Disaggregating between- and within-classroom variation in student behavior: A multilevel factor analysis of teacher ratings of student prosociality and aggression

Teacher ratings of student behaviors vary systematically both at the student and teacher/classroom level. Multilevel confirmatory factor analysis (ML-CFA) can disaggregate between- and within-teacher/classroom variance, identify an optimal psychometric model at each level, and test correlates of the resulting dimensions. In this study, 250 teachers (37% males) rated an average of 4.02 students (51% males; aged 10 years at Time 1 and 11 years at Time 2) from a normative sample of Swiss youth. Substantial and unidimensional between-teacher variation in ratings of both prosociality and aggression were identified, and this was stable across time. These dimensions were not associated at the between-teacher/classroom level with teacher gender or teacher–student relationships, although they were associated with teacher-student relationships at the within-teacher/classroom level. There was little between-teacher/classroom variation observed in student self-reports of prosocial and aggressive behavior, and multilevel CFA was not possible for these ratings. Future research should aim to identify sources of between-teacher/classroom variation. This should include factors that influence negative and positive teacher perceptions of and response biases related to student behavior as well as those that influence student behavior itself.