The Teaching Experience

** Also posted here on “Valerie Strauss’ Answer Sheet” in the Washington Post.

The topic of teacher experience is getting a lot of attention in education debates.  In part, this makes sense, since experience (years of service) does play several important roles in education policy, including teachers’ raises and transfer/layoff policies. 

Usually, experience is discussed in terms of its relationship to performance –whether more experienced teachers produce larger student test score gains than less experienced teachers.  There is a pretty impressive body of research on this topic, the findings of which are sometimes used to argue for policy changes that eliminate the role of experience in salary and other employment policies.  Proponents of these changes often argue that experience is only weakly related to performance, and therefore shouldn’t be used in determining salary and other conditions of work.  It is not unusual to hear people say that experience doesn’t matter at all.

As is often the case when empirical research finds its way into policy debates, the “weakly related” characterization of the findings on the experience/achievement relationship borders on oversimplification, while the claim that experience doesn’t matter is flat-out wrong.  The relationship is substantial but context-dependent, and blanket statements about it often hide as much as they reveal.

The dozens of analyses of teacher experience show that it matters a great deal in the early years on the job (also see here, here, here and here).  There is general consensus that the returns to experience are strongest in the first year of teaching.  Then the rate of improvement starts to level off quickly – usually stagnating within about 4-5 years (time frames vary a bit).  After that, most teachers tend to remain relatively stable in terms of their effects on student test scores (though a very large proportion leaves the profession before that point).

But these overall findings ignore the fact that the experience/achievement relationship differs a great deal by context.  For instance, the returns to experience appear to vary by where teachers work.  The relationship is more consistent among elementary school teachers (especially compared with those in high schools).  The effect of experience on teacher productivity may also be mediated by the quality of their peers in the same school – i.e., that novice teachers with more effective peers in the same school do better.

Similarly, there is evidence that experience matters less – or less consistently – in poorer schools (also see here).  There are several plausible explanations for this discrepancy, such as the possibility that teachers in poorer schools burn out more rapidly, or that there are difficulties in teaching lower-income children that are harder to adjust to.

The experience factor not only varies by where you teach, but also by what you teach.  Math teachers seem to improve more quickly (and consistently) than reading teachers, while newer evidence suggests that the same is true for teachers who remain in the same grade for multiple years.

Finally, it bears mentioning the obvious point that the effect of teacher experience might be totally different if we were able to look at outcomes other than test scores.  The idea that experience doesn’t matter after five or so years incorrectly implies that test scores are the only relevant outcome.  Nobody believes that is the case.  (And, for what it's worth, teachers with whom I’ve spoken find the idea that they stop improving after four or five years laughable.)

Teachers who produce test score gains are not always the same ones who are effective at imparting other types of skills.  It seems quite plausible (if not probable) that teachers do exhibit longer-term improvement in their students’ learning other skills, such as social/behavioral skills, that elude standardized tests.  Let’s also keep in mind that it remains an open question whether the returns to experience follow a similar pattern among teachers not in tested grades or subjects (which is roughly three in four teachers).

That said, experience is actually one of the very few observable teacher characteristics that is consistently correlated with achievement, and its effect is among the strongest, especially for some sub-groups, such as elementary school and math teachers.

Even those who think the magnitude of these returns is not commensurate with the role of experience in education policy cannot dispute that it is still a proven signal of quality, at least during the early years of teachers’ careers.  And it is virtually certain that teachers also improve in other ways that don't show up in their students' test scores.

So, unless we are going to design employment policies based strictly on test scores (which is both ridiculous and logistically impossible), we might recalibrate these policies to exploit the findings above, including using other measures along with experience, restructuring salary schedules, keeping teachers in the same grade over multiple years, or paying more attention to the important role of peers in shaping teachers' learning curves. Let's tone down the rhetoric, and try not to throw out the baby with the bathwater.