More on Value-Added Evaluation of Teachers

In this Harvard Educational Review article, Kimberlee Callister Everson, Erika Feinauer, and Richard Sudweeks (Brigham Young University) raise serious concerns about value-added evaluation of individual teachers. Such calculations “create estimates of teacher effectiveness that are relative to all other teachers at the school, district, or state levels,” say the authors. “These comparisons are a misguided venture, as they essentially hold the teacher responsible for how they are estimated to teach students they never encounter.” Everson, Feinauer, and Sudweeks also raise concerns about non-random assignment of students to teachers’ classrooms, the fact that teachers may be more effective in one setting than another, the negative impact value-added evaluation might have on teacher collaboration, and other methodological problems. 

So what should be used for teacher evaluation? “A teacher should be held accountable only for the job she was hired to do,” say the authors. They suggest focusing on gains made with the students teachers actually teach – what they call propensity score-based estimates, focusing on this question: “Did a  particular teacher do what he was hired to do?” However, the authors acknowledge several weaknesses in this model, especially if it’s used for high-stakes decisions on teachers. 

“If the right information is available,” they conclude, “linking teachers to test scores may be reasonable. High-quality tests do measure part of the learning process, and good teaching should predict higher scores, all else being equal. Unfortunately, all else is never equal, and tests are not always high quality. In some subjects, reliable and valid tests are not even available. The information that would be required to completely separate a teacher’s impact on test scores, including student and home variables, differences in resources across classrooms, and school-level inputs to student achievement, will never be available.” For high-stakes decisions about teachers, they conclude, test scores should not be used at all.

“Rethinking Teacher Evaluation: A Conversation About Statistical Inferences and Value-Added Models” by Kimberlee Callister Everson, Erika Feinauer, and Richard Sudweeks in Harvard Educational Review, Summer 2013 (Vol. 83, #2, p. 349-370), no e-link available

From the Marshall Memo #496