SAT'S PROBLEM IS THE STUDENTS, NOT THE TEST Jul. 11, 1993 Los Angeles Daily News Boys do better than girls on standardized tests such as the Scholastic Assessment Test, and critics say that results from "gender bias." I'm skeptical. Mostly I'm skeptical because a lot of the people advancing this claim have a proprietary interest to defend. A whole industry has sprung up to offer courses in how to do better on the SAT test. The College Board, which sponsors the SAT and changed its name from the Scholastic Aptitude Test in March, gives excellent advice on that subject, and it's free: Take a demanding program of academic courses in high school. But nobody will profit (except the students) if students take that advice. The bias hunters say they can sniff out questions that tend to favor boys or girls. But their notions - that questions that refer to sports favor boys, for instance - are simplistic. I know less about sports than almost anybody you could meet, but who needs to be a sports fan to figure out that basketball is to basket as soccer is to goal? Cooking questions don't necessarily favor girls, either. Men did better than women on this one. "A recipe calls for 1 cup of nuts, 5 cups of chocolate chips, and -1/3 cup of raisins. What is the ratio of nuts to chips to raisins?" (It's 3:15:1.) Because stereotypical ideas about who will find questions hard can so easily be wrong, the Educational Testing Service, which prepares the test, does a statistical analysis to identify questions that are answered differently by different groups of people with the same ability. Such questions are either eliminated or balanced with other questions. So if the bias doesn't lie in the individual questions, why do scores turn out differently? Most probably, because they are measuring a real difference, especially in mathematics. Women are less likely to take college prep courses, and they are more likely to be the first in their family to attend college, and both of these factors have a large effect on scores. On the verbal part of the exam, the 1992 average score for women was 419 (on a scale from 200 to 800) and for men, 428. That difference is quite small, compared with the gap that results from different high school preparation; from 348 (weak) to 464 (strong). Parents' education has an even stronger effect. In the mid-'70s, when more men than women took the SAT, there was almost no gap in verbal scores. The math gap is wider, from 456 to 499. But again, it's smaller than the range just for women alone, which runs from 388 (parents with no high school diploma) to 510 (parent with a graduate degree). Besides the gender differences in scores, there are sizable ethnic differences. There are also significant effects from household income, and from location, whether by state or urban vs. rural. And all these factors interact, in ways that are not all all obvious or intuitive. The "gender-bias" theorists like to make much of the fact that women generally earn higher grades than men with the same SAT scores, a phenomenon they describe as "underpredicting" the performance of women. One could just as logically say that the tests overpredict the performance of men, but that wouldn't be nearly as gratifying to people who want to feel aggrieved. "Promoting a test which underpredicts the performance of the majority of its consumers," writes chief theorist Phyllis Rosser, "is more than consumer fraud, it is irresponsible and damaging. After nearly a quarter of a century of this inequity, women cannot wait any longer to be equally included in the talent pool." Rosser is the director of the Equality in Testing Project in Holmdel, N.J., and author of "The SAT Gender Gap: Identifying the Causes," published in 1989 by the Center for Women Policy Studies. Among the ills Rosser attributes to the gender gap are a loss in women's self-esteem when they learn their SAT scores, lowered aspirations for college and, for all I know, the hole in the ozone layer and radioactive spinach. Changing the SAT to lower men's scores would eliminate the gender gap just as effectively as raising women's scores, but one would inflate women's self-esteem and the other wouldn't. Neither would improve the effectiveness of the test in predicting college performance. If the scores changed, the formula used to calculate performance would change too, but the predictions wouldn't. If you want to measure something three inches long, it doesn't matter where on the ruler you start. There's nothing wrong with courses that coach students in how to do better on the SAT, as long as they work. I taught a course like that five years ago, in Shanghai. It covered the Graduate Record Examination, which is very similar to the SAT, just harder. There wasn't any great secret to what we did in the class, which was to practice taking sample exams with the same time limits as real exams, and then to talk about why the answers were what the answer book said they were. And it did work. Almost everybody improved somewhat, and a few people made spectacular gains. Most of my students had never met a native speaker of English before. For that reason they needed a teacher, although otherwise they could probably have taught themselves. There were a lot of cultural gaps that someone had to fill in. Such gaps don't mean the test is biased, though, just that it is measuring something that's important for success in graduate work - namely, fluency in English. If I had to take a college admission test in Chinese, I'd register as brain-dead, because I can't read Chinese. But that's a fair estimate of my likelihood of succeeding as a student in a Chinese-speaking university (nil). The Chinese lesson I would like to leave you with is that there was no gender gap in mathematics in my Shanghai classes. Both the women and the men were so far above the U.S. average on the quantitative section of the GRE that we decided it wasn't important to practice that part of the test more than once. They weren't math and science students, either; most of them were in the humanities. The gap that worries me most, in other words, is not the gender gap - it's the nationality gap. Instead of fretting about whether standardized tests are equally fair to men and women, I think it's far more important to find out why both sexes do so badly.