Paper presented at the ECPR Joint Session on Personality and Turnout, Münster, Germany, March 2010. A bureaucratic mind? Personality traits and political attitudes among government employees in an international comparison Bernhard Kittel1 Markus Tepe2 Abstract3 This study explores the interplay between personality traits and government employment to explain voter turnout and party preferences in affluent democracies. Drawing on three thus far separated bodies of literature, we develop a stylized model in which political attitudes of government employees are considered to be the joint outcome of bureau voting and “bureaucratic” personality traits. We test this model for eight countries using international survey data. Estimation results obtained from multivariate analyses indicate that personality traits have a direct effect on partisan ideology. With respect to voting, the impact of personality traits is filtered via sector affiliation. Furthermore, exploring the individual determinants of sector choice, we find that government employment tends to attract persons with high levels of agreeableness, lending support to the notion of a “bureaucratic” personality. In sum, even though the public/private cleavage continues to matter, the underlying causal relationship appears to be subtler than the Bureau Voting Model suggests. Keywords: party ideology, personality traits, public administration, public/private sector cleavage, voter turnover 1 Carl von Ossietzky University Oldenburg, Center for Social Science Methodology, Ammerlaender Heerstr. 114-118, 26111 Oldenburg, [email protected] 2 Carl von Ossietzky University Oldenburg, Center for Social Science Methodology, Ammerlaender Heerstr. 114-118, 26111 Oldenburg, [email protected] 3 This study is part of an ongoing research project exploring ‘The Transformation of the State as Employer’ (Project D6) at the Collaborative Research Center ‘Transformation of the State (CRC) 597’, www.sfb597.uni-bremen.de. We would like to thank Sebastian Schneider for comments on an earlier version. We gratefully acknowledge financial support from the German Research Foundation and the Ministry for Science and Culture of Lower Saxony. -1- 1. MOTIVATION Government employees have a reputation for being different, both in their personality and their political attitudes. If jokes tell us anything about the public perception of a profession, one might conclude that government employment is associated with conformity, risk aversion and laziness. Textbook rational-choice accounts on government employees do not tend to present a brighter picture. Government employees are presumed to differ from the rest of the electorate in showing higher voter participation and holding stronger preferences for state interventionist parties. Taken both features together, government employees are presumed to cause growing public budgets (Bush & Danzau 1977, Cameron 1978, Garrett 1995, Iversen & Cusack 2000). Hence, it is not surprising that with the onset of economic downturn, politicians from all political camps regularly seek to utilize these views to create public support for retrenching government employment either in terms of scope, payment or working conditions (e.g. The Economist 2010: 9-10). Following these debates, one might question whether the differences between government employees and their private sector counterparts simply represent a chimera of public policies, or whether government employment is still a relevant political cleavage in affluent democracies. Moreover, even if government employees’ political attitudes are different, can we attribute these differences to a specific “bureaucratic” personality? In trying to conceptualize the nexus between personality traits, government employment and political attitudes we contrast two perspectives. Following the standard electoral cleavage perspective, government employees hold distinct political views as they expect to benefit from larger governments (Bureau Voting Model). Alternatively, -2- given the specific requirements of public employment certain personalities may have a predisposition to work for the government. In this case we might observe a distinct bureaucratic personality (Bureaucratic Personality Model). We develop and test a stylized model that considers political attitudes of government employees as the joint result of bureau voting and bureaucratic personality. We seek to contribute to the existing literature in a threefold manner: First, we explore the robustness of findings on bureau voting (Jensen et al. 2009, Wise & Szücs 1996, Blais et al. 1990, Garand et al. 1991) by re-estimating these models on a recent large-scale international survey. Second, we try to contribute to the growing literature exploring the role of personality traits for political attitudes (Schoen & Schumann 2007, Mondak & Halperin 2008, Mondak et al. 2010, Gerber 2009). Our main intention lies on the interplay between personality traits and government employment. Instead of presenting personality traits as a self-contained alternative towards the public/private cleavage approach, we suppose that the two explanations are interconnected. The empirical analysis relies on the ISSP Work Orientation III module 2005, covering eight OECD countries. The ISSP contains an instrument that measures the respondent’s personality on the basis of the Big Five personality traits concept. Results of the multivariate analysis can be summarized in three points: First, our estimation results are consistent with prior evidence on bureau voting. Second, personality traits significantly improve our ability to predict political attitudes. Moreover, the direction of certain personality dimensions is largely consistent with prior studies (Mondak & Halperin 2008, Mondak et al. 2010, Gerber 2009). Third, we find that government -3- employment tends to attract persons with high levels of agreeableness lending some tentative support towards the notion of a “bureaucratic” personality. The study proceeds as follows: In the next section we bring together three strands of literatures, which we use to derive a coherent and testable framework on the relationship between personality traits, government employment and political attitudes. Section three presents the dataset and method with a particular focus on how personality traits are measured. The empirical results are represented in section four. We conclude this study with a discussion of the benefits and limitations of our results and indicate some implications for further research on government employees’ political attitudes. - Figure 1 - 2. THEORETICAL FRAMEWORK There are three bodies of knowledge that are of help to answer our research question on the relationship between sector affiliation, personality traits and political attitudes: (1) the Bureau Voting Model (BVM), (2) the literature on personality traits and voting, and (3) research on Public Service Motivation (PSM). We will use these literatures to develop a testable framework that takes into account both, the public/private cleavage as well as psychological preconditions to explain voter turnover and party preferences among government employees. -4- 2.1 Rational Bureau Voting In contrast to Weber (1978), who conceptualized bureaucrats as loyal servants of the government, the BVM starts from the assumption that government employees behave as rational self-interested actors, who are not primarily interested in the goals of the organization, but in maximizing their own well-being (Downs 1967: 2). According to this line of reasoning, government employees have a vested interest in promoting the public sector. Downs’ (1967) and Niskanen’s (1971) economic theory of the bureaucracy focuses on a single bureaucrats willingness and capacity to manipulate the bureaus’ budget. Dunleavy (1991: 175) takes into account the interest group character of government employees. To this end, bureaucrats may choose on a continuum between individual (e.g. promotion) and collective strategies to manipulate public budgets. Voting is certainly the most collective strategy any interest groups can choose. In order to translate the fiscal preferences of government employees into aggregate public expenditure at least two conditions must be met: First, government employees must achieve higher turnout rates at elections in order to outweigh contrary preferences of private sector employees. And secondly, they must vote for a political party that increases government expenditure. Drawing on the concept of partisan budget manipulation in the spirit of Hibbs (1977) and Tufte (1978), the BVM predicts government employees to vote for left wing parties as these parties are more willing to expand the public sector. The link between government employment and partisan voting works as follows: If government employees are able to implement a leftist government “it will be easier to obtain a larger budget” (Blais et al. 1991: 205). -5- The BVM can be decomposed into two hypotheses: (H1) Government employees are more likely to participate in elections than private sector employees. (H2) Government employees are more likely to vote for left wing political parties than private sector employees. Following Garand, Parkhurst & Seoud (1991: 180) we argue that if the BVM is correct, each hypothesis should be supported empirically. Failure to find empirical support for any of these hypotheses would call into question the degree to which the BVM provides an adequate explanation of government employees’ political attitudes. The BVM has been the subject of numerous empirical studies, even though none of them showed that bureau voting is actually responsible for growing budgets. The majority of studies explored H1, employing data from national election studies (Bennett & Orzechowski 1983; Borcherding, Busch & Spann 1977, Corey & Garand 2002, Frey & Pommerehne 1982, Jensen, Sum & Flynn 2009). Although empirical evidence tends to support H1, for the Netherlands Jaarsma, Schram & Van Winden (1986: 186) point out that the effect is probably too small to outweigh conflicting preferences. Evidence on H2 is rather fragmented and less conclusive. For the USA, Garand et al. (1991) find that spending attitudes and partisan preferences increase the likelihood of being a public servant. Exploring a sample of 18 countries, Jensen, Sum and Flynn (2009) find empirical support for both hypotheses. A common shortcoming of the empirical investigations of the BVM refers to operationalization of bureaucrats. Available datasets do not allow differentiating between “real” bureaucrats working in the public offices and government employees working in general public service provision. Further doubt on the appropriateness of the research design stems from Dunleavy’s (1991) and Knutson’s (2005) objection that bureaucrats -6- are unlikely to behave as a homogeneous group. Particularity, the branch and position should matter for a government employees’ choice from the menu of “bureau budget manipulation”. Even though we are aware of these limitations, we postpone a more finegrained analysis of the BVM for the moment and try to re-produce findings on the two standard hypotheses H1 and H2 for a start. 2.2 Personality Traits and Voting Since Lipset & Rokkan (1967) political science has produced a rich theoretical and empirical literature exploring the relevance of various cleavage structures for voting behavior. Just recently, however, scholars embarked into a systematic study on the role of personality traits for political attitudes (Schoen & Schumann 2007, Mondak & Halperin 2008, Mondak et al. 2010, Gerber 2009, Vecchione & Caprara 2009). According to McCrae & Costa (1990: 23) personality traits are considered as consistent patterns of thought, feelings and actions. In contrast to values, which refer to “what a person considers as important”, personality traits describe “what a person is like” (Roccas et al. 2002). The dominant approach to represent differences in personality traits is the Big Five concept. The concept is based on the idea that “most of the socially-relevant and salient personally characteristics have become encoded in the natural language” (John & Srivastava 1999: 103). The Big Five personality traits are the result of decades of research trying to boil down an extensive list of expressions used to describe individual characters into a manageable and meaningful taxonomy of personality dimensions. The concept asserts that differences in human personality can be measured on five -7- dimensions: neuroticism, openness to experience, extraversion, agreeableness, and conscientiousness. The five traits are described by Roccas et al. (2002: 792) as follows: Individuals who score high on extraversion tend to be sociable, talkative, assertive, and active; those who score low tend to be retiring, reserved and cautious. Individuals who score high on agreeableness tend to be good-natured, compliant, modest, gentle and cooperative. Individuals who score low on this dimension tend to be irritable, ruthless, suspicious, and inflexible. Individuals who score high on openness tend to be intellectual, imaginative, sensitive and openminded. Those who score low tend to be down-to-earth, insensitive, and conventional. Individuals scoring high on conscientiousness tend to be careful, thorough, responsible, organized and scrupulous. Those who score low on this dimension tend to be irresponsible, disorganized and unscrupulous. Individuals scoring high on neuroticism tend to be anxious, depressed, angry and insecure. Those scoring low on Neuroticism tend to be calm, poised, and emotionally stable. In contrast to the BVM, which comes with a set of directional hypotheses, the theoretical mechanism that links personality dimensions to political attitudes is less clear-cut. Schoen & Schuman (2007: 476) link personality to voting on the basis of the “attractiveness paradigm”. They assume that “voters whose personality fits nicely with a party’s policy proposal and politicians in personality terms is likely to vote for it” (Schoen & Schuman 2007: 476). It is equally plausible, however, that voters do not try to maximize the match between the candidates’ personality and their own personality, but that they vote for a candidate that posseses those personality traits that are considered as being important for political leadership; e.g. someone might see herself as a good-natured character but also believes that politics is dirty business that requires politicians to possess an unscrupulous character. Mondak & Halperin (2008: 342) presents a set of -8- hypotheses relating the five personality traits to political attitudes. They presume that high levels of openness increase the identification with more liberal political ideologies, whereas low levels of openness are related to resistance to change and are therefore assumed to be more consistent with a conservative ideology. Likewise Mondak & Halperin (2009: 343) presume that high level of conscientiousness represent an emphasis on personal duty and responsibility, which is consistent with conservative ideology. High scores on extraversion are associated with a ‘can do’ quality (Mondak & Halperin 2009: 344). Therefore extraversion is presumed to be positively related to all aspects of groupbased political participation. With respect to emotional stability and agreeableness the basis for a directional hypothesis is even thinner. In order to understand how personality traits are translated into political attitudes we suppose to consider the nexus between personality and values as discussed in the psychological literature (Roccas et al. 2002; Dollinger et al. 1996; Luk & Bond 1993). Values, such as security, justice etc., are cognitive representations of abstract goals motivating human actions (Seligman, Olson, & Zanna, 1996). In contrast to biological studies that see personality traits as the result of genetic dispositions (Hatemi et al. 2009), Roccas et al. (2002) hold a more dynamic view of the personality concept. They presume that personality traits and values mutually influence each other. Inborn temperaments, for example, give rise to parallel traits and values vice versa (Roccas et al. 2002: 291). Following this line of reasoning, agreeableness is associated with the motivational goal of benevolence; extraversion with pursuing excitement, novelty, and challenge. Openness conflicts with the motivational goals of conformity, tradition, and security, whereas conscientiousness is positively associated with these goals. Neuroticism does not -9- facilitate the attainment of the motivational goal of any type of value (Roccas et al. 2002). The empirical analysis confirms this view. Roccas et al. (2002: 789) find that agreeableness correlates most positively with benevolence and tradition values, openness with self-direction and universalism values, extraversion with achievement and stimulation values, and conscientiousness with achievement and conformity values. Evidence from prior studies relying on the Big Five personality traits to explain voting behavior and political ideology can be summarized as follows: Schoen & Schumann (2007) investigate the effect of personality traits on partisan attitudes and vote choice in Germany using 12 items for each of the five dimensions. Their findings suggest that openness increases support for parties endorsing social liberalism, whereas low scores on conscientiousness increase the likelihood of voting for parties subscribing to economic or social liberalism, as do high levels on agreeableness (Schoen & Schumann 2007: 492). For the USA, Mondak et al. (2010) find that openness has a positive effect on voter turnout, whereas emotional stability has a negative effect. They were relying on a abbreviated 10-item battery to measure personality traits. In a prior study on the relationship between personality and political attitudes among US citizens, Mondak & Halperin (2008) used a 21-item battery. Respondents scoring high on agreeableness and openness were less likely to support left wing parties, whereas respondents with high scores of conscientiousness were more likely to favor the conservative party. Gerber et al. (2009) exploration of personality traits and political attitudes for the US supports the notion that extraversion, agreeableness and openness is linked to more liberal political views, while emotional stability and conscientiousness is linked to support for conservative candidates. In contrast to Mondak et al. (2010), they find that emotional - 10 - stability increases the probability to vote. Given these conflicting results and the fact that the impact of personality traits has yet been explored with a focus on the USA, we will test whether the Big Five concept provides a robust instrument to improve our ability to explain political attitudes in affluent democracies. Drawing on the interplay between personality traits and motivational values we explore the following two hypotheses: (H3) Individuals with high levels of conscientiousness are more likely to be motivated by conformity and traditional values and are therefore more likely to participate in elections. (H4) Individuals with high levels of agreeableness and openness and low levels of conscientiousness are more likely to possess motivational goals and values that are consistent with a left wing party ideology. We are aware that linking values and personality hints to the ambiguity in the meaning of the five personality dimensions. When Mondak & Halperin (2008) derive directional hypotheses on personality traits, they implicitly follow the same logic, with the slight difference that as political scientists we tend to speak of preferences rather than values or motivational goals. 2.3 Person-Environment Fit Let us now turn to the final causal relationship indicated in Figure 1; the effect of personality traits on sector affiliation. To draw the link between personality traits and government employment we begin with a review of the literature on Public Service Motivation (PSM). Perry & Wise’s (1990) exploration of theories on public sector work motivation suggests that public sector employees differ with respect to their responsiveness toward intrinsic and extrinsic work incentives. In contrast to private - 11 - employees, public employees are presumed to see their work as a duty rather than a job. PSM predicts that the public sector attracts individuals with particular normative orientations – “a desire to serve the public interest, loyalty to duty and to the government as a whole, and social equity” (Perry & Wise 1990: 369). Empirical evidence on the PSM argument is still rather mixed (see Wright 2001 for a review). Steen (2008), for example shows that many individuals working in the private sector are motivated by public service values. A particular shortcoming, however, results form the fact that the majority of empirical studies explore motivational differences between public and private sector employees in post employment-choice settings. Leisink & Steijn (2008: 122) point out that the PSM literature implicitly relies on the person-environment-fit framework to explain why motivational goals matter for employment choices. The person-environment-fit framework is an established concept in organizational studies. The general idea is that people are attracted to particular organizational settings and that they are more likely to stay in these settings if they ‘fit’ (Leisink & Steijn 2008: 120). According to Kristof-Brown et al. (2005: 281) personenvironment-fit is defined as “the compatibility between an individual and work environment that occurs when their characteristics are well matched.” They distinguish four domains of person-environment fit: person-job, person-group, person-organization and person-supervisor. Concerning government employment we presume the personorganization fit as most important. Perry & Wise (1990: 370) predict that “the greater an individuals public service motivation, the more likely the individual will seek membership in a public organization”. The person-organization-fit framework can be extended to personality traits. In analogy to the PSM approach we predict that workers - 12 - are trying to maximize the fit between their own personality traits and those values the employer or employment organization stands for. The question is, which characters might be attracted by government employment? Despite some more illustrative notions on the public perception of government employees’ working attitude, there is remarkable little systematic research on the relationship between personality and bureaucratic behavior (e.g. Bozeman & Rainey 1998: 167). Weber treated impersonality primarily as a feature of the bureaucratic production process. In his analysis impersonality is a precondition for the use of rules and the selection of personnel (Downs 1966: 65). Downs (1966) shifts toward the individual level, pointing out that loyalty to one’s superior plays a vital functional role in bureaus. For Downs (1966: 71) this is due to the fact that “all top level officials (and many others) are frequently in danger of being embarrassed by revelations of their illegal acts”. At this stage, however, it is not exactly clear why loyalty either as a personality trait or value should be more important in a bureau than in any other work environment. Downs (1966: 84) assumes that bureaucrats follow quite different motivational goals: Power, money income and prestige play a role but also convenience (resistance to change in behavior that increases personal effort), security (low probability in future losses in power) and the desire to serve the public interest. Based on these motives he provides a typology of bureaucrats. (1) Climbers, purely self-interested bureaucrats, which focus on power, income and prestige; (2) Conservers, with a focus on convenience and security; (3) Zealots, which are loyal to relatively narrow policies or concepts; (4) Advocates, which are loyal to a broader set of functions; and (5) Statesmen, which are loyal to society as a whole. In what follows has been commented by Dunleavy (1991: - 13 - 168) as “venturing into amateur social psychology”. Downs (1966: 89) assumes that besides the rank and position, psychological predispositions determine which particular type of official a bureau member is likely to be. He considers Zealots as more optimistic than Statesmen and Advocates. Advocates are assumed to be basically optimistic and normally quite energetic, even though they are more subject to influence by others than average. Finally, the statesmen personality is characterized to vary between extreme laziness and hyperactivity. The most systematic approach to explore the link between personality traits and bureaucratic behavior of which we are aware is presented by Bonzeman & Rainey (1998). They draw on the work of Thompson (1961) and Merton (1940) to test the existence of a “bureaucratic personality”. Merton (1940) assumes that personal characteristics influence the perception of organizational characteristics. In his view bureaucratic personalities are persons who have an inherent need for constraints and regulation. Probably the strongest version of this argument can be found in Thompson (1961), who suggests a theory of “bureaupathology”. In his perspective, a persons’ favor for bureaucratic behavior stems from insecurity. Hence, those behaviors most people would think of as being bureaucratic, are pathological distortions of the Weberian bureaucracy in the account of Thompson (1961) (see also Bonzeman & Rainey 1998: 167). Personal insecurity gives rise to managerial efforts to exert control over public policies beyond any reasonable degree. In this respect, Bonzeman & Rainey (1998: 168) describe Thompson’s (1961) theory as a “Weberian bureaucracy run amuck”. Drawing on the person-environment-fit model and the “bureaupathology” argument we complement our set of directional hypotheses: (H5) Respondents with high - 14 - levels of openness and extraversion are less likely to work for the government, whereas respondents with high levels of conscientiousness are more likely to choose working for the government. 2.4 A stylized model on personality traits, government employment and political attitudes The three bodies of literature are synthesised into a testable framework of attitudes towards electoral participation and partisan ideology (see Figure 1). We are going to explore three causal relationships: First, the link between sector affiliation and political attitudes (H1 and H2). Second, the link between personality traits and political attitudes (H3 and H4). And finally, the effect of personality traits on employment choice (H5). Table 1 summarizes the expected relationships. With this set up we seek to disentangle the joint impact of personality traits and bureau voting for electoral outcomes in affluent democracies. - Table 1 - 3. DATA AND METHODS We test our theoretical model on survey data for eight countries on which the variables of interest are available in the International Social Survey Programme 2005 (Module on Work Orientation III). All countries are advanced industrial democracies with a welldefined public sector. - 15 - 3.1 Dependent Variables We capture political attitudes with two dependent variables. First, electoral participation is measured with a simple binary variable. Second, partisan orientation is captured with a 5-point scale ranging from far right (1) to far left (5). The item measuring the respondent’s self-reported sector affiliation is used as an independent variable in those models contrasting the explanatory power of the BVM against the influence of personality traits. In the last step of the analysis, we will use self-reported sector affiliation as the dependent variable in order to assess whether certain personality traits predetermine the choice of the employment sector. Respondents were asked whether they regard themselves as (1) working for the government, (2) a public owned firm/national industry, (3) private firm or (4) whether they are self-employed. A difficulty arises with respect to the second category. Jensen et al. (2009) considered respondents in the category “public owned firm/national industry” as government employees. We suppose that the exact meaning of this category is likely to depend on country contexts. Since the dataset provides no information, whether respondents in that category belong to the public or private sphere, we have decided to follow the most conservative strategy and excluded this category from the analysis.1 3.2 Independent Variables The ISSP 2005 includes the abbreviated version of the Big Five Inventory (BFI) composed by Rammstedt & John (2007). The standard BFI usually consists of 44 short phrase items (John et al. 1999), rated on a five-step scale ranging from “1= disagree strongly” to “5= agree strongly”. Rammstedt & John (2007) selected two items for each - 16 - dimension. The exact spelling of the ten items is shown in Appendix Table 1. The selection procedure and reliability of the BFI-10 item battery compared to the BFI-44 item battery is discussed in Rammstedt & John (2007). Overall, their empirical analysis suggests that the BFI-10 leads to smaller effect sizes but is still sufficient for research settings with strong time constraints. In order to test how well the BFI-10 captures personality traits, we calculated the scale reliability coefficient for the pooled sample as well as for each country separately. With the exception of agreeableness, which yields a scale reliability coefficient of 0.13, our findings are by and large consistent with Rammstedt & John (2007: 206) who also find scale reliability coefficients below the critical value of 0.70. Besides this consistency and the fact that Mondak et al. (2010) also rely on the BFI-10, doubts about the reliability of the measurement concept remain. Our approach to encounter this shortcoming is the following: If the BFI-10 is an unreliable instrument, it should provide statistical artefacts. In this case, it should be very unlikely that we are able to reproduce findings from prior studies that used the extended BFI battery. If, however, the BFI-10 confirms studies that rely on the extended version, we can conclude that the real impact of personality traits is probably larger than the impact revealed on the basis of the abbreviated item battery. As the re-production of the BVM is the first stage in our research strategy, we have selected control variables in accordance with prior research in this topic. The statistical models include as control variables: age, gender, educational degree, social class, employment status, and urban background. These are standard predictors in the individual-level analysis of political attitudes (Lipset & Rokkan 1967). Since most - 17 - employee selection techniques take into account the applicants job-related abilities, we will also control for the respondents’ occupation. 3.3 Method We employ binary logit regression (with robust standard errors) to test the effect of personality traits and government employment on voter turnover. Since partisan ideology is measured on a 5-point right-to-left-scale, we use an ordered logit model (with robust standard errors) to test the effect of personality traits and government employment on partisan ideology. Finally, to explore the effect of personality traits on sector choice we rely on a multinomial logit model (with robust standard errors), using private sector employees as the reference category. Our interpretation of the impact of an independent variable on the dependent variable is based on the beta coefficient, statistical significance and its effect on the predicted probability as the selected independent variable moves from its minimum to its maximum.2 Since we use a cross-national dataset, each model includes a full set of country dummies. In order to contrast the explanatory power of the BVM against the personality traits concept, we rely on the Bayesian Information Criterion (BIC) and the Akaike Information Criterion (AIC). 4. EMPIRICAL ANALYSIS The empirical analysis will proceed in three steps. First, we estimate the effect of selfreported sector affiliation and personality traits on voter turnout (Table 2). Second, we estimate the effect of self-reported sector affiliation and personality traits on partisan orientation (Table 3). In doing so, we compare the capacity of the BVM and the - 18 - personality concept to explain political attitudes. Finally, we test the effect of personality traits on self-reported sector affiliation (Table 4). - Table 2 - 4.1 Determinants of voter turnout The statistical models on voter turnout are arranged according to the following logic: Model 1 seeks to reproduce prior empirical findings on the BVM by Jensen et al. (2009). Model 2 is a reduced model that only includes those variables measuring personality traits. Model 3 includes the five personality variables and the full set of control variables. Model 4 includes the full set of control variables and both concepts of theoretical interest – bureau voting and personality. To begin with, the estimation results for our control variables are consistent with prior findings on electoral participation (Model 1). Older respondents and respondents with a higher educational degree are more likely to participate in elections, whereas females and respondents in larger cities are less likely to participate. The latter is also the case with unemployment. Unemployment yields a negative effect on electoral participation, a finding, which lends empirical support to the growing literature on welfare state “outsiders” (Rueda 2007). Concerning effect sizes, ageing and educational degree have the largest impact on the predicted probability to vote. If age increases from its minimum to its maximum, keeping all other independent variables at their mean, the predicted probability to vote increases by 45 percent. Educational degree increases the probability to vote by 28 percent. - 19 - Let us now turn to the reproduction of findings on the BVM. Model 1 indicates that government employees are more likely to vote than private sector employees. Even though the substantive effect might appear to be rather small – government employment increases the predicted probability to vote by 6.8 percent – the result provides empirical support in favor of H1. Moreover, this finding is consistent with Jensen et al. (2009) both in terms of statistical significance and effect size. Results obtained from Model 2 provide empirical support in favor of H3. Respondents scoring high on the conscientiousness and agreeableness dimension are significantly more likely to participate in elections, whereas respondents scoring high on neuroticism are less likely to vote. However, this pattern vanishes in Model 3, which includes the full set of control variables. Conscientiousness has no longer a significant effect on voting. From these changes we conclude that our control variables absorb some of the variance explained by personality traits. Model 4 includes both explanatory concepts – BVM and personality traits. Still, government employment yields a statistically significant and positive effect on voting, whereas among the five personality dimensions only neuroticism continues to have an impact on voting. In substantive terms, increasing neuroticism from its minimum to its maximum, while keeping all other independent variables at their mean, decreases the likelihood to vote by 5.5 percent. With the inclusion of sector affiliation in Model 4, the estimation coefficient for agreeableness dimension looses its statistical significance. Some part of the variance on voting that was captured by agreeableness seems to be better captured by sector affiliation. According to Model 4, we reject H3. Nevertheless, this does not necessarily mean that the personality concept has little to say about electoral participation. The Wald Test as well as the Likelihood Ratio Test on the five personality - 20 - variables indicates that the personality concept is jointly significant. The BIC and AIC also show that including personality traits into the statistical model slightly improves the overall model fit. In sum, personality traits tend to exert an indirect rather than direct effect on voting. - Table 3 - 4.2 Determinants of partisan orientation Table 3 represents estimation results obtained from the ordered logit model on party ideology. This sample includes only those respondents who have reported their political party preferences. Those who did not answer this question or did not prefer any particular party are not considered. Moreover, the partisan orientation was not asked in France. Due to these limitations, the sample on partisan orientation is smaller than the sample on electoral participation. First, taking a look at our set of control variables indicates that women, students, urbanite and unemployed respondents are more likely to favor left wing parties, whereas respondents belonging to higher social classes support right wing parties. These results are broadly consistent with prior investigations on partisan orientation. Furthermore, Model 1 provides empirical evidence in support of H2. Consistent with Jensen et al. (2009), we find that government employees are more likely to favor left wing parties. On average, government employment increases the predicted probability to support a left wing party by 4.9 percent. Moreover, we see that self-employed are – compared to those employed in the private sector –more likely to support right wing parties. This pattern is - 21 - perfectly consistent with the idea that people vote in their economic self-interest. Model 2 presents the reduced model including only the five variables measuring personality traits. Respondents scoring high on the conscientiousness dimension are less likely to support a left wing party, whereas respondents scoring high on the openness and agreeableness dimension tend to support left wing parties. In order to test the robustness of this pattern, Model 3 includes the full-set of control variables. As a result, the effect size decreases, but the statistical significance of conscientiousness, openness and agreeableness is not affected. Finally, Model 4 estimates the joint effect of the BVM and personality traits on partisan orientation. Government employment still exerts a statistically significant and positive effect on partisan orientation (H2). Since entering self-reported sector affiliation into the statistical model does not alter the effect of personality traits, we conclude that personality appears to provide an explanation for partisan orientation that is independent from sector affiliation. The Wald and Likelihood Ratio Test for the five personality measures support this conclusion. In substantive terms the openness and agreeableness dimension increase the predicted probability to favor a left wing party by 6.3 and 5.7 percent, while conscientiousness decreases left wing support by 3.3 percent. Compared to the substantive effect of government employment (4.9 percent), personality traits can be considered to be relevant predictors of partisan orientation. Concerning our directional hypotheses on personality traits and partisan orientation, Model 4 partially confirms (H4) as openness and conscientiousness show to have the expected effect on partisan attitudes. The effect of agreeableness has not been captured by our set of hypotheses. - 22 - - Table 4 - 4.3 Determinants of sector choice Finally, we explore the effect of personality traits on self-reported sector affiliation. To do so, the sample includes only those who are actually working.3 We estimate two multinomial logit models. The first represents the “pure” personality traits model, whereas the second model includes a full set of control variables. Model 1 indicates that government employees score higher on the openness, agreeableness and neuroticism dimensions, whereas among the self-employed openness and conscientiousness tends to be higher. This pattern, however, is not robust toward the inclusion of our control measures (Model 2). With respect to our control variables, we find that compared to private sector employees government employees and the self-employed tend to be older. Moreover, females and respondents with a higher educational background are more likely to belong to the sector affiliation “government employment”. This pattern is consistent with public sector recruitment policies that tend to stress gender equality and formal qualifications (Disney & Gosling 1998, Peters & Hood 1995). The inclusion of control variables, particularly the inclusion of occupation, alters the impact of personality traits on sector affiliation. Despite agreeableness, none of the five personality measures have a statistically significant impact on sector affiliation. Thus, despite agreeableness, personality must have an indirect effect on sector affiliation since the Likelihood Ratio Test shows that including personality traits improves our statistical model. The direct effect of agreeableness is consistent with our conclusion obtained from Table 2. There we - 23 - found that the effect of agreeableness on voting disappears as soon as we control for sector affiliation. In sum, this pattern lends some tentative support towards the notion of a “bureaucratic” personality. Nevertheless, compared to the self-employed, the impact of personality traits is rather weak. Model 2 indicates that higher scores on extraversion and openness increase the probability to choose self-employment. In this respect one might conclude that extraversion and openness are the distinct feature of an entrepreneur. 5. CONCLUSIONS This study explored the nexus between personality traits, government employment and political attitudes. We develop and test a stylized model that considers political attitudes of government employees as the joint result of bureau voting and bureaucratic personality. The empirical findings obtained from multivariate regression analysis on eight affluent democracies can be summarized in three points: First, our estimation results are consistent with prior evidence on the BVM. Thus, government employees are more likely to vote and they are more likely to support left wing parties. Second, personality traits significantly improve our ability to predict political attitudes. The direction and substantive effect of certain personality dimensions is largely consistent with prior studies (Mondak & Halperin 2008, Mondak et al. 2010, Gerber 2009). Comparing the explanatory power of Bureau Voting and Bureau Personality, indicts that personality yield an indirect effect on voting and a direct impact on partisan ideology. Third, we find that government employment attracts persons with high levels of agreeableness lending support to the notion of a “bureaucratic” personality trait. This pattern tilts in the direction of the PSM framework, as Roccas et al. (2002) has shown - 24 - that the agreeableness dimension correlated with co-operative and altruistic values. Brewer (2003) explored differences in civic attitudes between public and private sector employees. He finds that government employment is a substantively important and highly significant predictor of civil participation. In the wake of the empirical evidence provide by this study, we can speculate that civic participation values and agreeableness traits mutually consolidate the difference between government employees and their private sector counterparts. Before we conclude with the theoretical implication of this exploration, we will mention some of the methodological limitations involved in the analysis: First, even though the BFI-10 generated results that are consistent with studies that relied on the extended version of the BFI, we suppose that there is some scope to improve the comparative measurement of personality traits. Second, the BVM and PSM literature evolves around the concept of bureaucrats rather than government employees - which we have used to capture the public/private cleavage. Using this definition, people working in social services (e.g. nurses, doctors, teachers, professors) are considered as government employees. Thus, our findings on the “bureaucratic” personality might be driven by social service provision. Further research might be able to provide a refined analysis that takes into account differences between government employees. Third, concerning the effect of personality traits on employment choices we also explored post-employmentchoice situations. Further research might test the robustness of any of our findings using pre-employment-choice data. Fourth, as the results presented by Norris (2003) suggest, there may be considerable cross-country differences in effect size and direction that are - 25 - related to administrative regimes (Tepe, Gottschall & Kittel 2010). Such cross-level effects should be explored in future research. The theoretical implications of this study are twofold. Firstly, with respect to the voting literature we can say that the public/private cleavage continues to matter, both with respect to political participation and party preferences. However, the underlying causal relationship appears to be more subtle than the BVM suggests. Personality traits have a direct and unique effect on partisan ideology and play a mediating role in government employees’ electoral participation. The impact of personality traits on voting is filtered via sector affiliation. Secondly, if at all, personality traits have received very limited attention in public administration research. We find some tentative support that the concept of personality improves our understanding of persistent differences between public and private sector employees’ work motivation. The latest OECD (2009: 33) report on public sector modernization concludes, that several reform instruments “have not resulted in the intended changes in behaviour and culture.” On this background, we speculate that designing successful public sector reforms might requires paying more attention to the compatibility between reform instruments and government employees’ personality traits. - 26 - ENDNOTES 1 Furthermore, we re-estimated each statistical model including a dummy variable controlling for respondents working in the national industry. Doing so did not alter any of our substantial results. 2 The statistical analysis uses Stata 10. In order to calculate changes in the predicted probability, we employ the prchange command from the SPOST package by Long & Freese (2006). The dataset and Stata command files are available for replication purposes. 3 Re-estimating the multinomial logit model on the full-sample did not alter the substantial findings. - 27 - REFERENCES Bennett, J.T., and Orzechowski, W.P. (1983) ‘The voting behavior of bureaucrats: Some empirical evidence’, Public Choice 41(2): 271-283. Blais, A., Dion, S., and Blake, D. (1990) ‘The Public/Private Sector Cleavage in North America’, Comparative Political Studies 23(3): 381-403. Blais, A., Blake, D. and Dion, S. (1991) ‘The voting behavior of bureaucrats’, in A. Blais and S. Dion (eds.) The budget-maximizing bureaucrat: Appraisals and evidence, Pittsburgh: University of Pittsburgh Press, pp. 205-230. Borcherding, T.E., Bush, W.C., and Spann, R.M. (1977) ‘The Effects on Public Spending of the Divisibility of Public Outputs in Consumption, Bureaucratic Power, and the Size of the Tax-Sharing Group’, in T. Borcherding (ed.), Budgets and Bureaucrats: The Sources of Government Growth, Durham: Duke University Press. Bozeman, B. and Rainey, H. (1998) ‘Organizational Rules and the "Bureaucratic Personality"’, American Journal of Political Science, 42(1): 163-189. Brewer, G.A. (2003) ‘Building Social Capital: Civic Attitudes and Behavior of Public Servants’, Journal of Public Administration Research and Theory 13(1): 5-26. Bush, W., and Danzau, A. (1977) ‘The voting behavior of bureaucrats and public sector growth’, in T. Borcherding (ed.), Budgets and Bureaucrats, Durham: Duke University Press, pp. 90-99. Cameron, D. R. (1978) ‘The expansion of the public economy: A comparative analysis’, American Political Science Review 72: 1243-1261. Corey, E.C., and Garand, J.C. (2002) ‘Are Government Employees more Likely to Vote?: An Analysis of Turnout in the 1996 U.S. National Election’, Public Choice 111(34): 259-283. Disney, R., and Gosling, A. (1998) ‘Does It Pay to Work in the Public Sector?’ Fiscal Studies 19(4): 347–374. Dunleavy, P. (1991) Democracy, bureaucracy and public choice: Economic explanations in political science, New York: Prentice Hall. - 28 - Dollinger, S., Leong, F., and Ulicni, S. (1996) ‘On traits and values: With special reference to Openness to Experience’, Journal of Research in Personality 30, 2341. Downs, A. (1967) Inside bureaucracy. New York: Little, Brown. Frey, B.S., and Pommerehne, W.W. (1982) ‘How powerful are public bureaucrats as voters?’, Public Choice 253–262. 38(3): 253-262. Garand, J.C., Parkhurst, C.T., and Seoud, R.J. (1991) ‘Bureaucrats, Policy Attitudes, and Political Behavior: Extension of the Bureau Voting Model of Government Growth’, Journal of Public Administration Research and Theory 1(2): 177-212. Garrett, G. (1995) ‘Capital mobility, trade and the domestic politics of economic policy’, International Organization 49:657–87. Gerber, A., Huber, G. A., Doherty, D. and Dowling, C. M. (2009) ‘Reassessing the Effects of Personality on Political Attitudes and Behaviors: Aggregate Relationships and Subgroup Differences.’ Unpublished manuscript, Yale University. Hatemi, Peter K., Sarah E. Medland, and Lindon J. Eaves. (2009) ‘Do Genes Contribute to the ‘GenderGap’?’ Journal of Politics 71(1): 262–76. Hibbs, D. (1977) ‘Political parties and macroeconomic policy’, American Political Science Review 71(4): 1467-1487. Iversen, T., and T. R. Cusack (2000) ‘The causes of welfare state expansion: Deindustrialization or globalization?’, World Politics 52 (April): 313–49. Jaarsma, B., Schram, A., and Winden, F.V. (1986) ‘On the voting participation of public bureaucrats’, Public Choice 48(2): 183-187. Jensen, J, Sum, P. and Flynn, D. (2009) ‘Political orientations and behavior of public employees: A cross-national comparison’, Journal of Public Administration Research and Theory forthcoming (doi:10.1093/jopart/mun031) John, O. P., and Srivastava, S. (1999). The Big 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 Press. pp. 102-138. - 29 - Knutsen, O. (2005) ‘The impact of sector employment on party choice: A comparative study of eight West European countries’, European Journal of Political Research 44(4): 593-621. Kristof-Brown, A. L., Zimmerman, R. D., and Johnson, E. C. (2005) ‘Consequences of individuals’ fit at work: A meta-analysis of person-job, person organization, person-group, and person-supervisor fit’, Personnel Psychology 58(2), 281–320. La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1998) ‘Law and Finance’, Journal of Political Economy, 106, 1113-1155. La Porta, R., Lopez-de-Silanes, F. and Shleifer, A. (2008) ‘The economic consequences of legal origin’, Journal of Economic Literature, 42(2), 285-332. Leisink, P. and Steijn, B. (2008) Recruitment, Attraction, and Selection. in James Perry and Annie Hondeghem (Eds.) Motivation in Public Management. The Call of Public Service, Oxford: Oxford University Press, pp.118-135. Lipset, S.; Rokkan, S. (1967) Party Systems and Voter Alignments. Cross-National Perspectives. New York: Free Press. Long, J. S., and J. Freese. (2006) Regression Models for Categorical Dependent Variables Using Stata, College Station Texas: Stata Press. Luk, C. L., and Bond, M. H. (1993) ‘Personality variation and values endorsement in Chinese university students’, Personality and Individual Differences 14: 429-437. McCrae R. R., and Costa, P. T. (1990) Personality in adulthood. New York: Guilford. Merton, R. (1940) Bureaucratic Structure and Personality, Social Forces 18: 560-668. Mondak, J. J., Hibbing, M. V., Canache, D., Seligson, M. A. and Anderson, M. R. (2010) ‘Personality and Civic Engagement: An Integrative Framework for the Study of Trait Effects on Political Behavior’, American Political Science Review (Feb.) Mondak, Jeffery J., and Karen D. Halperin (2008) ‘A Framework for the Study of Personality and Political Behavior’, British Journal of Political Science 38: 335– 62. Niskanen, W. (1971) Bureaucracy and representative government, Chicago: Aldine Atherton. - 30 - Norris, Pippa (2003) Is There Still a Public Service Ethos? Work Values, Experience, and Job Satisfaction among Government Workers. In John D. Donahue & Joseph S. Nye (Eds.), For the People: Can We Fix Public Service? New York: Brookings. OECD (2009) Government at a Glance 2009, Paris: OECD. Perry, J. L., and Wise, L. R. (1990) ‘The motivational bases of public service’, Public Administration Review 50(3), 367-373. Peters, B. and Hood, C. (1995) ‘Erosion and Variety in Pay for High Public Officials’, Governance 8:171-194. Pollitt, C., and Bouckaert, G. (2004) Public Management Reform: A Comparative Analysis, Oxford: Oxford University Press. Rammstedt, B.; John, O. (2007) ‘Measuring personality in one minute or less: A 10-item short version of the Big Five Inventory in English and German’, Journal of Research in Personality 41 (2007) 203–212. Roccas, S.; Sagiv, L.; Schwartz, S. and Knafo, K. (2002) ‘The Big Five Personality Factors and Personal Values’, Personality and Social Psychology Bulletin 28(6): 789-801. Rueda, D. (2007) Social Democracy Inside Out. Partisanship and Labor Market Policy in Advanced Industrialized Democracies, Oxford: Oxford University Press. Seligman, C., Olson, J. M., and Zanna, M. P. (Eds.). (1996). The psychology of values: The Ontario symposium (Vol. 8). Mahwah, NJ: Lawrence Erlbaum. Schoen, H., and Schumann, S. (2007) ‘Personality Traits, Partisan Attitudes, and Voting Behavior: Evidence from Germany’, Political Psychology 28 (4): 471–98. Steen, T.(2008) Not a Government Monopoly: The Private, Nonprofit, and Voluntary Sectors. In James Perry and Annie Hondeghem (Eds.) Motivation in Public Management. The Call of Public Service, Oxford: Oxford University Press, p.203222. Tepe, M., Gottschall, K. and Kittel, B. (2010). A Structural Fit of States and Markets? Public Administration Regulation and Market Economy Models in the OECD. Unpublished Manuscript. - 31 - The Economist (2010) Stop! The size and power of the state is growing and discontent is on the rise, January 23rd-28th: 9-10. Thompson, V. (1961) Modern Organization, New York: Knopf. Tufte, E. (1978) Political Control of the Economy, Princeton: Princeton University Press. Vecchione, M. and Caprara, G. V. (2009) ‘Personality Determinants of Political Participation: The Contribution of Traits and Self-Efficacy Beliefs’, Personality and Individual Differences 46: 487-492. Wise, L.R., and Szücs, S. (1996) ‘The Public/Private Cleavage in a Welfare State: Attitudes Toward Public Management Reform’, Governance 9(1): 43-70. Weber, M. (1978) Economy and Society: An Outline of Interpretive Sociology, Edited by Guenther Roth and Claus Wittich, University of California Press. Wright, B. E. (2001). Public-sector work motivation: A review of the current literature and a revised conceptual model. Journal of Public Administration Research & Theory, 11(4), 559–86. - 32 - Figure 1. Stylized relationship between sector, personality and political behavior. Socio-demographics & other control Var. Personality traits 2. Personality / value nexus 3. PersonEnvironment-Fit Public / private sector cleavage Political attitudes 1. Bureau Voting Model Table 1. Stylized Hypothesis on the relationship between sector, personality and political behavior. Independent Variables Sector: Government Personality traits Conscientiousness Extraversion Openness Agreeableness Neuroticism Electoral participation Dependent Variables Left wing party affiliation + + + + + - 33 - Sector: Government + + Table 2. The effect of sector and personality traits on electoral participation Model (1) Model (2) Model (3) Model (4) Delta Pr min->max “Did you vote in last election?” Yes=1, No=0 Government employee Self employed Age Female Edu. Degree Class Region Manager Academic Technician Office worker Service worker Agricultural worker Craftsmen Non Working Retired Student Unemployed 0.473*** [0.10] 0.0977 [0.10] 0.717*** [0.05] -0.131* [0.07] 0.414*** [0.05] 0.0525 [0.03] -0.239*** [0.04] 0.675*** [0.15] 0.291* [0.15] 0.494*** [0.12] 0.428*** [0.12] 0.194* [0.11] 0.419** [0.20] 0.155 [0.12] 0.0583 [0.11] -0.0249 [0.15] 0.0424 [0.24] -0.479** [0.23] Conscientiousness Extraversion Openness Agreeableness Neuroticism Observations Pseudo R-squared AIC BIC Wald Test (Big 5) LR Test (Big 5) 7281 0.173 6937.5 7116.7 0.111*** [0.04] -0.0126 [0.04] -0.0184 [0.04] 0.0835** [0.04] -0.0900** [0.04] 7281 0.081 7671.0 7760.6 0.728*** [0.05] -0.115 [0.07] 0.432*** [0.05] 0.0487 [0.03] -0.235*** [0.04] 0.639*** [0.15] 0.376** [0.15] 0.526*** [0.12] 0.456*** [0.12] 0.208* [0.11] 0.417** [0.20] 0.113 [0.12] 0.0538 [0.11] -0.0307 [0.15] 0.0780 [0.23] -0.472** [0.23] 0.0287 [0.04] 0.0396 [0.04] -0.0313 [0.04] 0.0610* [0.04] -0.0793** [0.04] 7281 0.171 6956.6 7156.5 0.475*** [0.10] 0.0925 [0.10] 0.709*** [0.05] -0.124* [0.07] 0.416*** [0.05] 0.0515 [0.03] -0.235*** [0.04] 0.656*** [0.15] 0.305** [0.15] 0.499*** [0.12] 0.432*** [0.12] 0.186* [0.11] 0.405** [0.20] 0.148 [0.12] 0.0757 [0.11] -0.0101 [0.15] 0.0907 [0.24] -0.470** [0.23] 0.0305 [0.04] 0.0425 [0.04] -0.0310 [0.04] 0.0580 [0.04] -0.0810** [0.04] 7281 0.175 6930.9 7144.6 12.49** 12.51** 0.0688 0.0143 0.4534 -0.0195 0.2848 0.0536 -0.1164 0.0876 0.0448 0.071 0.0616 0.0282 0.0569 0.0225 0.0117 -0.0016 0.0139 -0.084 0.0268 0.0299 -0.0222 0.0541 -0.0555 Note: Logit estimates (using sample weights); robust standard errors in brackets; * p<0.10 ** p<0.05 *** p<0.01; constant and country dummies included but not reported; sample includes: United States, Ireland, Czech Republic, Japan, France, Denmark, Switzerland, South Korea - 34 - Table 3. The effect of sector and personality traits on party affiliation Model (1) Model (2) Model (3) Model (4) Delta av. Pr min->max Party affiliation from far right (1) to far left (5) Government employee Self employed Age Female Edu. Degree Class Region Manager Academic Technician Office worker Service worker Agricultural worker Craftsmen Non Working Retired Student Unemployed 0.563*** [0.09] -0.168** [0.08] -0.0182 [0.04] 0.219*** [0.06] -0.0354 [0.04] -0.0689** [0.03] 0.168*** [0.04] -0.245** [0.12] 0.0671 [0.12] -0.0797 [0.11] -0.0945 [0.11] -0.121 [0.10] -0.155 [0.15] 0.0336 [0.11] 0.0692 [0.09] -0.0973 [0.12] 0.519** [0.22] 0.431** [0.19] Conscientiousness Extraversion Openness Agreeableness Neuroticism Observations Pseudo R-squared AIC BIC Wald Test (Big 5) LR Test (Big 5) 4782 0.025 12426.6 12607.8 -0.0840*** [0.03] -0.0264 [0.03] 0.160*** [0.03] 0.108*** [0.03] 0.0508 [0.03] 4782 0.015 12521.4 12618.5 -0.0165 [0.04] 0.240*** [0.06] -0.0295 [0.04] -0.0760*** [0.03] 0.160*** [0.04] -0.319*** [0.11] 0.112 [0.12] -0.103 [0.11] -0.0544 [0.11] -0.135 [0.10] -0.296** [0.14] -0.0367 [0.11] 0.0191 [0.08] -0.0744 [0.12] 0.499** [0.22] 0.338* [0.19] -0.0678** [0.03] -0.0377 [0.03] 0.145*** [0.03] 0.113*** [0.03] 0.0212 [0.03] 4782 0.023 12456.9 12657.6 0.545*** [0.08] -0.179** [0.08] -0.0206 [0.04] 0.212*** [0.06] -0.0606 [0.04] -0.0681** [0.03] 0.165*** [0.04] -0.256** [0.12] 0.0513 [0.12] -0.0894 [0.11] -0.103 [0.11] -0.141 [0.10] -0.144 [0.15] 0.0205 [0.11] 0.0514 [0.09] -0.103 [0.12] 0.495** [0.22] 0.388** [0.19] -0.0669** [0.03] -0.0289 [0.03] 0.144*** [0.03] 0.104*** [0.03] 0.0203 [0.03] 4782 0.028 12395.9 12609.5 34.72*** 34.89*** 0.0490 0.0171 0.0102 0.0200 0.0219 0.0425 0.0513 0.0246 0.0048 0.0085 0.0098 0.0135 0.0138 0.0019 0.0048 0.0098 0.0452 0.0349 0.0333 0.0120 0.0632 0.0571 0.0081 Note: Ordered logit estimates (using sample weights); robust standard errors in brackets; * p<0.10 ** p<0.05 *** p<0.01; cut-points, constant and country dummies included but not reported; sample includes: United States, Ireland, Czech Republic, Japan, Denmark, Switzerland, South Korea, Party affiliation has not been asked in France. - 35 - Table 4. The effect of personality traits on sector choice Model (1) Model (2) Delta av. Pr min->max “Working for private versus public sector?” Government(1), Private(2), Self employed(3) Governm. Self empl. Governm. Self empl. Governm. Private. Self empl. 0.811*** [0.07] -0.264*** [0.10] 0.0559 [0.06] 0.0852 [0.05] -0.0657 [0.06] 1.033*** [0.19] 0.780*** [0.22] 0.731*** [0.19] -1.465*** [0.36] 1.014*** [0.17] 2.833*** [0.24] 0.563*** [0.19] -0.00966 [0.05] 0.106** [0.05] 0.146*** [0.05] 0.0166 [0.05] 0.0343 [0.05] 0.0701 -0.5305 0.4604 0.0578 -0.0242 -0.0336 0.1853 -0.1813 -0.0040 -0.0649 0.0003 0.0646 -0.0195 0.0379 -0.0184 -0.0841 -0.0712 0.1553 0.0843 -0.1582 0.0739 0.0260 -0.1050 0.0790 0.1017 0.0007 -0.1024 0.0402 -0.1578 0.1176 -0.1218 -0.4441 0.5659 -0.1300 0.0396 0.0904 0.130*** [0.05] -0.0168 [0.05] 0.0937** [0.05] 0.0758 [0.05] -0.0221 [0.05] 0.323*** [0.05] 0.359*** [0.09] 0.393*** [0.06] -0.0558 [0.04] -0.0501 [0.05] -0.493*** [0.19] 0.656*** [0.15] 0.297** [0.14] 0.479*** [0.15] 0.456*** [0.15] -0.169 [0.40] -1.152*** [0.25] -0.0289 [0.05] -0.0459 [0.04] 0.0408 [0.04] 0.0822* [0.04] 0.0456 [0.04] -0.0223 0.0243 -0.0020 -0.0390 -0.0103 0.0493 0.0145 -0.0743 0.0597 0.0626 -0.0633 0.0007 0.0255 -0.0363 0.0108 Age Female Edu. Degree Class Region Manager Academic Technician Office worker Service worker Agricultural w. Craftsmen Conscientious. Extraversion Openness Agreeableness Neuroticism Observations Pseudo R2 AIC BIC LR Test (Big 5) -0.0129 [0.05] -0.0342 [0.04] 0.0712* [0.04] 0.0852** [0.04] 0.0968** [0.04] 5375 0.066 9218.6 9389.9 5375 0.174 8205.1 8534.5 29.37*** Note: multinomial logit estimates (using sample weights) employed only; reference category = 2 (Private); robust standard errors in brackets; * p<0.10 ** p<0.05 *** p<0.01; constant and country dummies included but not reported; sample includes: United States, Ireland, Czech Republic, Japan, France, Denmark, Switzerland, South Korea. - 36 - Appendix Table 1. Definition and coding of variables Name Dependent variables Vote Party affiliation Sector choice Definition and coding Did you vote in last election? Yes=1, No=0 Party affiliation from far right (1) to far left (5) Working for private versus public sector? Government (1), Private (2), Self employed (3) Independent variables Government employee Self employed Working for the government (Dummy) Self employed (Dummy) Personality traits (Big5) - Conscientiousness - Extraversion - Openness - Agreeableness - Neuroticism I see myself as someone how is … (1 to 5 scale) ... does a thorough job, ... tends to be lazy (R) ... is outgoing, sociable, … is reserved (R) ... has an active imagination, ... has few artistic interests (R) ... is generally trusting, ...tend to find fault with others (R) ... gets nervous easily, ... is relaxed, handles stress well (R) Control variables Age Female Edu. Degree Class Region Occupation Employment status Respondents age in years Gender Highest education level / degree from no formal qualification (1) to university degree 6) Top Bottom self-placement 10 pt scale (respondents deviation from the country mean response) Type of community from rural (1) to city (5) via ISCO88: Manager, Academic, Technician, Office worker, Service worker, Agricultural worker, Craftsmen (Reference category: Elementary workers) Employed (full-, part- and less than part-time employed), unemployed, student (student, school, vocational training), retired, other non-working (housewife,-man, home duties, other not in labor force, permanently disabled, helping family member) Note: R=reversed; Age, Degree, Class, Region and the Big 5 Personality traits have been z-standardized in order to compare within model effect sizes. Dataset: International Social Survey Program 2005: Work Orientation III, (ZA4350_F1.dta) - 37 - Appendix Table 2. Personality Traits: Dimensions and scale reliability Dimension Openness Conscientiousness Agreeableness Extraversion Neuroticism Country Pooled United States Ireland Czech Republic Japan France Denmark Switzerland South Korea Pooled United States Ireland Czech Republic Japan France Denmark Switzerland South Korea Pooled United States Ireland Czech Republic Japan France Denmark Switzerland South Korea Pooled United States Ireland Czech Republic Japan France Denmark Switzerland South Korea Pooled United States Ireland Czech Republic Japan France Denmark Switzerland South Korea Mean 7.07 7.26 7.01 6.93 6.43 7.01 7.71 7.26 6.50 8.25 8.55 8.50 7.42 7.93 8.76 8.86 8.40 7.35 7.55 7.53 7.89 6.72 7.73 7.09 7.66 7.72 7.80 6.76 6.65 6.82 6.12 5.97 7.18 7.88 7.22 5.83 5.14 4.89 4.67 5.36 5.97 5.40 4.19 4.87 6.16 Scale reliability 0.31 0.32 0.23 0.39 0.39 0.44 0.39 0.41 0.41 0.48 0.40 0.36 0.41 0.43 0.45 0.34 0.41 0.42 0.13 0.29 0.30 0.12 0.19 0.08 0.19 0.10 0.22 0.45 0.47 0.33 0.22 0.56 0.35 0.61 0.52 0.39 0.55 0.55 0.57 0.52 0.26 0.63 0.54 0.49 0.39 Note: Each personality dimensions is composed of two items (see Appendix Table 1). Each dimension represents an additive index ranging form 2 to 10. - 38 -
© Copyright 2026 Paperzz