Biological Influences on Personality

Research on the antecedents of personality reveals that all known personality traits are heritable.Footnote ⁴ Hatemi, Medland, and Eaves (Reference Hatemi, Medland and Eaves2009, 265) summarize the literature well, noting “There is overwhelming evidence, in numerous large scale studies across cultures and continents, that personality is substantially influenced by genes.” A familiar claim is that heritability for personality averages 0.50; the unshared environment accounts for the bulk of remaining variance. Recent research suggests that this 0.50 mark may understate biological influences, especially for the Big Five trait dimensions. In estimating heritability, effects of measurement error are categorized as aspects of the unshared environment (Medland and Hatemi Reference Medland and Hatemi2009). Thus, measurement error will tend to deflate estimates of heritability and of effects associated with the shared environment, while exaggerating the apparent influence of the unshared environment. Recognizing this, Riemann et al. (Reference Riemann, Angleitner and Strelau1997; see also Heath et al. Reference Heath, Neale, Kessler, Eaves and Kendler1992) measure the Big Five using both self-reports and peer reports. Using only self-report data, heritability was found to be between 0.42 and 0.56, but estimates rose to between 0.66 and 0.79 with corrections for measurement error.Footnote ⁵ Further evidence on this point is reported by McCrae et al. (Reference McCrae, Jang, Livesley, Riemann and Angleitner2001), who examine the antecedents of the Big Five structure itself rather than only the component traits. On decomposing the unshared environment into two subcomponents, the actual impact of the unshared environment and systematic error attributable to method bias (e.g., Borkenau Reference Borkenau1992), the authors find no impact of the actual unshared environment on the Big Five structure. Thus, apart from method bias, the only systematic correlate of the five-factor personality structure they identified is genetics.Footnote ⁶

The lesson we derive from these studies is that biological forces account for most of the variance in personality traits. Three interrelated questions arise from this claim: (1) in addition to biological factors, do environmental forces also influence personality?; (2) is it not the case that people's personalities change, and does this not imply a role for influences other than biology?; and (3) irrespective of other nonbiological effects, can political behavior alter an individual's personality? Brief attention to these questions should be sufficient to convey why we do not see these matters as problematic.

In Figure 1, one certain path by which environmental factors shape personality is acknowledged: if environmental forces alter biological processes, effects on personality may occur (e.g., Bagby et al. Reference Bagby, Levitan, Kennedy, Levitt and Joffe1999). In McCrae and Costa's (Reference McCrae, Costa, John, Robins and Pervin2008) Five-Factor Theory, environmental influences on personality are recognized only if those influences alter the biological bases of personality, and the authors note only one past study that, contrary to the tenets of Five-Factor Theory, offers persuasive evidence suggestive of direct environmental influences on personality. In that research (McCrae et al. Reference McCrae, Yik, Trapnell, Bond and Paulhus1998), a strong environmental shock was needed to induce discernible impact on personality. Specifically, acculturation was identified among Chinese students who had moved to Canada to attend college. In our view, the empirical record supports three conclusions. First, biological factors account for the vast majority of influence on personality traits, and especially on the five-factor structure. Second, environmental forces that operate through alteration of biological processes are acknowledged. Third, direct environmental influences on personality may occur (as represented by the dashed arrow in Fig. 1 from environmental factors to traits); however, to produce discernible impact on personality, the change in environment apparently must be quite stark—as in the case of students from China changing continents and cultures by attending college in Canada.

The second question noted previously concerns the possibility that people's personalities change over time. It is true that all humans exhibit characteristic changes in their personalities over the life cycle, and especially during adolescence, but the mere occurrence of change does not necessarily signify a noteworthy impact of environmental influences. First, the changes are slight. Costa and McCrae (Reference Costa and McCrae1988) find that the estimated stability of the Big Five trait dimensions averages 0.94 when measured at six-year intervals. Second, all change is not environmental. Costa and McCrae's (Reference Costa and McCrae2006) analysis reveals that biological influence—specifically, intrinsic maturationFootnote ⁷—offers the most plausible account for why life cycle changes in personality are identical across a wide array of cultural contexts (c.f. Roberts, Walton, and Viechtbauer Reference Roberts, Walton and Viechtbauer2006).

The final question is whether political behavior might alter an individual's personality. Although we have no means to issue an ironclad rejection of this possibility, the studies reviewed here provide ample grounds for skepticism. First, biological factors account for the vast majority of variance in personality, leaving little room for any environmental forces, let alone those based in politics, to operate. The high temporal stability of personality likewise leaves little space for environmental effects. Second, the sole study that McCrae and Costa (Reference McCrae, Costa, John, Robins and Pervin2008) concede might have identified a direct impact of the environment on personality involved a marked environmental change. Here, we examine acts of political participation such as contributing money to political candidates and affixing campaign bumper stickers to one's car. At least for dependent variables such as these, the parallel to immersion in a new culture is difficult to discern.

By viewing personality in context, we have seen that the bulk of variance in personality is rooted in biology, and that effects of political behavior on personality are unlikely. These points bring two important implications. First, statistical relationships between personality traits and facets of political behavior can reasonably be interpreted as signaling causal influence of the former on the latter. Most past research on personality and politics has been silent on this subject, apparently taking for granted that personality is causally prior to political behavior. By placing personality in context, and especially by reviewing key research regarding the antecedents of personality, we have established a much firmer and more explicit foundation for causal inference. Second, as articulated in Figure 1, because personality is substantially rooted in biology, any effects of personality on political behavior likely signal the mediated influence of biology. Scholars examining the heritability of political behavior have made passing reference to personality as a likely mechanism (e.g., Alford, Funk, and Hibbing Reference Alford, Funk and Hibbing2005; Fowler, Baker, and Dawes Reference Fowler, Baker and Dawes2008), but past research on personality and politics has neglected the possibility that personality effects imply biological influence.

The remaining components of Figure 1 to be discussed involve the effects of personality traits and environmental factors on political behavior. Here, the pivotal question is whether environmental and trait effects operate independently of one another, or in combination. Central to our argument is the proposition that full understanding of personality effects requires attention to possible environmental interactions.

Situational Expression of Trait Effects

In contemplating the political influences of environmental factors and personality traits, two types of processes can be considered. One scenario is characterized by isolated, homogeneous effects. For example, partisan mobilization efforts might be hypothesized to yield comparable returns irrespective of the attributes of the citizens targeted for mobilization. Likewise, the effects of personality traits might be presumed to be similar in form and magnitude regardless of the situation, as would be the case were we to posit that conscientiousness should exert a positive influence on the propensity to execute all duties of citizenship. The second scenario, in contrast, is characterized by situational, heterogeneous effects. Here, it is assumed that variation in people's psychological predispositions leads them to respond differently when exposed to common environmental stimuli, and, correspondingly, that the expression of personality traits will vary by situation.

In Figure 1, the dashed lines flowing from environmental factors and personality traits to political behavior acknowledge the possibility of isolated, homogeneous effects. Nonetheless, we contend that effects of this form are rare. The complexities of human psychology and of the social and political environments are such that heterogeneous effects should be the norm. When scholars report empirical models with predictors specified in simple additive form (as we ourselves have done countless times), our sense is that these specifications are rarely premised on an explicit rejection of heterogeneous influences, but instead imply only that theories and/or data do not yet allow for identification of more complex paths. As summarized in Figure 1, our framework highlights contingent effects, and especially personality × environment interactions. This emphasis marks what we believe is the essential course if attention to personality is to aid in the development of rich, nuanced accounts of political behavior.

The possibility of trait–environment interactions has received curious treatment by psychologists for several decades. In the mid-twentieth century, the norm in the field was for researchers to posit and test direct, additive effects of personality traits on a wide array of attitudes and behaviors. In the late 1960s, though, such inquiry faced a strong challenge from Walter Mischel (Reference Mischel1968), who argued that situational differences in human behavior were so pronounced as to make the very concept of personality traits essentially meaningless. Research on personality traits then stalled, with relatively little progress made for some two decades. It was only with the emergence of the Big Five framework, coupled with efforts by some scholars, including Mischel himself (e.g., Mischel Reference Mischel1979; Mischel and Shoda Reference Mischel and Shoda1995), to forge a détente built on recognition of person–situation interactions, that the field saw reinvigoration. Today, it is common to encounter statements in studies of personality traits acknowledging the likely significance of interactions between environmental factors and personality traits. As one example, McCrae and Costa (Reference McCrae, Costa, John, Robins and Pervin2008, 165) highlight these interactions in Postulate 5a of their Five-Factor Theory: “The social and physical environment interacts with personality dispositions to shape characteristic adaptations, and with characteristic adaptations, to regulate the flow of behavior.” In a recent review, Funder (Reference Funder, John, Robins and Pervin2008) stated the fundamental logic underlying attention to person–situation effects with admirable clarity: “What people do depends both on who they are—their dispositions such as personality traits—and the situation they are in” (568), and “situations and persons interact in a way that goes beyond the statistical sense of this term . . . (N)either can have any impact on the world at all without the contributions of both” (569).

Although students of trait psychology now commonly note that traits likely operate in interaction with situational variables, theory and research dedicated toward specifying such interactions has been rare. Hence, beyond a tip of the hat to the likelihood that personality effects are best understood in context, most empirical analyses differ little in form from those reported in the mid-twentieth century, with the norm once again being specification of isolated, additive effects. The problem, as Funder (Reference Funder, John, Robins and Pervin2008, 577) notes, is that attention to interactions is easier said than done: “Dispositions and situations interact to determine what people do. Which dispositions and which aspects of situations (specifically) affect which behaviors? The search for specific answers to this seemingly straightforward question lays out a formidable research agenda.” We concur with Funder both that it is essential that personality–environment interactions be awarded a central position in theories of human behavior and that the challenges associated with such a research agenda help explain why most scholars pay no more than lip service to these effects.Footnote ⁸

A true advance in scholarship in this area requires careful consideration of process. For both personality traits and environmental factors, we must detail in clear terms how and why effects on political behavior are expected to operate, and ultimately in what circumstances. Our view regarding how to proceed has much in common with Mischel and Shoda's (Reference Mischel and Shoda1995) “if-then” depiction of situational effects. When focused on environmental factors, the question to be entertained is whether the impact of a given factor may be contingent on the individual's psychological dispositions. Likewise, when focused on personality traits, the central issue is whether the expression of trait effects might depend on aspects of the situation. Funder (Reference Funder, John, Robins and Pervin2008) is correct that these seemingly straightforward tasks actually require formidable effort. Because of this, the empirical tests presented here should be seen only as illustrative examples, and certainly not as either definitive or exhaustive representations of personality × environment effects.

Rather than limiting our tests to specification of formal statistical interactions, we follow a more inclusive course. Likewise, situational variables are broadly construed. With focus on political participation, four sets of tests will be conducted. The first tests are of mediating influences. Here, the question is how and why one Big Five trait dimension, openness to experience, comes to exert positive influence on various facets of political participation. Invoking a variant of Mischel and Shoda's (Reference Mischel and Shoda1995) “if-then” logic, we posit that if openness to experience enriches the individual's resource base, particularly in terms of political information, then positive effects on participation will follow. The second tests are interactive, but with focus on the interplay between an individual's psychological predispositions and the person's attitudes. Specifically, we propose that any relationship between conscientiousness and participation will be contingent on the individual's views regarding whether political engagement constitutes an obligation of citizenship; that is, if the person perceives a sense of duty, then conscientiousness will compel engagement. The third tests are classic explorations of situational differences in trait expression. In one, we argue that extraversion will bring positive effects on participation in some circumstances, but not in others, with the key being whether the acts in question include elements of sociability. In the second, the impact of extraversion on protest activity is hypothesized to vary as a function of the nature and prevalence of protest within the context. Last, our final tests consider whether an environmental factor, the size of one's political discussion network, produces differing effects on patterns in social communication depending on levels of three traits, extraversion, agreeableness, and emotional stability.

Our objective in this section has been to reconsider personality effects on political behavior from the vantage point of a broader and more nuanced theoretical context. In doing so, we have highlighted both the powerful influence of biological forces on personality traits and the likelihood that there will be situational variance in the expression of trait effects. Testing of our thesis will be most illuminating with a wide array of dependent variables and a multiplicity of research contexts. Hence, our empirical analyses make use of data from national surveys conducted in the United States and two other nations, and our models begin with simple additive specifications before addressing possible situational variation. Last, we reiterate that because of clear evidence regarding the heritability of the five-factor trait structure, any findings we report regarding the effects of personality on political behavior will be strongly suggestive of a role for personality as a mechanism linking biology and politics.

Contingent Effects

The first effort to view personality in context does not require further analysis, but instead merely closer inspection of results in Table 2. Our thesis regarding extraversion holds that individuals scoring high on this trait will be inclined toward social interaction, which should prompt involvement in those forms of political participation that bring engagement with others. Conversely, in situations in which participation is more individualistic, extraversion should be inconsequential. Results in Table 2 provide evidence of the expected pattern. Coefficients for extraversion reach at least minimal levels of statistical significance in the six tests involving political discussion, attendance at political meetings, working for candidates, and contacting elected officials; conversely, the coefficients are insignificant for the four tests pertaining to voter turnout, displaying yard signs and bumper stickers, and contributing money to candidates and organizations. Prior research provides further evidence of the situational effect suggested here. For instance, Mondak and Halperin (Reference Mondak and Halperin2008) report significant effects for extraversion in seven of nine tests involving social forms of political participation.

Openness to experience is the only Big Five trait dimension to produce significant coefficients in all ten models in Table 2. Our next process-based test brings closer consideration of these results. We hypothesized positive effects of openness partly because persons high in openness should welcome novel aspects of political action, and partly because the high levels of information exposure characteristic of people who are open to experience should facilitate participation. We focus on the latter as our second test of process. The basic test is quite straightforward: if the impact of openness on participation is mediated by information, then the openness effects seen in Table 2 should be attenuated on inclusion of controls for political information. For the sake of simplicity, we add only two information-related variables. The first is an objective measure of civics knowledge, with values ranging from 0 to 5.Footnote ¹⁶ The second is a subjective indicator of internal political efficacy, or ability to comprehend the complexities of politics, with values ranging from 1 to 5.Footnote ¹⁷ When knowledge and internal efficacy are used as dependent variables in models with the same predictors as in Table 2, openness to experience produces significant positive effects for both (p < .01). Thus, any direct effect of knowledge or efficacy on participation would signal an indirect effect of openness.

The efficacy measure is only available on the panel survey, not on our alternate postelection survey. Therefore, in Table 3 we first replicate the openness effects from Table 2, this time using only those cases available via the panel survey (and for which valid responses are available on the knowledge and efficacy variables). We then report a second version of each model, now adding the knowledge and efficacy controls. In the first column of Table 3, nine of the ten coefficients again reach at least a minimal level of statistical significance. However, in the second column, we see that five of these nine effects slip to insignificance on introduction of the knowledge and efficacy variables, and all ten coefficients are reduced in magnitude. For their part, the coefficients for political knowledge reach at least a minimal level of significance in all ten models, along with seven for internal political efficacy. The average coefficient for openness in the first model is 1.39 versus 0.85 in the second column. Hence, it appears that, on average, approximately 40 percent of the impact of openness to experience on political engagement, at least in terms of coefficient size, is mediated by political knowledge and internal efficacy. People high in openness to experience are thoughtful, analytical, and welcoming of exposure to information. As a result, they fare well in terms of both the actual possession of political knowledge and self-confidence that they can navigate the complexities of the political arena. In turn, this knowledge and efficacy facilitates civic engagement.Footnote ¹⁸ Although the results in Table 3 are consistent with our thesis that the effects of openness on civic engagement are mediated by efficacy and political knowledge, caution is warranted due to the modest correlation between the two items used to construct our openness scale, especially as compared with the stronger alpha for the knowledge measure. Following our argument that openness is causally prior to knowledge and efficacy, results in Table 3 suggest mediation, but we cannot rule out the possibility that these results instead merely reflect the fact that knowledge is measured more precisely than is openness. This matter highlights the importance of attention to measurement issues, and especially reliability, in future research on personality and politics. We discuss this concern further later in the article.

TABLE 3.

Openness to Experience and Civic Engagement: Influence of Political Information

Source: 2006 CES.

Notes: Each row reports results from two separate models; the first entry in each row is the dependent variable. The first replicates the openness effects from Table 2 using only 2006 CES panel data, and the second adds political knowledge (0–5) and internal efficacy (1–5) as control variables. Cell entries are binomial logistic regression coefficients, with the exception of those for social network size (ordered logistic regression). Standard errors are in parentheses. Models include, as controls, conscientiousness, extraversion, agreeableness, emotional stability, age, and indicators for female and black.

***p < .001, **p < .01, *p < .05, ^#p < .10.

In Table 2, people scoring high in conscientiousness tended to fare poorly in terms of levels of civic engagement. One explanation for these results is perhaps individuals high in conscientiousness do not see civic engagement as a duty. Under this scenario, the conscientious would focus their time and effort on matters such as job and family, leaving little time for political participation. To test this interpretation, we rely on data from a three-category item asked of panel survey respondents regarding the importance of “being involved in election campaigns” (2 = very important, 1 = somewhat important, 0 = not very important). Our expectation is that an interaction between this variable and conscientiousness will produce a positive effect because any sense of the presence or absence of duty captured by the importance measure should resonate especially strongly with the conscientious. As a dependent variable, we sum the six dichotomous measures of campaign activity: working for a candidate, attending rallies or meetings, persuading others on how to vote, displaying yard signs or bumper stickers, contributing money to candidates, and contributing money to political groups.

Table 4 displays the relevant ordered logistic regression results.Footnote ¹⁹ Because the importance item is not available on the supplemental postelection survey, we again report two models. In the first, conscientiousness produces a negative coefficient of moderate magnitude. In the second model, where we introduce the interaction between conscientiousness and perceived importance of campaign activity, the negative effect of conscientiousness—the impact of conscientiousness for respondents who believe that involvement in election campaigns is unimportant—sharpens considerably. And, as expected, the interaction term produces a positive coefficient. This pattern of results suggests that the actions of individuals with high levels of conscientiousness are strongly shaped by a sense of the task's importance: where civic engagement is seen as highly important, the conscientious get involved, but where that engagement is viewed as unimportant, the conscientious focus their attention elsewhere.

TABLE 4.

Conscientiousness and Campaign Activity: Influence of Perceived Importance

Source: 2006 CES.

Notes: Cell entries are ordered logistic regression coefficients. The dependent variable is the sum of six campaign activities from Table 2: working for a candidate, attending campaign meetings or rallies, persuading others regarding how to vote, displaying yard signs or bumper stickers, contributing money to candidates, and contributing money to political groups. The first model includes the same predictors as those used in Table 2. The second model adds the perceived importance of campaign activity and its interaction with conscientiousness. Data are available from panel respondents only. Standard errors are in parentheses.

***p < .001, **p < .01, *p < .05, ^#p < .10.

Further clarification regarding this effect emerges in Figure 2, which reports the predicted probability that a respondent will engage in at least one form of campaign activity.Footnote ²⁰ The horizontal axis represents perceived importance of being involved in campaigns. Among respondents with low levels of conscientiousness, the perceived importance of campaign activity is of virtually no relevance in determining actual involvement; specifically, the predicted probability of participation increases from only 0.48 to 0.56. In sharp contrast, the predicted values rise from 0.14 to 0.62 among the highly conscientious. What is especially noteworthy in Figure 2 is that conscientious individuals who see campaign activity as important barely outpace respondents who are low in conscientiousness, but when campaign activity is seen as unimportant, the conscientious exhibit a striking hesitance toward civic engagement. When contemplating the creed “do it right, or don't do it all” as applied to civic engagement, the highly conscientious seemingly embrace the second half of the statement more strongly than the first.

FIGURE 2.

Conscientiousness, Sense of Duty, and Participation in Elections and Campaigns

Our next tests involve interactions between personality traits and respondents’ social communication networks. Much of the research on political discussion in recent years has examined communication that yields exposure to disagreement (e.g., Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004; McClurg Reference McClurg2006; Mutz Reference Mutz2006; Mutz and Mondak Reference Mutz and Mondak2006). These works have fared well in documenting the effects, mostly positive, of cross-cutting political discourse, and they also have shed light on the social and contextual factors that give rise to such conversations. But one point of dispute in this literature concerns the impact of network size. As networks grow, the likelihood that they will include at least one person with a differing point of view rises (Huckfeldt, Johnson, and Sprague Reference Huckfeldt, Johnson and Sprague2004), but it also appears that aggregate homogeneity increases as networks expand (Mutz Reference Mutz2006). In reconsidering this matter, we propose that the impact of network size on a person's exposure to disagreement may be conditional on personality. The key point is this: as network size increases, any tendency toward homogeneity or heterogeneity should not be assumed to be constant for all individuals. Instead, we posit that the person's enduring psychological tendencies may predispose the person to prefer homogeneity within the network, or to accept heterogeneity.

For the moment, we treat network size as an environmental factor fully exogenous to individual choice. We know, of course, that this assumption is at least partly incorrect; as noted in Figure 1, our framework acknowledges the influence of personality on the environment, and in Table 2, the evidence shows that personality strongly influences network size. Nonetheless, factors outside the realm of conscious choice also affect one's number of discussion partners, factors such as employment status and occupation, place of residence, and family composition. What is important for our purposes is that however a network came to be of a given size, the impact of that network on exposure to cross-cutting discourse may be contingent on personality.

Respondents were asked to indicate which way their discussion partners had voted in the local U.S. House race. We operationalize exposure to disagreement within the network with a dummy variable coded 1 if at least one of the respondent's discussion partners cast a House vote at odds with the respondent's partisan affiliation, and 0 if otherwise. With the analysis limited to respondents who named at least one discussant, a score of 1 is recorded in just under 40 percent of cases.Footnote ²¹ Our chief predictor is network size, but three of the Big Five variables may moderate its impact. First, the free-wheeling sociability of the extravert supports the hypothesis that extraverts will be relatively undiscriminating in their political conversations, in which case larger networks should magnify the odds of exposure to disagreement. Second, because people high in agreeableness tend to avoid conflict, we predict that the agreeable will surround themselves with like-minded discussion partners. Conversely, exposure to disagreement should be less disconcerting for individuals who are themselves disagreeable. Last, owing to their minimal need for social acceptance, we expect that individuals with high levels of emotional stability will tend to have heterogeneous networks.

Two logistic regression models are reported in Table 5. In the first, interactions between personality and network size are omitted. Among the personality variables, only openness produces even a hint of a direct effect on exposure to disagreement. As expected, a very crisp effect is found for network size. The second model adds interactions between network size and the Big Five. Two strong interactive effects are identified. A significant positive interaction is observed for extraversion, along with a significant negative interaction for agreeableness.

TABLE 5.

Personality, Network Size, and Exposure to Cross-cutting Political Discourse

Source: 2006 CES.

Note: Cell entries are binomial logistic regression coefficients with standard errors in parentheses. The dependent variable is exposure to cross-cutting political views in social communication (1 = exposure occurs, 0 = no exposure occurs).

***p < .001, **p < .01, *p < .05, ^#p < .10.

Predicted probabilities derived from the extraversion and agreeableness interactions are depicted in Figure 3.Footnote ²² The two patterns bear a strong resemblance to one another. First, the positive effect of network size on exposure to differing political preferences is quite modest for introverts (an 11-point swing as network size increases from one to four) and for individuals scoring high in agreeableness (a nine-point swing).Footnote ²³ Second, much more dramatic effects are seen for extraverts and those with low marks on agreeableness, with a 49-point swing found for the former and a 58-point swing for the latter. The link between network size and exposure to disagreement clearly represents more than the occurrence of a simple stochastic process. The extent to which an expansion in the size of a person's political discussion network translates into a higher probability of exposure to disagreement depends to a substantial extent on the individual's personality. In short, our understanding of the conditions under which exposure to cross-cutting discourse takes place grows markedly with attention to interactions between personality and network size.

FIGURE 3.

Conditional Impact of Network Size on Exposure to Disagreement

As a final note on Figure 3, recall from Table 2 that network size itself is strongly influenced by personality. In particular, we know from Table 2 that extraverts tend to have large networks. This point, of course, brings implications for the interpretation of the extraversion results in Figure 3 because introverts and extraverts are not distributed evenly across categories on the horizontal axis. Most important, extraverts are substantially more likely than introverts to have full four-person discussion networks.Footnote ²⁴ Large networks facilitate exposure to cross-cutting political discourse in part because the type of person who is likely to have a large network—i.e., the extravert—is the very type of person who is receptive to free-wheeling conversation.

Thus far, our consideration of how personality variables interact with other factors to influence patterns in political participation has shown that the impact of extraversion on participation is contingent on the nature of the participatory act in question, that the effects of openness are mediated partly by political information, that conscientiousness operates in concert with attitudes regarding civic duty, and that personality moderates the effects of network size on a person's exposure to cross-cutting social communication. Our final tests shift gears, this time assessing whether personality effects on participation may be contingent on the national context.

Applications of the Big Five in the study of politics mostly have made use of U.S. data, but the five-factor framework itself has been studied extensively cross-nationally. Evidence consistent with the five-factor model has been derived from personality batteries administered in numerous languages from multiple language families (Church Reference Church2000, Reference Church2001; McCrae and Costa Reference McCrae and Costa1997; Saucier and Goldberg Reference Saucier and Goldberg2001), and much of McCrae's recent research has involved further cross-cultural analysis (e.g., Allik and McCrae Reference Allik and McCrae2004; McCrae and Costa Reference McCrae, Costa, Mroczek and Little2006; Schmitt et al. Reference Schmitt, Allik, McCrae and Benet-Martinez2007). This research has also served to solidify claims regarding the biological basis of the five-factor structure. For instance, Yamagata et al. (Reference Yamagata, Suzuki, Ando, Kijima, Ostendorf, Riemann, Livesley, Jang, Spinath, Angleitner, Kimio and Ono2006) examine twin data from North America, Europe, and Asia to test whether there is a commonality of influence of genetics on personality across diverse geographic regions. Results are positive, supporting the conclusion that a universal genetic structure underlies the Big Five.

We provide a look at the possible influence of personality on political behavior outside the United States via examination of data from two national surveys conducted in 2007, one in Uruguay and the other in Venezuela.Footnote ²⁵ The surveys were virtually identical in form. Big Five indicators are again constructed using data from bipolar adjective pairs. Unfortunately, openness to experience is represented with data from only a single item, “intellectual/pragmatic.” Two items are available for each of the other four trait dimensions: “hard working/lazy” and “neat/sloppy”” for conscientiousness, “introverted/extraverted” and “talkative/shy” for extraversion, “sympathetic/unsympathetic” and “kind/rude” for agreeableness, and “relaxed/tense” and “calm/nervous” for emotional stability.

We provide a look at the possible influence of personality on political behavior outside the United States via examination of data from two national surveys conducted in 2007, one in Uruguay and the other in Venezuela.Footnote ²⁵ The surveys were virtually identical in form. Big Five indicators are again constructed using data from bipolar adjective pairs. Unfortunately, openness to experience is represented with data from only a single item, “intellectual/pragmatic.” Two items are available for each of the other four trait dimensions: “hard working/lazy” and “neat/sloppy” for conscientiousness, “introverted/extraverted” and “talkative/shy” for extraversion, “sympathetic/unsympathetic” and “kind/rude” for agreeableness, and “relaxed/tense” and “calm/nervous” for emotional stability.

Inclusion of data from Uruguay and Venezuela serves two purposes. First, we can determine whether any of our core findings from the United States replicate in other national contexts. Toward this end, our first dependent variable is a four-category indicator drawn from an item that asks about attendance at partisan political meetings (0 = respondent never attends partisan political meetings to 3 = respondent attends such meetings at least weekly). This item approximates our U.S. indicators from Table 2, and especially the measure of attendance at campaign meetings or election rallies. Attendance at meetings is a social form of political engagement, and thus the clearest hypothesis to consider is whether extraversion exerts a positive influence in these nations. Positive effects of openness are also to be expected.

Our second purpose is to test whether personality effects differ when relevant features of the national context vary. In this instance, the second dependent variable captures participation in protests (1 = respondent has participated in a protest in the past year, 0 = respondent has not participated in a protest). In Venezuela in this period, “political demonstrations occur regularly throughout the country” (Overseas Security Advisory Council 2006), and violent acts in those protests had become increasingly common (Overseas Security Advisory Council 2008). In 2007, in particular, Venezuela experienced a relatively high degree of social and political turmoil, and protests of various kinds were frequent. These protests include those both in support of and in opposition to President Hugo Chavez. Our analysis of media reportsFootnote ²⁶ identified numerous protest incidents in Venezuela in the year prior to the fielding of our survey, including incidents in August, September, October, and December of 2006, and April, May, June, and August of 2007. Some of these incidents lasted many days, and even weeks. Protests surrounded government response to crime, upcoming elections, the government closing of an opposition television station, and several labor-related matters. Many of these incidents—especially those in September 2006 and May 2007—turned violent. Some of the incidents summarized here gave rise to literally hundreds of protests. For example, Venezuelan Interior Minister Pedro Carreño reported in mid-June of 2007 that there had been more than 400 demonstrations in the prior three weeks over the closing of the opposition television station—243 in the first week, 123 in the second, and 73 in third (“Government Says Conflict on the Decline in Venezuela” 2007). In Uruguay, in contrast, the nation's enduring democratic stability was in evidence in 2007, and, with one anomalous exception, protest was rare. The exception occurred in March of 2007, when there were large-scale demonstrations in Uruguay surrounding U.S. President George Bush's visit to that nation. Outside of those protests, Uruguay is highly stable, and protest there is uncommon (see Overseas Security Advisory Council 2007). In our review of media coverage, the only other incidents we identified in Uruguay were in October 2006; these were actions in protest of government fuel and free trade policies.

Consistent with this history, only eight percent of respondents on the Uruguay survey indicated that they had participated in a protest in the past year versus more than fourteen percent in Venezuela.Footnote ²⁷ Our reading of protests in Venezuela in 2007 suggests that protest activity there was a commonplace occurrence. The same was not true in Uruguay. This distinction holds importance for extraversion. Specifically, we predict that any positive influence of extraversion on participation in protest will be strongest in Venezuela, where protest had, by 2007, become a familiar mainstream group-based behavior. Similar logic arguably could apply to conscientiousness. That is, because protest in Venezuela was so common, any negative stigma regarding protest activity may have become muted. In this scenario, the negative effect of conscientiousness on involvement in protest should be most pronounced in Uruguay. We expect similar effects in Uruguay and Venezuela for a third trait, agreeableness; high values should correspond with a lessened proclivity toward protest.

Results for the Uruguay and Venezuela models are reported in Table 6. The first two rows of results are those for our indicator of conventional participation, attendance at party meetings. Based on findings from the United States, our expectation for this variable would be a strong positive effect of extraversion and a moderate positive effect of openness to experience. Openness, which is measured in these data with only a single indicator, yields positive coefficients in both models, but neither reaches statistical significance. The expected strong influence of extraversion emerges in both nations. Substantively, the effects are quite pronounced. The predicted probability that a respondent will have attended a political meeting once or more in the past year (as opposed to never) rises from 0.06 to 0.13 across extraversion in Uruguay and from 0.10 to 0.27 across extraversion in Venezuela. Together with the U.S. results, these findings provide clear support for the contention that extraversion functions as a critical determinant of the individual-level propensity to join in group-based forms of political engagement.

TABLE 6.

The Big Five and Cross-national Political Behavior: Evidence from Uruguay and Venezuela

Sources: 2007 Uruguay and Venezuela LAPOP Surveys.

Notes: Each row reports results from a separate multivariate model; the first entry in each row is the dependent variable. Cell entries in the first two rows are ordered logistic regression coefficients, and cell entries in the third and fourth rows are binomial logistic regression coefficients. Models include age (age in years) and sex (female = 1, male = 0) as controls. Standard errors are in parentheses.

***p < .001, **p < .01, *p < .05.

Estimates for the protest models reveal that personality effects differ as a function of both the form of political participation and the nation's political context. We hypothesized that conscientiousness would be inversely related with the propensity to protest. This hypothesis is supported in both nations, although, contrary to expectations, the effect is no stronger in Uruguay than in Venezuela. The projected negative effects for agreeableness are also found, although the coefficient on this trait dimension reaches statistical significance only in Venezuela. Last, findings for extraversion match with expectations. In Venezuela, where political protests were common in 2007, extraversion acts much as it did in our other models of group-based political activity. Conversely, extraversion alone was not enough to prompt protest activity in Uruguay, a nation where, notwithstanding the events surrounding George Bush's visit, this form of engagement remained infrequent in 2007.Footnote ²⁸

Collectively, the empirical results presented here help establish the possible importance of renewed attention to personality and politics, particularly when seen in combination with findings from other recent studies. We see value in the incorporation of personality in our theoretical and empirical accounts of political behavior. However, we also see it as important that analysts proceed with caution and with awareness of some of the possible pitfalls that may be encountered.