Abstract
While it is well recognized that performance management can induce gaming responses, the extant literature has mostly adopted a sole-goal perspective and neglected the influences of the interplay among multiple goals. This research attempts to fill this gap by bringing in the multi-goal perspective and investigating how performance feedback on the focal goal and targets set by higher-up authorities for the conflicting goal affect, independently and interactively, public organizations’ data manipulation in the focal goal. We posit that negative performance feedback on the focal goal makes public organizations more inclined to engage in data manipulation, and the target of a conflicting goal set by their higher-up authorities exerts influences by positively moderating this effect. Utilizing the context of China’s ‘environment-economy’ nexus, our empirical analyses using a novel approach to measuring cities’ manipulation of air quality data find robust evidence strongly supporting the theoretical arguments. This study represents an effort to open up a research agenda that engages the multi-goal perspective in gaming research.
Acknowledgments
I thank Editor Kaifeng Yang and three anonymous reviewers for their constructive and helpful comments. Any remaining errors are those of the author.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 An assumption underlying our theory is that performance failure in the focal goal will incur blame or punishment. The intensity or risk of being blamed/punished for low performance could vary across a wide range of contingencies, including the features of elements of the performance management system designed for the focal goal (e.g., the formal incentives attached to the evaluations and the scope of the disclosure of performance information) and the higher-up authorities’ priority placed on the goal, among others. Therefore, the impacts of negative performance feedback on public organizations’ data manipulation can be conditioned by various contingencies (i.e., factors such as the disclosure of performance information, formal punishments associated with low performance, and higher-up authorities’ attention to the goal, among others, are potential sources for the blame/punishment and thus potential boundary conditions for the impact of negative performance feedback on data manipulation). However, given that this assumption holds, our prediction that negative performance feedback increases public organizations’ tendencies to engage in data manipulation (ceteris paribus) should be upheld. In other words, our prediction should be largely applicable to performance management systems where those performing poorly in the focal goal face the risk of being blamed or punished.
2 Discussions on and empirical analyses of the potential non-linear relationship between performance and data manipulation are presented in the appendices (Part A).
3 The coverage of ranked cities was 169 at first. However, a city (Laiwu) was incorporated into another city and removed from the ranking since January 2019, making the final number 168.
4 Results using an alternative measurement that captures the different extent of negative performance feedback are presented in the appendices (Part A). The results using the alternative measurement are highly robust and consistent. We present results using the dummy variable as the measurement in the main text because it was mostly used in the previous performance feedback literature.
5 Analyses using other approaches, such as the feasible generalized least squares (FGLS) estimations, were also performed to obtain heteroskedasticity- and autocorrelation-consistent standard errors (Wooldridge, Citation2010). Results are largely consistent with those using clustered robust standard errors.
6 In practice, the LATE at the cutoff is obtained by comparing conditional expectations of the outcome variable when approaching from the left () and from the right (
) of the cutoff.
7 Validity tests for the RD design, which includes the balance test of the predetermined covariates on either side of the cutoff and the density test of the forcing variable at the cutoff (Cattaneo et al., Citation2024), have been performed. All results, which are presented in the appendices (Part B), verify the validity of our RD design.
8 RD plots with higher-order specifications are presented in the appendices (Part B).
Additional information
Funding
Notes on contributors
Shaowei Chen
Shaowei Chen is Professor of Public Administration at the School of Public Affairs, Xiamen University, China. His research interests include environmental policy & politics, performance management, and central-local relations in China. His recent publications can be found in Public Management Review, Public Administration, Governance, International Public Management Journal, International Review of Administrative Sciences, Local Government Studies, and PNAS, among others.