The Slate article makes some good points. On the other hand, I almost stopped reading at this nonsense:
In a four-year study that started with nearly 3,000 college students, a team of Michigan State University researchers led by Neal Schmitt found that test score (SAT or ACTâ"whichever the student took) correlated strongly with cumulative GPA at the end of the fourth year. If the students were ranked on both their test scores and cumulative GPAs, those who had test scores in the top half (above the 50th percentile, or median) would have had a roughly two-thirds chance of having a cumulative GPA in the top half. By contrast, students with bottom-half SAT scores would be only one-third likely to make it to the top half in GPA.
This tells me almost nothing about a test's effectiveness, other than it can differentiate pretty well between "high-achievers" and "total morons." I could probably come up with a "test" that could satisfy this stat by talking to each student for a minute.
But hey, I gave the Slate article author the benefit of the doubt, so I tracked down the actual article he cited in this paragraph, which provides data summarized in more helpful and less dubious ways.
You can read that article, if you like, because it presents a much more nuanced picture. Long story short: correlation of SAT/ACT performance with college GPA is at a level of 0.53. That's exactly the same correlation (0.53) of high school GPA with college GPA. Meanwhile, high school GPA and SAT/ACT performance correlation with each other is 0.58.
Unfortunately, I can't seem to find a more thorough analysis in this article about where the mismatches occur. THAT would be really interesting. What are the characteristics of students who get high GPA in high school but low SAT scores -- how do they perform in college? And the reverse: high SAT, but low GPA? That's the only way we could actually tell how much information the SAT is actually adding... but alas, most such analyses don't look at the data that way.
Also, college GPA isn't everything. You also need to, well, FINISH college. According to that cited article, turns out high school GPA is a much better predictor in this regard than SAT scores. Having a better high school GPA gives you an odds ratio of 3.77 to actually graduate college, while having a high SAT has odds ratio of only 1.3. (This isn't mentioned in the Slate article, which only notes how high SAT correlates with graduation -- well, yes it does, but nowhere near as well as high school GPA.)
Anyhow, you can take from this what you will. I've read a number of such studies on SATs, and my conclusion is a little different from the Slate guy. Yes, SATs are correlated with college achievement. But so are some other things too (like high school grades). In borderline cases, having the SAT score may help with an admissions decision, but is the level of correlation high enough to justify letting student A in with an SAT score of X, while rejecting student B with an SAT of X-50? That's really the kind of question we need to ask if we want to justify the rampant use of SAT scores in admissions.
And I'm not sure if any research out there really can answer that question well. Certainly not most of the citations in the Slate article, which focus on broader correlations, and definitely not the Slate article itself, which cites some stats that barely qualify the SAT as useful as if they were truly revelatory.
SAT scores correlate closely with measured IQ
Here's the other problem -- the SAT was originally designed as a proxy for IQ, more or less. But over the past few decades, each reform of the SAT has actually moved it AWAY from an effective IQ proxy. The SAT used to have more than basic vocabulary: it used to have things like analogies and antonyms, which require a much more subtle comprehension of language that correlates more strongly with basic reasoning skills. But those were dropped, and recently the vocab level has been reduced further -- not that advanced vocabulary correlates exactly with IQ, but it's one way of measuring student's experience with reading as well as reasoning skills in being able to guess what a word might mean through context, etymology, etc. (Again, these are more basic reasoning skills that are generally not taught directly in school.) Similarly, the math section used to contain things like quantitative comparisons, which also requires reasoning skills... not just repeating the basic arithmetic and symbolic manipulation you learned in algebra I. But those were done away with too.
Basically, the SAT used to be like an IQ test, in that it measured "basic reasoning ability" which sometimes might not been seen in high school performance or grades as well. But reforms over the past few decades have made the SAT much more like a basic high-school test (probably in response to SAT critics who complained that kids "didn't learn these things in schools").
And if the SAT is basically just a high school test now, then I expect its correlation with high school GPA will continue to rise. And that makes us again ask the question: are we still justified in having people spend hundreds of millions of dollars each year in test fees and test-prep courses for a glorified high-school verbal and algebra test?