The SAT and ACT are predictors of success in college, and are quite good at it, but this is all they are meant to be. They need not at all measure whether one is good at mathematics, language, etc., still less whether they know what is most worth knowing in mathematics or language. But it is exactly these sorts of questions that teachers and designers of curricula need to ask.

I wonder if we might be better off judging college aptitude by completely random questions, the way the MMPI measures personality. Take ten thousand college students, ask them random questions (e.g. “I am bothered by constipation” or “I would like to be a singer”) and figure out which random questions select the successful and the unsuccessful.

Sure, it wouldn’t fly, but this is not for being less rational. But it would also free up high schools to teach what they think should be learned, without having to teach to a test; or free up colleges to have their own curriculum without having to keep a constant eye on grad schools, med schools, etc.



  1. November 19, 2014 at 1:48 pm

    Actually, the SAT and ACT are very poor predictors, no better than random guessing. See here.

    High school grades are much better.

    • November 19, 2014 at 1:52 pm

      Excellent! My information was old. Thanks

      • Paradox said,

        November 19, 2014 at 3:04 pm

        Well that’s not going to fly. It means that Ball State, where I had originally intended to transfer, is going to decide whether to let me in according to a meaningless test.
        Oh well, I might go to a different college anyway….

        Take care.

  2. PatrickH said,

    November 20, 2014 at 10:09 am

    Nope. The study did not show that the ACT and SAT are not good predictors, no better than random guessing. It was comparing test optional and non-optional admissions policies, not high school grades only, let alone random guessing. The Commission quoted in the article said standardized tests should not be used as sole predictors. They did not claim that standardized tests were no more predictive than random guessing. They mentioned “other tests” also being more useful because they are more achievement tests, like AP, International Baccalaureate, etc.

    If you want to show that standardized tests are no more predictive than random guessing, that’s what you compare, not test optional vs standardized test required admissions policies.

    I think you should strike out the strikeout.

    • November 20, 2014 at 12:10 pm

      I migh be missing your point, but the MMPI is not random guessing, but a predictive tool based on questions originally random but which are found to be good predictors of something (though no one knows why). If, for some inexplicable reason, we can better predict love of being an airplane pilot from ice-cream preferences, then it’s better to have a career exam for pilots based on ice-cream preference than on another test that tried to gauge our interest in flying, and other apparently obvious factors.

      • Peter said,

        November 20, 2014 at 6:53 pm

        To be rational, a college entry test needs to actually predict on the basis of scholastic aptitude and not merely irrelevant factors that happen to be statistically correlated with it. Otherwise you will likely find that your test is measuring things such as upbringing, cultural background, family income, social class, language spoken at home etc. Which may all perform well as predictors of academic success but don’t necessarily make for just entry criteria.

        In the case of a pilot exam, if we did find that for some reason great airline pilots tend to prefer vanilla icecream, do we really want to exclude someone who actually is a potentially good pilot who just happens to prefer chocolate?

      • PatrickH said,

        November 23, 2014 at 8:23 am

        Sorry, James. I was responding the first comment’s claim that the SAT and ACT do no better in predicting academic performance than random guessing (or high school GPA). The random *question* MMPI style approach is an intriguing possibility, although I have no idea how it could be implemented or if it would have any predictive validity at all.

        But the first comment said that the ACT and SAT are no better than random *guessing*, and that is not supported by decades of testing results. They are good at predicting college academic performance. And they are good at predicting mathematics performance and so on. I take it that random guessing is just doing the equivalent of throwing darts at a board with student IDs pasted on it.

%d bloggers like this: