Louis Freedberg/California WatchA wall is shown at a kindergarten classroom at Aspire Alexander Twilight Academy in Sacramento.
Passions ran high on the best way to evaluate teacher effectiveness at the first public discussion held in the wake of the Los Angeles Times' pathbreaking series, which included publishing the names of 6,000 elementary school teachers and ratings of their "effectiveness" in increasing students' test scores.
If there was any agreement, it was that the "value-added" methodology employed by the Times should not be used as the sole criterion on which to evaluate a teacher. Richard Rothstein, a visiting UC Berkeley professor from the Economic Policy Institute, argued that despite the L.A. Times insistence to the contrary, it has in effect encouraged using the method as a sole criterion by providing measures of a teachers' effectiveness based on test scores alone without providing any other measures of effectiveness.
There was also agreement that there is no agreement on what constitutes an "effective" teacher. Several of the panelists pointed out that if one doesn't know what constitutes an effective teacher, that will complicate considerably the task of trying to measure his or her effectiveness.
As the moderator, I was only able to take incomplete notes, so this posting is based on overall impressions. I'll update it when I can review the webcast, which can be viewed online beginning today.
The symposium opened with harsh critiques of the value-added methodology by two Berkeley statisticians, Mark Wilson and Sophia Rabe-Hesketh.
Wilson said the problems start with the tests used to measure student achievement, which in turn are used to measure teacher effectiveness. "A significant issue is what is not on the tests," said Wilson, who was on a National Academy of Sciences panel that urged caution in using value-added methodology.
Rabe-Hesketh was even more critical, noting that there is no way to measure directly how effective a teacher is with student Y compared with how effective he or she might be with student X if student X has a different family background or parental education, or attends schools with a different kind of student body, etc. To do so requires using complex statistical tools with uncertain results.
Value-added methods, she said, should not be used for "naming and shaming."
A contrary view came from the Hoover Institution's Eric Hanushek, who said that imperfect as the value-added methodology is, it is better than not doing anything at all – a view shared by L.A. Times investigative reporter Jason Felch, a principal author of the series. At least some of the flaws Rabe-Hesketh noted could be corrected, Hanushek said. Just listing the defects of value-added methodology, he said, gives an impression that value added methodology is "useless and it is not."
I asked Jason Felch whether the Times intended to update its database on teachers if a teacher improved his or rating, and he said they intended to do so. He also indicated that the Times now plans to move on to rating high school teachers, which he said was more complex than rating elementary school teachers. If the Times decided not to continue with the project and was unable to update the database, they would remove it from the web.
I also asked Felch if the Times had considered whether the series might discourage teachers from going into the profession or teaching at schools in tough neighborhoods. He said it was not journalism's role to anticipate the policy implications of their work.
Some written questions submitted by audience members underscored the passions, and the divisions, on the issue:
From a parent in the Oakland public school system for 17 years: Mr. Felch's introduction described how teachers long for feedback so they don't have to teach in isolation and improve their practice. Why don't "reforms" talk about providing these supports before using this [to] improve teaching approach?
From a San Francisco parent: Today we are hearing a lot of discussion of why L.A. Times did a bad thing, we have a war on teachers, etc. As least we are having a conversation. Where are the solutions? Where is the discussion of what evaluations work?
Parents cannot choose their teacher and have no information on what drives quality teaching. We need transparency and leadership from our educators!
A public school parent and teacher's wife: All urban principals I know are overwhelmed by their workload due to the lack of funding for administrators, as classroom needs take priority. Would adequate funding to meet schools' needs allow for more effective teacher evaluation?
David Plank, executive director of Policy Analysis for California Education, indicated that California lags behind many other states in the kinds of support and training it gives to teachers who need to upgrade their skills.
And state Sen. Carol Liu, D-Glendale, didn't hold out much hope that help would come from Sacramento, in this regard or any other. She suggested that because California did not succeed in the competition for Race to the Top funds, Sacramento may be less forceful in moving on its pledge to use student test scores to evaluate teachers. That was one of the requirements imposed by states applying for Race to the Top funds by the Obama administration.
Anthony Cody, a long-time Oakland educator and author of the Living in Dialog blog in Teacher Magazine, accused the L.A. Times of being one of the media institutions "waging a war against teachers." Cody, who earlier this year hosted a telephone conversation between teachers and U.S. Education Secretary Arne Duncan, cited the suicide last week of Rigoberto Ruelas, a teacher at Miramonte Elementary School in Los Angeles. Family members were reported as saying he was depressed at his being rated a "less effective" teacher by the L.A. Times. "He is the first victim of the war against teachers," Cody said.



Comments
The problem with most value added methods is that they don't take into account the demographics of the students. Give me a reasonable range of demographic data associated with test scores and I can do a better job of predicting how well a (large enough) group of students will do than just using the past performance of the students. Demographic data contains a number of proxies for factors outside of a teacher's control that have an effect on student performance.
Also, when using demographic data, something like a hierarchical model should be used. Why? Because a student who is poor, but not a minority will probably only do slightly worse than average. A student who is a minority, but not poor will also probably only do slightly worse than average. However, a student who is both a minority and living in poverty is at a very large disadvantage in school, as well as later in life. (This probably does not apply in a very homogeneous state, such as New Hampshire.) None of that is the fault of the teacher, but most methods for evaluating teachers do not take this into account. The combined effect of a teacher and a school is typically going account for less than half of a student's performance at school. It's just that our teachers and schools are the things that public officials can control, so guess who gets more than 50% of the blame.
People like to complain about tests, but when properly used, they can do a very good job of showing us not only where the worst problems are, but also the solutions are. The idea is to make predictions based on school demographics (and past performance too,) look for the schools that significantly exceed expectations on a regular basis, and then start using them as a model for other schools with similar demographics. Such a system would need to be partially implemented for a couple of years in order to work the bugs out before going full tilt, but looking at the places where there is success despite disadvantages is a good starting place for closing the achievement gap. The same system could be applied to individual teachers, but it would be best to use a rolling average over multiple years to introduce stability so that real trends aren't lost in the noise. Also, a school's administrators have an effect on the quality of education received by students, and if we're going to evaluate the teachers, administrators should be fair game too.
I do support assessing teachers and schools using test data (not as the sole measure, though) but only if the assessment methods remove as many factors as possible that are outside of the control of the teachers and schools and only if there is a built in mechanism for improvement.
via Twitter