As promised Monday, here are some facts on the source of variation in value-added estimates of teacher ability. I’ve condensed these down from a newly published paper by Dan Goldhaber and Michael Hansen. Here’s a picture.
What this means is that value-added measures (so long as they measure the educational accomplishments we want to measure) are plenty good enough for research purposes and are also useful for individual personnel actions. With regard to the latter, we’d want to adjust for predictable changes in teacher effectiveness and might want to use more than one year’s VAM measure in order to reduce the role played by measurement error.