The New York City schools have released value-added estimates for individual teaches over five years.
This should make for some lively discussion over the next few days.
For the technically inclined, the NYT reporters' mangling of basic statistics about 2/3 of the way through the article will provide some entertainment. In particular, the NYT reports seem quite innocent regarding the types of variation that will be picked up, or not, in the estimated standard errors.
The Brits report school-level, rather than teacher-level, performance measures (also based on "value-added" models), which reduces the micro-numerosity (i.e. small n) problem a bit. My sense is that principals in the UK have a lot more control over their schools than do principals in NYC, which also argues for school-level rather than individual reporting.
Related paper by yours truly here.
Hat tip: Sara Goldrick-Rab
Overall, I am glad to see the data published. Re: micro-numerousity, this line scares me: "over the five years, 521 teachers were rated in the bottom 5 percent for two or more years, and 696 were repeatedly in the top 5 percent." Really? That means that nearly 4,000 teachers (over 20% of all teachers!) were in the top 5% at least one year in 5 and almost the same at the other end.
ReplyDeleteSecond thought: To the extent that these measures come to matter, one may prefer teaching a class of students who scored poorly the previous year -- regression to the mean will make such a teacher look good.