MEASURING CHARISMA

The General Charisma Inventory (GCI), designed to measure influence and affability, was developed in a series of studies reported by Tshkay et al. (2018). An initial pool of items theoretically tapping the two dimensions of charisma that were generated through collecting public opinions comprised 408 indicators grouped into 100 categories. These 100 categories were further assessed in terms of the likelihood of being a good descriptor of charisma, which resulted in a pool of 40 items. In view of the results of exploratory factor analysis, which supported the existence of the underlying two-factorial structure, this pool was further reduced to 10 items. Next, based on the results of confirmatory factor analysis, the scale was trimmed to six items (three items on each scale) and yielded a good model fit. In four subsequent studies, Tshkay et al. (2018) claimed that the GCI is a valid and unique measure of charisma through the demonstration of a range of correlates of basic personality traits: emotionality, charismatic leadership and expressiveness, competence and warmth, and intelligence.

PARTICIPANTS

A total of N = 756 adult participants from Spain, aged between 18 and 49 years (M = 33.77, SD = 9.60, 71.6% females) participated voluntarily in the study. The Spanish GCI was entered into a Google form and distributed by means of a snowball sampling methodology to collect data, by which research subjects facilitate the contact of third subjects to participate in the research, and so on (Vogt & Johnson, 2011). This method of convenience sampling has been used to examine different populations and its application in quantitative studies has received empirical support (Cohen & Arieli, 2011). In our case, we provided the link of the survey to four classes of psychology undergraduates and asked them to re-send the link to their acquaintances through the smartphone, with the only requisite of being > 18 years old. They did not receive any compensation in exchange. Based on the original study of Tshkay et al. (2018), we attempted to find a similar size of around 750 participants. Accordingly, the sampling recruitment finished when we reached an approximate size. The participants had to read and accept an online informed consent form before completing the test. The final sample was divided into three age groups: n = 292 participants younger than 30 (M = 23.33, SD = 3.33, 73.6% females); n = 203 participants between 30 and 39 (M = 34.90, SD = 2.89, 70.4% females); and n = 261 respondents who were 40 or older (M = 44.58, SD = 2.79, 70.1% females). Based on previous research (e.g., Roberts et al., 2006), age thresholds were created arbitrarily to examine changes in personality traits. Gender distribution was equal across age groups (χ²(2) = 1.01, p = .604).

MEASURES

General Charisma Inventory (GCI). The GCI (Tshkay et al., 2018) is composed of six items designed to measure influence (α = .68; sample item: “Has a presence in a room”) and affability (α = .71; sample item: “Makes people feel comfortable”); respondents answer using a five-point Likert-type scale from 1 (strongly disagree) to 5 (strongly agree). The validation followed a multistep process for translating and adapting tests (Muñiz et al., 2016). We (a) translated the items into Spanish, (b) consulted its linguistic and cultural suitability, (c) had the items back-translated into English by two independent translators, (d) reported the results to the original authors, and (e) discussed and added the authors’ suggestions. First of all, the items were translated into first-person statements to make them more comprehensible in Spanish. During revision of the paper, we asked 8 Spanish native speakers to rate the suitability and comprehensiveness of two versions of the GCI using a ten-point scale: (a) a first-person translation of the GCI and (b) a third-person translation of the GCI. The first-person statements received slightly higher ratings (M = 54.38, SD = 7.21) than the third-person version (M = 52.25, SD = 6.48), but this difference was not significant (t(7) = 0.67, p = .526), suggesting that both versions are adequate.

STATISTICAL ANALYSES

The structure of the Spanish adaptation of the GCI across age and gender was verified using a multigroup confirmatory factor analysis, which was calculated using Mplus v. 7.2 (Muthén & Muthén, 2014). The results were estimated using a robust maximum likelihood method approach due to likely violation of the multivariate normality assumption. In order to evaluate the model fit, we used CFI and RMSEA. An acceptable model fit is indicated when CFI ≥ .95, TLI ≥ .95, RMSEA ≤ .08, left boundary of RMSEA confidence interval is ≤ .10, and SRMR ≤ .10 (Schermelleh-Engel et al., 2003). In multigroup analyses, it is possible to test for measurement invariance when three models are tested: configural (unconstrained), metric (with factor loadings constrained to be equal across groups), and scalar (with intercepts constrained to be equal across groups; Meredith, 1993). The data may be deemed as invariant when the difference in ΔCFI does not exceed .010 and the difference in ΔRMSEA does not exceed .015 (Chen, 2007).

RESULTS

Item characteristics. Overall, participants scored higher on each item measuring affability (items 4-6) than on any item measuring influence (items 1-3), which is also reflected in a significant difference in their scores (t(755) = –25.15, p < .001, d = 0.92). The corrected item-total correlations (i.e., the correlation of a single item with the total score, which excludes this particular item) ranged from .43 to .61 for influence and from .47 to .60 for affability. The tested model was well fitted to the data (χ²(8) = 41.19, p < .001, CFI = .954, TLI = .914, RMSEA = .074, 90% CI [.053, .097], SRMR = .046, BIC = 10929.01) and represented the GCI structure better than the one-factor model (which was itself poorly fitted: χ²(9) = 578.74, p < .001, CFI = .217, TLI = –.305, RMSEA = .289, 90% CI [.270, .310], SRMR = .135, BIC = 11316.34) as according to the BIC values and the difference in the χ² between the models, the two-factor model yielded superior fit. The strength of the standardized factor loadings ranged between .53 and .87, and the latent factor correlation was small (ρ = .22, p < .001) and similar to the correlation between observed scores (r = 18, p < .001). A summary of descriptive statistics along with the reliability estimates, corrected item-total correlations, and standardized factor loadings is presented in Table 1.

Assessment of the GCI structure across age and gender. Results of the tested models (presented in Table 2) indicate that the two-factorial model is well fitted to the data, better than the one-factorial model, which showed poor adjustment as CFI ranged from .379 (RMSEA = .294) to .719 (RMSEA = .138) in the age groups and from .433 (RMSEA = .274) to .698 (RMSEA = .184) in the gender groups.

According to the values of ΔRMSEA, the two-factorial structure of the GCI may be deemed invariant across age and gender; however, the values of ΔCFI exceeded acceptable recommendations. Thus, following the procedure proposed by Gregorich (2006), in order to achieve partial scalar invariance, we freed the constraint of equal intercepts of a single item, which significantly improved the fit of the scalar model. Across age and gender, the GCI was found to have the same factor loadings, and with some minor exceptions, the same intercepts, which makes a comparison of latent mean scores meaningful. The differences in the latent scores of influence and affability across the compared groups are presented in Table 3.

In comparison to the 18-29 age group, we noted a significant and stable decrease in influence in other groups. Affability, however, decreased more slowly and was only significantly lower in the 40-49 age group. In regards to gender differences, men scored higher on influence, but no difference was found in affability.

DISCUSSION

Our results demonstrated that higher levels of charisma are more characteristic of younger people. The path of change, nonetheless, seems to be to some extent different for influence and affability. The drop in influence is more drastic among respondents in their thirties, and it negligibly decreases onwards, whereas the drop in affability begins rather slowly and continues to increase with age. These results contradict prior research on basic personality traits, because traits associated with charisma (i.e., social dominance and agreeableness) tended to increase with age (Roberts et al., 2006). Therefore, future studies (preferably longitudinal) are needed to address whether the dynamics of charisma differ with age or whether those differences are the result of studied generations.

In regards to gender differences, men scored significantly higher on influence than women, and no significant difference was found for affability (although women tended to score higher). These results are in accordance with existing studies reporting that men generally have a more agentic orientation, expressed by domination and competence, for example, whilst women generally have a more communal orientation expressed by nurturance and warmth (Wojciszke & Szlendak, 2010). Our results may suggest that men and women have their own distinct ways of being perceived as charismatic.

PARTICIPANTS

The study was completed by N = 96 adults (71.9% females) from Spain aged between 20 and 72 (M = 47.58, SD = 12.79). Data were collected by means of a snowball sampling method and the participants were asked to complete the measures in a Google form. The participants had to read and accept an online informed consent form prior to completing the test. There were no missing responses throughout the study.

MEASURES

The General Charisma Inventory (GCI; Tshkay et al., 2018) was used to measure general charisma. On average, the participants scored M = 3.52 (SD = 0.80) on influence and M = 4.25 (SD = 0.53) on affability. The reliability estimates in the current study were as follows: α = .80 for influence and α = .66 for affability. Both dimensions were positively correlated at r = .36 (p < .001).

The Mental Health Continuum – Short Form (MHC-SF; Keyes, 1998; Spanish adaptation: Echeverria et al., 2017) was used to measure three components of well-being: emotional, social, and psychological. The scale is composed of 14 items which the respondents answer using a six-point Likert-type scale from 1 (never) to 6 (always). It has been validated across different cultures and demonstrates good psychometric properties (e.g., Karaś et al., 2014; Longo et al., 2017; Rogoza et al., 2018a). The reliability estimates in the current study were good: α = .81 for emotional well-being, α = .83 for social well-being, and α = .86 for psychological well-being.

STATISTICAL ANALYSES

We began with evaluation of the GCI structure via the means of confirmatory factor analysis in accordance with recommendations reported in Study 1. Further, to evaluate the criterion validity, we ran correlational analyses supplemented by linear regression models in which charisma dimensions were entered as predictors of well-being.

RESULTS

The measurement model of the Spanish version of GCI was mostly well fitted to the data (χ²(8) = 13.34, p = .101, CFI = .960, TLI = .925, RMSEA = .083, 90% CI [.000, .159], SRMR = .041) and the strength of the standardized factor loadings was adequate (range: .59-.89). The zero-order correlations between all studied variables are presented in Table 4.

Separate linear regression models with influence and affability as predictors of the three components of well-being were tested. The results are presented in Table 5.

All of the regression models were significant. Affability was a positive predictor of all of the well-being components, whereas influence did not predict emotional well-being (16% of variance was explained by the model), but was a positive predictor of social well-being (24% of variance was explained by the model) and psychological well-being (32% of variance was explained by the model). Although the direction of the effects was the same for both dimensions of charisma, affability was generally a stronger predictor of well-being.

DISCUSSION

The results obtained in this study support the distinction of subjective and psychological well-being since affability emerged as the only predictor of emotional (i.e., subjective) well-being, whereas both dimensions of charisma predicted the social and psychological components of eudaimonic well-being. Affability, the ability to make other people feel good, seems to be more important in maintaining warm and stable interpersonal relationships, whereas influence seems to be a resource for initiating and developing them. To summarise, both dimensions of charisma account for variance of well-being independently and uniquely, which supports their concurrent criterion validity and thus makes it meaningful to consider both dimensions.

PARTICIPANTS

The link to the study was sent to different local and student groups on a social networking site (Facebook). In total, N = 149 participants (76.5% female) completed the survey. Participants were mostly young adults aged between 18 and 50 (M = 23.91, SD = 4.68). No missing responses were recorded.

MEASURES

The General Charisma Inventory (GCI; Tshkay et al., 2018) was again used as a measure of general charisma. In the current study we applied the third-person version of the measure. On average, respondents scored M = 3.24 (SD = 1.01) on influence and M = 3.71 (SD = 0.95) on affability. The reliability estimates in the current study were good, i.e., α = .85 for influence and α = .80 for affability. Both dimensions were positively correlated at r = .46 (p < .001).

The Short Dark Triad (Jones & Paulhus, 2014; Spanish adaptation: Pineda et al., 2020) was used to measure three socially malevolent traits referred as the Dark Triad of personality. The scale is composed of 27 items on which the respondents answer using a five-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Although the scale is criticised for the high overlap between psychopathy and Machiavellianism (e.g., Persson et al., 2017; Rogoza & Cieciuch, 2017), in empirical research it provides distinct and theoretically valid results (Jones & Paulhus, 2017; Kowalski et al., 2018a, b). The reliability estimates in the current study were good: α = .78 for Machiavellianism, and α = 72 for both narcissism and psychopathy.

STATISTICAL ANALYSES

Similar to Study 2, we began with evaluation of the GCI structure by means of confirmatory factor analysis. To assess whether the first- and third-person version are interpreted in similar vein, that is, whether the same items are loading the same latent factors, we compared their factorial equivalence. Factorial equivalence may be claimed when the approximate fit indices fall within the acceptable range, that is: CFI ≥ .95, TLI ≥ .95, RMSEA ≤ .08, left boundary of RMSEA confidence interval is ≤ .10, and SRMR ≤ .10 (Schermelleh-Engel et al., 2003). As in the previous study, to evaluate the criterion validity, we ran correlational analyses supplemented by linear regression models in which charisma dimensions were entered as predictors.

RESULTS

The analysed model was well fitted to the data (χ²(8) = 3.57, p = .894, CFI = 1.00, TLI = 1.00, RMSEA = .000, 90% CI [.000, .043], SRMR = .020) and the strength of the standardized factor loadings was adequate, ranging from .68 to .89, confirming the two-dimensional structure of the GCI in the Spanish-speaking population.

The zero-order correlations between all studied variables are presented in Table 6.

As in the previous study, we tested three linear regression models of the relationship between dimensions of charisma and the Dark Triad traits. The results of the regression analyses are presented in Table 7.

All of the analysed regression models were significant. The first model explained almost half (42%) of the variance in narcissism variance, and given that affability did not account for variance in narcissism, this presumably reflects the criterion validity of the influence dimension. Influence and affability explained less of the variance in psychopathy and Machiavellianism (19% and 7% respectively) and they acted in opposite directions: influence was a positive predictor, whereas affability was a negative predictor, thus confirming our hypotheses.

ASSESSMENT OF THE FACTORIAL EQUIVALENCE BETWEEN THE THIRD- AND FIRST-PERSON VERSION

In the third study we used a different version of the measure, in which the items were presented in the third person. In order to test whether the respondents using the different versions of the measure understood charisma in the same way and whether, therefore, the two versions could be used interchangeably, we tested for factorial invariance across the data from Study 2 and Study 3. The model had a good fit to the data (χ²(16) = 15.70, p = .474, CFI = 1.00, TLI = 1.00, RMSEA = .00, 90% CI [.00, .08], SRMR = .030), supporting factorial equivalence of the two versions, indicating that both of them could be used in research on Spanish-speaking populations.

DISCUSSION

Narcissism was most strongly associated with charisma, which is not surprising since narcissists are described as possessing an immediately perceptible charm and charisma (Young & Pinsky, 2006). While we replicated these results in correlational analyses, when we controlled for the shared variance of the influence and affability, narcissism was related only to the latter, which may reflect the difference between narcissism and charisma. Both are superficially associated with boldness, visionary quality and self-confidence, but while extended contact leads to this impression of narcissists wearing off due to their arrogance, lack of concern for others and sense of entitlement (Fatfouta, 2018; Ong et al., 2016), it may be more persistent in the case of charismatic individuals due to their affability. Hence, although there are similarities between narcissism and influence, affability seems to differentiate them.

The lack of relationship between influence and Machiavellianism might be surprising at first glance, but it is theoretically plausible. The influence dimension of charisma reflects agentic attitudes (Tshkay et al., 2018). However, among the Dark Triad traits, only narcissism seems to be related to the agency-related traits such as intelligence. While this is incongruent with their actual abilities (Zajenkowski & Czarna, 2015; Zajenkowski et al., 2020), only narcissists are perceived by others as more agentic (Rauthmann & Kolar, 2013), which results in agency-related outcomes such as being nominated as a leader (Rogoza et al., 2020). Furthermore, narcissism in the context of the Dark Triad predominately taps its agentic facet, while antagonistic features are assessed to a limited extent (Rogoza et al., 2019). These antagonistic features have much more in common with Machiavellianism and psychopathy (Trahair et al., 2020); thus, it is not surprising to see a null relation (Rogoza & Fatfouta, 2020). The lack of relationship between Machiavellianism and affability is against our expectations; however, existing research suggests that Machiavellianism is less specific in its operationalisation than psychopathy (Rogoza et al., 2019). Therefore, seeing the same pattern of relationship but of lesser magnitude is understandable.