Value-added Modeling and the Power of Magical Thinking

This article explains the impetus in the United States behind the drive to extend test-based accountability to teachers and the growing interest in employing valueadded models to generate the indicators to be used in teacher evaluation. The empirical literature finds that teaching quality is the most important school-related determinant of student. Yet, in the main, teacher evaluations are done poorly – if at all – and compensation has been largely determined by seniority and credentials. Policy makers see strengthening teacher accountability as a priority. In particular, they are looking to increase the role of outputs in comparison to inputs. However, given the technical problems associated with status-based indicators of teacher and school effectiveness, the focus has turned to indicators based on some measure of the progress students have made during the academic year. Value-added analysis relies on sophisticated statistical models to generate estimates of the relative effectiveness of teachers, based on a measure related to student progress. This article provides a brief introduction to value-added models and summarizes key research findings. Although value-added estimates have some desirable properties, they do not represent a simple, neat solution to a complex evaluation problem. In this spirit, the article concludes by describing some of the many concerns regarding the use of value-added scores for high-stakes decisions and suggests some ways to enhance the likelihood that teacher accountability will contribute constructively to the improvement of teaching.