Monday, 17 June 2013

Ofsted needs to get its facts straight

Recent pronouncements by Sir Michael Wilshaw, Chief Inspector of Schools in England (above) - to the effect that large, mixed-ability comprehensives are 'failing' the brighest pupils - are as worryingly misleading as they are inaccurate.

No-one would dispute that big schools taking in everyone from the locality often experience problems dealing with the brightest and the best - though it's arguable that they 'fail' the less academic even more, a blind spot that England and Wales in particular have never really addressed. In his unending quest for 'standards', Sir Michael wants to focus on the most scholastically able.

But what he said won't wash. Here's the deal: Wilshaw said that forty per cent of pupils getting the best marks when they were tested at the age of eleven didn't emerge from school at sixteen with top GCSEs in core subjects. So the system must be going wrong somewhere.

Hmm. Well, that's one crude way of looking at the picture. Here's two counter-assertions, better supported by the data. One: tests at eleven aren't very good indicators of 'ability', or 'achievement', and in fact much lighter-touch tests - which don't claim to do anything but look at where pupils have got to by that time and at that moment - would be much more useful. The pitiful predictive inaccuracy of testing before children's academic identities have really formed is revealed by this data more than anyone's 'failings' in particular. Two: all sorts of schools have these problems, because tests at the end of primary school and GCSE test completely different things. Llots of grammar or other selective schools don't do all that well by this measure either. Forty one per cent of grammar school pupils 'fail' to get the best GCSEs when they did well at eleven. Is that a 'failure' too, then?

For some reason Sir Michael didn't refer to non-comprehensives' 'success' or 'failure' rates. That seems like a highly pregnant silence when you think about just how badly grammar schools did at integrating working class children in the 1950s and 1960s - a time when they were supposed to be boosting social mobility, but during which they actually had a pretty good chance of leaving withou taking any formal public exams at all. Early Leaving, one 1954 report on the matter, estimated that only a third of working class children attending grammar schools attained a pass in three or more O-Levels.

This column has argued again and again that you have to understand how numbers are built - the raw material that you're being fed - before you can use any numbers with confidence. This rather embarrassing episode fails that test.

At a time when a lot of experts are very worried about the public's poor understanding of statistical data, it's critical that we speak and write in a manner that is both accessible and accurate. This kind of pronouncement looks accessible, but isn't, because it lacks both historical and statistical context. If you hadn't studied public statistics, and you didn't know about grammar schools' relative failure in the 1950s and 1960s, you'd just read this story and say 'tut, terrible old comprehensives, eh?' And that's a very muddled, and not I hope a deliberately deceptive, picture.

In the dawning age of big data, that just isn't good enough.