Dr. Jekyll or Mr. Hyde? President Donald Trump’s Personality Profile as Perceived from Different Political Viewpoints

The present research used an empirical, crowdsourced trait profiling approach to describe the personality of President Donald Trump (hereafter Trump) that accounts for political views. We recruited participants who voted for Hillary Clinton (N = 120; hereafter Clinton) and Trump (N = 118), and asked them to rate Trump’s personality on the 30 facets of the Five Factor Model. Participants also provided perceived helpfulness and harmfulness ratings of the facets before and after the election. We treated these facet level ratings as trait profiles, which were transformed into estimates of personality disorders (PDs) and complex trait-based constructs based on expert profiles. Results suggest only modest agreement between Clinton and Trump voters on Trump’s personality. Clinton voters perceived much greater antagonism, lower conscientiousness, and higher levels of impairment in Trump’s personality than did Trump voters who primarily perceived high levels of extraversion and emotional stability (i.e., low neuroticism). At the level of PDs and complex traits, there was some convergence with both groups seeing Trump as high in narcissism and psychopathy.


Introduction
There is currently a great deal of debate and discussion surrounding the personality of President Donald Trump (hereafter Trump). These debates are characterized by four primary themes. The first is the issue of Trump's narcissism that has repeatedly been discussed in public forums (e.g., Campbell, 2016;Krugman, 2017;Nutt, 2016;Singer, 1997). The second is the issue of Trump's impulsivity; this is not necessarily linked to narcissism, and rather has been discussed regarding hypomania and psychopathy (e.g., Brown, 2017). The third is the issue of impairment, or the perception that Trump's personality traits harm his interpersonal life and performance as president (i.e., love and work). Indeed, there have been some calls for Trump to be removed from office under the auspices of the 25 th amendment (section 4) because of this perceived impairment (e.g., Silver, 2017). Fourth, there is a debate about making psychiatric diagnoses, with psychiatrists and psychologists discouraged from diagnosing Trump from "afar" by the Goldwater rule (see Lilienfeld, Miller, & Lynam, 2017;cf. Davis, 2017), although this has not discouraged some mental health professionals from doing so.
In the current research, we used a crowdsourced trait profiling approach to estimate Trump's personality and impairment in a way that is quantitative and can account for political differences. Our approach is grounded in a realistic assessment model (RAM;Funder, 1995). We start from the position that perceptions of personality are open to and influenced by observed behavioral, reported, or trace data; these include peer-reports, thin slices of behavior, social media activity, selected physical environment (e.g., offices), etc. We do not assume that these personality observations are necessarily fully accurate, as they can be biased by the information available to perceivers as well as the perceivers' own biases. In the present study, we attempt to investigate the degree of this potential bias through recruiting diverse observers. Specifically, we recruited individuals who reported voting for Trump as well as those who voted for Hillary Clinton (hereafter Clinton) from Amazon's Mechanical Turk (see Mason & Suri, 2012;cf. Goodman, Cryder, & Cheema, 2013) and asked them to rate Trump's personality by collecting their perceptions of Trump's standing on general traits (i.e., Five Factor Model [FFM] facets). Previous work by Wright and Tomlinson (2018) suggests that voter identification is related to differing perceptions of Trump's personality, specifically in the trait domains of Agreeableness and Conscientiousness, such that Clinton voters (but not Trump voters) viewed Trump as being very low in these domains. Building on this work, we gathered FFM facet-level data regarding perceptions of Trump's personality, and we also assessed perceptions of the degree to which these traits caused impairment to Trump pre-and post-election.
Additionally, by gathering perceptions of basic trait profiles, we can estimate the degree to which perceptions of Trump's personality comport with complex trait profiles (e.g., narcissistic personality disorder, psychopathy) without directly asking about them. We utilized a profile matching approach, such that we compared trait profiles of Trump generated by Clinton voters and Trump voters to existing expert trait profiles of various personality constructs (e.g., personality disorders [PDs]). This allowed us to estimate the variability in perceived personality by voter identification, as well as the variability in perceived impairment at both the overall and facet level across time and context (i.e., pre-post election; occupational vs. interpersonal functioning).
We next invited 200 Clinton voters and 200 Trump voters to participate in the trait and impairment ratings and received 120 responses from Clinton voters and 118 responses from Trump voters. In the current sample, Clinton voters (mean age = 41.5, SD = 12.7) and Trump voters (mean age = 39.2, SD = 12.8) were not significantly different in age and gender composition (73% and 64% female, respectively). These participants provided ratings of Trump's personality traits and the associated helpfulness and harmfulness of each trait in multiple psychosocial domains. We paid participants $.05 for completing phase 1 and $1.50 for phase 2; all aspects of this research were approved by the Institutional Review Board.

Impairment Ratings
The helpfulness and harmfulness of Trump's standing on each FFM trait were measured as it pertained to pre-election Trump, post-election Trump, and Trump's personal life. For example, a sample item was "How much did Trump's anxiousness or lack of anxiousness help him in the 2016 presidential election (i.e., before he was president)?". Items were rated on a 1 (Not at all) to 7 (Very much) scale. Results are presented as pre-election scores minus post-election scores, such that a positive score indicates that the trait has been more helpful/ harmful after the election.

Profile Matching Analyses
To conduct profile matching analyses, we considered the column of FFM facet ratings by Trump and Clinton voters to be personality "profiles" that are representative of how these voters view Trump's personality as a whole. Then, by conducting double entry intraclass correlations between these ratings and expert ratings of other relevant personality disorder profiles (e.g., Narcissistic PD), we are able to quantify the level of absolute similarity between Trump's personality profile as generated by Trump and Clinton voters and expert-generated profiles of commonly studied personality constructs (see McCrae, 2008).

Within Group Agreement
We first examined the within-group agreement for the Clinton and Trump voters at the overall and facet level. There was modest agreement about Trump's personality trait profile from Clinton (IRR = .47) and Trump (IRR = .33) voters. At the facet level, Clinton voters also showed a pattern of generally higher agreement than Trump voters, suggesting that Clinton voters had a more unified perception of Trump than did Trump voters (Supplemental Tables 1 and 2), particularly for traits from the domain of Agreeableness.

Mean Facet Ratings by Voter Identification
We next compared the mean facet level ratings of Trump separated by voter identification (see Table 1). We calculated product moment and intraclass (ICC) correlations between the FFM ratings of Trump made by Clinton voters and Trump voters to provide an estimate of the agreement between personality profiles of Trump. The profiles were moderately similar (r = .43; r ICC = .26). An examination of the facets show clear disagreement between Trump and Clinton voters (see Figure 1) on traits of Agreeableness and Conscientiousness such that Trump voters see Trump as much more Agreeable and Conscientious than do Clinton voters, with the smallest disagreement on the facet of Modesty, with both groups viewing Trump as very low on Modesty. The differences in perceived traits were substantial; Cohen's d's were large for 18 of 30 facets. The mean trait differences in descending order were for Conscientiousness (mean d = 1.46), Agreeableness (mean d = 1.19), Neuroticism (mean d = .83), Openness (mean d = .68), and Extraversion (mean d = .64).
Next, we examined which traits were seen as most prototypical of Trump based on voter identification (i.e., furthest away from midpoint; Table 2). In general, Clinton voters rated Trump as low in facets of Agreeableness, particularly Modesty, Tender-mindedness, Altruism, Compliance, and Trust, as well as facets Note: Facet values ≥ 4 and ≤ 2 are bolded; 1 Cohen's d presented as absolute value, and large effect sizes (i.e., values ≥ .80) are bolded; 2 this superscript indicates that independent sample t-tests suggest that the mean ratings for a given trait as rated by Clinton and Trump voters are not statistically significantly different at p < .01.
of Conscientiousness, particularly Deliberation and Dutifulness. Clinton voters also rated Trump as high in Assertiveness (from Extraversion) and Anger/hostility (from Neuroticism) and low in Warmth (from Extraversion). Trump voters rated Trump as high in Assertiveness and Impulsiveness, as well as low in Modesty. However, they also rated Trump as particularly low in Depressiveness and Vulnerability (both from Neuroticism), high in Gregariousness and Activity (both from Extraversion), high in Achievement and Dutifulness (both from Conscientiousness), and high in Straightforwardness (from Agreeableness).

Profile Matching
The FFM trait profiles of Trump were matched to existing expert-based FFM profiles of PDs (e.g., Lynam & Widiger, 2001). As can be seen in Figure 2, Clinton voters' average trait profile is highly similar to expert ratings for a suite of related constructs: narcissistic PD, psychopathy, and antisocial PD. The average Trump profile, as rated by Trump voters, also matched these constructs, but at a descriptively lower level. Similarly, ratings by Clinton and Trump voters alike were highly dissimilar to expert ratings of dependent PD, avoidant PD, and schizoid PD, although the similarity was smaller for Clinton voters. The highest profile similarity for Trump voters was the expert-rated profile of successful psychopathy, which differs from more traditional psychopathy in that it assumes a lack of major difficulties in conscientiousness/disinhibition (see Mullins-Sweatt, Glover, Derefinko, Miller, & Widiger, 2010), unlike it less adaptive counterpart.

Impairment
Tables 3 and 4 provide ratings of perceptions of how helpful and harmful each trait was to Trump before and after the 2016 U.S. presidential election, as well as change scores across that time period. 1 In general, Clinton voters rated Trump's traits, on average, as being moderately helpful prior to the election (mean = 4.28; SD = .72) but generally rated these traits as less helpful following
In general, there were small differences in perceptions of helpfulness of Trump's personality pre-election ( 77) after the election as Clinton voters perceived these traits as being much less adaptive during the governing rather than campaigning period.

Discussion
In the current study, we estimated President Trump's personality using ratings from Trump and Clinton voters who rated Trump on the 30 facets of a popular and well-validated model of general personality traits, which allowed us to examine differences in perception of Trump's personality across political groups. We then compared the general trait profiles to configurations of traits associated with various expert personality and personality disorder profiles. Finally, we tested perceptions among these two voter groups of whether these traits were viewed as more helpful/hurtful before or after the election, as well as the degree of interpersonal impairment associated with each trait.
We found a modest degree of relative agreement (r = .43) and more limited absolute agreement (r ICC = .26) between profiles for each type of voter. At the specific trait/facet level, agreement was found primarily with traits related to assertiveness, immodesty, and +impulsiveness. Overall, there were dramatic differences in the personality traits perceived across the groups with 18 of 30 facets demonstrating a large difference (Cohen's d > .80). In fact, for several traits the group means differed by more than 1.5 standard deviations; these traits included ratings of   perceived him to be avergage or relatively high on them, replicating previous findings (Wright & Tomlinson, 2018). Reflecting the modest agreement between the two trait profiles of Trump, both mean FFM profiles were associated with pathological trait profiles associated with narcissistic, antisocial, and psychopathic ("successful" and "non-successful") PDs. The degree of match with these constructs was somewhat higher for the Clinton voters, which is not surprising given their stronger endorsement of antagonistic and disinhibited traits, both of which are central to these PDs (e.g., Lynam & Widiger, 2001). Of note, we believe that esteem for Trump's salient indicators of success (e.g., wealth, celebrity status) are likely contributors to Trump voter's stronger endorsement of his personality as emblematic of successful psychopathy. Previous research has examined the relations between psychopathy, narcissism, and presidential performance (Lilienfeld et al., 2012;Watts et al., 2013) finding that these traits are often trade-offs for performance. Aspects of both are linked to better performance regarding ratings of leadership, persuasiveness, communication in some domains but poorer performance regarding risk-taking and ethical lapses/impeachment resolutions. Narcissism in leadership is also linked to conflict and turbulence within an organizational system (e.g., government; Sedikides & Campbell, 2017), and given their theoretical and empirical overlap, this link likely extends to psychopathy. Consistent with the perceived narcissism identified by both Clinton and Trump voters, one might expect that Trump's traits would be viewed more favorably initially than over time -as this is the typical perception of more narcissistic individuals in leadership positions (Campbell & Campbell, 2009;Ong, et al., 2016). There was evidence of this trend in these data such that both sets of voters believed Trump's traits were more helpful in the campaign than they have been in governing. Although both groups of voters saw his traits as being less helpful and more impairing in his role as President vs. party nominee, there were substantial differences in size such that the "prepost" comparisons for Clinton voters were enormous (helpful: d = 2.42; hurtful: d = 2.75) but more moderate for Trump voters (helpful: d = .76; hurtful: d = .44). This perceived increase in impairment is also consistent with the actual polling data, showing that Trump's approval ratings have decreased since an initial brief honeymoon immediately post-election (FiveThirtyEight, 2018).

Implications
This work shows the very different ways that individuals can perceive personality. Indeed, even within Trump voters, our results suggest there is only modest within group agreement on his personality profile. One possibility this raises is that there may be different clusters of Trump voters, such that some pockets of voters view Trump in a more or less favorable light, although the current analyes cannot speak to this possibility. A second possibility is that there is just merely less consensus about features of Trump's personality that are not especially salient (e.g., facets of Openness), and thus there is more variability and poorer reliability within these ratings. Alternately, the reliability scores for the FFM domains Agreeableness and Conscientiousness are relatively good, suggesting that there is more consensus on Trump's standing on these traits for both Clinton and Trump voters. What we do not know, however, is the source of these differences. Individuals could be seeing different representations of Trump (i.e., consuming news from different sources) and/or distorting the representations that they see in a way that fits their preferred image of Trump. A similar bias is found throughout the social psychological literature, whereby people gather and report information about themselves or a related group (e.g., a sports team) in a way that is self-enhancing (e.g., Hastorf & Cantril, 1954). Likewise, in political psychology, an individual's tribal bias (e.g., Republican, Democrat) has been shown to dramatically influence their perception of political events (Bartels, 2002).

Limitations
This research also has several limitations. First, we are using personality estimates -these are not expert judgments, or well-informed informant ratings (i.e., Connelly & Ones, 2010). Our approach is based on the idea that multiple raters will converge on a reasonable profile, but it is possible that this crowd-sourced estimate is far off what experts (e.g., academic biographers of Trump) would determine. Additionally, we are using single items to estimate facet level data about Trump's personality, and it is likely that using a more robust measures may yield profiles that are more in line with expert ratings. Fortunately, two recent studies have asked experts in U.S. politics and personality psychology to provide ratings of Trump's personality (Nai & Meier, 2018;Visser, Book, & Volk, 2017). Results of these studies suggest that like community members in the current study, experts in U.S. politics and personality psychology alike rate Trump's personality as very low in Agreeableness, low in Conscientiousness, and high in Extraversion. In fact, when Wright and Tomlinson (2018) compared the Clinton voters' ratings of Trump to the expert ratings from Nai and Meier (2018), they found the experts rated Trump as even lower in Agreeableness and Conscientiousness than did the Clinton voters.
Second, due to the deep, growing partisan divide within the current United States political realm (e.g., Doherty, 2017), a related limitation is that the Trump ratings provided by Trump and Clinton voters may be somewhat determined by a motivation to portray Trump (and concordantly, the Republican party) in a more positive or negative light, respectively. In support of this point, a recent survey of politically moderate voters suggests that they tend to view Trump's personality traits as somewhere in between the view of Trump voters and Clinton voters (Wright et al., 2018). Third, the ratings were collected from MTurk, which has proven to be a reasonable source of data (e.g., Shapiro, Chandler, & Mueller, 2013), but cannot be viewed as a random sample of American voters especially since MTurk participants tend to be relatively young (e.g., mid-30s), well-educated, and unrepresentative of the full racial, ethnic, and socioeconomic status of the American population.
Fourth, while we can glean from these data where sources of agreement (e.g., modesty) and disagreement (e.g., conscientiousness) lie, there is no way to determine if either set of ratings are closer to the "truth" as to the President's personality. Our raters only have access to public reports of Trump and his public actions, and thus they are rating Trump's persona, or public personality, but not necessarily his private personality. We have no way of knowing if there is a private Trump who remains hidden from view, and, if so, how the private Trump might differ. However, of note, there are multiple journalistic and bibliographic efforts that attempt to describe Trump's