The right way to evaluate teacher performance
There is much written and talked about when it comes to evaluating and compensating teachers based on student performance on standardized tests. Those in favor of the approach emphasize the importance of incentives and culling of bad teachers and rewarding of really good ones. Those opposed emphasize natural variation in student background and skills, that standardized tests can be a poor reflection of actual learning, that teachers will teach to the test instead of developing a broader, deeper curriculum, and possible cheating by teachers.
There are simple solutions to these problems:
1) Evaluate students based on performance on standardized tests in the subsequent grade or class.
2) Randomly assign students to teachers.
3) Use student performance in the previous class as a baseline, so that only "value added" measures of student performance are attributed to any given teacher.
These aren't my ideas, but I rarely see them talked or written about, especially by widely recognized leaders of school reform, like Michelle Rhee and Bill Gates.
(1) Makes it nearly impossible for teachers to cheat and removes nearly all incentive to "teach to the test," unless there are true long-term benefits to such a strategy. A broader curriculum may facilitate the kind of deep learning that shows up on standardized tests in subsequent years, even if the test is imperfect.
(2) Helps for evaluation purposes, since all baselines will typically be the same up to a random error that can be quantified. Perhaps more importantly, it instills a level of fairness, so that kids of the most over-bearing parents don't always get the best teachers. This could present a challenge if there are benefits from tracking students by level or ability. I'm a little dubious about the benefits of tracking. (I'd be curious to see good evidence on this--does it exist?) My anecdotal experience is that tracking is done for the benefit of a few parents who want special treatment for their kids, and by veteran teachers who want their pick of students. But if there are true benefits to tracking, random assignment could still work so long as the tracks weren't so small that a single teacher was responsible for an entire track.
(3) While random assignment of students would facilitate unbiased teacher assessment, using past student performance as a baseline, the so-called "value added method," could make assessments more accurate. Or, if random assignment in (2) weren't possible, the value added approach may serve as a substitute. The hyperlink goes to a story where value-added measures for individual teachers were published in the LA Times, which was understandably controversial. Such measures need not be made public for individual teachers, but can and should be made public for informative aggregates, like school and grade-level. And individual measures can and should be used for incentive pay, and for publicly acknowledging the very best teachers.
So, why aren't we doing these things already?
There are simple solutions to these problems:
1) Evaluate students based on performance on standardized tests in the subsequent grade or class.
2) Randomly assign students to teachers.
3) Use student performance in the previous class as a baseline, so that only "value added" measures of student performance are attributed to any given teacher.
These aren't my ideas, but I rarely see them talked or written about, especially by widely recognized leaders of school reform, like Michelle Rhee and Bill Gates.
(1) Makes it nearly impossible for teachers to cheat and removes nearly all incentive to "teach to the test," unless there are true long-term benefits to such a strategy. A broader curriculum may facilitate the kind of deep learning that shows up on standardized tests in subsequent years, even if the test is imperfect.
(2) Helps for evaluation purposes, since all baselines will typically be the same up to a random error that can be quantified. Perhaps more importantly, it instills a level of fairness, so that kids of the most over-bearing parents don't always get the best teachers. This could present a challenge if there are benefits from tracking students by level or ability. I'm a little dubious about the benefits of tracking. (I'd be curious to see good evidence on this--does it exist?) My anecdotal experience is that tracking is done for the benefit of a few parents who want special treatment for their kids, and by veteran teachers who want their pick of students. But if there are true benefits to tracking, random assignment could still work so long as the tracks weren't so small that a single teacher was responsible for an entire track.
(3) While random assignment of students would facilitate unbiased teacher assessment, using past student performance as a baseline, the so-called "value added method," could make assessments more accurate. Or, if random assignment in (2) weren't possible, the value added approach may serve as a substitute. The hyperlink goes to a story where value-added measures for individual teachers were published in the LA Times, which was understandably controversial. Such measures need not be made public for individual teachers, but can and should be made public for informative aggregates, like school and grade-level. And individual measures can and should be used for incentive pay, and for publicly acknowledging the very best teachers.
So, why aren't we doing these things already?
Comments
Post a Comment