This paper examines the role of social class and prior achievement composition effects on pupils' progress and the judgements made about school performance. It finds that there are small but significant compositional effects on pupil progress raising questions about how pupils can best be allocated to schools. Comparisons between the official contextual value added model and one that includes composition variables show that school rankings are significantly changed in the latter suggesting that official value added analyses are misleading. This study, therefore, poses a fundamental challenge to policy makers over the determinants of pupil progress and school performance and the way schools are judged, pointing to the need for a re-appraisal of policy relating to these matters. (Contains 4 figures and 8 tables.)