Educational accountability is becoming more important and is often an issue for policy and financial debate especially in public schooling systems. In order to justify the return of investment, investment receipts need to show that the investment on them have made substantive impacts. These often lead to ranking of school performance (in whatever metrics system) with a consideration of low achiever of being deprived of new or existing resources. In such a case, measures of school achievement need to be highly reliable so that high-stakes quality assurance and decision accordingly can be made legitimately. This article is significant to point out that in a nested structure of the school systems (school – classroom – student), classroom-level variance is crucial and ignoring it will result in decrease of reliability, or the generalizability, of the school performance measures. Here one of the variables concerned within the school-level variance is the quality of classroom climate, i.e. the teacher-student interaction, and a class random effect model is an appropriate model. However, it should be acknowledged that, besides the classroom-level variables, there may be some other relevant and important variables worth consideration, such as the clusters of teaching teams and special arrangement of methods of pedagogies.