CVA is an attempt to isolate the progress from their prior level of attainment made by pupils in each school. In fact, around 80 per cent of the variation in school outcomes is explained by the prior attainment of their pupils. This leaves only around 20 per cent of the variation that could be attributable to schools in a value-added measure.

Since value-added scores are based on the difference between prior and subsequent attainment, at least half of that 20 per cent is still explained by prior attainment, leaving only 10 per cent for a possible school effect. Using the characteristics of pupils, such as poverty, first language, sex or ethnicity improves the predicted score for each pupil and school, and so reduces the 10 per cent variation attributable to schools even further.

Perhaps only 4 per cent of the variation in school outcomes could be due to the actions of the schools. But this 4 per cent also covers supra-school influences such as parnerships, local authorities, and regional finance. More importantly, this 4 per cent also includes the error component stemming from missing data, measurement and assessment errors, and mistakes in entering, copying and recording data.

There is simply not enough “loose change” left to explain systematic differences within schools caused by their curriculum, pedadogy or leadership. To judge, punish or to close schools on the basis of CVA is little better than superstition.

In general, most people who understand CVA do not use it, and many people who use it just believe it without comprehension.

Professor Stephen Gorard, School of Education, University of Birmingham.