Although much of the leadership research has focused on constructive leader behaviors, recently researchers has shown their interests in dysfunctional or destructive leader behaviors that can be referred as abusive supervision. Abusive supervision, defined as “subordinates’ perceptions of the extent to which their supervisors engage in the sustained display of hostile verbal and nonverbal behaviors, excluding physical contact”, is one of the most common types of dysfunctional leadership. Previous studies have found that abusive supervision correlates negatively with subordinates’ job and life satisfaction, task performance, and OCB. But, knowledge sharing, defined as individual behaviors that involve sharing task-related ideas, information, and suggestions with others at workplace including coworkers and supervisors, is the fundamental means for accomplishing organizational effectiveness, there is a paucity of studies examining how abusive supervision may become a barrier of individuals’ knowledge sharing. So, in this study, we examine abusive supervision as a barrier to knowledge sharing, both because of this close relationship with discretionary behaviors and the destructive outcomes that occur when employees perceive they are receiving abusive treatment from their supervisors. Furthermore, our study examine this relationship drawing on insights from Psychological Reactance Theory. Psychological Reactance Theory contends that any persuasive message may arouse a motivation to reject the advocacy and when ones freedom is lost he or she may engage in a variety of behaviors in order to regain or prevent the loss of freedom. Abused subordinates often experience frustration along with a diminished sense of personal control, and one way they can restore this sense of autonomy and freedom is by intentionally withholding discretionary behaviors like knowledge sharing. And in order to clarify and investigate the mechanisms between these relationships, we consider the Mediating role of Leader Endorsement and Moderating effect of Self-Efficacy. Out of the process, we developed and empirically tested the following hypotheses: H 1: Abusive supervision is negatively related to knowledge sharing. H 2: Leader endorsement mediates the relationship between abusive supervision and knowledge sharing. H 3: Self-Efficacy moderates the negative relationship between abusive supervision and leader endorsement, such that the negative relationship will be stronger when self-efficacy is high than when it is low. H 4: Self-Efficacy moderates the indirect relationship between abusive supervision and knowledge sharing, such that the negative relationship becomes stronger when self-efficacy is high than when it is low. Data was collected by giving questionnaire to 360 pairs of a supervisor and his/her immediate employee. 300 questionnaires were collected. Among those, 255 dyadic data were used for analysis. The result of this study showed that abusive supervision was significantly and negatively related to employee’s job performance, suggesting the support of Hypothesis 1. Also, to test our hypotheses regarding the mediating role of leader endorsement, the result of mediation analysis suggested the support of Hypothesis 2. And there was a significant interaction effect of Self-Efficacy in the relationship between abusive supervision and leader endorsement, so Hypothesis 3 was supported. Finally, our hypotheses regarding the moderated mediating effect of self-efficacy was supported, suggesting the support of Hypothesis 4. In sum, Hypotheses 1, 2, 3, 4 were supported. This study’s implications, limitations and future research are discussed.