Knowing that I’ve been writing a fair amount about various methods for attributing student achievement to their teachers, several colleagues forwarded to me the recently released standards of the Council For the Accreditation of Educator Preparation, or CAEP. Specifically, several colleagues pointed me toward Standard 4.1 Impact on Student Learning:
4.1.The provider documents, using value-added measures where available, other state-supported P-12 impact measures, and any other measures constructed by the provider, that program completers contribute to an expected level of P-12 student growth.
Now, it’s one thing when relatively under-informed pundits, think tankers, politicians and their policy advisors pitch a misguided use of statistical information for immediate policy adoption. It’s yet another when professional organizations are complicit in this misguided use. There’s just no excuse for that! (political pressure, public polling data, or otherwise)
The problems associated with attempting to derive any reasonable conclusions about teacher preparation program quality based on value-added or student growth data (of the students they teach in their first assignments) are insurmountable from a research perspective.
Worse, the perverse incentives likely induced by such a policy are far more likely to do real harm than any good, when it comes to the distribution of teacher and teaching quality across school settings within states.
First and foremost, the idea that we can draw this simple line below between preparation and practice contradicts nearly every reality of modern day teacher credentialing and progress into and through the profession:
one teacher prep institution –> one teacher –> one job in one school –> one representative group of students
The modern day teacher collects multiple credentials from multiple institutions, may switch jobs a handful of times early in his/her career and may serve a very specific type of student, unlike those taught by either peers from the same credentialing program or those from other credentialing programs. This model also relies heavily on minimal to no migration of teachers across state borders (well, either little or none, or a ton of it, so that a state would have a large enough share of teachers from specific out of state institutions to compare). I discuss these issues in earlier posts.
Setting aside that none of the oversimplified assumptions of the linear diagram above hold (a lot to ignore!), let’s probe the more geeky technical issues of trying to use VAM to evaluate ed school effectiveness.
There exist a handful of recent studies which attempt to tease out certification program effects on graduate’s student’s outcomes, most of which encounter the same problems. Here’s a look at one of the better studies on this topic.
Specifically, this study tries to tease out the problem that arises when graduates of credentialing programs don’t sort evenly across a state. In other words, a problem that ALWAYS occurs in reality!
Researchy language tends to downplay these problems by phrasing them only in technical terms and always assuming there is some way to overcome them with statistical tweak or two. Sometimes there just isn’t and this is one of those times!