The ability to examine teacher performance on an annual basis is a potentially powerful approach to learning more about which teachers perform well and what efforts on the parts of schools and districts support greater teacher improvement. However only about half of teachers who have value-added scores in their first year have value-added estimates in two of the next four years (and only 20 percent had value-added in all of next four years). We are struck by this remarkably high rate of missing information about teacher performance among teachers who were clearly still members of the teacher workforce throughout the first five years.

There are a number of clear reasons that could account for missing value-added scores—e.g., permanent attrition from the workforce, switching to a non-tested subject or grade, insufficient numbers of tested students in a given year, unexpected leaves of absence. Of more interest, however, is the possibility that some of those explanations could be strategic on the part of teachers and/or principals. Strategic teacher re-assignments may occur when principals confront the statewide test-based accountability pressures that are typically specific to math and reading subjects in third through eighth grade. For example, principals could be strategic about which teachers are assigned to tested subjects and grades, in which case we expect that teachers who appear more often in these classrooms might also be relatively more effective. Further, if there is evidence that principals attempt to be strategic about how they assign teachers of different effectiveness, then one might wonder whether these strategies are beneficial to student achievement. To our knowledge, no research has attempted to distinguish between the inevitable noisy movement of teachers within a school system and the possibility that teacher re-assignments could be more strategic or systematic in nature.

In this paper, we use value-added scores of New York City teachers over the last ten years to document the frequency and patterns with which new teachers fail to have a complete history of performance measures over time. These patterns provide a first signal regarding the randomness of missing value-added. We examine whether there is any initial evidence that those teachers with fewer value-added scores during their tenure in New York City are of different early career skill than those with more value-added scores. We further explore additional hypotheses that are consistent with the hypothesis that teachers and/or principals make strategic decisions in their movements within the profession. For instance, teacher movements to non-tested subjects or grades might be more strategic if we observe that relatively low performing teachers are moved out of tested subjects and grades permanently, either by separating them from the system or reassigning them to other grades or subject, rather than sporadically. It may also be true that some principals engage in this behavior more than others. Finally, we hypothesize that schools under greater pressure from No Child Left Behind should also be more likely to allocate the best teachers to tested grades.