Defining Program Effects: A Distribution-Based Perspective

In an age of accountability, it is critical to define and estimate the effects of teacher education and professional development programs on student learning in ways that allow stakeholders to explore potential reasons for what is observed and to enhance program quality and fidelity. Across the suite of statistical models used for program evaluation, researchers consistently measure program effectiveness using the coefficients of fixed program effects. We propose that program effects are best characterized not as a single effect to be estimated, but as a distribution of teacher-specific effects. In this article, we first discuss this approach and then describe one way it could be used to define and estimate program effects within a value-added modeling context. Using an example dataset, we demonstrate how program effect estimates can be obtained using the proposed methodology and explain how distributions of these estimates provide additional information and insights about programs that are not apparent when only looking at average effects. By examining distributions of teacher-specific effects as proposed, researchers have the opportunity to more deeply investigate and understand the effects of programs on student success.