The aim of the current study was to examine mean-group differences on behavior rating scales and variables that may predict such differences. Sixty-five teachers completed the Clinical Assessment of Behavior-Teacher Form (CAB-T) for a sample of 982 students. Four outcome variables from the CAB-T were assessed. Hierarchical linear modeling was used to analyze variance components across three levels; examine mean-group differences across outcome variables for student gender, race/ethnicity, and free or reduced-price lunch status; and examine whether teacher variables predicted teacher-specific differences in ratings. Results revealed that a significant amount of variance was attributable to teacher- and school-level variables. Several mean-group differences emerged, and some teacher-specific differences in ratings across groups were predicted by teacher self-efficacy for behavior management and teacher age, but not for teacher race/ethnicity, gender, or years of experience.