Association for Public Policy Analysis & Management

Performance management is often criticized for causing gaming behaviors. One of them is cream skimming. For instance, teachers may focus on improving the performance for students who score just below or above the average level to boost the overall pass rate even though the other students especially those at the bottom are ignored.No Child Left Behind (NCLB) has a specific requirement to ensure equality and mitigate the problem. NCLB was passed in 2002 and required each state to report the results like “adequate yearly progress”(AYP), for both the student population and for particular “subgroups” of students, including English-learners and students in special education, racial minorities, and children from low-income families. This paper examines whether performance management after NCLB mitigate the problem of cream skimming in U.S high schools (focusing on students with the highest potential for meeting proficiency while ignoring others). In this paper, performance management refers to a series of activities of setting standards, continuously monitoring performance, periodically rating principals and teachers’ performance and linking the test scores and the evaluation results with rewards and sanctions.

As schools which implement performance management better can be different from schools lagging behind in this aspect, we used TWANG package in R to match schools in top quartile, second quartiles, third quartile and bottom quartiles of performance management scores on several school-level and aggregate individual-level measures.Using multilevel models with weight created in the matching, we found that performance management reduced the overall standardized math score slightly, but it moderated the negative effect of being in the minority and speaking English as a second language. The result remains robust after filling in missing values.

The paper contributes to the literature of public management in at least three ways. First, it provides a systematic and empirical examination on whether the requirements for “subgroup population” reporting achieve its effect on mitigating cream skimming. Despite the abundant anecdotes of gaming behaviors caused by NCLB, there have been very few studies systematically examining the issue. Filling the gap is necessary. Otherwise, policymakers could only rely on anecdotes to evaluate the success of NCLB, and, by extension, performance management.

Second, the paper sheds light on the possibility that relationship between extrinsic and intrinsic motivation is not confined to being crowding-out, a common view in the public management literature. It supports the view in organizational theory that the relationship between intrinsic and extrinsic motivation depends on the organizational and institutional contexts.

Third, the multilevel models allow us to distinguish effects between different units of analysis and examine how they could interact with each other. Studies on performance management usually aggregate individual measure at the school level and fail to distinguish between the within-group and the between-group effect of being in certain demographic groups. Using multilevel models, the paper is able to tease out the effects at different levels and focus on how school-level performance management affects individual standardized math performance.