ABSTRACT
The majority of studies of educational effects-whether classroom experiments, or evaluations of programs, or surveys-have collected and analysed data in ways that conceal more than they reveal.
(Cronbach et al., 1976, p.l) This conclusion was based on the fact that studies of school effects, for the most part, had not adequately addressed the hierarchical or multilevel nature of schooling data (see also Burstein, 1980a, 1980b). In a large part this has been due to the fact that statistical methods for handling multilevel data typical of school effects studies have only recently become available. According to Raudenbush and Bryk (1986),
Research on school effects has been plagued by both methodological and conceptual problems. In our view the two are closely related…there is a natural hesitancy to form a judgement when it remains unclear how to test the fruits of that conceptualisation.
