Does the number of perpetrators matter? An extension and re-analysis of workplace bullying as a risk factor for exclusion from working life

Based on a nationally representative sample ( N = 1,613) and a true prospective design, we show that the link between self-labelled workplace bullying and exclusion from working life (i.e., becoming a non-participant in working life) over a 5-year time lag becomes stronger with increasing numbers of perpetrators involved. The amount of exposure to bullying behaviours could not explain the added effect of multiple perpetrators, indicating that the presence of multiple perpetrators is significant in and of itself. A post hoc descriptive cross-tabulation analysis also showed that when one or two perpetrators were involved at T1, the rate of exclusion from working life 5 years later did not differ substantially from the normal population's exclusion rate. When three or more perpetrators were involved, on the other hand, the exclusion rate approached 50%, indicating that ≥ 3 perpetrators could denote a critical cut-off point for a distinguishable group-bullying phenomenon, at least with respect to exclusion from working life.


| INTRODUCTION
Workplace bullying denotes a significant social work stressor, comprising exposure to ongoing negative and unwanted behaviour by superiors or colleagues, which the targeted individual has difficulty defending against due to a real or perceived power imbalance (cf. Einarsen, Hoel, Zapf, & Cooper, 2011). While bullying can be enacted by both individual perpetrators and by groups, the issue of bullying as carried out by a group has been largely neglected in the empirical literature (Tepper & Henle, 2011). This holds true in spite of the fact that scholars have recognized the existence of specific types of bullying and related phenomena involving multiple perpetrators. For instance, scapegoating involves attributing the blame for all frustration in the group to one individual, with the goal of driving him or her away (cf. Thylefors, 1987). Similarly, the black sheep-effect involves targeting and expelling norm deviants who obscure the identity of a group (cf. Marques & Paez, 1994), and the term predatory alliances denotes bullies conjoined as a 'wolf pack', targeting individual employees in cooperation (Hutchinson, Vickers, Jackson, & Wilkes, 2006). Additionally, the term mobbing has been used by English speaking researchers when referring to harassment and bullying carried out by a group of perpetrators that aim to force away undesired co-workers (Davenport, Schwartz, & Elliot, 1999;Sperry, 2009). In the present paper, we apply the rationale behind these phenomena to the empirical literature on workplace bullying, looking specifically at the effect of number of perpetrators on the targets' risk of becoming excluded from working life.

| WORKPLACE BULLYING AND RISK OF EXCLUSION
Among the detrimental outcomes of workplace bullying (see, e.g., Hoel, Sheehan, Cooper, & Einarsen, 2011;Hogh, Mikkelsen, & Hansen, 2011), scholars suggest that targets face an increased risk of exclusion from working life. This outcome must not be confused with ostracism or other exclusionary negative interpersonal treatment, but refers to exclusion from working life itself, that is, changing status to being non-participants in ordinary working life in the aftermath of exposure to workplace bullying. The assumption itself is relatively old, and was first explicitly presented by Leymann (1992). In his theorized process model, workplace bullying is construed as a gradually evolving phenomenon starting with low-intensity conflicts or incivility, and ending with the target's exclusion from the workplace or from working life itself, if victimization is allowed to escalate and management fails to adequately intervene (see Leymann, 1990Leymann, , 1992. However, thorough empirical support for the assumption of Leymann has only been presented more recently, showing for instance that workplace bullying can predict long-term sickness absence, working disability and unemployment (Glambek, Skogstad, & Einarsen, 2015;Hogh, Hoel, & Carneiro, 2011;Niedhammer, Chastang, Sultan-Taïeb, Vermeylen, & Parent-Thirion, 2013;Nielsen, Emberland, & Knardahl, 2017).
Additionally, scholars have begun to explore the explanatory mechanisms of the relationship between bullying and exclusion from working life, for example, suggesting that passive-avoidant leadership (i.e., laissez-faire leadership) facilitates exclusion of bullying targets (Glambek, Skogstad, & Einarsen, 2018). However, this line of research is yet scarce, and several potential mechanisms remain unexplored. In the present study, aiming to address this knowledge gap, we propose that the number of perpetrators involved in a given bullying scenario can be construed as one of the most potent explanatory factors in this regard.

| THE IMPACT OF NUMBER OF PERPETRATORS
Theoretically, there are several reasons to expect that the number of perpetrators involved in a given case affect the risk that targets may end up as non-participants in working life. For instance, when several individuals enact bullying, the power imbalance may be far more overwhelming, making it even harder to cope with or defend against the behaviour. As such, any associated strain is likely to yield more severe health consequences, reasonably affecting the targets' ability to maintain their position as employed over time. In line with this, with a group of bullies, it is also possible that social support is less available in the work environment, and that the threshold to intervene among possible bystanders becomes higher. Accordingly, such a situation may reflect not only a more severe threat to individual safety, but also to basic social needs, such as the need to belong (Baumeister & Leary, 1995). This, in turn, may possibly further increase any related negative health outcomes, in line with research demonstrating high distress levels among those socially excluded at work (Wu, Yim, Kwan, & Zhang, 2012).
In addition to these possibilities, in the case that several perpetrators collectively target one or a few individuals, the situation may itself be more likely to take the form of an expulsion process. Specifically, rejection and potentially eviction of an individual group member is a common theme in the group phenomena mentioned above, such as 'black sheep'-processes and scapegoating. These are established group-level processes with potentially aggressive and antisocial connotations, where certain individuals are attributed with unfavourable characteristics. This, in turn, can spur negative interpersonal treatment and eviction from the group, for example as the target is blamed with any frustration or misfortunes evident, or because they are perceived to obscure the group's social identity (e.g., Hogg, 2005;Marques & Paez, 1994;Thylefors, 1987). In support of this, one study found that negative gossip in the workplace appears centered on a few individuals ('scapegoats') (Ellwardt, Labianca, & Wittek, 2012). Evidence from the bullying literature also seem to support the general idea that bullies frequently conjoin in their effort to victimize certain individuals (Hutchinson et al., 2006). Additionally, consciously or unconsciously, groups of bullies are also believed particularly motivated to force their targets out of the workplace (Davenport et al., 1999;Duffy & Sperry, 2007).
Being expelled from the workplace does not necessarily imply that targets are forced out of working life itself.
However, targets of bullying who are forced to quit or who leaves more or less voluntarily, and particularly those targeted by several individuals, will likely exit their job at a time when their resources are depleted, and many will reasonably have a hard time finding their way back to or further on to ordinary employment (Leymann, 1992). Based on this notion, and the rationale reviewed further above, we investigate whether the number of perpetrators of workplace bullying predicts the risk of exclusion from working life among self-labelled bullying targets, hypothesized as follows: Hypothesis 1 The association between workplace bullying and subsequent exclusion from working life will be stronger with higher numbers of perpetrators. the Norwegian working force using the T1 measure (Nielsen et al., 2009) as well as prospective associations between workplace bullying, job insecurity and exclusion from working life (Berthelsen, Skogstad, Lau, & Einarsen, 2011;Glambek et al., 2015Glambek et al., , 2018Hauge, Skogstad, & Einarsen, 2007). The present study thus represents an extension and re-analysis of these data.
The Regional Committee for Medical Research Ethics in Norway approved the study.

| Measures
Exposure to workplace bullying was established using a standard self-labelling procedure (Einarsen & Skogstad, 1996;Nielsen, Notelaers, & Einarsen, 2011). Respondents indicated whether they had been exposed to workplace bullying during the past 6 months, based on a research definition, presented as follows: Bullying (e.g., harassment, torment, freeze-out or hurtful teasing) is a problem in some workplaces and for some employees. To be able to call something bullying, it has to occur repeatedly over a certain period of time, and the bullied person has difficulty in defending him-or herself. It is not bullying when two persons of approximately equal 'strength' are in conflict, or if it is an isolated situation.
Respondents subsequently indicated the number of perpetrators involved. A single continuous variable was constructed, starting with one perpetrator, and ending with ≥12 perpetrators. All respondents who reported ≥12 perpetrators were collapsed in to this latter value (coded as 12), in order to eliminate possible effects of statistical outliers.
Exclusion from working life was established by asking respondents to indicate their employment status at the follow-up measure 5 years after the baseline measure. The response alternatives 'out on sick leave', 'on leave with salary', 'on vocational rehabilitation', 'recipient of disability pension', 'unemployed', 'retired' and 'under full-time education' were collapsed into the category excluded (coded as 1), while 'full-time employed', 'part-time employed' and 'self-employed' were collapsed into the reference category employed (coded as 0).
Covariates included age and gender and exposure to bullying behaviours. The latter covariate was included in the final step, in order to assess whether any relationship between number of perpetrators and exclusion from working life could simply be ascribed to the amount of bullying behaviours enacted. Exposure to bullying behaviours was measured using the Negative Acts Questionnaire-Revised (Einarsen, Hoel, & Notelaers, 2009), a 22-item scale comprising exposure to typical bullying behaviours over the past 6 months (e.g., 'exposed to exaggerated teasing and joking'), yet without any reference to the terms bullying or harassment. Items were assessed using a 5-point Likert-type scale ranging from (1) never to (5) daily. Acceptable internal stability was obtained for this measure (α = .90). Additionally, as the normal retirement age in Norway is 67 years we excluded respondents aged 62 or older at T1 to account for retirement under ordinary circumstances at follow-up 5 years on, and included only respondents working a minimum of 15 hr per week at T1.

| Analyses and results
We used logistic regression analyses to test the study hypothesis based on a true prospective design. Data were analysed using IBM SPSS statistics 25.
Descriptive statistics were retrieved for those aged 61 or younger at T1, and showed that 4.2% (N = 101) selflabelled as targets of workplace bullying. In addition, 84.2% (N = 1,253) of the overall sample reported being employed at follow-up, while 15.8% (N = 235) reported an employment status indicative of exclusion from working life.
In support of Hypothesis 1, the results showed that as the number of perpetrators became higher, so did the probability of exclusion from working life 5 years on (OR = 1.25, p = .024), controlling for age and gender. Moreover, the full regression model explained between 11.8% (Cox and Snell R 2 ) and 17.9% (Nagelkerke R 2 ) of the variance in exclusion, correctly classifying 75.4% of the cases. The test for model fit yielded some support for the full model with a borderline significant chi-square (χ 2 = 7.15, df = 3, p = .067) and a non-significant Hosmer and Lemeshow test (χ 2 = 5.89, df = 8, p = .659). Furthermore, upon adding the measure of overall exposure to bullying behaviours to the control measures in a second step, the effect remained (OR = 1.3, p = .018). The results of the logistic regression analyses are displayed in further detail in Table 1.
Finally, we carried out a post hoc descriptive analysis in order to inspect how the exclusion rate was distributed with respect to number of perpetrators. As shown in Table 2, among the 31 targets reporting bullying carried out by a single perpetrator, four were excluded from working life at T2, while bullying by two perpetrators led to exclusion in one out of 13 targets. Together, this amounts to an exclusion rate of approximately 12%, which does not differ substantially from that of the normal population (approximately 16%). However, as further shown in Table 2, risk of exclusion appear to severely increase when targets report bullying by more than two perpetrators at T1. For example, in the case of three perpetrators, three in five targets report exclusion, in the case of four perpetrators, the figure is one in two and in the case of five perpetrators, two in five. In fact, exclusion from working life is reported in almost 50% of the cases where more than two perpetrators are involved, possibly indicating that ≥3 perpetrators may denote an important cut-off for a group bullying phenomenon with significant implications for survival in working life.

| DISCUSSION
The results of the present study show that the previously established link between workplace bullying and exclusion form working life is affected by the number of perpetrators involved. Specifically, the odds of becoming a nonparticipant in working life before reaching normal retirement age over a 5-year time lag appears to escalate when workplace bullying is enacted by several perpetrators. Importantly, this effect remains when controlling for the overall exposure to bullying behaviours. In addition, as shown in a post hoc analysis, the rate of targets excluded from T A B L E 1 Logistic regression analysis with number of workplace bullying perpetrators as a predictor of exclusion from working life over a 5-year time lag B S E Wald OR 95% CI Step Out on sick-leave n = 2, on disability benefits n = 1, early retirement n = 1. b Early retirement n = 1. c On disability benefits n = 2, under full time education n = 1. d On disability benefits n = 1. e On vocational rehabilitation n = 1, unemployed n = 1. f Unemployed n = 1. g Out on sick-leave n = 1. h On disability benefits n = 1, early retirement n = 1.
working life seems to increase radically when at least three perpetrators are involved, while the difference between one and two perpetrators is less pronounced.
There are several potential explanations for the present results. First, a high number of perpetrators may add to the risk of exclusion from working life via health impairment. The perception of being targeted by a group may for instance represent a particular threat to the social needs of the target, obscuring the fundamental need for belongingness and social inclusion (Baumeister & Leary, 1995). In fact, being bullied by a group is possibly strongly associated with ostracism and social exclusion, phenomena known to yield significant negative health consequences, with short term reactions involving pain and distress (Eisenberger & Lieberman, 2004;Williams, 2007), and long term outcomes involving resignation and withdrawal (Spoor & Williams, 2007;Williams & Nida, 2011). Possibly, such outcomes are not as prominent in the case of bullying by one or two perpetrators, but may leave targets of group bullying in poorer health and with a more reduced work capacity. Related to this, it is also likely that bullying carried out by a group is characteristic of a situation entailing less social support from other organizational members, leaving the target more vulnerable. Furthermore, the presence of several perpetrators may affect the power imbalance present in the situation, with targets finding it more difficult to defend against the mistreatment or otherwise resolve the situation. Specifically, this may increase the probability of organizational exit at a time when the targets' resources are severely reduced, thereby further decreasing the probability that they will find and maintain new employment (Leymann, 1992). Related to this, similar and overlapping group-level phenomena such as scapegoating, black sheep-processes and mobbing are often explained by social groups' ability to expel certain members (Hogg, 2005;Hogg, Fielding, & Darley, 2005;Marques & Paez, 1994;Sperry, 2009). Thus, a conscious or unconscious motivation or intention to force the target away may be characteristic of group bullying as a phenomenon per se, increasing the likelihood that targets will exit the organization, and subsequently risk exclusion from working life itself.
The present study's emphasis on the number of perpetrators involved in bullying represents a rare initiative in the bullying and harassment literature (Tepper & Henle, 2011). However, as outlined here, there are strong theoretical arguments for such a perspective, with respect to both antecedents and outcomes of bullying. Future research should therefore address group bullying from both of these angles. Additionally, we recommend empirical investigations of bystander roles in this regard. Vignette studies have shown that fear of so-called stigma by association is conditioned by whether other employees partake in the bullying situation, which increases the likelihood that bystanders will also enact antisocial behaviour towards the target (Mulder, Pouwelse, Lodewijkx, Bos, & van Dam, 2016). Hence, a closer look at bystanders of group bullying versus bullying enacted by individuals may yield important insights with high practical and theoretical value.
Limitations of the present study include the use of secondary analyses and our exclusive use of self-report data, which may entail a danger of common method variance (Podsakoff & Organ, 1986). With respect to the latter limitation, however, the use of a long time lag with 5 years separating the measurement points will offer significant protection against related biases. In addition, the measure of exclusion from working life used in this study refers to an objective employment status with little need for emotional or cognitive appraisal and judgement. Together with the use of a true prospective design with new cases of exclusion from working life as an outcome at follow-up, this likely further prevents any problematic aspects of common method variance in this study (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). Finally, even though the number of bullying targets is sufficient for the present analyses and is likely to reflect a representative incidence rate, the relatively low number of bullying targets may also represent a limitation.
In conclusion, we note that bullying carried out by several perpetrators appears to represent a distinctive and significant form of workplace mistreatment, particularly as this number reaches at least three individuals. With respect to the detrimental outcome of exclusion from working life, this form of group bullying appears to represent an extreme risk factor compared with bullying carried out by one or two perpetrators. As an outcome of bullying, exclusion from working life is severe, entailing significant detriment to the individual target, to next-of-kin, to organizations and to society. Our finding should therefore be highly relevant for organizations, organizational practitioners and clinicians working with current and previous targets of workplace bullying, and denote a starting point for future research on group bullying as a phenomenon as it occurs in the context of working life.