Why Strip Mining Might Shrink the Pie
(This post also appears on Rick Hess Straight Up.)
Yesterday, I discussed the recent Ed Next forum between Kati Haycock and Rick Hanushek, noting that I agree with Haycock’s focus on sensible strategies to get more good teachers into high-poverty schools but that I worry about the casual heavy-handedness with which some advocates tackle the issue. In particular, I suggested that reflexive efforts to shift “effective” teachers from high-performing schools and classrooms to others–such as attempts potentially countenanced in some language proposed for NCLB reauthorization–may actually reduce the pool of effective teachers. This would turn strip mining from an effort to redistribute the pie into a strategy that would actually shrink the size of the “good teaching” piece. Why might that be?
First, efforts to redistribute seemingly effective teachers may shift teachers from schools and classrooms where they are effective to those where they are not. Florida International University’s Lisa Delpit has noted that the skills which make a teacher effective with proficient, affluent students will not necessarily translate to schools serving a different student population. Moreover, there is substantial evidence that teacher effectiveness is contextual. Hanushek and Swarthmore’s Tom Dee have reported, for instance, that students appear to benefit from having a teacher of the same race. Duke University economists Charlie Clotfelter, Sunny Ladd, and Jacob Vigdor reported a few years ago that the effects of teacher experience in North Carolina varied with student race and family income. (I discuss a lot of this research here).
There is good reason to believe that teacher effectiveness is partly a function of some teachers being better suited for some students, schools, and contexts. As practitioner CCBACHEMDMS commented on my blog yesterday, “A teacher may be well-suited, even gifted, when facing one population, and completely lost in another.” To the extent that this is true, seemingly desirable teacher movement may bring much disruption for little gain. Indeed, encouraging teachers to move without attention to context and constraints could readily reduce the overall quality of teaching. This is not intended to counsel against finding ways to steer teachers where we think they’re needed, but it does suggest that such efforts should be carefully designed with an appreciation for perverse incentives (and that they should not be steered by directives from Washington).
Second, ill-conceived efforts to move seemingly effective teachers to more disadvantaged schools may prompt them to leave the profession at higher rates. The consequence would be to push out exactly those teachers we most want to retain. The University of Pennsylvania’s Richard Ingersoll has observed that teachers in high-poverty schools are almost twice as likely to leave teaching as teachers in medium-poverty schools. This is a well-documented finding, echoed by researchers at RAND and elsewhere. It would be a self-defeating, short-sighted strategy to systematically encourage effective teachers to work in those schools if the result is to accelerate the rate at which they leave the profession. Again, avoiding unintended consequences requires that strategies directing teachers to certain schools be pursued with careful attention to incentives, retention, and context.
Third, in determining the allocation of “effective” teachers, a critical problem is that while we know that good teachers matter enormously and have confidence in our ability to identify good teachers in various ways, we don’t have any reliable way to consistently identify good teachers from state capitals–much less from Washington. The “highly qualified teacher” provision of NCLB does not identify effective teachers. It identifies those with particular credentials, though there is much evidence that those credentials do almost nothing to predict student achievement.
Why not just judge teachers using value-added scores? For one thing, even in states that have the requisite data systems, such scores are only available for a minority of teachers. A more fundamental problem is that–even if we stipulate that these systems are capturing the things we care most about–the measures can be imprecise and of uncertain reliability when just two or three years worth of data are being used to judge individual teachers. Finally, equating effectiveness at boosting basic math and reading proficiency with broader teacher effectiveness presumes that math and reading are reliable proxies for teacher excellence more generally. To date, there is no evidence supporting this notion and much cause for sensible caution. I’m all in favor of using value-added metrics, but let’s be smart about it and not romanticize their power or be blind to their limitations. (For more on this count, see here).
The desire to ensure that we have lots more terrific teachers in high-poverty, low-performing schools is a laudable one. It’s absolutely worth acting on, aggressively. But let’s not undermine successful schools or systems along the way. And, let’s take care not to unintentionally undercut efforts to attract and retain more good teachers.