- Email: [email protected]
Machiavellianism has a dimensional latent structure: Results from taxometric analyses
Personality and Individual Differences 113 (2017) 57–62
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
Machiavellianism has a dimensional latent structure: Results from taxometric analyses☆ Johannes Beller ⁎,1, Stefanie Bosse Braunschweig University of Technology, Germany
a r t i c l e
i n f o
Article history: Received 8 January 2017 Received in revised form 20 February 2017 Accepted 7 March 2017 Available online xxxx Keywords: Taxometric analysis Machiavellianism MACH-IV Dirty Dozen Dark triad
a b s t r a c t Despite the importance of Machiavellianism, no study has examined the basic issue of its latent status: Is Machiavellianism a dimensional or a categorical construct? Or equivalently, do people differ in the extent to which they are Machiavellian, or do Machiavellianists differ categorically from non-Machiavellianists? To answer these questions, we analyzed two large online samples of N1 = 10,918 participants who completed the MACH-IV questionnaire and N2 = 40,265 participants who completed the Machiavellianism subscale of the Dirty Dozen questionnaire. Via taxometric methods, we found that Machiavellianism encompasses quantitative rather than qualitative differences in both samples. Hence, people differ quantitatively to the extent to which they are Machiavellian. These findings have important practical and theoretical implications regarding assessment, classification, causality, and labeling. © 2017 Elsevier Ltd. All rights reserved.
1. Introduction The idea of Machiavellianism dates back to the Italian philosopher and politician Niccolò Machiavelli. According to the theme “the ends justify the means”, Machiavelli argued that moral aspects should be disregarded in favor of effectiveness and power (Christie & Geis, 1970). Four centuries later, psychologists have begun studying Machiavellianism as a trait. Individuals scoring high in Machiavellianism—often referred to as Machs—manipulate, control and exploit others to further their own intrinsically motivated goals (Jones & Paulhus, 2009; Paulhus & Williams, 2002). Consequently, Machiavellianism predicts a wide range of real-world anti-social and malevolent behavior (e.g. Baughman, Dearing, Giammarco, & Vernon, 2012; Chabrol, Van Leeuwen, Rodgers, & Séjourné, 2009; O'Boyle, Forsyth, Banks, & McDaniel, 2012) and correlates with mental health issues such as alexithymia and interpersonal maladjustment (Hawley, 2006). At the same time, Machiavellianism is also related to positive outcomes: Machs are more successful on the job (Jones & Paulhus, 2009), and individuals scoring high in Machiavellianism are seen as more effective, desirable and charismatic leaders (Deluga, 2001). Recent evolutionary research has even argued that the “Machiavellian Intelligence” of being able to manipulate our social environment enabled the
☆ We would like to thank Eric Jorgenson for sharing his data with us. ⁎ Corresponding author at: Department of Developmental, Personality and Forensic Psychology, TU Braunschweig, Humboldtstraße 33, 38106 Braunschweig, Germany. E-mail address: [email protected] (J. Beller). 1 Johannes Beller is supported by the German National Academic Foundation.
http://dx.doi.org/10.1016/j.paid.2017.03.014 0191-8869/© 2017 Elsevier Ltd. All rights reserved.
evolutionary success of Homo sapiens as a species (Maestripieri, 2007). Thus Machiavellianism represents an important psychological construct. 1.1. The latent status of Machiavellianism One central issue for every latent construct (like Machiavellianism) is its latent nature: Is the construct best described as a continuum or as qualitative different categories? Answering this question bears several important implications regarding, inter alia, classification, assessment, causality and labeling (Ruscio, Haslam, & Ruscio, 2006). First, the latent status of a construct is important in classifying individuals. If the underlying construct is continuous rather than categorical—as implied by the practice of forming sum scores—then, any classification in dichotomous groups needs to be considered very carefully and the whole process of classifying individuals based on sum-scores might be questioned. On the other hand, if a true categorical latent structure exists, taxometric analysis can provide and justify different cut scores. Second, information about the latent status is important for the development of assessment procedures. If the latent structure is categorical one would focus on using items which maximally discriminate between groups. Contrary, if the construct under consideration was continuous one would need to include items over the whole spectrum of the latent continuum. Third, the latent status is important in labeling. Whether a construct is communicated as categorical or dimensional in nature, impacts the perceptions, attitudes and behavior of the lay public (Prentice & Miller, 2007). For example, viewing Machiavellianism as categorical—as implied by researchers using the terms “Machs” and “Non-Machs”—may lead the lay public to construe Machiavellianism
58
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
as more stable and resistant to change than if it were viewed as dimensional. These beliefs may in turn shape the way members of this supposed category are perceived, approached and evaluated (Prentice & Miller, 2007). Fourth, the latent status might further theoretical insights about the construct, its antecendents and consequences (e.g. Beller & Baier, 2013). Ruscio et al. (2006), for example, suggest that categorical latent constructs might result from specific etiological factors, threshold effects, nonlinear interactions or developmental bifurcation, while dimensional constructs tend to result from numerous additive influences. Thus, analyzing the latent status of Machiavellianism may provide important insights. Due to these possible important insights numerous taxometric studies have explored the latent status of diverse psychological constructs (see e.g. Kliem et al., 2014; Beller & Kröger, 2016; or for Machiavellism related constructs: Edens, Marcus, Lilienfeld, & Poythress, 2006; Foster & Campbell, 2007). However, no such study exists for Machiavellianism. In the literature Machiavellianism has largely been considered to be a dimensional construct per fiat, but evidence might also suggest Machiavellian categories for at least three reasons: First, evolutionary approaches to Machiavellianism have conceptualized human populations as a mixture of cooperators (or Non-Machs) and exploiters (or Machs) (e.g. Maestripieri, 2007; Wilson, Near, & Miller, 1998). In the same vein, a large part of the vast literature on game theory has been concerned with analyzing categorically distinct cooperative and conflictive strategies, including deception (e.g. Brown, Garwood, & Williamson, 2012; Ettinger & Jehiel, 2010). Therefore it could be argued that these evolutionary strategies might constitute latent categories of Machiavellianism. Second, Machiavellianism has been shown to result mostly from environmental effects, but also from genetics (heritability factor of 0.31; Vernon, Villani, Vickers, & Harris, 2008). Following the aforementioned argument by Ruscio et al. (2006), Machiavellianism could start out as a dimensional construct, but might become a qualitatively different state only when specific genetic and environment or personality factors interact in a non-linear way, thus forming a categorical Mach condition. Third, the Machiavellianism test scores are often dichotomized. For example, Verbeke et al. (2011) suggest that a cut off score might be used to divide the participants into low and high Machiavellians. Other studies have used further strategies like a median split or a certain sum score range to classify participants in distinct Machiavellianism categories (e.g. Porter, Bhanwer, Woodworth, & Black, 2014; Bereczkei & Czibor, 2014; Láng & Birkás, 2014). Thus evolutionary theories and research practices might support a categorical view of Machiavellianism. But despite the importance of Machiavellianism no study exists in which the latent status of Machiavellianism is empirically determined. The current study strives to fill this gap. 1.2. Current study The current study contributes to the literature by clarifying whether the latent status of Machiavellianism is dimensional or categorical. Clarifying the latent status might provide important insights regarding classification, assessment, causality and labeling. Towards this end three non-redundant taxometric methods are applied to two large online samples (N1 = 10,918, N2 = 40,265). We ask: Is the latent status of Machiavellianism categorical or dimensional? 2. Method 2.1. Participants and procedure 2.1.1. Sample 1 Data were collected via an online survey provided on www. personality-testing.org. Participants answered questions regarding the MACH-IV questionnaire and demographic data (N = 13.156). Data collection began in January 2012 and ended June 2012. All participants explicitly agreed that their data might be used for scientific analyses. Prior
to the analyses we removed participants who indicated that they were younger than 18 or older than 80 years. After additionally deleting all participants with missing values on the MACH-IV scale a final sample size of 11,702 participants (65.7% male) ranging in age from 18 to 80 (M = 30.79, SD = 11.41) was obtained. 2.1.2. Sample 2 Data were also collected via an online survey from www. personality-testing.org. Participants answered questions regarding the Dirty Dozen questionnaire and demographic data (N = 53.981). Data collection began in July 2012 and ended in December 2013. Only participants who agreed that their data might be used for further scientific analyses were included in the sample. Additionally, as in the first sample, all participants who indicated that they were younger than 18 or older than 80 years old were removed prior to analyses. After deleting all missing values regarding the Machiavellianism subscale of the Dirty Dozen questionnaire, a final sample size of N2 = 40,165 participants (65.15% male) ranging in age from 18 to 80 (M = 28.12, SD = 10.53) was obtained (the results reported in this study do not change significantly when no participants are excluded in the analyses). 2.2. Measures 2.2.1. MACH-IV The 20 item MACH-IV scale by Christie and Geis (1970) has been the most widely used instrument to measure Machiavellianism (Jones & Paulhus, 2009). In previous studies, the Machiavellianism scale test scores showed good psychometric properties with acceptable internal consistencies (e.g. α = 0.71; Christie & Geis, 1970). Previous studies, however, differed in the proposed factor structure of the MACH-IV scale (Rauthmann, 2013). Originally, the MACH-IV was designed to encompass three sub-scales (interpersonal tactics, cynical view of human nature, disregard for conventional morality), which might be combined to form a total score. Subsequent studies challenged this factor structure (Rauthmann, 2013, for an overview). For example, Calvete and Corral (2000) found via confirmatory factor analyses that a four-factor structure (positive interpersonal tactics, negative tactics, positive view of human nature, cynical view of human nature) best fitted their data. Example items of the MACH-IV scale include “Never tell anyone the real reason you did something unless it is useful to do so”, “Most people are basically good and kind” (R) and “It is hard to get ahead without cutting corners here and there”. Participants responded on a five point scale (coded from 1 to 5), with reversed items recoded so that higher values represented stronger Machiavellianism believes. Additionally, a short version of the MACH-IV, the MACH* has recently been developed based on item response theory (Rauthmann, 2013). Thus, because of the widespread use, high validity and reliability of the MACH-IV scale scores, it seems tenable to use the MACH-IV to examine the latent structure of Machiavellianism. 2.2.2. Dirty Dozen The 12 item Dirty Dozen questionnaire has been a popular measure of the Dark Triad (Jonason & Webster, 2010) and thus Machiavellianism. Despite its conciseness with 4 items, the Machiavellianism subscale of the Dirty Dozen questionnaire has been shown to have good psychometric properties (e.g. α = 0.72; Jonason & Webster, 2010). The participants indicated how much they agreed (1 = not at all, 5 = very much) with statements such as “I tend to manipulate others to get my way”. Thus it seems also tenable to use the Machiavellianism subscale of the Dirty Dozen questionnaire to examine the latent structure of Machiavellianism. 2.3. Taxometric analyses Regarding the first sample (MACH-IV), we combined the single items into item sum score indicators in accordance with Ruscio et al.
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
(2006). To form the sum scores, we first conducted an exploratory factor analysis with oblimin rotations. Both a scree plot of the eigenvalues and a Parallel analysis (Horn, 1965) suggested a three-factor solution: Items 1, 2, 5, 8, 12, 13, 15, 18 and 20 constituted a negative personal tactics and views about human nature factor, items 3, 6, 7, 9, 10, 16 and 19 were combined into a positive personal tactics factor, and items 4, 11, 14, and 17 formed a positive views about human nature factor. This is in line with previous research, which also found the three aforementioned factors, sometimes splitting the first factor in positive tactics and positive views about human nature (e.g., Calvete & Corral, 2000). The factors accounted for 38% of the variance. According to the RMSEA and TLI criteria a very good model fit was obtained, RMSEA = 0.04, TLI = 0.95 (Hu & Bentler, 1999). Regarding the second sample, we decided to use the items themselves as indicators for taxometric analyses, due to the conciseness of the measurement scale. As recommended by Ruscio, Walters, Marcus, and Kaczetow (2010) we applied the MAMBAC, MAXEIG and L-Mode taxometric procedures (Meehl & Yonce, 1994; Waller & Meehl, 1998). If a categorical structure is present the graphical output will yield a peaking curve for the MAMBAC and MAXEIG analyses and a multi-modal distribution curve for the L-Mode procedure. In conducting these analyses, we followed the default settings in Ruscio's taxometric program (Ruscio, 2014; see Ruscio et al., 2006 for a comprehensive introduction). For taxometric analysis to have sufficient power to detect possible categorical differences between putative groups, the standardized mean differences between the hypothetical categorical groups should be sufficiently large (e.g. d ≥ 1.25; Meehl, 1995). Furthermore, all indicators should correlate substantially with each other, but the correlation should be substantially smaller in the hypothetical categorical groups. Additionally, the sample size should be reasonably large. Recently it has been shown that sample characteristics of the data, for example skew and inter-correlations, influence the shape of the graphical output, which might potentially bias the reseacher's judgement when relying on these graphical decision procedures. To mitigate these issues, we employed the more objective resampling comparison curve fit index approach in this study (CCFI; Ruscio, Ruscio, & Meron, 2007). In the case of the CCFI approach to taxometric analyses, categorical and dimensional data that match the sample in several important aspects (e.g., inter-correlations and skewness) are simulated and then compared in its relative fit to the actual data. A resulting CCFI value above 0.50 denotes a better fit for a categorical latent structure and a value below 0.50 denotes a better fit for dimensional latent structure. We used the mean of the three taxometric procedures as the indicator of the latent status of the construct. The accuracy of this decision rule is supported by a large simulation study of Ruscio et al. (2010). Ruscio et al. (2010) found that using the mean CCFI of the three aforementioned taxometric procedures with a threshold of 0.5 achieved an accuracy of 98% in correctly classifying the latent status of a construct. To assign cases to putative groups we chose to empirically calculate the prevalences in our samples via preliminary MAMBAC and MAXEIG runs, because no estimates of the prevalence of Machiavellianism have yet been established. In our first sample (MACH-IV) we found a base rate of 39.20% and a base rate of 43.15% regarding our second sample (Machiavellianism subscale of the Dirty Dozen), which were used in all three subsequent taxometric analyses. All analyses were conducted in the open statistic Software R (R Development Core Team, 2014) with the code for taxometric analyses provided by Ruscio (2014). 3. Results 3.1. Descriptive characteristics 3.1.1. Sample 1 The overall mean of the MACH-IV sum score was 66.92 (SD = 13.84) for the total sample, 62.93 (SD = 13.07) for women and 68.98 (SD =
59
13.77) for men. Men had significantly larger sum scores than women, t(7994.2) = 22.48, p b 0.001. Age correlated negatively with the MACH-IV sum score, with a medium effect size of r = − 0.24, p b 0.001. The mean sum scores of the three empirically determined sub-scales of the MACH-IV were 29.02 (SD = 7.33), 23.01 (SD = 5.84) and 14.89 (SD = 3.06) for the total sample, 27.10 (SD = 7.09), 21.59 (SD = 5.50) and 14.24 (SD = 2.99) for women, and 30.02 (SD = 7.25), 23.74 (SD = 5.88) and 15.22 (SD = 3.04) for men. 3.1.2. Sample 2 The overall mean sum score of the Machiavellianism subscale of the dirty dozen questionnaire was 12.72 (SD = 4.03) for the total sample, 11.68 (SD = 3.95) for women and 13.27 (SD = 3.96) for men. Men had significantly larger sum scores than women, t(28404) = 38.27, p b 0.001. Age correlated negatively with the sum score, with a small effect size of r = −0.17, p b 0.001. The means of the four items used as indicators in the second sample were 3.07 (SD = 1.32), 3.68 (SD = 1.17), 3.34 (SD = 1.26) and 2.62 (SD = 1.23) for the total sample, 2.81 (SD = 1.33), 3.46 (SD = 1.21), 3.17 (SD = 1.26) and 2.23 (SD = 1.15) for women, and 3.20 (SD = 1.30), 3.80 (SD = 1.13), 3.44 (SD = 1.25) and 2.83 (SD = 1.22) for men. 3.2. Taxometric analyses 3.2.1. Sample 1 We observed a mean standardized difference of d = 1.86 between putative groups (d = 1.82, d = 2.03, d = 1.72 for the three indicators respectively). Additionally, an average correlation of r = 0.56 (SD = 0.05) for the total sample was obtained. As requested, a much smaller correlation in the hypothetical categorical groups (r = 0.20, SD = 0.11; r = 0.21, SD = 0.07) was found. Thus the data are adequate for taxometric analyses. Fig. 1 depicts the results of the taxometric procedures. In the graphs, the simulated dimensional data seem to fit the actual curve shape better than the categorical solution. Consequently, all CCFI values were substantially lower than CCFI = 0.50 (CCFIMAXEIG = 0.30, CCFIMAMBAC = 0.33, and CCFIL-Mode = 0.21, mean CCFI value = 0.28). Thus the results of our first sample strongly suggest a dimensional latent structure of Machiavellianism. 3.2.2. Sample 2 We observed a mean standardized difference of d = 1.82 between putative groups (d = 2.18, d = 1.52, d = 1.45, d = 2.13 for the four indicators respectively). Additionally, an average correlation of r = 0.54 (SD = 0.08) for the total sample was obtained. As requested, a much smaller correlation in the hypothetical categorical groups (r = 0.19, SD = 0.11; r = 0.19, SD = 0.11) was found. Thus the data of our second sample are also adequate for taxometric analyses. Fig. 2 depicts the results of the taxometric procedures of the second sample. In the graphs, the MAXEIG, MAMBAC and L-Mode simulated dimensional data seem to fit the actual curve shape better than the categorical solution. Consequently, all CCFI values were substantially lower than CCFI = 0.50 (CCFIMAXEIG = 0.24, CCFIMAMBAC = 0.42, and CCFILMode = 0.33, mean CCFI value = 0.28). Thus the results of our second sample strongly suggest a dimensional latent structure of Machiavellianism as well. 4. Discussion The present study evaluated the latent status of Machiavellianism in two large samples using two popular measures of Machiavellianism. In both samples, we found that the mean CCFI values were significantly smaller than 0.5. Therefore, Machiavellianism seems to have a dimensional latent status. Hence, our results are in accordance with previous studies. Both of the other dark triad traits—psychopathy and narcissism—are associated
60
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
Fig. 1. Average MAXEIG (top), MAMBAC (middle), and L-Mode (bottom) plots for the MACH-IV scale assessing Machiavellianism in a large online sample (N = 10.918). The plots in the left and right column show the actual research data (bold line) and two thin lines that depict the maximum and minimum of the simulated expected averaged categorical (left) or dimensional (right) comparison data. The shaded areas represent the middle 50% of simulated data points. The simulated dimensional data fit the actual data better than the simulated categorical data, thus suggesting a dimensional latent status of Machiavellianism.
with clinical disorders and might thus support categorical classifications. Machiavellianism, on the other hand, is not per se associated with a clinical diagnosis (but see Wastell & Booth, 2003) and therefore does not lend itself to categorical classifications. Multiple studies have evaluated the latent status of psychopathy and narcissism, but no such study had existed in the case of Machiavellianism. It was simply implicitly assumed in Machiavellianism research that Machiavellianism represents a dimensional construct. Thus, our results should comfort Machiavellianism researchers in that they suggest that the previously untested implicit assumption of a dimensional latent status has likely been correct.
Although taxometric analyses cannot identify causal mechanisms, they can inform the process of finding them. Here, results point to a dimensional latent construct. This dimensional solution is according to Meehl (1995) most likely to result from many small factors that add to each other and interact. In the case of Machiavelliansm numerous predictors have been empirically established (e.g. Alexithymia; Wastell & Booth, 2003). Most importantly, the dimensional latent structure suggests that no single predictor establishes a necessary or sufficient condition to form Machiavellianism: To shift a person on the latent dimension from low to high Machiavellianism multiple factors have to combine their effects.
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
61
Fig. 2. Average MAXEIG (top), MAMBAC (middle), and L-Mode (bottom) plots for the Machiavellianism sub-scale of the Dirty Dozen questionnaire in a large sample (N = 40.265). The plots in the left and right column show the actual research data (bold line) and two thin lines that depict the maximum and minimum of the simulated expected averaged categorical (left) or dimensional (right) comparison data. The shaded areas represent the middle 50% of simulated data points. The simulated dimensional data fit the actual data better than the simulated categorical data, thus suggesting a dimensional latent status of Machiavellianism.
The results have additional important implications for assessment and classification. Categorical latent constructs can typically be measured with fewer items that dimensional constructs, when they are strategically placed around the border between the different latent states. For a dimensional latent construct, however, it is important to have many items that, when combined in a dimensional sum score, allow fine differentiations across the whole latent dimension of the construct. Therefore, those instruments that are currently mostly used to measure Machiavellianism and that have been practically verified can and still
should be beneficially used. Furthermore researchers should be careful when dividing participants in Machs and Non-Machs or Low Machiavellians and High Machiavellians. The results of the current study suggest that this classification is essentially arbitrary and artificial. Regarding theory, researchers should prefer approaches which include the possibility of non-categorical dynamic strategies or at least include multiple different strategies instead of a strategy-dichotomy. Regarding research practice, researchers should avoid using cut-off scores, median splits and other procedures which arbitrarily classify participants, as these
62
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
procedures could at best add noise to real effects and at worst lead to false findings. Instead the original dimensional sum score should be used. Last the results have important implications for labeling. As noted previously, whether a construct is communicated as categorical or dimensional in nature, impacts the perceptions, attitudes and behavior of the lay public (Prentice & Miller, 2007). Here, results point to a dimensional latent structure. This implies that researchers should be careful in choosing labels implying a categorical latent status—such as “Machs”. For example, people could view the behavior of their interaction partners as more stable and less changeable when they are referred to as “Machs” which might in turn negatively influence cooperative or help-seeking behavior. Thus scientist should be careful in their language when communicating with other scientists and especially so when communicating with the lay public. One limitation of the current study is that only one data-source, answers on online questionnaires, was used. This implies that the results could suffer from a mono-methodological bias (Shadish, Cook, & Campbell, 2002). However, by using two different online questionnaires we could obtain two comparatively large sample sizes, which is seen as pivotal in taxometric studies. Additionally, we were able to analyze two of the most widely used instruments to measure Machiavellianism. Thus, although future studies must try to replicate our findings using other measures as well as other samples and by examining putative effects of culture, we have strong faith in the accuracy of our results. Summing up, although Machiavellianism constitutes an important phenomenon, no study to date analyzed the basic issue of its latent status. Analyzing the latent status in two large samples using two popular measures of Machiavellianism, we found that Machiavellianism encompasses dimensional differences. Thus, researchers should prefer dimensional conceptualizations of Machiavellianism in theory and research practices. Future studies should replicate these findings with additional measures and in further samples. References Baughman, H. M., Dearing, S., Giammarco, E., & Vernon, P. A. (2012). Relationships between bullying behaviours and the Dark Triad: A study with adults. Personality and Individual Differences, 52, 571–575. Beller, J., & Baier, D. (2013). Differential effects: Are the effects studied by psychologists really linear and homogeneous? Europe's Journal of Psychology, 9, 378–384. Beller, J., & Kröger, C. (2016). Is religious fundamentalism a dimensional or a categorical phenomenon? A taxometric analysis in two samples of youth from Egypt and Saudi Arabia. Psychology of Religion and Spirituality Advance online publication. http://dx. doi.org/10.1037/rel0000085 Bereczkei, T., & Czibor, A. (2014). Personality and situational factors differently influence high Mach and low Mach persons' decisions in a social dilemma game. Personality and Individual Differences, 64, 168–173. Brown, C., Garwood, M. P., & Williamson, J. E. (2012). It pays to cheat: Tactical deception in a cephalopod social signalling system. Biology Letters. http://dx.doi.org/10.1098/ rsbl.2012.0435. Calvete, E., & Corral, S. (2000). Machiavellianism: Dimensionality of the Mach IV and its relation to self-monitoring in a Spanish sample. The Spanish Journal of Psychology, 3, 3–13. Chabrol, H., Van Leeuwen, N., Rodgers, R., & Séjourné, N. (2009). Contributions of psychopathic, narcissistic, machiavellian, and sadistic personality traits to juvenile delinquency. Personality and Individual Differences, 47, 734–739. Christie, R., & Geis, F. L. (Eds.). (1970). Studies in Machiavellianism. San Diego, CA: Academic Press. Deluga, R. J. (2001). American presidential Machiavellianism: Implications for charismatic leadership and rated performance. The Leadership Quarterly, 12, 334–336.
Development Core Team, R. (2014). R [computer software]. Retrieved from http://www. R-project.org/ Edens, J. F., Marcus, D. K., Lilienfeld, S. O., & Poythress, N. G., Jr. (2006). Psychopathic, not psychopath: Taxometric evidence for the dimensional structure of psychopathy. Journal of Abnormal Psychology, 115, 131–144. Ettinger, D., & Jehiel, P. (2010). A theory of deception. American Economic Journal: Microeconomics, 2, 1–20. Foster, J. D., & Campbell, W. K. (2007). Are there such things as “narcissists” in social psychology? A taxometric analysis of the Narcissistic Personality Inventory. Personality and Individual Differences, 43, 1321–1332. Hawley, P. H. (2006). Resource control strategies inventory – Revised. Lawrence, KS: University of Kansas. Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30, 179–185. Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1–55. Jonason, P., & Webster, G. (2010). The dirty dozen: A concise measure of the dark triad. Psychological Assessment, 22, 420–432. Jones, D. N., & Paulhus, D. L. (2009). In M. R. Leary, & R. H. Hoyle (Eds.), Handbook of individual differences in social behavior (pp. 93–108). New York, NY: Guilford Press. Kliem, S., Beller, J., Kröger, C., Birowicz, T., Zenger, M., & Brähler, E. (2014). Dimensional latent structure of somatic symptom reporting in two representative population studies: Results from taxometric analyses. Psychological Assessment, 26, 484–494. Láng, A., & Birkás, B. (2014). Machiavellianism and perceived family functioning in adolescence. Personality and Individual Differences, 63, 69–74. Maestripieri, D. (2007). Machiavellian intelligence: How rhesus macaques and humans have conquered the world. Chicago University Press. Meehl, P. E. (1995). Bootstraps taxometrics: Solving the classification problem in psychopathology. American Psychologist, 50, 266–275. Meehl, P. E., & Yonce, L. J. (1994). Taxometric analysis: I. Detecting taxonicity with two quantitative indicators using means above and below a sliding cut (MAMBAC procedure). Psychological Reports, 74, 1059–1274. O'Boyle, E. H., Jr., Forsyth, D. R., Banks, G. C., & McDaniel, M. A. (2012). A meta-analysis of the dark triad and work behavior: A social exchange perspective. Journal of Applied Psychology, 97, 557–579. Paulhus, D. L., & Williams, K. M. (2002). The Dark Triad of personality: Narcissism, Machiavellianism, and psychopathy. Journal of Research in Personality, 36, 556–563. Porter, S., Bhanwer, A., Woodworth, M., & Black, P. J. (2014). Soldiers of misfortune: An examination of the Dark Triad and the experience of schadenfreude. Personality and Individual Differences, 67, 64–68. Prentice, D. A., & Miller, D. T. (2007). Psychological essentialism of human categories. Current Directions in Psychological Science, 16, 202–206. Rauthmann, J. F. (2013). Investigating the MACH–IV with item response theory and proposing the Trimmed MACH*. Journal of Personality Assessment, 95, 388–397. Ruscio, J. (2014). Taxometric programs for the R computing environment: User's manual. Retrieved from http://www.tcnj.edu/~ruscio/taxometrics.html Ruscio, J., Haslam, N., & Ruscio, A. M. (2006). Introduction to the taxometric method: A practical guide. Mahwah, N.J: Lawrence Erlbaum Associates. Ruscio, J., Ruscio, A. M., & Meron, M. (2007). Applying the bootstrap to taxometric analysis: Generating empirical sampling distributions to help interpret results. Multivariate Behavioral Research, 42, 349–386. Ruscio, J., Walters, G. D., Marcus, D. K., & Kaczetow, W. (2010). Comparing the relative fit of categorical and dimensional latent variable models using consistency tests. Psychological Assessment, 22, 5–21. Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. Boston: Houghton-Mifflin. Verbeke, W. J., Rietdijk, W. J., van den Berg, W. E., Dietvorst, R. C., Worm, L., & Bagozzi, R. P. (2011). The making of the Machiavellian brain: A structural MRI analysis. Journal of Neuroscience, Psychology, and Economics, 4, 205–216. Vernon, P. A., Villani, V. C., Vickers, L. C., & Harris, J. A. (2008). A behavioral genetic investigation of the Dark Triad and the Big 5. Personality and Individual Differences, 44, 445–452. Waller, N. G., & Meehl, P. E. (1998). Multivariate taxometric procedures: Distinguishing types from continua. Thousand Oaks, CA: Sage. Wastell, C., & Booth, A. (2003). Machiavellianism: An alexithymic perspective. Journal of Social and Clinical Psychology, 22, 730–744. Wilson, D. S., Near, D. C., & Miller, R. R. (1998). Individual differences in Machiavellianism as a mix of cooperative and exploitative strategies. Evolution and Human Behavior, 19, 203–212.
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
Machiavellianism has a dimensional latent structure: Results from taxometric analyses☆ Johannes Beller ⁎,1, Stefanie Bosse Braunschweig University of Technology, Germany
a r t i c l e
i n f o
Article history: Received 8 January 2017 Received in revised form 20 February 2017 Accepted 7 March 2017 Available online xxxx Keywords: Taxometric analysis Machiavellianism MACH-IV Dirty Dozen Dark triad
a b s t r a c t Despite the importance of Machiavellianism, no study has examined the basic issue of its latent status: Is Machiavellianism a dimensional or a categorical construct? Or equivalently, do people differ in the extent to which they are Machiavellian, or do Machiavellianists differ categorically from non-Machiavellianists? To answer these questions, we analyzed two large online samples of N1 = 10,918 participants who completed the MACH-IV questionnaire and N2 = 40,265 participants who completed the Machiavellianism subscale of the Dirty Dozen questionnaire. Via taxometric methods, we found that Machiavellianism encompasses quantitative rather than qualitative differences in both samples. Hence, people differ quantitatively to the extent to which they are Machiavellian. These findings have important practical and theoretical implications regarding assessment, classification, causality, and labeling. © 2017 Elsevier Ltd. All rights reserved.
1. Introduction The idea of Machiavellianism dates back to the Italian philosopher and politician Niccolò Machiavelli. According to the theme “the ends justify the means”, Machiavelli argued that moral aspects should be disregarded in favor of effectiveness and power (Christie & Geis, 1970). Four centuries later, psychologists have begun studying Machiavellianism as a trait. Individuals scoring high in Machiavellianism—often referred to as Machs—manipulate, control and exploit others to further their own intrinsically motivated goals (Jones & Paulhus, 2009; Paulhus & Williams, 2002). Consequently, Machiavellianism predicts a wide range of real-world anti-social and malevolent behavior (e.g. Baughman, Dearing, Giammarco, & Vernon, 2012; Chabrol, Van Leeuwen, Rodgers, & Séjourné, 2009; O'Boyle, Forsyth, Banks, & McDaniel, 2012) and correlates with mental health issues such as alexithymia and interpersonal maladjustment (Hawley, 2006). At the same time, Machiavellianism is also related to positive outcomes: Machs are more successful on the job (Jones & Paulhus, 2009), and individuals scoring high in Machiavellianism are seen as more effective, desirable and charismatic leaders (Deluga, 2001). Recent evolutionary research has even argued that the “Machiavellian Intelligence” of being able to manipulate our social environment enabled the
☆ We would like to thank Eric Jorgenson for sharing his data with us. ⁎ Corresponding author at: Department of Developmental, Personality and Forensic Psychology, TU Braunschweig, Humboldtstraße 33, 38106 Braunschweig, Germany. E-mail address: [email protected] (J. Beller). 1 Johannes Beller is supported by the German National Academic Foundation.
http://dx.doi.org/10.1016/j.paid.2017.03.014 0191-8869/© 2017 Elsevier Ltd. All rights reserved.
evolutionary success of Homo sapiens as a species (Maestripieri, 2007). Thus Machiavellianism represents an important psychological construct. 1.1. The latent status of Machiavellianism One central issue for every latent construct (like Machiavellianism) is its latent nature: Is the construct best described as a continuum or as qualitative different categories? Answering this question bears several important implications regarding, inter alia, classification, assessment, causality and labeling (Ruscio, Haslam, & Ruscio, 2006). First, the latent status of a construct is important in classifying individuals. If the underlying construct is continuous rather than categorical—as implied by the practice of forming sum scores—then, any classification in dichotomous groups needs to be considered very carefully and the whole process of classifying individuals based on sum-scores might be questioned. On the other hand, if a true categorical latent structure exists, taxometric analysis can provide and justify different cut scores. Second, information about the latent status is important for the development of assessment procedures. If the latent structure is categorical one would focus on using items which maximally discriminate between groups. Contrary, if the construct under consideration was continuous one would need to include items over the whole spectrum of the latent continuum. Third, the latent status is important in labeling. Whether a construct is communicated as categorical or dimensional in nature, impacts the perceptions, attitudes and behavior of the lay public (Prentice & Miller, 2007). For example, viewing Machiavellianism as categorical—as implied by researchers using the terms “Machs” and “Non-Machs”—may lead the lay public to construe Machiavellianism
58
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
as more stable and resistant to change than if it were viewed as dimensional. These beliefs may in turn shape the way members of this supposed category are perceived, approached and evaluated (Prentice & Miller, 2007). Fourth, the latent status might further theoretical insights about the construct, its antecendents and consequences (e.g. Beller & Baier, 2013). Ruscio et al. (2006), for example, suggest that categorical latent constructs might result from specific etiological factors, threshold effects, nonlinear interactions or developmental bifurcation, while dimensional constructs tend to result from numerous additive influences. Thus, analyzing the latent status of Machiavellianism may provide important insights. Due to these possible important insights numerous taxometric studies have explored the latent status of diverse psychological constructs (see e.g. Kliem et al., 2014; Beller & Kröger, 2016; or for Machiavellism related constructs: Edens, Marcus, Lilienfeld, & Poythress, 2006; Foster & Campbell, 2007). However, no such study exists for Machiavellianism. In the literature Machiavellianism has largely been considered to be a dimensional construct per fiat, but evidence might also suggest Machiavellian categories for at least three reasons: First, evolutionary approaches to Machiavellianism have conceptualized human populations as a mixture of cooperators (or Non-Machs) and exploiters (or Machs) (e.g. Maestripieri, 2007; Wilson, Near, & Miller, 1998). In the same vein, a large part of the vast literature on game theory has been concerned with analyzing categorically distinct cooperative and conflictive strategies, including deception (e.g. Brown, Garwood, & Williamson, 2012; Ettinger & Jehiel, 2010). Therefore it could be argued that these evolutionary strategies might constitute latent categories of Machiavellianism. Second, Machiavellianism has been shown to result mostly from environmental effects, but also from genetics (heritability factor of 0.31; Vernon, Villani, Vickers, & Harris, 2008). Following the aforementioned argument by Ruscio et al. (2006), Machiavellianism could start out as a dimensional construct, but might become a qualitatively different state only when specific genetic and environment or personality factors interact in a non-linear way, thus forming a categorical Mach condition. Third, the Machiavellianism test scores are often dichotomized. For example, Verbeke et al. (2011) suggest that a cut off score might be used to divide the participants into low and high Machiavellians. Other studies have used further strategies like a median split or a certain sum score range to classify participants in distinct Machiavellianism categories (e.g. Porter, Bhanwer, Woodworth, & Black, 2014; Bereczkei & Czibor, 2014; Láng & Birkás, 2014). Thus evolutionary theories and research practices might support a categorical view of Machiavellianism. But despite the importance of Machiavellianism no study exists in which the latent status of Machiavellianism is empirically determined. The current study strives to fill this gap. 1.2. Current study The current study contributes to the literature by clarifying whether the latent status of Machiavellianism is dimensional or categorical. Clarifying the latent status might provide important insights regarding classification, assessment, causality and labeling. Towards this end three non-redundant taxometric methods are applied to two large online samples (N1 = 10,918, N2 = 40,265). We ask: Is the latent status of Machiavellianism categorical or dimensional? 2. Method 2.1. Participants and procedure 2.1.1. Sample 1 Data were collected via an online survey provided on www. personality-testing.org. Participants answered questions regarding the MACH-IV questionnaire and demographic data (N = 13.156). Data collection began in January 2012 and ended June 2012. All participants explicitly agreed that their data might be used for scientific analyses. Prior
to the analyses we removed participants who indicated that they were younger than 18 or older than 80 years. After additionally deleting all participants with missing values on the MACH-IV scale a final sample size of 11,702 participants (65.7% male) ranging in age from 18 to 80 (M = 30.79, SD = 11.41) was obtained. 2.1.2. Sample 2 Data were also collected via an online survey from www. personality-testing.org. Participants answered questions regarding the Dirty Dozen questionnaire and demographic data (N = 53.981). Data collection began in July 2012 and ended in December 2013. Only participants who agreed that their data might be used for further scientific analyses were included in the sample. Additionally, as in the first sample, all participants who indicated that they were younger than 18 or older than 80 years old were removed prior to analyses. After deleting all missing values regarding the Machiavellianism subscale of the Dirty Dozen questionnaire, a final sample size of N2 = 40,165 participants (65.15% male) ranging in age from 18 to 80 (M = 28.12, SD = 10.53) was obtained (the results reported in this study do not change significantly when no participants are excluded in the analyses). 2.2. Measures 2.2.1. MACH-IV The 20 item MACH-IV scale by Christie and Geis (1970) has been the most widely used instrument to measure Machiavellianism (Jones & Paulhus, 2009). In previous studies, the Machiavellianism scale test scores showed good psychometric properties with acceptable internal consistencies (e.g. α = 0.71; Christie & Geis, 1970). Previous studies, however, differed in the proposed factor structure of the MACH-IV scale (Rauthmann, 2013). Originally, the MACH-IV was designed to encompass three sub-scales (interpersonal tactics, cynical view of human nature, disregard for conventional morality), which might be combined to form a total score. Subsequent studies challenged this factor structure (Rauthmann, 2013, for an overview). For example, Calvete and Corral (2000) found via confirmatory factor analyses that a four-factor structure (positive interpersonal tactics, negative tactics, positive view of human nature, cynical view of human nature) best fitted their data. Example items of the MACH-IV scale include “Never tell anyone the real reason you did something unless it is useful to do so”, “Most people are basically good and kind” (R) and “It is hard to get ahead without cutting corners here and there”. Participants responded on a five point scale (coded from 1 to 5), with reversed items recoded so that higher values represented stronger Machiavellianism believes. Additionally, a short version of the MACH-IV, the MACH* has recently been developed based on item response theory (Rauthmann, 2013). Thus, because of the widespread use, high validity and reliability of the MACH-IV scale scores, it seems tenable to use the MACH-IV to examine the latent structure of Machiavellianism. 2.2.2. Dirty Dozen The 12 item Dirty Dozen questionnaire has been a popular measure of the Dark Triad (Jonason & Webster, 2010) and thus Machiavellianism. Despite its conciseness with 4 items, the Machiavellianism subscale of the Dirty Dozen questionnaire has been shown to have good psychometric properties (e.g. α = 0.72; Jonason & Webster, 2010). The participants indicated how much they agreed (1 = not at all, 5 = very much) with statements such as “I tend to manipulate others to get my way”. Thus it seems also tenable to use the Machiavellianism subscale of the Dirty Dozen questionnaire to examine the latent structure of Machiavellianism. 2.3. Taxometric analyses Regarding the first sample (MACH-IV), we combined the single items into item sum score indicators in accordance with Ruscio et al.
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
(2006). To form the sum scores, we first conducted an exploratory factor analysis with oblimin rotations. Both a scree plot of the eigenvalues and a Parallel analysis (Horn, 1965) suggested a three-factor solution: Items 1, 2, 5, 8, 12, 13, 15, 18 and 20 constituted a negative personal tactics and views about human nature factor, items 3, 6, 7, 9, 10, 16 and 19 were combined into a positive personal tactics factor, and items 4, 11, 14, and 17 formed a positive views about human nature factor. This is in line with previous research, which also found the three aforementioned factors, sometimes splitting the first factor in positive tactics and positive views about human nature (e.g., Calvete & Corral, 2000). The factors accounted for 38% of the variance. According to the RMSEA and TLI criteria a very good model fit was obtained, RMSEA = 0.04, TLI = 0.95 (Hu & Bentler, 1999). Regarding the second sample, we decided to use the items themselves as indicators for taxometric analyses, due to the conciseness of the measurement scale. As recommended by Ruscio, Walters, Marcus, and Kaczetow (2010) we applied the MAMBAC, MAXEIG and L-Mode taxometric procedures (Meehl & Yonce, 1994; Waller & Meehl, 1998). If a categorical structure is present the graphical output will yield a peaking curve for the MAMBAC and MAXEIG analyses and a multi-modal distribution curve for the L-Mode procedure. In conducting these analyses, we followed the default settings in Ruscio's taxometric program (Ruscio, 2014; see Ruscio et al., 2006 for a comprehensive introduction). For taxometric analysis to have sufficient power to detect possible categorical differences between putative groups, the standardized mean differences between the hypothetical categorical groups should be sufficiently large (e.g. d ≥ 1.25; Meehl, 1995). Furthermore, all indicators should correlate substantially with each other, but the correlation should be substantially smaller in the hypothetical categorical groups. Additionally, the sample size should be reasonably large. Recently it has been shown that sample characteristics of the data, for example skew and inter-correlations, influence the shape of the graphical output, which might potentially bias the reseacher's judgement when relying on these graphical decision procedures. To mitigate these issues, we employed the more objective resampling comparison curve fit index approach in this study (CCFI; Ruscio, Ruscio, & Meron, 2007). In the case of the CCFI approach to taxometric analyses, categorical and dimensional data that match the sample in several important aspects (e.g., inter-correlations and skewness) are simulated and then compared in its relative fit to the actual data. A resulting CCFI value above 0.50 denotes a better fit for a categorical latent structure and a value below 0.50 denotes a better fit for dimensional latent structure. We used the mean of the three taxometric procedures as the indicator of the latent status of the construct. The accuracy of this decision rule is supported by a large simulation study of Ruscio et al. (2010). Ruscio et al. (2010) found that using the mean CCFI of the three aforementioned taxometric procedures with a threshold of 0.5 achieved an accuracy of 98% in correctly classifying the latent status of a construct. To assign cases to putative groups we chose to empirically calculate the prevalences in our samples via preliminary MAMBAC and MAXEIG runs, because no estimates of the prevalence of Machiavellianism have yet been established. In our first sample (MACH-IV) we found a base rate of 39.20% and a base rate of 43.15% regarding our second sample (Machiavellianism subscale of the Dirty Dozen), which were used in all three subsequent taxometric analyses. All analyses were conducted in the open statistic Software R (R Development Core Team, 2014) with the code for taxometric analyses provided by Ruscio (2014). 3. Results 3.1. Descriptive characteristics 3.1.1. Sample 1 The overall mean of the MACH-IV sum score was 66.92 (SD = 13.84) for the total sample, 62.93 (SD = 13.07) for women and 68.98 (SD =
59
13.77) for men. Men had significantly larger sum scores than women, t(7994.2) = 22.48, p b 0.001. Age correlated negatively with the MACH-IV sum score, with a medium effect size of r = − 0.24, p b 0.001. The mean sum scores of the three empirically determined sub-scales of the MACH-IV were 29.02 (SD = 7.33), 23.01 (SD = 5.84) and 14.89 (SD = 3.06) for the total sample, 27.10 (SD = 7.09), 21.59 (SD = 5.50) and 14.24 (SD = 2.99) for women, and 30.02 (SD = 7.25), 23.74 (SD = 5.88) and 15.22 (SD = 3.04) for men. 3.1.2. Sample 2 The overall mean sum score of the Machiavellianism subscale of the dirty dozen questionnaire was 12.72 (SD = 4.03) for the total sample, 11.68 (SD = 3.95) for women and 13.27 (SD = 3.96) for men. Men had significantly larger sum scores than women, t(28404) = 38.27, p b 0.001. Age correlated negatively with the sum score, with a small effect size of r = −0.17, p b 0.001. The means of the four items used as indicators in the second sample were 3.07 (SD = 1.32), 3.68 (SD = 1.17), 3.34 (SD = 1.26) and 2.62 (SD = 1.23) for the total sample, 2.81 (SD = 1.33), 3.46 (SD = 1.21), 3.17 (SD = 1.26) and 2.23 (SD = 1.15) for women, and 3.20 (SD = 1.30), 3.80 (SD = 1.13), 3.44 (SD = 1.25) and 2.83 (SD = 1.22) for men. 3.2. Taxometric analyses 3.2.1. Sample 1 We observed a mean standardized difference of d = 1.86 between putative groups (d = 1.82, d = 2.03, d = 1.72 for the three indicators respectively). Additionally, an average correlation of r = 0.56 (SD = 0.05) for the total sample was obtained. As requested, a much smaller correlation in the hypothetical categorical groups (r = 0.20, SD = 0.11; r = 0.21, SD = 0.07) was found. Thus the data are adequate for taxometric analyses. Fig. 1 depicts the results of the taxometric procedures. In the graphs, the simulated dimensional data seem to fit the actual curve shape better than the categorical solution. Consequently, all CCFI values were substantially lower than CCFI = 0.50 (CCFIMAXEIG = 0.30, CCFIMAMBAC = 0.33, and CCFIL-Mode = 0.21, mean CCFI value = 0.28). Thus the results of our first sample strongly suggest a dimensional latent structure of Machiavellianism. 3.2.2. Sample 2 We observed a mean standardized difference of d = 1.82 between putative groups (d = 2.18, d = 1.52, d = 1.45, d = 2.13 for the four indicators respectively). Additionally, an average correlation of r = 0.54 (SD = 0.08) for the total sample was obtained. As requested, a much smaller correlation in the hypothetical categorical groups (r = 0.19, SD = 0.11; r = 0.19, SD = 0.11) was found. Thus the data of our second sample are also adequate for taxometric analyses. Fig. 2 depicts the results of the taxometric procedures of the second sample. In the graphs, the MAXEIG, MAMBAC and L-Mode simulated dimensional data seem to fit the actual curve shape better than the categorical solution. Consequently, all CCFI values were substantially lower than CCFI = 0.50 (CCFIMAXEIG = 0.24, CCFIMAMBAC = 0.42, and CCFILMode = 0.33, mean CCFI value = 0.28). Thus the results of our second sample strongly suggest a dimensional latent structure of Machiavellianism as well. 4. Discussion The present study evaluated the latent status of Machiavellianism in two large samples using two popular measures of Machiavellianism. In both samples, we found that the mean CCFI values were significantly smaller than 0.5. Therefore, Machiavellianism seems to have a dimensional latent status. Hence, our results are in accordance with previous studies. Both of the other dark triad traits—psychopathy and narcissism—are associated
60
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
Fig. 1. Average MAXEIG (top), MAMBAC (middle), and L-Mode (bottom) plots for the MACH-IV scale assessing Machiavellianism in a large online sample (N = 10.918). The plots in the left and right column show the actual research data (bold line) and two thin lines that depict the maximum and minimum of the simulated expected averaged categorical (left) or dimensional (right) comparison data. The shaded areas represent the middle 50% of simulated data points. The simulated dimensional data fit the actual data better than the simulated categorical data, thus suggesting a dimensional latent status of Machiavellianism.
with clinical disorders and might thus support categorical classifications. Machiavellianism, on the other hand, is not per se associated with a clinical diagnosis (but see Wastell & Booth, 2003) and therefore does not lend itself to categorical classifications. Multiple studies have evaluated the latent status of psychopathy and narcissism, but no such study had existed in the case of Machiavellianism. It was simply implicitly assumed in Machiavellianism research that Machiavellianism represents a dimensional construct. Thus, our results should comfort Machiavellianism researchers in that they suggest that the previously untested implicit assumption of a dimensional latent status has likely been correct.
Although taxometric analyses cannot identify causal mechanisms, they can inform the process of finding them. Here, results point to a dimensional latent construct. This dimensional solution is according to Meehl (1995) most likely to result from many small factors that add to each other and interact. In the case of Machiavelliansm numerous predictors have been empirically established (e.g. Alexithymia; Wastell & Booth, 2003). Most importantly, the dimensional latent structure suggests that no single predictor establishes a necessary or sufficient condition to form Machiavellianism: To shift a person on the latent dimension from low to high Machiavellianism multiple factors have to combine their effects.
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
61
Fig. 2. Average MAXEIG (top), MAMBAC (middle), and L-Mode (bottom) plots for the Machiavellianism sub-scale of the Dirty Dozen questionnaire in a large sample (N = 40.265). The plots in the left and right column show the actual research data (bold line) and two thin lines that depict the maximum and minimum of the simulated expected averaged categorical (left) or dimensional (right) comparison data. The shaded areas represent the middle 50% of simulated data points. The simulated dimensional data fit the actual data better than the simulated categorical data, thus suggesting a dimensional latent status of Machiavellianism.
The results have additional important implications for assessment and classification. Categorical latent constructs can typically be measured with fewer items that dimensional constructs, when they are strategically placed around the border between the different latent states. For a dimensional latent construct, however, it is important to have many items that, when combined in a dimensional sum score, allow fine differentiations across the whole latent dimension of the construct. Therefore, those instruments that are currently mostly used to measure Machiavellianism and that have been practically verified can and still
should be beneficially used. Furthermore researchers should be careful when dividing participants in Machs and Non-Machs or Low Machiavellians and High Machiavellians. The results of the current study suggest that this classification is essentially arbitrary and artificial. Regarding theory, researchers should prefer approaches which include the possibility of non-categorical dynamic strategies or at least include multiple different strategies instead of a strategy-dichotomy. Regarding research practice, researchers should avoid using cut-off scores, median splits and other procedures which arbitrarily classify participants, as these
62
J. Beller, S. Bosse Personality and Individual Differences 113 (2017) 57–62
procedures could at best add noise to real effects and at worst lead to false findings. Instead the original dimensional sum score should be used. Last the results have important implications for labeling. As noted previously, whether a construct is communicated as categorical or dimensional in nature, impacts the perceptions, attitudes and behavior of the lay public (Prentice & Miller, 2007). Here, results point to a dimensional latent structure. This implies that researchers should be careful in choosing labels implying a categorical latent status—such as “Machs”. For example, people could view the behavior of their interaction partners as more stable and less changeable when they are referred to as “Machs” which might in turn negatively influence cooperative or help-seeking behavior. Thus scientist should be careful in their language when communicating with other scientists and especially so when communicating with the lay public. One limitation of the current study is that only one data-source, answers on online questionnaires, was used. This implies that the results could suffer from a mono-methodological bias (Shadish, Cook, & Campbell, 2002). However, by using two different online questionnaires we could obtain two comparatively large sample sizes, which is seen as pivotal in taxometric studies. Additionally, we were able to analyze two of the most widely used instruments to measure Machiavellianism. Thus, although future studies must try to replicate our findings using other measures as well as other samples and by examining putative effects of culture, we have strong faith in the accuracy of our results. Summing up, although Machiavellianism constitutes an important phenomenon, no study to date analyzed the basic issue of its latent status. Analyzing the latent status in two large samples using two popular measures of Machiavellianism, we found that Machiavellianism encompasses dimensional differences. Thus, researchers should prefer dimensional conceptualizations of Machiavellianism in theory and research practices. Future studies should replicate these findings with additional measures and in further samples. References Baughman, H. M., Dearing, S., Giammarco, E., & Vernon, P. A. (2012). Relationships between bullying behaviours and the Dark Triad: A study with adults. Personality and Individual Differences, 52, 571–575. Beller, J., & Baier, D. (2013). Differential effects: Are the effects studied by psychologists really linear and homogeneous? Europe's Journal of Psychology, 9, 378–384. Beller, J., & Kröger, C. (2016). Is religious fundamentalism a dimensional or a categorical phenomenon? A taxometric analysis in two samples of youth from Egypt and Saudi Arabia. Psychology of Religion and Spirituality Advance online publication. http://dx. doi.org/10.1037/rel0000085 Bereczkei, T., & Czibor, A. (2014). Personality and situational factors differently influence high Mach and low Mach persons' decisions in a social dilemma game. Personality and Individual Differences, 64, 168–173. Brown, C., Garwood, M. P., & Williamson, J. E. (2012). It pays to cheat: Tactical deception in a cephalopod social signalling system. Biology Letters. http://dx.doi.org/10.1098/ rsbl.2012.0435. Calvete, E., & Corral, S. (2000). Machiavellianism: Dimensionality of the Mach IV and its relation to self-monitoring in a Spanish sample. The Spanish Journal of Psychology, 3, 3–13. Chabrol, H., Van Leeuwen, N., Rodgers, R., & Séjourné, N. (2009). Contributions of psychopathic, narcissistic, machiavellian, and sadistic personality traits to juvenile delinquency. Personality and Individual Differences, 47, 734–739. Christie, R., & Geis, F. L. (Eds.). (1970). Studies in Machiavellianism. San Diego, CA: Academic Press. Deluga, R. J. (2001). American presidential Machiavellianism: Implications for charismatic leadership and rated performance. The Leadership Quarterly, 12, 334–336.
Development Core Team, R. (2014). R [computer software]. Retrieved from http://www. R-project.org/ Edens, J. F., Marcus, D. K., Lilienfeld, S. O., & Poythress, N. G., Jr. (2006). Psychopathic, not psychopath: Taxometric evidence for the dimensional structure of psychopathy. Journal of Abnormal Psychology, 115, 131–144. Ettinger, D., & Jehiel, P. (2010). A theory of deception. American Economic Journal: Microeconomics, 2, 1–20. Foster, J. D., & Campbell, W. K. (2007). Are there such things as “narcissists” in social psychology? A taxometric analysis of the Narcissistic Personality Inventory. Personality and Individual Differences, 43, 1321–1332. Hawley, P. H. (2006). Resource control strategies inventory – Revised. Lawrence, KS: University of Kansas. Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30, 179–185. Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1–55. Jonason, P., & Webster, G. (2010). The dirty dozen: A concise measure of the dark triad. Psychological Assessment, 22, 420–432. Jones, D. N., & Paulhus, D. L. (2009). In M. R. Leary, & R. H. Hoyle (Eds.), Handbook of individual differences in social behavior (pp. 93–108). New York, NY: Guilford Press. Kliem, S., Beller, J., Kröger, C., Birowicz, T., Zenger, M., & Brähler, E. (2014). Dimensional latent structure of somatic symptom reporting in two representative population studies: Results from taxometric analyses. Psychological Assessment, 26, 484–494. Láng, A., & Birkás, B. (2014). Machiavellianism and perceived family functioning in adolescence. Personality and Individual Differences, 63, 69–74. Maestripieri, D. (2007). Machiavellian intelligence: How rhesus macaques and humans have conquered the world. Chicago University Press. Meehl, P. E. (1995). Bootstraps taxometrics: Solving the classification problem in psychopathology. American Psychologist, 50, 266–275. Meehl, P. E., & Yonce, L. J. (1994). Taxometric analysis: I. Detecting taxonicity with two quantitative indicators using means above and below a sliding cut (MAMBAC procedure). Psychological Reports, 74, 1059–1274. O'Boyle, E. H., Jr., Forsyth, D. R., Banks, G. C., & McDaniel, M. A. (2012). A meta-analysis of the dark triad and work behavior: A social exchange perspective. Journal of Applied Psychology, 97, 557–579. Paulhus, D. L., & Williams, K. M. (2002). The Dark Triad of personality: Narcissism, Machiavellianism, and psychopathy. Journal of Research in Personality, 36, 556–563. Porter, S., Bhanwer, A., Woodworth, M., & Black, P. J. (2014). Soldiers of misfortune: An examination of the Dark Triad and the experience of schadenfreude. Personality and Individual Differences, 67, 64–68. Prentice, D. A., & Miller, D. T. (2007). Psychological essentialism of human categories. Current Directions in Psychological Science, 16, 202–206. Rauthmann, J. F. (2013). Investigating the MACH–IV with item response theory and proposing the Trimmed MACH*. Journal of Personality Assessment, 95, 388–397. Ruscio, J. (2014). Taxometric programs for the R computing environment: User's manual. Retrieved from http://www.tcnj.edu/~ruscio/taxometrics.html Ruscio, J., Haslam, N., & Ruscio, A. M. (2006). Introduction to the taxometric method: A practical guide. Mahwah, N.J: Lawrence Erlbaum Associates. Ruscio, J., Ruscio, A. M., & Meron, M. (2007). Applying the bootstrap to taxometric analysis: Generating empirical sampling distributions to help interpret results. Multivariate Behavioral Research, 42, 349–386. Ruscio, J., Walters, G. D., Marcus, D. K., & Kaczetow, W. (2010). Comparing the relative fit of categorical and dimensional latent variable models using consistency tests. Psychological Assessment, 22, 5–21. Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. Boston: Houghton-Mifflin. Verbeke, W. J., Rietdijk, W. J., van den Berg, W. E., Dietvorst, R. C., Worm, L., & Bagozzi, R. P. (2011). The making of the Machiavellian brain: A structural MRI analysis. Journal of Neuroscience, Psychology, and Economics, 4, 205–216. Vernon, P. A., Villani, V. C., Vickers, L. C., & Harris, J. A. (2008). A behavioral genetic investigation of the Dark Triad and the Big 5. Personality and Individual Differences, 44, 445–452. Waller, N. G., & Meehl, P. E. (1998). Multivariate taxometric procedures: Distinguishing types from continua. Thousand Oaks, CA: Sage. Wastell, C., & Booth, A. (2003). Machiavellianism: An alexithymic perspective. Journal of Social and Clinical Psychology, 22, 730–744. Wilson, D. S., Near, D. C., & Miller, R. R. (1998). Individual differences in Machiavellianism as a mix of cooperative and exploitative strategies. Evolution and Human Behavior, 19, 203–212.