The Political Self: How Identity Aligns Preferences With Epistemic Needs

Abstract

Numerous studies have indicated that the need for closure predicts political preferences. We examined a potential moderator of this relationship: political-identity centrality, or the extent to which individuals’ political preferences are central to their self-concept. We tested three hypotheses. First, we predicted that need for closure would be more strongly related to political identity (symbolic ideology and party identification; Hypothesis 1) and issue positions (operational ideology; Hypothesis 2) among individuals who see their political preferences as more self-central. Then we predicted that the stronger relationship between need for closure and issue positions among individuals high in centrality would be accounted for by stronger relationships between need for closure and political identity and between political identity and issue positions (Hypothesis 3). Data from a nationally representative survey provide evidence for these hypotheses, suggesting that the relationship between epistemic needs and political preferences differs as a function of how self-relevant politics is.

Keywords

political psychology need for closure ideology identity open data

Liberals and conservatives in the United States have become bitterly polarized in recent years (Abramowitz, 2010; Iyengar, Sood, & Lelkes, 2012). As competing ideological groups have become more distinct, cohesive, and mutually antagonistic, behavioral scientists interested in political polarization have focused increasingly on the deeper psychological differences between those drawn to the political left and those drawn to the political right. In this vein, research has suggested that individuals high in need for closure are more likely to report conservative identities and attitudes, whereas those low in need for closure lean in a liberal direction (Jost, Glaser, Kruglanski, & Sulloway, 2003). But the conditions under which this relationship obtains may be more specific than initially believed. Filling this gap, we explored the role of political-identity centrality, or the extent that one’s political preferences are central to the self-concept. We present evidence that the need for closure is more strongly related to two political identifications (ideological self-identification, or symbolic ideology, and party identification) and issue attitudes (or operational ideology) among individuals who see their political preferences as more central to their identities. Our results suggest that psychological differences may polarize primarily those liberals and conservatives who see politics as a means of self-expression.

The Epistemic Bases of Ideology

As noted, individual differences in epistemic motivation have a robust relationship with political preferences (Jost, Federico, & Napier, 2009, 2013). The need for cognitive closure (Webster & Kruglanski, 1994)—a motivation to arrive at decisions quickly and to adhere to those decisions once made—is one example. People who show the “seizing” and “freezing” tendencies characteristic of high need for closure are more likely to identify as conservatives and to adopt conservative issue preferences (Jost et al., 2003). Conversely, those low in need for closure—who show an active preference for deliberation and “keeping their options open”—are more likely to identify as liberals and to adopt liberal issue attitudes. Note that need for closure has been studied as a motive that differs both as a function of situations and individual differences in personality; here, we focused on the role of need for closure as a personality difference. Thus, we regard an individual’s need for closure as a trait-level aspect of the self.

Moderators of the Relationship Between Epistemic Needs and Ideology

But just how fundamental and pervasive is the relationship between need for closure and politics? We contend that political preferences will be more polarized as a function of need for closure among individuals whose political identifications and attitudes are more central to the self. For those whose self-concepts are defined by their political convictions, politics presents an opportunity for self-expression (Cohen, 2003; Kahan, 2015; Sears, 1993). They can make statements about who they are by adopting political identities and supporting policies that align with fundamental aspects of the self, such as those embodied in personality. For the many others who do not invest their self-concept in politics, political preferences may say little about the self, even if they may desire one political outcome over another for practical reasons. In the language of functional theories of attitudes, the preferences of individuals high in identity centrality are more likely to serve an expressive function: They say more about who the person is by reflecting core aspects of personality more strongly (Katz, 1960; Shavitt, 1989; Smith, Bruner, & White, 1956).

The idea that political affiliations can comprise a cherished aspect of oneself is an old one (Campbell, Converse, Miller, & Stokes, 1960). However, recent work provides evidence that the centrality of political preferences to the self-concept has important correlates in polarized political behavior. Stronger social identification with labels like “liberal” and “conservative” predicts participation in campaigns (Huddy, Mason, & Aarøe, 2015), label-consistent voting behavior (Devine, 2015), and label-consistent policy positions (Malka & Lelkes, 2010).

As one aspect of the self, personality traits—and dispositional need for closure in particular—may therefore be expressed in polarized political preferences to a greater extent when politics is central to the self-concept. People high in need for closure and high in identity centrality might attach themselves to conservative groups and policies because doing so signals that they are efficient, decisive, and steadfast. Those low in need for closure and high in identity centrality might attach themselves to liberal groups and policies to signal that they are circumspect, complex, and open-minded. That is, if need for closure is a meaningful part of individuals’ personalities, then those individuals who construe politics as an expression of who they are should be more inclined to adopt political preferences that comport with their dispositional need for closure. Those low in political-identity centrality, in contrast, should adopt political preferences without regard for their implications for the self-concept. Among these individuals, there is no reason that political preferences would be related to any aspect of their self-concept, need for closure included.

Although this moderating role of identity centrality has not yet been directly examined, the research on the moderating effects of political information is consistent with it. People vary considerably in the amount of information they possess regarding political institutions, issues, and figures (Delli Carpini & Keeter, 1996; Lupia, 2015), and those who are better informed more often sort themselves into different political positions on the basis of differences in need for closure, strengthening the relationship between need for closure and political preferences (Federico, Deason, & Fisher, 2012; Federico & Goren, 2009; see also Federico, Fisher, & Deason, 2011; Johnston, Lavine, & Federico, 2017; Malka, Soto, Inzlicht, & Lelkes, 2014). Johnston et al. (2017) argued that informed individuals are not merely more knowledgeable about what positions match their psychological characteristics—they are also more motivated to adopt political preferences that express aspects of the self. In other words, the moderating role of information observed in prior work may be partly due to political preferences being more identity central among the informed. In this study, we disentangled the two by examining the moderating role of identity centrality while also accounting for the moderating role of information.

Though we examined the role of identity centrality in the association between need for closure and political orientation, the moderation that we predicted might also imply a broader point: Construing politics as a means of self-expression might help to create more polarized political groups, who differ from one another not only in their expressed political identities and attitudes but also in their underlying psychological needs and motives. That said, our predictions also imply that these more fundamental differences are limited to people who see politics as an expression of who they are.

Hypotheses

We hypothesized that need for closure predicts polarized left-right orientations most strongly among people whose political preferences are central to their self-concept. In testing this, we considered several dependent variables. First, we looked at two broad political identifications: symbolic ideology and party identification. Symbolic ideology refers to self-identification with an ideological group label symbolizing a general philosophical posture—that is, whether someone thinks of himself or herself as “liberal” or “conservative” when asked (Ellis & Stimson, 2012). Party identification refers to identification with labels representing the competing organized political parties in a political community—the Democratic and Republican parties in the present American case. Second, we looked at a summary index of individuals’ issue attitudes: operational ideology. Operational ideology refers to whether a person adopts left-leaning or right-leaning attitudes across issues; it is the primary latent left/right dimension underlying issue attitudes (Ellis & Stimson, 2012). Though operational ideology is correlated with symbolic ideology and party identification, they are theoretically distinct. Symbolic ideology and party identification refer to social identifications in the political realm, whereas operational ideology reflects attitudes toward specific issues. Moreover, the three variables often do not align empirically, whether within individuals, within nations, or across history (Malka & Lelkes, 2010). We therefore considered the three variables separately.

We made several specific predictions. First, the relationship between need for closure and the two political identifications will be stronger among individuals higher in identity centrality. When their political preferences are more central to the self, individuals with high need for closure will be more likely to identify as conservatives and Republicans, whereas individuals with low need for closure will be more likely to identify as liberals and Democrats (Hypothesis 1). Second, the relationship between need for closure and operational ideology will be stronger among individuals higher in identity centrality. When their political preferences are more central to the self, individuals with high need for closure will be more likely to adopt conservative issue positions, whereas individuals with low need for closure will be more likely to adopt liberal issue positions (Hypothesis 2).

Finally, we argued that the stronger linkage between need for closure and operational ideology among individuals high in identity centrality will be accounted for by stronger linkages (a) between need for closure and the two political identifications and (b) between the two political identifications and operational ideology among those high in identity centrality (Hypothesis 3). That is, we predicted a pattern of mediated moderation. This hypothesis implies that identity centrality strengthens the relationship between need for closure and operational ideology by increasing the likelihood that individuals varying in need for closure will sort into different ideological and partisan identities and then adopt issue positions consistent with those identities. This follows directly from our predictions in Hypothesis 1 and from experimental (e.g., Cohen, 2003) and longitudinal (e.g., Goren, 2005) evidence that suggests that ideological and partisan identities have a downstream influence on issue positions.

Method

Data

Data came from a nationally representative survey of 1,511 American adults conducted by Knowledge Networks (now GfK) in fall 2008. This sample size was set and requested in advance of data collection to be comparable with sample sizes in major American election studies (American National Election Studies, n.d.). Potential panel respondents were chosen from a probability sample initially contacted via random-digit dialing (for evidence on the representativeness of the Knowledge Networks panel, see Chang & Krosnick, 2009). The actual survey relied on a probability sample of all panel members 18 or older. Among panel members randomly selected for the survey, 65.7% completed the survey. Taking into account the household recruitment rate for the web panel (20%) and the rate at which at least one individual in each household finished the profile survey (54.5%), the cumulative response rate for the survey was 7.2%. In each analysis, all respondents to the survey with complete data on the relevant variables were included. Note that the survey also included an experimental manipulation aimed at examining separate hypotheses; however, the manipulation did not include any of the variables used in this analysis and was administered in the survey after all of the variables analyzed in the present study were assessed.

Measures

We describe our measures in the sections that follow; further details on the survey items can be found in the Supplemental Material available online, and intercorrelations between key variables can be found in Table 1. Unless otherwise indicated, all measures were recoded to run from 0 to 1.

Table 1.

Intercorrelations Among Key Variables

Variable	1	2	3	4	5
1. Symbolic ideology	—
2. Party identification	.71***	—
3. Operational ideology	.71***	.68***	—
4. Information	.01	.05^†	.10***	—
5. Identity centrality	–.002	.03	.04	.37***	—
6. Need for closure	.16***	.14***	.17***	–.13***	–.04

Note: The table presents Pearson correlation coefficients. Higher scores on both ideology measures indicate greater conservatism.

†

p < .10. ***p < .001.

Need for closure

Need for closure was assessed using a revised 14-item version of the Need for Closure scale (Pierro & Kruglanski, 2006; see also Federico et al., 2012). Validation data indicate that this scale has good psychometric properties, with high reliability (α = .81) and a high disattenuated correlation with the full 42-item Need for Closure scale (r = .92). Participants responded to all items using a scale ranging from 1 (strongly disagree) to 6 (strongly agree); the text of the items can be found in the Supplemental Material. Higher scores indicate higher need for closure (α = .81; M = .44, SD = .14).

Symbolic ideology

Symbolic ideology—self-identification with an ideological label—was operationalized using a standard 7-point measure of ideological self-description (e.g., Zaller, 1992): 1 (liberal, strong), 2 (liberal, not very strong), 3 (moderate/neither, lean liberal), 4 (moderate/neither), 5 (moderate/neither, lean conservative), 6 (conservative, not very strong), and 7 (conservative, strong). Higher scores indicate greater conservatism (M = .57, SD = .34).

Party identification

Symbolic ideology is only one type of political identification individuals might adopt. Party identification is another and is certainly no less important (Sears, 1993; Zaller, 1992). Though symbolic ideology and party identification are correlated in present-day America, they are not the same thing in conceptual terms, and members of the same political party can in fact vary in their symbolic ideology (Huddy et al., 2015; Sears, 1993). For this reason, we examined party identification and symbolic ideology as distinguishable political identities. Note, however, that we anticipated a similar pattern of results for each variable because ultimately they are both examples of political identities. Party identification was operationalized using a 7-point measure ranging from 1 (strong Democrat) to 7 (strong Republican), which was recoded to run from 0 to 1. Higher scores indicate a greater Republican tilt (M = .50, SD = .36).

Operational ideology

A measure of operational ideology—the average left/right tilt of a respondent’s issue attitudes—was constructed from items measuring attitudes toward eight different policies covering the domains of economics, social welfare, defense, and social issues. Responses to branching items for each issue were used to create eight attitude measures, each on a 5-point scale. Responses to each item were recoded to run from 0 to 1 and reverse-scored when needed so that higher scores always indicated a more conservative position. Because operational ideology is conceptually and empirically defined in the literature as a general left/right factor underlying all issue preferences (Ellis & Stimson, 2012), we subjected these items to a principal-axis factor analysis to create a measure. A scree plot indicated that the eight items had a strongly unidimensional structure: The eigenvalues for the first four factors were 2.47, 0.50, 0.07, and 0.02, with a ratio of 4.95 between the first and second factors; the items also formed a reliable scale (α = .77). Thus, we computed factor scores for the dominant first factor, which were recoded to run from 0 to 1 (M = .50, SD = .22). These recoded scores served as our index of operational ideology; higher scores indicate greater policy conservatism.

Though the analysis strongly implied that all eight issues form a single dimension rather than separating into different issue domains, we performed one additional check to be sure that need for closure did not function differently for economic attitudes (as it has in some previous studies; see Malka et al., 2014). Specifically, we created two separate subscales corresponding to the two economic issues (i.e., regarding services and spending and jobs) and the six noneconomic issues. We then ran Model 2 (as specified in Tables 2, 3, and 4) separately using each of the two subscales as the dependent variable. The Need For Closure × Identity Centrality interaction was significant in the economic-issues model, b = 0.66, 95% confidence interval (CI) = [0.27, 1.05], p = .001, and the noneconomic-issues model, b = 0.40, 95% CI = [0.15, 0.66], p = .002. Thus, our key prediction about identity centrality as a moderator generalizes across issue domains, suggesting that the use of one unidimensional composite is not obscuring asymmetries on the basis of issue domain.

Note on the link among symbolic ideology, party identification, and operational ideology

The correlations among symbolic ideology, party identification, and operational ideology in our data (see Table 1) were similar to those found in other studies (e.g., Malka & Lelkes, 2010). Though these correlations are relatively strong, the constructs are conceptually distinct and are typically treated as different variables. They are also empirically distinct, despite their correlation. In particular, each construct has an independent effect on judgments about novel political stimuli (Malka & Lelkes, 2010), and there is considerable variation in operational ideology among individuals with the same symbolic ideology or party identification (Ellis & Stimson, 2012).

To confirm the independent significance of each construct, we regressed evaluations of Barack Obama and John McCain (M = .50, SD = .34, and M = .51, SD = .28, respectively, both recoded 0–1) on the three variables in multiple regressions. All three variables independently predicted evaluations of Obama (symbolic ideology: b = −0.10, 95% CI = [−0.16, −0.03], p = .001; party identification: b = −0.33, 95% CI = [−0.38, −0.27], p < .001; operational ideology: b = −0.57, 95% CI = [−0.65, −0.48], p < .001) and McCain (symbolic ideology: b = 0.13, 95% CI = [0.07, 0.19], p < .001; party identification: b = 0.33, 95% CI = [0.28, 0.38], p < .001; operational ideology: b = 0.15, 95% CI = [0.07, 0.23], p < .001). Moreover, in these regressions, the variance inflation factors for the three variables were all less than 3, below the most stringent cutoff for excessive correlation cited in the literature (i.e., 5; O’Brien, 2007). Given these results, we felt comfortable treating the three variables as distinct constructs.

Political information

We also controlled for the documented moderating role of political information to isolate the role of political-identity centrality. Information was measured using eight factual knowledge items (e.g., Delli Carpini & Keeter, 1996) to which the participants responded using a multiple-choice format. The questions asked about the positions held by various political figures, party control of the Senate and House, and government procedure (e.g., the length of a senator’s term). Responses were coded as correct (0) or incorrect (1) and averaged to form a scale (α = .65; M = .71, SD = .24).

Identity centrality

This was measured using two items adapted from the Identification subscale of Luhtanen and Crocker’s (1992) Collective Self-Esteem Scale: (a) “My political attitudes and beliefs are an important reflection of who I am” and (b) “In general, my political attitudes and beliefs are an important part of my self-image.” Participants responded to the items on a scale ranging from 1 (strongly agree) to 7 (strongly disagree). Responses to the two items were averaged and reverse-coded to run from 0 (low identity centrality) to 1 (high identity centrality; α = .84; M = .59, SD = .26).

Demographics

Several standard demographic controls that are commonly included in models predicting political attitudes in public opinion research (e.g., Federico et al., 2012; Malka et al., 2014) were also considered. This allowed us to gauge the interactive effects of need for closure and identity centrality independent of demographic correlates of political attitudes and account for the fact that variables related to political engagement (such as information and identity centrality) often vary with demographic characteristics (Delli Carpini & Keeter, 1996). These were age (in years), male gender (0 = no, 1 = yes), two dummy variables indicating whether the respondent was Black or Latino, education (six ordered categories: no high school degree, high school degree, some college, associate’s degree, bachelor’s degree, graduate degree), and income (in thousands of dollars per year). Note that we obtained identical results when the demographics were excluded. These analyses are summarized in the Supplemental Material.

Results

Symbolic ideology and party identification as a function of need for closure and identity centrality

Hypothesis 1 predicted that need for closure would be more strongly related to symbolic ideology and party identification among individuals whose political attitudes are more central to the self. We tested this hypothesis with respect to the two political identifications using two regression models. In Model 1, we regressed each dependent variable on demographics, information, identity centrality, and need for closure. In Model 2, we added the key Need For Closure × Identity Centrality interaction; we also included the Need For Closure × Information interaction in Model 2 to distinguish any moderating effect of identity centrality from that of information. Robust standard errors were used in all analyses (Long & Ervin, 2000). Given the 0 to 1 coding of all variables, each regression coefficient can be interpreted as the proportion change (or percentage change when multiplied by 100) in the dependent variable associated with moving from the lowest (0) to the highest (1) value of a continuous predictor or from one group to the other for a dichotomous predictor (Baguley, 2009). Further details on the specification of all statistical models can be found in the Supplemental Material.

The results for symbolic ideology are summarized in Table 2. Looking first at Model 1, we found that individuals low in need for closure identified as more liberal, and those high in need for closure identified as more conservative (b = 0.34, p < .001). In Model 2, we found the predicted Need For Closure × Identity Centrality interaction (b = 0.82, p < .001); the Need For Closure × Information interaction found in prior work (e.g., Federico & Goren, 2009) was not significant (p > .250) once the other interaction was accounted for. To unpack the Need For Closure × Identity Centrality interaction, we estimated simple slopes for need for closure across the full range of identity centrality values from 1 standard deviation below (low identity centrality) to 1 standard deviation above (high identity centrality) the mean. These estimates are plotted in the left panel of Figure 1; values on the y-axis represent the simple-slopes estimates. At low identity centrality, the simple slope for need for closure was weak and nonsignificant, b = 0.10, 95% CI = [−0.07, 0.26], p > .250. At high centrality, the simple slope was significant and more than five times stronger, b = 0.52, 95% CI = [0.35, 0.69], p < .001. Given our variable codings, these simple-slopes estimates indicate the proportion change in symbolic ideology associated with going from the lowest to the highest value of need for closure at a particular level of identity centrality. Thus, going from the minimum to the maximum need for closure was associated with only a 10% shift in the conservative direction at low centrality but a 52% change in the same direction at high centrality. To provide a more conservative effect-size estimate, we also computed the change in symbolic ideology associated with going from 1 standard deviation below to 1 standard deviation above the mean of need for closure. A shift of this magnitude in need for closure was associated with a 3% change in the conservative direction at low centrality but a 15% change at high centrality.

Table 2.

Symbolic Ideology as a Function of Need for Closure (NFC) and Identity Centrality

Predictor	Model 1			Model 2
Predictor	b	95% CI	p	b	95% CI	p
Age	0.14	[0.06, 0.22]	.001	0.14	[0.06, 0.23]	< .001
Male gender	0.06	[0.02, 0.09]	.001	0.06	[0.02, 0.09]	.001
Black	–0.17	[−0.24, −0.11]	< .001	–0.18	[−0.24, −0.12]	< .001
Latino	–0.14	[−0.20, –0.08]	< .001	–0.13	[−0.19, −0.07]	< .001
Education	–0.14	[−0.20, −0.0]	< .001	–0.13	[−0.19, −0.08]	< .001
Income	0.05	[−0.04, 0.13]	> .250	0.05	[−0.03, 0.14]	.196
Information	0.02	[−0.06, 0.09]	> .250	–0.10	[−0.33, 0.13]	> .250
Identity centrality	–0.01	[−0.08, 0.06]	> .250	–0.36	[−0.57, −0.15]	.001
NFC	0.34	[0.21, 0.46]	< .001	–0.36	[−0.73, 0.01]	.056
NFC × Information				0.26	[−0.22, 0.75]	> .250
NFC × Identity Centrality				0.82	[0.37, 1.27]	< .001
Intercept	0.39	[0.29, 0.48]	< .001	0.68	[0.51, 0.86]	< .001

Note: Model 1: F(9, 1501) = 16.62, p < .001, R² = .081; Model 2: F(11, 1499) = 14.53, p < .001, R² = .092. The table presents ordinary least squares regression coefficients and 95% confidence intervals (CIs) based on robust standard errors. Higher scores on the symbolic-ideology measure indicate greater conservatism (N = 1,511 in both models). Bolded coefficients are significant (p < .05).

Fig. 1.

Simple slopes for need for closure at levels of identity centrality ranging from 1 standard deviation below to 1 standard deviation above the mean. Results are shown separately for symbolic ideology (left panel) and party identification (right panel). Values on the y-axes represent simple-slopes estimates; the error band indicates the 95% confidence intervals around these estimates. Higher scores indicate greater conservatism and Republican identification.

The results for party identification are summarized in Table 3. Looking first at Model 1, we found a positive relationship between need for closure and identification with the Republican Party, b = 0.32, 95% CI = [0.19, 0.45], p < .001. In Model 2, we found the expected Need For Closure × Identity Centrality interaction, b = 0.72, 95% CI = [0.25, 1.20], p = .003; the Need For Closure × Information interaction was marginally significant, b = 0.49, 95% CI = [−0.03, 1.01], p = .062. To unpack the Need For Closure × Identity Centrality interaction, we estimated simple slopes for need for closure across the full range of identity-centrality values from 1 standard deviation below (low identity centrality) to 1 standard deviation above (high identity centrality) the mean. These estimates are plotted in the right panel of Figure 1. At low identity centrality, the simple slope for need for closure was weak and nonsignificant, b = 0.10, 95% CI = [−0.07, 0.28], p > .250. In contrast, at high centrality, the simple slope was significant and more than four times stronger, b = 0.48, 95% CI = [0.30, 0.65], p < .001. These estimates indicated that going from the minimum to the maximum need for closure was associated with only a 10% shift in the Republican direction at low centrality but a 48% change in the same direction at high centrality. As with symbolic ideology, we also computed the change in party identification associated with going from 1 standard deviation below to 1 standard deviation above the mean of need for closure. A shift of this magnitude in need for closure was associated with a 3% change in the Republican direction at low centrality but a 13% change in this direction at high centrality.

Table 3.

Party Identification as a Function of Need for Closure (NFC) and Identity Centrality

Predictor	Model 1			Model 2
Predictor	b	95% CI	p	b	95% CI	p
Age	–0.01	[−0.10, 0.07]	> .250	–0.01	[−0.10, 0.07]	> .250
Male gender	0.06	[0.02, 0.10]	.001	0.06	[0.03, 0.10]	.001
Black	–0.32	[−0.38, −0.26]	< .001	–0.33	[−0.38, −0.27]	< .001
Latino	–0.19	[−0.25, −0.13]	< .001	–0.18	[−0.24, −0.12]	< .001
Education	–0.11	[−0.17, −0.04]	.001	–0.10	[−0.17, −0.04]	.001
Income	0.13	[0.04, 0.22]	.003	0.14	[0.05, 0.23]	.002
Information	0.05	[−0.04, 0.13]	> .250	–0.18	[−0.43, 0.07]	.165
Identity centrality	0.03	[−0.05, 0.10]	> .250	–0.28	[−0.50, −0.06]	.012
NFC	0.32	[0.19, 0.45]	< .001	–0.48	[−0.88, −0.09]	.017
NFC × Information				0.49	[−0.03, 1.01]	.062
NFC × Identity Centrality				0.72	[0.25, 1.20]	.003
Intercept	0.29	[0.18, 0.39]	< .001	0.63	[0.44, 0.82]	< .001

Note: Model 1: F(9, 1477) = 24.93, p < .001, R² = .094; Model 2: F(11, 1475) = 21.83, p < .001, R² = .104. The table presents ordinary least squares regression coefficients and 95% confidence intervals (CIs) based on robust standard errors. Higher scores on the party-identification measure indicate a greater affinity for the Republican Party (N = 1,487 in both models). Bolded coefficients are significant (p < .05).

Operational ideology as a function of need for closure and identity centrality

In turn, Hypothesis 2 predicted that need for closure would be more strongly related to operational ideology among individuals whose political attitudes were more central to the self. We tested this hypothesis with models similar to those used to examine Hypothesis 1. Results are summarized in Table 4. In Model 1, we found that individuals low in need for closure adopt more liberal issue positions on average, and those high in need for closure adopt more conservative positions on average (b = 0.24, p < .001). In Model 2, the Need For Closure × Identity Centrality interaction was significant as predicted (b = 0.52, p < .001). The Need For Closure × Information interaction was also significant and positive (b = 0.32, p = .040), indicating that need for closure more strongly predicts policy conservatism among the more informed. We broke down the significant Need For Closure × Identity Centrality interaction as we did in the symbolic-ideology model. These estimates are plotted in Figure 2. At low identity centrality, the simple slope for need for closure was again weak, b = 0.08, 95% CI = [−0.02, 0.18], p = .115. However, at high centrality, the simple slope was significant and more than four times stronger, b = 0.35, 95% CI = [0.24, 0.46], p < .001. These estimates indicated that going from the minimum to maximum need for closure was associated with an 8% shift in the direction of conservative operational ideology at low centrality but a 35% move in the same direction at high centrality. Again, for a more conservative effect-size estimate, we computed the change in operational ideology associated with going from 1 standard deviation below to 1 standard deviation above the mean of need for closure. A shift of this magnitude in need for closure was associated with a 2% change in the conservative direction at low centrality but a 10% change at high centrality.

Table 4.

Operational Ideology as a Function of Need for Closure (NFC) and Identity Centrality

Predictor	Model 1			Model 2
Predictor	b	95% CI	p	b	95% CI	p
Age	0.05	[0.00, 0.10]	.067	0.05	[0.00, 0.10]	.061
Male gender	0.05	[0.03, 0.08]	< .001	0.06	[0.03, 0.08]	< .001
Black	–0.15	[−0.18, −0.11]	< .001	–0.15	[−0.19, −0.12]	< .001
Latino	–0.09	[−0.13, −0.04]	< .001	–0.08	[−0.12, −0.04]	< .001
Education	–0.12	[−0.15, −0.08]	< .001	–0.11	[−0.15, −0.08]	< .001
Income	0.04	[−0.01, 0.09]	.142	0.05	[−0.01, 0.10]	.088
Information	0.10	[0.05, 0.14]	< .001	–0.05	[−0.20, 0.10]	> .250
Identity centrality	0.01	[−0.03, 0.05]	> .250	–0.21	[−0.35, −0.07]	.003
NFC	0.24	[0.16, 0.31]	< .001	–0.32	[−0.55, −0.08]	.007
NFC × Information				0.32	[0.02, 0.63]	.040
NFC × Identity Centrality				0.52	[0.22, 0.81]	.001
Intercept	0.32	[0.26, 0.38]	< .001	0.56	[0.44, 0.67]	< .001

Note: Model 1: F(9, 1501) = 29.62, p < .001, R² = .120; Model 2: F(11, 1499) = 25.38, p < .001, R² = .133. The table presents ordinary least squares regression coefficients and 95% confidence intervals (CIs) based on robust standard errors. Higher scores on the operational-ideology measure indicate greater conservatism (N = 1,511 in both models). Bolded coefficients are significant (p < .05).

Fig. 2.

Simple slopes for need for closure at levels of identity centrality ranging from 1 standard deviation below to 1 standard deviation above the mean. Results are shown for operational ideology. Values on the y-axis represent simple-slope estimates; the error band indicates the 95% confidence intervals around these estimates. Higher scores indicate greater operational conservatism.

Mediated moderation analysis

Hypothesis 3 predicted that the stronger relationship between need for closure and operational ideology among individuals high in identity centrality would be accounted for by stronger relationships between need for closure and the two political identifications (symbolic ideology and party identification) and stronger relationships between the two political identifications and operational ideology. The pattern of associations we predicted corresponds to what Muller, Judd, and Yzerbyt (2005) referred to as mediated moderation. Because symbolic ideology and party identification are related, we examined the mediating role of both symbolic ideology and party identification simultaneously to better distinguish their respective associations with operational ideology.

We tested this prediction by performing a median split on identity centrality (in the final model: low centrality n = 737, high centrality n = 750) and then estimating a multigroup structural equation model. We specified both need for closure and operational ideology as latent variables. For need for closure, three item parcels were constructed to serve as indicators; construction of the parcels is described in the Supplemental Material. The loading for the first parcel was fixed to 1 to set the metric of the factor. For operational ideology, the eight individual items were used as indicators; the loading for the services-and-spending item was fixed to 1. Given that we had only single indicators for party identification and symbolic ideology, we were unable to generate latent variables for these constructs. Thus, they were entered as observed variables. In each group, the structural model was as follows: (a) Need for closure was allowed to have direct effects on party identification and symbolic ideology; (b) need for closure, party identification, and symbolic ideology were allowed to have direct effects on operational ideology; and (c) the disturbances for the two political identifications were allowed to covary. In the structural equations predicting party identification, symbolic ideology, and operational ideology, the demographics were included as covariates. Finally, the exogenous variables—need for closure and the demographics—were also allowed to covary. Importantly, this model assumed that symbolic ideology and party identification were causally prior to operational ideology. This is consistent with both experimental research and longitudinal analyses indicating that political identities tend to shape issue positions more strongly and more often than issue positions shape identity (Cohen, 2003; Goren, 2005; Lenz, 2012; Malka & Lelkes, 2010), and that these identities also tend to be more stable than issue positions over time (Converse & Markus, 1979; Goren, 2005).

The model was estimated in Mplus (Version 8; Muthén & Muthén, 2017). Maximum likelihood estimation was used, and the variances of all direct and indirect effects were estimated using a bootstrapping procedure (with 1,000 replications). As a preliminary step, we tested whether the factor loadings for need for closure and operational ideology were invariant across groups. Compared with a model in which all parameters were allowed to vary across groups, a model in which the loadings for need for closure were constrained to equality across groups did not worsen the fit of the model, Δχ²(2) = 2.17, p > .250. Similarly, constraining the operational-ideology loadings to equality across groups did not worsen model fit, Δχ²(9) = 9.12, p = .240. Therefore, in the primary model, all latent-variable loadings were constrained to equality. This model provided an acceptable fit to the data, χ²(237) = 835.31, p < .001, comparative fit index = .92, root-mean-square error of approximation = .058, though the chi-square value was significant because of the large sample size. On the basis of modification-index results, we also allowed a correlation between the error variances for two of the issue items dealing with social welfare matters: job guarantees and aid to African Americans. The estimates from this model are summarized in Figure 3. For visual clarity, factor loadings are not shown. However, all free loadings for need for closure were in excess of 0.70 (with a mean of 0.94), whereas all free loadings for operational ideology (except for one, at 0.32) were in excess of 0.70 (with a mean of 0.82). In additional, all free loadings for both latent variables were significant (p < .001). This model served as the baseline model for the cross-group tests we report in the following sections. In interpreting the estimates, recall that all observed variables have been recoded to run from 0 to 1, including the indicators for need for closure and operational ideology whose loadings were set to unity.

Fig. 3.

Multigroup structural equation model from the mediated moderation analysis. Unstandardized estimates are presented; standardized estimates for the covariance between the party-identification and symbolic-ideology disturbances are shown in parentheses. Factor loadings are not shown. Significance tests were based on bootstrapped standard errors. Symbols indicate marginally significant (^†p < .10) and significant (***p < .001) paths.

These estimates revealed several patterns consistent with the idea that identity centrality promotes greater alignment between need for closure and political preferences and between the various political preferences themselves. First, confirming Hypothesis 1, we found that the coefficients for the regression of the two political identifications on need for closure were both larger in the high-centrality group. Indeed, constraining these two coefficients to equality across groups significantly worsened the fit of the model compared with the baseline, Δχ²(2) = 10.06, p = .007. Moreover, the coefficients for the relationships between the two political identifications and operational ideology were also larger in the high-centrality group, consistent with the notion that identity centrality also strengthens the relationship between political identifications and issue preferences. Constraining these two coefficients to equality across groups also significantly worsened the fit of the model, Δχ²(2) = 32.66, p < .001. Finally, as Figure 1 indicates, the residual correlation between the disturbances for party identification and symbolic ideology was stronger in the high-centrality group, Δχ²(1) = 55.83, p < .001, suggesting the two political identifications overlap more as centrality increases.

Importantly, the direct relationship between need for closure and operational ideology was negligible at both centrality levels (β = 0.08, p < .250, in the low-centrality group; β = 0.12, p < .076, in the high-centrality group). Moreover, constraining the direct path from need for closure to operational ideology to equality across the two groups did not produce a significant decrement in model fit, Δχ²(1) = 0.15, p > .250, suggesting little variation in this relationship as a function of centrality once the mediating role of party identification and symbolic ideology was accounted for.

To account for the mediating roles of the two political identifications, we estimated conditional indirect effects of need for closure on operational ideology via both party identification and symbolic ideology in each group. We also report the total conditional indirect effects via both identifications in each group. These estimates are summarized in Table 5, with bootstrapped 95% bias-corrected CIs. As the estimates indicated, the magnitude of the conditional indirect effects of need for closure via the two political identifications increased as a function of identity centrality. In the low-identity-centrality group, the indirect effects of need for closure via both mediators were low in magnitude and failed to reach significance, indirect effect = 0.059, 95% CI = [−0.009, 0.139], via symbolic ideology; indirect effect = 0.028, 95% CI = [−0.023, 0.085], via party identification. Moreover, the total conditional indirect effect via both mediators failed to reach significance in the low-centrality group, b = 0.087, 95% CI = [−0.022, 0.201]. However, in the high-centrality group, the indirect effects of need for closure via both mediators were significant and notably stronger, indirect effect = 0.280, 95% CI = [0.166, 0.412], via symbolic ideology; indirect effect = 0.187, 95% CI = [0.115, 0.284], for party identification. In addition, the total conditional indirect effect via both mediators was significant and more than five times larger than it was at the lowest identity-centrality level, indirect effect = 0.467, 95% CI = [0.292, 0.654]. Finally, the differences in the two indirect effects and the total indirect effects across groups were all significant, as none of their bias-corrected bootstrapped CIs included 0. Thus, our results were consistent with Hypothesis 3.

Table 5.

Results From the Mediated Moderation Analysis: Conditional Indirect Effects of Need for Closure on Operational Ideology via Symbolic Ideology and Party Identification

Identity centrality	Via symbolic ideology		Via party identification		Total indirect effect
Identity centrality	Indirect effect	95% CI	Indirect effect	95% CI	Indirect effect	95% CI
Low	0.059	[−0.009, 0.139]	0.028	[−0.023, 0.085]	0.087	[−0.022, 0.201]
High	0.280	[0.166, 0.412]	0.187	[0.115, 0.284]	0.467	[0.292, 0.654]
Difference	0.221	[0.087, 0.358]	0.159	[0.070, 0.264]	0.380	[0.174, 0.597]

Note: The table presents unstandardized coefficients. Confidence intervals (CIs) are based on 1,000 bootstrapped samples. Bolded coefficients are significant (p < .05).

Additional mediated moderation analyses

One shortcoming of the multigroup test of mediated moderation presented above is that it did not allow us to assess the simultaneous moderating effects of identity centrality and political information. Though we focus on the structural equation analyses here, we also present an analogous mediated moderation model based on Hayes’s (2013) conditional process analysis approach in the Supplemental Material. This model considers both multiple mediators (party identification and symbolic ideology) and multiple moderators (identity centrality and political information). This analysis produced conclusions identical to those produced by the multigroup structural equation analysis. Finally, for completeness, we also present reversed versions of the conditional process model in the Supplemental Material that treat symbolic ideology and party identification as the final outcomes and operational ideology as the mediator. These models suggest some evidence for the reversed pattern as well. This should be taken into consideration in interpreting our mediated moderation results, though we caution that reversing the arrows in mediation models does not allow for valid model comparisons using cross-sectional data (Thoemmes, 2015) and that the preponderance of public opinion research points toward symbolic ideology and party identification as causes of issue attitudes as opposed to vice versa (as noted earlier).

Discussion

In sum, we found support for our three hypotheses. Of course, our study has some limitations. First, our data were correlational. This leaves open the possibility of a causal ordering that is the reverse of the one implied by Hypothesis 3. Indeed, additional analyses presented in the Supplemental Material suggest that this alternative hypothesis cannot entirely be ruled out. However, existing evidence indicates that symbolic ideology and party identification influence concrete issue positions more consistently than vice versa (Goren, 2005; Lenz, 2012), suggesting that our model is the more appropriate one. Nevertheless, longitudinal data are needed to more adequately test Hypothesis 3. Second, our data came from a single sample in a single country at a single point in time. Further work will be necessary to test the robustness of our findings across time periods and nations in which the personality-politics link differs from that in contemporary America.

Despite these limitations, our results have important implications. Above all, they complicate the notion of a simple, universal relationship between personality and political preferences. With a few exceptions, research on the psychology of political preferences has focused overwhelmingly on main-effect relationships between psychological variables and political attitudes (Jost et al., 2003). This work has usefully reasserted the significance of psychological variables for understanding political preferences, but it has devoted far less attention to processes that forge the link between psychology and politics or account for why that link might not be forged at all in some individuals. Our work helps fill this gap. While our findings confirm that personality matters for political preferences, they also suggest that it does not matter equally for all. Rather, political preferences appear to be more related to deep-seated personal characteristics among individuals who regard their political identifications and attitudes as central to the self and thus expressive of identity. When politics reflect the self, they also reflect personality to a greater extent—a pattern that may extend to other traits (such as openness to experience; see Gerber, Huber, Doherty, & Dowling, 2011).

On the one hand, then, our findings suggest that individuals with opposed political preferences are not always fundamentally different from one another. On the other hand, our findings suggest that there are subsets of individuals with liberal and conservative preferences whose psychological predispositions are intensely different and that these are the people who are most passionately invested in their political allegiances and most likely to participate in politics. It is these individuals who contribute most significantly to the polarization of the mass public in contemporary politics. Thus, some of the most divisive debates in the American polity may amount to duels between highly politicized subpopulations whose attitude dynamics bear only a limited resemblance to those in the public at large. Researchers interested in understanding the deeper psychological bases of political division would therefore do well to consider how central political attitudes are to citizens’ identities.

Supplemental Material

FedericoOpenPracticesDisclosure – Supplemental material for The Political Self: How Identity Aligns Preferences With Epistemic Needs

Supplemental material, FedericoOpenPracticesDisclosure for The Political Self: How Identity Aligns Preferences With Epistemic Needs by Christopher M. Federico and Pierce D. Ekstrom in Psychological Science

Supplemental Material

FedericoSupplementalMaterial – Supplemental material for The Political Self: How Identity Aligns Preferences With Epistemic Needs

Supplemental material, FedericoSupplementalMaterial for The Political Self: How Identity Aligns Preferences With Epistemic Needs by Christopher M. Federico and Pierce D. Ekstrom in Psychological Science

Footnotes

Action Editor

Leaf Van Boven served as action editor for this article.

Author Contributions

C. M. Federico developed the study concept. Both authors contributed to the study design and analysis plan. C. M. Federico was responsible for the design of the original survey and contracting for the data collection. C. M. Federico analyzed and interpreted the data, with contributions from P. D. Ekstrom. Both authors were responsible for the drafting of the manuscript and subsequent revisions. Both authors approved the final manuscript for submission.

Declaration of Conflicting Interests

The author(s) declared that there were no conflicts of interest with respect to the authorship or the publication of this article.

Supplemental Material

Additional supporting information can be found at

Open Practices

All data have been made publicly available via Harvard Dataverse and can be accessed at https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/OE9BNP. Materials have not been made publicly available, and the design and analysis plans for the study were not preregistered. The complete Open Practices Disclosure for this article can be found at https://journals-sagepub-com.web.bisu.edu.cn/doi/suppl/10.1177/0956797617748679. This article has received the badge for Open Data. More information about the Open Practices badges can be found at .

References

Abramowitz

A. I.

(2010). The disappearing center: Engaged citizens, polarization, and American democracy. New Haven, CT: Yale University Press.

American National Election Studies. (n.d.). The ANES guide to public opinion and electoral behavior. Retrieved from http://www.electionstudies.org/nesguide/nesguide.htm

Baguley

(2009). Standardized or simple effect size: What should be reported? British Journal of Psychology, 100, 603–617. doi:10.1348/000712608X377117

Campbell

Converse

P. E.

Miller

W. E.

Stokes

D. E.

(1960). The American voter. New York, NY: Wiley.

Chang

Krosnick

J. A.

(2009). National surveys via RDD telephone interviewing versus the Internet: Comparing sample representativeness and response quality. Public Opinion Quarterly, 73, 641–678.

Cohen

G. L.

(2003). Party over policy: The dominating impact of group influence on political beliefs. Journal of Personality and Social Psychology, 85, 808–822.

Converse

P. E.

Markus

G. B.

(1979). Plus ça change . . .: The new CPS election study panel. American Political Science Review, 73, 32–49.

Delli Carpini

M. X.

Keeter

. (1996). What Americans know about politics and why it matters. New Haven, CT: Yale University Press.

Devine

C. J.

(2015). Ideological social identity: Psychological attachment to ideological in-groups as a political phenomenon and a behavioral influence. Political Behavior, 37, 509–535.

10.

Ellis

Stimson

J. A.

(2012). Ideology in America. New York, NY: Cambridge University Press.

11.

Federico

C. M.

Deason

Fisher

E. L.

(2012). Ideological asymmetry in the relationship between epistemic motivation and political attitudes. Journal of Personality and Social Psychology, 103, 381–398. doi:10.1037/a0029063

12.

Federico

C. M.

Fisher

E. L.

Deason

(2011). Expertise and the ideological consequences of the authoritarian predisposition. Public Opinion Quarterly, 75, 686–708.

13.

Federico

C. M.

Goren

(2009). Motivated social cognition and ideology: Is attention to elite discourse a prerequisite for epistemically motivated political affinities? In Jost

J. T.

Kay

A. C.

Thorisdottir

(Eds.), Social and psychological bases of ideology and system justification (pp. 267–291). New York, NY: Oxford University Press.

14.

Gerber

A. S.

Huber

G. A.

Doherty

Dowling

C. M.

(2011). The Big Five personality traits in the political arena. Annual Review of Political Science, 14, 265–287.

15.

Goren

(2005, April). Ideological identification and core political values. Paper presented at the annual meeting of Midwest Political Science Association, Chicago, IL.

16.

Hayes

A. F.

(2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York, NY: Guilford.

17.

Huddy

Mason

Aarøe

(2015). Expressive partisanship: Campaign involvement, political emotion, and partisan identity. American Political Science Review, 109, 1–17.

18.

Iyengar

Sood

Lelkes

(2012). Affect, not ideology: A social identity perspective on polarization. Public Opinion Quarterly, 76, 405–431.

19.

Johnston

C. D.

Lavine

H. G.

Federico

C. M.

(2017). Open versus closed: Personality, identity, and the politics of redistribution. Cambridge, England: Cambridge University Press.

20.

Jost

J. T.

Federico

C. M.

Napier

J. L.

(2009). Political ideology: Its structure, functions, and elective affinities. Annual Review of Psychology, 60, 307–333.

21.

Jost

J. T.

Federico

C. M.

Napier

J. L.

(2013). Political ideologies and their social psychological functions. In Freeden

Sargent

L. T.

Stears

(Eds.), The Oxford handbook of political ideologies (pp. 232–250). New York, NY: Oxford University Press.

22.

Jost

J. T.

Glaser

Kruglanski

A. W.

Sulloway

(2003). Political conservatism as motivated social cognition. Psychological Bulletin, 129, 339–375.

23.

Kahan

D. M.

(2015). Climate-science communication and the measurement problem. Advances in Political Psychology, 36, 1–43.

24.

Katz

(1960). The functional approach to the study of attitudes. Public Opinion Quarterly, 24, 163–204.

25.

Lenz

G. S.

(2012). Follow the leader? How voters respond to politicians’ policies and performance. Chicago, IL: University of Chicago Press.

26.

Long

J. S.

Ervin

L. H.

(2000). Using heteroscedasticity consistent standard errors in the linear regression model. American Statistician, 54, 217–234.

27.

Luhtanen

Crocker

(1992). A Collective Self-Esteem Scale: Self-evaluation of one’s social identity. Personality and Social Psychology Bulletin, 18, 302–318.

28.

Lupia

(2015). Uninformed: Why people seem to know so little about politics and what we can do about it. New York, NY: Oxford University Press.

29.

Malka

Lelkes

(2010). More than ideology: Conservative–liberal identity and receptivity to political cues. Social Justice Research, 23, 156–188.

30.

Malka

Soto

C. J.

Inzlicht

Lelkes

(2014). Do needs for security and certainty predict cultural and economic conservatism? A cross-national analysis. Journal of Personality and Social Psychology, 106, 1031–1051.

31.

Muller

Judd

C. M.

Yzerbyt

V. Y.

(2005). When moderation is mediated and mediation is moderated. Journal of Personality and Social Psychology, 89, 852–863.

32.

Muthén

L. K.

Muthén

B. O.

(2017). Mplus user’s guide (8th ed.). Los Angeles, CA: Author.

33.

O’Brien

R. M.

(2007). A caution regarding rules of thumb for variance inflation factors. Quality & Quantity, 41, 673–690.

34.

Pierro

Kruglanski

A. W.

(2006). [Validation of a revised Need for Cognitive Closure scale]. Unpublished data.

35.

Sears

D. O.

(1993). Symbolic politics: A socio-psychological theory. In Iyengar

McGuire

W. J.

(Eds.), Explorations in political psychology (pp. 113–149). Durham, NC: Duke University Press.

36.

Shavitt

(1989). Operationalizing functional theories of attitude. Attitude Structure and Function, 3, 311–337.

37.

Smith

M. B.

Bruner

J. S.

White

R. W.

(1956). Opinions and personality. Oxford, England: Wiley.

38.

Thoemmes

(2015). Reversing arrows in mediation models does not distinguish plausible models. Basic and Applied Social Psychology, 37, 226–234.

39.

Webster

D. M.

Kruglanski

A. W.

(1994). Individual differences in need for cognitive closure. Journal of Personality and Social Psychology, 67, 1049–1062.

40.

Zaller

(1992). The nature and origins of mass opinion. New York, NY: Cambridge University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.77 MB

0.33 MB