I’ve been meaning to comment on the public release of teacher performance data since the LA Times did a major data release several months ago. It is a complicated issue, and there are many reasons not to release the data. Bill Gates discussed some of them in a NY Times editorial yesterday.
Today the NY Times released its own batch of data, after the teachers union failed in its effort to block the school district from doing so. From reading the comments on the web site (many of them by teachers) I have come up with my own performance metric: every teacher who displays by their comments a fundamental misunderstanding of value-added measures should receive a negative evaluation. Some of this is willful misunderstanding of the sort characterized by Upton Sinclair—“It is difficult to get a man to understand something when his salary depends on his not understanding it”—but much of it is just plain old-fashioned ignorance of the type people go to college to avoid.
Most these misunderstandings take the form of “so many factors affect test scores that the teacher doesn’t have control over.” True enough. But value-added measures (when they are based on sufficient sample sizes) by and large remove these factors since the value-added measures capture the change in performance from year to year. The favorite thing to blame is socioeconomic status. But a teacher who wants to improve their value-added scores would prefer to teach poor, troubled kids, since there is way more potential upside to improve scores—and it is only the improvement from year to year that matters in value-added measures. The spoiled brats will do well regardless of teacher, so there is likely a much lower marginal return for teaching the privileged.
It is dispiriting to hear such important, revealing measures being disparaged by people who have every incentive to understand them, but clearly don’t. Equally unsettling is the notion that test scores don’t measure anything of value. It has become conventional gospel among the teaching establishment and, more generally, critics of NCLB, that “teaching to the test” is this horrible thing that no decent educator would want to do. But doesn’t it depend on the test and the subject? Standardized test scores are not by any means perfect measures of learning, but they do some things fairly well. And it seems hard to make the case that someone is a good teacher if his/her students consistently underperform on standardized tests. Is there really the trade-off between test performance and other educational outcomes not meausured by tests? I doubt it (and I doubt there is any evidence for it).
Put another way, is it too much to expect that our kids can think creatively and analytically and be able to divide fractions. I would like my kids to understand things like why we would want to divide fractions in the first place and how fractions relate to the real world they live in. But if they can’t actually divide fractions, aren’t the higher level analytical skills sort of meaningless? And isn’t the best way to determine if kids can divide fractions is to give them some fractions on a test and see if they can divide them. If teaching to the test means kids can actually take take two fractions and divide them and come up with the answer, while not teaching to the test means they can’t, then please, let us have teachers who teach to the test!
I have worries about disparaging and dispiriting teachers—most of whom work hard and do reasonably well. But evidence is accumulating in study after study that improving teacher performance is the key to educational reform. How to do that certainly isn’t as simple as publicly releasing data. However, on balance, efforts to bring accountability into the classroom have to be promoted, however painful.
And one clear signal that releasing this type of data is a good idea is that teachers’ unions oppose it. The best reform agenda that any public body could adopt is to go through the list of policy options that teachers unions hate and implement each one of them.