Thompson: Ratings Underplay School, Student, Assignment

 

Let's not get carried away by the headline of the recent NYT story, Teacher Quality Widely Diffused, Experts Say, by Fernanda Santos and Robert Gebeloff.  The biggest issue with using value-added for teacher evaluations is whether good teachers in less effective (or even dysfunctional) schools can be expected to increase test scores as much as equally good teachers in effective schools. If not, test-driven evaluations amount to collective punishment for educators who commit to battling against our toughest challenges. Even the Times article concluded that, overall, "the teacher data reports tended to be highly correlated to the schools’ grades," and that "79 percent of high-performing math teachers worked in “A” or “B” schools."  A large database published by the Tulsa Public Schools, using three years of data for all schools found the same thing.  More than 70 percent of high schools that were above average in achievement had higher than average value-added.  Almost 70 percent of lower performing schools had lower value-added.  Are we supposed to conclude that the big majority of U.S. History teachers in the top schools are great while a similar percentage of their colleagues in lower performing schools are not measuring up?-JT (@drjohnthompson)Image via. 

 

Comments

[Ari]

John, I know you don't like to let people comment on your posts, but maybe you should look at the math. Of course schools that raise test scores are going to contain more teachers that raise test scores. That's how it works. If it didn't, then the calculations would be useless.

[john thompsonj]

The point of the graph is that the difference in teacher quality can't possibly explain such a stark pattern. In theory, Tulsa might have just hired slugs to teach history in the low performing schools and great teachers in the high performing schools, but the basic pattern holds across schools and subjects.

Also, I love getting comments, and I agree that the chart gives more evidence that those test score calculations are useless.