Multilevel modelling in school effectiveness research

Fifteen years ago, Burstein (1980) argued that the key to methodological progress in studies of classroom and school effects depended on the development of appropriate models and methods for the analysis of multilevel data. Considerable progress has been made in the intervening years such that anyone familiar with the growing school effectiveness research literature will have encountered the methodological imperative: 'Pay attention to the multilevel organisational structure in which schooling occurs' (i.e., students within classes within schools). Results are now available from a number of studies that have employed multilevel modelling to investigate school and teacher effectiveness. In the main, these results suggest that variation between classes is far more significant than variation between schools, although in detail the evidence often appears to be contradictory and open to a variety of interpretations. This article considers why different studies generate different findings, identifies some key issues in the design of studies for investigating the relative importance of class and school effects and relates this to the presentation and discussion of findings from an Australian study of school and teacher effectiveness.