绩效管理改革的效果:来自美国联邦政府的证据

从1993年的《政府绩效与结果法案》(GPRA)以及布什政府的项目评估定级工具(PART)已经建立了一种新的政府管理惯例,两者的目的均在于促进美国联邦政府绩效管理实务的发展。但从绩效信息的使用方面来看,两者所做出的努力都极其有限,即一旦考虑其它因素,机构雇员不论是否参与PART评审,是否参与GPRA的实施进程,他们几乎都没有影响到绩效信息的使用。本文采用顺序Probit回归模型,通过对美国联邦审计总署基于1996、2000、2003和2007年的数据进行调研,最终得出依赖于难以观测的官僚行为来实现的政府改革是具有局限性的以及在政府绩效改善的过程中绩效信息使用的重要性。对数据的深层分析还揭示了一系列影响绩效信息使用的组织因素,这包括领导对结果的承诺、监管者引导学习惯例、工作动机的性质、将测量与行动链接起来的能力、管理的自由裁量权以及利益相关者之间的政治冲突。

The Effects of Performance Management Reforms:Evidence from the United States Federal Government

The objective of this paper is to illustrate and examine "The Government Performance and Results Act(GPRA) " of 1993 and the Bush administration’s "Program Assessment Rating Tool(PART) " have not accomplished the performance information use.The GAO administered surveys in 1996,2000,2003,and 2007 to collect data on the implementation of performance management reforms in federal agencies.We employ ordered probit regression models due to the categorical nature of the dependent variables.Using data from two surveys,we find that the involvement of agency employees with PART reviews and GPRA processes generally had little direct effect on performance information use once other factors are accounted for.The main exception is that involvement in GPRA processes and PART reviews succeeded in encouraging employees to use performance data to further refine measures and goals.This reflects the limits of government-wide reform efforts that depend upon bureaucratic behavior that is difficult to observe.We also find that a series of organizational factors———leadership commitment to results,learning routines led by supervisors,the motivational nature of the task,and the ability to link measures to actions,managerial discretion,and political conflict among stakeholders are positive predictors of performance information use.We find that measurement and motivation matter to performance information use.When managers report that measurement is more difficult or when it is more difficult to link measures to actions,performance information use is lower.This reflects the fact that while all programs may be awash with data,whether the data is insightful or not may vary quite a bit.