SAT’s not fair predictor of college success

Across the country, high-school seniors are biting their nails, wondering whether they will get into the college of their choice.

Coincidentally, the Supreme Court heard oral arguments yesterday on the University of Michigan affirmative-action cases. A top issue: Is affirmative action a necessary counterbalance to the bias of standardized tests?

Both events spotlight the far-too-powerful role of standardized tests in deciding who gets what in this society. It’s time, in fact, to dump the SAT and ACT as requirements for college admission – as more than 400 colleges have done.

The SAT, which is the granddaddy of college-entrance exams, purports to demonstrate merit and ability and to predict first-year college grades.

Nonsense. How well you do on the SAT has far more to do with your family’s income – so much so that the higher the family income, the higher the average test scores.

In 2002, for example, official figures show college-bound students with a family income of more than $100,000 had an average SAT score of 1,123, compared to a score of 931 for students with a family income of $20,000 to $30,000.

But bias due to family income is only the tip of the iceberg. The SAT also discriminates against blacks and Latinos and against women.

According to FairTest, an advocacy group based in Boston, “Even when students are matched for academic preparation, there are still large gender and racial gaps in their SAT scores.”

Women, for example, scored almost 40 points below men, on average, in the 2002 SAT. Blacks scored more than 200 points lower. And guess what? White males on average do the best.

Yet can anyone in this day and age seriously argue that white guys inherently are smarter and have more ability than anyone else on the planet?

Some might argue that white males do better on the SAT because they have taken harder courses or attended better schools. But the SAT makers say the test is designed to measure “ability” and not the quality of one’s high-school education.

The Educational Testing Service (ETS), which runs the SAT, works hard to eliminate bias in its questions. But following is a question that passed all of ETS’s sensitivity reviews:

“Runner is to marathon as (A) envoy is to embassy (B) martyr is to massacre (C) oarsman is to regatta (D) referee is to tournament (E) mortgages is to homeowner.” (The correct answer is (C) runner is to marathon as oarsman is to regatta.)

Not surprisingly, 53 percent of whites answered this question correctly compared to 22 percent of blacks. Research has shown that test-takers do better on questions set in familiar situations. Perhaps someone at the testing service failed to notice that regattas are a predominantly white sport.

And then there is the question of coaching. Expensive test prep courses – $899 can buy you 24 hours of tutoring for the SAT – can raise test scores by 100 points or more.

The SAT is validated (in other words, its only reliable use) for predicting first-year college grades. But it fails even in that area. High-school grade-point averages or class rank have been shown to be the best predictors of first-year grades, despite differences in high-school courses.

One of the problems with the SAT is that it rewards people who can guess quickly and with a high degree of accuracy. Research has shown that more than 40 percent of reading comprehension items on the SAT can be answered correctly without reading the passage.

Indeed, the SAT works against people who take time to develop thoughtful answers or who look for shades of meaning. And it particularly works against those who do not speak English as their first language.

Dumping the SAT and ACT as a requirement for college admission is not yet widespread but it is growing. In the meantime, take heart, all of you who get mediocre grades on standardized tests; it doesn’t mean you’re a failure.

The Rev. Martin Luther King Jr., arguably the most eloquent and effective civil-rights leader of the 20th century, took the Graduate Record Exam – and scored in the bottom percentile.