31 Dec 2011

If not this, then what?

Submitted by Karl Hagen
Standardized tests in general, and the SAT in particular, get a lot of bad press. Companies like Princeton Review build their entire marketing strategy on trash talk about how horrible the test is. Organizations like Fair Test campaign for abandoning the use of the SAT (and the ACT) in college admissions, claiming that it is both biased and ineffective.

Now there are some ways in which I sympathize with this attitude. The experience of taking the SAT is understandably stressful, and I sympathize with those who feel oppressed by the whole process. Also, I completely agree with Fair Test that badly constructed standardized tests are pernicious, and bad uses of standardized tests--whether they are well written or not--create significant harm.

I want to make the case, however, that the SAT is not badly constructed. When examined carefully, it actually turns out to be an extremely good test. Sure, you can argue about alternate question formats, but given the basic format, I would claim that it does about as good a job as it is possible to do. To my mind, the tests that deserve the greatest criticism are those used for NCLB accountability.

Most of the criticism about the SAT is either outdated or based on an inadequate understanding of how tests work and what they can accomplish. Moreover, I will argue that while SAT scores are in some cases put to inappropriate uses, any careful consideration of alternative regimes is bound to lead us to a system that, at least in its general outline, resembles the one we already have.

I will make my case in a series of posts, but the basic question I will pose is this: "If not this, then what?"

In other words, if there's something you don't like about the SAT, or college admissions testing in general, how would you do it better?

Perhaps you think that there should be no standardized testing at all in college admissions. Many schools don't require the SAT or ACT now. But they aren't the most selective institutions. At the schools that are highly selective, how do we fairly compare grading standards at different schools around the country? What do you do about grade inflation? And if you think grade inflation is bad now, what will happen in a world where course grades are the sole academic index used for college admissions?

Perhaps you concede the need for standardized testing, but think that the test is unfair. OK, so how would you do it better? What steps would you take to eliminate bias that aren't currently being taken? This description of the ACT's fairness-review process is typical of what all the most prominent testing programs do these days. How would you improve on it? If you take the stance that all standardized testing is inherently biased, I would ask you to quantify the bias of non-standardized testing (IOW, the ordinary process of giving in-class tests and other assignments). How many teachers out there are consciously or unconsciously skewing their students grades because of different kinds of bias? What is the grounds for believing that ordinary grades are less biased than standardized tests?

Perhaps you think the test is not sufficiently predictive of college performance. OK, what can you suggest that would improve the decision-making ability of college admissions officers?

Perhaps you think that the multiple-choice format is weak and we should substitute some other question type. Or perhaps you have problems with the specific content tested. Again, how would you do it better?

Perhaps you think the content is OK but find many questions to be flawed. Show me that you can write better questions. (It's clear from the published practice tests put out by test-prep companies that few people can.)

In future posts, I'll tackle each of these issues in more detail, but in all of them, I want you to consider what the consequences of doing things differently would be.