and your measured score is significantly lower just because you happened to fall on the left part of your own personal score distribution for the real test, maybe you decide not to submit your score. If that happens systematically, but SAT score really is correlated with success, you would expect nonsubmitters to outperform their score.
Spot on. An interesting point to explore would be to see if students who has a low score and took the test again and got a higher score, and has similar high school performance, performs at the same level as the non-submitters with a low score.
> Furthermore, our data indicated that high school GPA had a stronger correlation with college success for Non- Submitters than the ACT/SAT (for the 27% of Non-Submitters for whom we had test scores)
The problem is, is high school GPA measured the same way across different states and schools?
> The test scores continued to most strongly correlate with family income.
I have no doubts about this. But I'm pretty sure you would get the same result by looking at family stability and parent's educational achievements.
Sure, I've only dug into this far enough to be pretty sure NPR isn't misrepresenting what the research they're citing found. The research itself may not be very good, for a bunch of possible reasons. I don't know.
Global correlation with college GPA is a completely useless metric because it doesn't control for the fact that higher SAT's get you into harder schools.
Again, it's entirely possible the linked research is garbage, that's just a different thing from NPR stating something at-odds with what its own source claims.
"Furthermore, our data indicated that high school GPA had a stronger correlation with college success for Non- Submitters than the ACT/SAT (for the 27% of Non-Submitters for whom we had test scores) -- both in terms of college cumulative GPA and graduation rate. While test scores had a generally stronger relationship with college GPAs for the Submitters, for the Non-Submitters they tended to show a weaker relationship, essentially under- predicting the college GPA. The test scores continued to most strongly correlate with family income."
So they may be right or may be wrong, and the research may be good or it may be crap, but NPR does not appear to be misrepresenting its own sources.