Undercover Economist

What lessons can schools learn from streaming by ability?

Monday is a big day in the Harford household: my oldest daughter will start school. That is a cue for the full spectrum of middle-class parental emotions: nostalgia for the toddler she once was; pride at seeing her reach a new stage of independence; and, of course, anxiety that the school will not be good enough for our little darling.

We have been given few reasons to fret about the quality of the teaching, but like many parents we’re nervous about the impact our daughter’s peers may have on her, many of whom are from deprived backgrounds or homes where nobody speaks English. Will the teacher be distracted by the need to teach the class skills she already has?

I have written before about “peer effects” in education, which are the influences, positive and negative, that classmates and school friends have on each other. They are hard to identify with much certainty. Bright children might make friends with each other without actually improving each other’s test scores. Or pushy middle-class parents might all flock to the same popular school. Or a class of smart kids might attract a good teacher. All these situations would produce clusters of high and low achievement, yet no true peer effects need be at play.

Still, there are occasions on which classmates are assigned absolutely at random. For example, in North Carolina, 120,000 children were randomly assigned classmates over the period of a decade. Using such situations, economists think they are identifying peer effects.

Caroline Hoxby, a Stanford professor of economics and a leading figure in the field, explained the emerging consensus to me in an interview last year. First, peer effects exist. Second, they are not nearly as important as good teachers: given the choice between the best class in town and the best teacher in town, parents should choose the best teacher any day.

Third, peer effects take the form of what Hoxby describes as “the sports model”. If you were looking to improve at a sport, you would typically seek to play against people who were a little better than you, because they would drag you up to their level. The same appears to be true in a classroom: children benefit from having classmates who are just a little ahead of them.

This is useful to know, since there are many other plausible models of peer effects, including the “rainbow” (children benefit from having a wide range of abilities around them), the “bad apple” (if the troublemakers can be deterred, cured or excluded, their classmates will be fine), the “shining star” (one classroom genius inspires everyone) and the “boutique” (what matters is that the whole class is at much the same level, so the teacher’s lessons can satisfy everyone). Since it is mathematically impossible for every child to have slightly superior classmates, the “boutique” model seems to be the next best thing, and that suggests some form of streaming by ability is wise.

There is new evidence of that from Kenya, whose education system is at the cutting edge of a newly popular type of economic analysis, the randomised controlled trial. In the latest example – studied by the economists Esther Duflo, Pascaline Dupas and Michael Kremer – 121 Kenyan schools were given a grant to hire an extra teacher and so split one large reception class into two smaller classes. In 61 randomly chosen schools the students were streamed by ability; in the other 60, they were randomly assigned to their classes. The result: better grades for everybody in the streamed classes, whether they were originally judged to be high, medium or low ability. It is not a shocking conclusion but it is good to have the gold-standard of the randomised trial to support it. If only there were more such trials outside east Africa.

Also published at ft.com.