Using SAT scores — Uplink Archive

Journalists discuss the challenges of comparing SAT scores across states, noting that participation rates and demographic factors significantly affect results and that simple statistical adjustments may not effectively normalize the data.

Using SAT scores

These are a few takes from a recent discussion on the NICAR-L listserv regarding using SAT scores. See page fifteen for information on subscribing to NICAR-L.

I'd like to run a map showing SAT scores by state, but I seem to remember that state scores are affected by how many students actually take the test. More students, lower scores. So, is there a way to adjust scores using the information provided - score and percent taking the test? Anna Byrd Davis, The Commercial Appeal, Memphis, Tenn.

Probably the best way to adjust these scores is by running a regression with number of students testing as the independent variable. But I've heard doing this takes so much variation out of the scores that your map would be all one color, or close to it.- Geoff Dougherty, St. Petersburg Times, Fla.

Geoff (Dougherty) is right about the result of using regression, and that's because the larger the population, the lower the variability. The key question may be why, implicitly, larger states have lower scores. The ACT reports a similar phenomenon. In our one high school here and statewide, there is a higher score for those whose coursework points toward college prep. Bring in those who may not have thought about college until later in high school, not completed work, or are simply not up to snuff, and those can also explain the population-score relationship. -George E. Schwarz, Rio Grande Sun, Española, N.M.

There are a bunch of other factors that may affect SAT/ACT scores. For instance, if your state has a high school graduate scholarship program that uses one of the tests as an eligibility requirement, you end up getting more kids taking the test who may not have been college-bound in previous years, or compared to other states. -Mark Schleifstein, The Times-Picayune, New Orleans, La.

Yes. It's not the number who take the test that matters, it's the proportion of high school seniors. The effect comes from the fact that states whose public colleges do not require the test get high scores because those who do take it are trying for competitive colleges. - Philip Meyer, University of North Carolina.