Evaluating Educational Interventions Using Multilevel Growth Curves: The Case of Reading Recovery

Abstract

Educational interventions are notoriously difficult to evaluate. This article describes an application of multilevel modelling to repeated measures as part of an evaluation of Reading Recovery in England. Growth curve and multivariate models are applied to the within and between school aspects of the design, which also included a phonological intervention. Reading Recovery is shown to be effective in the short term but not in the long term. A measure of social disadvantage is a powerful predictor of reading growth in the long term.