QUOTAS FOR SMART KIDS
(FORTUNE Magazine) – One senses uneasily that the headline above will engender a certain amount of dismay in Norma Cantu, who heads the Office for Civil Rights (OCR) in the U.S. Department of Education and will instantly intuit that our lengthy exchange of views did not result in one's buying her version of recent amazing events in the world of test scores. Main event: the agreement, brokered by OCR and gamely/glumly acquiesced in by the College Board and the Educational Testing Service (ETS), to change the Preliminary Scholastic Assessment Test (PSAT). PSAT scores are the sorting mechanism used in selecting the 15,000 semifinalists for National Merit Scholarships--an extremely smart group of kids in the upper half of the 99th percentile, implying that they have minimum IQs around 140. Point of the change: to ensure that more girls make it into this group, which on the latest available data was only 44% female.
Here at the Keeping Up social policy desk, we look at that figure and note that it looks about right. Down there in the District, Norma's office says the figure raises big, big questions about Title IX of the 1972 Civil Rights Act, which bars sex bias in federally aided education.
Ms. Cantu denies that she wants absolute gender parity in the semifinalist totals but admits to being unhappy with a 56 to 44 ratio and expects it to move substantially closer to fifty-fifty after next year's change in the PSAT. This test, taken annually by 1.2 million high school juniors hoping to latch on to the $27 million of Merit Scholarships, has a verbal and math section--just like the Scholastic Assessment Test (SAT) taken by college-bound seniors. The National Merit Scholarship Corp. has long double-weighted the verbal test, which obviously benefits the girls, whose verbal scores are well above their math scores. (The boys, in contrast, do a lot better in math.) What's now being added to the PSAT, with OCR's encouragement, is yet another verbal test--of writing skills, on which girls do far better than boys.
We like that 56 to 44 sex ratio because it is so logical. Just about all psychologists agree that males and females are on average equally smart. But scholars are increasingly aware of another proposition: that there is more variability in the male scores. There are more men at the tails of the distribution curve--which means they are overrepresented among the retarded, and also among gifted groups like the Merit Scholars. Higher variance for males also means that the further out you go at either extreme, the higher the proportion of males you find. The latest (1992) edition of Nathan Brody's Intelligence, a text widely identified as fair minded, summarizes many studies and states flatly, "Males generally have higher variance on ability scores." Brody adds, "If selective institutions use high cutting scores on the SAT as one means of selecting their students, they will find more males meeting this criterion than females." Last year Larry V. Hedges and Amy Nowell of the University of Chicago reported in Science on their analysis of six huge studies with data on the cognitive skills of American teenagers. They reported greater male variability in test scores in all the studies and virtually all tests, and served up many examples of increasing male dominance as the cutoffs rose. In the Project Talent study, the sex ratio in math scores was 1.3 to 1 in the top 10% but 7 to 1 in the top 1%. How come our government doesn't know this? Ditto for the New York Times editorialists weighing in under the weirdish headline, RIGGING TEST SCORES JUSTIFIABLY. Let's hear it for justifiable rigging.
The above-referenced editorialists and numerous others keep telling you that the PSAT (like the SAT) is supposed to be a predictor of college grades, but that it underpredicts women's grades, so it must be biased. The data do indeed show that, on the standard 1 to 4 scale, women get grades 0.10 point higher than the test scores predict. Is that a big deal? Okay, let's assume it is. But much of that gap vanishes when you control for course difficulty--for the fact that men are overrepresented in courses with tough grading standards (physics) and women are overrepresented in other courses (education) with softer standards. The public education director at FairTest, the advocacy group that initially brought the Merit Scholarship complaint, allows that controlling for these differences reduces the gap by about half. Studies at Dartmouth and Stanford show the gap largely disappearing when you control for course difficulty.
In any event, a gap between test scores and college grades does not prove test bias. Some scholars believe it's the grades that are biased. And why shouldn't the profs be biased in favor of female students--a group notorious for getting assigned projects in on time and not shooting off its mouth in class? H.D. Hoover, professor of education at the University of Iowa (and director of the state's Basic Skills Testing Program), believes there is more evidence of grade bias favoring girls than of test bias favoring boys. How about a Title IX investigation of grading bias against boys?
Just a thought.