Personality Traits and the Propensity to Vote Strategically: An Experimental Inquiry. Delia Dumitrescu

Transcription

1 Personality Traits and the Propensity to Strategically: An Experimental Inquiry Delia Dumitrescu André lais Université de Montréal Department of Political Science Abstract: The paper examines whether the propensity to vote strategically is affected by one s personality through an original experiment in which college students participated in a series of five FPTP elections. We focus on the ig Five personality traits. We define a strategic vote as one based partly on one s preferences and partly on one s expectations about the likely outcome of the election, as when one is inclined not to support a party/candidate whose chances of winning are perceived to be nil or very weak. We find that some personality traits (in particular conscientiousness) do affect how people perceive parties chances in elections, and they have a direct impact on the decision for whom to vote. Paper prepared for presentation at the EPR Workshop Personality and Turnout, Münster, March 22-27, This is work in progress (please do not cite without permission). 1

2 The paper examines whether the propensity to vote strategically is affected by one s personality through an original experiment, described below, in which undergraduate students were invited to participate in a series of five elections held under first past the post (FPTP). We define a strategic vote as one that is at least partly based on the motivation to have one s vote count (ox 1997). More specifically, in the case of FPTP elections, a strategic vote is one based partly on one s preferences and partly on one s expectations about the likely outcome of the election, that is, one is inclined not to support a party/candidate whose chances of winning are perceived to be nil or very weak (Abramson et al., 2009). We look at the relationship between the ig Five personality traits and strategic voting. According to the ig Five approach, there are five basic personality traits: openness to experience, conscientiousness, extraversion, agreeableness, and emotional stability. These traits offer a concise approach to personality, in so far as they encompass characteristics of almost any personality construct (Funder, 2001, 200), and they have become the dominant typology in research on personality (John et al., 2008). A number of studies have recently showed correlations between personality traits and vote choice, ideological orientation or political engagement (Mondak and Halperin 2008; Gerber et al. 2008, 2009; Vecchione and aprara 2009; Mondak et al. 2010). Recent research has also identified a number of interactive effects between individuals personality traits and context affecting people s political engagement (Mondak et al. 2010). We propose to consider a particular type of political behavior: the decision to strategically desert one s preferred option if and when that option is perceived to be unviable. We conceive of the decision of how to vote as a function of the context in which one finds themselves, one s personality, and their interaction. Given the complexity implied by strategic voting (Abramson et al., 2009), this interaction is of particularly strong interest for our dependent variable, vote choice. Some contexts are more conducive to strategic voting than others, and the decision to vote strategically requires people explicitly to take these contexts under consideration. onsequently, we first approach the inquiry into personality and strategic voting from the angle of perceptions about the context: How does personality affect expectations under various electoral contexts? We then examine how personality affects how people act, given these perceptions. In a nutshell, we find that some personality traits (in particular conscientiousness) do affect how people perceive parties chances in elections, and they have a direct impact on the decision for whom to vote. The paper proceeds as follows. We lay out some expectations about strategic voting and personality. We then describe the experimental setting and present a number of descriptive results. The next section focuses on strategic voting in our sample, by examining the relationship between preferences, perceptions about the race and vote choice. We then look at the indirect impact of personality on the decision to vote strategically, by means of their effect on perceptions as a function of the characteristics of the race. The last section examines the direct impact of personality traits on vote choice in each context. 2

3 Personality and Political ehavior: Theoretical Premises and Expectations Research on personality has long recognized the importance of the interaction between contextual factors and individuals personality characteristics for behavioural responses (Funder 2008, p. 568). Mondak et al. (2010) make a strong argument for the importance of the interaction between personality and contextual factors in political participation decisions. In particular, they find that more conscientious people are more likely to participate the more important they perceive participation to be. Similarly, people with high scores on agreeableness appear to be more likely to shy political disagreements, no matter how big their discussion network, whereas less agreeable individuals are substantially more likely to be exposed to political disagreements the bigger the number of their discussion partners. Yet, as Mondak et al. (2010, 7) note, empirical studies focusing on this interaction have been relatively rare. We seek to add to this stream of research by experimentally creating a number of different electoral contexts in which people have to make a vote choice influencing the allocation of real resources to a cause, which we describe in the next section. Using this experimental design, we are able to look at how personality affects how people behave in various electoral contexts, more specifically whether some personality types react differently from others to variations in the competitiveness of the election. We advance minimal expectations at the outset with regard to which personality traits affect the decision to vote or not strategically. Previous research would indicate that particular attention should be given to three traits: conscientiousness, agreeableness and openness. According to John & Srivastava (1999, 121, cited by Gerber et al., 2009, 7) conscientiousness describes socially prescribed impulse control that facilitates task and goal directed behaviour such as thinking before acting, delaying gratification, following norms and rules, and planning, organizing and prioritizing tasks. Since strategic voting requires voters to undertake a more serious analysis of the context than just voting for one s preferred option, we expect conscientiousness to play a role in this process. Similarly, according to John & Srivastava (1999, 121), agreeableness * + includes traits such as altruism, tender-mindedness, trust & modesty. We therefore expect people who score higher on agreeableness to be more sensitive to situations when successful coordination among voters is critical to maintain their preferred electoral option in the race. Finally, openness has been linked to greater propensity towards liberalism (arney et al, 2008), leading us to expect it to be correlated to preferences for parties in the experiment, as we describe below. We specify these expectations further in the context of the experimental setting. The Strategic Voting Experiment Two hundred undergraduate students in the department of political science at the Université de Montréal were recruited to take part in a study on elections. The study took place over the internet. Participants were required to go at a specified address and vote in a series of elections, as well as answer some questions about their political preferences, interest, and, importantly for this study, personality. The study took about 15 minutes to complete. 3

4 The experimental setup involved putting the respondents in a context in which they would vote to allocate a sum of money ($600) to either an environmental NGO or to be distributed to the participants themselves. More specifically, participants had to choose between three parties, generically denoted as A,, or. The victory of each party entailed a different distribution of the sum: if Party A won the election, $500 would go to the NGO and $100 to participants (resulting in an additional gain of 50 cents per participant). If Party won the election, the sum would be split in the middle, with $300 going to the NGO and $300 to participants, and if Party won the election, the NGO would receive $100, with the remaining $500 being distributed to participants. The NGO was generally known to participants (only 7% of the sample said they had not heard of it before). The use of a real and well known NGO, as well as the choice of a salient political issue, the environment, was intended to encourage participants to consider the collective consequences of their vote, and increase the external validity of the study. Prior to each election, respondents were made aware that 9000 votes had already been cast, that two hundred more votes were to be cast by the participants (including themselves) and that the winner would be the party with the most support based on the cumulated 9200 votes. The distribution of these 9000 ballots varied by election and constituted the main experimental manipulation. To examine the propensity of casting a strategic vote as a function of a party s chances of winning, the experiment used a 2(Distance between A and : small v. large) x 2 (Distance between and : small vs. large) within subjects design. A fifth election placed Party A in a distant (impossible to win) third place. With the exception of the fifth election (which always came last), the order of the other four was randomized across individuals. Respondents were reminded that each election should be treated independently. The payoff for both the NGO and the participants would be the mean payoff across all five elections. The five 9000 vote distributions appear in Table 1. In essence the manipulation was intended to examine most particularly the strategic vote behavior of Party A supporters, and the manipulations are listed, in the order of the increased incentive for Party A supporters to consider defecting to Party. In condition lose A, lose (), there is a close contest between both A and and and, and so there is little reason for not voting sincerely for one s preferred party. In condition Far A, Far (FF), Party A has practically no chance of winning (this would require all 200 study participants to vote for that party and then that party to win the draw), and so party A supporters should be very tempted to vote for party. In condition lose A, Far (F), there is a close contest between A and for second place and in condition Far A, lose (F), there is a close contest between and for first place while the distance between A and is the same, relatively large though not huge. Finally, in the fifth election, party A is certain not to win, and so there seems to be no reason to stick with party A. [Table 1 here] Generally speaking, we expected perceptions of Party A s chances of winning to decrease as the number of votes required to ensure its victory increased. We also expected Party A supporters to be more likely to cast a strategic vote, the lower the chances of this party to win the election. In terms of personality, we had three expectations. First, we expected the more conscientious to give lower estimates for Party A s chances of winning the greater the strategic vote incentives. We also 4

5 expected them to be less likely to vote for Party A the greater the pressure on Party A supporters to defect to a less preferred option. Second, more open people should be more likely to prefer Party A to all others. Third, we expected people high on agreeableness to be more likely to stick with Party A in the F and FF election, where significant coordination among voters would be required to hand this party a victory, than those who score lower on this personality trait. Descriptive Findings: Election Results and Perceptions Table 2 presents the outcome of the five elections. In each of the five elections, Party gets the least votes among participants and Party A gets the most (except in the fifth). Party A gets its greatest support in the lose A, lose election and is weakest in the fifth; the reverse holds for Party. Support for Party A is the most variable across elections (range of 45) and support for Party is the most constant (range of 12). [Table 2 here] Party wins three of these elections, Party A and Party each one. The party that received the least support among participants, Party, wins most elections, because it was ahead to start with, considering the 9000 votes that had been cast. This suggests some lack of coordination among the participants, a point that we revisit below. The main contrast in the participants vote is between the first two elections and the last three. In the first two elections, and F, Party A gets more than twice as many votes as Party while the vote is almost evenly divided between these two parties in the F, FF and fifth elections. The first two elections are the only ones where there is only a small distance (in the 9000 votes already cast) between Party A and Party. This suggests that whether Party A is close or not to Party matters more than the distance between and A, or and. The other piece of evidence that is required is participants perceptions of each party s chances of winning. We asked participants to indicate how they rated those chances on a 0 to 10 scale, where 0 means no chance and 10 means certain to win. These ratings are standardized so that they add up to 1 for each individual. Table 3 shows the mean score given to A,, and in each of the five elections. With one exception (F and F, which we discuss further), the mean ratings for A are significantly different from each other (t>3.74, p<.000), indicating a significant treatment effect on perceived chances of victory for the most preferred party in the sample. The reverse applies to the perceived chances of Party, which increase from the first to the fifth elections. Interestingly, the perceived chances of Party hardly vary across elections. Finally, we note that many participants were unwilling to give a rating of 0 to Party A s chances in the fifth election even though they were objectively absolutely nil. [Table 3 here] These descriptive findings allow us to conclude that context had a significant effect on the vote, and that this effect is unequal across the sample. On the one hand, perceptions of Party A s chances of winning decrease, as we would expect, the farther behind this party is in the race. Nevertheless, some people 5

6 appear to cast a sincere vote for A in all contexts, and some people refuse to accept that in some contexts Party A had zero chances to win the race. To examine to what extent participants vote strategically, and how personality affects the strategic vote calculus, we conduct a series of multinomial estimations of vote choice. Measures: Personality and Preferences Personality. Our main interest in this paper concerns the relationship between personality traits and vote choice. As indicated above, we have decided to focus on the ig Five personality factors: openness to experience, conscientiousness, extraversion, agreeableness, and emotional stability. To that effect, we use the TIPI scale (Gosling et al. 2003). Participants were asked to indicate the extent to which pairs of adjectives associated with each trait describe them on a scale from 1 to 7. Table 4 presents the distribution of their responses. The advantage of the battery is its high degree of validation in multiple studies. In our case, however, the necessity to translate the scale into French might have affected its reliability, and this might have reflected into relatively lower correlations than usual for certain items in the scale. All the correlations are nevertheless significant, allowing us to use the two combined items for each personality trait as indications of participants personality characteristics. [Table 4 here] Preferences. Participants were asked at the outset on the study to rate their preference on a scale from 0 to 10 for each party winning the election. Unfortunately this allowed us to compute strong preference only for 70% of the sample, limiting significantly the utility of these measures for the study of strategic voting. We therefore use one of the votes as a proxy for party preferences, specifically, participants actual vote in the lose A lose election. We use the election as the reference election, because in this election, as can be seen in Table 3, each of the three parties is perceived to have some chance of winning. We assume that the vote in that election is sincere and that strategic considerations come into play only in the other elections, in which there is greater distance between at least two of the contenders. Strategic hoice across Elections Table 5 shows the relationship between vote choice in the election and the others. Readers should keep in mind that the first four elections were held in a random order. [Table 5 here] Table 5 indicates that indeed some of those who vote for party A move to another party, usually party, and that the propensity to desert increases from the second to the fifth, as the distance (among votes already cast) between Party A and Parties and increases. This is exactly as we would expect. The weakest Party A appears to be, the highest the likelihood of strategic desertion. Note, however, that even in the fifth election, where party A has absolutely no chance, 60% of its supporters stick with it. There is also some movement away from party or party, which suggests that something else is going on. Still, we find it reasonable to infer that most of the shift in vote choice between the first election and 6

7 the following ones is due to strategic considerations arising from differences in the distribution of the 9000 initial votes. Table 6 presents a multinomial estimation of vote choice in all but the first election which we use to infer the participants sincere preferences. This is a simple strategic vote choice model in which the decision to vote for party or rather than party A is assumed to depend on one s preference (measured by vote choice in the first election) and on the parties perceived chances of winning. As one would expect, those who like party A are less inclined to move to party or. What is more important for our purposes, however, is that we see that, even controlling for these preferences, vote choice is affected by perceptions of competitiveness, in accordance with the strategic vote model: the better party A s perceived chances of winning, the weaker the inclination to vote for party or. 7 [Table 6 here] The implications are illustrated in Figure 1. We focus here only on those who voted for party A in the election. The figure shows how the propensity to stick with party A is conditional on whether the party is perceived to be a viable option or not. More precisely, the predicted probability of voting for party A among its supporters is typically only about 47% when the party is perceived to have no chance at all and increases to over 70% in almost all elections when the perceived chances are over 30% (with the exception of FF and the fifth election when this probability decreases to about 60%). These results clearly support our strategic vote choice model. [Figure 1 here] The next step is to examine how personality may affect participants behavior in these elections. onsequent with our model, personality may affect participants preferences, their perceptions of the parties chances of winning, and their final decision. We examine each of these three possibilities. Preferences and Personality The first question that we examine is whether there is a relationship between personality and sincere preferences. Given the environmental theme of our experiment and its ideological implications, we would expect more open participants to side with Party A, as compared to those who are less open. As already noted, we use vote choice in the election, which we assume to be devoid of strategic considerations, as our measure of preference. [Table 7 here] Model 1 in Table 7 shows that those who score high on conscientiousness are more prone to vote for Party A while those high on agreeableness are more inclined to vote for Party. ontrary to our expectations, openness seems not to be related to party preference. This could well be an artifact of the relatively low correlation level for the openness scale. At the same time, an alternative explanation for this result lies in the sample itself: college students might be particularly pro-environmental irrespective of personality traits, as suggested by the fact that fully 60% of the sample votes for Party A in the election. One would assume that those who care the most about the environment are the most prone

8 to support party A, since more money would go the environmental group Equiterre if Party A wins the election. To check this assumption, we constructed a 10 point scale of pro-environmental attitudes (including individuals environmental policy positions and the emotional attachment to the environment). 1 The correlations between this environmental scale and personality traits are insignificant (under.07 for each trait). As expected however, we find indeed that 70% of those above the median on the environment scale vote for Party A, against 45% of those below the median. Moreover, environmental attitudes are a significant predictor of vote choice when we include them in the estimation of vote choice in Model 2. The strong pro-party A coefficient for conscientiousness is more unexpected. This coefficient stays significant even when we include environmental attitudes as a predictor for party preference in the election in our analysis. This indicates that support for Party A is very strong among those who are both conscientious and pro-environment. onsidering that conscientiousness has been linked to more conservative and sometimes authoritarian ideological orientations (see Gerber et al. 2009, for review), this result is surprising. Our interpretation is that conscientious individuals might be more sensitive to the practical consequences entailed by each party winning the election in the contest, where any result was possible. More specifically, a victory of Parties and would only increase participants gains by an insignificant amount (up to 2.5$), but would significantly limit the gain (and subsequently the practical use of the money) for the environmental NGO Equiterre. Mondak et al s (2010) results indicate that more agreeable people are more likely to avoid disagreements. In our case, they might be more reluctant to prefer a party which might be perceived as antagonizing, and this is one likely explanation accounting for the relationship between agreeableness and preference for Party, which stays significant even when controlling for environmental preferences. Shy of conflict, more agreeable people may feel more at ease with the middle position represented by Party, which ensures that neither participants, nor Equiterre receive any special treatment in the allocation of money. Perceived hances and Personality To what extent do personality traits influence perceptions? We start with simple correlations between perceptions of chances of A for each election and the five personality traits in the sample. We focus on Party A only, as this was consistently trailing the other two. These correlations are presented in Table 8. [Table 8 here] 1 The first environmental question was: There s a lot of talk about greenhouse gases nowadays. Some say that the 3% reduction target proposed by the anadian federal government is more than sufficient to prevent future global warming (these people are at one end of the scale below, at point 0 ). Others say that this 3% reduction target is totally insufficient if anada wants to fight global warming. (These people are at the other end of the scale below, at point 10). Of course that other people have opinions in between these points. Where would you place yourself on this scale?. The second question was: Do you agree or disagree with the following statement: When I think about the environment, I feel extremely emotional, with 0=Totally disagree, 10=Totally agree. 8

9 The results generally confirm our expectations with regard to the role of conscientiousness. People high on this trait are significantly more likely to estimate Party A s chances to the downside in three of the five elections. The fifth election is one of them, where Party A s chances were effectively nil. One election in which this correlation does not reach significance is the sincere one, where Party A could win, and indeed it did so easily. In the other election, Far A, Far, this correlation approaches significance (p=.06, one tailed). Two more traits exhibit significant correlations with perceptions: agreeableness and emotional stability. More agreeable participants are less likely to credit Party A with chances to win the election in the sincere vote and in the Far A, lose vote. In this latter election, greater emotional stability is also conducive to more conservative estimates of Party A s chances of winning. Interestingly, F is also the election when Party, which we presume to be the second option for most voters, has the highest objective chances of winning the election against Party. Yet, the correlation between agreeableness and perceived chances of winning for Party A in the F context runs contrary to our expectations, as it appears that more agreeable people are more distrustful of the success of a possible coordination among Party A supporters. In the next step of the analysis, we retain the three most important traits appearing to influence perceptions, and examine their joint influence on perceived chances of Party A s winning each election. Table 9 presents the OLS regression results, with and without party preference as a control. [Table 9 here] The impact of conscientiousness is robust when we consider the simultaneous impact of the three personality traits. Highly conscientious people are significantly more likely to rate Party A s chances lower in three out of four elections. The effect persists even when controlling for Party A support. oth emotional stability and agreeableness have the same negative effect on perceptions of Party A s chances in the F election, but agreeableness loses its significance. When we observe the joint impact of these three traits, agreeableness becomes positively associated with perceptions in two other elections: FF and the fifth. The FF one in particular places the highest demands for coordination on voters to ensure that Party A stays in the race, and goes in the direction of our expectations. The effect of this trait in the fifth election is more puzzling, as Party A has absolutely no chance of winning. Personality and Strategic Voting There is therefore ground to posit a weak albeit significant indirect impact of some personality traits on perceptions of Party A s chances of winning (with conscientiousness being the most robust of these effects). We now turn our attention to their direct impact. [Table 10 here] Table 10 presents the results of the multinomial estimation of vote choice with the three personality traits as direct predictors, and vote for Party A as the reference outcome. Two models are estimated: one including perceptions of Parties A and s chances, the other without. The models also control for 9

10 party preferences using the vote in the lose A, lose election. With the exception of the FF election, conscientiousness is significantly and positively related to defection from Party A. As we would expect, this result is more robust when perceptions are not included in the analysis. This result mirrors the previously found influence of this personality trait on perceptions of Party A s winning. While conscientious participants appear to act based on their perceptions of the context, it is also interesting to note that the direct influence of conscientiousness on defection from Party A does not necessarily result in strategic voting. In all elections, more conscientious people end up defecting to Party too. This result could be explained by the fact that about a third of the respondents in the sample did not have a strict order of preference between the three parties, and particularly did not express a strong preference between Party and Party. At the outset, we expected agreeableness to prevent people from defecting when voter coordination is critical to maintain their preferred party in the race (particularly in the Far A, lose and Far A, Far contexts). Despite their more conservative estimations of Party A s chances of winning, more agreeable individuals appear to be indeed more likely to stick with this party in the F election. Yet, agreeableness appears to no effect on defection in the FF election, despite previous results which indicated a propensity among more agreeable people to rate party A s chances of winning higher than the rest of the respondents. This would seem to indicate that perceptions of chances enter less in the vote calculus for more agreeable people. This is however only one of the likely explanations and more research would be needed to set light on this matter. onclusions This paper provides a first examination of the relationship between personality traits and strategic voting by means of an original experiment which asked participants to vote in a series of electoral contexts with various degrees of electoral competition. This research allows us to draw two preliminary, yet important conclusions. The first is related to the experimental design. y linking respondents decisions to a payoff to a real nongovernmental organization (which could potentially yield real policy consequences), and by asking them to behave as in a larger group, we believe that this design allows for an increase of the external validity of this experimental research. This design could potentially be applied to a host of voting related questions, and allows for great flexibility in manipulating voting contexts at relatively low costs. The second conclusion relates to the likely impact of personality traits on strategic voting. In particular, conscientiousness appears to have a consistent and significant impact on individuals decision to rate their preferred party s chances as lower, and defect to a less preferred party when their first choice appears to be out of the race. This is consistent with prior theoretical developments focusing on this trait and deserves further investigation. The problem of strategic voting might pose a challenge for those high on agreeableness too, and indicates the need for further theorizing and empirical analyses of the relationship between personality traits and this particular type of political behavior. 10

11 Works cited: Abramson, P.R., Aldrich J. H., lais, A., Diamond, M., Diskin, A., Indridason, I. H., Lee, D. & Levine, R..(2009). omparing Strategic Voting Under FPTP and PR Systems. omparative Political Studies, 43: arney, D. R., Jost, J. T., Gosling, S. D. & Potter, J.. (2008). The Secret Lives of Liberals and onservatives: Personality Profiles, Interaction Styles, and the Things They Leave ehind. Political Psychology 29: ox, G.W.. (1997). Making s ount: Strategic oordination in the World's Electoral Systems. ambridge: ambridge University Press. Funder, D... (2008). Persons, Situations, and Person-Situation Interactions. In O. P. John, R. W. Robins, & L. A. Pervin (eds), Handbook of Personality: Theory and Research. New York: Guilford. Funder, D... (2001). Personality. Annual Review of Psychology 52: Gerber, A., Huber, G. A., Raso,. & Ha, S.. (2008). Personality and Personal ehavior. Unpublished manuscript, Yale University. Gerber, A., Huber, G. A., Doherty, D. & Dowling,. M.. (2009). Reassessing the Effects of Personality on Political Attitudes and ehaviors: Aggregate Relationships and Subgroup Differences. Unpublished manuscript, Yale University. John, O. P. & Srivastava, S. (1999). The ig Five Trait Taxonomy: History, Measurement, and Theoretical Perspectives. In L.A. Pervin & O.P. John (eds), Handbook of Personality: Theory and Research. New York: Guilford. John, O. P., Naumann, L. P. & Soto,. J.. (2008). The ig Five Trait Taxonomy: History, Measurement, and Theoretical Perspectives. In O. P. John, R. W. Robins, & L. A. Pervin (eds), Handbook of Personality: Theory and Research. New York: Guilford. Mondak, J. J. & Halperin, K. D. (2008). A Framework for the Study of Personality and Political ehavior. ritish Journal of Political Science 38: Mondak, J. J., Hibbing, M. V., anache, D., Seligson, M. A. & Anderson, M. R. (2010). Personality and ivic Engagement: An Integrative Framework for the Study of Trait Effects on Political ehavior. American Political Science Review. Vecchione, M. & aprara, G. V. (2009). Personality Determinants of Political Participation: The ontribution of Traits and Self-Efficacy eliefs. Personality and Individual Differences 46:

12 Election Experimental condition Table 1. Experimental manipulation of parties chances of winning lose A, lose () lose A, Far (F) Far A, lose (F) Far A, Far (FF) 5TH Election Distribution of previous 9000 votes Party A Party Party Table 2. Elections Results lose A, lose () lose A, Far (F) Far A, lose (F) Far A, Far (FF) Participants 119 (59.5%) Total (9200) 3109 (33,79%) Participants 108 (54%) Total (9200) 3068 (33.34%) Participants 88 (44%) Total (9200) 3018 (32.80%) Participants 86 (43%) Total (9200) 2986 (32.45%) 5TH Election Participants 74 (37%) Total (9200) 2924 (31.78%) Party A Party Party Winner 50 (25%) 3050 (33,15%) 52 (26%) 3022 (32.84%) 81 (40.5%) 3111 (33.81%) 78 (39%) 3078 (33.45%) 83 (41.5%) 3108 (33.78%) 31 (15.5%) 3041 (33,06%) 40 (20%) 3110 (33.80%) 31 (15.5%) 3071 (33.38%) 36 (18%) 3136 (34.08%) 43 (21.5%) 3168 (34.43%) Party A Party Party Party Party 12

13 Election Table 3. Mean Perceived hances of Winning for Party A lose A, lose () lose A, Far (F) Far A, lose (F) Far A, Far (FF) Party A Party Party Table 4. Personality Scale 5TH Election Scale components Mean (Std. dev) Pearson r Sig. Openness Open to new experiences, omplex Not conventional, not uncreative 5.61 (1.27) 5.09 (1.44) onscientiousness Dependable, self-disciplined Not disorganized, not careless 5.77 (1.12) 5.19 (1.62) Agreeableness Sympathetic, Warm Not critical, Not quarrelsome 5.67 (1.19) 3.73 (1.75) Extraversion Extraverted, Enthusiastic Not reserved, not quiet 4.79 (1.55) 3.63 (1.79) Emotional stability alm, emotionally stable 4.82 (1.55) Not anxious, not easily upset (1.65)

14 Table 5. Party hoices across Elections (N=200) lose A Far hoice in lose A lose election Party A Party Party Party A Party Party Far A lose Party A Party Party Far A Far Party A Party Party Fifth election Party A Party Party

15 Table 6. Strategic Voting (Dependent Variable: hoice) lose A Far Far A lose Far A Far Fifth election Party A preference -3.21** (1.23) -5.87** (1.12) -2.81* (1.20) -6.47** (1.21) # (.92) -5.20** (.87) (972.7) (972.7) Party preference 1.40 (1.28) -3.06* (1.24) 2.13 (1.55) -3.84* (1.86) 2.18* (1.08) # (1.06) (972.7) (972.7) Party A Perceived chances of winning -7.45* (2.54) -9.14** (2.87) -4.92* (2.1) -8.47* (3.32) -5.47** (1.66) # (2.51) -6.14** (1.67) -9.84** (2.77) Party perceived chances of winning 8.10* (3.40) (3.08) 9.85** (3.22) -.41 (3.51) 6.68** (2.36) (2.68) 4.22 # (2.29).57 (2.81) onstant -.05 (.67) 5.51** (1.4) -.87 (1.72) 5.24** (1.68) -.63 (1.21) 3.99** (1.14) (972.7) Pseudo R N Multinomial logit regression results. ase outcome: Probability to vote for Party A. ** p<.01, * p<.05, # p< ** (1.05) 15

16 Far A, Far Election lose A, Far Election Perceptions of Party A's hances of Winning Probability to for Party A Probability to for Party Perceptions of A's hances of Winning Probability to for Party A Probability to for Party Far A, lose Election Fifth Election (Party A annot Win) Perceptions of Party A's hances of Winning Probability to for Party A Probability to for Party Perceptions of Party A's hances of Winning Probability to for Party A Probability to for Party Figure 1. Predicted Probabilities to for A and : Party A Supporters Only 16

17 Table 7: Personality Traits and Party Preferences Openness in lose A, lose Election Model 1 Model (.17) -.19 (.20) -.16 (.18) -.14 (.21) onscientiousness -.35* (.16) -.37 (.19) -.38* (.17) -.42* (.20) Extroversion -.03* (.12).11 (.14) -.02 (.12).12 (.15) Agreeableness.35* (.16) -.19 (.19).37* (.17) -.15 (.19) Emotional stability -.12 (.14).21 (.16) -.12 (.14).23 (.17) Environmental attitudes * (.09) -.39* (.10) onstant 1.07 (1.44) 1.13 (1.71) 3.41* (1.64) 3.66 # (1.91) Pseudo-R N Multinomial logit regression results. Standard errors in parentheses. *p<.05 # p<.10 17

18 Perceptions of hances of Winning of Party A Table 8. Personality traits and perceptions of party A s chances of winning: correlations Personality traits Openness onscientiousness Extroversion Agreeableness Emotional stability lose A lose lose A Far Far A lose * * * # -.19* Far A Far Fifth election N=200 *p<.05, # p< *

19 Table 9. Dependent Variable: Perceptions of hances of Winning of Party A lose A Far Far A lose Far A Far Fifth election Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 onscientiousness -.019* (.008) -.026* (.008) # (.009) -.022* (.009) (.009) # (.009) -.021* (.009) -.022* (.009) Agreeableness.000 (.007).002 (.008) (.008) (.008).017* (.008).017* (.008).014 # (.007).014 # (.007) Emotional stability.000 (.007).000 (.007) # (.009) -.017* (.008) (.008) (.008) (.007) (.007) Party A preference (Yes=1) -.076* (.018) -.079* (.02) -.02 (.02) -.07 (.02) onstant.34** (.07).32** (.06).41** (.07).39** (.07).22** (.07).21** (.07).23** (.06).23** (.06) R N OLS regression results. Robust standard errors in parentheses. **p<.01, *p<.05, # p<

20 Table 10. Impact of Personality on Strategic Voting Decisions (Dependent Variable: hoice) lose A Far Far A lose Far A Far Fifth election Without perceptions With Perceptions Without Perceptions With Perceptions Without Perceptions With Perceptions Without Perceptions With Perceptions onscientiousness.74* (.27).59* (.25).54* (.26).46 # (.27).19 (.20).75* (.38).08 (.24).65 (.40).07 (.18).14 (.27) -.05 (.19).10 (.29).28 (.19).65* (.25).19 (.22).56* (.28) Agreeableness -.21 (.25) -.23 (.29) -.20 (.22) -.19 (.25) -.28 # (.16) -.14 (.39) -.32 # (.18) -.04 (.42) -.11 (.17) 18 (.28) -.01 (.19).32 (.27) -.36* (.17).03 (.24) -.29 (.18).25 (.27) Emotional stability -.05 (.22).16 (.21) -.01 (.22).16 (.24).15 (.17).36 (.28).00 (.19).21 (.31) -.12 (.15) -.12 (.2) -.17 (.16) -.30 (.23).02 (.15) -.13 (.20) -.01 (.17) -.32 (.22) Pseudo-R N Multinomial logit results. ase outcome: Probability to vote for Party A. Other variables included in each model: for Party A in the election, for Party in the election. *p<.05; # p<