SWIMMING WITH SHARKS, READING WITH WHIZZES

Saturday, June 1, 2002

If you've ever tried to improve at some activity that takes both skill and practice, you know how much your progress is influenced by the performance of those around you. It's not much fun playing tennis or bridge with people who are all a lot better at it than you are, or a lot worse.

Apart from the times when you're actually taking lessons -- in which case you want the best teacher you can get, and whether the teacher is having fun playing with you is beside the point -- you probably learn the most when your partners are roughly as good as you are, maybe a little bit better. Then holding your own is a challenge, but an achievable one.

If that's been your experience, as it has mine, consider the implications for that important activity we call school.

Well, you still want the best teacher you can get for your children. Clearly it's a mathematical impossibility for every child's classmates to be "a little bit better" in school than he or she is. And no matter who their classmates are, some children will do better academically than others, because some children are smarter than others.

I know it's considered in bad taste to mention that, though no one seems to have any problem acknowledging the role of talent in music or athletics. But it matters. Any educational policy that pretends it doesn't, or any educational research that fails to control for individual differences, is guaranteed to go off the rails very quickly.

However, it's still worth asking about the effects of the social trinity: race, class and sex.

It's well known that poor children (as indicated by their participation in the school-lunch program) do less well, on average, than nonpoor children. Average, mind you; poverty is not destiny, though sometimes it looks that way.

But Denver's Piton Foundation recently released a study of performance on the state's reading tests, which found that children who attended elementary schools with a high percentage of poor children also did less well, regardless of whether they themselves were poor.

You can't assume this is simple cause and effect, because the children coming into those schools bring differences with them, individual as well as family and neighborhood. But no one should be surprised to discover that how well children do in school is influenced by how well their classmates do. It's just that the ways in which they're influenced are complicated and unpredictable.

Caroline Hoxby, a professor of economics at Harvard, writes about peer effects in the summer issue of the journal Education Next (available online at educationnext.org). The central problem with estimating peer effects, she says, is that families have many ways of selecting their children's peers -- by where they buy a house, for example, or by enrolling their children in a particular school or classroom.

"If savvy parents believe that a certain third-grade teacher is particularly good," Hoxby says, "they may get their children assigned to her class, thereby creating a classroom of children whose parents care about education to an unusual degree." Outcomes that look like peer effects, in other words, might really be effects of the process by which the children became peers.

Hoxby studied the results on Texas' state tests for grades three through six during the 1990s. Just by chance, a particular grade and school will occasionally have a different mix of students -- an unusually high or low number of girls, for instance.

Girls are typically better readers than boys; on the third-grade reading test they averaged about half a standard deviation higher than boys. Exploiting the natural variation in the classroom, Hoxby found that "both boys and girls tend to perform better in reading when they are in classes with larger shares of girls." The numbers suggest, she says, that being surrounded by peers who score one point higher "on average raises a student's own score by 0.3 to 0.5 points, depending on the grade.

Hoxby also finds that black, Hispanic and white third-graders all do worse in reading and math when there are more black students in their classes. "What's particularly interesting," she says, "is that having more black peers appears to be most damaging to other black students." (She is careful to point out that race and poverty in Texas are strongly correlated so the effect may have as much to do with income as with race.)

Having more Hispanic classmates lowers scores of Hispanic students if the class is less than one-third Hispanic; but raises them if the class is more than two-thirds Hispanic. In the middle third, it hardly matters. But for black students, the negative effect is strongest in the middle third.

None of this is very comforting, to be sure. But it's essential information for education reformers who want every child to perform at the top of his game.