In an NBER Working Paper released in November, Marianne Bitler, Thad Domina, Emily Penner, and I use statistical models commonly used in research, policy, and practice to estimate teacher “effects” on student height. We find effects about as large as those found on math and reading achievement. Do we really believe teachers affect their students’ height? No. Rather, this paper provides a cautionary tale for the use of “value-added” models in settings less obvious than ours. As others have pointed out, estimated teacher effects contain considerable sampling error, and our ability to separate the “signal” (true teaching effectiveness) from “noise” (sampling and other year-to-year variability) in real-world applications is limited.
For a less technical discussion see our related blog post at the Brookings Institution Brown Center Chalkboard.