Research on workplace aggression has added new insights to our understanding of its antecedents, consequences, and underlying processes. Despite the progress, scholars have raised concerns regarding the current measurement methods (Aquino & Thau, Reference Aquino and Thau2009; Hershcovis, Reference Hershcovis2011; Hershcovis & Reich, Reference Hershcovis and Reich2013; Nielsen, Notelaers, & Einarsen, Reference Nielsen, Notelaers, Einarsen, Einarsen, Hoel, Zapf and Cooper2020). Notably, they have argued that although many aggression factors are relevant, most studies have focused on the frequency of aggression with little or no attention directed toward the severity of these behaviors (Aquino & Thau, Reference Aquino and Thau2009; Hershcovis, Reference Hershcovis2011).
To date, the most common way of measuring workplace aggression is by asking respondents to self-report the extent to which they have experienced aggression. This method implies a variable-oriented approach focusing on how the frequency of aggression varies in the population and identifying its antecedents and consequences (Craig & Smith, Reference Craig and Smith2000; Gabriel, Daniels, Diefendorff, & Greguras, Reference Gabriel, Daniels, Diefendorff and Greguras2015). Studies adopting this approach have demonstrated that aggression is detrimental to employee well-being (Schat & Kelloway, Reference Schat and Kelloway2003, Reference Schat, Kelloway, Barling, Kelloway and Frone2005), an idea supported by meta-analytic findings (Bowling & Beehr, Reference Bowling and Beehr2006). Studies considering the severity of aggression are still rare (for an exception, please see Escartín, Rodríguez-Carballeira, Zapf, Porrúa, & Martín-Peña, Reference Escartín, Rodríguez-Carballeira, Zapf, Porrúa and Martín-Peña2009; Nixon, Arvan, & Spector, Reference Nixon, Arvan and Spector2021). Theoretically, the severity of aggression makes meaningful sense because severe forms such as a physical attack, which may rarely occur, have a stronger adverse impact (Zapf, Knorz, & Kulla, Reference Zapf, Knorz and Kulla1996). Yet, we are unaware of studies attempting to explore the antecedents and outcomes of the severity of workplace aggression. We, therefore, try to fill this gap by considering both the frequency and severity dimensions of aggression for a better understanding of its antecedents and consequences.
Not only does it seem obvious that frequency and severity affect victims but how these two factors combine to have these effects also seems obvious. However, the previous research on the interaction effect of frequency and severity using a variable-oriented approach has surprisingly yielded little support. For example, in the health-protective behavior literature, when these variables are used to predict change in intentions or behaviors, the simple main effect of one or both is usually observed, but the interaction effect is almost always not significant (Weinstein, Reference Weinstein2000). This trend was found in correlational (Ronis, Reference Ronis1992) and experimental studies (Beck, Reference Beck1984; Kaplan, Hammel, & Schimmel, Reference Kaplan, Hammel and Schimmel1985). Indeed, the failure to find the significant interaction effect has guided Rogers (Reference Rogers1975) to drop the interaction assumption in the revised version of the protection motivation theory.
Although researchers have identified many reasons for the failure to find the interaction effect, the most compelling evidence suggests that the relationship between frequency and severity is something other than a simple multiplicative interaction (Weinstein, Reference Weinstein2000). To fully understand the effects of these two factors, one needs to consider the different configurations of frequency and severity. For instance, some may experience a high prevalence of severe aggression, while others may not experience aggression, suggesting that distinct groups of individuals may exist. The variable-oriented approach does not account for such heterogeneity in the sample. To investigate this possibility, we have adopted the person-oriented approach (using latent profile analysis [LPA]) that shifts the focus from variables to individuals (Notelaers, Einarsen, Witte, & Vermunt, Reference Notelaers, Einarsen, De Witte and Vermunt2006). The person-centered approach uncovers population heterogeneity, revealing subgroups or patterns characterized by shared perceptions of the severity and frequency of aggression. Neglecting these variations within theoretical frameworks may result in disregarding subtle nuances in how constructs function (Gabriel et al., Reference Gabriel, Daniels, Diefendorff and Greguras2015). By employing LPA (Vermunt, Reference Vermunt2010), this study seeks to uncover diverse victim profiles and identify whether any of these profiles may negatively relate to individual outcomes. Examining workplace aggression through the lens of LPA, our research makes a significant contribution by advancing our understanding of the relationship between distinct victim profiles, their antecedent conditions, and subsequent well-being outcomes.
In this research, we integrate rational choice theory and victim precipitation theory to establish a relationship between victim profiles and their antecedents, which may be sensitive to variation within victim profiles. Specifically, drawing on rational choice theory (Becker, Reference Becker1968), which posits that individuals seek pleasure over pain, we examine whether victim’s age, gender, and tenure predict their perception of exposure and severity of workplace aggression. Simultaneously, we employ victim precipitation theory (Olweus, Reference Olweus and Hodgins1993) to investigate the influence of two crucial personality traits – negative affect and psychopathy – as predictors of distinct victim profiles. While previous research has explored the impact of targets’ negative affect on their exposure to workplace aggression (Aquino, Grover, Bradfield, & Allen, Reference Aquino, Grover, Bradfield and Allen1999), a notable research gap exists concerning the victimization of employees with psychopathic traits and their subjective interpretation of these experiences. Accordingly, this research contributes to the field of workplace aggression by introducing psychopathy as a potential antecedent, thereby expanding our knowledge in this area.
We also draw on learned helplessness (LH) theory (Abramson, Seligman, & Teasdale, Reference Abramson, Seligman and Teasdale1978) to investigate if these distinct victim profiles can be differentiated in terms of their experiences of helplessness, as indicated by general health outcomes. While numerous studies have tried to understand the impact of aggression on victims’ internalizing symptoms like anxiety and depression (Nielsen & Einarsen, Reference Nielsen and Einarsen2012; Presti, Pappone, & Landolfi, Reference Presti, Pappone and Landolfi2019), this study is the first to explore the effect of perceived severity on victims’ such symptoms. Integrating rational choice, victim precipitation, and LH theories establishes a crucial foundation for our hypotheses, offering a unique perspective on workplace aggression and its associated factors. Rational choice theory explores decision-making processes in target selection, victim precipitation theory examines how victim traits precipitate aggression, and LH theory explains psychological responses to adversity. By incorporating these theories, we gain insights into conditions under which characteristics of aggression (frequency and severity) may associate with diminished employee well-being. Consequently, our findings provide actionable insights for managers to customize interventions and support mechanisms, addressing workplace aggression by considering diverse victim characteristics and their perceptions of severity.
The remainder of the paper proceeds as follows. We start with a brief conceptualization of the frequency and severity of aggression by focusing on the person-oriented approach (LPA) as an analytical lens. Next, we propose hypotheses related to how distinct victim patterns are related to antecedents and the victim’s general health outcomes. Finally, we discuss key findings and implications for both theory and practice.
Frequency and severity of aggression
We conceptualize aggression as a broad spectrum of behaviors that (a) hold the potential to cause harm, (b) the target is motivated to avoid, and (c) occurs in the work-related context (Schat & Frone, Reference Schat and Frone2011; Schat & Kelloway, Reference Schat, Kelloway, Barling, Kelloway and Frone2005). Although many factors might be relevant, most aggression definitions primarily focus on frequency and severity.
The perceived frequency refers to the number of times a behavior has occurred during the specified time (generally in the last 6 months). Apart from self-labeling global measures of aggression that ask respondents whether they have been victims, the most widely used methods employ formative measures using frequency anchors. These measures constitute a list of behavioral items aggregated to measure the level of aggression (Hershcovis, Reference Hershcovis2011). For example, the Workplace Harassment Questionnaire consists of 24 behavioral items (Björkqvist, Östermanm, & Hjelt-Bäck, Reference Björkqvist, Österman and Hjelt-Bäck1992), averaged to use as a continuous measure of workplace harassment. This method has amassed a wealth of knowledge, showing to impact employee behaviors (e.g., performance, citizenship behavior), attitudes (e.g., job satisfaction), and health (e.g., physical health, stress). The downside of this method is that it relies on two assumptions: (1) all episodes of aggression are equal in severity and (2) the higher the frequency, the worse the aggression outcome (Hershcovis & Reich, Reference Hershcovis and Reich2013). These assumptions can be particularly questionable, because a small body of literature suggests that aggressive behaviors vary in severity (Escartín et al., Reference Escartín, Rodríguez-Carballeira, Zapf, Porrúa and Martín-Peña2009).
Perceived severity refers to the intensity or harmfulness the victim attributes to aggressive behavior (Hershcovis, Reference Hershcovis2011). Researchers have long advocated for the severity of aggression because seemingly mild behaviors may adversely affect victims. Although it is still evolving in the workplace literature, severity is a well-researched area in the school bullying literature. This stream consistently shows physical abuse as the most severe form and relational abuse as the least severe form (Chen, Liu, & Cheng, Reference Chen, Liu and Cheng2012; Ellis & Shute, Reference Ellis and Shute2007). In the workplace literature, this topic has recently gained momentum. Rodríguez-Carballeira, Solanelles, Vinacua, García, and Martín-Peña (Reference Rodríguez-Carballeira, Solanelles, Vinacua, García and Martín-Peña2010) conducted a Delphi study with 30 experts and developed six bullying categories. The study identified emotional abuse (attacking and sneering at workers’ feelings and emotions) as the most severe form, while social isolation was the least severe form of bullying. Another study by Escartín et al. (Reference Escartín, Rodríguez-Carballeira, Zapf, Porrúa and Martín-Peña2009), with a sample of 300 workers from four different organizations in Spain, used the hierarchical categorization developed by Rodríguez-Carballeira et al. (Reference Rodríguez-Carballeira, Solanelles, Vinacua, García and Martín-Peña2010) and obtained similar findings. Nixon et al. (Reference Nixon, Arvan and Spector2021) showed that employees who reported higher intensity of workplace mistreatment reported increased psychological strain and intentions to turnover while also reporting reduced affective commitment.
Unlike severity, the research on frequency has yielded accurate and conclusive results using the variable-centered approach. However, such an approach ignores the possibility that the frequency and severity can cooccur in numerous configurations in which individuals form combined perceptions of two factors. In contrast, a variable-based approach that accounts for the average relationship between variables does not account for unobserved heterogeneity in the target population (Mattern, Marini, & Shaw, Reference Mattern, Marini and Shaw2012). Therefore, scholars have repeatedly called for a person-oriented approach (e.g., LPA) that accounts for within-individual variation (Einarsen, Hoel, & Notelaers, Reference Einarsen, Hoel and Notelaers2009; Notelaers et al., Reference Notelaers, Einarsen, De Witte and Vermunt2006). Therefore, in this study, we employ a person-oriented approach that provides an analytical lens to explore issues that are otherwise difficult to address.
A person-oriented approach to workplace aggression
The strategy of classifying individuals into subgroups based on their observed characteristics is not new to the aggression literature. For example, applying latent class models, Notelaers et al. (Reference Notelaers, Einarsen, De Witte and Vermunt2006, Reference Notelaers, Vermunt, Baillien, Einarsen and De Witte2011) conducted a series of studies that identified six distinct patterns of victims who differ in frequency and nature of reporting bullying. They have observed that certain demographic and occupational groups are at increased risk of bullying. For example, employees between the ages of 35 and 54, public servants, blue-collar workers, and employees working in food and manufacturing industries are at elevated risk of becoming victims (Notelaers, Vermunt, Baillien, Einarsen, & De Witte, Reference Notelaers, Vermunt, Baillien, Einarsen and De Witte2011). In another study, using a sample from 24 different workplaces, Leon-Perez, Notelaers, Arenas, Munduate, and Medina (Reference Leon-Perez, Notelaers, Arenas, Munduate and Medina2014) identified six unobserved latent clusters consistent with the findings from Notelaers et al. (Reference Notelaers, Einarsen, De Witte and Vermunt2006), (Reference Notelaers, Vermunt, Baillien, Einarsen and De Witte2011)). However, when employing a short-negative act questionnaire, Notelaers, Van der Heijden, Hoel, and Einarsen (Reference Notelaers, Van der Heijden, Hoel and Einarsen2019) have observed a four-cluster pattern.
The person-oriented approach captures the unobserved heterogeneity in the sample by identifying the subgroup that shares similar configurations (also known as patterns or profiles) (Arnold, Connelly, Gellatly, Walsh, & Withey, Reference Arnold, Connelly, Gellatly, Walsh and Withey2017; Foti, Bray, Thompson, & Allgood, Reference Foti, Bray, Thompson and Allgood2012). In the context of workplace aggression, these subgroups (patterns or clusters) are victims who share similar perceptions of frequency and severity. Such different profiles across groups would remain undetected by employing a variable-oriented approach. Therefore, the first objective of this study is to identify the victims who share a similar profile by applying LPA. Using a broad sample of employees, we examine for the presence of patterns of the perceived severity and frequency of aggression.
Possible victim profiles
We expect to identify two primary sets of victim patterns. The first set of patterns will likely have variable scores that are all consistently high or low – where the frequency and severity of aggression are all expressed at approximately the same level. A review of 31 studies showed that the prevalence of severe forms of bullying in Europe, occurring every week or daily, is between 1% and 4% (Zapf, Einarsen, Hoel, & Vartia, Reference Zapf, Einarsen, Hoel, Vartia, Einarsen, Hoel, Zapf and Cooper2003). Schat, Frone, and Kelloway (Reference Schat, Frone, Kelloway, Kelloway, Barling and Hurrell2006) documented similar findings and reported that nearly 1.3% of the US workforce experiences physical violence every week. These results reiterate that many victims experience severe forms of aggression regularly. Therefore, we expect to see a group of victims who face aggression frequently (four or more times in the last 6 months) also perceive it as severe (see Cluster#6, Notelaers et al., Reference Notelaers, Einarsen, De Witte and Vermunt2006). In addition, we also expect some employees to have never experienced any form of aggression. Generally, workplace aggression studies report a large group of non-victims. For example, Notelaers et al. (Reference Notelaers, Einarsen, De Witte and Vermunt2006) reported 35% of the study participants as non-victims, and Leon-Perez et al. (Reference Leon-Perez, Notelaers, Arenas, Munduate and Medina2014) reported 32% as non-victims. Therefore, we expect to find a group of non-victims in our study sample. Because it is illogical to expect non-victims to perceive aggressive behaviors as severe, we believe they perceive workplace aggression as less severe. As a result, we hope to find a cluster where the participants experience significantly less or no aggression (see Cluster #1, Notelaers et al., Reference Notelaers, Einarsen, De Witte and Vermunt2006) and perceive these behaviors as less severe.
The second set will comprise patterns where the constituent variable scores are mixed. Instead of reducing bullying to an either-or phenomenon, we consider that a varied nature and extent of bullying may occur. For example, Keashly and Jagatic (Reference Keashly, Jagatic, Einarsen, Hoel, Zapf and Cooper2011)reported that 20% of the sample indicated bullying as less frequent and experiencing some forms of mistreatment but not others (see Clusters 2–5, Notelaers et al., Reference Notelaers, Einarsen, De Witte and Vermunt2006). Therefore, we expect victims to exhibit varying levels of frequency and severity, such as a high frequency of low-severe behavior or a low frequency of high–severe behaviors. Thus, we expect to find patterns where the frequency and severity scores are all consistent (high vs. low) and mixed patterns with a variable degree of frequency and severity.
Hypothesis 1: The nature of patterns is characterized by a consistent pattern (low or high on both dimensions) and a variable pattern (high on one dimension relative to the other).
Demographic predictors of victim profiles
A clear assessment of risk groups is vital to develop effective tailor-made prevention and intervention strategies (Notelaers et al., Reference Notelaers, Vermunt, Baillien, Einarsen and De Witte2011). However, one of the aspects that we know little about is the role of victim demographics (Escartín et al., Reference Escartín, Rodríguez-Carballeira, Zapf, Porrúa and Martín-Peña2009; Escartín, Salin, & Rodríguez-Carballeira, Reference Escartín, Salin and Rodríguez-Carballeira2011). Therefore, this paper explores how employee demographics such as age, gender, and organizational tenure are associated with the likelihood of a victim’s membership to aforementioned subgroups.
The existing empirical evidence on the relationship between victim characteristics and aggression provides inconclusive findings. The most influential meta-analysis showed no association between age and aggression, while it showed a significant relationship between gender and aggression (Bowling & Beehr, Reference Bowling and Beehr2006). A study by Notelaers et al. (Reference Notelaers, Vermunt, Baillien, Einarsen and De Witte2011) reported that employees between the ages of 35 and 54 are at risk for aggression. In contrast to previous findings, they observed that employees younger than 25 years and temporary contract employees are at least risk. Also, they did not find any significant difference in prevalence rates among different genders. However, the authors reported that public servants, blue-collar workers, and employees working in the food and manufacturing sectors are at elevated risk of becoming victims. Regarding the severity of aggression, Escartín et al. (Reference Escartín, Salin and Rodríguez-Carballeira2011) reported gender as a significant predictor and showed that men and women perceived different behaviors as severe.
The rational choice theory, also referred to as the effect-danger ratio in aggression literature (Björkqvist, Österman, & Lagerspetz, Reference Björkqvist, Österman and Lagerspetz1994), provides a possible explanation for demographic differences in aggression (Jensen, Patel, & Raver, Reference Jensen, Patel and Raver2014). According to this theory, people engage in activities when they envision pleasure over pain. Drawing on the rational choice theory, we argue that victim demographics could influence perpetrators’ evaluations of the potential costs and benefits because certain demographic groups, such as women and those in lower tenure, are considered ‘safe victims’ (Bowling & Beehr, Reference Bowling and Beehr2006). The main reason for this is the perceived power deficit. Older and experienced men generally occupy higher levels of organizational hierarchy and gain higher social power (Cortina & Magley, Reference Cortina and Magley2009). As a result, they pose a significant risk to perpetrators. However, women and lower-tenure employees hold less power and become easy targets (Carli, Reference Carli1999). Perpetrators are also less afraid of these groups because of their inability to defend or retaliate (Salin, Reference Salin2003).
There is also a difference in perceived severity among different groups. For instance, men and women face different societal norms and expectations for acceptable behaviors (Ely & Padavic, Reference Ely and Padavic2007). It is reasonable to argue that most cultures reward men for being aggressive, expect to feel less threatened, and perceive aggressive acts as more acceptable. On the other hand, women are expected to show interpersonal sensitivity and warmth (Franke, Crown, & Spake, Reference Franke, Crown and Spake1997). Women who behave aggressively are perceived as a threat to these expectations and may face backlash. Therefore, compared to men, women perceive these behaviors as more severe. Similarly, younger employees are more likely to classify aggression as severe due to lower possibilities of defending themselves. Therefore, we hypothesize the following:
Hypothesis 2a: Age is negatively associated with patterns characterized by high levels of one or both frequency and severity.
Hypothesis 2b: Employee tenure is negatively associated with patterns characterized by high levels of one or both frequency and severity.
Hypothesis 2c: Gender is positively associated with patterns characterized by high levels of one or both frequency and severity such that women are more likely to belong to these patterns.
Disposition as predictor of victim patterns
The research based on the victim precipitation theory (Olweus, Reference Olweus and Hodgins1993) suggests that a victim’s negative emotional traits play a vital role in precipitating workplace aggression (Aquino et al., Reference Aquino, Grover, Bradfield and Allen1999). The two important negative emotional traits that may precipitate aggression are psychopathy and negative affectivity (NA). Although substantial research has explored the relationship between NA, psychopathy, and aggression (Bowling & Beehr, Reference Bowling and Beehr2006), most of these studies have focused on the perpetrators’ perspective, explicitly examining whether individuals with high psychopathy scores are more likely to engage in workplace aggression. There is limited research from the victim’s perspective as only a few studies have explored the role of psychopathy in predicting the individual’s experience of aggression. However, these studies have either examined nonworking populations (Daigle & Teasdale, Reference Daigle and Teasdale2018) or did not directly assess these relationships.
Psychopathy is characterized by low negative emotionality with deficits in affect (e.g., callousness), narcissism (constant need for admiration), and self-control (e.g., impulsiveness) (Fanti & Kimonis, Reference Fanti and Kimonis2012; Hicks & Patrick, Reference Hicks and Patrick2006; Jones & Paulhus, Reference Jones and Paulhus2014). Research on psychopathy suggests that characteristics of psychopathy, including impulsivity, callous-unemotional traits (Fanti & Kimonis, Reference Fanti and Kimonis2012), low self-control (Pratt, Turanovic, Fox, & Wright, Reference Pratt, Turanovic, Fox and Wright2014), and impaired emotion regulation (Garofalo, Neumann, & Velotti, Reference Garofalo, Neumann and Velotti2021), predispose psychopathic individuals to behave in ways that exploit and exhibit a disregard for the well-being of others (Boddy, Reference Boddy2006). These behaviors elicit negative-aggressive responses from those they affront (Daigle & Teasdale, Reference Daigle and Teasdale2018; Valentine, Fleischman, & Godkin, Reference Valentine, Fleischman and Godkin2018). Fanti and Kimonis (Reference Fanti and Kimonis2012) showed that youth with high callous and impulsivity traits faced peer victimization. Daigle and Teasdale (Reference Daigle and Teasdale2018) showed that individuals with psychopathic traits are at increased risk of experiencing recurring victimization because these individuals cannot see the consequences of their actions and are likely to place themselves in situations leading to aggression. This sequence may explain why individuals with high psychopathic traits experience higher levels of aggression.
We also argue that people with higher levels of psychopathy do not perceive aggression or other harmful acts as severe. We draw on the fearlessness model of psychopathy to justify our argument (Fowles & Dindo, Reference Fowles and Dindo2009; Lykken, Reference Lykken1995). According to this model, psychopaths are characterized by a high fear threshold, suggesting they are less likely to experience fear and may not respond to risks like others. Supporting this notion, research by Hosker-Field, Molnar, and Book (Reference Hosker-Field, Molnar and Book2016) showed that individuals with high psychopathy had lower risk perception across several domains, including ethical, financial, health/safety, and social. Accordingly, individuals high in psychopathy are expected to evaluate aggressive behavior as being less severe.
NA refers to an individual’s tendency to experience high levels of negative emotions (Watson & Clark, Reference Watson and Clark1984). The extant evidence shows a positive association between NA and the experience of workplace aggression (Aquino & Thau, Reference Aquino and Thau2009; Penney & Spector, Reference Penney and Spector2005). We offer three potential explanations for the positive association, both of which are based on victim precipitation theory, which suggests that employees with certain personality traits attract aggression because they tend to violate social norms or threaten others’ identities (Glomb, Reference Glomb2002). First, interactions with those that exhibit higher levels of negative affect may be experienced as unpleasant or irritating which may increase the likelihood that high NA individuals experience mistreatment. A second possibility is that employees high in NA focus on negative aspects of themselves and the world in general (Watson & Clark, Reference Watson and Clark1984). Because of their heightened sensitivity to threat, these individuals are ‘hypervigilant’ and analyze every social interaction for hostile intent (Aquino & Thau, Reference Aquino and Thau2009; Bies, Tripp, & Kramer, Reference Bies, Tripp, Kramer, Giacalone and Greenberg1997). As a result, high NA employees are more likely to make hostile attribution for ambiguous behavior and report more workplace aggression than employees low in NA. Third, it may be due to high-NA employees being perceived as incompetent. NA can drain the emotional resources required to complete work-related tasks and make high-NA employees perform less effectively. Such employees become easy targets as they are perceived as lacking the ability or resources to defend themselves from or retaliate against unfair treatment. Therefore, we argue that high-NA employees report experiencing high aggression and perceive these behaviors as more severe because of their negative emotionality.
Taking all the discussion together, we expect that high-NA individuals are more likely to belong to patterns consisting of high frequency and high severity profiles. In contrast, individuals scoring high on psychopathy are more likely to belong to high-frequency but low-severity groups. Consistent with this, we hypothesize the following:
Hypothesis 3a: NA is positively associated with patterns characterized by high levels of one or both frequency and severity.
Hypothesis 3b: Psychopathy is positively associated with patterns characterized by high frequency but negatively associated with patterns characterized by high severity.
Learned Helplessness (LH): Implications for emotional well-being
Workplace aggression is an organizational stressor associated negatively with victims’ physical and emotional well-being (Bowling & Beehr, Reference Bowling and Beehr2006; Schat & Kelloway, Reference Schat and Kelloway2000, Reference Schat and Kelloway2003). Emotional well-being is the overall effectiveness of an employee’s psychological functioning (Wright & Cropanzano, Reference Wright and Cropanzano2004). Despite the growing research on employee well-being, only a few studies have focused on victim internalizing problems – defined as overcontrol of emotions – including social withdrawal, feeling worthless, and dependency (Guttmannova, Szanyi, & Cali, Reference Guttmannova, Szanyi and Cali2008). Therefore, we explore the negative outcome of aggression by considering the within-victim configurations as a possible explanation. Drawing on LH theory (Abramson et al., Reference Abramson, Seligman and Teasdale1978; Seligman, Reference Seligman, Friedman and Katz1974), we examine the relationship between workplace aggression and victims’ internalizing problems. Specifically, we focus on three outcomes indicative of internalizing problems: anxiety, social dysfunction, and loss of confidence (Graetz, Reference Graetz1991; Shevlin & Adamson, Reference Shevlin and Adamson2005).
The LH theory explains why victims are at increased risk for internalizing symptomatology. The central assertion of this theory is that when individuals face persistent stressful events, they perceive the events as out of control. The sense of uncontrollability creates a deficit in motivational, cognitive, and emotional functioning (Abramson et al., Reference Abramson, Seligman and Teasdale1978). As a result, victims think, feel, and behave like they are helpless. Applying this to the workplace aggression context, we argue that victims who experience persistent aggression and cannot foresee an escape can develop deficits in psychological processes, putting themselves at risk of internalizing problems. First, the cognitive deficiency comprising difficulty in learning and problem-solving makes one feel worthless and inadequate by hurting their self-confidence (Abramson et al., Reference Abramson, Seligman and Teasdale1978; Vartia, Reference Vartia2001). Next, the motivational deficiency comprises retarded initiation of voluntary responses resulting in the victim’s passivity and intellectual slowness. As a result, motivational deficits interfere with the victim’s ability to function normally (Schat & Kelloway, Reference Schat and Kelloway2000). Finally, emotional deficiency stemming from the uncontrollability of the situation produces depressed affect and anxiety (Vartia, Reference Vartia2001; Zapf et al., Reference Zapf, Knorz and Kulla1996).
Only a few studies have investigated the relationship between workplace aggression and internalizing problems. Although these studies have not modeled the severity explicitly, the research shows that victims of severe forms of aggression present lower psychological health (Zapf et al., Reference Zapf, Knorz and Kulla1996), suggesting the adverse impact of severity on victims’ emotional well-being. Thus, we predict that individuals who exhibit high frequency and severity patterns will experience the highest anxiety levels, loss of confidence, and social dysfunction. In contrast, individuals who experience no or low aggression do not feel helpless. Similarly, victims who experience less severe forms also may not see an imminent threat and, therefore, do not feel their situation as uncontrollable. Accordingly, we believe that individuals who exhibit low frequency and severity patterns report the lowest anxiety levels, loss of confidence, and social dysfunction.
Thus, drawing on patterns from the previous hypothesis, we propose that a consistent high pattern will exhibit the highest levels of loss of confidence, anxiety, and social dysfunction. We also expect that consistent low patterns will show the lowest loss of confidence, anxiety, and social dysfunction.
Hypothesis 4a: Patterns characterized by high levels of both frequency and severity will be associated with the highest levels of anxiety, loss of confidence, and social dysfunction.
Concerning the mixed patterns (i.e., showing unique configurations of high and low levels), we expect that these patterns are related to low levels of emotional well-being. As discussed earlier, there is a hierarchy to the severity of aggression, and some behaviors are considered more severe than others (Escartín et al., Reference Escartín, Rodríguez-Carballeira, Zapf, Porrúa and Martín-Peña2009; Zapf et al., Reference Zapf, Knorz and Kulla1996). For instance, the previous research shows that attacking a person’s private life is a more powerful predictor of psychosomatic complaints than social isolation, which was only weakly related. Because severe behaviors can create the anxiety of repeating similar events in the future, they may have a lasting impact on victims. Thus, we predict that a pattern characterized by high exposure to less severe behavior reports relatively low levels of emotional well-being (but not as low as the low–low pattern). We also expect that the negative effect of aggression patterns is more pronounced with behaviors of high severity. Accordingly, we hypothesize the following:
Hypothesis 4b: Mixed patterns characterized by high levels of severity will be associated with higher levels of loss of confidence, anxiety, and social dysfunction than patterns characterized by high-frequency levels.
Method
Research design and participants
The data for this study came from organizations across North America. The participants were recruited using an online panel from Qualtrics. Online panels are widely used in organizational behavior research (Connelly, Zweig, Webster, & Trougakos, Reference Connelly, Zweig, Webster and Trougakos2012). All participants completed an electronic survey with informed consent information explained on the first page of the survey. The participants were offered a small incentive to participate.
Several a priori measures were taken in the study design to address common method bias. We used measures that have been validated in previous research using both self and other reports. We completely randomized the order of measures and the sequence of questions on all surveys and assured participants that the survey was anonymous (Podsakoff, Mackenzie, Lee, & Podsakoff, Reference Podsakoff, Mackenzie, Lee and Podsakoff2003). Furthermore, participants showing extreme speeding compared to others (relative to the sum of median response time) were dropped subsequently by Qualtrics. To mitigate the sampling bias, we recruited respondents from 25 different sectors. Finally, to reduce the response bias, we explicitly stated in the respondent information sheet that the respondents had the choice to omit questions that were not related to them or made them feel uncomfortable answering.
Respondent characteristics
Our sample represents participants from 25 sectors, with the most popular categories being healthcare and medical (13%), education (12%), retail (10%), construction (8%), and customer services (7%). A total of 330 participants were contacted to participate in the study. However, after excluding those who failed to complete the survey, we ended up with a sample of 200 with a response rate of 60%. Fifty-eight percent of the participants were female and 42% were male. The average age of the participants was between 30 and 50 years, and the average experience was between 10 and 19 years. Sixteen percent of the participants did not experience any aggression.
Measures
Workplace aggression
Workplace aggression was measured using a 25-item scale. We followed a rigorous process to generate a list of workplace aggression items. Two research assistants carried out a thorough literature review to pool items from various workplace aggression constructs ranging from low severity to high severity. They independently coded the items for redundancy, and in instances where disagreement arose, a consensus was achieved through discussion. The constructs include perceived severity of bullying (Escartín et al., Reference Escartín, Rodríguez-Carballeira, Zapf, Porrúa and Martín-Peña2009), incivility (Cortina, Magley, Williams, & Langhout, Reference Cortina, Magley, Williams and Langhout2001), social undermining (Duffy, Ganster, & Pagon, Reference Duffy, Ganster and Pagon2002), bullying (Negative Acts Questionnaire; Einarsen & Raknes, Reference Einarsen and Raknes1997), and workplace aggression (Glomb, Reference Glomb2002; Schat, Reference Schat2004). After a thorough review and excluding all the repetitive items, we finalized 29 items. However, it is essential to note that we dropped four items related to cyberbullying from this study as cyberbullying was outside of the focus of this study. Sample items of the scale include ‘Being insulted in front of others’ and ‘Being glared at or given a dirty look’. The respondents were asked to specify the frequency with which they encountered each behavior in their workplace over the past 6 months. The frequency scale was anchored on a 4-point scale from 0 (never) to 3 (4 or more times). The coefficient alphas for the frequency scale are 0.96. The respondents were asked to rate the severity of the behaviors. The severity was anchored on a 10-point scale from 1 (not at all severe) to 10 (extremely severe). The coefficient alpha for this scale was 0.95.
General Health Questionnaire
We adopted the General Health Questionnaire (GHQ; Shevlin & Adamson, Reference Shevlin and Adamson2005) for measuring emotional well-being constructs – anxiety, loss of confidence, and social dysfunction. The scale consists of 12 items. The sample item is ‘have you felt under strain’. The scale was anchored on a 7-point scale from 1 (not at all) to 7 (all the time). The coefficient alpha value for the GHQ was 0.91
Psychopathy
We adopted a short version dark triad (SD3) scale to measure psychopathy (Jones & Paulhus, Reference Jones and Paulhus2014). The sample items include, ‘I will say anything to get what I want’. The scale was anchored on 5 points from 1 (strongly disagree) to 5 (strongly agree). The coefficient alpha value for the scale is 0.86.
Negative affectivity
NA was measured using a 10-item shortened PANAS scale (Watson, Clark, & Tellegen, Reference Watson, Clark and Tellegen1988). The scale consists of five items, each for positive affectivity and NA. The respondents indicated the degree to which they generally feel on five descriptors of negative emotions (such as ‘distressed,’ ‘upset,’ and ‘nervous’) on a 5-point scale from 1 (not at all) to 5 (extremely). The coefficient alpha for the negative affect scale is 0.89.
Analytic strategy
We used LPA to identify groups of individuals who exhibited a similar pattern of scores on the frequency and the severity. We later compared these groups with covariates and hypothesized outcomes (Arnold et al., Reference Arnold, Connelly, Gellatly, Walsh and Withey2017). We estimated LPA models using a robust maximum likelihood estimator within Mplus 8.2 (Muthén & Muthén, Reference Muthén and Muthén1998-2017). Following the recommendations of Nylund, Asparouhov, and Muthén (Reference Nylund, Asparouhov and Muthén2007), we began by testing and evaluating the fit of a two-pattern model to the data. We then repeated this process by adding an additional pattern each time until it became clear that increasing the number of patterns could no longer be justified (Foti et al., Reference Foti, Bray, Thompson and Allgood2012). Once an optimal number of patterns was determined, we modeled the relations between these discrete classes and various auxiliary variables. For each individual, we determined the most likely class membership based on the distribution of classification probabilities from the LPA (Asparouhov & Muthén, Reference Asparouhov and Muthén2014). We later assessed the relations between predictors, auxiliary variables, and the pattern solution.
We adopted the R3STEP and DU3STEP in Mplus to model antecedents and outcomes, respectively. R3STEP command (Asparouhov & Muthén, Reference Asparouhov and Muthén2014) conducts a series of multinomial logistic regression to assess whether an increase in a predictor variable result in a higher probability that an individual belongs to one class over another (Arnold et al., Reference Arnold, Connelly, Gellatly, Walsh and Withey2017). DU3STEP command provides a test of mean equality across different groups for each outcome variable. We analyzed antecedents (R3STEP) and distal outcomes (DU3STEP) separately.
Results
Descriptive statistics
Descriptive statistics, Cronbach’s alpha coefficients and correlations for all the study variables are represented in Table 1. As predicted, demographics and personal traits are correlated with either frequency or severity of aggression and with outcome variables. We observed moderate to strong correlations between aggression and the outcome variables.
Table 1. Descriptive statistics and correlations
N = 200, where appropriate, scale reliabilities are presented in the diagonal.
* p< .05, **p <.01.
Preliminary measurement models
To test the preliminary model consisting of frequency and severity of workplace aggression, we conducted a series of confirmatory factor analyses using maximum likelihood estimation within Mplus 8.2 (Muthén & Muthén, Reference Muthén and Muthén1998-2017). We assessed and compared the three nested models. In the first model, we specified all indicator variables to load on a single latent factor (χ2 = 4088.08, df = 1160; CFI = .53; RMSEA = .11; SRMR= .20). In the second model, the indicator variables loaded on two latent factors (χ2 = 2579.52, df = 1159; CFI = .77; RMSEA = .07; SRMR= .08). The results suggested that the two-factor model provided a reasonably good fit to our data, and the results also suggested that there is no difference between two-factor oblique and orthogonal models. Therefore, we retained the two-factor measurement model (frequency and severity) to be consistent with the full range of aggressive behaviors.
We conducted confirmatory factor analyses using maximum likelihood estimation for the GHQ. We assessed and compared the three nested models. Previous literature on GHQ supports both the two- and the three-factor models. Based on the recommendation by Politi, Piccinelli, and Wilkinson (Reference Politi, Piccinelli and Wilkinson1994), we first specified all indicator variables to load two latent factors (χ2 = 117.81, df = 53; CFI = .96; RMSEA = .08; SRMR= .05). In the second model, the indicator variables loaded on three factors as suggested by Martin (Reference Martin1999)(χ2 = 302.20, df = 51; CFI = .83; RMSEA = .16; SRMR= .11). In the third model based on Graetz (Reference Graetz1991), we again loaded on three factors but with different indicators loaded on to these factors (χ2 = 98.26, df = 51; CFI = .97; RMSEA = .06, SRMR= .04). The results suggested that the three-factor model suggested by Graetz (Reference Graetz1991) provided a good fit for our data. Therefore, we retained the three-factor measurement model to be consistent with the previous literature.
LPA and test of hypotheses
To identify groups with distinct levels of aggression, we conducted a series of latent profile models. We performed LPA with two to six classes on the full sample. To determine the number of classes, we examined each model using several relative fit indices (Arnold et al., Reference Arnold, Connelly, Gellatly, Walsh and Withey2017; Gabriel et al., Reference Gabriel, Daniels, Diefendorff and Greguras2015), including the log-likelihood (LL), Akaike information criterion (AIC), the Bayesian information criterion (BIC), and sample size–adjusted Bayesian information criterion (SSA-BIC), where lower values indicate the more parsimonious models. To show the accuracy of the classification process, we examined Entropy, where higher values indicate higher accuracy, with values >.80 considered higher (Clark & Muthén, Reference Clark and Muthén2009). Finally, we used Lo–Mendell–Rubin–adjusted likelihood ratio test (LMR; Lo, Mendell, & Rubin, Reference Lo, Mendell and Rubin2001) and bootstrap likelihood ratio test (Nylund et al., Reference Nylund, Asparouhov and Muthén2007), which compares the estimated model (K classes) with a model having one less class (K – 1 class). A significant p-value (p < .05) suggests that the model with the K pattern fits the data well than the model with the K – 1 pattern.
Table 2 provides the fit statistics of latent profile models. Considering all the fit statistics, the four-pattern model represents the best choice given that it displayed lower LL, AIC, BIC, and SSA-BIC than the other models. Although five and six-pattern models presented lower fit statistics, nonsignificant Lo–Mendell–Rubin–adjusted likelihood ratio test and bootstrap likelihood ratio test values show that the model is not a better fit than the four-pattern model. It was also observed that the four-pattern model has adequate membership in all profiles ranging from 15% to 43% of the total sample. In addition, the four-pattern model shows that members are rightly classified with average latent class probabilities for most likely pattern membership ranging from 0.91 to 0.99.
Table 2. Model fit statistics for latent pattern structure
Notes: LL = log-likelihood; AIC = Akaike information criterion; BIC = Bayesian information criterion; SSA-BIC= Sample size–adjusted Bayesian information criterion; LMR = Lo–Mendell–Rubin–adjusted likelihood ratio test; BLRT = bootstrap likelihood ratio test.
Figure 1 presents a graphic representation of the four-pattern model. Table 3 displays the estimated means and confidence intervals for the frequency and severity of aggressive behaviors in each of the four groups. We label pattern 1 as ‘low–low’ (43%) because the group members experienced little to no aggression and perceived aggressive behaviors as less severe. We label the members of pattern 2 as ‘high–low’ (26.5%), and members of this group experienced aggressive behaviors but perceived these behaviors as less severe. We label the third pattern as ‘moderate–moderate’ (15.5%), where members experienced high levels of aggression and perceived these behaviors as severe. Finally, we labeled the fourth pattern as ‘high–high’ (15%), where the group members experienced the highest levels of severe aggression. The results confirmed the existence of patterns composed of different levels of frequency (high vs. low exposure group) and severity (high vs. low severity group). Thus, we found support for the consistent patterns by obtaining groups with high and low levels of both variables. Regarding the mixed patterns, we derived a group with high frequency and low severity levels. However, we did not derive a group with low frequency and high severity, thus providing partial support for Hypothesis 1.
Figure 1. Latent profiles of frequency and severity.
Table 3. Descriptive information for four latent patterns of frequency and severity of aggression
Note: Frequency = frequency of aggression; severity = perceived severity of aggressive behaviors; CI = confidence interval.
The results from covariate analysis reveal that several predictors have implications for pattern membership. Table 4 shows the results of the predictor analysis. We ran all analyses using the R3STEP option in Mplus. Examination of multinomial logistic regression coefficients (using pattern one, low–low, as a reference group) showed that all the predictors have contributed to predicting at least one pattern membership. For ease of interpretation, we present the logistic regression coefficients as odds ratio (OR). An OR of >1 show that for every 1-unit change in the predictor variable, the probability of being classified in the group is more relative to the comparative group. In testing Hypothesis 2a, the OR associated with age shows that older employees are 0.79 and 0.71 times, respectively, to belong to the moderate–moderate and high–high victims’ groups. In addition, the regression coefficient for these groups (−.25 and −.35, p < .05) are negative, suggesting age associates negatively with frequency and severity, thus supporting Hypothesis 2a. Next, consistent with Hypothesis 2b, our results showed that the risk of facing severe forms of aggression is higher for women than men. Compared to the reference group (low–low), women are 4.06 and 3.15 times more likely to belong to the high–high and moderate–moderate groups, respectively. In terms of organizational tenure, our results showed that experienced employees are 0.64 times less likely to belong to the high–high victim group. However, the results with other subgroups yielded non-significant results, thus partially supporting Hypothesis 2c.
Table 4. Predicting pattern membership from predictors (R3STEP)
Note: Pattern 1 (low–low) is the reference group. Coefficient = multinomial logistic regression coefficients; SE = Standard error of the coefficients; OR = odds ratio. Bolded coefficients are statistically significant, p< .05.
In terms of dispositional predictors, we found support for NA as a predictor of victim profiles. The results from Table 4 showed that NA relates positively with high frequency and severity groups (.53 and .86, p < 0.05). In addition, OR values reveal that people with high NA are 2.36 and 1.71 times more likely to belong to high−high and moderate−moderate groups, respectively. Thus, showing support for Hypothesis 3a. Regarding psychopathy, results reveal that it is positively related to frequency and severity, which contradicts our hypothesis. Consistent with our hypothesis, we found that psychopathy is positively associated with frequency. However, as shown in Table 1, we also found that psychopathy relates positively to severity. These results are further supported by covariate analysis, showing that compared to the reference group, employees scoring high on psychopathy are 2.87 times more likely to belong to the high−high group, 1.84 times more likely to belong to a moderate–moderate group, and 1.56 times more likely to belong to high–low groups. Thus, Hypothesis 3b is not supported.
Table 5 shows the results from outcome analyses. We ran all analyses using the DU3STEP option in Mplus. The test of mean equality across the pattern was performed using the three-step method with 3 degrees of freedom for the overall chi-square (χ2) test. Therefore, to examine if the four patterns differ regarding internalizing problems, we referred to the Wald chi-square (χ2) test of mean equality. For all three outcomes, the means differed across all four patterns: social dysfunction (χ2 (3) = 11.57, p <.05); anxiety (χ2 (3) = 8.04, p <.01); and loss of confidence (χ2 (3) = 15.91, p <.05). The mean differences for consistent pattern groups (low−low and high−high) suggest that both frequency and severity of aggression relate positively with anxiety, loss of confidence, and social dysfunction. The results show that the high−high victim group has the highest means, and the low–low victim group has the lowest means for all three outcome variables. Thus, supporting Hypothesis 4a. Regarding mixed patterns, our data produced a high frequency and low severity profile (high–low group) but did not produce a low frequency of high-severity behaviors groups. Instead, we witnessed a moderate–moderate group. Therefore, we could not test mixed patterns regarding outcome variables. However, we should note that for the high–low group, the means of social dysfunction (2.85) and loss of confidence (2.58) are higher than the low–low group but higher than the moderate–moderate group. However, for anxiety, the mean for the high–low group (3.41) is higher than the moderate–moderate group (3.30), thus partially supporting Hypothesis 4b.
Table 5. Three-step results for distal outcomes (DU3STEP): Evaluating the effect of latent pattern membership on emotional well-being
Note: All analyses are run using DU3STEP procedure in Mplus. The mean values for social dysfunction, anxiety, and loss of confidence for each pattern group are displayed. A test of mean equality across the pattern groups was performed using the three-step procedure for the overall chi-square (χ2).
Discussion
The primary goal of this research was to examine the underlying patterns of frequency-severity dimensions of aggression. LPA provided evidence for four distinct victim groups, thus suggesting that aggression occurs in various configurations distinguishable by the severity and frequency of acts involved. Contrary to our prediction, the data did not support mixed patterns. We did not find support for the low-frequency–high-severity behaviors. The reason may be that aggression is not a random act but rather occurs in an orderly fashion progressing from less to more severe acts – referred to as the escalation hypothesis (Glomb, Reference Glomb2002). Indeed, workplace aggression is described as a continuum that begins with mild acts such as disrespect and discourtesy and escalates into verbal assault and intimidation, and lastly, results in physical aggression (Einarsen, Reference Einarsen1999; Einarsen, Hoel, Zapf, & Cooper, Reference Einarsen, Hoel, Zapf, Cooper, Einarsen, Hoel, Zapf and Cooper2020; Weatherbee & Kelloway, Reference Weatherbee, Kelloway, Kelloway, Barling and Hurrell2006). The family violence research that supports the escalation hypothesis shows that psychological aggression often precedes physical aggression (Murphy & O’Leary, Reference Murphy and O’Leary1989; Schat & Kelloway, Reference Schat and Kelloway2000). Therefore, it is reasonable to expect that individuals experience severe aggression only after experiencing mild forms first. Our results are consistent with previous research on the prevalence of workplace aggression that showed 95.7% of US workers who experienced physical violence also reported experiencing psychological aggression (e.g., social isolation, withholding work-related information). Moreover, only 0.4% reported exposure to workplace violence among those who did not experience psychological aggression (Schat et al., Reference Schat, Frone, Kelloway, Kelloway, Barling and Hurrell2006).
Next, the study presents the importance of demographic and personality characteristics in predicting workplace aggression profiles. Compared with younger employees, the risk of belonging to the high−high group is lower for older employees. The organizational tenure also yielded similar results suggesting that experienced employees are less likely to become victims of severe aggression. Further, the study supported the idea that women are safe victims. Our data showed that the risk of facing persistent and severe aggression is higher for women than men. Compared with the reference group (low–low), women are four times more likely to belong to the high−high category and nearly three times more likely to belong to the moderate–moderate category. Regarding dispositional factors, people with high NA perceived they experienced frequent episodes of severe aggression. The results also offered interesting insights into psychopathy. People who scored high on psychopathy are almost three times more likely to belong to the high−high pattern and almost twice more likely to belong to the moderate–moderate group. In contrast to our hypothesis, our results corresponding to psychopathy suggest that such individuals not only reported experiencing high levels of aggression but also viewed these behaviors as severe. There are two plausible explanations for the tendency of individuals with psychopathic traits to perceive workplace aggression as more severe when directed at them. First, it may be that although these individuals tend to lack empathy for others, they are highly emotionally sensitive to behaviors that could be considered as threatening to their self-interest and personal well-being. This might lead them to evaluate aggressive behavior directed at themselves as particularly severe. Recent research on psychopathy substantiates our argument and questions the notion of a complete absence of empathy in primary psychopathy, as these individuals often exhibit high agentic positive emotionality (Lilienfeld et al., Reference Lilienfeld, Patrick, Benning, Berg, Sellbom and Edens2012), self-esteem (Falkenbach, Howe, & Falki, Reference Falkenbach, Howe and Falki2013), and affective empathy (Lishner et al., Reference Lishner, Vitacco, Hong, Mosley, Miska and Stocks2012). Second, employees with psychopathic traits frequently exhibit a heightened sense of self-importance and narcissistic tendencies. Therefore, they may perceive aggression directed at them as being particularly offensive, and therefore rate it as more severe. Because psychopathy is a multifaceted personality trait encompassing callousness, narcissism, and impulsivity, additional research is necessary to better understand and more thoroughly unravel the motivations behind why people with psychopathic traits tend to interpret instances of aggression directed at them as more severe.
Finally, the study offered implications of victim profiles for employee emotional well-being. As predicted, individuals in the high–high group reported the highest anxiety, loss of confidence, and social dysfunction, while the low–low category reported the lowest levels. Further, the moderate–moderate category reported the second-highest levels of loss of confidence and social dysfunction. However, this group reported low levels of anxiety than the high–low victim group, which relates positively to social dysfunction and loss of confidence. The negative implications of frequency to health outcomes we witnessed in this study are consistent with the previous research. However, the high- and low-severity patterns that we identified are also psychologically and behaviorally distinct. Therefore, the study offered valuable insights regarding the negative implications of the severity. The results suggest that victims of severe mistreatment are more likely to develop internalizing problems because of the long-lasting effects. In addition, fear of repeating such incidents in the future may drain victim resources. Eventually, the resource drain may interfere with the proper functioning of cognition and emotional processes.
Contributions
The study makes several notable contributions. The aggression literature has focused predominantly on the prevalence and used a variable-centered approach to identify the adverse outcomes. We contribute to the aggression literature by demonstrating how frequency and severity cooccur in patterns with varying levels. Although a few studies have examined the severity, these studies have only highlighted its importance by demonstrating how aggressive behaviors differ in their severity. The implications of severity on employee outcomes are seldom examined. Therefore, the finding that perceptions of frequency and severity coexist has implications for researchers and practitioners. Indeed, there were no patterns that represented the high or low levels of only one dimension. Therefore, research investigating aggression without considering these dimensions may inaccurately portray how victims perceive aggression. This study takes a significant first step in studying the implications of frequency and severity on employee emotional well-being using latent class analysis.
The study also makes theoretical contributions. Jointly considering rational choice theory and victim precipitation theory provided a framework for investigating demographic and individual difference characteristics as potential antecedents of aggression profiles. Consistent with the rational choice theory, we found that female, young, and new employees are at risk of belonging to high frequency and high severity patterns. Consistent with the victim precipitation theory, we showed that negative emotional traits are positively related to the frequency and severity of aggression. Thus, we argue that based on victims’ characteristics and the levels of social power earned because of their demographic roles, aggression is perceived differently by distinct groups. Therefore, instead of seeing these differences as an undesirable bias, these differences should be captured using items with high differentiating ability that can elicit a unique response from distinct groups (Escartín et al., Reference Escartín, Salin and Rodríguez-Carballeira2011).
From the outcome standpoint, our research contributes to the LH theory by supporting it as a framework to assess the impact of aggression on victims’ cognitive, emotional, and motivational deterioration. Building on the LH theory, we showed that victims who experience severe aggression report worse internalizing problems than those who encounter fewer moderate forms of aggression. Overall, the results showed that anxiety, loss of confidence, and social dysfunction increased with increased exposure to severe forms of aggression. Although theoretical evidence supports the importance of severity (Einarsen et al., Reference Einarsen, Hoel and Notelaers2009), its implications are not based on empirical evidence (Keashly & Jagatic, Reference Keashly, Jagatic, Einarsen, Hoel, Zapf and Cooper2011). Therefore, ours is the first study to provide empirical evidence for the impact of severity on emotional well-being.
The insights from the study open new avenues for future research by suggesting a person-oriented approach to conceptualizing workplace aggression. For instance, the nonexistence of a pattern with a low frequency of high-severity behaviors shows that workplace aggression is hierarchical, as proposed by many scholars (Lutgen‐Sandvik, Tracy, & Alberts, Reference Lutgen‐Sandvik, Tracy and Alberts2007). Perhaps, the victim of physical violence is also the victim of incivility. In other words, lower-order aggression (e.g., incivility) may be a precursor for higher-order aggression (e.g., physical violence). Researchers can explore these possibilities using a person-oriented approach (Leon-Perez et al., Reference Leon-Perez, Notelaers, Arenas, Munduate and Medina2014).
Finally, our results suggest that modeling severity and frequency are imperative in assessing workplace aggression. This finding is substantiated by showing how victims from different profiles experienced distinct emotional well-being outcomes. Therefore, future research should focus on comparing the broad range of aggression constructs (e.g., incivility, social undermining, and bullying), which is beyond the scope of the present study (Nixon et al., Reference Nixon, Arvan and Spector2021).
Practical implications
The finding that severity and frequency are crucial for understanding aggression has several implications for practitioners. First, instead of seeing workplace aggression as an either-or phenomenon to distinguish between victims and non-victims (Notelaers et al., Reference Notelaers, Vermunt, Baillien, Einarsen and De Witte2011), one must examine the varied nature of harmful acts. Such information help managers in developing effective intervention strategies and differentiate their efforts between prevention and assistance strategies. For example, victims of the low–low group are not exposed to aggression. Therefore, providing information on organization policies and resources may be sufficient for such victims. However, victims of the high−high category may require a comprehensive intervention strategy. These include re-examining anti-bullying policies, training, and, if possible, conflict resolution at an individual case level.
Acknowledging that victim demographic and personality characteristics influence conceptualizing workplace aggression also has implications for practice. Ignoring these factors may lead to underestimating the consequences of workplace aggression. Therefore, in the committee that tackles aggression issues, we emphasize having a group representing different demographic segments of the workforce. A committee comprising diverse members benefits from having a comprehensive view of aggression policies considering the risks, consequences, and implications of these policies for various workgroups.
Limitations
Despite the contributions of our study, when interpreting the results, one must remember several limitations. First, the sample size for this study was only moderately large; therefore, future studies should consider a larger sample. Despite this, we found significant results for both antecedents and consequences. In addition, our data are a good representative of the workforce covering multiple sectors, contributing to the generalizability of the meaning of the latent class solution. However, it is advantageous to conduct controlled experiments that help determine causality, which was not possible to assess in this study.
Next, although our study has focused on multiple social demographic antecedents, we could not test complex models involving interaction effects among these variables because of the multicollinearity issues (Notelaers et al., Reference Notelaers, Vermunt, Baillien, Einarsen and De Witte2011). A large sample size would allow the interaction effect between social and demographic variables. Further, we only focused on aggression from a single source, coworkers. The perceived severity of behaviors may change when perpetrated by different sources. For instance, yelling and shouting by an outsider (customer) may not be as severe as yelling by a supervisor. Therefore, more research is needed to understand the severity of aggression coming from different sources.
Finally, although studies have shown how aggression impacts organizations, we could not offer insights into organizational-level outcomes. In this study, we examine the consequences for the victim outcomes. Therefore, future research should extend this work by considering the team and organizational outcomes, such as performance and organizational citizenship behavior.
Conclusion
Overall, this study is the first to adopt a within-individual-based approach considering both severity and frequency dimensions of aggression. In doing so, we empirically provide evidence that both the frequency and severity of negative acts are imperative in determining the full consequences of aggression. Therefore, this study adds value to both theory and practice.
Competing interests
The authors declare none.
Hamsa Gururaj (hgururaj@ubishops.ca) is now an Assistant Professor of Human Resource Management at Bishop’s University. She received her doctorate in Business Administration from McMaster University. Her research focuses on determinants and outcomes of workplace aggression.
Aaron Schat (schata@mcmaster.ca) is an Associate Dean of the DeGroote School of Business at McMaster University. His research focuses on workplace aggression, emotional labour, and employee stress. He has published his research in top peer-reviewed journals such as the Journal of Applied Psychology, the Journal of Occupational Health Psychology, and the Journal of Organizational Behaviour.
Open access
