The COVID-19 pandemic as a contextual factor
The events system theory (Morgeson et al., 2015) explains that majorly disruptive and novel
events can shape employee experiences. The COVID-19 pandemic is one such hazardous global
event that has caused serious disruptions to peoples’ lives and work (Karabag, 2020). Further, the
AET posits that events do not trigger emotions in a vacuum; the environment also influences
individuals’ emotional reactions to affective events (Weiss and Cropanzano, 1996). Thus, the
pandemic may constitute a wider exogenous affective event (see Min et al., 2021) that can
punctuate context (Johns, 2006), shaping supervisors’ experience and responses to deviance.
Specifically, the COVID-19 pandemic introduces new threats to life and financial
instability, which may affect supervisors. This threatening environment indicates an exogenous
emotional climate of extreme negative valence, defined as ‘an objective group phenomenon
that can be palpably sensed as and when one enters a party or a city and feels an attitude of […]
depression, […] or fear’ (De Rivera, 1992, p. 197). Past research confirms that large-scale
hazardous events, such as hurricanes and earthquakes are associated with a wide range of
negative emotions, such as anguish and sorrow (Cuddy et al., 2007, Sun et al., 2013). Recent
research further indicates the pandemic as a source of negative affective states, such as health
anxiety (Trougakos et al., 2020) and prolonged stress (Carstensen et al., 2020). A negative
exogenous affective event of such a magnitude is likely to be overwhelming, with the potential
to magnify the negative affective experiences of supervisors. Thus, we expect the felt leader
identity threat due to supervisor-directed deviance to be further exacerbated as the pandemic
continues.
Hypothesis 3a (H3a): The pandemic context moderates the indirect effect of supervisor-
directed deviance on retaliation against the subordinate through felt leader identity
threat by strengthening the relationship between supervisor-directed deviance and felt
leader identity threat compared to normal circumstances.
Hypothesis 3b (H3b): The pandemic context moderates the indirect effect of supervisor-
directed deviance on retaliation against the subordinate through felt leader identity threat
by progressively strengthening the relationship between supervisor-directed deviance
and felt leader identity threat from the first wave to the third wave of the pandemic.
In addition, exogenous hazardous events have often been associated with emotions such as
concern towards social groups affected by such events (Cuddy, Rock, and Norton, 2007; Sun,
Zagefka and Goodwin, 2013). Empathic concern, or feelings of warmth and compassion towards the
unfortunate (see Davis, 1980) is an ‘other-oriented’ emotion that is particularly relevant in the
context of the pandemic. Empathic concern constitutes a positive emotion towards people suffering
or at the risk of major health issues or job or income loss due to the pandemic. Thus, supervisors
whose self-definition as leaders is threatened by deviance but are also concerned about the effect of
the pandemic on others would be less likely to respond harshly to this threat and experience less
desire to punish a deviant subordinate. Past research confirms that empathic concern relates to
employees ‘helping and showing courtesy towards co-workers’ (Kalshoven,
Den Hartog, and De Hoogh, 2013, p. 212). Thus, we expect that supervisors’ empathic
concern during the pandemic will reduce the effects of felt leader identity threat on retaliation
against deviant subordinates.
Hypothesis 4 (H4): COVID-19-induced empathic concern moderates the indirect effect
of supervisor-directed deviance on retaliation against the subordinate through felt leader
identity threat, such that the relationship between felt leader identity threat and
subordinate-directed retaliation is stronger when empathic concern is low rather than
high.
Overview of studies
Study 1 initially tested H1 and H2 by manipulating supervisor-directed deviance in an
experiment conducted before the first COVID-19 case reported in the United States. The
pandemic presented an opportunity for a quasi-experimental approach in which individuals are
further assigned to two conditions (i.e., pandemic context vs. normal circumstances) which are
naturally occurring rather than manipulated (Grant and Wall, 2009). Thus, approximately one
year later, we re-ran the experimental study, which allowed for estimating the effect of
supervisor-directed deviance by comparing pre- with post-COVID-19 changes of the outcome
for the experimental and the control groups (i.e., examining the interaction of supervisor-
directed deviance and time, also referred to as difference-in-difference; see Sieweke and
Santoni, 2020). Study 2 replicated the experiment as the pandemic progressed in the US with a
Wave 1–Wave 3 (T1–T2) pandemic study design. In Study 3, we established external validity
through a field study during the third wave of the pandemic and tested the moderating role of
COVID-19-induced empathic concern. Each study’s design and results are described below.
Method
Sample and procedure. We collected data before the pandemic (Pre: 1 December 2019)
and towards the end of the third wave (Post: 14 February–6 April 2021) from American
supervisors recruited through Mechanical Turk Prime. From the 161 completed pre-COVID-19
online survey responses, we excluded two participants who did not pass the attention check,
resulting in a sample of 159 supervisors (41% of females). In the third wave, the 159
supervisors were contacted and assigned to the same experimental vs. control condition out of
which 97 completed the study (61% response rate). 38% were female and, on an average, they
had 22 years of work experience in various industries such as retail, education, and technology.
Only participants who completed both experiments were included in the analyses. Thus, our
final sample consisted of 97 participants (45 in the experimental group).
To test for the potential impact of attrition, we compared the pre-COVID-19 responses of
stayers (those present at both pre- and post-COVID-19 period) vs. leavers (those present only at pre-
COVID-19 period). A series of one-way analyses of variance (ANOVA) did not indicate any
significant differences between the two groups in terms of demographic or the main variables.
We also conducted a logistic regression with a dichotomous variable representing stayers vs. leavers
used as the dependent variable and all other variables as independent variables. None of these
variables were significant. Therefore, we concluded that our data was missing at random, and that
participant attrition was unlikely to affect our results (Goodman and Blum, 1996). Experimental
manipulation: Supervisor-directed deviance. We proxied supervisor-directed deviance with its
typical form of social undermining, defined as behaviour intended to damage an individual’s work
success, relationships, and reputation (Duffy et al., 2002). Participants were randomly assigned to
two conditions — supervisor-directed deviance and the control; they read two pilot-tested (n =
85) scenarios about a fictitious subordinate1. To control for the subordinate’s gender, we also
manipulated the names and pronouns used in the scenarios, such that the participants were
assigned randomly to read the behaviour of subordinates: Mike or Mary (see Appendix A). Next,
the participants responded to questions about the scenario and their demographics.
Manipulation check. We used six items with the highest factor loadings from the scale
by Duffy et al. (2002) to evaluate the manipulation. A sample item was, ‘your subordinate
insulted you’ (1 = strongly disagree, 5 = strongly agree; Pre: α = 0.90; Post: α = 0.93).
Measures.
Retaliation. We measured supervisor retaliation in the form of career-related penalties.
After reading the scenarios, the supervisors indicated their recommended subordinate’s bonus
(1= no bonus, 5 = full bonus) and job performance rating (1 = extremely poor, 5 = exceptional;
both items reverse-coded).
Felt leader identity threat. We measured leader identity threat by adapting Henderson
and O’Leary-Kelly’s (2012) seven-item self-identity threat scale. We asked, ‘considering the
scenario you have just read, to what extent do the statements below describe how you feel?’.
Sample items included ‘devalued as a leader’ and ‘respected as a leader’ (1 = hardly
describes my feelings, 5 = clearly describes my feelings; reverse-coded).
Although leader identity threat and supervisor-directed deviance are theoretically distinct,
our manipulation may have inherently influenced the participants’ felt leader identity threat, creating
the potential for spurious mediation. Thus, we conducted a series of confirmatory factor analyses for
each timepoint (CFA; Thompson, 2004) including leader identity threat and our manipulation check
measures, which showed that two reverse-worded items had to be removed from the leader
identity threat scale, and established discriminant validity2. Thus the five-item felt leader identity
threat scale was used in further analyses.
Control variables. We controlled for supervisor and subordinate gender (see Appendix
A), to account for different experiences of men and women during the pandemic (Gausman and
Langer, 2020) and potential biases leading towards discrimination against women in the
workplace (Heilman and Eagly, 2008). We also controlled for tenure in supervisory role to
proxy supervisor confidence, because greater tenure is expected to increase supervisor
confidence, diminishing felt leader identity threat.
Further, we controlled for the following variables that may have influenced the
supervisors’ stress and therefore, their affective responses to deviance. Because childcare may
be more demanding during work-from-home conditions, we included a dichotomous variable
where ‘1’ indicated having children. Recent research also suggests that remote work creates
digital exhaustion (Leonardi, 2021). Thus, we asked the participants to report their teleworking
experience before and after the pandemic (from 1 = not at all to 6 = between 80 and 100% of my
work time). We also accounted for the respondents’ location with a dichotomous variable
representing New York State, an early epicentre of the COVID-19 pandemic in the US3.
Moreover, we controlled for felt anger, the most pertinent emotional expression of revenge
(Bies and Tripp, 2005), which is also an established mediator in the wrongdoing– retaliation
relationship (Jackson et al., 2019). After reading the scenario, participants could select an emoji
depicting anger, we thus created a dichotomous variable where ‘1’ signified ‘anger’4. Additionally,
COVID-19 pandemic can be considered a macro-level affective context reflected in the
organisational climate (Spicer, 2020; see also Ashkanasy, Wilderom, and Peterson, 2010), which
may influence or offset revengeful behaviour (Restubog et al., 2015). Thus, we controlled for the
organisational climate of welfare before and after COVID-19, with the following item adapted
to the pandemic (see Patterson et al., 2005): ‘To what extent does your company care about its
employees’ welfare?’.
Finally, we controlled for negative reciprocity because of its potential to influence
retaliation (Mitchel and Ambrose, 2007). We used six items with the highest factor loadings
from a scale developed by Eisenberger et al. (2004), as used in previous studies (Bordia et al.,
2014). An example item was ‘if someone says something nasty to you, you should say
something nasty back’.
Results and discussion
ANOVA (analysis of variance) for both pre- and post-COVID-19 data showed that the
manipulation was successful. That is, the participants’ perception of supervisor-directed
deviance through social undermining was greater in the experimental than in the control group
(Pre: 3.95 vs. 1.96, F = 141.77, p < 0.001; post: 4.17 vs. 1.75, F = 321.36, p < 0 .001).
H1 predicted a positive relationship between supervisor-directed deviance and
retaliation. ANOVA for both pre- and post-COVID-19 data suggested that retaliation was
significantly higher in the experimental than in the control condition (Pre: 4.14 vs. 2.88, F =
46.27, p < 0.001; Post: 4.36 vs. 2.39, F = 134.34, p < 0.001)6; thus, supporting H1.
H2 and H3a predicted a first-stage moderated mediation model, whereby supervisor-directed
deviance (for experimental vs. control group) and the pandemic context (pre- and post-COVID-19)
would induce leader identity threat and in turn elicit retaliation. We first ran separate mediation
analyses for the pre- and post-COVID-19 data using a bootstrapping procedure on 5,000 samples to
construct 95% bias-corrected confidence intervals (CI; Hayes, 2018). The results (presented in Table
2) showed a positive and significant effect of supervisor-directed deviance on leader identity threat
(Pre: unstandardized Beta (B) = 0.61, standard error (SE) =
0.27, CI: [0.07; 1.14]; Post: B = 1.26, SE = 0.23, CI: [0.79; 1.72])7 and leader identity threat was
positively and significantly related to retaliation (Pre: B = 0.21, SE = 0.09, CI: [0.03; 0.41];
Post: B = 0.28, SE = 0.09, CI: [0.08; 0.46]). Further, although we did not find a significant
indirect effect of supervisor-directed deviance on retaliation via leader identity threat before
COVID-19 (Pre: B = 0.13, SE = 0.09, CI: [-0.04; 0.31]), the effect was positive and significant
after COVID-19 (Post: B = 0.35, SE = 0.13, CI: [0.08; 0.62]) and when the whole sample was
considered (B = 0.11, SE = 0.08, CI: [0.06; 0.38]).
--Insert Table 2 and Figure 1a about here--
Additionally, the indirect effect of the interaction between supervisor-directed deviance
and time was positive and significant (B = 0.03, SE = 0.01, CI: [0.01; 0.07])8. As demonstrated
in Figure 1a, the pandemic context strengthened the effect of supervisor-directed deviance on
leader identity threat, with supervisors feeling more threatened after COVID-19 than before,
thus supporting H2 and H3a.
In summary, we found that the subordinates’ supervisor-directed deviance elicits
retaliatory behaviours due to leader identity threat experienced by supervisors. This indirect
effect was particularly evident during the pandemic, providing support for our rationale that
supervisors’ felt threat as targets of deviant behaviour is contextualized. As our findings
further demonstrated, the pandemic context moderates this indirect relationship, with
supervisors being more threatened during the pandemic than under normal circumstances.
To replicate the findings and confirm the role of the pandemic, we conducted the following
study.
References
Aquino, K. and Douglas, S. (2003). 'Identity threat and antisocial behavior in organizations:
The moderating effects of individual differences, aggressive modeling, and hierarchical status',
Organizational Behavior and Human Decision Processes, 90, pp. 195-208.
Aquino, K., Tripp, T. M. and Bies, R. J. (2006). 'Getting even or moving on? Power,
procedural justice, and types of offense as predictors of revenge, forgiveness, reconciliation,
and avoidance in organizations', Journal of Applied Psychology, 91, pp. 653-668.
Ashforth, B. E. and Schinoff, B. S. (2016). 'Identity under construction: How individuals come
to define themselves in organizations', Annual Review of Organizational Psychology and
Organizational Behavior, 3, pp. 111-137.
Ashkanasy, N. M., Wilderom, C. P. M., and Peterson, M. F. (2010). The handbook of culture
and climate (2nd ed.), pp. 3-10. Thousand Oaks, California: Sage
Basch, J. and Fisher, C. D. (2000). 'Affective events–emotions matrix: A classification of work
events and associated emotions'. In J. Basch, C. E. Hartel and W. J. Zerbe (eds), Emotions in
the Workplace: Research, Theory, and Practice, pp. 36-48. Quorum Books/Greenwood.
Baumeister, R. F., Bratslavsky, E., Muraven, M., and Tice, D. M. (1998). ‘Ego depletion: Is the
active self a limited resource?’, Journal of Personality and Social Psychology, 74(5), p.1252.
Bies, R. J., and Tripp, T. M. (2005). ‘The study of revenge in the workplace: Conceptual,
ideological, and empirical issues.’ In S. Fox and P. E. Spector (eds.), Counterproductive Work
Behavior: Investigations of Actors and Targets, pp. 65-81. Washington, DC: APA Press.
Bordia, P., Kiazad, K., Restubog, S.L.D., Difonzo, N., Stenson, N. and Tang, R. (2014). ‘Rumor
as Revenge in the Workplace’, Group and Organization Management, 39, Pp. 363-388.
Bradfield, M and Aquino, K. (1999). ‘The effects of blame attributions and offender
likableness on forgiveness and revenge in the workplace’, Journal of Management, 25, pp.
607-631.
Camps, J., Stouten, J., Euwema, M. and De Cremer, D. (2020). 'Abusive supervision as a
response to follower hostility: a moderated mediation model', Journal of Business Ethics, 164,
pp. 495-514.
