TY - JOUR
T1 - Learning from Performance Information
AU - Andersen, Simon Calmar
AU - Nielsen, Helena Skyt
PY - 2020
Y1 - 2020
N2 - Years of research on performance management has generally concluded that performance information is seldom used purposefully by public managers and that it does not improve performance as intended. More recently, both theoretical and empirical work have begun to focus on situations in which performance management may facilitate internal organizational learning. In this study, we focus on one key component in performance management systems, namely generation of performance information. Based on a Bayesian learning model, we argue that generation of performance information at the individual level may create performance improvements because both users and frontline workers may learn where to prioritize their efforts. To test the isolated effect of this key component of any performance management system, we use as-good-as-random variation in exposure of students to testing because of a technical breakdown in an IT system. We identify the effect of testing on student learning measured two years after the breakdown. Results show positive and statistically significant effects of about 0.1 standard deviations, which is comparable to much more expensive interventions, and effects tend to persist after four years. We find larger effects for students with low socioeconomic status but also that schools with many students from this group are more reluctant to measure their performance. Implications and limitations in terms of increasing the level of student testing are discussed.
AB - Years of research on performance management has generally concluded that performance information is seldom used purposefully by public managers and that it does not improve performance as intended. More recently, both theoretical and empirical work have begun to focus on situations in which performance management may facilitate internal organizational learning. In this study, we focus on one key component in performance management systems, namely generation of performance information. Based on a Bayesian learning model, we argue that generation of performance information at the individual level may create performance improvements because both users and frontline workers may learn where to prioritize their efforts. To test the isolated effect of this key component of any performance management system, we use as-good-as-random variation in exposure of students to testing because of a technical breakdown in an IT system. We identify the effect of testing on student learning measured two years after the breakdown. Results show positive and statistically significant effects of about 0.1 standard deviations, which is comparable to much more expensive interventions, and effects tend to persist after four years. We find larger effects for students with low socioeconomic status but also that schools with many students from this group are more reluctant to measure their performance. Implications and limitations in terms of increasing the level of student testing are discussed.
UR - http://www.scopus.com/inward/record.url?scp=85092065692&partnerID=8YFLogxK
U2 - 10.1093/jopart/muz036
DO - 10.1093/jopart/muz036
M3 - Journal article
SN - 1053-1858
VL - 30
SP - 415
EP - 431
JO - Journal of Public Administration Research and Theory
JF - Journal of Public Administration Research and Theory
IS - 3
ER -