A Causal Inference of Employee Empowerment on Organizational Performance in the US Federal Agency, Using a Propensity Score Matching Method

Ha, Hyesong; Ha, Hyesong

Employee empowerment is a vital principle of NPR and NPM reforms, and it has been popularly utilized in organizations of the public as well as private sectors to innovate organizational structure and environment and to improve its outcomes such as performance, job satisfaction, and commitment. Despite its widespread popularity, however, the understanding and effectiveness of employee empowerment have been hindered by a lack of empirical studies in the public sector. Even though there have been many empirical outputs on its use and consequences published by public management scholars, the critique might be still reasonably applicable to current studies since almost all of them have focused on correlations and limited in causality.

The goal of this paper is to make causal inference with more confidence on employee empowerment’s organizational performance, using a non-experimental method of propensity score matching (PSM) on 2012 Federal Employee Viewpoint Survey (FEVS) data. We assumed each employee’s perceived different levels of employee empowerment as the difference in treatment intensity of employee empowerment provided by managers of each agency. Based on this assumption, the employees’ perceived empowerment indexes through the principal component analysis (PCA) are divided into four levels (least, fair, better, best). Next, these four levels are paired into three groups, each of which consists of one of three levels (the least, the better, or the best treatment ) referenced to the fair level of empowerment treatment: (least vs. fair) (better vs. fair) (best vs. fair). So, in each paired group, the former works as a treatment group while the latter does as a reference or control group. Then, for each paired group, we matched employees (in the treatment vs. reference groups) to have same propensity score in employee empowerment, using agency’s demographic and organizational characteristics as observable selection variables. After checking common support areas and balances on propensity score and observable selection variables for each three matched paired groups, various estimation techniques are applied including doubly robust propensity score weighting regression. Additionally, generalized propensity score and dose-response function for continuous empowerment treatment are applied to conduct a sensitivity test.

Results from various matching estimations for four levels of discrete empowerment treatment and continuous treatment showed that there are significant causal effects on organizational performance depending on treatment intensity of employee empowerment. For example, from the doubly robust propensity score weighting regression estimation, the first level that felt poor empowerment treatment showed half of performance (-0.506) perceived, compared to the second level that felt fair empowerment treatment as a reference level. The third or fourth levels that felt better or best revealed 0. 196 or 0.392 times higher perceived performance than the reference level, respectively. Also, the continuous treatment effect estimation showed that 1 unit increase in empowerment treatment perceived by individuals induced 0.467 unit increase in organizational performance perceived.

Finally, how to improve this first attempt in the public sector to estimate a causal effect of employee empowerment on organizational performance using PSM method is discussed.

Association for Public Policy Analysis & Management

Panel Paper: A Causal Inference of Employee Empowerment on Organizational Performance in the US Federal Agency, Using a Propensity Score Matching Method