Comparison of value‐added models for secondary‐school effectiveness

Abstract

The issue of value‐added measures of school effectiveness is reviewed in relation to the Government's requirement for schools to publish raw examination results. The study reports a series of multilevel analyses of the 1993 GCSE examination results in Lancashire employing a wide variety of pupil intake and school context variables as well as a range of 1991 Census measures attached to each school's catchment area and each pupil's home area. The methodology involves evaluating and comparing the results of several different value‐added models controlling for different pupil and school background factors. The findings indicate that a substantial percentage of school level variation in pupil outcomes can be explained by pupil intake factors. In addition, comparisons between the different value‐added models indicate that when rich and wide‐ranging pupil level data (for example, prior attainment measures in different areas) are available and taken into account in the analysis, school‐context factors (such as percentage of pupils entitled to free school meals) are not significant in predicting pupil outcomes. However, when pupil prior attainment data are lacking, contextual factors are more useful and may be viewed as adequate but approximate measures of pupil intake. Within individual schools there is also evidence of inconsistent departmental effectiveness and differential effectiveness for pupils of different levels of prior attainment. The implications and limitations of the findings are discussed, as is the question of how the methodology may be employed to stimulate approaches to school improvement and to widen the scope of value‐added measures.