Value-Added

A Few Points About The Instability Of Value-Added Estimates

One of the most frequent criticisms of value-added and other growth models is that they are "unstable" (or, more accurately, modestly stable). For instance, a teacher who is rated highly in one year might very well score toward the middle of the distribution – or even lower – in the next year (see here, here and here, or this accessible review).

Some of this year-to-year variation is “real." A teacher might get better over the course of a year, or might have a personal problem that impedes their job performance. In addition, there could be changes in educational circumstances that are not captured by the models – e.g., a change in school leadership, new instructional policies, etc. However, a great deal of the the recorded variation is actually due to sampling error, or idiosyncrasies in student testing performance. In other words, there is a lot of “purely statistical” imprecision in any given year, and so the scores don’t always “match up” so well between years. As a result, value-added critics, including many teachers, argue that it’s not only unfair to use such error-prone measures for any decisions, but that it’s also bad policy, since we might reward or punish teachers based on estimates that could be completely different the next year.

The concerns underlying these arguments are well-founded (and, often, casually dismissed by supporters and policymakers). At the same time, however, there are a few points about the stability of value-added (or lack thereof) that are frequently ignored or downplayed in our public discourse. All of them are pretty basic and have been noted many times elsewhere, but it might be useful to discuss them very briefly. Three in particular stand out.

The Year In Research On Market-Based Education Reform: 2012 Edition

** Reprinted here in the Washington Post

2012 was another busy year for market-based education reform. The rapid proliferation of charter schools continued, while states and districts went about the hard work of designing and implementing new teacher evaluations that incorporate student testing data, and, in many cases, performance pay programs to go along with them.

As in previous years (see our 2010 and 2011 reviews), much of the research on these three “core areas” – merit pay, charter schools, and the use of value-added and other growth models in teacher evaluations – appeared rather responsive to the direction of policy making, but could not always keep up with its breakneck pace.*

Some lag time is inevitable, not only because good research takes time, but also because there's a degree to which you have to try things before you can see how they work. Nevertheless, what we don't know about these policies far exceeds what we know, and, given the sheer scope and rapid pace of reforms over the past few years, one cannot help but get the occasional “flying blind" feeling. Moreover, as is often the case, the only unsupportable position is certainty.

The Sensitive Task Of Sorting Value-Added Scores

The New Teacher Project’s (TNTP) recent report on teacher retention, called “The Irreplaceables," garnered quite a bit of media attention. In a discussion of this report, I argued, among other things, that the label “irreplaceable” is a highly exaggerated way of describing their definitions, which, by the way, varied between the five districts included in the analysis. In general, TNTP's definitions are better-described as “probably above average in at least one subject" (and this distinction matters for how one interprets the results).

I’d like to elaborate a bit on this issue – that is, how to categorize teachers’ growth model estimates, which one might do, for example, when incorporating them into a final evaluation score. This choice, which receives virtually no discussion in TNTP’s report, is always a judgment call to some degree, but it’s an important one for accountability policies. Many states and districts are drawing those very lines between teachers (and schools), and attaching consequences and rewards to the outcomes.

Let's take a very quick look, using the publicly-released 2010 “teacher data reports” from New York City (there are details about the data in the first footnote*). Keep in mind that these are just value-added estimates, and are thus, at best, incomplete measures of the performance of teachers (however, importantly, the discussion below is not specific to growth models; it can apply to many different types of performance measures).

Creating A Valid Process For Using Teacher Value-Added Measures

** Reprinted here in the Washington Post

Our guest author today is Douglas N. Harris, associate professor of economics and University Endowed Chair in Public Education at Tulane University in New Orleans. His latest book, Value-Added Measures in Education, provides an excellent, accessible review of the technical and practical issues surrounding these models. 

In particular, how should “value-added” measures be used in teacher evaluation? President Obama’s Race to the Top initiative expanded the use of these measures, which attempt to identify how much each teacher contributes to student test scores. In doing so, the initiative embraced and expanded the controversial reliance on standardized tests that started under President Bush’s No Child Left Behind.

In many respects, The Race was well designed. It addresses an important problem - the vast majority of teachers report receiving limited quality feedback on instruction. As a competitive grants program, it was voluntary for states to participate (though involuntary for many districts within those states). The Administration also smartly embraced the idea of multiple measures of teacher performance.

But they also made one decision that I think was a mistake.  They encouraged—or required, depending on your vantage point—states to lump value-added or other growth model estimates together with other measures. The raging debate since then has been over what percentage of teachers’ final ratings should be given to value-added versus the other measures. I believe there is a better way to approach this issue, one that focuses on teacher evaluations not as a measure, but rather as a process.

Value-Added, For The Record

People often ask me for my “bottom line” on using value-added (or other growth model) estimates in teacher evaluations. I’ve written on this topic many times, and while I have in fact given my overall opinion a couple of times, I have avoided expressing it in a strong “yes or no” format. There's a reason for this, and I thought maybe I would write a short piece and explain myself.

My first reaction to the queries about where I stand on value-added is a shot of appreciation that people are interested in my views, followed quickly by an acute rush of humility and reticence. I know think tank people aren’t supposed to say things like this, but when it comes to sweeping, big picture conclusions about the design of new evaluations, I’m not sure my personal opinion is particularly important.

Frankly, given the importance of how people on the ground respond to these types of policies, as well as, of course, their knowledge of how schools operate, I would be more interested in the views of experienced, well-informed teachers and administrators than my own. And I am frequently taken aback by the unadulterated certainty I hear coming from advocates and others about this completely untested policy. That’s why I tend to focus on aspects such as design details and explaining the research – these are things I feel qualified to discuss.  (I also, by the way, acknowledge that it’s very easy for me to play armchair policy general when it's not my job or working conditions that might be on the line.)

That said, here’s my general viewpoint, in two parts. First, my sense, based on the available evidence, is that value-added should be given a try in new teacher evaluations.

Does It Matter How We Measure Schools' Test-Based Performance?

In education policy debates, we like the "big picture." We love to say things like “hold schools accountable” and “set high expectations." Much less frequent are substantive discussions about the details of accountability systems, but it’s these details that make or break policy. The technical specs just aren’t that sexy. But even the best ideas with the sexiest catchphrases won’t improve things a bit unless they’re designed and executed well.

In this vein, I want to recommend a very interesting CALDER working paper by Mark Ehlert, Cory Koedel, Eric Parsons and Michael Podgursky. The paper takes a quick look at one of these extremely important, yet frequently under-discussed details in school (and teacher) accountability systems: The choice of growth model.

When value-added or other growth models come up in our debates, they’re usually discussed en masse, as if they’re all the same. They’re not. It's well-known (though perhaps overstated) that different models can, in many cases, lead to different conclusions for the same school or teacher. This paper, which focuses on school-level models but might easily be extended to teacher evaluations as well, helps illustrate this point in a policy-relevant manner.

Do Top Teachers Produce "A Year And A Half Of Learning?"

One claim that gets tossed around a lot in education circles is that “the most effective teachers produce a year and a half of learning per year, while the least effective produce a half of a year of learning."

This talking point is used all the time in advocacy materials and news articles. Its implications are pretty clear: Effective teachers can make all the difference, while ineffective teachers can do permanent damage.

As with most prepackaged talking points circulated in education debates, the “year and a half of learning” argument, when used without qualification, is both somewhat valid and somewhat misleading. So, seeing as it comes up so often, let’s very quickly identify its origins and what it means.

A Look At The Changes To D.C.'s Teacher Evaluation System

D.C. Public Schools (DCPS) recently announced a few significant changes to its teacher evaluation system (called IMPACT), including the alteration of its test-based components, the creation of a new performance category (“developing”), and a few tweaks to the observational component (discussed below). These changes will be effective starting this year.

As with any new evaluation system, a period of adjustment and revision should be expected and encouraged (though it might be preferable if the first round of changes occurs during a phase-in period, prior to stakes becoming attached). Yet, despite all the attention given to the IMPACT system over the past few years, these new changes have not been discussed much beyond a few quick news articles.

I think that’s unfortunate: DCPS is an early adopter of the “new breed” of teacher evaluation policies being rolled out across the nation, and any adjustments to IMPACT’s design – presumably based on results and feedback – could provide valuable lessons for states and districts in earlier phases of the process.

Accordingly, I thought I would take a quick look at three of these changes.

Schools Aren't The Only Reason Test Scores Change

In all my many posts about the interpretation of state testing data, it seems that I may have failed to articulate one major implication, which is almost always ignored in the news coverage of the release of annual testing data. That is: raw, unadjusted changes in student test scores are not by themselves very good measures of schools' test-based effectiveness.

In other words, schools can have a substantial impact on performance, but student test scores also increase, decrease or remain flat for reasons that have little or nothing to do with schools. The first, most basic reason is error. There is measurement error in all test scores - for various reasons, students taking the same test twice will get different scores, even if their "knowledge" remains constant. Also, as I've discussed many times, there is extra imprecision when using cross-sectional data. Often, any changes in scores or rates, especially when they’re small in magnitude and/or based on smaller samples (e.g., individual schools), do not represent actual progress (see here and here). Finally, even when changes are "real," other factors that influence test score changes include a variety of non-schooling inputs, such as parental education levels, family's economic circumstances, parental involvement, etc. These factors don't just influence how highly students score; they are also associated with progress (that's why value-added models exist).

Thus, to the degree that test scores are a valid measure of student performance, and changes in those scores a valid measure of student learning, schools aren’t the only suitors at the dance. We should stop judging school or district performance by comparing unadjusted scores or rates between years.

Colorado's Questionable Use Of The Colorado Growth Model

I have been writing critically about states’ school rating systems (e.g., Ohio, Florida, Louisiana), and I thought I would find one that is, at least in my (admittedly value-laden) opinion, more defensibly designed. It didn't quite turn out as I had hoped.

One big starting point in my assessment is how heavily the systems weight absolute performance (how highly students score) versus growth (how quickly students improve). As I’ve argued many times, the former (absolute level) is a poor measure of school performance in a high-stakes accountability system. It does not address the fact that some schools, particularly those in more affluent areas, serve  students who, on average, enter the system at a higher-performing level. This amounts to holding schools accountable for outcomes they largely cannot control (see Doug Harris' excellent book for more on this in the teacher context). Thus, to whatever degree testing results can be used to judge actual school effectiveness, growth measures, while themselves highly imperfect, are to be preferred in a high-stakes context.

There are a few states that assign more weight to growth than absolute performance (see this prior post on New York City’s system). One of them is Colorado's system, which uses the well-known “Colorado Growth Model” (CGM).*

In my view, putting aside the inferential issues with the CGM (see the first footnote), the focus on growth in Colorado's system is in theory a good idea. But, looking at the data and documentation reveals a somewhat unsettling fact: There is a double standard of sorts, by which two schools with the same growth score can receive different ratings, and it's mostly their absolute performance levels determining whether this is the case.