Personality, Ideological Attitudes, and Group Identity as Predictors of Political Behavior in Majority and Minority Ethnic Groups

Authors


Abstract

Prior research on personality and politics has largely investigated relationships using national samples from North America and Europe. In contrast, we used multigroup path analysis to assess how Big Five personality, ideological attitudes (RWA, SDO), and group identities (National and Ethnic Identification) predicted right versus left Political Behavior (party support, past voting, present voting intention) across majority and minority ethnic groups in a New Zealand national sample (N= 6,333). The effects of personality on ideological attitudes and group identities were mostly invariant across ethnic groups and consistent with prior findings. In contrast, the effects of ideological attitudes on Political Behavior varied across ethnic groups being moderately strong for the European majority but nonsignificant for the minorities. Group identities had little effect on Political Behavior. We discuss cultural and contextual factors that might account for this disconnect between ideology and politics among the minority ethnic groups.

During the past two decades, researchers have begun to systematically investigate the relationship between personality and left- versus right-wing Political Orientation and Behavior. This became possible with the development of the Big Five (Neuroticism, Extraversion, Openness, Agreeableness, and Conscientiousness) as broad-bandwidth personality dimensions consistently identified using multiple measurement methods (e.g., Costa & McCrae, 1992). Research has focused on three main issues, that is, how the Big Five predict Political Behavior (such as voting for left-right parties or leaders) or self-rated Political Orientation (as liberal or left versus conservative or right), how they predict sociopolitical values or ideological attitude dimensions, and how sociopolitical values or ideological attitudes might mediate the effects of Big Five personality on Political Behavior or Orientation.

We therefore, first, briefly review the research on these issues noting that they have provided a reasonably consistent pattern of findings. Second, we go on to note that these findings have been obtained almost entirely using national samples from North America and Europe in which ethnic majority groups have predominated and that there are indications that the research findings might not be invariant across ethnic majority and minority groups. We also suggest that the possible role of group membership raises the issue of group identification as a possible predictor of Political Behavior over and above any effects of personality and ideological attitudes. And finally, we introduce new research investigating to what extent the predictive effects of Big Five personality on ideological attitudes and group identities, and of all these three factors on Political Behavior, might vary across ethnic majority and minority groups.

Research Findings on Personality and Politics

Studies of how the Big Five predict voting for left-right political parties or candidates have been conducted in North America and Western Europe. Openness, associated with voting left, has been the strongest and most consistent predictor, while Conscientiousness, associated with voting right, and Agreeableness, associated with voting left, have also been consistent predictors (Barbaranelli, Caprara, Vecchione, & Fraley, 2007; Caprara, Barbaranelli, & Zimbardo, 1999; Caprara, Francescato, Mebane, Sorace, & Vecchione, 2010; Caprara, Schwartz, Capanna, Vecchione, & Barbaranelli, 2006; Schoen & Schumann, 2007; Vecchione, Schoen, Castro, Cieciuch, Pavlopoulos, & Caprara, 2011). Less consistent effects have been shown for Extraversion, associated with voting right in Italy (Caprara, Barbaranelli et al., 1999) and the United States (Barbaranelli et al., 2007), and Neuroticism, for which there have been weaker and party or candidate specific associations with left voting in Germany (Schoen & Schumann, 2007) and the United States (Barbaranelli et al., 2007).

Studies investigating the Big Five as predictors of Political Orientation, that is, peoples' self-definition of themselves as politically left-wing (or liberal) versus right-wing (or conservative), have produced broadly similar findings. Thus, a meta-analysis of 73 studies (N = 71, 895) found two Big Five dimensions nontrivially (r = >.10) correlated with Political Orientation. Openness was associated with more left orientation (r = .18) and Conscientiousness weakly associated (r = .10) with more right orientation (Sibley, Osborne, & Duckitt, 2012).

Many studies have also investigated the Big Five as predictors of peoples' ideological attitudes along the two dimensions of Right Wing Authoritarianism (RWA) or Social Conservatism and Social Dominance Orientation (SDO) or Economic Conservatism. A meta-analysis (71 studies with 22,068 participants) found that RWA was significantly associated with lower Openness and higher Conscientiousness, while SDO was associated with lower Agreeableness, but also weakly with lower Openness, although the latter effect was eliminated when RWA was controlled (Sibley & Duckitt, 2008).

Finally, research has investigated the role of ideological attitudes (RWA and SDO) and basic social values, particularly as measured by the Schwartz Values Inventory (SVI; Schwartz, 1992), in predicting left-right Political Orientation and Behavior and how values or ideological attitudes might mediate the effects of Big Five personality dimensions on Political Orientation and Behavior. The findings showed that ideological attitudes, such as RWA (or Social Conservatism) and SDO (or Economic Conservatism), were moderate to strong predictors of left-right voting or Political Orientation in most countries studied, though the effects were sometimes weaker or absent, notably in East European postcommunist countries (Barnea, 2003, cited in Barnea & Schwartz, 1998; Wojcik, Cislak, & Formannowics, 2011).

In addition, basic social values comprising the two higher-order SVI dimensions of Self-enhancement versus Self-transcendence and Conservation versus Openness (which generally correlate moderately to strongly with SDO and RWA respectively; see, e.g., Lee, Ashton, Ogunfowora, Bourdage, & Shin, 2010) have also been strong predictors of right-left voting or Political Orientation (Barnea & Schwartz, 1998; Caprara, Francescato, Mebane, Sorace, & Vecchione, 2010; Caprara, Schwartz et al., 2006; Caprara, Vecchione, & Schwartz, 2009; Piurko Schwartz, & Davidov, 2011). Moreover, several studies in Italy have found that the effects of Big Five on left-right voting or Political Orientation seemed to be fully or substantially mediated via these value dimensions. For example, Caprara, Vecchione et al. (2009) in two studies found that the effects of Big Five Agreeableness and Openness were mediated via Universalism (i.e., Self Transcendence) while the effects of Conscientiousness and Openness was mediated via Security (i.e., Conservation) values. Similar findings, also in Italy, were obtained by Caprara, Francescato et al. (2010) and Caprara, Schwartz et al. (2006).

To sum up, prior research has produced a reasonably consistent pattern of findings. Three of the Big Five Personality dimensions (notably Openness, Conscientiousness, and Agreeableness) have been reasonably consistent predictors of left-right Political Behavior and Orientation. Even stronger effects have been found for ideological attitude (RWA and SDO) and closely correlated value dimensions (Conservation and Self-Enhancement respectively) on Political Behavior or Orientation, and these value dimensions (or ideological attitudes) have been shown to wholly or largely mediate the effects of Big Five Personality on left-right voting and Political Orientation in the one country studied (i.e., Italy).

Personality and Politics: Unresolved Issues

A major gap in research on personality and Political Behavior has been the almost exclusive use of national samples which inevitably therefore consisted primarily of the ethnic majority in each country. Only one study in the United States compared effects of Big Five on political attitudes and orientation for majority (Whites) and minority (Blacks) ethnic groups separately (Mondak & Halperin, 2008) with the findings suggesting there might be important differences with effects tending to be weaker for the minority. However, the nature of these effects were not entirely clear, partly because they were often weak and only statistically significant due to the large sample sizes, partly because effects were not consistent across many criterion dimensions, and also because only one minority group was used raising the possibility that the results might be specific to that particular ethnocultural group.

In addition, although a number of studies have shown that ideological attitudes, such as RWA (or Social Conservatism) and SDO (or Economic Conservatism), were moderate to strong predictors of left-right voting or Political Orientation in North American or European countries, two studies that compared effects across countries did note that these effects sometimes tended to be weaker or absent, notably in East European postcommunist countries (Barnea, 2003, cited in Barnea & Schwartz, 1998; Wojcik et al., 2011). This again raised the possibility that the consistent pattern of moderately strong effects for personality or ideological attitudes on Political Behavior might not generalize across ethnocultural groups and might be notably weaker in ethnic minority or nonwestern groups than in the ethnic majority and western samples that have thus far been predominantly studied.

A focus on the possible role of ethnocultural group differences in the prediction of Political Behavior also raises the issue of possible effects of group identification, particularly with those ethnocultural groups or with the national group. A great deal of research has shown the importance of group identities for social behavior in general (see, e.g., Brown, 2000) and for political involvement and action in particular (e.g., Huddy, 2013). It has been shown that Big Five personality factors do reliably predict identities, such as national identity (e.g., Sagiv, Roccas, & Hazan, 2012). Moreover, although there has been surprisingly little research on the issue, it has been argued that ethnic and national identities might have important effects on left-right Political Behavior and Orientation for majority and minority groups (Amit-Talai & Knowles, 1998). Since right-wing parties almost universally tend to focus more on national cohesion and left-wing parties on minority rights, it seems likely that National Identity would be more strongly associated with support for right-wing parties (for both majority and minority ethnic groups). In contrast, Ethnic Identity would be more strongly associated with support for left-wing parties for minorities and for right-wing parties for the majority ethnics (e.g., Amit-Talai & Knowles, 1998; Conover & Feldman, 2004; Farnen, 1994).

These findings and this reasoning suggests the possibility that group identities such as national and ethnic identity might have a similar role to ideological attitudes in mediating effects of Big Five personality on left- versus right-wing Political Behavior such as party support and voting but with effects over and above those of ideological attitudes. This also raises the possibility that if the effects of ideological attitudes on Political Behaviors are indeed weak or absent for minority ethnocultural groups, this may be because group identity, and particularly ethnic identity, may a more important predictor in these minorities.

The primary objective of the current research was therefore to systematically compare how the Big Five personality dimensions predicted both ideological attitudes (RWA and SDO) and group identities (National and Ethnic Identity) and how personality, ideological attitudes, and group identities together predicted left versus right Political Behavior (party support and voting) for the European origin ethnic majority and the main ethnic minorities in New Zealand.

The Current Research: Testing a Multigroup Model

The current research analyzed data from a large national sample of New Zealand voters using a multigroup path analysis to compare effects of personality, ideological attitudes, and group identities on Political Behavior across majority and minority ethnic groups. The model (shown in Figure 1) embodied several assumptions derived from theory and prior findings.

Figure 1.

The path analytic model for Big Five Personality, Ideological Attitudes (RWA, SDO), and Group Identification (National, Ethnic), and Political Behavior (Pol-Behav) with broken arrows indicating the paths tested in the multigroup analysis.

The primary assumption was that personality influences the value-based ideological attitudes of RWA and SDO, which then influence Political Behavior (PB), with ideological attitudes (RWA, SDO) partly or wholly mediating effects of personality (see, e.g., Barnea & Schwartz, 1998; Caprara, Schwartz et al., 2006; Caprara, Vecchione et al., 2009; Caprara, Francescato et al., 2010; Piurko et al., 2011; Van Hiel & Mervielde, 2002). We added to this the assumption that Ethnic and National Identity might play a similar mediating role to that of RWA and SDO.

This multigroup path analysis would then enable us to assess how the Big Five would predict PB either directly or indirectly via effects on ideological attitudes and group identities. Following the prior research already reviewed, we hypothesized that the effects of the Big Five on PB would be largely or entirely indirect with Openness and Conscientiousness having effects via RWA and Agreeableness via SDO. Following Sagiv et al.'s (2012) findings, we hypothesized that lower Neuroticism and higher Agreeableness and Conscientiousness would predict National and Ethnic Identification, which in turn would predict PB with National Identification associated with right (conservative) PB for all groups, and Ethnic Identification with left (liberal) PB at least for the ethnic minority groups.

Finally, the multigroup analysis would indicate if the effects of personality on ideological attitudes and group identities, and of ideological attitudes and group identities on PB, might differ systematically across ethnic majority and minority groups. We expected the effects for the ethnic majority (European-Pakeha) to be similar to those from previous research which had almost entirely used samples consisting predominantly of ethnic majority members in western societies. However, because several studies have suggested the effects of ideological attitudes or values on PO or PB might be weaker in nonwestern than in western societies (Barnea, 2003, cited in Barnea & Schwartz, 1998; Wojcik et al., 2011), we tentatively expected that these effects might be weaker in the primarily nonwestern ethnic minorities in New Zealand. Finally, we tentatively expected that effects of group identity and particularly ethnic identity on PB might be stronger for the ethnic minorities.

Method

Sample

The data were drawn from the 2009 New Zealand Attitudes and Values Study (New ZealandAVS-09), which comprised a postal questionnaire sent to 40,500 participants randomly sampled from the 2009 New Zealand (New Zealand) electoral roll. The overall response rate was 16.6% (6,723). This low response rate may be at least partially due to the fact that people were opting into a 20-year longitudinal study, of which the data analyzed here were the first wave. While this response rate was low, response rates of this magnitude have not been uncommon for postal questionnaires (see, e.g., Kerlinger, 1986; Krosnick, 1999). The analyses used those respondents who classified their ethnicity as one of European/Pakeha (n = 4629), Maori (n = 1112), Pacific Nations (n = 280), or Asian (n = 312), giving a total sample of 6,333. These are the four main ethnic groups in New Zealand accounting for approximately 99% of the population (Ministry of Social Development, 2010). Participants who reported a different ethnicity, who reported multiple identities, or who did not report their ethnicity were discarded.

The mean age of the sample was 48.0 years (SD = 15.8), 59.5% were female, and the mean household income was $New Zealand 83,182 (SD = New Zealand$ 62,770). In educational level, 31.9% had or were studying toward a tertiary qualification (university/college degree/diploma), 26.4 had completed high school, 30.7% had some high school, and 11.1% less than high school or unreported. The sample thus provided a broad cross section of the New Zealand adult population, reasonably representative in ethnic composition (Ministry of Social Development [2010] estimates were 77% European/Pakeha, 15% Maori, 7% Pacific Nations, and 10% Asian in 20061 ), though tending to be somewhat more female, older, better educated, and higher income than the general population.

Initial analyses revealed that the effects for the Maori and Pacific Nations groups were extremely similar. This was not surprising since Maori and Pacific peoples are both ethnically Polynesian and similar culturally and linguistically. Because of this, we combined these two groups and report effects only for this combined Maori-Pacific group (n = 1, 392). Thus, the primary analyses reported effects across three ethnic samples, the majority European-Pakeha group and the Maori-Pacific and Asian ethnic minority groups.

Measures of Personality, Ideological Attitudes, and Group Identification

  • 1)RWA and SDO were each measured using short six-item balanced scales with items randomly sampled from Altemeyer's (1996) RWA scale and Pratto, Sidanius, Stallworth, and Malle's (1994) SDO scale. Both shortened scales had previously been used to assess RWA and SDO in New Zealand (e.g., Duckitt, 2006). The alpha coefficients in the full sample were .70 and .69 respectively and were similar across the three subgroups, as was the case for all the other measures used here.2
  • 2)Big Five Personality was measured using the Big Five Mini-IPIP scales developed by Donnellan, Oswald, Baird, and Lucas (2006) with four-item balanced scales for each of the constructs. Illustrative items are “Get upset easily” (Neuroticism), “Am the life of the party” (Extraversion), “Have a vivid imagination” (Openness), “Sympathize with others' feelings” (Agreeableness), and “Get chores done right away” (Conscientiousness). The alphas were .64 (Neuroticism), .71 (Extraversion), .67 (Openness), .66 (Agreeableness), and .65 (Conscientiousness).
  • 3)National and Ethnic Identification. National Identification was measured using two items from Kosterman and Feshbach's (1989) Patriotism scale adapted for New Zealand (“I feel a great pride in the land that is our New Zealand” and “Although at times I may not agree with the government, my commitment to New Zealand always remains strong.”), with an alpha of .69. Ethnic Identification was measured using Leach et al.'s (2008) Centrality of Identity scale adapted for ethnic identity. The items were “I often think about the fact that I am a member of my ethnic group”; “The fact that I am a member of my ethnic group is an important part of my identity”; and “Being a member of my ethnic group is an important part of how I see myself.” The alpha was .84.

All the scales were responded to on 7-point response scales ranging from very accurate to very inaccurate for the personality scales and strongly agree to strongly disagree for the ideological attitude (RWA, SDO) and group identity scales.

Political Behavior

Political Behavior was assessed by an index comprised of political party support, party voted for in the previous election, and present voting intention.

  • 1)Political party support. Participants rated their level of support for each of the parties represented in the New Zealand parliament at the time of the research (National, Labour, Greens, ACT, United Future, Maori, Progressive) on a scale ranging from 1 (strongly oppose) to 7 (strongly support). These ratings had been used in prior New Zealand research by Sibley and Wilson (2007).
  • 2)Voting: Previous and future intended voting behavior was assessed using two items. One item asked participants to indicate which of the political parties noted above they had voted for in the past election, and a second item asked which of these parties they would vote for “if an election were to be held today.”
  • 3)Scoring party support, voted last election, and would vote now. Two main political groupings organized around two large parties have dominated New Zealand politics for most of its history. A liberal or left coalition has been centred on the Labour Party (34% support last election) and a conservative or right coalition centred on the National Party (44.9% support last election). The smaller Green Party (6.7% support last election) and Progressive Party (0.9% support last election) have been consistent Labour allies, while the small ACT Party (3.7% support last election) has been a consistent National ally. In most elections, these two consistent coalitions can be expected to gain over 90% of the vote. The remaining two small parties, the Maori Party (2.4% support last election) and United Future (0.9% support last election) have not been consistent allies of these two large blocs, having in recent years each joined coalition governments with both Labour and National. Political Party support was therefore scored by computing support for Conservative versus Liberal parties by subtracting mean support for the consistent Left/liberal parties (Labour, Greens, Progressive) from the mean support for the consistent Right/Conservative (National, ACT) parties. Vote in the past election and Would vote now were simply scored as having voted for or intending to vote for the Right (National Party dominated) bloc (scored as 1) or for the Left (Labour party dominated) bloc (scored as 0).
  • 4)The Political Behavior (PB) index. The Party Support, Would Vote Now, and Voted Past Election indices were highly intercorrelated in both the total sample (range .75–.85) and the three ethnic samples separately (range .61–.77) giving an alpha coefficient in the total sample of .92 (alphas for the three ethnic groups separately varied from .88 to .92). Because the three indices were scaled differently, they were combined into a single PB index using the first common factor from a principal factors analysis, which was then used in the correlational and multigroup analyses.

Results

We first report the correlations between personality, ideological attitudes, group identities, and PB, followed by the multigroup path analysis to assess the effects of personality mediated via ideological attitudes and group identities on PB across the three ethnocultural groups. We did not use the usual 5% or 1% statistical tests for significance because the large samples meant that even trivial effects would be significant. We therefore used Cohen's (1988) well-known conventions for assessing effect size, so that a correlation or a standardized3 beta path coefficient would need to be at least .10 to be viewed as a weak but significant effect (see also Kline, 2011). In the multigroup analysis, therefore, all paths where no standardized3 effect exceeded .10 were deleted as nonsignificant, unless deleting them produced a significant decrease in model fit (i.e., an increase in RMSEA ≥ .01, following Chen, 2007).

Correlational Findings

The correlations between the Big Five, RWA, SDO, Ethnic Identification, National Identification, and PB for the full sample are shown in Table 1. In general, these correlations were consistent with expectation and prior research. Thus, the correlations between the Big Five were all weak (the strongest being .25) as were those between RWA, SDO, National Identification, and Ethnic Identification (the strongest being .19 between RWA and SDO), indicating the relative independence among these constructs.

Table 1. Correlations among Big Five personality (N, E. O, A, C), Right Wing Authoritarianism (RWA), Social Dominance Orientation (SDO), National Identification (Nat Id), Ethnic Identification Eth Id), and Political Behavior (PB)
VariablesNEOACRWASDONat IdEth Id
  1. Note. N = Neuroticism, E = Extraversion, O = Openness to Experience, A = Agreeableness, C = Conscientiousness
E−.08       
O−.02.25      
A−.03.21.25     
C−.12.01.01.15    
RWA−.01−.07−.24.00.11   
SDO−.01.00−.17−.29−.02.19  
Nat Id−.12.07−.00.14.10.12−.12 
Eth Id.05.04−.05−.01.01.18.01.15
PB−.07.02−.12−.06.10.22.26.04−.14

As hypothesized, and consistent with prior meta-analysis (Sibley & Duckitt, 2008), the strongest Big Five personality predictor of RWA was low Openness (−.24) with a weaker effect for high Conscientiousness (.11), while the strongest predictor of SDO was low Agreeableness (−.29) with a weaker effect for low Openness (−.17). Consistent with Sagiv et al.'s (2012) findings, lower Neuroticism and higher Agreeableness and Conscientiousness predicted National Identification (−.12, .14, .10 respectively), but contrary to expectation, no personality factors predicted Ethnic Identification.

Also as expected from prior research, the strongest correlates of PB were RWA (.22) and SDO (.29). Only two Big Five personality dimensions had correlations with PB equal to or exceeding .10, that is, Conscientiousness (.10) and Openness (−.12), which was also consistent with recent meta-analytic findings for the Big Five and Political Orientation (Sibley et al., 2012). Contrary to expectation, National Identification did not predict PB significantly, but as expected for minorities, Ethnic Identification did as have an overall significant negative effect (−.14).

Multigroup Path Analysis

The multigroup path analysis was used to compare the effects of personality, ideological attitudes, and group identities on PB across the three ethnic groups (the model tested is shown in Figure 1). The advantage of the multigroup analysis was that it enabled all effects including possible mediation of personality via ideological attitudes and group identities to be computed and compared across groups in a single integrated analysis.

We followed the generally recommended procedure for multigroup path analysis (Byrne, 2006; Caprara et al., 2010). We initially tested path analytic models for each of the groups separately, testing all possible paths and deleting nonsignificant paths. In this way, we obtained good-fitting models for each of the three groups separately using Hu and Bentler's (1999) recommendations that good fit would be indicated by values close to or better than .06 for RMSEA, .08 for SRMR, and .95 for GFI and CFI. We then tested a multigroup model for all three groups incorporating all paths that had been significant in any group, with all paths, correlated errors, and error variances allowed to vary across groups. This unconstrained model had good fit with: χ2 = 311.83, df = 88, RMSEA = .035 (.031–.039), SRMR = .069, GFI = .97, CFI = .94.

Next, we systematically constrained effects to be equal across the groups, unless this resulted in a significant decrease in fit, in which case the effects were allowed to vary. The criterion for a significant decrease in fit was an increase of RMSEA ≥ .01, which has been recommended with large samples for which the chi-square difference test would be too sensitive (Chen, 2007). Paths where all effects were nonsignificant were deleted from the model. In this way we obtained a final multigroup model with good fit (χ2 /df = 300.38/107, RMSEA = .029 (.026–.033), SRMR = .073, GFI = .96, CFI = .95).

This final multigroup model is shown in Figure 2. In order to enable direct comparisons across groups, unstandardized3 path coefficients are shown for the European/Pakeha, Maori/Pacific, and Asian ethnic groups in that order. Equal path coefficients or only one coefficient for a path indicates that those paths were constrained to be equal without a significant decrease in model fit. To show more clearly which paths were constrained equal across all groups, these paths have darker bolded arrows. Correlated errors were permitted between RWA, SDO, National Identity, and Ethnic Identity, but they are not shown in order to simplify the diagram.

Figure 2.

Unstandardized coefficients for the multigroup path analysis of Big Five personality, ideological attitudes (RWA, SDO), and group identities (NAT ID, ETH ID) on more conservative Political Behavior (Pol-Beh) across three ethnic groups (European/Pakeha, Maori/Pacific, and Asian, with unstandardized path coefficients for these groups given in that order for each path). Equal betas or only one beta for a path indicate paths that were constrained to be equal without a significant decrease in model fit. Dark bold arrows indicate paths where all effects were constrained to be equal, while lighter arrows indicate paths where effects varied between groups. Paths that could be constrained equal but where the common metric standardized beta was <.10 were treated as nonsignificant and deleted from the model unless they resulted in a significant decrease in model fit. Correlated errors were permitted between RWA, SDO, Nat ID and Eth ID but are not shown in order to simplify the diagram.

+indicates path coefficients that were nonsignificant (i.e., with a common metric standardized Beta <.10).

Looking at the effects of personality, there appear to be four notable findings. First, all the significant paths obtained were generally consistent with expectation and prior findings. Thus, low Agreeableness was the strongest predictor of SDO, followed by weaker effect for low Openness. The strongest predictor of RWA was low Openness, followed by an overall weaker effect for Conscientiousness. National Identity was predicted by lower Neuroticism and higher Agreeableness, which was consistent with the prior findings of Sagiv et al. (2012), with both effects relatively weak (standardized3 betas of −.12 and .13), but as for the correlational findings, there were no significant effects for any of the Big Five personality on Ethnic Identity.

Second, all but one of the six significant paths for personality on ideological attitudes and group identity could be constrained to be equal across all three groups with only one effect varying significantly across groups. Thus, with only one exception, the effects of personality on ideological attitudes and group identities did not vary across groups. Third, the one exception to this was for the effect of Conscientiousness on RWA. In this case the effects for two of the ethnic groups, the majority European-Pakeha and the minority Maori-Pacific groups, were relatively weak and similar in magnitude to the effects obtained in prior research (cf. Sibley & Duckitt, 2008), but the effects for the minority Asian ethnic group were significantly and markedly stronger.

Fourth, the results showed there were no significant direct effects of Big Five personality on PB. Thus, consistent with expectation, any effects of personality on PB would necessarily have to be indirect and mediated via ideological attitudes or groups identities. In this case, the pattern of effects for ideological attitudes and group identities still to be described indicates that the mediation could only be via ideological attitudes (RWA and SDO).4

In striking contrast to the direct effects for personality which were generally invariant across groups, the effects for ideological attitudes and group identities on PB all varied significantly across groups. In the case of ideological attitudes, the findings were exactly as had been hypothesized. The effects for RWA and SDO on PB were significant for the majority European-Pakeha ethnic group and in magnitude similar to those obtained in prior research in North America and Western Europe (standardized betas of .21 and .26 respectively, with a multiple correlation on PB of .35). In contrast, the effects in both the two minority ethnic groups (Maori-Pacific and Asian) were significantly weaker and indeed all nonsignificant.

The findings for group identity on PB, however, were generally inconstant with our expectations with little or no significant effects evident. Thus, as for the correlational findings, there was no significant effect for National Identification on PB. In the case of Ethnic Identification, there was a significant effect for stronger Ethnic Identity to be associated with more liberal PB for the Maori-Pacific minority, but this was not very powerful in magnitude (standardized beta = −.18), and the effects were nonsignificant for both the other minority (Asian) and the European-Pakeha majority group. Thus, the only effect for group identity on PB was a relatively weak effect for Ethnic Identity specific to only the Maori-Pacific minority.

Alternative Models

We also assessed a series of alternative models. First, a fully constrained model, in which all effects in our final model (shown in Figure 2) were invariant across groups, had fit indices inferior to those for our final model, being: χ2 = 479.03, df = 85, RMSEA = .047 (.043–.051), SRMR = .093, GFI = .94, CFI = .89. The 90% confidence intervals for RMSEA for this model did not overlap with the final partially constrained model.

We next tested a version of our final model which included the four major sociodemographic factors of age, gender, educational level, and household income in order to assess if their inclusion might alter the effects observed. This model also had good fit (χ2 = 489.28, df = 111, RMSEA = .040 (.037–.044), SRMR = .060, GFI = .97, CFI = .95). The sociodemographic variables had significant effects on RWA, SDO, National and Ethnic Identity, but left the pattern of significant effects of personality on ideological attitudes and group identities and of all variables on PB unchanged with only slight changes in the magnitudes of the coefficients. There was only one significant effect of a sociodemographic variable on PB, with higher income being a weak predictor of more conservative PB for the European-Pakeha group, a somewhat stronger predictor for the Maori-Pacific group and nonsignificant for the Asian group.5 Thus, the inclusion of the sociodemographic variables did not notably alter any of the effects obtained in our analysis or any of the conclusions suggested.

We also tested a model in which the personality variables mediated effects of ideological attitudes and group identities on PB, which as expected had very poor fit. Further alternative models were tested in which either group identities or ideological attitudes or both were modelled as exogenous variables (with personality), but the basic effects obtained from our original analysis also emerged clearly in these altered models to the extent that these models incorporated them, providing some support for their robustness.

Discussion

The primary objective of the research was to investigate how personality, ideological attitudes, and group identities predicted PB across majority and minority ethnic groups. The multigroup path analysis suggested three main sets of conclusions with all involving important cross group similarities or dissimilarities. These pertain to, first, the effects of personality on ideological attitudes and PB; second, the effects of personality on Group Identities and of Group Identities on PB; and third, the effects of ideological attitudes on PB.

Effects of Personality on Ideological Attitudes and PB

There were three particularly striking features of the results here. The first was that the findings for the effects of personality on ideological attitudes were highly consistent with prior findings and that the effects were largely invariant across the three groups. The second was that there was one interesting case where effects did vary across groups; and the third was that the effects of personality on PB were fully mediated.

The findings here that low Openness was the strongest predictor of RWA and that high Conscientiousness was an overall weaker predictor, as well as that low Agreeableness was the strongest predictor of SDO, with low Openness a secondary and weaker predictor were fully consistent with prior meta-analytic findings (Sibley & Duckitt, 2008). These prior findings, however, were derived almost entirely from western (North America and Europe) samples predominantly comprised of ethnic majorities in their societies. The current research therefore extended these findings by showing that these effects were in most cases invariant across the western ethnic majority group (European-Pakeha) and our two nonwestern ethnic minority groups (Maori-Pacific and Asian). This suggests that the way in which personality traits express themselves in people's sociopolitical or ideological values and attitudes may indeed be similar across ethnic majority and minority groups.

Second, there was, however, one interesting exception to this cross-group invariance. For two of our groups (the ethnic majority European-Pakeha and the ethnic minority Maori-Pacific group), the effects of low Conscientiousness on RWA was consistent with the pattern in prior studies of weak but significant effects. However, the effects for the Asian ethnic minority group, although in the same direction, were significantly stronger, and for this group was the strongest predictor of RWA. While this effect clearly needs replication, Conscientiousness has been noted to be a particularly highly valued personality trait in East Asian Collectivist cultures (see, e.g., Kim, Atkinson, & Yang, 1999). This may, therefore, account for its greater importance in predicting ideological attitudes for Asian peoples.

The third noteworthy finding was that there were no significant direct effects for personality on PB. Thus, the effects of personality on PB were fully mediated via RWA and SDO.4 This finding is reasonably consistent with the prior research in Italy finding that personality effects on left-right Political Orientation or voting were largely or entirely mediated by Schwartz's value dimensions closely associated with RWA and SDO (Caprara, Schwartz et al., 2006; Caprara, Vecchione et al., 2009; Caprara, Francescato et al., 2010).

Role of Group Identities

The multigroup path analysis showed two significant effects of personality on Group Identities that were invariant across all three groups, with higher Agreeableness and lower Neuroticism predicting National Identity. These effects were weak with standardized betas of .12 and −.12. No relevant prior studies for Big Five personality and Ethnic Identity were traced, but one prior study had reported significant correlations of higher Agreeableness and lower Neuroticism on stronger National Identity in an Israeli sample (Sagiv, Roccas, & Hazan, 2012). Sagiv et al.'s study had also obtained significant correlations for Conscientiousness on National Identity, and this was also replicated by the correlational analyses in this research, but this effect did not reach significance in the multigroup path analysis. The consistency of these findings for the personality predictors of National Identity across these two studies, and across very different ethnic samples and contexts does provide some support for the generality of the findings.

Group Identities also had little effect on PB with only one significant and relatively weak effect for stronger Ethnic Identity predicting more left PB in only the Maori-Pacific group and not the other two groups. It is not difficult to explain this particular effect. In New Zealand, Maori and Pacific Nations people have had a long history of involvement with the Labour Party both as individual members and through trade union membership, and the Labour Party (the dominant party of the political left) has also been a strong advocate for Maori and Pacific interests. This has not been the case for the Asian minority, which has had no particular involvement with the Labour Party and for which there has been no particular advocacy from the political left, which would account for the lack of any effect for this group.

More surprising, perhaps, might be the relative weakness of the effect for Ethnic Identity for the Maori-Pacific group given the very powerful ethnic loyalties that can characterize identity politics (Posner, 2005). This could possibly be explained by the New Zealand Labour Party despite its association with and advocacy for the Maori-Pacific minority, not being in any way an ethnic party as such.

The conclusion this suggests is that ethnic group identification may not be generally associated with left-right political behavior unless a particular ethnic group has a strong and clear association with a particular party on either the left or right which has been active in representing the group's interests. The strength of any such association may then depend on the degree to which the party directly identifies itself with and represents that ethnic group specifically.

Effects of Ideological Attitudes on Political Behavior

The most striking finding from this research was that the effects of the ideological variables (RWA, SDO) on PB varied significantly across groups, being moderately strong in magnitude for the European-Pakeha group (i.e., with multiple R = .35) and roughly consistent with prior research in western predominantly majority group samples (Caprara, Schwartz et al., 2006; see also Caprara et al., 2009; Caprara, Francescato et al., 2010) but nonsignificant for the two ethnic minority groups. This was in sharp contrast to the effects of personality on the ideological variables which were mostly invariant across groups and consistent with prior research. The important conclusions suggested by these findings are that peoples' personality dispositions seem to influence ideological attitudes and orientations, which tend to be expressions of personal values, in much the same way across majority and minority ethno-cultural groups. However, whereas these value-based ideological attitudes and orientations have relatively strong effects on political behavior for western majority groups, their effect seems to be much weaker and often negligible for nonwestern minority groups.

Although no prior research has looked directly at possible cross-group differential effects of ideological attitudes or personal values on PB, some research has been reported on cross national differences for Political Orientation. Both these investigation have reported that the effect of ideological attitudes on Political Orientation did indeed vary across national samples in Europe with the effects being stronger in Western Europe and tending to be weaker or even absent in postcommunist East European countries and certain south European countries, such as Portugal (Barnea, 2003, cited in Barnea & Schwartz, 1998; Piurko et al., 2011; Wojcik et al., 2011). These prior findings suggest that a disconnect between ideological attitudes and PB may not be due to majority-minority status per se but due to other ethno-cultural or contextual factors that happened to be associated with minority status in this study.

Two possible explanations seem noteworthy. One is in terms of cultural individualism-collectivism. European-Pakeha New Zealanders and most of the North American and West European samples studied in prior research on personality and politics tend to fit the individualist cultural pattern (Hofstede, 1980; Schwartz, 1992), while the Maori-Pacific and Asian ethnic minorities in New Zealand tend to fit a more collectivist cultural pattern (Liu, McCreanor, McIntosh, & Teaiwa, 2005). It seems likely that people from both individualist and collectivist cultures might express their personality dispositions in their personal values (and hence ideological attitudes). However, because ideological attitudes and orientations are expressions of personal values (and individual personality), they might affect social and collective behaviors, such as politics, only for individualists and not for collectivists. This explanation would also broadly fit the pattern observed in the prior research of differential effects between West and East (and certain South) European countries by Piurko et al. (2011) and Wojcik et al. (2011).

If these differences can be attributed to differences in cultural individualism-collectivism, this raises the question of what social or collective factors might influence political behavior for collectivists? One obvious possibility would be strength of group identity. However, our findings show that ethnic identity had only a relatively weak effect on PB for only one of the two ethnic minorities (Maori-Pacific) and none at all for the second (Asian), which seems to largely eliminate this possibility. Another possibility is that significant others in peoples' primary social groups, such as members of their extended family and work groups, might have strong influences on political behavior for collectivists but not for individualists.

A second possible explanation for the group differences in effect of ideological attitudes on political behavior is that threat may moderate the effects of value-based ideological attitudes on political behavior. There is evidence that minorities experience higher levels of threat or insecurity than majority groups (e.g., Henry, 2011). There is also evidence that Big Five Openness is significantly related to Political Orientation when threat levels are low but not when they are high (Sibley et al., 2012). Extrapolating from this suggests that ideological attitudes may be more strongly related to political behavior for groups experiencing low threat and less strongly related for groups experiencing high threat. In the latter case, when groups (or individuals) experience high levels of threat, variables related to threat experience and coping may become more important determinants of political behavior than personal dispositions such as personality or ideological attitudes. The differential effects between West and East (and certain South) European countries by Piurko et al. (2011) and Wojcik et al. (2011) also seems compatible with this explanation since postcommunist East European countries and economically disadvantaged south European countries seem likely to have experienced higher threat levels than more prosperous and stable West and North European counties. Clearly, more research is needed here since there are numerous factors that might influence how groups (and individuals within groups varying in political engagement) might respond politically (along different behavioral dimensions) to threats of varying kinds and intensities.

Conclusions

The findings suggest three important conclusions. First, the effects of personality on ideological attitudes and orientation were consistent with prior findings and mostly invariant across three ethno-cultural groups, suggesting that these effects may indeed be general across majority and minority ethnic groups. The exception was that the effect of Conscientiousness on RWA was stronger for the Asian group, possibly because this trait is particularly highly valued in Asian cultures and may therefore be more likely to be expressed in ideological attitudes.

Second, there were significant but weak effects of personality on National Identity, which were consistent with prior research and were invariant across groups. However, the only significant though relatively weak effect for group identities on PB was for stronger ethnic identity predicting more liberal PB for the only the Maori-Pacific ethnic minority, possibly due to its strong historical association with the New Zealand Labour Party. This suggests that group identities may have little effect on political behavior unless there is some direct political connection between political party and ethnic group, as in the case of identity politics.

Third, the effects of ideological attitudes (RWA, SDO) on PB varied markedly across groups, being moderately strong for the majority European and nonsignificant for the two ethno-cultural minorities. This did parallel prior research suggesting a similar disconnect between ideological attitudes and politics for postcommunist East European countries and certain South European countries in comparison to West and North European countries. It was suggested this might be related to group differences in individualism-collectivism or levels of social threat.

Acknowledgments

John Duckitt and Chris G. Sibley, University of Auckland, Auckland, New Zealand. Correspondence should be addressed to John Duckitt, Psychology Department, University of Auckland, Private Bag 92019, Auckland, New Zealand. E-mail: j.duckitt@auckland.ac.nz.

Footnotes

  1. 1

    The total exceeds 100% because the estimates for each group include persons indicating membership in more than one ethnic group.

  2. 2

    The lowest and highest alphas for the three ethnic groups were as follows: RWA (.64–.71), SDO (.61–.73), National identification (.62–.72), Ethnic Identification (.79–.83), Extraversion (.62–.74), Openness (.62–.69), Agreeableness (.61–.69), Conscientiousness (.61–.67), and Neuroticism (.57–.67).

  3. 3

    We used LISREL's common metric completely standardized betas in order to gauge effect size. Since these betas are computed from the weighted average of covariance matrices across groups, Joreskög and Sörbom (2001, pp. 290–293) have argued they can be directly compared across groups. Others (e.g., Kline, 2011), however, have suggested that unstandardized betas are still preferable for direct cross group comparisons. We therefore reported only unstandardized beta coefficients in Figure 2 where direct cross group comparisons are relevant, but since standardized betas give more easily interpretable indications of actual effect size, we have noted the common metric standardized betas in text when the issue of the magnitude of effects is relevant.

  4. 4

    The strongest indirect effects for personality on PB were for Agreeableness for which the common metric standardized betas (unstandardized in parentheses) for European-Pakeha, Maori-Pacific, and Asian were −.07 (−.06), −.02 (−.02), −.02 (−.02) respectively, and for Openness for which the respective betas were −.08 (−.05), −.02 (−.01), −.02 (−.01). Given the failure of RWA and SDO to predict PB for the two minorities described in the next paragraph, it is evident that the indirect effects for Agreeableness and Openness on PB though generally weak were far stronger for the majority European-Pakeha group than for the two minorities where they were virtually zero.

  5. 5

    The common metric standardized betas were .15, .24, and −.02 respectively.

Ancillary