Levels of workplace bullying and escalation – a new conceptual model based on cut-off scores, frequency and self-labelled victimization

ABSTRACT

The aim of this study was to identity levels of workplace bullying based on cut-off scores, frequency and self-labelled victimization, and to use these levels to identify the escalation of workplace bullying in terms of onset of different negative acts. Data were collected from a representative sample of the Swedish workforce (n = 1856). Bullying was measured using the Negative Acts Questionnaire–Revised together with self-labelled victimization based on a definition. Other variables included aspects of work experience, perception of the organization, and health and well-being. The results showed differences between the suggested levels of workplace bullying (Risk for bullying and Incipient bullying; Ongoing bullying; and Severe and Extreme bullying) for these variables, and also that different areas were pronounced at different levels of workplace bullying. Further, the onset of different negative acts depended on the level of workplace bullying, at early stages only work-related negative acts, but at higher levels more person-related negative acts. The study contributes to the understanding of workplace bullying and the escalation process. The suggested new levels of bullying also have practical and pedagogical value making it easier to grasp and to convey to, e.g. HR personnel, and organizational psychologists.

KEYWORDS:

• Workplace bullying
• escalation
• organizational conditions
• negative exposure
• work experience

Introduction

Workplace bullying is a rather common problem in many organizations all over the world. There are many studies on workplace bullying from the Scandinavian countries, the rest of Europe, North America, Australia, and Asia showing it to be a serious risk factor for the exposed individual, for example, mental health problems, reduced job satisfaction, burnout, long sickness absence, unemployment, and suicide ideation (Dollard, Dormann, Tuckey, & Escartín, 2017; Glambek, Skogstad, & Einarsen, 2015; Nielsen & Einarsen, 2012; Nielsen, Magerøy, Gjerstad, & Einarsen, 2014; Nielsen, Nielsen, Notelaers, & Einarsen, 2015; Slany et al., 2014). It can also involve problems for the organization at large in terms of, for example, role ambiguity, disarray, a reduction in efficiency and productivity, and high turnover rate (Einarsen, Hoel, Zapf, & Cooper, 2011; Mathisen, Einarsen, & Mykletun, 2008; Salin & Notelaers, 2017). It is important to be able to both identify individuals at risk of being bullied and of course those who are bullied, but also to identify organizational conditions that could lead to bullying. Currently there is no clear, uniform or agreed upon way of measuring workplace bullying (Zapf, Escartín, Einarsen, Hoel, & Vartia, 2011). The different measures listed by Zapf and colleagues typically only include one type of measure, for example, self-labelling (with or without a definition) or the frequency of negative acts (e.g. at least weekly exposure). Using cut-off scores is another example of how to separate victims from non-victims (Notelaers & Einarsen, 2013). In this study the main focus is on identifying workplace bullying based on a combination of measures and to use this to show how bullying behaviours can increase and change in focus depending on the level of bullying.

There are different definitions of workplace bullying with slightly different focus, e.g. whether the perpetrator has an intent to bully. In a European tradition the definition posed by Einarsen et al. (2011) is probably the most widespread:

“Bullying at work means harassing, offending, or socially excluding someone or negatively affecting someone’s work. In order for the label bullying (or mobbing) to be applied to a particular activity, interaction, or process, the bullying behavior has to occur repeatedly and regularly (e.g. weekly) and over a period of time (e.g. about six months). Bullying is an escalating process in the course of which the person confronted ends up in an inferior position and becomes the target of systematic negative social acts. A conflict cannot be called bullying if the incident is an isolated event or if two parties of approximately equal strength are in conflict.” (p. 22)

The definition includes many different kinds of possible negative acts and bullying behaviours. It also highlights the importance of a systematic and frequent exposure together with a situation in which the target no longer can defend him- or herself. In this definition both the term bullying and mobbing is mentioned. Based on the definition it is not clear if the two terms are used interchangeably, or if they refer to separate, but related aspects of the phenomenon. In some cases or contexts (see for example Zapf, 1999), bullying and mobbing are used to describe two different ways in which a person can be exposed, where bullying refers to the actions of one person, the bully, against another, while mobbing refers to many people (the mob) gathering up against one person. Leymann (1990, 1996)) used the term mobbing, but in Swedish there is only one term for bullying/mobbing: “mobbning”. The Swedish word “mobbning” refers to bullying behaviours of one or more, or initially from one, where others join in. Leymann (1990, 1996)) argued for using “mobbing” rather than “bullying” for all cases – no matter if there was only one bully or many in a mob. In this article we use the term bullying as something that can involve one or many as the bully or bullies with a focus on those exposed to bullying behaviours.

Criteria for workplace bullying

We began the article by saying that bullying is a common problem. The truth is that we really do not know its full extent. This is mainly because there are a multitude of ways of measuring and categorising bullying, which makes it almost impossible to find a common ground when comparing different results showing prevalence. Bullying has mainly been measured using self-labelling with or without a definition – the self-labelling method, or exposure to a number of negative acts – the behavioural experience method (Einarsen et al., 2011). There are a few different inventories that measure negative acts, e.g. the Leymann Inventory of Psychological Terror (LIPT, Leymann, 1996) and the Negative Acts Questionnaire–Revised (NAQ-R, Einarsen, Hoel, & Notelaers, 2009). The scales used for the self-labelling method have often been either a simple yes or no, or some kind of frequency scale (e.g. never, now and then, monthly, weekly and daily). The behavioural experience method often involves a frequency scale. Often the duration of the bullying or bullying behaviours is also included (often 6 months or 12 months), but sometimes time is not mentioned. Einarsen et al. (2011) presented studies showing the frequency of workplace bullying, but very few studies used the same combination of the above-mentioned alternatives.

There is also no agreement as to which criterion or criteria should be used for assessing workplace bullying. Examples of criteria include weekly exposure, different cut-off scores, or self-labelling. Leymann (1990, 1996)) suggested that it takes at least a weekly exposure of bullying behaviours to call it bullying.¹ This method is often referred to as the operational classification method. Mikkelsen and Einarsen (2001) argued that the Leymann criterion resulted in too large a discrepancy between the number who self-labelled themselves as bullied and those categorized based on the criterion. They suggested exposure to at least two negative acts at least weekly. A third alternative based on the Leymann criterion was put forth by Agervold (2007) who argued that the Leymann criterion can lead to an overestimation of the prevalence of bullying and concluded that at least three negative acts a week would be a better criterion.

Notelaers, Einarsen, De Witte, and Vermunt (2006) took a different approach and used latent clusters with the Negative Acts Questionnaire and identified six clusters from those not bullied to victims. They framed the sometimes bullied, the work-related bullied and the victims as targets, adding up to about 20%. The latent cluster approach has also been used elsewhere (e.g. Leon-Perez, Notelaers, Arenas, Munduate, & Medina, 2014). Another approach by Notelaers and Einarsen (2013) focussed on receiver operation characteristic (ROC analysis) to find cut-offs for NAQ-R scores. Their results showed a lower threshold at 33 or higher which they called “occasional bullying” and a higher threshold at 45 or higher which they called “victims of bullying”.

Bullying as an escalating process

Bullying has been described as an escalating process (Einarsen et al., 2011; Zapf & Gross, 2001), but also as an end state (Leymann, 1990). The question is, what is bullying or maybe when do we call something bullying? These questions are not new; they have been discussed previously (e.g. Einarsen, 1999; Hoel & Beale, 2006; Leymann, 1990, 1996; Zapf, 1999), but they are still relevant as there seem to be disagreement as to whether bullying is the process leading to becoming a victim and onwards, or if it is just something that occurs when becoming a victim? The process leading to becoming bullied could be described as an escalating process of exposure to negative acts. Leymann (1990) seems to have advocated that bullying occurs when one becomes a victim and not the process leading up to it. He described bullying as hardly ever having degrees: one is either a victim of bullying or not. At the other end “bullying seems to exist on a continuum from occasional exposure to negative behaviours to severe victimization resulting from frequent and long-lasting exposure to negative behaviours at work” (Einarsen et al., 2011, p. 12). These contrasting viewpoints may create some confusion, because they represent not only very different ways of assessing bullying, but also distinct ways of framing what bullying is. The difference is probably not related to the essence of the phenomenon, but to different views regarding what is needed to call something bullying. The problem and the solution probably lie in the definition, so once again: what is bullying? The reason for asking this question is also to better be able to distinguish bullying from other unfair treatment or more isolated examples of exposure to negative acts, and also to not risk diluting the concept of workplace bullying. In this context the use of the NAQ-R (Einarsen et al., 2009) and its cut-off scores (Notelaers & Einarsen, 2013) may contribute to this by calling the lower threshold “occasional bullying”. It is almost a contradiction in terms; according to the definition, for it to be bullying the negative treatment must be frequent, at least weekly – not merely occasional. However, the exposure to negative acts, no matter the frequency, probably means a heightened risk for the exposed. Einarsen et al. (2011) suggested that the exposure to negative behaviours, without clarifying the level as long as it is systematic and occurs over time, should be viewed as bullying, and that it should be framed as severe bullying if the duration is six or more months. The process of exposure to negative acts, ultimately leading to an increased risk of being bullied, is not always linear (Kwan, Tuckey, & Dollard, 2016; Zapf & Gross, 2001). Zapf and Gross used a conflict escalation model to indicate at what point bullying is in “its final stage” (p. 502). Their study showed four different courses of conflict escalation trajectories in which bullying escalated over time. The study by Kwan et al. (2016) identified five courses of escalation and a difference depending on the psychological safety climate in the organization.

One way of dealing with the somewhat different ways of describing workplace bullying is, on the one hand, to focus on negative acts that can occur in isolation or more systematically, and on the other, to ask what it entails to become a victim of bullying. Based on the levels of conflict (Glasl, 1982) and the notion of conflict escalation as a pathway to workplace bullying (e.g. Baillien et al., 2015; Leon-Perez, Medina, Arenas, & Munduate, 2015) the process could be described as an escalation of negative acts. At some level the risk of this escalating process leading to bullying increases, and at some point, the systematics of exposure of negative acts increases to a level where the ability to defend oneself drops to an extent that it is relevant to talk about becoming a victim of bullying. It is highly relevant to identify exposure to negative acts at an early stage, long before one ends up a victim of bullying, in order to facilitate de-escalation. If the process has lasted too long, the probability of reversing it quickly approaches zero. This is one of the reasons for conducting the current study: to more clearly describe consequences of negative exposure at different levels of workplace bullying, and to show when different negative acts are more likely to appear in that process. To use more than one type of measure, to provide more information on the aspects of workplace bullying covered in the definition (Einarsen et al., 2011), was also an important starting point for this study.

Aim

The aim of this study was to identity levels of workplace bullying based on cut-off scores, frequency and self-labelled victimization. Further, to use these levels to identify how workplace bullying can change in focus depending on the level of bullying in terms of onset of different negative acts.

Methods

Participants

The study was based on a representative sample of the Swedish workforce. Data were collected in the autumn of 2017 using a simple random selection from the Swedish workforce (ages 18–65 and working at workplaces with at least 10 employees). The overall purpose of the survey was to investigate how organizational factors affect employees’ experience of work, and how this relates to workplace bullying, health and well-being. The government agency Statistics Sweden (https://www.scb.se/en) handled the selection process and distribution of the questionnaire. A total of 1856 questionnaires were completed. The mean age was 48.2 years (SD = 10.9), ranging from 21 to 67 years. There were 57% women in the sample. A majority had some university or college education (58%); one third (34%) had ten to twelve years of education while the rest (7.6%) had nine years or less. About half were married (54%) and 52% had at least one child. The majority were born in Sweden (88%). Of those not born in Sweden, 20% came from the other Nordic countries, 43% from the rest of Europe, 20% came from Asia, and the remaining 16% came from Africa, North and South America, and Oceania. Finally, 14% worked in some form of leading position, and 95% had a fixed contract.

Measures

In the study we used seven subscales from the Psychosocial Work Environment Questionnaire (PSYWEQ, Rosander & Blomberg, 2018). We measured salutogenic health using the Salutogenic Health Indicator Scale (SHIS, Bringsén, Andersson, & Ejlertsson, 2009), and anxiety and depression using the Hospital Anxiety and Depression Scale (HAD, Zigmond & Snaith, 1983). Workplace bullying was measured using the Negative Acts Questionnaire–Revised (NAQ-R, Einarsen et al., 2009) as well as self-labelling based on a definition of bullying, which was “Bullying occurs when a person, repeatedly and over time, is subjected to negative treatment from one or more persons, in situations where the victim has difficulty defending him- or herself. It’s not bullying if two equally strong persons are in conflict with each other”. The answers were on a frequency scale (never, now and then, monthly, weekly, and daily). All measures were selected based on previous validation in a Swedish context.

Subscales from the PSYWEQ (Rosander & Blomberg, 2018) included Individual work experience (IWE), Organizational conditions for work (OCW), Roles in the organization (RIM), Quality & efficiency (Q&E), Joint job satisfaction (JJS), Negative exposure at work (NEW), and Stress exposure (WOSEQ). All subscales were based on items using a seven-point Likert scale.

Individual work experience (IWE)

This is a combined variable of six subscales from PSYWEQ (I1-I6, Rosander & Blomberg, 2018), which in total comprises 28 items, all on an individual level. They capture different aspects of individual perception of one’s work, including perceived supportive leadership and support from close co-workers, job satisfaction, energy level, perceived influence and conditions for work (α = .96). Previous research has shown that being exposed to workplace bullying has consequences for one’s experience of work in terms of commitment, job satisfaction, and absenteeism (Nielsen & Einarsen, 2012).

Organizational conditions for work (OCW)

This is also a combined variable of six other subscales from PSYWEQ (G1-G6, Rosander & Blomberg, 2018). In total it comprises 34 items, which are all expressed from a group perspective: how “we” do things and how “our” business works. It captures the level of active and constructive leadership, cooperation and climate among the closest co-workers as well as the whole workplace, aspects of a well-functioning organization (e.g. roles, goals and systematic approach in how to work with environment issues), the social work environment, and staffing (α = .96). The “work environment hypothesis” (e.g. Leymann, 1996; Salin & Hoel, 2011) points to organizational conditions as the main cause of workplace bullying, that is, anybody can be exposed to bullying behaviours depending on the organizational conditions. The rationale is that in a poor work environment the constraints and norms that creates order and predictability are less visible and, in many ways, allow for incivility and a negative treatment of co-workers. The work demands also play an important role for the occurrence of workplace bullying (Lewis, Megicks, & Jones, 2017).

Roles in the organization (RIM)

Six items measuring aspects of role clarity in the organization, including a clear division of tasks, clear roles and role expectations as well as an orderly organization with well-functioning routines (α = .89). Unclear roles and expectations in an organization increase the risk of workplace bullying (Lewis et al., 2017; Salin & Hoel, 2011).

Quality & efficiency (Q&E)

Two items focussing on the efficiency of the organization, where time and resources are used in the best way, and that the work carried out leads to high quality products or services (α = .76). The conditions that foster workplace bullying, such as unclear roles and a passive leadership (Einarsen et al., 2011), are also signs of a less effective organization. To our knowledge no clear direct effect of workplace bullying on organizational effectiveness has been shown, however, many have shown the indirect effect described above.

Joint job satisfaction (JJS)

Two items that capture a more general experience of job satisfaction: that all who work together enjoy and have fun together at work (α = .89). As workplace bullying often involves social exclusion and ostracism (Einarsen et al., 2011) the victim’s perception of the social work climate and job satisfaction will suffer (Arenas et al., 2015).

Negative exposure at work (NEW)

Seven items representing different negative exposures at work, including sexual harassment, discrimination, threats and violence from within the organization, but also the perception of unequal treatment and feelings of being unsafe at work (α = .76). The inclusion of this factor in the study was to add a broadened concept of mistreatment as workplace bullying often is regarded as distinct from, for example, sexual and ethnical discrimination (Einarsen et al., 2011), but at the same time workplace bullying and other forms of harassment often go hand in hand (Lim & Cortina, 2005).

Stress exposure (WOSEQ)

Exposure to stress was measured using an instrument called Workplace Stress Exposure Questionnaire (WOSEQ, Rosander & Blomberg, 2018), containing 15 items that focus on different possible stressors, e.g. aspects of workload and ambiguity, powerlessness and vulnerability, and lack of control over one’s work situation (α = .78). Several studies have shown a positive association between perceived job stressors and workplace bullying (Attell, Kummerow Brown, & Treiber, 2017; Lewis et al., 2017).

Salutogenic health indicator scale (SHIS)

This is a scale developed by Bringsén et al. (2009) measuring salutogenic health indicators with 12 items. The items use six-point semantic differential scales. Based on a factor analysis and data from different organizations in Sweden (n = 4343), Rosander and Blomberg (2018) presented two subscales: self-confident and capable, and vigorous and energetic. However, in this study the total SHIS with all twelve items was used (α = .95). Many studies have shown the negative consequences of bullying on a person’s health and well-being (e.g. Bonde et al., 2016; Nielsen et al., 2014; Rodríguez-Muñoz, Moreno-Jiménez, & Sanz-Vergel, 2015).

Hospital anxiety and depression scale (HAD)

This is a widely used scale created by Zigmond and Snaith (1983). It contains two subscales: anxiety (HAD-A, α = .85) and depression (HAD-D, α = .84). Both anxiety and depression have previously been connected to exposure to workplace bullying. In a meta-analysis Nielsen and Einarsen (2012) showed firm support for the connection to mental health problems including anxiety and depression based on 33 studies and a combined sample size of more than 30,000. More recently, also in a meta-analysis, Verkuil, Atasayi, and Molendijk (2015) showed the same result.

Negative acts questionnaire–revised (NAQ-R)

The questionnaire assesses the frequency (never, now and then, monthly, weekly or daily) of exposure to 22 negative acts during the past 6 months. The acts can be categorized as work-related, person-related or physically intimidating behaviours (Einarsen et al., 2009). The cut-off scores 33 and 45 proposed by Notelaers and Einarsen (2013) were used in the study. In this study the Swedish version of NAQ-R (Rosander & Blomberg, 2018) was used (α = .91).

Analyses

The analyses of the study involved a number of steps, (a) determining how to categorize exposure to negative acts, (b) investigating how different aspects of work experience and perception of the organization, and health and well-being differ between different levels of exposure, and (c) studying the onset of exposure of different negative acts.

Categorization of exposure to negative acts

The proposed method for categorising into levels of exposure to bullying involved combining three different ways of assessing bullying, (a) cut-off scores for NAQ-R based on Notelaers and Einarsen (2013) study, (b) the Leymann (1996) criterion and variations of that principle (i.e. the number of negative acts occurring at least weekly), and (c) self-labelling as bullied. MANOVA was used as a way to test how each way of assessing bullying and their respective cut-offs could differentiate the variables from PSYWEQ about aspects of e.g. individual work experience, organizational conditions for work, negative exposure at work, as well as health and well-being measured by SHIS and HAD. The tests were between the different levels of bullying excluding all “not bullied” according to the three assessment methods. The reason for this exclusion was that the group “not bullied” was many times larger than the group “bullied” and the differences between “not bullied” and “bullied”, regardless of assessment method, were distinct. For NAQ-R the cut-off scores (Notelaers & Einarsen, 2013) and a division of self-labelling with at least weekly exposure are reasonable divisions based on previous studies. However, for the third way of assessing bullying that was used, that is, the number of negative acts that occur at least weekly (the operational classification method), the division is not so clear cut. Leymann (1996) suggested one per week, Mikkelsen and Einarsen (2001) suggested two per week, and Agervold (2007) suggested at least three per week as a criterion. In this study we tested two alternative limits for different levels of bullying, one based on Agervold’s at least three weekly negative acts, and one based on Mikkelsen and Einarsen’s at least two weekly negative acts. In Table 1 the lower and higher level of severity of bullying for the three ways of assessing bullying is presented. To further investigate which of the two alternatives of operational classification to use, we conducted a discriminant analysis for each.

Table 1. Lower and higher level of severity according to three ways of assessing bullying.

	Not bullied	Lower level	Higher level
NAQ-R cut-offs	< 33	≥ 33 (occasionally bullied)	≥ 45 (victims of bullying)
Self-labelling as bullied	No	Now and then or Monthly	Weekly or Daily
Frequency of weekly exposure	0 negative acts weekly	1–2 negative acts weekly	≥ 3 negative acts weekly	Alt. 1
		1 negative act weekly	≥ 2 negative acts weekly	Alt. 2

Work experience and perception of the organization at different levels

We used MANOVA to test if the levels and subcategories could differentiate using the variables of the study, with a post-hoc test (LSD) in which the focus only was on differences between adjacent levels or categories.

Onset of exposure to different negative acts

To identify onset of negative acts, we used pairwise comparisons of column proportions using Bonferroni correction adjusting for multiple comparisons.

Results

The results are divided into three parts. First, we present descriptive statistics for the variables of the study, and investigate the three different ways of categorising workplace bullying and their different levels. Then we present and test the suggested new levels of bullying. Finally, based on these new levels, we present the escalation of workplace bullying in terms of onset of different negative acts.

In Table 2 we present the descriptive statistics for the variables of the study together with their intercorrelations. The result showed many moderate or high correlations between the three ways of categorising bullying and the variables measuring, e.g. individual work experience and organizational conditions for work. This indicates that the ten chosen variables were relevant.

Table 2. Descriptive statistics and intercorrelation between the variables of the study.

	Mean	SD	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.
1. NAQ-R	27.79	8.13
2. Self-label.	1.10	0.44	.61
3. Weekly exp.	0.53	1.63	.85	.58
4. IWE	5.38	1.10	−.55	−.34	−.37
5. OCW	5.07	1.15	−.52	−.31	−.35	.90
6. RIM	5.10	1.32	−.42	−.22	−.29	.71	.80
7. O&E	4.90	1.43	−.32	−.16	−.22	.60	.68	.70
8. JJS	5.25	1.40	−.48	−.32	−.33	.76	.75	.60	.54
9. NEW	1.49	0.76	.58	.40	.46	−.45	−.44	−.29	−.25	−.40
10. WOSEQ	3.15	0.98	.49	.18	.32	−.50	−.52	−.49	−.34	−.44	.44
11. SHIS	3.97	1.22	−.37	−.19	−.25	.54	.47	.43	.33	.46	−.31	−.46
12. HAD-A	5.64	4.07	.37	.17	.25	−.44	−.38	−.37	−.28	−.39	.31	.46	−.69
13. HAD-D	3.61	3.32	.43	.21	.30	−.52	−.45	−.43	−.33	−.48	.32	.45	−.68	.70

Note: NAQ-R (Negative Acts Questionnaire–Revised), Self-label (Self-labelled as bullied based on a definition), Weekly exp (Weekly exposure to negative acts), IWE (Individual Work Experience), OCW (Organisational Conditions for Work), RIM (Roles in the organisation), Q&E (Quality & Efficiency), JJS (Joint Job Satisfaction), NEW (Negative exposure at Work), WOSEQ (Workplace Stress Exposure Questionnaire), SHIS (Salutogenic Health Indicator Scale), HAD-A (Hospital Anxiety and Depression scale, Anxiety subscale), and HAD-D (Hospital Anxiety and Depression scale, Depression subscale)

We conducted four separate MANOVAs, for each of the different ways of categorising into a lower and a higher level of bullying (see Table 1), with two separate analyses for weekly exposure. The results are presented in Table 3. All three ways of categorising workplace bullying including both alternatives for weekly exposure showed significant results. Self-labelling was least successful in terms of the number of dependent variables that significantly differed between the lower and higher level of workplace bullying.

Table 3. MANOVAs testing three ways of assessing a lower and higher level of workplace bullying (including the two alternative categorisations for weekly exposure).

	NAQ-R cut-offs	Self-labelling	Weekly exposure (alt 1)	Weekly exposure (alt 2)
Pillai’s trace (V)	.30	.13	.25	.20
n	308	128	407	407
df	10, 297	10, 117	10, 396	10, 396
F	12.79***	1.73^†	12.96***	9.89***
Partial eta squared	.30	.13	.25	.20
Between-subjects effects (F-value)
Individual work experience (IWE)	27.12***	6.41*	62.73***	57.57***
Organizational conditions for work (OCW)	19.00***	7.92**	39.67***	33.65***
Roles in the organization (RIM)	9.44**	5.57*	28.32***	23.55***
Quality and efficiency (Q&E)	6.10*	3.00 (ns)	15.47***	11.97***
Joint job satisfaction (JJS)	26.82***	5.88*	48.29***	23.02***
Negative exposure at work (NEW)	100.76***	6.28*	90.03***	62.63***
Stress exposure (WOSEQ)	37.44***	1.46 (ns)	43.87***	39.94***
Health (SHIS)	23.46***	0.12 (ns)	24.35***	22.25***
Anxiety (HAD-A)	26.50***	0.35 (ns)	32.25***	26.32***
Depression (HAD-D)	22.31***	0.21 (ns)	34.94***	28.35***

* p < .05, ** p < .01, *** p < .001

^† p = .081

Note: All categorised as Not bullied by respective method for assessment were excluded in this analysis.

For the two alternatives of weekly exposure categorization, the MANOVAs indicated that alternative 1 was the better alternative. However, to further investigate this the analyses for the two alternatives were followed up using discriminant analysis. For alternative 1 the discriminant function had a canonical R = .50 and could differentiate the two groups, Λ = .75, χ²(10) = 113.25, p < .001, with 74.4% of original grouped cases correctly classified. For alternative 2 the discriminant function had a canonical R = .45 and could differentiate the two groups, Λ = .80, χ²(10) = 89.18, p < .001, with 69.5% of original grouped cases correctly classified. Taken together, although both alternatives showed significant results, alternative 1 seems to be the better alternative of the two. In the following, weekly exposure will be divided into the following three levels: 0; 1–2; and 3 or more negative acts per week.

The three different ways of assessing workplace bullying, cut-off scores for NAQ-R, self-labelling and weekly exposure, were not completely overlapping. There were a number of examples of people categorized as bullied according to one of the assessment methods, but not by others. For example, 43% of those who self-labelled themselves as bullied at least “now and then” had no negative acts at least weekly, based on NAQ-R (see Table 4); 31% of those who were “not bullied” according to the lower NAQ-R cut-off at 33 labelled themselves as bullied at least “now and then” based on a definition (see Table 5); and 27% of those categorized as at least “occasionally bullied” (NAQ-R ≥ 33) had no negative acts at least weekly (see Table 6). Further, 20% of those who labelled themselves as “not bullied” reported at least one negative act occurring at least weekly (see Table 4) and 13% of these had an NAQ-R score of 33 or more (see Table 5). Finally, 12% of those “not bullied” according to the NAQ-R cut-off, that is a NAQ-R score below 33, reported one or more negative acts occurring at least weekly (see Table 6). In Tables 4, 5 and 6 we present the frequencies and percentages for all combinations of NAQ-R cut-off scores, self-labelling based on a definition and weekly exposure to negative acts.

Table 4. Frequencies and percentages for a combination of Weekly exposure and Self-labelling.

	Self-labelling based on a definition
Weekly exposure	Not bullied		Now and then or Monthly		Weekly or Daily		Total
0/w	1375	80.3%	51	48.6%	5	19.2%	1431
1–2/w	279	16.3%	23	21.9%	4	15.4%	306
≥ 3/w	58	3.4%	31	29.5%	17	65.4%	106
Total	1712	100%	105	100%	26	100%	1843

Table 5. Frequencies and percentages for a combination of NAQ-R and Self-labelling.

	Self-labelling based on a definition
NAQ-R	Not bullied		Now and then or Monthly		Weekly or Daily		Total
Not bullied	1492	87.1%	36	34.3%	5	19.2%	1533
≥ 33–44	189	11.0%	39	37.1%	5	19.2%	233
≥ 45	31	1.8%	30	28.6%	16	61.5%	77
Total	1712	100%	105	100%	26	100%	1843

Note: NAQ-R (Negative Acts Questionnaire–Revised)

Table 6. Frequencies and percentages for a combination of Weekly exposure and NAQ-R.

	NAQ-R
Weekly exposure	Not bullied		≥ 33–44		≥ 45		Total
0/w	1353	88.0%	78	33.5%	5	6.5%	1431
1–2/w	183	11.9%	114	48.9%	9	11.7%	306
≥ 3/w	2	0.1%	41	17.6%	63	81.8%	106
Total	1538	100%	233	100%	77	100%	1843

Note: NAQ-R (Negative Acts Questionnaire–Revised)

Levels of workplace bullying

The results presented in Table 3 showed that all three different ways of categorising workplace bullying could differentiate different levels of bullying, and Table 4–6 showed that the categorizations overlapped to some degree, but also contributed with unique information about the categorization of workplace bullying. The definition of workplace bullying includes a systematic exposure to negative behaviours (Einarsen et al., 2011). This systematic exposure is often operationalized as at least a weekly exposure (Leymann, 1996). Table 6 shows that only relying on the cut-off scores (based on a ROC analysis, Notelaers & Einarsen, 2013) to differentiate the bullied from the not bullied more than one in four categorized as bullied had no weekly exposure, that is, they were not bullied based on the definition. A similar reasoning can be made if only relying on the self-labelling method. As Table 4 shows, almost half of those reporting they had been bullied at least to some degree had no weekly exposure to negative behaviours, and almost one in five had no weekly exposure although they reported they had been bullied at work on a weekly basis. And finally, almost one in three of those self-labelling as bullied, at least to some degree, were not bullied if relying on the cut-off scores for categorization. To overcome the above-mentioned limitations of the three methods for classification used in this study and to better comply with the definition of workplace bullying, we used them in combination and in this included each method’s lower and higher level of bullying as presented in Table 1. As shown in Table 3 the lower and higher level for each method could, at least to some degree, differentiate the organizational and individual measures included in the study. The basis for creating our levels of bullying was the cut-off scores presented by Notelaers and Einarsen (2013). For Level 1 we included those with a score above the lower cut-off at 33, but with no or just the lower level of weekly exposure (less than three negative acts per week). We categorized this as a risk of bullying or incipient forms of bullying. For Level 2 we added the higher level of weekly exposure, that is, at least three weekly negative acts (based on the results presented in Table 3). Many exposed to negative acts, also on a weekly basis, do not see themselves as being bullied (possible reasons for this is discussed in the discussion section). So, for Level 3 we also included self-labelling as we believe that when exposed to many negative acts and many of them on at least a weekly basis, for a person to really be a victim of bullying (NAQ–R ≥ 45) one also needs to see oneself as bullied, at least to some degree. Therefore, as a combination of the cut-off scores for NAQ-R (≥ 33–44 and ≥ 45, Notelaers & Einarsen, 2013), the number of negative acts occurring at least weekly (0, 1–2 or 3 or more), and self-labelling based on a definition (now & then or monthly, and weekly or daily), we suggest three levels of bullying. Level 1 includes Risk of bullying and Incipient bullying (two subcategories), level 2 is called Ongoing bullying, and level 3 includes Severe bullying and Extreme bullying (two subcategories).

In all, five categories of workplace bullying were created. In this suggested model, those not bullied are those with an NAQ-R score below 33 (Notelaers & Einarsen, 2013). Comparing those below 33 with those at 33 and above using a MANOVA showed a very clear difference for all of the variables of the study, V = .32, F(10, 1800) = 86.11, p < .001, η² = .32. For the first level of workplace bullying there are two subcategories, Risk of bullying and Incipient bullying. For Risk of bullying the NAQ-R score must be 33 or higher, but with no exposure to negative acts at least weekly. For Incipient bullying there must also be one or two negative acts occurring at least weekly. For the next level, Ongoing bullying, the NAQ-R score still has to be 33 or higher, together with at least 3 negative acts weekly. Finally, for the third level of bullying, the NAQ-R score must be 45 or above (the upper cut-off according to Notelaers & Einarsen, 2013), together with at least 3 negative acts at least weekly, and self-labelling at least “now and then”. The third level has two subcategories, Severe bullying and Extreme bullying. For Extreme bullying to occur the self-labelling must be at least weekly. The categories are clarified in Table 7.

Table 7. Categorisation of workplace bullying based on NAQ-R cut-off scores, weekly exposure and self-labelling.

Main levels	Subcategories	NAQ-R cut-off score	Weekly exposure	Self-labelling
Not bullied		< 33	–	–
Level 1	Risk of bullying	≥ 33	No weekly negative acts	–
	Incipient bullying	≥ 33	1–2 weekly negative acts	–
Level 2	Ongoing bullying	≥ 33	≥ 3 weekly negative acts	–
Level 3	Severe bullying	≥ 45	≥ 3 weekly negative acts	Now and then or Monthly
	Extreme bullying	≥ 45	≥ 3 weekly negative acts	Weekly or Daily

Note: NAQ-R (Negative Acts Questionnaire–Revised)

One objection to the use of the number of negative acts at least weekly (weekly exposure) as an indication of bullying has been that there may be individuals that are exposed to a wide range of negative behaviour, but not one and the same on a weekly basis (Notelaers & Einarsen, 2013; Notelaers et al., 2006). The suggested categories use information about weekly exposure, but do not use it as the sole indicator, and the category “Risk for bullying” includes individuals with no weekly exposure. Of the 83 individuals with an NAQ-R score ≥ 33 and no weekly exposure (see Table 6), 94% had an NAQ-R score < 45 and 59% had an NAQ-R score between 33 and 35. Only five individuals have an NAQ-R score ≥ 45, and only one of them self-labelled as bullied on at least a weekly basis.

When categorising levels of bullying, one interesting aspect is in what way adjacent categories differ from each other. The reason for focussing on adjacent levels is that although, for example, joint job satisfaction differs significantly when comparing the not bullied to those exposed to severe workplace bullying, it is not a good indication that the suggested categories are useful. Two MANOVAs using the seven factors from PSYWEQ, and SHIS, HAD-A and HAD-D as dependent variables with the suggested levels of bullying (three main levels or five subcategories) as the fixed factor showed significant results for both. For the five subcategories, using Pillai’s trace, V = .48, F(50, 9015) = 19.31, p < .001, partial η² = .10. For the three main levels V = .46, F(30, 5409) = 32.71, p < .001, partial η² = .15. Table 8 presents the differences between adjacent sublevels of workplace bullying. The asterisks indicate significant differences in a negative direction between adjacent levels. Table 9 presents the corresponding results for differences between adjacent main levels of workplace bullying.

Table 8. Significant differences between adjacent sublevels of workplace bullying based on a post-hoc test after a MANOVA. The asterisks indicate a significant difference between adjacent sublevels.

	Not bullied compared to Risk	Risk compared to Incipient	Incipient compared to Ongoing	Ongoing compared to Severe	Severe compared to Extreme
Organizational conditions for work (OCW)	***	**			***
Roles in the organization (RIM)	***	**
Quality and efficiency (Q&E)	***	*
Joint job satisfaction (JJS)	***		*	*	**
Anxiety (HAD-A)	***		*	*
Depression (HAD-D)	***		*
Negative exposure at work (NEW)	***			***	***
Stress exposure (WOSEQ)	***			***
Health (SHIS)	***			**
Individual work experience (IWE)	***			*	*

* p < .05, ** p < .01, *** p < .001

Table 9. Significant differences between adjacent main levels of workplace bullying based on a post-hoc test after a MANOVA. The asterisks indicate a significant difference between adjacent levels.

	Not bullied compared to Level 1	Level 1 compared to Level 2 (Ongoing)	Level 2 (Ongoing) compared to Level 3
Anxiety (HAD-A)	***	**	*
Depression (HAD-D)	***	**	*
Joint job satisfaction (JJS)	***	*	***
Quality and efficiency (Q&E)	***	*
Individual work experience (IWE)	***		***
Negative exposure at work (NEW)	***		***
Stress exposure (WOSEQ)	***		***
Organizational conditions for work (OCW)	***		***
Health (SHIS)	***		**
Roles in the organization (RIM)	***		**

* p < .05, ** p < .01, *** p < .001

The results presented in Tables 8 and 9 were not only relevant as an indication of differences between adjacent categories, they also showed that different variables were relevant at different levels. All variables differed significantly when comparing Not bullied and Risk for bullying. When comparing Risk of bullying and Incipient bullying, the differences were mainly organizational aspects, that is, conditions for work, role clarity, and the perception of efficiency. The comparison of Incipient bullying and Ongoing bullying showed that the differences involved joint job satisfaction, and anxiety and depression symptoms. Ongoing bullying and Severe bullying differed in terms of negative exposure, stress, health and the individual work experience. Differences between the two level-three categories showed even more negative exposure, a worsening of organizational conditions for work, and individual work experience.

Onset of negative acts

Using the suggested levels of bullying with a focus on onset of the various negative acts of NAQ-R revealed an interesting pattern. In this study, onset was defined as the level at which the column proportion for weekly exposure differed significantly from zero. Once onset was established, additional differences for higher levels were not regarded as relevant as onset already has been established. For each level, the negative acts with onset at that level were sorted from the highest to the lowest proportion. Table 10 presents the onset of the different negative acts.

Table 10. Onset of the 22 negative acts from NAQ-R occurring at least weekly across the levels and sublevels of workplace bullying (bold numbers indicate onset). Percentage reporting at least weekly exposure.

					Level 1		Level 2	Level 3
			Total	Not bullied	Risk of bullying	Incipient bullying	Ongoing bullying	Severe bullying	Extreme bullying
3	Work below competence level	W	12.6	6.8	–	52.8	62.1	56.5	60.0
1	Information withheld	W	5.7	1.8	–	24.4	45.5	43.5	60.0
21	Unmanageable workload	W	5.9	2.1	–	20.3	51.5	43.5	46.7
16	Unreasonable deadlines	W	3.6	0.8	–	10.6	42.4	30.4	40.0
4	Responsibility removed or trivial tasks	W	2.9	0.5	–	12.2	30.3	30.4	33.3
14	Opinions ignored	W	2.2	0.2	–	2.4	24.2	39.1	60.0
18	Excessive monitoring	W	2.2	0.3	–	2.4	16.7	56.5	66.7
6	Ignored or excluded	P	1.8	0.2	–	4.9	18.2	17.4	60.0
5	Gossip and rumours	P	1.8	0.0	–	4.1	16.7	34.8	66.7
2	Humiliation or ridicule	P	1.7	0.1	–	3.3	12.1	39.1	60.0
7	Insulting or offensive remarks	P	1.4	0.1	–	0.8	9.1	34.8	60.0
13	Persistent criticism of errors or mistakes	P	1.1	0.0	–	0.0	9.1	30.4	46.7
8	Target of spontaneous anger	PI	1.5	0.1	–	4.1	13.6	26.1	46.7
11	Repeated reminders of errors or mistakes	P	1.4	0.1	–	0.0	10.6	26.1	80.0
12	Ignored or getting hostile reactions	P	1.5	0.2	–	0.8	12.1	17.4	73.3
19	Pressured to not claim what you are entitled to	W	0.9	0.0	–	2.4	9.1	17.4	20.0
15	Practical jokes by people you don’t get along with	P	0.8	0.0	–	0.8	3.0	17.4	53.3
17	Allegations against you	P	0.9	0.0	–	0.0	6.1	13.0	60.0
10	“You should quit your job”	P	0.8	0.0	–	0.8	3.0	8.7	60.0
20	Subject to excessive teasing and sarcasm	P	0.7	0.0	–	0.8	4.5	13.0	40.0
9	Intimidating behaviours	PI	0.8	0.1	–	0.8	9.1	4.3	33.3
22	Threats of violence or physical abuse	PI	0.4	0.1	–	0.0	9.1	0.0	6.7

Note: From left to right, the order of the item in NAQ-R (Negative Acts Questionnaire–Revised), a short description, if the item is work-related (W), person-related (P) or physically intimidating bullying (PI), percentage reporting at least weekly exposure in total and for each category.

At Incipient bullying there was onset of three negative acts, all work-related. At Ongoing bullying there was onset of six negative acts; the first four were also work-related and for the final two the first examples of person-related acts appeared. At Severe bullying all but two of the nine negative acts were person-related. Finally, at Extreme bullying two of the negative acts were person-related and one was physically intimidating behaviour. For the final negative act, “Threats of violence or physical abuse”, too few actually reported this at least weekly and a significant onset could not be identified.

In Appendices 1 and 2 we present the distribution of demographic and occupational groups across the three levels and the five subcategories of workplace bullying.

Discussion

One aim of this study was to identify and test a new way of categorising levels of bullying combining three different ways of assessing bullying that have been used previously (weekly exposure, e.g. Leymann, 1996; NAQ-R cut-off scores, Notelaers & Einarsen, 2013; and self-labelled victimization). The results showed that each of them could, in different ways, successfully differentiate a number of important organizational and work experience factors, as well as aspects of health and well-being. For weekly exposure the difference between one and two negative acts at least weekly, that is the Leymann (1996) criterion and the criterion suggested by Mikkelsen and Einarsen (2001), was not as evident as when comparing them to three negative acts at least weekly (Agervold, 2007). There was a lot of overlap between the three ways of assessing bullying. But there were also discrepancies, for example, a number of individuals that self-labelled as bullied although categorized as not bullied according to NAQ-R, and reverse, a number of individuals exposed to negative acts to the extent that they were classified as bullied according to NAQ-R who did not see themselves as bullied (according to the self-labelling method). One reason for this discrepancy may be understood in terms of feelings of shame (Lewis, 2004) and threats to the self-esteem of the victims (Einarsen et al., 2011). Viewing exposure to negative acts as an escalating process in which, for example, a conflict escalation eventually turns into bullying (Zapf & Gross, 2001), the point at which a person acknowledges to him- or herself that he or she is a victim of bullying may also mean acknowledging defeat.

The results showed that the suggested categorization including weekly exposure as one of the components can fail to correctly identify individuals as bullied if exposed to many different negative acts, but not one and the same on a weekly basis, which is in line with the critique posed by Notelaers and Einarsen (2013) and Notelaers et al. (2006). The assumption is that these individuals may be bullied on a weekly basis, but are exposed to a different kind of negative act each week. However, the results showed that very few individuals in this situation actually see themselves as bullied on a weekly basis (only one individual in this study). One suggestion for further research is to include one more question at the end of NAQ-R: “Of the above-mentioned negative behaviours, how often have you been exposed to at least one of them in the last 6 months?”. The response alternatives should be the same as for the other 22 items of NAQ-R (never, now and then, monthly, weekly, and daily). This would provide information on the frequency of exposure for those exposed to many different negative acts although no single one on a weekly basis. Another suggestion for NAQ-R on the lack of information on power imbalance would be to add a question focussing on that: “If you have been subjected to one or more of the negative acts mentioned above, to what extent have you had the opportunity to influence or put an end to it/them?” The response alternative here would probably be a Likert scale with the end points “Never” and “Very often”.

The new categorization proposed in this study, of three main levels and five subcategories, is a way to illustrate the process of workplace bullying, from risk for bullying to extreme bullying. The categorization is based in part on the operational criterion method, which does not need to merely include two groups – bullied and not bullied – as it is often framed (Einarsen et al., 2011). In this study the levels or categories of bullying are used as a way to illustrate the escalating process. We believe the idea of bullying as an escalating process is important, in order to identify a risk of bullying at an early stage, as this increases the chances of reversing the process. However, separating risk of bullying from actual bullying is also important, so as not to dilute the concept. In this study we differentiate exposure to negative acts which can occur infrequently or more systematically, and being bullied which includes the requirements of the definition in terms of weekly exposure. Being exposed to negative acts does not mean that one is the victim of bullying, but it increases the risk. In the model, levels two and three mean that an individual is being bullied, and level one is exposure to negative acts, which involves the risk of escalation and becoming bullied. This is similar to Nielsen’s (2009) division into targets of bullying behaviour and victims of bullying.

When testing the ability to differentiate between adjacent levels or categories of bullying on a number of central organizational factors, aspects of work experience, and health and well-being, there were some interesting findings. Previous studies have shown correlations between workplace bullying and both health- and well-being related outcomes and job-related outcomes (e.g. Dollard et al., 2017; Nielsen & Einarsen, 2012; Nielsen et al., 2014; Slany et al., 2014). First of all, the results showed that as soon as one is exposed to negative acts to a degree (NAQ-R ≥ 33), even though the frequency of any one negative act is less than weekly there is a difference in the way the individual perceives the organization and the experience of work. The result also showed clear consequences for health and well-being already for those categorized as at risk of being bullied compared to those not bullied. When adding frequency as a requirement for categorization (at least one negative act at least weekly), and comparing risk of bullying to incipient bullying, mainly organizational factors differed. The perception of the organizational conditions for work, role clarity and a perception for efficiency in the organization dropped. In line with the “work environment hypothesis (e.g. Salin & Hoel, 2011) organizational conditions play an important role in understanding workplace bullying and exposure to negative behaviours at work. For the next level, ongoing bullying, more negative acts on a weekly basis are required for categorization. The difference to incipient bullying at level one showed a drop in the joint job satisfaction as well as an increase in anxiety and depression symptoms. The connection between bullying and mental health problems has got strong support (e.g. Nielsen & Einarsen, 2012). Finally, level three, which includes both a higher NAQ-R score (≥ 45) and also self-labelling as at least bullied now and then, showed a clear difference to level two regarding negative exposure at work (e.g. threats, violence, sexual harassment and discrimination), stress exposure, health and well-being, and the individual work experience. Several studies have shown bullying as a major stressor with detrimental consequences for health and well-being (e.g. Attell et al., 2017; Lewis et al., 2017), and co-occurrence with other forms of negative social exposures (Lim & Cortina, 2005). In all, based on the results, we can conclude that different things happen regarding both health and well-being, and job-related outcomes, at different levels of workplace bullying when exposed to negative acts, or at least the perception of the exposed differs significantly between levels. Previous research has shown that exposure to workplace bullying has consequences for the exposed. The current study adds to this, but also shows that not everything happens at once – at different levels different consequences are in the foreground. Based on the overall result a plausible escalation process could be that the reaction is initially focussed on the organization and that something is wrong, maybe together with feelings of being treated unfairly. Next, there are feelings of being excluded or at least of being less included in the social interaction at the workplace. Anxiety and depression symptoms increase. Finally, more concrete negative exposure becomes apparent, stress level increases, health and well-being suffer, and the individual work experience drops, including support from supervisor and close co-workers.

The definition of workplace bullying used in this study define it as “an escalating process” (Einarsen et al., 2011, p. 22), but even so, as the current data is cross-sectional our results do not show a process. We can only show that individuals at different levels of workplace bullying exposed to, for each level, an increased frequency of negative behaviours experience increased negative consequences for their health and well-being, and their perception of their work environment. However, the nature of the temporal dynamics connected to workplace bullying needs more focus. There have been a few studies focussing on individual trajectories (Zapf & Gross, 2001), and more recently on the temporal dynamics of cyberbullying (Soni & Singh, 2018). When studying cyberbullying the continuous tracking of negative treatment is possible allowing for a more thorough study of the dynamics involved. The corresponding tracking of negative behaviours at work outside text-based interactions on the internet is not as easy to accomplish. The time-lag when studying workplace bullying longitudinally is often more than a year (Cole, Shipp, & Taylor, 2016; Nielsen & Einarsen, 2012). There are studies with a day-to-day tracking of negative behaviours, for example, Hoprekstad et al. (2019) studied naval cadets on an eleven-week sea voyage (measurement 33 consecutive days). The study design is probably intertwined with special conditions of the study and may be hard to transfer to other contexts. Ethically and methodologically there would be some issues, such as what it would mean to the individual, both victim and alleged perpetrator, in an organization to continuously highlight this. The continuous focus on identifying negative acts may also have consequences for the measurement as such.

The second part of the aim of this study involved illustrating the escalation of workplace bullying, focussing on the onset of different negative acts occurring at least weekly. Using the suggested levels or categories of workplace bullying, the onset of the 22 negative acts described in NAQ-R showed an interesting pattern. At first, for incipient bullying, the negative acts that were most frequent were all work-related: being assigned work below one’s competence level, having information withhold, and being given an unmanageable workload. Using the latent cluster approach Notelaers et al. (2006) found work-related bullying behaviours as a cluster below victims of bullying, something that resonates with the current finding. At the next level, ongoing bullying, there are still work-related bullying behaviours such as being given unreasonable deadlines and excessive monitoring, but also indirect or passive forms of person-related bullying, such as being excluded and having rumours and gossip about oneself spread. At level three, severe bullying, there is an onset of almost exclusively person-related bullying behaviours including more direct and active forms, such as being humiliated, insulted, or ignored, and having allegations made against oneself. For extreme bullying, even more intimidating bullying behaviours become frequent.

Although not presented earlier it is interesting to compare the result involving the onset of the different negative acts occurring at least weekly to the selection of items used in the short version of NAQ-R called SNAQ (Notelaers, van der Heijden, Hoel, & Einarsen, 2018). Of the nine items of SNAQ, six have an onset at severe bullying according to the suggested levels of workplace bullying. Of the other three, one has onset at incipient bullying and two at ongoing bullying. This supports the idea of SNAQ as a measure of exposure to severe bullying. The inclusion of only one item in SNAQ covering work-related bullying could prove to be a problem for identifying those in risk of becoming bullied. However, that is something for future research to investigate.

The self-labelling method does not inform of what negative acts are involved, and because people may have different thresholds for what being bullied entails, the interpretation of the results are more uncertain. For the behavioural experience method, using a number of relatively clear negative acts without mentioning bullying gives an idea of the bullying behaviours involved, but does not, in its current form, indicate whether the exposure is related to a power imbalance – something that the self-labelling method can do by way of the definition the question is based on. Using cut-off scores (Notelaers & Einarsen, 2013) or latent clusters (Leon-Perez et al., 2014; Notelaers et al., 2006) for the behavioural experience method does not provide information about the frequency of the negative acts one is exposed to. However, asking about exposure to a number of negative acts can provide information about the frequency if one adds the operational criterion method in which the criterion can be exposure to, e.g. one (Leymann, 1996), two (Mikkelsen & Einarsen, 2001), or three or more (Agervold, 2007) negative acts at least weekly. Using latent clusters can give an insight into the escalation process. Duration is captured using the instruction for both the self-labelling method and the behavioural experience method, e.g. “the last six months”. Taken together, all three sources of information, the self-labelling method, the behavioural experience method, and the operational criterion method, can provide information about all aspects of workplace bullying covered in the definition. The suggested levels of bullying use all three sources to assess workplace bullying and thus provide a better fit to the theoretical definition of workplace bullying. Empirically we have shown that the main levels and subcategories are meaningful divisions showing significant differences between adjacent categories, and provide insight into the escalation of workplace bullying, both in terms of different outcomes, and of onset of different negative acts at different points in the process.

Limitations

It is important to stress that the cut-off scores used in this article are based on Norwegian data (Notelaers & Einarsen, 2013). However, the two countries are neighbours and probably have a lot in common in terms of how society works and is perceived. Using the cut-off scores also showed clear differences when comparing groups above and below them. When constructing the levels of workplace bullying presented in this study, we assumed that exposure to negative acts escalates with time, but previous studies (Zapf & Gross, 2001) have shown that the bullying process can have other trajectories. However, they also showed that in a large majority of the cases there was a steady increase over time. Finally, the data in the current study, although a representative sample of the Swedish workforce, are not longitudinal, which means we can only show differences between individuals within the different levels or categories of workplace bullying suggested in this study, not the actual escalation process. One way to get a better idea of workplace bullying as a process would be to focus on the individuals exposed to negative behaviours at the first measurement point and then track them over time, a total of at least three observations (Cole et al., 2016), to see how the perception of the organization and working condition, as well as health and well-being change over time. To further understand the processes also the workgroups in which they work could be included to better understand the context of the exposure. A difficulty in studying individual cases of workplace bullying longitudinally is that it involves a small group of people, and a risk of drop-out as a result of a heightened risk for the bullied to leave the organization, being on long-time sick leave or unemployment (Glambek et al., 2015).

Conclusions and practical implication

Workplace bullying is a serious problem that can have detrimental effects for the individual in terms of a reduction in health and well-being (Nielsen et al., 2014), the risk of unemployment (Glambek et al., 2015), and ultimately even suicide (Nielsen et al., 2015). Occurrence of workplace bullying can also have a negative effect on the organization itself (Mathisen et al., 2008; Salin & Notelaers, 2017). Workplace bullying can be described as an escalating process from occasionally being exposed to negative acts to ending up a victim of severe bullying. In this study we have argued for a new way of categorising different levels of workplace bullying to better capture this escalation process. We did this by combining different ways of assessing workplace bullying to better match the criteria of the definition of workplace bullying. The results showed different outcomes and focus, depending on the level of bullying. We have also illustrated the escalation by identifying the onset of the various negative acts that are regarded as potential bullying behaviours. Taken together, we believe the suggested levels of workplace bullying can help both practitioners to better understand escalation and the different kinds of exposure at different levels, and be a new way for researchers to understand workplace bullying.

Using the suggested levels of workplace bullying can have a clearer pedagogical value than, for example, the latent clusters method, in that the levels more clearly identify the escalation of negative acts and what negative exposure can be expected at different points. This can have practical value in that it is easier to grasp and to convey to, e.g. HR personnel, managers and organizational psychologists. The added information about onset can also provide support to these groups to identify bullying behaviours early in the process and to use it as a way to structure the preventive work together with a focus on organizational factors that can be a breeding ground for workplace bullying.

Data-availability

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

The research project was approved by the Regional Research and Ethics Committee at Linköping University, Sweden (#2017/336–32).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the AFA Försäkring under Grant number 160285.

Notes

1. Leymann (1990) wrote: “These actions take place often (almost every day)” (p. 120) and 1996 “These actions occur on a very frequent basis (statistical definition: at least once a week)” (p. 168), that is, a reference to the frequency needed to call it bullying, but actually nothing about how many negative acts are required. The Leymann criterion is understood as at least one negative act per week. It is often referred to using one or both of these references.

Appendix 1. Distribution of demographic and occupational groups across the three levels of workplace bullying.

Table

		Not bullied	Risk of bullying	Ongoing bullying	Severe bullying	Ongoing or Severe bullying
Total		80.6	12.8	4.0	2.8	6.8
Gender	Male	76.5	16.3	5.0	2.2	7.2
	Female	84.5	9.2	3.1	3.3	6.4
Age	Less than 40 years	77.9	13.3	4.5	4.3	8.8
	40-49 years	81.8	12.7	2.8	2.6	5.4
	50-59 years	79.9	13.4	5.4	1.3	6.7
	60 years or older	88.1	9.0	2.3	0.7	3.0
Education	Max elementary school	72.9	16.3	3.2	7.6	10.8
	High school	80.1	13.8	4.2	2.0	6.2
	University degree	82.5	11.1	4.1	2.4	6.5
Married	Not married	79.2	13.7	4.7	2.4	7.1
	Married	82.2	11.5	3.2	3.1	6.3
Children	No children	79.3	13.1	5.3	2.4	7.7
	One or more children	81.8	12.3	2.9	3.0	5.9
Country of birth	Sweden	82.3	11.6	3.9	2.2	6.1
	Not Sweden	70.8	18.8	4.7	5.8	10.5
Type of contract	Fixed contract	80.8	13.1	3.7	2.4	6.1
	Temporary	76.5	8.5	8.1	7.0	15.1
Leading position	Leading position	80.0	16.6	2.5	1.0	3.5
	Not leading position	80.6	12.3	4.1	3.0	7.1
Seniority	0- 3 years	82.4	9.7	3.9	3.9	7.8
	3-10 years	77.0	15.5	5.1	2.5	7.6
	10-19 years	77.5	15.4	4.2	2.9	7.1
	19-50 years	85.5	11.1	2.9	0.5	3.4
Monthly income	Less than 25,000^‡	79.0	11.2	4.8	5.0	9.8
(SEK)	25-30,000	79.1	14.6	3.9	2.3	6.2
	30-40,000	81.9	13.5	3.5	1.3	4.8
	More than 40,000	83.2	12.2	3.5	1.1	4.6
Sector	Government	81.6	14.1	2.3	1.9	4.2
	Municipality	83.6	10.2	3.6	2.6	6.2
	Health	82.7	7.5	7.3	2.6	9.9
	Private	78.8	14.4	4.0	2.9	6.9
Number of employees	10-29	79.7	13.7	3.6	3.0	6.6
	30-59	82.6	12.0	2.7	2.6	5.3
	60-129	81.3	10.6	6.4	1.7	8.1
	130-499	78.7	15.8	2.5	3.0	5.5
	500 or more	80.7	10.8	5.4	3.1	8.5
City/town population	Less than 25,000	78.5	14.4	4.3	2.9	7.2
	25-75,000	84.0	10.8	3.8	1.5	5.3
	75-150,000	78.0	14.5	4.3	3.3	7.6
	Over 150,000	81.4	11.4	3.8	3.4	7.2

^‡ SEK = Swedish crowns (SEK 10 ≈ EUR 1)

Note: A calibrated weight based on gender, age, marital status, country of birth, income, education level, work sector and the number employed at the workplace was used for the above results (Särndal & Lundström, 2005).

Appendix 2. Distribution of demographic and occupational groups across the levels and sublevels of workplace bullying.

Table

		Not bullied	Risk of bullying	Incipient bullying	Ongoing bullying	Severe bullying	Extreme bullying
Total		80.6	5.6	7.2	4.0	1.8	1.0
Gender	Male	76.5	7.5	8.8	5.0	1.2	1.0
	Female	84.5	3.6	5.5	3.1	2.3	1.0
Age	Less than 40 years	77.9	5.6	7.7	4.5	3.6	0.7
	40-49 years	81.8	6.3	6.4	2.8	0.7	1.9
	50-59 years	79.9	5.7	7.8	5.4	0.6	0.7
	60 years or older	88.1	3.5	5.4	2.3	0.3	0.4
Education	Max elementary school	72.9	9.3	6.9	3.2	6.2	1.4
	High school	80.1	6.4	7.4	4.2	1.1	0.9
	University degree	82.5	4.1	7.0	4.1	1.4	1.0
Married	Not married	79.2	5.9	7.8	4.7	1.5	0.9
	Married	82.2	5.2	6.4	3.2	2.0	1.1
Children	No children	79.3	5.9	7.2	5.3	1.8	0.6
	One or more child	81.8	5.3	7.1	2.9	1.7	1.3
Country of birth	Sweden	82.3	4.8	6.8	3.9	1.4	0.8
	Not Sweden	70.8	9.5	9.3	4.7	3.7	2.1
Type of contract	Fixed contract	80.8	5.9	7.2	3.7	1.5	0.9
	Temporary	76.5	1.7	6.8	8.1	5.2	1.8
Leading position	Leading position	80.0	8.7	7.9	2.5	0.4	0.6
	Not leading position	80.6	5.2	7.1	4.1	2.0	1.0
Seniority	0- 3 years	82.4	4.5	5.2	3.9	3.0	0.9
	3-10 years	77.0	7.3	8.2	5.1	1.4	1.1
	10-19 years	77.5	5.8	9.6	4.2	1.6	1.3
	19-50 years	85.5	5.1	6.0	2.9	0.4	0.1
Monthly income	Less than 25,000^‡	79.0	5.0	6.3	4.8	3.7	1.3
(SEK)	25-30,000	79.1	6.4	8.3	3.9	0.9	1.4
	30-40,000	81.9	5.4	8.1	3.5	0.6	0.7
	More than 40,000	83.2	5.9	6.3	3.5	0.8	0.3
Sector	Government	81.6	7.8	6.3	2.3	1.8	0.1
	Municipality	83.6	4.5	5.7	3.6	1.2	1.4
	Health	82.7	3.7	3.8	7.3	1.5	1.1
	Private	78.8	6.1	8.2	4.0	2.0	0.9
Number of employees	10-29	79.7	4.5	9.2	3.6	2.1	0.9
	30-59	82.6	5.0	7.0	2.7	2.4	0.2
	60-129	81.3	6.5	4.2	6.4	1.0	0.7
	13-499	78.7	6.6	9.2	2.5	0.9	2.1
	500 or more	80.7	5.5	5.4	5.4	2.3	0.8
City/town population	Less than 25,000	78.5	6.2	8.3	4.3	2.1	0.8
	25-75,000	84.0	4.4	6.5	3.8	0.7	0.8
	75-150,000	78.0	7.8	6.7	4.3	2.1	1.2
	Over 150,000	81.4	4.0	7.4	3.8	2.3	1.1

^‡ SEK = Swedish crowns (SEK 10 ≈ EUR 1)

Note: A calibrated weight based on gender, age, marital status, country of birth, income, education level, work sector and the number employed at the workplace was used for the above results (Särndal & Lundström, 2005).