In this study, we chose to differentiate between perceived support from two different sources within the workplace, that is, close co-workers and immediate supervisors. We were not investigating social support or social climate in general, but the potential effects of perceived support from two clearly differentiated sources. There may still be other sources of support in the organization for example HR personnel or supportive systems in the organization, such as guidelines and procedures that an individual can consult, but they were not included in this study.
The finding that exposure to bullying behaviours is negatively associated with health and well-being replicates what many other studies have showed (for example Einarsen and Nielsen 2015; Theorell et al. 2015; Hoel et al. 2004; Trépanier et al. 2013), but what is more interesting is the interaction effects found.
Perceived support from close co-workers moderated the negative association between exposure to bullying behaviours, and health and well-being. This interaction was negative indicating that perceived support from close co-workers reduced, but did not eliminate, the negative effects of bullying behaviours for the overall health and well-being. The negative interaction also implies that the reduction in negative consequences for health and well-being are likely to eventually disappear at higher levels of bullying behaviours. It is, however, clear that one is less likely to suffer negative effects on health and well-being as a result of workplace bullying if exposure to bullying behaviours arises when one perceives that one has a high level of support from close co-workers.
It is interesting to compare our result with the study by Rousseau et al. (2014). They also found a moderating effect of support from co-workers when investigating how trust in management (as an organizational resource) and role overload (as an organizational demand) affected workplace bullying. They showed that the effects on workplace bullying from both were dependent on factors such as autonomy, employee participation, and support from co-workers. In their study, trust in management had a lowering simple effect on workplace bullying. In cases of low trust, support from co-workers had a buffering (moderating) effect. Even if we did not use the same design, and did not investigate the exact same factors, there are probably important common findings here. They investigated trust in management whereas we investigated perceived supportive leadership, which include dimensions of trust and feelings of security in relation to one’s immediate supervisor. The two factors are similar, however, Rousseau’s et al. (2014) concept is on a general organisational level (general trust) whereas our concept is on an individual level. Another difference is that they investigate the effects on workplace bullying, whereas we investigate the potential moderating effect on the negative association between workplace bullying, and health and well-being. But even if there are clear difference, what is common is the findings that different sources of support and trust in the organisation interact. So, when investigating this phenomenon, one must be careful as the different sources interact and influence each other in important ways.
Surprisingly, in our full model (Model 5) there was no interaction between perceived supportive leadership and exposure to bullying behaviours with respect to health and well-being. This result is the opposite of what one might expect, based on studies which have reported that supportive leadership or trust in management lowers the risks for workplace bullying (see for example Gardner et al. 2013; Rousseau et al. 2014; Van den Brande et al. 2016). However, those studies indicated that supportive leadership reduces the incidence of bullying, whereas our study focused on the effects of workplace bullying on health and well-being. This is more similar to a study by Clausen et al. (2019) who showed a buffering effect of supportive leadership on the risk of workplace bullying leading to disability pensioning. In the current study the Perceived supportive leadership (PSL) and the Roles in the organization (RIM) were highly correlated (r = 0.55) and as we used RIM as a covariate we controlled for the effect of conflicting and ambiguous roles in the organization. The lack of a protective effect of supportive leadership in our analysis was at least in part due to the inclusion of RIM as a covariate. This implies that the effect of perceived supportive leadership is linked to how clear the roles in an organization are. In other words, the presence of a perceived supportive leadership may have indirect protective effects mediated by its impact on the organization, rather than having a direct effect at an individual level.
It is also interesting that when the Perceived supportive leadership (PSL) was tested as a single moderator like Clausen et al. (2019) (in Model 3) it actually had a significant moderating effect, but when the Perceived support from close co-workers (PSC) was included (in the Model 4 and 5) the significant moderating effect of PSL disappeared. That, once again, clearly underscores the importance of not investigating support in general and that support from different sources can have different effects, and that there may be significant interactions between these sources of support as we see in our three-way interaction model.
A possible explanation for our finding that there was an interaction between perceived support from close co-workers and exposure to bullying behaviours with respect to health, but not for perceived supportive leadership, can be found in social exchange theory (see for example Coyle-Shapiro and Shore 2007; Cropanzano and Mitchell 2005). Parzefall and Salin (2010) noted that there is a growing body of evidence suggesting that co-workers have an important influence on employees’ perceptions of social relationships. Korte (2009) pointed to the importance of frequency and Wanous (1992) to the quality of social interactions. Svensson (2010) also highlighted the importance of the distance between interaction partners and the regularity of their interactions. One would expect people to interact more frequently with close co-workers than with their immediate supervisor, hence one might expect perceived social support from close co-workers—which may be categorized as a high-quality and high-frequency interaction—to have a stronger protective effect on health and well-being.
Svensson (2010) also showed that proximity and regularity of interaction are important preconditions for bullying behaviours, suggesting that it is important for victims to get out of the way and find a safe place to which they can retreat. One could argue that being exposed to bullying behaviours by people that are physically or socially close, that is, people that one has to interact with on a regular basis, is more damaging than being bullied by people one is more distant to. This implies that if one is bullied by one or more close co-workers it may be very difficult to get the social support that would otherwise have reduced the associated health risks.
Whilst we did not find an interaction between exposure to bullying behaviours and perceived supportive leadership with respect to health and well-being, we did find a three-way interaction between exposure to bullying behaviours, perceived support from close co-workers and perceived supportive leadership. The interesting point is that the moderation of the negative association between exposure to bullying behaviours and health by perceived support from close co-workers is conditional on perceived supportive leadership. By distinguishing between perceived support from co-workers and perceived supportive leadership (see Zapf et al. 1996) we have been able to show that the effects of one are contingent on the other. We also found that when perceived supportive leadership is low the interaction effect disappears. In other words, the health risks associated with being exposed to bullying behaviours in the workplace are reduced if one perceives a moderate or a highly supportive leadership together with support from close co-workers. Conversely, on the lowest 12.6% of the range of perceived supportive leadership, there is no interaction. So, our moderated moderation analysis indicated that the effect of perceived support from close co-workers depends on the level of perceived supportive leadership and, in addition, when the level of the perceived supportive leadership is low perceived support from close co-workers does not moderate the health risks associated with exposure to workplace bullying. Perceiving support from close co-workers will only reduce the health risks associated with exposure to bullying behaviours if one trusts one’s supervisor and feels safe in that relationship.
The finding that lack of trust or security in one’s relationship with one’s supervisor may block the beneficial effects of perceived co-worker support may reflect situation in which the supervisor not only fails to provide supportive leadership but actually acts as a bully (Zapf et al. 2011) which is also discussed by Clausen et al. (2019). It is reasonable to believe that scores in the bottom 12.6% of the range on a scale measuring perceived supportive leadership will include cases in which the respondent is bullied by his or her workplace supervisor. But supervisors are not responsible for all workplace bullying (Zapf et al. 2011). A low score for supportive leadership may reflect circumstances and factors other than a bullying leader. For example, passive and absent leadership, also called laissez-faire leadership, has been shown to be strongly associated with workplace bullying through its effects on role ambiguity, interpersonal conflicts, etc. (Skogstad et al. 2007a).
This highlights the question of who the bully is if one is being exposed to bullying behaviours yet receiving support from both close co-workers and one’s supervisor. In most workplaces, however, there are many organizational levels and many workgroups or social groups, so it is entirely possible that one could be receiving support from close co-workers and one’s immediate supervisor, yet still being exposed to bullying behaviours in the organization (Zapf et al. 2011).
This also opens for questions about the direct and moderated effects on health and well-being of workplace bullying from different sources. It is reasonable that the negative effect of exposure to bullying behaviours may be different depending on if the source is some or all of your co-workers, your supervisor, a client, or a combination of several sources. For example, Törek et al. (2016) showed that employees exposed to workplace bullying from leaders experienced more severe depressive symptoms compared to those that were bullied by co-workers or clients. This result somehow contradicts the reasoning connected to social exchange theory (above) where, for example, proximity (Svensson 2010) and the quality of relationships (Wanous 1992) were of particular importance. One may, however, reason that being bullied is one thing and being protected from its negative effects is another. If perceived support from co-workers is a buffer and protection for the effects on health and well-being from exposure to bullying behaviours, our data suggest that this buffering loses its effect when the perceived supportive leadership is very low. This may suggest that if you are bullied by your supervisor, or that your supervisor knows that you are bullied but doesn’t care, it really doesn’t matter how much of support you get from your co-workers, your health and well-being will suffer anyhow.
It would be interesting to investigate a combination of different settings. For example, bullying from co-workers moderated by perceived support from a supervisor, bullying from some co-workers moderated by perceived support from other co-workers, bullying from a supervisor moderated by perceived support from co-workers etc. Such a refined analysis would, however, demand a longitudinal setting, a very large data sample and very specific questions about who the bullies are. Perhaps it would be better to perform deep interviews and qualitative analysis to further investigate and clarify different aspects of these questions.
Nevertheless, our results call for more research into the questions of how different sources of perceived support and different kinds of support (emotional, instrumental, informal, appraisal; see Foster 2012) may moderate the negative effects on health and well-being of workplace bullying from different sources. However, our study suggests that there may be important differences between perceived supportive leadership and perceived co-worker support, and their protective effects on health and well-being when one is exposed to bullying behaviours.
Limitations
This study was based on data from a self-report questionnaire. Relying on a single data collection method may lead to common method bias and threaten construct validity (Donaldson and Grant-Vallone 2002) even if such a problem seems to be rarer than has been assumed (see for example Spector 2006). The tendency of an employee to bias his or her response may be evaluated based on four factors (a) the true state of affairs, (b) the sensitivity of constructs, (c) dispositional characteristics, and (d) situational characteristics (Donaldson and Grant-Vallone 2002). On this basis, we conclude that there is little risk that our participants’ responses were biased. For example, participants were not asked to self-report any socially undesirable behaviours among themselves and the constructs were not in itself sensitive for the respondents as we did not use self-definition of being exposed to bullying but the Negative acts questionnaire-revised (NAQ-R)—see Nielsen et al. (2011) for a discussion of the advantages and disadvantages of different workplace bullying measurement methods. Furthermore, the data were collected in the context of a regular work environment survey to which the participants were used to submitting information. This also indicates that data were collected in a context where there was little situational pressure to give socially desirable answers. Also, testing a common latent variable showed only 1.7% common variance among the variables in the study (Podsakoff et al. 2003).
Another limitation is that all our data are cross-sectional which means the directions of the associations between the investigated factors are unknown. Our results were, however, consistent with the theoretical reasoning and other studies behind our hypotheses. Nevertheless, it is possible that, for example, being in poor health could lead to an employee being exposed to behaviours and acts such as being assigned uninteresting and uninspiring work tasks or redeployed because his or her performance has deteriorated and that he or she perceives these events as amounting to workplace bullying. Longitudinal research in which data on bullying and health have been collected on several occasions has concluded that the direction of influence is mainly from workplace bullying to health (see for example Einarsen and Nielsen 2015) although there are circumstances under which the associations may be bi-directional (Einarsen and Nielsen 2015). For example, a chain of events may start with workplace bullying having a negative influence on health, after which the influence flows in both directions because worsening health increases exposure to negative workplace behaviours. We nevertheless conclude from this study that it is more plausible that the negative influence flows mainly from bullying to health, rather than vice versa.
It has been argued that in a moderation analysis the candidate moderators and the predictor variable should be uncorrelated (see Hayes 2018). Clearly, there may be problems with multicollinearity and a high variance inflation factor when these correlations are high. Hayes (2018) argued that it is always a good idea to do what you can to reduce the correlation between the predictor variable and the moderators, but also stated that the non-correlation criterion should be treated as an ideal rather than a requirement. In our data, the correlations between the predictor and the moderators were between 0.38 and 0.45 which might be regarded as less than ideal, but cannot be regarded as high, given that it leaves about 80–85% of variance unaccounted for.
There is only small, although significant increase in R2 when adding the interactions when all variables are included in the model. This could of course be due to a rather large sample size. In the model, there are many variables that all, but the interaction between negative acts and perceived supportive leadership, significantly contribute to the explained variance. The final model is the model that explains most variance of the tested models.
A final limitation on our findings relates to their representativeness because our data are cross-sectional and were collected from a single cohort of workers in a governmental department in Sweden. Our results need to be replicated in other employment sectors and other countries.