Information and Employee Evaluation

The impact of providing [objective performance] information [on teachers] is greater when [it is] ...more precise and is smaller when principals have already supervised their teachers for a greater number of years.

As part of the incentives built into its $4.3 billion Race to the Top Fund, the federal government is prodding states and school districts to use data on growth in student achievement to measure teacher effectiveness, and then to implement policies to "recruit, develop, reward, and retain effective teachers." But some would argue that teaching is a multidimensional task with varying goals and that basing employee evaluations strictly on objective performance measures may lead to dysfunctional behaviors, including cheating on standardized tests.

In Information and Employee Evaluation: Evidence from a Randomized Intervention in Public Schools, (NBER Working Paper No. 16240), co-authors Jonah Rockoff, Douglas Staiger, Thomas Kane, and Eric Taylor examine how school principals gauge the performance of individual teachers in their schools, and how such a process is affected by the addition of objective data on teachers' success in raising students' standardized achievement test scores. This study is based on the results of a randomized pilot program conducted by the New York City Department of Education in its public schools during the school year 2007-8. Roughly one quarter of eligible school principals participated in the study, with half of them selected at random to receive objective performance data and training in how these "value-added" performance metrics were created.

Rockoff and his co-authors first document a positive correlation between objective estimates of teacher performance based on student test data and the principals' prior subjective beliefs. The strength of this correlation rises with the precision of the objective estimates and with the length of the principals' relationship with the teacher. Principals provided with this objective data incorporate it into their subsequent evaluations, and do so in a way that is consistent with economic theories about employer learning. The impact of providing information is greater when value-added estimates of teachers are more precise and is smaller when principals have already supervised their teachers for a greater number of years.

As a result of principals incorporating the new performance data into their evaluations, those teachers with lower value-added estimates are more likely to exit teaching in the school. There is also a small but marginally significant improvement in student test scores the following year, consistent with these changes in selective retention of the better teachers. The authors' results suggest that objective performance data provides useful information to principals in constructing employee evaluations and using these evaluations to improve productivity.

-- Frank Byrt