[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Fw: Willingham: ealuating teachers by student test scores is 'A terrible idea' - The Answer Sheet

In concise and clear fashion, Willingham points out fatal flaws in using "growth" models to evaluate teachers. He does not explicitly point out that testing in fall and spring instantly doubles the amount of standardized testing! In order to get data that measures little and is plagued by error - and in order to more thoroughly control teachers, students, curriculum and instruction by Washington, test companies, and perhaps state ed departments.


Willingham: 'A terrible idea'
My guest today is cognitive scientist Daniel Willingham, a professor at the University of Virginia and author of "Why Don't Students Like School?"

By Daniel Willingham
Randi Weingarten's recent speech at the National Press Club garnered a great deal of press attention, almost all of it on her openness to student achievement data being part of an evaluation scheme for teachers.

This is a terrible idea.

When people talk about using student achievement measures to evaluate teachers, they are usually talking about growth models, sometimes called value-added measures.

Simply measuring what a child knows at the end of the year is obviously a measure not just of that year's teacher, but of all of the teachers the child has had to that point, the parents, the neighborhood, etc. So growth models measure kids in the fall and the spring, and look at the change across the year.

Here is an incomplete list of problems:

1) Teacher effectiveness is influenced by factors outside of teacher's control, e.g., the principal and other administrators of the school, the parents.

2) Testing kids in the fall doesn't account for the fact that some kids are easier to teach than others. Some kids are rowdy and disruptive.

3) Students learn more when their peers add value to the classroom. It's better for a student to be in a class with peers who excel than in a classroom with peers who struggle.

4) Tying salary or promotion decisions to student growth over the course of year encourages teachers not to worry about future years. Why should I lay the groundwork for next year's work?

5) Growth models yield scores that are unstable. Teachers who look pretty good one year might look pretty bad the next. This problem may be inherent in growth scores because fall and spring scores tend to be highly correlated. Once you've accounted for fall scores, there may not be much variability left in the spring scores that is not due to error.

6) Whatever the initial intention, there is usually a tendency to rely more heavily on standardized assessments than on "softer" assessments, probably due to the apparent specificity of a number and the ease with which numbers can be compared. (Think of how the SAT test is used at some colleges.) But everyone agrees that standardized tests capture only part of what we want students to learn.

7) Test prep is fine by me, if teachers are prepping students for a good test. Which test are we talking about using?

Some of these points can be seen on a video that I made (titled "Merit Pay, Teacher Pay and Value Added Measures") and available by clicking here.

What was Randi Weingarten, president of the American Federation of Teachers, thinking when she suggested that student test scores might contribute to teacher evaluation?

I hope she was thinking that she's playing ball with the Obama administration, and that locals will do whatever they want anyway. Growth models don't yield meaningless data, but they are not good enough to evaluate individual teachers. Not yet.