Statistical Modelling Issues in School Effectiveness Studies

The assessment of school effectiveness in educational research studies is considered from the viewpoint of statistical modelling. A variety of models are applied to a set of data on 907 pupils in 18 schools from one Local Education Authority. We argue for the general use of variance component or "random parameter" models for the analysis of such studies involving clustered observations. For the data examined, the model which regresses school mean outcome on school mean intake (i.e the "means on means" model) is shown to give estimated school effects considerably different from those produced by other models. In the light of our results, we comment on several recent large-scale British studies of school effectiveness.