Exploring the Relationships between Self-Evaluations, Personality Traits and Disorders, and Political Skill ()
1. Introduction
Our ability to befriend and influence people in the workplace is central to both job satisfaction and success. An individual’s political skill can reflect a legitimate influence outside of organisation authority and expertise (De Luca, 1999; Mintzberg, 1985) , and may be considered fundamental to both leadership emergence and effectiveness. The politically skilled at work are confident, calm, and usually admired; they understand social environments and influence them for their own, and others’ benefit (Ferris et al., 2005) . Hence political skill is associated with promotion and organisational seniority (Furnham et al., 2013) . There is however a “dark-side” to political skills often referred to as “office politics” which has very negative connotations associated with corruption, Machiavellianism and manipulation (Furnham, 2018a, 2018b) .
Perhaps the greatest factor in the development of the concept occurred fifteen years ago when Ferris et al. (2005) reported on the development of a Political Skill Inventory (PSI) which has generated a great deal of research, and has since become the most popular measure of its kind (Ferris et al., 2012; Zettler & Lang, 2015) . It has been applied in organizational settings attempting to explain how it relates to job success (Bing et al., 2011; Ferris et al., 2005; Liu et al., 2007; Treadway et al., 2004; Zivnuska et al., 2019) and varies across cultures (Cepas-González et al., 2020; Chen & Lin, 2014; Shi & Chen, 2012) . The Political Skill Inventory breaks political skill into four factors. The first, Networking Ability, reflects how an individual forms friendships and connections with influential people at work. The second, Social Astuteness, measures how well an individual can understand interpersonal situations and the emotions and motivations of others. The third, Interpersonal Influence is how well an individual can adapt to their environment to evoke desired behaviour in others. The fourth, Apparent Sincerity, is how well an individual manages their perceived intentions, so they are seen as genuine, open and honest.
Various studies looked at the relationship between “bright-side” personality and political skill (Ferris et al., 1999) . In their seminal paper, Ferris et al. (2005) showed how each Political Skills factor correlated with different personality traits. Conscientiousness was positively linked to each factor, Trait Anxiety was negatively correlated with Social Astuteness and Apparent Sincerity, but positively correlated with Interpersonal Influence and Networking Ability. Later, Munyon et al. (2015) created meta-analytic correlation tables and showed that Extraversion was the strongest predictor of overall political skill (r = .48), with Agreeableness, Conscientiousness, and Emotional Stability also having notable positive correlates (all r > .20).
Being political at work means a lot of social interaction, which explains the importance of trait Extraversion, while people also need to be warm and apparently empathic, which explains the relationship with Agreeableness. Equally, politically skillful people need to be resilient (low Neuroticism), with high Emotional Stability to withstand the rejections and problems at work. Similarly, political skill involves being planful, organized and being forward-looking, which explains the role of Conscientiousness
There have been a number of convergent/divergent and incremental validity studies which have attempted to explain how political skills predict job performance above classic measures of intelligence and personality (Blickle et al., 2013) . Inevitably, to determine the convergent and divergent validity of the measures, it has been correlated with well-known tests like the Big Five personality traits. Recent research has shown that political skill positively relates to likability and sociability (Wang & Hall, 2019) and impression management (Maher et al., 2018) .
In this study we focus on a number of potential correlates of political skill to answer the question as to their incremental validity at both the domain (total score) and facet (scale) level. First, we attempt to replicate the studies on Big-Five correlates of political skill using the short TIPI measure. Next, we explore personality disorder or dark-side personality correlates of political skill. There is a growing literature which shows, paradoxically, that dark-side traits particularly associated with the dark-triad and Cluster B, are associated with management success (especially leader emergence; Furnham, 2018a, 2018b ). Third, we also explore self-esteem correlates of political skill as assessed by self-ratings. There have been a number of studies on self-ratings of intelligence (Furnham & Grover, 2020) as well as other factors like attractiveness and intelligence which are closely intercorrelated as well as reliable and robust measures of self-esteem. We hypothesise that like past studies, Extraversion, Agreeableness, Emotional Stability and Openness will positively predict political skill.
Additionally, this paper also re-examines the factor structure of the CFI with a more rigorous confirmatory factor analysis. Ferris et al. (2005) original paper used confirmatory factor analysis to measure how well this four-factor model fit the data and found that it had reasonable fit (met all but the CFI = .95 minimum criteria; Kline, 2016 ), but did not specify what correlation matrices or estimation method was used in the analysis. Other confirmatory factor analyses of this scale (Lvina et al., 2012; Smith & Webster, 2017) have either also not specified their methods or chosen maximum likelihood estimation which is suboptimal for the Likert-based, interval data received (Li, 2016; Mîndrilă, 2010) . This study’s analysis will use least square estimation and polychoric correlations in the confirmatory factor analysis, which have been shown to have the best accuracy for Likert data (Forero et al., 2009; Holgado-Tello et al., 2008) .
2. Method
2.1. Participants
506 adults were recruited to complete the questionnaire; all adults were British residents who were also born in the UK. Four participants failed an attention check in the study and were removed. Of the remaining 502 participants, 251 identified as male, 248 as female, and 3 as non-binary. Two hundred and seventy-six participants had a university degree. The average participant age was 28.8 years old (SD = 9.02) and was politically liberal (M = 6.11, SD = 1.78) and optimistic (M = 5.74, SD = 2.04; the latter two on 1 - 9 scales). Participants were all British nationals with a good command on English which was part criterion for taking part. Our analyses with 502 participants achieves .80 power with a = .05, where our regression effect sizes are above R2 > .0157, and correlation coefficients are >.124 (Faul et al., 2009) .
2.2. Measures
Political Skill Inventory (PSI; Ferris et al., 2005 ). This 18-item inventory was used which comprises four dimensions was used. They are: Networking Ability (α = .910), Interpersonal influence (α = .870), Social Astuteness (α = .836), and Apparent Sincerity (α = .784). A 7-point Likert scale recorded responses of this and the following two measures from “Strongly Disagree” to “Strongly Agree”. Mean scores were created for each factor in addition to a total political skill score (scale α = .925).
The Ten Item Personality Inventory (TIPI; Gosling et al., 2003 ). This 10-item scale measures five-factor model personality traits Extraversion (α = .712), Neuroticism (α = .702), Conscientiousness (α = .502), Agreeableness (α = .283) and Openness to Experience (α = .393) with 2 items each, one of which is reversed. While these alpha scores are low on some traits, this is by design. Our goal was not to maximise CFA fits or internal reliability, but to instead focus on validity while maintaining brevity. Indeed, the TIPI has been found to achieve better validity than other brief five-factor trait measures (Furnham, 2008) . Mean scores were created for each trait.
Structured Assessment of Personality Abbreviated Scale (SAPAS) (Lange et al., 2012) is a twelve-item screening interview for personality disorder. When clustering three are usually made which we combined: A: Odd/Eccentric (Self-defeating, Passive-aggressive, Paranoid, Schizoid, Schizotypal) (α = .67); B: Dramatic/Emotional/Erratic (Antisocial, Borderline, Histrionic, Narcissistic) (α = .74) and C: Anxious/Fearful (Avoidant, Dependent and Obsessive-Compulsive) (α = .63). It has been used in a number of studies (Merlhiot et al., 2014) .
Self-Evaluations (Furnham & Grover, 2020) were measured across three questions, which asked participants to rate their physical attractiveness, intelligence (IQ), and emotional intelligence (EQ) respectively, each on 1 - 100 scales. A mean score was calculated from the three variables which correlated .38 < r < .70 (α = .712).
2.3. Procedure
Ethical approval was sought and obtained (CEHP/514/2017). Data collection was conducted online, using Prolific as a source for participants to get a broad sample of the British population. Prolific is an alternative to the more widely used Amazon’s Mechanical Turk; both platforms have been shown to yield reasonable quality data, above that of university students (Peer et al., 2017) , but Prolific tends to have greater usability and quality (Palan & Schitter, 2018) . Participants were recruited on a first-come-first-serve basis, aiming for an even number of male and female participants. We specified they should be over 25 years old and in employment. Participants gave informed consent before answering any questions and were paid £1.70 for their time. Participants filled out the measures in the order presented above. This research was not funded by any outside body.
2.4. Analysis
R was used in this analysis to assess the data and conduct factor analyses (R Core Development Team, 2013) . Several of its packages were used. Multiple imputation used the mice package to replace the missing data following their identification (van Buuren et al., 2020) . The mvn package was used to test variable normality (Korkmaz et al., 2014) and the lavaan package was used to run the factor analyses (Rosseel, 2012) . SPSS 26 was used to calculate the regressions and correlation coefficients (IBM Corp, 2019) .
3. Results
3.1. Preliminary Analysis
Missing data were minimal, with 63 points missing over the 73 variables (.2%). Visual interpretation of the missing rate between data showed similarity across the survey, so data were assumed to be missing at random (Heitjan & Basu, 1996; Rubin, 1976) . Multiple imputation was used to replace the missing data.
The data were then screened for outliers using Mahalanobis distances; outliers could be present in the data as participants were paid a flat sum for participation, some may have rushed through the survey to increase their rate of reward without understanding the questions. Mahalanobis distance calculations returned 27 cases which exceeded the cut-off value (104.716, for χ2 (73) at p < .001). These were removed from the data set.
3.2. Factor Analysis
The factor structure of the Political Skill Inventory (Ferris et al., 2005) was then tested. As previously mentioned, unweighted least square estimation and polychoric correlations were used in calculation. The model had acceptable fit, meeting all of Kline (2016) ’s recommended thresholds, an improvement to that previously shown by Ferris et al. (2005) ’s original analysis (χ2/df = 1.78, CFI = .993, RMSEA = .041, SMSR = .053).
3.3. Descriptive Statistics and Correlations
First, Shapiro-Wilk tests were again used to analyse normality of the new mean variables; all variables other than SMS scores were significantly non-normal (p < .05), so non-parametric correlation coefficients were calculated. Spearman’s rho coefficients were chosen over Kendall’s tau to avoid inaccuracies caused by equal values in calculation, especially likely to between personality variables as were only calculated from two items (See Table 1).
All five (PSI total score, plus its four facets) correlations with self-evaluations were positive (r = .172 to .420). The pattern of correlations for the Big-5 showed Neuroticism was negatively associated with political skill (r = −.093 to −.286), but all the other factors positively associated with political skill. The strongest correlations were with Extraversion (r = .187 to .549). Whilst there were clear
![]()
Table 1. Descriptive statistics and spearman’s rho correlation coefficients.
**p < .01, *p < .05. Sex coded as Male = 1, Female = 2.
similarities between the four facets of political skill, Apparent Sincerity was the most different of the four in its associations and their strengths. The correlations between the PSI factors were high ranging from .28 < r < .65. On scales of 0 - 100, the average participant rated their physical health 69.3 (SD = 20.36), their intelligence 74.10 (SD = 13.60), emotional intelligence 68.50 (SD = 20.45), and physical attractiveness 59.59 (SD = 20.24). Personality disorder factor 1 was the only one of the factors to be significantly associated with all personality trait variables, with negative relationships with each. Factors 2 and 3 had comparatively weaker associations.
3.4. Multiple Regressions
Hierarchical multiple regressions were carried out first with each of the four factors and a total mean score of the political skill inventory as the outcome variables. Hierarchical regressions were chosen to see if adding additional personality traits and then personality disorders explained any further, incremental variance in the political skill variables. Ninety-five percent confidence intervals were generated. Influential cases were assessed using Cook’s distance, were the maximum value for any case across regressions was .08791, below the threshold of one (Field et al., 2012) . Durbin-Watson values ranged between 2.018 - 2.118, suggesting independent errors. Variance inflation scores ranged between 1.038 - 1.497, suggesting minimal multicollinearity (Bowerman & O’Connell, 1990) . Therefore, these results suggest that the regression models used meet their assumptions (Table 2).
Table 2 shows the results of the five step-wise regressions. Three things are apparent from these regressions. First, the predictor variables account for between 25% and 41% of the variance in the criterion variables. Second, in three of the five regressions, adding PD variables did not add any explained variance. Third, in four of the analyses Agreeableness, Extraversion and Conscientious were significant correlates (always positive), while in three Neuroticism was associated with lower political skill. Equally in four Self-Evaluation was positively associated with the political skill scores in four of the five analyses. The regressions suggested that similar personality factors were associated with three of the five regressions: Total score, Network analysis and Interpersonal influence.
4. Discussion
This study replicated and extended the literature in this field. First, we replicated studies that looked at personality trait correlates of political skill (Ferris et al., 2005; Munyon et al., 2015) , and gave further, stronger evidence for the factor structure of the PSI. The politically skilled individual is an Agreeable, Conscientious, Stable, and Extravert. It may surprise observers about these findings given the reputation of “political” people at work: that they are agreeable, meaning tender-minded, empathic and warm rather than the image of being cold, disagreeable and manipulative, which is more the image of being political rather than savvy or skilled (De Luca, 1999) . It is interesting to note that trait Agreeableness was not related to Social Astuteness though it was related to all of the other factors.
Next, sex and age seemed little related to any of the facets of political skill. It may be assumed that some older people have learnt to become politically skilled as they navigated their working lives. Equally it has often been speculated that there are distinct gender differences in office politics (De Luca, 1999; Furnham, 2018a) . However, this was not found in our results.
The study highlighted the advantage of analysing political skill at the facet level as the facets did reveal different patterns of correlations. Of the factors the one which often provokes most comment is “Apparent Sincerity” at a time when Authenticity at Work is often portrayed as a virtue. Apparent sincerity may be better conceived in terms of a social monitoring skill requiring the ability to be diplomatic and socially charming. Interestingly, in this study, two of the personality disorder factors were related to this factor.
The study added a number of new features to this literature. The first was to show that self-evaluations are related to political skills there are a number of explanations for this finding: self-confident people develop better political skills; the politically skilful become more confident as a consequence of using these skills; there is reciprocal causation between these two variables; the association
**p < .01, *p < .05.
**p < .01, *p < .05.
**p < .01, *p < .05.
**p < .01, *p < .05.
**p < .01, *p < .05.
is inflated by impression management. All of these explanations are plausible, which suggests that those interested in teaching political skills should consider the important role of self-evaluations in the process. It is a finding that is worth replicating using other measures of self-esteem and worth.
This study also introduced the idea that the personality disorders may be related to political skill. It explored the idea that, paradoxically, some disorders are associated with leadership emergence and short-term leadership success (Furnham, 2018a, 2018b) . However, the correlational results suggest that the disorders were negatively associated with political skill, but in 3/5 of the regressions, Factor 3 (Moving Toward People/Cluster 3/OCD and Dependent PD) was positively related to political skill. Again, this clearly suggests that political skill is very different from many popular conceptions of Machiavellian manipulativeness. Interestingly for Social Astuteness, the regression suggested that Cluster two PDs (Narcissistic, Anti-Social, Histrionic) were positive. This makes sense as these PDs are often associated with the ability to understand and exploit people’s weakness and vulnerabilities. There is indeed a growing literature on the paradoxical finding from studies which show dark-side traits are associated with work success as well as failure (Dilchert et al., 2014; Furnham et al., 2013) . Many think that some of these dark traits (e.g. Narcissism) are initially very useful at getting a job (i.e. leadership emergence) but that they are associated with long term management derailment and failure. Thus, at a subclinical level, dark-side traits may be beneficial in certain jobs at certain times in a person’s career.
Major limitations of this study included method invariance which usually inflates associations, and the use of very short measures (TIPI, PDs) which can obscure subtle and important relationship between variables. Ideally, a study would use measures of the FFM that assessed domains as well as facets. Further we used a very short measure of the PDs that had an unclear factor structure. Ideally, a robust measure of the PDs would have been used with a confirmed factor structure. More importantly perhaps, it would be desirable to have some observer as well as behavioural measures of political skill, as many findings could be inflated by impression management and self-delusion.