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

TLDR

Value‑added measures of teacher impact on student test scores are debated for their potential bias in estimating causal effects. The study tests for bias in value‑added estimates using unobserved parent characteristics and a quasi‑experimental design based on staff changes. The authors employ a quasi‑experimental design that leverages changes in teaching staff and incorporates previously unobserved parent characteristics. Value‑added models that control for prior test scores produce unbiased forecasts of teacher impacts on student achievement, as demonstrated with data on more than one million children. JEL codes: H75, I21, J24, J45.

Abstract

Are teachers' impacts on students' test scores (value-added) a good measure of their quality? One reason this question has sparked debate is disagreement about whether value-added (VA) measures provide unbiased estimates of teachers' causal impacts on student achievement. We test for bias in VA using previously unobserved parent characteristics and a quasi-experimental design based on changes in teaching staff. Using school district and tax records for more than one million children, we find that VA models which control for a student's prior test scores provide unbiased forecasts of teachers' impacts on student achievement. (JEL H75, I21, J24, J45)