Let me quote Kate Walsh:
One of the obstacles to states putting in new evaluation systems which use any test score data is the pushback from teachers. Teachers worry a lot that their rating will depend more on who they teach, than how well they teach.
Walsh went on to share some very reassuring news from Florida
Florida has found near-zero correlation between teachers’ evaluation scores and the percentages of their students who are poor, nonwhite, gifted, disabled or English language learners. None. The fact that a teacher only is responsible for growing student learning from the baseline scores she inherits from the previous year mitigates any need to make accommodations for students’ background.
Teachers might dislike being evaluated based on teacher test scores for two reasons.
1. What students bring to class has a bigger effect on test scores than what the teacher does in the class.
→Go read Walsh’s message above. Using a well-designed value-added system, this doesn’t happen.
2. Test scores don’t measure everything that a teacher should do.
→True, and important. But that test scores do measure some of the things that a teacher should do.
All-in-all, good news from Florida. Next, time a more wonky take on the news.
Florida’s value-added model is a random effects model that controls for student disability, ELL status, and gifted course enrollment. So it isn’t exactly surprising that teachers’ estimated value-added is uncorrelated with these characteristics — it is built into the estimation procedure!
Or:
3. VAM are inaccurate even with respect to test scores.
Walsh’s point about correlations doesn’t do much to alleviate those concerns.