The Problems With Value-Added Assessment *
You asked what keeps me running, which I assume means how I find the energy to stay on the road week after week, speaking to teachers, parents, school board members, and concerned citizens. These days, I am running because of an inner rage at the attacks on teachers and public education. I see one of our most important public institutions under siege by people who want to privatize it, turn it into profit centers, and treat children as data points on a chart. This is wrong, and it will end badly. Critics say I defend the status quo, but nothing could be further from the truth. The status quo is awful, but the demonizing of teachers and the vilification of public education are even worse.
Last week, I was in Los Angeles. I spoke to L.A. teachers, who were shamed by the Los Angeles Times' disgraceful release of test-score data and ratings of 6,000 elementary teachers as more or less effective. I had previously believed that such ratings (value-added assessment) might be used cautiously by supervisors as one of multiple measures to evaluate teacher performance. The L.A. Times persuaded me that the numerical scores—with all their caveats and flaws—would drown out every other measure. And, in fact, the L.A. Times database contained only one measure, based on test scores.
And so I concluded that value-added assessment should not be used at all. Never. It has a wide margin of error. It is unstable. A teacher who is highly effective one year may get a different rating the next year depending on which students are assigned to his or her class. Ratings may differ if the tests differ. To the extent it is used, it will narrow the curriculum and promote teaching to tests. Teachers will be mislabeled and stigmatized. Many factors that influence student scores will not be counted at all.
The latest review of value-added assessment was written by New York University economist Sean Corcoran. He examines value-added assessment in Houston and New York City. He describes a margin of error so large that a teacher at the 43rd percentile (average) might actually be at the 15th percentile (below average) or the 71st percentile (above average). What is the value of such a measure? Why should it be used at all? Please read this important and well-written study.
While I was in Los Angeles, a teacher committed suicide. Rigoberto Ruelas, 39, had taught 5th graders for 14 years. He was known as unusually dedicated and caring; he worked in a gang-ridden, impoverished neighborhood. Most students in his school were English-language learners. Friends and family said he was depressed by the poor rating he received in the L.A. Times. No one will ever know what caused him to despair and take his own life. Colleagues and former students wrote beautiful tributes to him. They thought he was a wonderful teacher.
It's worth noting, however, that Los Angeles Deputy Schools Superintendent John Deasy said that Mr. Ruelas had a "great performance review" from his supervisors, but Mr. Deasy couldn't release the personnel records because they are confidential. So only the test scores were released to the media, not the laudatory reviews by professionals who observed his work.
Now I hear that more districts, prodded on by U.S. Education Secretary Arne Duncan and Race to the Top principles, want to release value-added rankings. More teachers will learn that they are subpar or superior when judged by flawed, dubious, inaccurate measures.
How many other ways can we discover to ruin teachers' reputations and encourage teachers to abandon their profession? Why isn't there a public outcry that such tactics undermine professionalism and the quality of education? When will we learn that we have turned education into a numbers racket, and we may lose the best teachers along with the worst?
In this week after NBC's one-sided slam against teachers, unions, and public education, I am furious. And it keeps me running.
*Editor's note: The headline on this blog has been changed at Diane Ravitch's request. Diane did not write the original headline.