LEAGUE tables that rank schools based on just one year鈥檚 performance are deeply misleading, according to a model that compares the quality of a school鈥檚 teaching with its standing in the tables.
Academic leagues based on one year鈥檚 results are 鈥渆ssentially meaningless for the vast majority of schools鈥�, says mathematician Rebecca Hoyle of the University of Surrey. Yet tables of this type are published in England, the US and Australia, while other countries, including Italy, are considering doing the same.
Hoyle and James Robinson from the University of Warwick used a simple mathematical model to simulate the results of pupils from 10 schools over 15 years. In the model, the quality of the school determined 10 per cent of each child鈥檚 exam results, with the remainder divided between prior achievement (60 per cent) and family background (30 per cent). These weightings were taken from previous studies of real schools.
Advertisement
The researchers will report in Proceedings of the Royal Society B that the position of a school in the league does not necessarily reflect its quality of education. Even when all the virtual schools were identical, some established themselves at the top of the table, while others quickly sank to the bottom.
Parents react to small random differences between the schools鈥� test results, and this creates a feedback effect. Better-off families can afford to move house or pay for transport to send their children to a school they perceive to be better. And because family background affects exam results, this can boost the performance of their chosen school. Less popular schools end up with more children from poorer backgrounds, so their results get worse.
鈥淭hey have shown that parental choice leads to unfairness,鈥� says Harvey Goldstein, a statistician at the Institute of Education in London. The schools become socially segregated, the researchers showed, and each school鈥檚 position in the league table ends up reflecting only the demographics of its pupils.
The 鈥渧alue-added tables鈥� being piloted in England attempt to correct for the students鈥� background. But these are unreliable too, says Hoyle, because once background is taken away, the gap in quality between the best and worst schools is only slightly bigger than the random differences from year to year.
Kathryn Doherty, research director for the US news magazine Education Week, admits that league tables are a 鈥渃rude measure鈥�. But she points out that not publishing the tables at all wouldn鈥檛 necessarily be helpful. 鈥淚t鈥檚 important to hold schools accountable,鈥� she says.
One alternative is to base tables on several years鈥� data. Hoyle and Robinson found that averaging the schools performances over the previous four years smoothed out the random variations and represented the schools鈥� quality hierarchy more accurately.