MULTILEVEL MODELING IN SCHOOL EFFECTIVENESS RESEARCH

Fifteen years ago, Burstein (1980) argued that the key to methodologic al progress in studies of classroom and school effects depended on the development of appropriate models and methods for the analysis of mul tilevel data. Considerable progress has been made in the intervening y ears such that anyone familiar with the growing school effectiveness r esearch literature will have encountered the methodological imperative : 'Pay attention to the multilevel organisational structure in which s chooling occurs' (i.e., students within classes within schools). Resul ts are now available from a number of studies that have employed multi level modelling to investigate school and teacher effectiveness. In th e main, these results suggest that variation between classes is far mo re significant than variation between schools, although in detail the evidence often appears to be contradictory and open to a variety of in terpretations. This article considers why different studies generate d ifferent 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.