Leadership is a multifaceted and continuously evolving research field (Clark and Harrison, 2018), and the sub-field of destructive leadership, which has evolved in parallel to post-heroic leadership research, is no exception (Mackey et al., 2019). The interest in destructive leadership is mainly based on the extent and severity to which these behaviors influence employe outcomes negatively (Zhang and Liao, 2015). The emphasis on destructive leadership as consequential for employe outcomes is also reflected in the commonly used definition of destructive leadership, which refers to behaviors that undermine employe motivation, well-being, or job satisfaction (Einarsen et al., 2007). Over the last two decades, discussions on what constitutes destructive leadership has developed from merely including leaders’ aggressive behaviors toward employes to taking a wider perspective (Skogstad et al., 2017). A wider perspective involves a more holistic approach in which leaders’ passive behaviors, susceptible followers, and the environment, also contribute to the understanding of the destructive leadership process (Thoroughgood et al., 2018).

Currently, researchers seem to agree that destructive leadership is a multidimensional phenomenon (Skogstad et al., 2017). However, there is still some debate on the dimensions encompassed by the concept (Thoroughgood et al., 2012). For example, the commonly used definition developed by Einarsen et al. (2007) has been repeatedly criticized for its inclusion of the supportive-disloyal dimension (i.e., organizational directed leadership behaviors) of destructive leadership (Krasikova et al., 2013; Schyns and Schilling, 2013). Leadership literature generally suggests that as leadership in the workplace involves employe-targeted influence, it should not include non-employe targeting behaviors (e.g., counter-productive work behaviors such as supportive-disloyal behaviors; Schyns and Schilling, 2013). Consequently, a recent study on the differential effects of both employe- and organizational-directed destructive leadership shows that negative employe effects are produced primarily by the former (Schmid et al., 2018).

An employe-oriented perspective on the definition of destructive leadership is also valid from a theoretical perspective. For example, Conservation of Resources (COR; Hobfoll, 1989) has been suggested as a useful framework to understand how destructive leadership may influence employe outcomes (e.g., Otto et al., 2018). COR depicts that objects (e.g., tools), personal characteristics (e.g., emotional stability), conditions (e.g., social support), or energies (e.g., money) are valued resources that employes’ strive to obtain, retain, foster and protect (Hobfoll and Shirom, 2001). The theory further proposes loss of resources as the defining principle for stress. Hence, stress ensues either if individuals perceive a threat to resources, actual loss of resources, or if they are unable to regain a resource loss (Hobfoll, 1989). As the quality of the social (i.e., leader-employe)relationship is understood as a major resource, destructive leadership behaviors places employes at risk for high levels of stress. In turn, this may negatively influence a range of employe well-being (e.g., in terms of job satisfaction) and performance (e.g., in terms of work-role performance) outcomes (Inceoglu et al., 2018).

We agree with the view of leadership at work as a social process, and that therefore non-employe targeted behaviors should not be included in the definition of destructive leadership. However, whereas some exclude passive/indirect leadership behaviors from their definition of destructive leadership (Schyns and Schilling, 2013), we share the view that the destructiveness of a leadership process should be determined by the actual harm it causes employes (Skogstad et al., 2007; Thoroughgood et al., 2018). That is, not by the extent to which a leader’s destructive behaviors are intentional or perceived as expressions of aggressiveness by employes. Such a view is in line with empirical findings of the kinds of behaviors employes describe as destructive leadership behaviors when interviewed (Schilling, 2009; Shaw et al., 2011; Larsson et al., 2012). In other words, forms of more or less active/direct, as well as passive/indirect employe-directed leadership behaviors are included in employes’ appraisals. It is also in line with theoretical suggestions on what kind of leadership behaviors that may cause negative employe outcomes. In addition to explain negative employe outcomes of active destructive leadership, COR theory has also been used to explain the relation between passive destructive leadership and negative employe outcomes. From an employe perspective, passive destructive leadership stands in the way of guidance and social support needed to perform job tasks satisfactory, which may cause resource depletion and thereby be a cause of workplace stress (Einarsen et al., 2007).

Hence, with few exceptions (e.g., Thoroughgood et al., 2012), most recent studies also exclude non-social behaviors, and organize destructive leadership behaviors into two main categories: active/direct and passive/indirect (e.g., Fosse et al., 2019; Kaluza et al., 2019; Trépanier et al., 2019; Lundmark et al., 2020).

There are few easy-to-use measures to evaluate the multidimensional nature of destructive leadership (Krasikova et al., 2013). Destructive leadership measures are either very extensive (Shaw et al., 2011), or combine a multitude of separate scales to capture different dimensions (Aasland et al., 2010; Lundmark et al., 2020), but without exploring the potential existence of a global factor (Larsson et al., 2012). An exception is the measure developed by Thoroughgood et al. (2012). However, they include organizationally directed deviant behaviors (e.g., steeling), as well as sexual harassment behaviors (e.g., “brings inappropriate material to work”). Thus, although including a global factor with distinct subdimensions, it covers a wider range of leaders’ workplace deviance, moving beyond the core of the destructive leadership definition as employe-directed behaviors. The only measure we have found that aligns with an employe-directed approach to destructive leadership, includes both passive and active destructive leadership aspects, is easy to use (i.e., not too long), and includes a global factor with subdimensions is Destrudo-L (Larsson et al., 2012).

Destrudo-L consists of five subdimensions of destructive leadership: arrogant/unfair, threats/punishments/overdemands, ego-oriented/false, passive/cowardly, and uncertain/unclear/messy (Larsson et al., 2012). The five factors are translatable into the labels or dimensions of destructive leadership that other researchers have used in their conceptual work (e.g., Tepper, 2000; Einarsen et al., 2007; Skogstad et al., 2007; Schilling, 2009). The subdimensions arrogant/unfair and threats/punishments/overdemands in Destrudo-L describe behaviors similar to those described by a tyrannical leadership, as suggested by Einarsen et al. (2007). The ego-oriented/false subdimension, on the other hand, resembles what Einarsen et al. (2007) call a derailed leadership. Finally, the passive/cowardly and uncertain/unclear/messy subdimensions consider more passive/indirect leadership behaviors, comparable to the laissez-faire, avoidant, or failed subdimensions of destructive leadership suggested by Skogstad et al. (2007). Thus, this indicates that on a conceptual level, there is substantial overlap between the subdimensions of Destrudo-L and subdimensions identified in literature on destructive leadership (for a full compilation, see Larsson et al., 2012).

There are two important sources of construct-relevant multidimensionality to consider in relation to constructs that are hierarchical, with subdimensions conceptually related to each other (such as the subdimensions of Destrudo-L). First, it is necessary to consider the hierarchical nature, by acknowledging the presence of an overarching global factor, which can be identified using bifactor models (Morin et al., 2016). Second, most items are, to some degree, imperfect and have some systematic association with other constructs, a phenomenon referred to as the fallible nature of indicators (Morin et al., 2016). Therefore, it is typically possible to justify cross-loadings based on substantive theory or item content in multidimensional measures (Asparouhov and Muthén, 2009). In the independent clusters model confirmatory factor analysis (ICM-CFA), cross-loadings are constrained to zero, which often inflates and biases factor correlations and worsens the construct validity of the factor structure (Marsh et al., 2014). In exploratory structural equation modeling (ESEM), all cross-loadings are estimated and assumed to be close to zero but not exactly zero, which takes into account the fallible nature of indicators, reduces factor correlations, and improves discriminant validity (Marsh et al., 2014; Morin et al., 2016). In line with previous recommendations (e.g., Marsh et al., 2014), we compare ICM-CFA models with ESEM models to identify the model that provides the best fit to the data.

As Destrudo-L was developed primarily for use in military settings, neither its psychometrical properties nor its relations with outcomes has been tested in a general work context. However, there have been calls to examine its reliability and validity in a general work context (Larsson et al., 2012). Additionally, there have been no examinations of the convergent and discriminant validity of the Destrudo-L, nor has its suggested factor structure, with one global factor and five subdimensions, ever been replicated.

Beyond finding effective ways to evaluate destructive leadership, there have been calls to find out more about the unique and relative association between different forms of destructive leadership behaviors and employe outcomes (Schmid et al., 2018). To articulate a meaningful multidimensional understanding of destructive leadership, it is important to know how destructive leadership behaviors are expressed in practice and how these expressions relate to outcomes. Such knowledge can help detect and guide interventions on these behaviors in organizations (Schmid et al., 2018). One way to further explore the dimensionality, the co-occurrence of different subdimensions, and the ways that different dimensions of a destructive leadership relate to outcomes is to use person-oriented approaches, such as cluster analysis or latent class/profile analysis (Aasland et al., 2010; Shaw et al., 2011).

Profile analysis is a relatively new concept within destructive leadership research; the studies by Aasland et al. (2010) and Shaw et al. (2011) have led the way. In their study, Aasland et al. (2010) based the latent classes/profiles on their four suggested dimensions of destructive leadership: Tyrannical, Derailed, Supportive-disloyal, and Laissez-faire. They identified six meaningful latent classes, and determined that (except for a non-destructive cluster) classes dominated by laissez-faire were by far most common. In their cluster analysis, Shaw et al. (2011) combined their overall categorization of leaders as bad or not bad leaders with 22 destructive leadership factors. Shaw et al. (2011) found seven meaningful clusters. They found six meaningful clusters using the same 22 factors in a second study with another sample (Shaw et al., 2014). Based on the results of these studies, Shaw et al. (2011, 2014) concluded that most of the leaders could not be described as destructive in an overall sense, but rather in some part (using a few destructive behavioral categories).

In Destrudo-L, it is possible to understand the global factor of destructive leadership as respondents’ generalized “the leader is, or is not, bad” evaluations (Larsson et al., 2012). Therefore, using an exploratory approach similar to that of Shaw et al. (2011), but combining the global factor and the result of the different subscales, can help reveal how destructive leadership is perceived from the employe perspective.

Previous studies using clustering or latent class/profile approaches did not relate the clusters to outcomes (Aasland et al., 2010; Shaw et al., 2011, 2014). Thus, whether and how clustering contributes to our understanding of the relations between destructive leadership and outcomes is still unknown. Shaw et al. (2011) suggested that only in some clusters could the leader been seen as truly destructive, based on the overall judgment of employes. In evaluating the profiles, we therefore suggest that one way to examine the importance of the overall judgment is to determine whether the higher-order factor is decisive for the profiles’ relation with outcomes.

The prevalence of destructive leadership is often used as an argument favoring the study of the phenomenon (e.g., Schyns and Schilling, 2013). However, only one study (Aasland et al., 2010) has examined its prevalence as a multidimensional phenomenon. A few other studies have conducted prevalence research on manager aggression (Hubert and van Veldhoven, 2001; Schat et al., 2006), which, from a multidimensional perspective, can be considered as one aspect of destructive leadership.

Aasland et al. (2010) study established the prevalence of destructive leadership as 83.7% (overall prevalence – exposure of any kind), 33.5% (using operational class method; OCM – cut-off criteria were “often”), and 61% (using latent class cluster; LCC – sum of respondents in destructive clusters) in a Norwegian sample. Schat et al. (2006) studied the overall prevalence of workplace aggression among mangers using six questions in a weighted representative U.S. sample. They found an overall prevalence of 13.5%. Hubert and van Veldhoven (2001) used only one question to evaluate exposure to managers’ aggression, and with an OCM criterion of “sometimes,” rendering a prevalence of 9% in a large Dutch sample. Thus, the subject of study (i.e., what kind of leadership or managerial behaviors), and the methods used to conduct the study, seem to influence the results quite substantially.

Results from a meta-analysis show that constructive (i.e., transformational) leadership is more common in the public sector than in the private sector (Lowe et al., 1996). These results have been explained mainly by the fact that leaders in the public sector are more limited in terms of motivational tools (e.g., rewards), and instead, must rely more on normative appeals and creating an attractive vision (An et al., 2019). However, there has been no such comparison of prevalence rates between the public and private sectors in terms of destructive leadership.

We used Statistics Sweden, a government agency that produces official statistics, to draw a stratified random sample, representative of the Swedish workforce, based on the Swedish Occupational Register in Sweden. We invited 3000 employes to participate in our study, 750 from each of the four sectors in Swedish working life: private, municipal, state, and government. A paper survey, which took around 20 min to complete, was distributed by mail to the respondents’ home addresses. After three reminders, we received 1132 responses, yielding an overall response rate of 39%. Of the participants, 218 were employed in the private sector, 329 by a municipality, 273 by the state, and 312 by the region. The sample had an average age of 48 years, 66% of which were female, and the respondents had an average tenure in their organization of 13 years.

For the prevalence analysis, we used population weights, provided by Statistics Sweden, to estimate responses of a population total (N = 3100282). The population weights compensate for sampling error, non-response error, and coverage error (Lundström and Särndal, 2001). The weights were constructed based on the total number of respondents for the entire survey (N = 1132), however the response rate for the Destrudo-L scale was somewhat lower (n = 1121). Consulting Statistics Sweden, we therefore estimated these missing responses (n = 11 in actual sample, n = 23521 in weighted sample) based on average scores on each item of the total sample. To control for potential differences caused by the method used for estimation, we compared results based on the total weighted sample with results based on the weighted sample without those (n = 11; n = 23521) that were missing. No differences were found on a 1.0% level, and we therefore based all prevalence analysis on the total weighted sample.

Descriptive statistics of the items and omega reliability coefficients are presented in Table 1. All items were positively skewed and had a non-normal response pattern, which shows the importance of using a robust estimator in subsequent analyses. Reliability coefficients ranged from 0.88 to 0.91 and the bootstrap confidence intervals were all narrow, indicating a relatively high precision of the point estimate. Following recommendations in the literature (Marsh et al., 2009b), we first compared the model fit of an ICM-CFA model with that of the ESEM. As seen in Table 2, the ESEM showed a better fit to the data, as indicated by a lower chi-square, higher CFI and TLI, and lower RMSEA and SRMR. The latent factor correlations (see Supplementary Table 1) were lower in the ESEM, ranging from 0.41 to 0.80 (M = 0.62), compared to the ICM-CFA, in which they ranged from 0.66 to 0.90 (M = 0.81). A comparison of the bifactor ICM-CFA and the bifactor ESEM also showed a superior model fit of the bifactor ESEM (Table 2).

To address research question 1, we examined the factor loadings of the bifactor ESEM model (Table 3). The results indicated a well-defined global destructive leadership factor with standardized factor loadings (λ) ranging from 0.69 to 0.88 (M = 0.80). The standardized target factor loadings (λ) for the specific factors were weaker than those of the global factor, but they– still indicated relatively well-defined factors (arrogant/unfair, λ range 0.29–0.45, M = 0.37; threats/punishments/overdemands, λ range 0.25–0.42, M = 0.37; Ego/False, λ range −0.11 to 0.45, M = 0.16; passive/cowardly, λ range 0.23–0.58, M = 0.40; and uncertain/unclear/messy, λ range 0.31–0.52, M = 0.44), with few cross-loadings larger than 0.30. Overlap was most evident between the arrogant/unfair factor and threats/punishments/overdemands factor, where one item per subdimension (i.e., “Is unpleasant”; “Shows violent tendencies”) showed cross-loadings around 0.30. The ego-oriented/false subdimension was the least well-defined of the five subdimensions, with two items (i.e., “Does not trust his/her subordinates”; “Does not keep promises”) showing low and not statistically significant factor loadings on the target factor. To summarize, we were able to replicate the original bifactor structure of Destrudo-L using an ESEM approach, which provided a better fit to the data than the traditional ICM-CFA.

To address research question 2, we estimated latent factor correlations between the destructive leadership factors and the outcome variables (Table 4). We found a clear pattern of associations between the global destructive leadership factor and the outcome variables. We observed positive correlations between global destructive leadership and the negative outcomes (i.e., turnover intentions and burnout), whereas the correlations between global destructive leadership and the positive outcomes (i.e., organizational citizenship behaviors, work-role performance, role clarity, job satisfaction) were negative. The global destructive leadership factor also had a negative correlation with transformational leadership and a positive correlation with abusive supervision. Most of the correlations between the specific destructive leadership factors and the outcome variables were small (rs < 0.30). The latent factor correlations, using a variable-centered approach, supports the convergent validity of the global destructive leadership factor, whereas the correlations with specific factors provide weak support for their convergent validity.

Based on the latent factor scores from the bifactor ESEM, we estimated a series of nested LPA models with increasing complexity (i.e., more profiles) and compared the model fit between these nested models to answer research question 3. Thus, the LPA input variables included the specific subdimensions and the global destructive leadership factor (cf. Morin et al., 2017). As seen in Table 5, the model fit improved for each added class. However, the 7-profile solution was chosen as the final model, because adding a seventh profile provided a theoretically intelligible and meaningful additional class, compared to a model with six profiles. Adding an eighth profile, however, resulted in an arbitrary division of an existing profile into smaller profiles that differed mostly in a quantitative way from each other. Further, the eighth class was small (∼ 4% of the sample) and did not add anything theoretically meaningful.

The following is a description of the seven latent profiles (see Figure 1). Following previous studies (e.g., Gustafsson et al., 2018), the descriptions are based on standardized z-scores and represent standard deviation (SD) units above (i.e., positive values) or below (i.e., negative values) the sample mean (which is 0). We define larger than ± 1 SD as very low/high, ±0.5 to 1.0 SD as low/high, and values from -0.5 to 0.5 SD as slightly below/above average.

Profile 1 was labeled consistently non-destructive and characterized by average to very low scores on all destructive leadership behaviors. Profile 2 was labeled somewhat messy and characterized by low scores on the active behaviors (i.e., arrogant/unfair, threats/punishments/overdemands and ego-oriented/false) and high scores on the specific subdimension uncertain/unclear/messy. Profile 3 was labeled passive and characterized by low scores on the active behaviors and very high scores on the passive behaviors (i.e., passive/cowardly and uncertain/unclear/messy). Profile 4 was labeled unfair-destructive and characterized by high scores on the specific subdimension arrogant/unfair, low scores on the ego-oriented/false factor, as well as a z-score above 0.5 on the global destructive factor. Profile 5 was labeled active destructive and was characterized by very high scores on the active subdimensions arrogant/unfair and threats/punishments/overdemands, and low to very low on the passive forms, with a z-score above 0.5 on the global destructive factor. Profile 6 was labeled ego-oriented passive destructive and was characterized by high scores on the Ego-oriented/False factor and on passive destructive behaviors, as well as low scores on the active subdimensions arrogant/unfair and threats/punishments/overdemands. Z-score for the global destructive factor was again above 0.5. Profile 7 was labeled ego-oriented active destructive and characterized by very high scores (i.e., z-score ≥1.0) on the global destructive factor and high scores on the active behaviors.

Comparisons between the latent profiles on the outcome variables (i.e., turnover intentions, burnout, organizational citizenship behaviors, work-role performance, role clarity, and job satisfaction), to answer research question 4, are presented in Supplementary Table 2 in the Supplementary Material and in Figure 2. Due to the large number of comparisons, the significance level was set to 1%, hence, a p-value lower than 0.01 was considered statistically significant. The omnibus Wald test indicated statistically significant differences between the profiles on all outcome variables (p < 0.005).

Pairwise comparisons showed that the consistently non-destructive profile (Profile 1) reported lower turnover intentions and burnout scores compared to all other profiles except the somewhat messy profile (Profile 2). Profile 1 also reported higher levels of organizational citizenship behaviors compared to the messy (Profile 2), ego-oriented passive destructive (Profile 6), and ego-oriented active destructive (Profile 7) profiles, higher work-role performance than the unfair destructive profile (Profile 4), and higher role clarity and job satisfaction compared to all other profiles.

The somewhat messy profile (Profile 2) reported lower turnover intentions and burnout scores than the passive (Profile 3), unfair destructive (Profile 4), active destructive (Profile 5), ego-oriented passive destructive (profile 6), and the ego-oriented active destructive (Profile 7) profiles. The somewhat messy profile (Profile 2) also reported higher work-role performance than the unfair destructive profile (Profile 4), higher role clarity than the ego-oriented passive destructive (Profile 6), and higher job satisfaction than the active destructive (Profile 4), ego-oriented passive destructive (Profile 6), and the ego-oriented active destructive (Profile 7) profiles. No other statistically significant differences were observed between the profiles.

The consistently non-destructive profile (Profile 1) showed the most favorable profile in relation to the outcomes variables, closely followed by the somewhat messy profile (Profile 2). Both of these profiles were characterized by relatively low levels of destructive leadership. The remaining profiles, particularly Profiles 4 to 7, reported high levels of destructive leadership, and these profiles also showed a more negative pattern on the outcome variables.

To address research question 5, we estimated the prevalence of destructive leadership behaviors reported by the weighted sample, both in numbers and percent (Table 6). For item-level prevalence and prevalence of behaviors exposed to for each subdimension using the OCM criterion “often,” see Supplementary Tables 3, 4 in the Supplementary Material. In total, 90.2% of the weighted Swedish workforce sample reported exposure to destructive leadership behaviors of any kind. Using the OCM criterion of “often” or more frequently, we found that 36.4% of our sample was exposed to at least one destructive leadership behavior. The prevalence of the different types of destructive leadership behaviors in our sample varied widely: 8.7% for threats/punishments/overdemands, 16.9% for ego-oriented/false, 21.0% for arrogant/unfair, 23.5% for passive/cowardly, and 23.9% for uncertain/unclear/messy.

Prevalence based on the results of the above presented LPA (Figure 1), with the global destructive leadership factor as an indicator of the leader being perceived as a destructive leader, yields somewhat different results. Using population weights: Profile 1 represented 34.9% (N = 1082047); Profile 2 represented 13.7% (N = 424952); Profile 3 represented 7.9% (N = 245521); Profile 4 represented 8.2% (N = 252537); Profile 5 represented 5.5% (N = 171685); Profile 6 represented 10.1% (N = 312314); and Profile 7 represented 19.7% (N = 611226). Thus, 23.8%, representing Profiles 4, 5, and 6, were considered as exposed to a moderately frequent (z-value above 0.5) destructive leader, and 19.7%, representing Profile 7, were considered as exposed to a highly frequent (z-value above 1.0) destructive leader. In total, 43.5% of the weighted sample was exposed to a moderate to highly frequent destructive leader. The association between the profiles and outcomes also show that these four profiles are most clearly associated with adverse outcomes (Figure 1).

Finally, to answer research question 6, we compared the private and public sectors (see Table 7), using the OCM criterion of “often” or more frequently, and found that exposure to destructive leadership behaviors of any kind was more common in the public sector (38.2%, compared to 35.1%, χ² = 4727,95, df = 1, p < 0.001). Looking at the different types of destructive leadership behaviors, more active/direct forms were somewhat more frequent in the public sector, whereas passive/indirect destructive leadership was more frequent in the private sector.

The findings are relevant to organizations in a number of ways. First, our results offer additional evidence in support of the reliability and validity of Destrudo-L, suggesting that it can be used to assess employes’ perceptions of destructive leadership behaviors in a general work context. Second, our study demonstrates that destructive leadership is prevalent in all sectors of working life, and is therefore a potential problem that all organizations must address. Identifying destructive leaders, however, is not enough and in line with a framework recently developed by Pradhan and Jena (2018), we suggest that organizations take a broad approach in their efforts to reduce the prevalence of destructive leadership behaviors in organizations. This approach includes primary interventions, such as tailoring the recruitment and selection process to facilitate identification of personality traits that are associated with destructive leadership are identified. In addition, HR policies that reflect zero tolerance for abuse, as well as orientation programs in which these policies are explained to new leaders may be implemented. Secondary interventions include ensuring that the climate in the organization does not tolerate mistreatment or avoidance, as well as strategies to identify destructive leadership behaviors in the organizations, such as the inclusion of Destrudo-L in employes’ surveys, or establishing a HR hotline or other such complaint mechanisms. Finally, tertiary interventions may consist of counseling, problem solving and stress management, leadership training, withdrawal of benefits, and if nothing else works, termination. Intervention may also target employes in different ways, mitigating the negative effects of destructive leaders.

This study is not without limitations. First, the data is cross-sectional. Although this is less of a problem in the context of investigating the factor structure (RQ1), profiles (RQ3), and prevalence (RQ5 and RQ6), it may be an issue for the interpretation of relations between destructive leadership and outcomes (RQ2 and RQ4). Particularly, Skogstad et al. (2014) found that, although active destructive leadership was worse than passive destructive leadership, in the short term (6 months), passive destructive leadership may build up to cause more adverse outcome over a longer period of time. Thus, we interpret the relations to outcomes with caution, primarily using them to enhance our understanding of the construct of destructive leadership. We recommend that future studies use experimental and longitudinal designs to examine the relations between destructive leadership and employe outcomes.

The Destrudo-L uses a 6-point response scale ranging from 1 (never/almost never) to 6 (always). This is in line with the ambition to capture the respondents’ general perception of the frequency stimuli. However, it also means that there is room for multiple interpretation of what constitutes “always” or “sometimes,” and the scale does not provide information on severity. There is a need for future studies that investigate how people interpret different response categories. Testing and implementing response scales that captures frequency as well as perceived severity is an additional opportunity in future studies.

This study relies on self-ratings, with destructive leadership being rated by employes. Although we used employes’ experience of their leaders, not leaders’ own ratings of their leadership behaviors, there is a risk of common-source bias, primarily affecting RQ2 and RQ4, which investigate relations between destructive leadership and outcomes. Using others’ ratings of leadership, such as employes’ ratings of their leader, as has been done here, is usually considered an advantage over self-ratings, as it minimizes the risk of self-serving bias. However, although ratings of others reflect the behavior of the rated, they may also be colored by the rater’s expectations, as well as by how much information they have on which base their judgment (Lornudd, 2015). In line with this, different raters have different experiences of the person they rate, and thus, additional raters have been shown to add to the variance explained (Oh and Berry, 2009). Thus, future studies, using multiple sources (e.g., supervisor-, employe- and self-ratings), are warranted.

The seven profiles are data-driven and exploratory, and must be replicated in future studies. This includes future studies to illuminate whether the profiles are representative for other working populations, which profiles may be considered primary or peripheral, and whether the profiles are stable over time. The culture in North/Western Europe differs somewhat from South/Eastern Europe and societies in other parts of the world. The difference is mainly found in terms of a lower power distance (i.e., the extent to which power in a society is equally/unequally shared), which in turn is reflected in the disapproval or endorsement of an autocratic leadership style (Koopman et al., 1999). Thus, the destructiveness of certain leadership behaviors may be interpreted in different ways depending on the specific societal culture where it is performed. Future studies should therefore strive to examine perceptions of destructive leadership in country contexts beyond North/Western Europe to advance our understanding of cross-cultural aspects of destructive leadership. We also encourage future studies where different subpopulations (e.g., different branches of work, older and younger employes) are compared.