Cross-Cultural Examination of the Bifactor Structure and Network Invariance of Dark Triad Items Across Four Countries

In recent decades, the dark triad of personality has captured the attention of both researchers and the general public. Since the conception of the construct by Paulhus and Williams (2002), it has been intensively studied in thousands of academic publications (Dinić & Jevremov, 2021). The Dark Triad of personality is composed of a pattern of traits that have been categorized as socially aversive: subclinical psychopathy, subclinical narcissism, and Machiavellianism. These traits tend to occur together as they are characterized by a disregard for social norms and an interpersonal orientation to take advantage of others, albeit with differentiated motivations and strategies (Vize et al., 2018). Likewise, people with these traits report less agreeableness and emotional closeness towards others’ needs, and have even been associated with aggressive behaviors (Muris et al., 2017), which represents a set of detrimental behaviors in human relationships (Moraga, 2015).

Machiavellianism is related to a strategic interpersonal behavior that disregards ethical norms and adapt to the demands of the context in order to exploit others and achieve personal gain (Koehn et al., 2019). It consists of two components: manipulating others and having a cynical view of human nature. Those with this trait tend to be analytical and calculating in social relationships and may use deception to achieve their goals (Muris et al., 2017).

Narcissism is characterized by arrogance, grandiosity, entitlement and an idealized self-perception of superior abilities (Krizan & Herlache, 2018). Generally, it can be described by three perspectives: a self-description of uniqueness, the imposition of one's needs on others, and preserving one's image (Campbell & Miller, 2011). Narcissistic individuals may be sensitive to criticism and feel the need to validate their desires and ways of seeing the world to strengthen their beliefs of being special or superior to others (Krizan & Herlache, 2018).

Psychopathy is a set of interpersonal, affective and behavioral components including insincerity, lack of empathy, and impulsivity (Jones & Paulhus, 2014). In secondary psychopathy, this is also associated with risky, emotionally unstable and criminal behavior (Yildirim & Derksen, 2015). In contrast, subclinical psychopathy involves superficial relationships for personal advantage and a lack of empathy when harming others in various contexts of domestic, academic, or work life (LeBreton et al., 2006). These individuals often use their understanding of interpersonal relationships to control and act selfishly due to their lack of emotional understanding, and may be charming and skilled at manipulating others, especially seeking out the weaknesses of others (Wai & Tiliopoulos, 2012).

Subclinical aversive behavior, characterized by maximizing self-interest at the expense of others, is estimated to be highly prevalent globally (LeBreton et al., 2018; Sanz-García et al., 2021) in different age groups, especially in young people (Barlett & Barlett, 2015; Zettler et al., 2021), and is manifested in everyday life contexts such as the pursuit of multiple partners, academic dishonesty, and unethical work behaviors (Moraga, 2015; Muris et al., 2017). These traits are likely to be common even in socioeconomic environments of stability or uncertainty, favoring competitiveness for power and hoarding available resources (Jonason et al., 2019; 2020). In this way, they may manifest in particular ways as they identify a context of perceived interpersonal benefit or utility (Moshagen et al., 2018).

Although they have been extensively studied, it is important to examine the precision of the measurement of dark personality traits in diverse societies and cultures. From the conception of this model, there is a greater predominance of studies in contexts such as the United States and European countries (Jonason & Webster, 2010; Jones & Figueredo, 2013; Maneiro et al., 2019; Pechorro et al., 2019), societies that stand out for a higher level of individualism, which may favor certain selfish or aversive behaviors (Jonason et al., 2020). However, it is possible that dark traits are also particularly manifest in collectivistic societies, and are even perceived as acceptable behaviors in the face of the group demands of these environments (Jonason et al., 2019; Robertson et al., 2016). Therefore, although there are recent studies in this regard in societies culturally different from the West such as South America or Asia (Bonfá-Araujo et al., 2021; Kawamoto et al., 2020; Özsoy et al., 2017), there is still a need to continue this direction of research in order to validate instruments and detect similarities or differences in measuring the aforementioned attributes. Furthermore, this cross-cultural examination is relevant in the context of the recent pandemic, where behaviors prioritizing personal benefit at the expense of others have had negative impacts on public, mental, and socio-political health across different continents, revealing that these patterns are broad and emerge in different situations (Farzanegan & Hofmann, 2021; Zettler et al., 2022; Zirenko et al., 2021).

Several combined measures of the dark triad (e.g. Dirty Dozen, Short Dark Triad) have been examined worldwide, which evidence a dimensional structure that can be summarized as a positive covariance between subclinical Machiavellianism, psychopathy and narcissism, according to two meta-analytic papers (Muris et al., 2017; Vize et al., 2018). However, a basic requirement for comparing instrument scores according to attributes such as gender or cultural context is to analyze measurement invariance. Invariance is relevant, as it ensures the comparability of measurements across different groups or contexts, which is essential for making valid and generalizable inferences (Milfont & Fischer, 2010). In contrast, the lack of evidence of invariance suggests that the observed differences could be attributed to variations in instrument characteristics rather than actual divergences in the psychological constructs measured. These variables are often inherently complex and are not always equivalent in people of different cultures and ages (Dong & Dumas, 2020). In the analysis of differences in complex variables such as personality, a more detailed and beneficial invariance methodology, at the item level, is network analysis. Network invariance has several advantages, including the examination of the distribution of items in the overall structural network, differences between specific associations of items, and the importance of the items that interconnect and sustain the entire network (van Borkulo et al., 2023). Therefore, these analyses seem more appropriate when assessing dark personality traits, whose best-known scales represent indicators of very similar traits and others that are more differentiated, even across cultures (Maples et al., 2014; Luo et al., 2023; Vize et al., 2018).

While previous studies have examined the psychometric analysis of dark triad instruments at the multicultural level (Aluja et al., 2022; Jonason et al., 2020), few works have combined multiple measures to verify the feasibility of alternative models and provide reliable interpretation for comparative precision in this type of work. In previous studies, different models were estimated to analyze whether the variety of dark personality traits cluster into a single aversive personality factor (e.g., Factor D, Moshagen et al., 2018). For this purpose, the two best-known dark triad measures (e.g., Dirty Dozen and Short Dark Triad; Jonason & Webster, 2010; Jones & Paulhus, 2014) were considered. Among the antecedents referring to general factor models, Moshagen et al. (2020) examined at least 12 combined measures of aversive traits such as psychopathy, Machiavellianism, narcissism, sadism, psychological arrogance and others, where all items loaded on a common core factor, which appeared to represent the core features of the dark personality. Similarly, Volmer et al. (2019) evaluated correlated factor models and an overall factor, using items from a set of dark triad instruments, including the two best-known scales (e.g. Short Dark Triad., Dirty Dozen). The results suggested that it is possible to obtain a better fit of the data using a general dark factor representation underlying other more specific traits. Also, other previous research using exploratory graph network analysis (EGA) methodology demonstrated that it is possible to model dark triad trait clusters to obtain dimensions as well as a higher order factor using long and short triad measures (Truhan et al., 2021).

Thus, the literature suggests that the conceptualization of a general dark factor is feasible through quantitative modeling that captures the common attributes of socially aversive traits, beyond their unique or particular characteristics (Moshagen et al., 2018; Zettler et al., 2021). Following these findings, it is important to examine whether the integration of available measures of dark triad can provide support for the conceptualization of a dark general personality factor also in cultural contexts as differentiated as South America or Europe. Therefore, the present research aimed to evaluate the proposed model of a general dark trait tendency based on the bifactor model and to analyze the structure and network difference of the most representative measures of dark triad at the multicultural level. This is important as the alternative models need to be replicated further to test their extension to diversified international contexts, which have not yet been adequately represented.

Method

Results

Regarding descriptive data by country, the USA consisted of participants from the general population (58% women, n = 682, mean age = 39.1, SD = 11.9); Germany examined university students and workers (61% women, n = 477, mean age: 23.1, SD = 5.9); in Serbia, university students and general population (50% female, n = 221, mean age: 28.1, SD = 6.6) and in Peru university participants were recruited (61% female, n = 190, mean age: 24.5, SD = 4.8).

On the other hand, the Short Dark Triad (SDT) questionnaire was parceled to 9 variables for better fit in models’ estimation. The bifactor structural equation model (Bifactor SEM) of the dark triad in both questionnaires was evaluated (Figure 1, Table 1). All models had factorial solutions. The general model fit indices (with all countries) showed good values (CFI = 0.959; RMSEA = 0.071); while in its reliability the general hierarchical omega was (G = 0.862), and the omega indices of the dimensions, Machiavellianism (M = 0.868); Psychopathy (P = 0.836), and Narcissism (N = 0.828).

Click to enlarge

Figure 1

Bifactor SEM of Dark Triad Scales

Note. G = general factor of dark personality traits; P = psychopathy; N = narcissism; M = Machiavellianism.

In models by countries, US (CFI = 0.964; RMSEA = 0.070; G = 0.857); Germany (CFI = 0.947; RMSEA = 0.077; G = 0.832), and Serbia (CFI = 0.928; RMSEA = 0.083; G = 0.849) obtained good fit indices and reliability indices, while Peru model showed good fit indices but low omega indices in the Psychopathy dimension (CFI = 0.914; RMSEA = 0.078; G = 0.800), all detailed results can be found at Table 1.

Then, we performed a network analysis (Figure 2), where SDT1 and SDT2 (r = 0.291); SDT4 and SDT5 (r = 0.354); and Machiavellianism (MAQ) and Psychopathy (PSI; r = 0.299) showed the highest relationships. The clusters were explored with the Spinglass algorithm. In addition, the nodes with highest predictability were Machiavellianism (MAQ; r² = 0.52), SDT2 (0.46), and Psychopathy (PSI; r² = 0.45), while SDT3 (r² = 0.38), SDT5 (r² = 0.37), and SDT9 (r² = 0.30) were the lowest. In the network accuracy and stability through 5000 bootstraps the most of estimated edge were greater than zero and not overlap with other edges; the edges are maintained even after removing large proportions from the sample, the Coefficient of Stability (CS = 0.75) demonstrates an adequate stability of the edges in the network.

Click to enlarge

Figure 2

Network Plot of Dark Triad Measures

Note. Blue paths are positive relationships, Green Cluster = Narcissism SDT; Turquoise Cluster = Psychopathy SDT; Orange Cluster = Machiavellianism SDT; Yellow Cluster = Dirty Dozen Scale.

In the centrality indices of the Dark Triad network analyzed (Figure 3), the Bridge-expected centrality (i.e., a variant of centrality that takes communities of nodes into account), and expected influence (i.e., the sum of raw values of edge weights connected to each node) for each node showed that Machiavellianism (MAQ) had the highest bridge expected influence in the network, while SDT5 was the lowest.

Click to enlarge

Figure 3

Centrality Plots (z-scores) of Dark Triad Network

Note. The first two graphs refer to the centrality of expected influence, while the last two refer to the centrality of bridge-expected influence. Values greater than 1 indicate higher centrality.

In the network analysis by country (Figure 4), all networks presented positive relationships, Peru, Serbia and Germany had the highest predictability (r²) in Machiavellianism (Peru = 0.316; Serbia = 0.397; Germany = 0.505), while the USA showed Psychopathy (PSI) as highest (r² = 0.612). In addition, the centrality indices of USA, Germany and Serbia shows a consensus in Machiavellianism (MAQ) as the highest value (US = 1.738; GE = 2.461; SE = 1.834), and SDT9 (US = -1.628; GE = -1.267; SE = -1.267) as the lowest; while in Peru, SDT2 has the highest value (EI = 2.390) and SDT6 the lowest (EI = -1.053).

Click to enlarge

Figure 4

The Dark Triad Network by Country and Centralities

Note. Blue paths are positive relationships. Green Cluster = Narcissism SDT; Turquoise Cluster = Psychopathy SDT; Orange Cluster = Machiavellianism SDT; Yellow Cluster = Dirty Dozen Scale. Values greater than 1 indicate higher centrality. In the centrality plot Germany = red line; Peru = Green line; Serbia = Blue line; USA = Purple line.

Finally, a network invariance between countries was conducted (Table 2), this confirms that there are significant differences in the edges of all networks, except for Germany and Serbia (p = 0.85) that showed an insignificant value. While in the centrality indices compared by the Expected Influence, can observe that all networks have significant differences, except for USA and Serbia (p = 0.08), and Germany and Serbia (p = 0.63).

Discussion

Previous research has often examined the Short Dark Triad (SDT) and the Dirty Dozen (DD) scales using a three-factor model, but this model has not always been supported in different contexts (McLarnon & Tarraf, 2017; Vize et al., 2018). This may be because the traits being measured share both common and specific variances, highlighting the need for a greater focus on the complexity of these traits and their methodological analyses (Bonfá-Araujo et al., 2021; Rogoza et al., 2021). By separating the common and specific variance, it may be possible to better understand the individual traits and the more universal attributes that have not been adequately considered in previous models. This study aimed to address this limitation and assess the feasibility of this approach as shown by previous studies (Jones & Figueredo, 2013; Moshagen et al., 2018).

To evaluate both the general and specific factors, in the present study we used a bifactor model which provided an optimal fit to the data, similar to other studies that used a variety of measures of dark traits (Gamache et al., 2018; Moshagen et al., 2020) with the difference that this study tested a model across general and specific countries. This generalized factor model explained part of the variance of all items and captures unique variance within subdomain-specific groups of items. This allows for interpretation of two conjoint measures and comparison of differences at the general or factor level (Chen et al., 2006; Reise et al., 2010). In this way, our results found that it is possible to use two measures to adopt the model of a single aversive disposition that can give rise to other specific traits, even when these traits have often been expressed in complex ways across different cultures and contexts (Moshagen et al., 2018).

By transferring this concept to statistic modeling, Moshagen et al. (2020) and Volmer et al. (2019) found that a set of items from different measures of dark traits were strongly correlated and could be combined to form a single aversive personality factor (e.g., Factor D, Moshagen et al., 2020). Moreover, there is evidence that this approach can be used to measure dark triad traits in both Western (Gamache et al., 2018; Jonason & Luévano, 2013) and South American samples (Bonfá-Araujo et al., 2021; Copez Lonzoy et al., 2020). We therefore build on these findings to test the generalization and subsequent comparison of that model in diverse contexts, including underrepresented regions, such as South America (Vitriol et al., 2020).

Additionally, we examined the conceptualization of dark traits in four countries using a network model, that allowed us to identify strongly connected traits and specific subgroups (Epskamp et al., 2018). We used partial correlations to examine these associations because they are best suited for variables with similar interpersonal orientation characteristics, such as the dark triad. This methodology also allows us to verify the contribution and independent conditional relationships between variables (Waldorp & Marsman, 2021).

The overall network structure evidenced a stable and unique organization of medium and high positive partial correlations between the two dark triad measures, suggesting dense interconnections of the same socially aversive nature. Likewise, the Machiavellianism domain of the Dirty Dozen scale was found to be the highest centrality measure, with the highest predictability and the most consistent across all estimated network indices (e.g. expected influence and bridge-expected influence). This may indicate that it is the core interconnection in the systematic representation of the dark factor with the measures used. In addition, this finding may imply a broad nature of manipulative behavior that is common among several traits that maximize personal benefit (Jones & Figueredo, 2013; Vize et al., 2020).

Another research found that a very similar trait such as interpersonal manipulation was highly correlated with Machiavellian dominance on the Dirty Dozen scale (Truhan et al., 2021), and at the same time was a central measure in other network studies in different countries, suggesting an important "dark core" emerging from the dark triad (Marcus et al., 2018; Wehner et al., 2021). Previous studies have noted that strategic and personal benefit-oriented interpersonal tendencies tend to occur across a wide number of countries, in nonclinical populations, university, and organizational settings (Hussain et al., 2021; Jonason et al., 2020). Thus, these social patterns can manifest opportunistically and adaptively to take advantage of others, even in differentiated individualistic contexts such as North America and Europe; or collectivistic such as South America (Robertson et al., 2016).

That is, an important manipulative feature of individuals with high levels of dark triad may involve a base of behaviors focused on violating established norms in an environment and the will of others to achieve personal desires relatively easily. However, their strategies may vary depending on the specific motivations of each dark trait (Vize et al., 2018). For example, it is necessary to distinguish between impulsivity leading to violence in psychopathy (Paulhus, 2014), Machiavellianism using less aggressive and more calculative strategies for power (Koehn et al., 2019), while narcissists are often motivated by obtaining attention and special treatment from others so they could use flattery and temporary affection to maintain superiority over potential victims (Chatterjee & Pollock, 2017).

On the other hand, to understand how cultural environments may influence the use of malevolent behavior, we conducted a cross-cultural network comparison to examine differences in the structure and centrality of connections between dark traits. Although other studies addressed the model of a general dark factor (Moshagen et al., 2020; Zettler et al., 2021), no multicultural-level invariance analyses were performed, which hinders a reliable assessment of the representativeness and stability of a construct's measurement in heterogeneous contexts (Milfont & Fischer, 2010), as proposed in the generalizable concept of dark traits.

In the present study we found differences in dark triad network structures between the North American and European sample (US-GE; US-SE); also, between Peruvian and European participants (GE-PE; SE-PE), while the characteristics of the German and Serbian sample (GE-SE) resulted in similar patterns of connections. These findings suggest that the expression and reinforcement of aversive dispositions may vary based on the proximity of countries and their cultural contexts (e.g. continents).

In Germany and Serbia, cultural patterns promote conformity and restrictive behaviors (Hofstede, 2022). The Machiavellian trait of the Dirty Dozen scale was the most central in these countries, reinforced by some psychopathic characteristics of the SDT scale. A previous study in 49 countries also suggests manipulative behaviors are more likely to thrive in advanced or restrained societies (Jonason et al., 2020). In Germany, such behaviors may be used to achieve professional success and increase social status and self-esteem (Schmitt et al., 2020; Spurk et al., 2016). In Serbia, social tactics focused on obedience and work responsibility may be promoted to achieve prestige and better resources, and some people may have traditional and rigid views of partner roles, leading others to accept greater control (Lavrič et al., 2019). Normative restrictions may also lead some individuals with selfish motivations to engage in calculated behaviors with the intention of harming others or rebelling against authority.

By contrast, in the American continent, many students and adults in general report a greater tendency toward instantaneous satisfaction of affective-interpersonal needs, attachment to established traditions, and resistance to social change (Hofstede, 2022). In both networks in this continent, there were differences related to a greater number of connections to the Machiavellian components of the Dirty Dozen scale in the US, and the SDT scale in the Peruvian environment. This may indicate that most dark triad traits were associated with broader manipulation characteristics in the North American context. In this environment, there is a greater emphasis on individual freedoms, short-term goals, and competitiveness (Hofstede, 2022), which can lead to manipulative behaviors in a wide variety of everyday life contexts, such as a greater need for non-monogamous relationships that require less effort and commitment (Muris et al., 2017). On the other hand, in the workplace, some managers may be authoritarian and unconcerned with the well-being of employees, while some employees may resort to unethical practices to stand out and gain benefits (Kisamore et al., 2010).

On the other hand, in the Peruvian network, strategic and calculating interpersonal exchange characteristics were more relevant in activating other dark traits. This is coherent in a context characterized by power inequality, collectivism, and low compliance with rules (Hofstede, 2022). In this way, many individuals find these patterns appealing to gain advantages in various situations. Even during the recent pandemic, manipulative tactics were observed, where some people in Peru demonstrated little empathy by resorting to bribery to access scarce resources and violate health rules (Muñoz, 2021). In addition, considering the sample of students, during this period, academic dishonesty may have been normalized as a way to succeed in university (Valdivia et al., 2020).