Monday, October 17, 2011

Does “value-added” add value?

A couple of months back, the Center for Public Education (the research arm of NSBA) came out in support of “value-added” teacher evaluation models that track student standardized test scores, tied to individual teachers. The issue has become particularly relevant in Pennsylvania, where such a model is being piloted state-wide.  However, their arguments strike me as unpersuasive and frankly, somewhat disconcerting.

Their central point is that the current system “is lacking” – as they see it, almost by definition – because only about 1% of teachers nationwide are identified each year as “unsatisfactory”. Therefore, their argument goes, as flawed as “value-added”  might be, anything would be an improvement over current practice. In fact, they directly acknowledge the unreliability of value-added data – only about a third of teachers ranking in the top 20% one year are similarly ranked the following year – but, well, it’s better than what we’ve got.

“It’s better than what we’ve got” sets the bar pretty low, don’t you think?  One thing that’s overlooked is that the current, flawed system is doing exactly what it was intended to do:  it identifies the small percentage of teachers who are performing so poorly that they deserve to be fired. It was not designed for the purpose of improving instruction and student learning.

Almost certainly, there are far better ways to accomplish those goals than to rely on test scores that don’t even attempt to measure much of what’s really important. (Note: no mention is made of how we’re going to evaluate teachers of subjects that aren’t currently tested – let’s not give them any ideas – or teachers on teams, or kindergarten teachers…) I bet that if we asked teachers, they could suggest evaluation models that would actually help to make them better teachers!

A related argument that seems to be gaining credibility is that overall teaching quality would be greatly improved if the 5-10% of our worst “performing” teachers were cycled out of the profession every year. (i.e., fired)

Really? I’m not sure we’ve thought this all the way through. Are we also in favor of “firing” 5-10% of our doctors every year? How about policeman? An organization that needs to fire 5% of its employees every year is not doing a very good job of professional development, or of hiring the right people in the first place.

The case for “value-added” violates a fundamental rule of research: you cannot use a measurement tool that was developed for one purpose (student “achievement”) to measure something else (teacher effectiveness) unless you have validated it for that purpose. (A pet peeve: why do we say we’re measuring student “achievement”, when what we’re measuring is a relatively narrow range of student ‘knowledge’ and/or skills? But I digress.)

Another point they make – presumably in support of their argument -  is that “improving teacher effectiveness can dramatically impact student learning”. Well, duh!  The rather important missing link is that there’s no evidence that using value-added scores improves teacher effectiveness!  Astonishingly, they also claim that “teachers have the single greatest impact on student performance, more than family background, socioeconomic status, or school.”  What’s astonishing is that the statement is flat-out wrong.

Finally, they actually say “there are ways to improve value-added models” - also allegedly in support of their main argument. Here’s a thought: why then, don’t we try improving the models before we go about implementing something that could easily do more harm than good?  It makes my head hurt.

The conventional wisdom among many so-called education reformers has been that teachers resist all kinds of evaluation, but in fact they're open to a number of ideas – especially if they’re designed to inform and improve instruction, and not used as swords of Damocles.