Association for Public Policy Analysis & Management

The effects of performance management on closing gap in math scores between students' subgroups have been mixed. On the one hand, proponents of performance management argue that attaching consequences to subgroups’ performances will galvanize teachers and administrators to improve not only average score but also subgroups’ score. On the other hand, skeptics of performance management believe that performance management will breed gaming behaviors entrenching the gap. For instance, the focus on proficiency goals will incentive teachers to focus their efforts on students in the middle of score distribution at the expense of students at both ends of score distribution. This paper tries to reconcile competing findings. Using data from Educational Longitudinal Study, the paper directly measures various aspects of performance management from standard setting, availability of teachers’ evaluation, enhancing principals’ autonomy to attaching consequences to performance for both teachers and students. It finds that performance management did not have an impact on 2004 math standardized math scores of students scoring above their school average in 2002, but it reduced 2004 math scores for those whose 2002 scores were below their school average. Also, performance management had mixed effects on historically disadvantaged subgroups of students. On the one hand, it has positive effects on students ever participating in English as the second language program, students ever participating in special education program and Native American students . On the other hand, it has a negative effect on black students.

Compared with previous studies, this paper directly measures the implementation of various aspects of performance management and their impacts on students' math score. Previous studies usually utilize the difference in difference strategies to gauge the impact of performance management and then attributes the difference between treatment and control group to performance management. Despite efforts to rule out competing explanations, studies are susceptible to the omitted variable bias because of the difficulty of exhausting all the plausible explanations. By contrast, the paper directly measures elements of performance management and their impacts on students’ math score. The elements include standard setting, evaluation of teachers, the frequency of tests, empowerment of principals,the connection between performance and consequences for both teachers and students.

Moreover, the multilevel model enables us to accurately and efficiently estimate the effect of school-level performance management on individual math test score. Most public management researches use aggregate variables at school level to examine the impact of management variables on performance. The approach can ignore the difference between within school effect and between school effect. In this paper, we are interested in whether school-level performance management measures can moderate the negative correlation of individual characteristic such as race, limited English proficiency, status of ever participating in special education, and low socioeconomic status with math test score.In this case, distinguishing the difference between within group and between group effect is critical to ensure unbiased estimates. Also, the multilevel model can address the paucity of individual observations in some schools by relying on the information provided by other schools.