Previous research on educational productivity has decomposed the variance in student test scores into school and class effects.In this paper, we extend this work to include differences attributable to teachers as well as to schools and classes. Using data drawn from the National Educational Longitudinal Study of 1988, we find that unobservable School, teacher, classroom characteristics are important factors in explaining 10th-grade mathematics achievement, and account for the majority of the variation that is explained by educational variables.