The tone at the top: a trickle-down model of how manager anger relates to employee moral behaviour

ABSTRACT

The question of how leaders’ expressions of anger influence employees have been the subject of considerable scholarly debate. So far, however, research on the consequences of angry leadership has predominantly focused on the effects of supervisor expressions of anger, neglecting the potential influence of higher-level managerial anger. In this study, we integrate the emotions as social information theory with the adapted elaboration likelihood model to examine how manager anger trickles down across organizational hierarchical levels (i.e., managers, supervisors, and employees) to affect employee moral behaviour. Results of a multi-source field study conducted in Chile demonstrate that perceptions of manager moral behaviour and supervisor servant leadership serially mediate a negative relationship between manager anger and employee moral behaviour. Furthermore, counter to our predictions, trait negative affectivity of supervisors did not moderate the trickle-down relation of manager anger on employee moral behaviour. Our research elucidates the process by which manager anger can “set the tone” in an organization and trickle down across hierarchical levels to predict the moral behaviour of employees.

KEYWORDS:

• Anger
• moral behaviour
• servant leadership
• trickle-down model
• multilevel structural equation modelling

Data availability statement

The data that support the findings of this study are available from the corresponding author upon request.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Out of the reported fit indices for our study model, the CFI falls slightly below.90 (Browne & Cudeck, 1993; Schweizer, 2010), however, this model fit is likely due to the complexity of our model as well as the sample size of our study being below N = 250. Previous research indicates that CFI values tend to worsen as model complexity increases (Kenny & McCoach, 2003) and that the accuracy to evaluate fit of a number of indices such as the RMSEA and the SRMR decreases if sample size is below N = 250 (Hu & Bentler, 1999; Kenny et al., 2014). As a result methodologists caution to treat fit indices as “golden rules” and suggest to determine the adequacy of fit of a given model in comparison with alternative models (Marsh et al., 2004, p. 321) – an approach we adopted in the present paper.

2. As many of our study variables were self-rated, this may raise concerns regarding common method bias (CMB). We tackled these concerns in a number of ways. First, we provided a cogent theoretical rationale for the direction of our hypotheses based on the emotions and moral leadership literature. Second, we conducted a series of MCFAs to attest to the distinctiveness of our study variables and results did not suggest a common method factor indicative of CMB. Third, we minimized the risk of CMB in our study design (see Podsakoff et al., 2003) by separating predictors and outcome variables in the survey and randomizing items as part of our surveys. Forth, we conducted a marker-variable analysis (see Lindell & Whitney, 2001) following the approach of previous research (e.g., Rofcanin, de Jong et al., 2018). Specifically, we used the item “Because I need the income” in response to the question “Why are you motivated to do your work” tapping into employee extrinsic motivation taken from Grant and Berry (2011) as a marker variable. Conceptually, there is no plausible link between employee extrinsic motivation and manager anger. We then used the lowest observed correlation between the marker variable and our study variables as a proxy for CMB, subtracted it from the correlations among study variables, and divided the resulting coefficient by 1 – the lowest observed correlation to produce CMB-adjusted correlations. Following this method, large differences between unadjusted and CMB-adjusted correlations are indicative of a CMB issue. In our sample, however, those difference were minimal, ranging between 0.002–0.003. Therefore, it can be concluded that CMB did not affect our analyses.

3. We additionally probed the interaction effect using the Johnson-Neyman technique because regions of significance can exist in the absence of a significant interaction effect (Bauer & Curran, 2005). Our results indeed showed that for grand-mean centred values of supervisor trait negative affectivity between −0.86 and 2.79 the manager anger – manager moral behaviour relationship is significant and becomes more negative. To ascertain whether these results inform our conclusions about moderated mediation (i.e., our H5), we calculated the index of moderated mediation (Hayes, 2015) which, if significant, would indicate that any two conditional indirect effects defined by different values of the moderator are statistically different. Conversely, nonsignificance would denote that no two conditional indirect effects can be deemed different from each other and thus moderated mediation cannot be established. Results showed that the index of moderated mediation was not significant (γ = −0.02, 95% CI Low = −0.042; CI High = 0.030), thereby corroborating our conclusion that moderated mediation cannot be established.