Reformers want to change teaching quality with carrots (merit pay) and sticks (fire teachers with lousy evaluations). Teacher unions emphasize training and teacher support, and want all teachers to be treated identically. Okay…I oversimplify…but not by a whole lot.
So how about a compromise: We evaluate teachers, but the evaluations focus on providing teachers specifics about what they’re doing that works and about what they’re doing that doesn’t work. Positive or negative consequences may follow from the evaluations, but the consequences are mostly small. The reformers get teachers to do a better job. Unions get support for teachers to do a better job.
For this to work, evaluations have to in fact lead to teachers doing better. Eric Taylor and John Tyler have a piece coming out in the American Economic Review, “The Effect of Evaluation on Teacher Performance,” that shows this is exactly what happens in the Cincinnati teacher evaluation system. The authors have written a long, quite readable, description of their work in the Fall 2012 Education Next. Here’s a quick summary of results.
Cincinnati evaluates mid-career teachers using a very structured and very detailed rubric. Multiple class visits are made by both trained peer evaluators and administrators. Teachers receive detailed feedback on how they’ve done. While both positive and negative consequences are possible, they’re relatively rare and mostly small. Evaluations take place about every five years. So what happens?
- Students of teachers who are evaluated improve their math performance by 4.5 percentile points compared to what would otherwise be expected. That’s a lot.
- Future classes of evaluated teachers do even better, suggesting teachers use what they’ve learned to make long-lasting changes.
- Since teachers know they won’t be evaluated in the immediate future, we can infer these longer run changes are not responses to external incentives. (After the evaluation year there aren’t any further incentives.)
- The evaluations make a difference for math scores, but not for reading.
- The effect may be larger for teachers who were initially lower performers.
- This suggests, but doesn’t prove, that teachers are really learning something from the evaluations. Possibly the better performing teachers already knew what the others learned, so don’t gain aa much from the feedback. Possibly teachers already know more about teaching reading than they do about teaching math.
- The program is not cheap, but neither is it outrageously expensive. My back of the envelope calculation is that it costs roughly 3 percent of the teacher salary pool.