Measuring the Impacts of Teachers I: Evaluating Bias in Teacher Value-Added Estimates

Issue/Topic: Teaching Quality--Evaluation and Effectiveness
Author(s): Chetty, Raj; Friedman, John; Rockoff, Jonah
Organization(s): Harvard University, Department of Economics; Harvard Kennedy School; Columbia University
Publication: National Bureau of Economic Research
Published On: 2013

The use of value added (VA) measures to determine teacher quality is on the rise despite several arguments against their use for this purpose. A core argument about VA measures focuses on whether the measures are biased by student sorting. If VA measures are biased by student sorting, they do not serve as valid measures of teacher quality. On the other hand, unbiased VA measures would likely provide important insights into teachers' quality.

To find out if the VA model developed by the authors is biased by student sorting or captures the causal impact of teachers

  • The authors found that their value-added model, which controlled for student characteristics including prior year test scores, is not biased by student sorting and can provide accurate forecasts of teachers' mean impacts on achievement.

Policy Implications/Recommendations:
  • Value-added measures that control for student characteristics, including past performance, are unbiased with regards to the sorting of students. These measures provide accurate forecasts of teachers' impacts on student achievement despite the prevalence of student sorting.

  • While existing achievement gaps are largely driven by factors other than teacher quality, a sequence of good teachers can significantly raise students' scores.

  • Any school district can easily apply the techniques from this study to evaluate its own value-added models. The authors provide a step-by-step guide to implementing their method in Appendix A.

  • In future work, it would be valuable to develop richer measures of teacher quality that go beyond the mean test score impacts.
Full text: http://www.nber.org/papers/w19423

Research Design:
Secondary data analysis

2.5 million students, plus teachers in grades 3-8

Year data is from:
Data for teachers and students spanned school years 1988-1989 through 2008-2009; for tax data 1996-2011


Data Collection and Analysis:
Data spanned school years 1988-1989 through 2008-2009 and covered roughly 2.5 million children in grades 3-8. Data included test scores and teacher assignments from a large urban district and parent characteristics based on tax data spanning 1996-2011. Identifiers such as names were stripped. Analysis included calculating VA measures for teachers using authors' model, predicting teachers' VA using data from other years and comparing them.


Reference in this Web site to any specific commercial products, processes or services, or the use of any trade, firm or corporation name is for the information and convenience of the public, and does not constitute endorsement or recommendation by the Education Commission of the States. Please contact Kathy Christie at 303-299-3613 or kchristie@ecs.org for further information regarding the information posting standards and user policies of the Education Commission of the States.