Duh or not Duh? That is the question.

See all posts

Duh. That was our first reaction when we read the broad coverage of the much talked about Chetty, Friedman and Rockoff paper concerning the long-term impact of great teachers. It's certainly not news that students with relatively high test scores face a brighter future (see here, here, and here) compared to their lower performing peers. But it seemed to be BIG news to a whole range of folks--so we found ourselves wondering what was all the fuss about?

Turns out that the headlines failed to grab the significance of this study, which in fact does break new ground, made all the stronger by its magnitude, looking at the progress made by some 2.5 million children. Ever heard the expression "RTFM"? (For decency's sake, we won't write it out.) A slight alteration of that term applies here: RTFP(aper).

For starters, this study provides more evidence that, yes, the impact of great teachers does appear to fade over time when looking at how well students score on later tests. But it also puts forward new evidence that economists' gloomy diagnosis of a "fade out effect" may be all wrong. Instead, great teachers clearly have a lasting impact, contributing to very real and very valuable life outcomes such as higher retirement savings, higher salaries and lower teenage pregnancy rates. The fact that a teacher's contribution may not show up in a test administered a few years after she's taught is more likely a problem with the tests themselves and their own innate though not fully explained insensitivity.

Just as importantly, the study also takes on the issue of causality by testing whether value added measures (VAM) are inherently biased. There's always been a lot of handwringing by researchers over what causes high test scores, theorizing that there are characteristics about students who earn high test scores that lead schools to assign certain teachers to them, termed "nonrandom" assignment.

In other words, is it Ms. Superstar who can get the credit for little Johnnie's 18-month gain--or was it the fact that little Johnnie is a very hard worker and that the school tends to assign Ms. Superstar all the really hard workers?

Chetty et al. tackle that chronically sticky problem in a particularly smart and novel way, one that didn't require them to persuade a school district to let them randomly assign teachers to students--which is almost impossible anyway. Instead, they exploit the fact that teachers transfer in and out of grades, schools and school districts pretty frequently.

Using teacher transfers to set up a quasi-experimental scenario, they were effectively able to trace the learning loss across a whole grade of students when a highly effective teacher left that grade. In fact, the loss in student learning that took place after her departure was a mirror image of the gains that a great teacher might have been expected to produce (based on her average value-added score over six years' time.) By looking at what happened both before and after a particularly great (or particularly bad) teacher entered and exited a grade, and by determining that students tended to perform pretty much as expected absent that teacher, they demonstrate rather definitively that teachers, not students, should get most of the credit or blame for student learning.

In sum, the paper is tremendously significant--not because of the old news it contains about higher test scores leading to higher life outcomes--but because 1) it's now clear that kids could have been well served by that great 4th grade teacher regardless of what any later test scores appear to indicate; and 2) great teachers are rightly given the credit for big spikes in learning, not some characteristics on the part of the students we can't measure. That is certainly worth the fuss.