Why don't schools and teachers seem to matter? Assessing the impact of unobservables on educational productivity

Using data drawn from the National Educational Longitudinal Study of 1988, which allows students to be linked to particular teachers and classes, we estimate the impact of observable and unobservable schooling characteristics on student outcomes. A variety of models show some schooling resources (in particular, teacher qualifications) to be significant in influencing tenth-grade mathematics test scores. Unobservable school, teacher, and class characteristics are important in explaining student achievement but do not appear to be correlated with observable variables in our sample. Thus, our results suggest that the omission of unobservables does not cause biased estimates in standard educational production functions.
I. Introduction Dating back to the 1966 Coleman report (Coleman et al. 1966), hundreds of studies by economists and others have investigated the relationship between educational productivity and schooling expenditures. [...]