Page 197 - 33-3

NTU Management Review Vol. 33 No. 3 Dec. 2023

behaviors towards others, their behaviors would be positively attributed due to their
physical attractiveness (Hawley, Johnson, Mize, and McNamara, 2007). Finally, the
politician literature also suggests that attractive politicians do get a “break” when involved
in scandals (Stockemer and Praino, 2019). Therefore, we also expect that leaders with

high attractiveness can attenuate the negative abusive supervision-leader effectiveness
relationship, or even make it positive (Hypothesis 2).
Third, based on the perspectives of abusive supervision as political activity (Tepper

et al., 2012), leaders might engage in abusive behaviors to control subordinates’ behaviors.
Therefore, we added subordinates’ supervisor-directed helping as the outcome of leader
effectiveness, and investigated whether abusive supervision from politically skilled or
physically attractive leaders can increase subordinates’ enactment of supervisor-directed
helping via enhanced perceived leader effectiveness (Hypotheses 3 and 4).

2. Research Design/Methodology

To test the proposed hypotheses, we collect data from 113 leaders and 367
subordinates at two different time points from diversified occupations and industries in
Taiwan, resulting in 367 valid paired responses. At Time 1, we ask the subordinates to
evaluate abusive supervision (five items) (Mitchell and Ambrose, 2007), whereas the
leaders rate their political skills (six items) (Ferris, Davidson, and Perrewé, 2005). Two

weeks later (Time 2), we ask the subordinates to rate physical attractiveness (four items)
(Tsai, Huang, and Yu, 2012) and leader effectiveness of their leaders (four items) (Hsu
and Cheng, 2004), whereas the leaders rate the subordinates’ supervisor-directed helping
behaviors (six items) (Dalal, Lam, Weiss, Welch, and Hulin, 2009). Finally, demographic

variables (e.g., leaders’ and subordinates’ gender, tenure with the leader), subordinates’
perceived leader power (three items) (Wu, 2008), negative emotions (three items) (Bono,
Foldes, Vinson, and Muros, 2007) and Leader-member Exchange (LMX) (seven items)
(Graen and Uhl-Bien, 1995) are also collected as control variables.

As our data are nested in nature (i.e., subordinates’ data are nested within the leaders),
we perform multilevel path-analysis to test our hypotheses using Mplus 7.4. Specifically, at
the Level-1 (subordinate-level), we specify subordinates’ perceived abusive supervision as
predictor, leader physical attractiveness as moderator, leader effectiveness as the mediator,

189