New Study on Teacher Coaching

This is University of Virginia basketball coach Tony Bennett. His team is trounced by Duke fairly regularly. Here's what he tells the guys while they struggle:

You look at them in the locker room and you say: 'You're fighting. You're getting close.' And that can sound old. But I said: 'What's the alternative? Just don't go backwards.' I want them to be the best they can be, and if that's scrapping in there and fighting, I said: 'At some point we're going to push through. Don't know if it's this year. Don't know if it's next year.'

A new study on coaching teachers, from UVa, examines the question of this year versus next year.

An EdWeek story about the study was written by Stephen Sawchuk (thanks Cormac). He makes a number of good observations. Worth reading.

The study itself was done by UVa Ed School team including dean Bob Pianta. (A few years ago, Malcolm Gladwell wrote about Bob in this New Yorker article.)

I met Bob a year ago in talking about our proposed Ed School. He was very kind in talking to Bob Manning and me, and offered lots of food for thought.

The study went like this. They lined up a bunch of high school teachers. Divided them randomly. The "treated" teachers got a mix of coaching -- via a website and some 20 to 30 minute phone 1-on-1 conversations.

1. An Observation The difference was 0.22 standard deviations. That's equivalent to students (those taught by teachers who received coaching) moving from the 50th percentile to the 59th. Sizable.

The intervention failed to improve teacher performance during the actual year of coaching. It was only a year later that the students' scores went up.

That is, let's say a teacher gets coaching in September, in November, in February....and then the kids take a test in April. Let's further suppose the coached teacher hits her peak in February. What will the test capture?

The test will only capture only 2 months worth of "peak" teacher level. Meanwhile, most of the kids' test scores will reflect teacher performance from September to January, or 5 months.

If the coaching "sticks," however, the following year should have all 7 months of peak teacher performance. That appears to be what happened in this study.

1b. Our evaluator for New Orleans, Matt Kraft, leapt at this finding. He believes we should study the NOLA teachers we are coaching right now as to how they do during next school year. If the UVa study is any indication, it may be a full year before the coaching "works."

Matt pointed out that my current thinking -- we need our NOLA coaching to work this year -- would have been a big mistake if applied to the UVa experiment. The conclusion would have been "No impact" because we'd wrap up measurement too quickly.

2. A Question The study itself is published in the journal Science. So you can't see it unless you're a subscriber. (A friend sent me my copy).

That's too bad. American taxpayers paid $2.6 million for this study. I appreciate the need for Science to pay its bills. But should access should be restricted to federally-funded studies?

3. An Opinion Overall, this is a terrific experiment. We need much, much more of this sort of rigorous research on how to improve teachers and teaching. A credit both to the research team and Institute of Education Sciences, which funded it.

If this were cancer research, you'd see a bunch of studies use this one as a jumping off point for more experiments. If we change dosage, can we move the 0.22 SD gain to 0.32? What happens if we combine this intervention with, say, some sort of merit pay incentive? What if teachers get this coaching as part of grade-level teams, not as individuals?