Britain has long favoured an odd school system whereby well-to-do parents buy an education at the better state schools by giving money to homeowners who live near those schools, rather than by giving the money to the schools themselves. This is not very satisfactory, and there are two logical responses. One is to let the parents give the money to the schools. The other is to prevent people from buying a place at a good school through the housing market, and instead assign places from a much wider area using a lottery. This bold new experiment is about to be tried in Brighton and Hove.
Some parents are understandably livid: they paid for a service (albeit indirectly) and suddenly discover it’s being handed out like a raffle prize. Their houses will probably lose value. Little Jeremy may not even go to that wonderful school at all. But Brighton’s dispossessed parents are also worried by the same thing that worries parents all over the country: that if their school allows too many of the ”wrong” type of children in the door, Little Jeremy’s performance will suffer.
What these parents are worrying about is what an economist would call a ”peer effect”. Peer effects are what happen when you hang around in the wrong company. Yet the evidence for their existence is slimmer than the nation’s parents assume.
The difficulty is this. If Jeremy hangs around with the ”right” kids and does well, why? The obvious explanation is that he did well because his peers were a good influence on him, but it is just as plausible to suggest that he chose those peers, or had those peers chosen, because he was one of the ”right” kids, too. Does John Terry play great football because he is surrounded by great footballers, or is he surrounded by great footballers because he plays great football?
Clever researchers can disentangle some of these effects. The economist Bruce Sacerdote used the same technique that medical researchers would use to test a new headache remedy: a randomised trial. He realised that students at Dartmouth College had roommates assigned largely at random. There was some selection at work based on sex, smoking and preferences for hours of work - but mostly, the assignments were the result of a lottery.
Sacerdote found a modest - and statistically robust - peer effect. Being assigned a roommate with a higher grade-point average improves your own. If your roommate is at the top of the grade-point distribution you’ll tend to be about 5 per cent better than average. If they are 20 per cent below average you’ll tend to be 1 per cent below average. Sacerdote doesn’t know what the cause is, but since students did not choose their peers, it must be a genuine peer effect.
Most studies of peer effects are not so careful, however. In a clever paper based on studies of North Carolina’s public schools, Thomas Nechyba and Jake Vigdor highlight the pitfalls. They provide what appears to be strong evidence of peer effects - but then demonstrate that these apparent effects are at work before the peers ever appear. That is, by looking at Jeremy’s fifth-grade classmates you can work out how Jeremy performed, with different classmates, in the fourth-grade. Similarly, the quality of John Terry’s team-mates is a sign that Terry was a good footballer before he joined Chelsea.
Nechyba and Vigdor also show that the peer effects evaporate once they consider the quality of teaching. Students with smart peers are also students with better teachers. Perhaps the good folk of Brighton and Hove should worry less about Jeremy falling in with the wrong sort of classmate, and more about him falling in with the wrong sort of teacher.
