Do SAT scores accurately predict graduation rates?

I'm not sure if we'll be discussing the issues surrounding basic writing during any of our remaining evenings together, but there was an interesting op-ed piece in today's New York Times about whether SAT scores predict graduation rates more accurately than high school grade-point averages. You can find the editorial here.

The author of the piece, Peter D. Salins, a political science professor and and former provost at State University of New York, notes that, "For some years now, many elite American colleges have been downgrading the role of standardized tests like the SAT in deciding which applicants are admitted, or have even discarded their use altogether. While some institutions justify this move primarily as a way to enroll a more diverse group of students, an increasing number claim that the SAT is a poor predictor of academic success in college, especially compared with high school grade-point averages." Salins concludes that SAT scores do provide an accurate measure of whether or not a student will get a degree or not

I'm not sure that I agree with Salins, although the numbers he presents seem to support his conclusion. For one thing, some of my friends in high school who did poorly on their SAT/ACT tests went to a state school for a year or two to improve their grades and then transferred to a private college--usually one which they could not have attended right out of high school. It seems that Salins' metric would not count those students as having graduated.

There is also the issue of how the education benefits the student. From a purely business-oriented focus, schools will obviously want to keep students attending and paying for their classes for a full four years. But a student may attend a couple of years, land a job that they're happy with and leave. Or decide to travel, or enter the military or peace corp, and defer graduation until they finish. SAT scores may be a good metric to use from a school's cost-benefit analysis, but may not be best for indicating how many students in an incoming class will actually benefit from the schooling they are paying for.