Performance-oriented managerial practices and job attitude in public organizations

Abstract

With the growing popularity of pay-for-performance (PFP) program as a performance-oriented managerial practice in the public sector, many public administration scholars are raising concerns that the use of PFP in the public sector could have negative consequences for employees and organizations. This study mainly investigates how pay systems affects the job attitude of employees. Drawing on theory and prior studies, this paper hypothesizes that public manager’s job attitudes differ between agencies where pay-for-performance mechanism is adopted (PFP system) and those with no such system (GS system), expecting that public employees’ attitude is more influenced by monetary rewards in the former than in the latter. Hierarchical linear modeling with large-scale survey data is utilized to test the hypotheses. Consistent with expectations, the analytical results reveal that employee work satisfaction is more determined by extrinsic rewards in the PFP than in the GS system. This study discussed the empirical findings and implications in the paper.

Keywords:

• Pay-for-performance
• job attitude
• job satisfaction

Notes

1. The examples of job characteristics are resource and job quality, and those of job environment include pay, size of organization, discrimination, and family-friendly policy (Hopkins, 1983).

2. In some cases, job satisfaction is measured by a single questionnaire item in empirical studies: ‘All in all, how satisfied are you with your job?’ However, this uni-dimensional measure has some disadvantages: (1) it ignores other important features besides work itself as pay, relationships with supervisor, job security, and so forth: (2) the single measure may overestimate the degree of job satisfaction (Kalleberg, 1974).

3. In ML, this statement is true when the sample is large and it is normally distributed, which results in unbiased estimators. If this is not met, variance of coefficient is biased downward, which is a shortcoming of ML. Thus, we assume that our sample is large enough and the estimated variable is normally distributed in this model.