Estimating Causal Effects Using School-Level Data Sets
Education researchers, practitioners, and policymakers alike are committed to identifying interventions that teach students more effectively. Increased emphasis on evaluation and accountability has increased desire for sound evaluations of these interventions; and at the same time, school-level data have become increasingly available. This article shows researchers how to bridge these two trends through careful use of school-level data to estimate the effectiveness of particular interventions. The author provides an overview of common methods for estimating causal effects with school-level data, including randomized experiments, regression analysis, pre–post studies, and nonexperimental comparison group designs. She stresses the importance of careful design of nonexperimental studies, particularly the need to compare units that were similar before treatment assignment. She gives examples of analyses that use school-level data and concludes with advice for researchers.