SEARCH

SEARCH BY CITATION

Keywords:

  • ideology;
  • need to evaluate;
  • political expertise;
  • attitude structure

Abstract

  1. Top of page
  2. Abstract
  3. Overview of the Analyses
  4. Results
  5. Discussion
  6. ACKNOWLEDGMENT
  7. REFERENCES

Political psychologists have typically argued that ideological commitments are structured in a bipolar fashion, where a positive evaluation of conservative objects implies a negative evaluation of liberal objects (and vice versa). Individual differences in conformity to this pattern are usually attributed to an ability-related variable, i.e., political expertise. Departing from this strict focus on ability, this study examines the hypotheses that an important motivational variable—the need to evaluate, or the desire to form opinions of objects as “good” or “bad”—would (1) predict deviations from ideological bipolarity, even controlling for expertise; and (2) moderate the relationship between expertise and deviations from bipolarity. Data from two national surveys provided evidence for these hypotheses and indicated that the results extended to deviations from bipolarity in evaluations of presidential candidates and political parties.

Political psychologists have engaged in a number of critical debates about the underlying structure of citizens’ ideological commitments. The most prominent of these debates have focused on whether members of the mass public use the higher-order distinction between “conservatism” and “liberalism” to structure their opinions (Campbell, Converse, Miller, & Stokes, 1960; Converse, 1964; Judd & Krosnick, 1989; Kinder & Sears, 1985; see also Achen, 1975; Judd, Krosnick, & Milburn, 1981; Zaller, 1992) and whether ideology consists of a single left-right dimension or multiple but interrelated left-right dimensions corresponding to different policy agendas (Asher, 1980; Knoke, 1979; Layman & Carsey, 2002; Luttbeg, 1968; Shafer & Claggett, 1995; Weisberg & Rusk, 1970; see also Duckitt, 2001; Schwartz, 1992). While these two debates have generated highly important work in the literature on political attitudes, they have often obscured what is perhaps an even more fundamental question about the dimensional structure of ideology. This is the question of whether ideology has a bipolar structure, where conservatism and liberalism are opposed ends of a single evaluative dimension; or a bivariate structure, where conservatism and liberalism are separate evaluative dimensions (Conover & Feldman, 1981; Green, 1988; Kerlinger, 1984).

Most work on the structure of ideology has assumed that ideological commitments are organized in a bipolar fashion: to the extent that an individual is attracted to conservatism and evaluates it more positively, he or she should be less attracted to liberalism and evaluate it less positively (and vice versa; see Campbell et al., 1960; Conover & Feldman, 1981; Converse, 1964; Green, 1988; Jost, Glaser, Kruglanski, & Sulloway, 2003; Kinder & Sears, 1985; Wilson, 1973). This assumption is evident in the construction of many policy items in the National Election Studies and other survey series, which present respondents with a bipolar response scale anchored by a conservative option at one end and a liberal option at the other. Moreover, the bipolarity assumption is taken for granted by almost all participants in the debates alluded to previously (Conover & Feldman, 1981). Regardless of the extent to which different researchers believe that citizens actually rely on ideological categories, most of them assume that ideological thinking will manifest itself in a bipolar form. Similarly, regardless of how many policy dimensions different researchers believe ideology consists of, they tend to see each of these dimensions as being a bipolar evaluative continuum with conservatism at one end and liberalism at the other.

Nevertheless, several lines of inquiry have suggested that ideological commitments may be organized in a bivariate fashion. These perspectives suggest that conservatism and liberalism are relatively independent categories of political evaluation, rather than opposite ends of a single bipolar spectrum. At the simplest level, researchers have noted that much of the mass public is unable to grasp the difference between “conservatism” and “liberalism” or accurately distinguish between positions which represent these two ideological categories (Campbell et al., 1960; Levitin & Miller, 1979). While this may indicate an absence of knowledge about conservatism and liberalism (Delli Carpini & Keeter, 1996), it may also indicate that many citizens do not see these two categories as being invariably opposed in the evaluative sense (as questions about differences between them implicitly assume; see Conover & Feldman, 1981). Moreover, other studies have suggested that there are individual differences in how conservatism and liberalism are distinguished from one another even among citizens who do make such distinctions, with some distinguishing between the two in terms of attitudes toward the concentration of power and others distinguishing between them in terms of attitudes toward the preservation of the status quo (Brown & Taylor, 1973; Conover & Feldman, 1981; Lane, 1962; Marcus, Tabb, & Sullivan, 1974).

Extending this argument, other researchers have directly suggested that conservatism and liberalism refer to different attitude systems with different reference points. In particular, Kerlinger (1967, 1972, 1980, 1984) has argued that conservatism and liberalism have different sets of “criterial referents,” i.e., goals, values, and concerns that are central to what it means to have a conservative or liberal outlook. According to this account, the referents that are “criterial” for conservatism are largely irrelevant to liberalism, and vice versa. This implies a noticeable deviation from evaluative bipolarity: one's level of conservatism should have little or nothing to do with one's level of liberalism, since the bases of each attitude system are effectively independent of one another. Consistent with this argument, exploratory and confirmatory factor analyses have shown (1) that evaluations of “conservative” and “liberal” referents load onto different factors and (2) that the factors corresponding to these two sets of referents are relatively orthogonal to one another, rather than being negatively related (as the bipolar model would assume; see Kerlinger, 1972, 1980; see also Conover & Feldman, 1984). Similarly, other analyses have shown that feeling-thermometer ratings of “conservatives” and “liberals” (Conover & Feldman, 1981) and Republicans and Democrats (Weisberg, 1980) have only a weak negative correlation with another.

However, factor-analytic studies using different model specifications have qualified these results (Green, 1988; Sidanius & Duffy, 1988). For example, after correcting for nonrandom error introduced by the use of a common response format, Green (1988) found that latent variables corresponding to evaluations of liberals and conservatives were in fact highly negatively correlated, suggesting a bipolar pattern. These studies—along with a growing literature on the polarization of distinctions between conservatives and liberals (e.g., Hetherington, 2001; see also Fiorina, 2005; Poole & Rosenthal, 1997)—have shed increasing doubt on the notion that the evaluative structure of citizens’ ideological commitments deviates significantly from bipolarity.

Interestingly, few studies in this body of work have asked whether there may be individual differences in the extent to which citizens’ ideological commitments fit the dominant bipolar model. Nevertheless, the question of how different citizens might evaluatively organize their ideological commitments is an important one. First, if some citizens’ commitments are bivariate rather than bipolar, then this suggests that some conservatives and liberals may not simply disagree with another; instead, they may be using distinct, incomparable frames of reference. As Conover and Feldman (1981) put it, conservatives and liberals may “view the political world not from different sides of the same coin,” but “from the perspective of entirely different currencies” (p. 624). Second, if conservatism and liberalism are not evaluative opposites for all citizens, then processes of political judgment may be different and more complex among such individuals: namely, they would be faced with aggregating conclusions from two different dimensions of evaluation, rather than one.

So, what variables might predict individual differences in the evaluative structure of citizens’ ideological commitments? On this score, a handful of studies point toward the importance of political expertise or the possession of well-organized systems of factual political knowledge (Delli Carpini & Keeter, 1996; Zaller, 1992). In particular, a more extensive knowledge of abstract political ideas may provide informed citizens not only with a better conceptual awareness of what conservatism and liberalism are, but also a better awareness of the fact that conservatism and liberalism are organized and treated as evaluative opposites in the elite political discussions responsible for the definition of these concepts (Converse, 1964; Hamill, Lodge, & Blake, 1985; McClosky & Zaller, 1984; Stimson, 1975; Zaller, 1992; see also Jennings, 1992). This suggests that political experts may be less likely to deviate from evaluative bipolarity in their ideological commitments, while novices may be more likely to do so. Consistent with this, multigroup confirmatory factor analyses conducted by Sidanius and Duffy (1988) found that the bipolar model fit better among Swedish and American students who were high in expertise. Conversely, while their results indicated that the bivariate model fit poorly at all expertise levels, they also suggested that its fit was especially poor among those high in expertise.

Thus, the limited evidence available tends to depict the structuring of ideological commitments as a matter of ability: Citizens are less likely to deviate from evaluative bipolarity if they have acquired the domain-relevant skills and knowledge needed to do so. However, a growing body of work suggests that motivation—in the form of generalized needs, goals, and wants—may determine if and how prior information is used to form and organize judgments (Higgins & Sorrentino, 1990; Kruglanski, 1996; Lavine, 2002). While extant work on political expertise has more broadly acknowledged the significance of motivational variables at a theoretical level (e.g., Delli Carpini & Keeter, 1996; Luskin, 1987), little or no empirical attention has been devoted to the role of motivation. Thus, an exclusive reliance on ability-related variables like expertise may leave us with only a partial account of how citizens evaluatively structure their ideological commitments. But how might motivation matter in this context? One motivational construct that has received a great deal of attention in recent years is the need to evaluate (Jarvis & Petty, 1996), i.e., the degree to which an individual is spontaneously motivated to form opinions of various objects as “good” or “bad.” Individuals with a high need to evaluate think more frequently in evaluative terms and are more likely to have formed attitudes toward a variety of objects, while those with a low need to evaluate engage in less evaluative effort of this sort. Accordingly, research suggests those high in the need to evaluate are more likely to spontaneously evaluate various objects and represent these evaluations in memory as “precomputed” attitudes that can be easily accessed when an opinion needs to be expressed (Jarvis & Petty, 1996; Tormala & Petty, 2001). In the political domain, other studies suggest that those with a high need to evaluate are more likely to display highly crystallized preferences, such as relatively extreme ideological commitments (Bizer et al., 2004; Federico, 2004).

While none of these analyses dealt with the question at hand, several characteristics of those high in the need to evaluate suggest a motivational pathway to different types of ideological structure. In particular, formal and institutional work on the functional significance of ideology suggests that the latter is useful primarily because it helps individuals reach a set of evaluative conclusions about politics and map them onto simple but fundamentally evaluative political choices (e.g., voting; Hinich & Munger, 1994; Sniderman & Bullock, 2004). Importantly, most of these choices are bipolar in nature, so the possession of ideological commitments that are also structured in an evaluatively bipolar fashion may greatly facilitate political judgment. Thus, if individuals with a high need to evaluate are more likely to be interested in forming attitudes toward political objects—and choosing between them—then a bipolar structuring of ideology may be more useful to them. Prior to any political decision, their commitments will be organized in a fashion which makes comparison and relative judgment easier: They will know that there is a “conservative” option and a “liberal” option and that one is likely to be more attractive than the other in light of one's own evaluative commitments. In contrast, citizens who approach politics without a strong motive to evaluate should have less of a need for ideological structures that facilitate judgment and choice. As such, a bipolar structure may be less functional for them, leading to greater deviation from bipolarity. Thus, those with a high need to evaluate may be less likely to deviate from ideological bipolarity, while those with a low need to evaluate may be more likely to do so.

However, given this logic, the need to evaluate may have more than a main effect on the structure of citizens’ ideological commitments. Instead, it may also determine the extent to which expertise allows individuals to organize their ideological commitments in a bipolar fashion. While expertise may provide citizens with a better awareness of the fact that conservatism and liberalism are evaluative opposites—at least as they are defined by elites—this knowledge may be more useful to those who approach politics with a strong need to evaluate. This suggests that the negative relationship between expertise and deviations from ideological bipolarity may be more pronounced among individuals with a high need to evaluate. Since these individuals will approach politics with a stronger motivation to evaluate objects and choose between them, we might expect them to use whatever understanding of ideological concepts they possess to form bipolar commitments that better prepare them for these judgments and choices. In contrast, the negative association between expertise and deviations from bipolarity may be weaker among those who are not strongly motivated to evaluate the objects they encounter. These individuals should be less motivated to establish the bipolar ideological commitments that facilitate political judgment and choices. In this case, even an expert understanding of political ideas may have a relatively small influence on ideological structure.

Thus, expertise and the need to evaluate may have important main and interactive effects on the evaluative structure of ideological commitments. However, given the organizing role ideology is believed to play vis-à-vis other attitudes, it is possible that these effects of expertise and the need to evaluate may extend to deviations from bipolarity in citizens’ evaluations of presidential candidates and political parties as well. Like evaluations of conservatism and liberalism, evaluative attraction to the Republican and Democratic candidates—and the parties that field them—can potentially be organized in either a bipolar fashion or a bivariate fashion (Green, 1988; Weisberg, 1980). In light of the increasingly strong relationship between ideological and partisan attitudes (Fiorina, 2005; Hetherington, 2001; Poole & Rosenthal, 1997), individuals who deviate more from ideological bipolarity may also deviate more from bipolarity in their attitudes toward the presidential candidates and political parties. This suggests several additional hypotheses. First, it suggests that expertise and the need to evaluate should be negatively related to deviations from bipolarity in evaluations of candidates and parties. Second, it suggests that the negative relationship between expertise and deviations from bipolarity in evaluations of these pairs of objects should be stronger among those with a high need to evaluate. Finally, given the organizing role of ideology, it suggests a pattern in which the interactive effect of expertise and the need to evaluate on deviations from bipolarity in evaluations of candidates and parties is mediated by deviations from ideological bipolarity.

Overview of the Analyses

  1. Top of page
  2. Abstract
  3. Overview of the Analyses
  4. Results
  5. Discussion
  6. ACKNOWLEDGMENT
  7. REFERENCES

In sum, the goal of the present study was to test two main hypotheses: (1) that the need to evaluate would be negatively associated with deviations from ideological bipolarity, even after controlling for the more familiar effect of political expertise; and (2) that the need to evaluate would moderate the negative relationship between expertise and deviations from bipolarity. These basic tests were supplemented by additional analyses aimed at determining (3) whether the main and interactive effects predicted for deviations from ideological bipolarity extended to deviations from bipolarity in evaluations of presidential candidates and the two major political parties; and (4) whether the interactive effects of expertise and the need to evaluate on deviations from bipolarity in evaluations of candidates and parties were mediated by reduced deviation from ideological bipolarity. Finally, since the need to evaluate tends to be moderately correlated with the need for cognition—or a general “tendency to engage in and enjoy effortful cognitive endeavors” (Bizer et al., 2004; Cacioppo, Petty, Feinstein, & Jarvis, 1996)—the effects of the latter were also taken into account. In general, the need for cognition is associated with a stronger tendency to make judgments based on the “systematic” processing of larger amounts of attitude-relevant information. Since attitude-relevant thought has been shown to increase consistency among related attitudes (Lavine, Thomsen, & Gonzales, 1997; Millar & Tesser, 1986), the general tendency toward cognitive elaboration associated with the need for cognition may also reduce deviations from bipolarity in evaluations of various objects. Therefore, in all analyses, the need for cognition and its interaction with expertise were also considered.

Method

The hypotheses were examined using data from the 2000 and 2004 National Election Studies (NES). These surveys contained measures of attitudes, political expertise, and a short form of the Need to Evaluate scale (see Bizer et al. 2004). Both the 2000 NES (N = 1,807) and the 2004 NES (N = 1,212) interviewed respondents before and after the 2000 and 2004 elections using a nationally representative sample. In 2000, respondents were randomly assigned to be interviewed either face-to-face or via telephone; in 2004, all survey interviews were carried out on a face-to-face basis. In 2000, the preelection interviews produced a response rate of 64.3% in the face-to-face mode and 56.5% via telephone. In the postelection panel, 1,555 of the preelection respondents were interviewed again, with a response rate of 86% in the face-to-face mode and 85.9% via telephone. In 2004, the preelection interviews produced a response rate of 66.1%. In the postelection panel, 1,066 of the preelection respondents were re-interviewed, with a response rate of 88%.1

Independent Variables

Political expertise. Expertise was measured using several factual-knowledge items, consistent with work suggesting that the latter are the most valid indicators of the differences in political cognition that should accompany the possession of well-developed political schemas (Delli Carpini & Keeter, 1996; Fiske, Lau, & Smith, 1990; Zaller, 1992). In 2000, eight items were included in this measure. These asked: which party controlled the House of Representatives prior to the 2000 election; (2) which party controlled the Senate prior to the election; the office held by Trent Lott; (4) the office held by William Rehnquist; the office held by Tony Blair; the office held by Janet Reno; which state George W. Bush lived in at the time of the 2000 election; and which state Al Gore was from. These items formed a reliable scale (KR-20 = .88; M = .40, SD = .29). In 2004, seven indices were used. These asked: which party controlled the House prior to the 2004 election; which party controlled the Senate prior to the election; which party was more conservative at the national level; the office held by Dennis Hastert; the office held by Dick Cheney; the office held by Tony Blair; and the office held by William Rehnquist. These items also formed a reliable scale (KR-20 = .82; M = .46, SD = .32).

Need to evaluate. This was assessed using a short version of the original scale constructed by Jarvis and Petty (1996). Both the 2000 and 2004 NES included two items. The first asked, “Some people have opinions about almost everything; other people have opinions about just some things; and still other people have very few opinions. What about you?” Responses included “almost everything,”“about many things,”“about some things,” or “about very few things.” The second asked, “Compared to the average person, do you have fewer opinions about whether things are good or bad, about the same number of opinions, or more opinions?” Those who indicated that they had “fewer” or “more” opinions than average received a follow-up question asking if they had “a lot” or “somewhat” fewer or more opinions. Responses were recoded to run from 0 to 1, where higher scores indicated a higher need to evaluate. In each dataset, the items were averaged to form a composite (α = .70; M = .56, SD = .23, in 2000; α = .64; M = .58, SD = .21, in 2004).

Need for cognition. In order to control for respondents’ general tendency to process information thoroughly, a short form of the need for cognition was included as well (Bizer et al., 2004; Cacioppo et al., 1996). In both datasets, the measure was constructed from two questions. The first asked, “Some people like to have responsibility for handling situations that require a lot of thinking, and other people don't like to have responsibility for situations like that. What about you?” Respondents were offered the option to say that they neither liked nor disliked situations involving thinking. Those who responded that they did “like” or “dislike” thinking situations were given a follow-up question where they were asked if they liked or disliked thinking situations “a lot” or “somewhat.” The second question asked, “Some people prefer to solve simple problems instead of complex ones, whereas other people prefer to solve more complex problems. Which type of problem do you prefer to solve: simple or complex?” Responses were recoded to run from 0 to 1; higher scores indicated a higher need for cognition. In each dataset, the items were averaged to form a composite (α = .67; M = .60, SD = .35, in 2000; α = .63; M = .58, SD = .35, in 2004). 2

Ideology and partisanship. In order to control for possible ideological and partisan differences in ideological structure, measures of these two predispositions were considered in both datasets. In both years, ideology was assessed using respondents’ self-placement on the 7-point NES ideology scale. Responses were recoded to run from 0 to 1, with higher scores indicating greater conservatism (M = .56, SD = .27, in 2000; M = .55, SD = .24, in 2004). Similarly, in both years, partisanship was assessed using respondents’ self-placement on the 7-point NES party identification scale. Responses were recoded to run from 0 to 1, with higher scores indicating a stronger identification with the Republican party (M = .46, SD = .35, in 2000; M = .48, SD = .35, in 2004).

Demographics and education. In both datasets, several demographics were considered: age (in years), income (in thousands of dollars per year), race (−1 = nonwhite, 1 = white), and gender (−1 = female, 1 = male). Finally, since both expertise and the need to evaluate correlate with education (Bizer et al., 2004), an index of educational attainment was also included. Earlier work has focused on the completion of a college degree as the critical experience responsible for the development of complex attitude structures in the political domain (Sniderman, Brody, & Tetlock, 1991), so a dummy variable indicating whether the respondents had completed a college degree (1 = yes; n = 556, in 2000; n = 362, in 2004) or not (−1 = no; n = 1,251, in 2000; n = 850, in 2004) was used.

Dependent Variables

Deviations from ideological bipolarity. In order to operationalize this construct, some index of the degree to which respondents fail to give oppositely valenced responses to conservative and liberal political objects—and are thus equally attracted to both—was needed. A response pattern of this sort would suggest that respondents are evaluating and gravitating toward conservative and liberal objects independently of one another—and more importantly, that they are organizing conservatism and liberalism as separate unipolar dimensions, rather than opposite ends of a single bipolar dimension. In order to provide the raw data for this index, respondents’ thermometer ratings of conservatives and liberals were used. These ratings were the simplest, most general questions about ideology in the NES surveys that also asked about conservatism and liberalism separately. Responses to these items were left on their original 0–100 scales, with higher scores indicating a “warmer” (i.e., more positive) rating of the ideological object (M = 58.65, SD = 20.00, for conservatives in 2000; M = 54.78, SD = 21.68, for liberals in 2000; M = 60.49, SD = 20.64, for conservatives in 2004; M = 55.40, SD = 20.71, for liberals in 2004).

The next step was to construct an individual-difference measure of deviations from ideological bipolarity using these indices. To this end, Thompson, Zanna, & Griffin's (1995) formula for the assessment of ambivalence was adapted. Typically, the “input” for this formula consists of ratings of the same object on separate positive and negative affect dimensions (i.e., scales ranging from no positive or negative affect at one end to intense positive or negative affect at the other). In turn, the measure produced by the formula gives an indication of the extent to which individuals express both positive and negative feelings about an object at the same time and with a similar degree of intensity, i.e., ambivalence (Craig et al., 2002; Priester & Petty, 1996). Despite its origins in the ambivalence literature, the formula itself does nothing more than provide an index of the extent to which individuals give similar responses of equally high magnitude on the scales substituted into it. As such, it can be used to quantify the extent to which respondents rate two different objects similarly and equally attractively on a single evaluative scale (i.e., one ranging from negative at one end to positive at the other, like the NES feeling thermometer). Accordingly, the Thompson et al. formula was modified to generate a deviation measure: Thermometer ratings of separate conservative and liberal objects—rather than positive and negative ratings of a single object—served as the two input quantities:

  • image

where C is the respondent's rating of conservatives and L is his/her rating of liberals. This formula also has the benefit of accounting for two different factors that contribute to deviations from bipolarity. First, the absolute difference subtracted by the second term of the equation taps the extent to which respondents evaluate conservative and liberal objects similarly. As respondents’ evaluations of the objects diverge from one another, subtraction of this term from the total reduces deviation scores. Second, the average in the equation's first term indicates the extent to which respondents give highly positive evaluations to both objects. As respondents’ average rating of both liberals and conservatives increases—indicating a positive attraction to both—overall deviation scores increase as well.3 Scores on the index increase as respondents’ ratings of conservatives and liberals deviate more from bipolarity. Once the measure was computed in each dataset, scores were recoded to run from 0 to 1 (M = .57, SD = .19, in 2000; M = .57, SD = .19, in 2004).

Deviations from bipolarity in candidate evaluations. In order to extend the analysis, measures of the extent to which respondents deviated from bipolarity in their evaluations of the 2000 and 2004 presidential candidates were also constructed. The same adapted formula used to assess deviations from ideological bipolarity was employed in the construction of these measures. The only difference is that evaluations of the Republican candidate (George W. Bush in both years) were substituted for ratings of conservatives, while evaluations of the Democratic candidate (Al Gore in 2000, John Kerry in 2004) were substituted for ratings of liberals. The evaluation score for each candidate was based on a composite of: (1) the candidate's thermometer rating; (2) the candidate's average rating on seven trait dimensions; (3) the difference between the number of positive emotions evoked by the candidate and the number of negative emotions evoked by the candidate; and (4) the difference between the number of “likes” and the number of “dislikes” listed for the candidate in open-ended items. For each candidate, these evaluations formed a reliable composite; scores were averaged and recoded to run from 0 to 1 (α = .89; M = .55, SD = .21, for Bush in 2000; α = .88; M = .58, SD = .21, for Gore in 2000; α = .80; M = .51, SD = .22, for Bush in 2004; α = .90; M = .55, SD = .23, for Kerry in 2004). These composites were then substituted into the aforementioned formula to generate final deviation indices for the candidate evaluations; these indices were recoded to run from 0 to 1 (M = .54, SD = .19, in 2000; M = .47, SD = .20, in 2004).

Deviations from bipolarity in party evaluations. Finally, measures of the extent to which respondents deviated from bipolarity in their evaluations of the two major parties in 2000 and 2004 were also generated. Again, these measures were constructed by substituting (1) evaluations of the Republican party for evaluations of conservatives and (2) evaluations of the Democratic party for evaluations of liberals in the original formula for deviations from ideological bipolarity. In each year, the evaluation of each party was based on a composite of two measures: (1) the party's thermometer rating and (2) the difference between the number of “likes” and the number of “dislikes” listed for the party in open-ended items. For each party, these evaluations formed a reliable scale; scores were averaged and recoded to run from 0 to 1 (α = .67; M = .51, SD = .18, for Republicans in 2000; α = .70; M = .56, SD = .19, for Democrats in 2000; α = .74; M = .50, SD = .20, for Republicans in 2004; α = .73; M = .56, SD = .19, for Democrats in 2004). These scores were then substituted into the formula described earlier to generate deviation indices for the party evaluations; they were then recoded to run from 0 to 1 (M = .60, SD = .17, in 2000; M = .58, SD = .21, in 2004).

Results

  1. Top of page
  2. Abstract
  3. Overview of the Analyses
  4. Results
  5. Discussion
  6. ACKNOWLEDGMENT
  7. REFERENCES

Intercorrelations between the six key variables are presented in Table 1 for both datasets. Expertise, the need to evaluate, and the need for cognition were all moderately correlated in the two datasets (with rs between .20 and .35; all ps < .001). Similarly, the three indices of deviation from bipolarity were positively correlated (ps < .001), with the indices for candidate evaluations and party evaluation having the strongest correlations (i.e., r = .60 in 2000, and r = .58 in 2004). Finally, expertise and the need to evaluate were negatively correlated with all three indices of deviation from bipolarity (all ps < .001). The need for cognition was also negatively correlated with all three deviation indices, albeit less strongly (all rs less than or equal to .15).

Table 1.  Intercorrelations for Study Variables
Variable12345
  • Note. All coefficients are Pearson correlations.

  • *

    p < .05.

  • **

    p < .01.

  • ***

    p < .001.

2000 National Election Study
1. Political expertise    
2. Need to evaluate.23***   
3. Need for cognition.23***.33***  
4. Deviation from Bipolarity: Ideology−.19***−.14***−.03 
5. Deviation from Bipolarity: Candidates−.28***−.22***−.12***.34***
6. Deviation from Bipolarity: Parties−.23***−.18***−.05*.40***.60***
2004 National Election Study
1. Political expertise    
2. Need to evaluate.32***   
3. Need for cognition.24***.31***  
4. Deviation from Bipolarity: Ideology−.30***−.19***−.08** 
5. Deviation from Bipolarity: Candidates−.21***−.21***−.15***.37***
6. Deviation from Bipolarity: Parties−.25***−.17***−.06***.43***.58***

Expertise, Need to Evaluate, and Deviations from Ideological Bipolarity

The hypotheses (1) that the need to evaluate would predict deviations from ideological bipolarity and (2) that the need to evaluate would moderate the relationship between expertise and deviations from bipolarity were tested in each dataset using a series of ordinary least-squares regression models. In these models, the index of deviations from ideological bipolarity was regressed on expertise, the need to evaluate, need for cognition, and the Expertise × Need to Evaluate and Expertise × Need for Cognition interactions. Age, income, race, gender, the college-degree indicator, ideology, and partisanship were included in each model as well. In order to guard against possible effects of heteroskedasticity, Huber-White robust standard errors were used in these analyses (Long & Ervin, 2000). Finally, all predictors were centered (see Aiken & West, 1991).

The results for the 2000 NES are summarized in Table 2. Model 1 examined the main effects of expertise, need to evaluate, and need for cognition, while controlling for the covariates. As expected, those high in expertise (b = −.11, p < .001) and the need to evaluate (b = −.09, p < .01) were less likely to exhibit deviations from ideological bipolarity, even after the covariates were considered. Moreover, partisanship was also related to the deviation index (p < .001), such that Republican identifiers were less likely to deviate from bipolarity. Finally, need for cognition was marginally related to deviations from bipolarity (p < .10), such that those higher in need for cognition were more likely to deviate from bipolarity in their ideological sentiments. In turn, Model 2 added the Expertise × Need to Evaluate and Expertise × Need for Cognition interactions to the model. While the key interaction between expertise and the need to evaluate was significant and in the predicted negative direction (b = −.31, p < .01), the interaction between expertise and the need for cognition did not reach significance (p > .10). Importantly, the result for the first interaction suggests that the relationship between expertise and deviations from ideological bipolarity was stronger among those high in the need to evaluate. In order to probe this interaction, simple slopes for the relationship between expertise and the deviation index were computed at need-to-evaluate levels one standard deviation above and below the variable's mean, using Aiken and West's (1991) method. These analyses indicated that the relationship between expertise and deviations from bipolarity was more strongly negative when the need to evaluate was high (b = −.17, SE b = .03, p < .001) than when the latter was low (b = −.02, SE b = .03, p > .40).

Table 2.  Expertise, Need to Evaluate, and Deviation from Ideological Bipolarity (2000 NES)
PredictorDeviation From Ideological Bipolarity
Model 1Model 2
bSE bbSE b
  • Note. Entries are unstandardized OLS regression coefficients and Huber-White robust standard errors.

  • +

    p < .10.

  • *

    p < .05.

  • **

    p < .01.

  • ***

    p < .001.

Age.0002(.0003).0002(.0003)
Income.00002(.002)−.0001(.002)
Race−.003(.01)−.003(.01)
Gender−.001(.01)−.001(.01)
College degree.003(.01).003(.01)
Ideological self-description−.02(.02)−.02(.02)
Partisanship−.08***(.02)−.07***(.02)
Political expertise−.11***(.02)−.09***(.02)
Need to evaluate−.09***(.03)−.06***(.03)
Need for cognition.03+(.02).02(.02)
Expertise × Need to evaluate−.31**(.10)
Expertise × Need for cognition.08(.06)
Constant.59***(.01).59***(.01)
F (degrees of freedom)7.04 (10, 1157)***6.75 (12, 1155)***
R2.063.071
N1,1681,168

The results for the 2004 NES are summarized in Table 3; the analyses were run in exactly the same fashion. Again, Model 1 indicated that those high in expertise (b = −.15, p < .001) and the need to evaluate (b = −.11, p < .01) were less likely to exhibit deviations from ideological bipolarity, even after the covariates were taken into account. Moreover, the estimates for this model also indicated that whites were slightly less likely to exhibit deviations from bipolarity (p < .05). In turn, the estimates for Model 2 indicated that the Expertise × Need to Evaluate interaction was significant (b = −.38, p < .001), but that the Expertise × Need for Cognition was not (p > .10). Simple slope analyses were again used to probe the significant Expertise × Need to Evaluate interaction. These analyses indicated that the relationship between expertise and deviations from bipolarity was more strongly negative when the need to evaluate was high (b = −.24, SE b = .04, p < .001) than when the need to evaluate was low (b = −.08, SE b = .03, p < .05). Thus, the aforementioned hypotheses were supported with regard to ideological bipolarity.

Table 3.  Expertise, Need to Evaluate, and Deviation from Ideological Bipolarity (2004 NES)
PredictorDeviation From Ideological Bipolarity
Model 1Model 2
bSE bbSE b
  • Note. Entries are unstandardized OLS regression coefficients and Huber-White robust standard errors.

  • +

    p < .10.

  • *

    p < .05.

  • **

    p < .01.

  • ***

    p < .001.

Age.001(.0004).001(.0004)
Income−.001(.001)−.001(.001)
Race−.02*(.01)−.02*(.01)
Gender−.003(.01)−.003(.01)
College degree−.01(.01)−.01(.01)
Ideological self-description−.04(.03)−.04(.03)
Partisanship−.03(.02)−.03(.02)
Political expertise−.15***(.03)−.16***(.03)
Need to evaluate−.11***(.03)−.07*(.03)
Need for cognition−.002(.02)−.001(.02)
Expertise × Need to evaluate−.38***(.11)
Expertise × Need for cognition−.08(.06)
Constant.58***(.01).59***(.01)
F (degrees of freedom)12.48 (10, 846)***11.25 (12, 844)***
R2.128.144
N857857

Deviations from Bipolarity in Evaluations of Presidential Candidates and Political Parties

Given the overlap between ideology and evaluations of key political objects, the hypotheses outlined earlier suggest that the foregoing effects should also extend to deviations from bipolarity in citizens’ evaluations of presidential candidates and political parties. This prediction was examined in a second set of OLS regression analyses. In these models, indices of the extent to which respondents deviated from bipolarity in their evaluations of (1) the two major presidential candidates and (2) the two major political parties in 2000 and 2004 were regressed on the same of independent variables used in the analyses summarized in Tables 2 and 3.

For the 2000 NES, the results of these analyses are shown in Table 4. Again, for each dependent variable, Model 1 simply looked at the main effects of expertise, the need to evaluate, the need for cognition, and the covariates. For the candidate-evaluations index, the estimates indicated that those high in expertise (b = −.13, p < .001) and the need to evaluate (b = −.15, p < .001) were less likely to exhibit deviations from bipolarity. Moreover, there was a small, marginally significant tendency for those with higher income to deviate less from bipolarity (p < .10). For the party-evaluations index, the results were similar: Those high in expertise (b = −.12, p < .001) and the need to evaluate (b = −.12, p < .001) were less likely to exhibit deviations from bipolarity, while those with a higher need for cognition were marginally more likely to deviate from bipolarity (p < .10).

Table 4.  Expertise, Need to Evaluate, and Deviations from Bipolarity in Candidate and Party Evaluations (2000 NES)
PredictorCandidate EvaluationsParty Evaluations
Model 1Model 2Model 1Model 2
bSE bbSE bbSE bBSE b
  • Note. Entries are unstandardized OLS regression coefficients and Huber-White robust standard errors.

  • +

    p < .10.

  • *

    p < .05.

  • **

    p < .01.

  • ***

    p < .001.

Age−.001(.0003)−.001(.0003)−.0003(.0003)−.0002(.0003)
Income−.003+(.002)−.003+(.002)−.002(.002)−.002(.002)
Race−.004(.01)−.005(.01)−.0003(.01)−.001(.01)
Gender.01(.01).01+(.01).001(.01).002(.01)
College degree.001(.01).002(.01)−.003(.01)−.003(.01)
Ideological self-description−.01(.02)−.01(.02).003(.02).003(.02)
Partisanship.01(.02).01(.02).03(.02).03+(.02)
Political expertise−.13***(.02)−.13***(.02)−.12***(.02)−.11***(.02)
Need to evaluate−.15***(.02)−.15***(.02)−.12***(.02)−.12***(.02)
Need for cognition−.03(.02)−.01(.02).03+(.01).03+(.01)
Expertise × Need to evaluate−.18*(.08)−.11+(.07)
Expertise × Need for cognition−.02(.05).02(.05)
Constant.57***(.01).58***(.01).60***(.01).60***(.01)
F (degrees of freedom)16.78 (10, 1,421)***14.54 (12, 1,419)***12.31 (10, 1,421)***10.29 (12, 1,419)***
R2.106.110.077.079
N1,4321,4321,4321,432

In turn, for each dependent variable, Model 2 added the Expertise × Need to Evaluate and Expertise × Need for Cognition interactions. The Expertise × Need to Evaluate was in the correct negative direction for both dependent variables, although it was only marginally significant with regard to the party-evaluations index (i.e., b = −.18, p < .001, for the candidate-evaluations analysis; b = −.11, p < .10, for the party-evaluations index). The Expertise × Need for Cognition interaction did not reach significance for either dependent variable (both ps > .10). Simple-slope analyses were then used to probe each significant interaction. For both dependent variables, these analyses indicated that the relationship between expertise and deviations from bipolarity was more strongly negative when the need to evaluate was high (b = −.17, SE b = .02, p < .001, for candidate evaluations; b = −.14, SE b = .02, p < .001, for party evaluations) than when the need to evaluate was low (b = −.09, SE b = .03, p < .001, for candidate evaluations; b = −.09, SE b = .02, p < .001, for party evaluations).

For the 2004 NES, the results of these analyses were quite similar; they are summarized in Table 5. The results for Model 1 are discussed first. For the candidate-evaluations index, the estimates indicated that those high in expertise (b = −.13, p < .001) and the need to evaluate (b = −.12, p < .001) were less likely to exhibit deviations from bipolarity. Moreover, there were marginally significant tendencies for men and Republican identifiers to deviate more from bipolarity (p < .10), as well as a tendency for those high in need for cognition to deviate less from bipolarity (p < .05). For the party-evaluations index, the results were similar: Those high in expertise (b = −.19, p < .001) and the need to evaluate (b = −.09, p < .01) were less likely to exhibit deviations from bipolarity, while men (p < .01) and Republican identifiers (p < .001) were more likely to deviate from bipolarity. Those with high income were also marginally less likely to deviate from bipolarity (p < .10).

Table 5.  Expertise, Need to Evaluate, and Deviations from Bipolarity in Candidate and Party Evaluations (2004 NES)
PredictorCandidate EvaluationsParty Evaluations
Model 1Model 2Model 1Model 2
bSE bbSE bbSE bbSE b
  • Note. Entries are unstandardized OLS regression coefficients and Huber-White robust standard errors.

  • +

    p < .10.

  • *

    p < .05.

  • **

    p < .01.

  • ***

    p < .001.

Age−.0002(.0004)−.0002(.0004)−.001(.0004)−.001(.0004)
Income−.001(.001)−.001(.001)−.002+(.001)−.002*(.001)
Race.0004(.01)−.0002(.01)−.01(.01)−.01(.01)
Gender.01+(.01).01+(.01).02**(.01).02**(.01)
College degree−.002(.01)−.001(.01)−.01(.01)−.01(.01)
Ideological self-description.01(.03).01(.03)−.004(.03)−.004(.03)
Partisanship.04+(.02).04+(.02).10***(.02).10***(.02)
Political expertise−.13***(.03)−.13***(.03)−.19***(.03)−.19***(.03)
Need to evaluate−.12***(.03)−.09**(.03)−.09**(.03)−.05(.03)
Need for cognition−.05*(.02)−.05**(.02).01(.02).003(.02)
Expertise × Need to evaluate−.27*(.11)−.43***(.11)
Expertise × Need for cognition.02(.07).03(.07)
Constant.47***(.01).58***(.01).57***(.01).58***(.01)
F (degrees of freedom)8.83 (10, 881)***7.81 (12, 879)***15.54 (10, 881)***15.07 (12, 879)***
R2.088.094.140.145
N892892892892

Again, for each dependent variable, Model 2 added the pair of two-way interactions. The Expertise × Need to Evaluate interaction was significant in the correct negative direction for both dependent variables (i.e., b = −.27, p < .05, for the candidate-evaluations analysis; b = −.43, p < .001, for the party-evaluations index). The Expertise × Need for Cognition interaction did not reach significance for either dependent variable (both ps > .10). Simple-slope analyses were then used to probe the significant interactions. For both dependent variables, these analyses indicated that the relationship between expertise and deviations from bipolarity was more strongly negative when the need to evaluate was high (b = −.19, SE b = .04, p < .001, for candidate evaluations; b = −.28, SE b = .04, p < .001, for party evaluations) than when the need to evaluate was low (b = −.08, SE b = .04, p < .05, for candidate evaluations; b = −.10, SE b = .04, p < .01, for party evaluations). Thus, the hypotheses were supported for deviations from bipolarity in evaluations of candidates and parties.

Mediation Analyses

Finally, the hypotheses suggest that any interactive effect of expertise and evaluative motivation on deviations from bipolarity in evaluations of the two “distal” but ideologically linked objects—candidates and parties—should be mediated by the interactive effect of expertise and evaluative motivation on deviations from ideological bipolarity. This suggests a pattern of mediated moderation (Baron & Kenny, 1986). In the present context, three conditions would need to be met in order for this pattern to obtain: (1) expertise and the need to evaluate must interact to predict deviations from bipolarity in candidate and party evaluations; (2) expertise and the need to evaluate must interact to predict the hypothesized mediator, i.e., deviations from ideological bipolarity; and (3) for models predicting deviations from bipolarity in candidate and party evaluations, the interaction between expertise and the need to evaluate must be significantly reduced in magnitude in a regression containing both this interaction and the mediator, with the mediator remaining significant (Baron & Kenny 1986; Wegener & Fabrigar, 2000).

These analyses are summarized in Figure 1 (for the 2000 NES) and Figure 2 (for the 2004 NES). In both datasets, the first condition is satisfied by the results summarized in Tables 4 and 5; these indicate significant or marginally significant interactive effects of expertise and the need to evaluate on deviations from bipolarity in evaluations of candidates and parties. Moreover, the second condition is satisfied in both datasets by the results summarized in Tables 2 and 3; these results indicate significant interactive effects of expertise and the need to evaluate on deviations from ideological bipolarity. Finally, the third condition was examined by adding the mediator—the index of deviations from ideological bipolarity—to the Model 2 equations for each “distal” deviation variable in both datasets. In all four analyses summarized in Figures 1 and 2, the mediator was significantly related to deviations from bipolarity in evaluations of candidates, as shown in the top panel of each figure, and parties, as shown in the bottom panel of each figure (all ps < .001). Moreover, in 2000, the interaction between expertise and the need to evaluate had a negative indirect effect on deviations from bipolarity in both candidate evaluations (IE = −.09, Sobel z = 2.99, p < .01) and party evaluations (IE = −.11, Sobel z = 3.05, p < .01). Similarly, in 2004, the same key interaction had a negative indirect effect on deviations from bipolarity in candidate (IE = −.17, Sobel z = 3.15, p < .01) and party evaluations (IE = −.20, Sobel z = 3.20, p < .001).4 In fact, in three of the four analyses, the interaction was reduced to nonsignificance once the mediator was added to the model; estimates for the effects of the interaction in the presence of the mediator are shown in parentheses in the figures. Thus, the mediated-moderation hypotheses were supported in both datasets.

image

Figure 1. Mediated Moderation Analyses for Deviations from Bipolarity in Evaluations of Candidates and Parties: 2000 NES. (Note: *p < .05, **p < .01, ***p < .001.)

Download figure to PowerPoint

image

Figure 2. Mediated Moderation Analyses for Deviations from Bipolarity in Evaluations of Candidates and Parties: 2004 NES. (Note: *p < .05, **p < .01, ***p < .001.)

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Overview of the Analyses
  4. Results
  5. Discussion
  6. ACKNOWLEDGMENT
  7. REFERENCES

Political psychologists have long been interested in the structural organization of citizens’ ideological commitments. Among other things, they have recurrently engaged in discussion about whether ideology is organized in a bipolar fashion, where conservatism and liberalism are understood as opposed ends of a single evaluative spectrum, or a bivariate fashion, where conservatism and liberalism correspond to independent dimensions of political evaluation (Conover & Feldman, 1981; Green, 1988; Kerlinger, 1984). While most work suggests that the bipolar structure is the psychological norm in mass publics, few studies have investigated the possibility of individual differences in the extent to which the evaluative structure of citizens’ ideological commitments conform to bipolarity or deviate away from it. In particular, not much work has focused on the potential motivational antecedents of individual differences in the structure of citizens’ ideological commitments; instead, most studies have focused on ability-related variables like political expertise. The present study attempted to fill these gaps by examining two primary hypotheses: (1) that an important motivational variable—the need to evaluate, or the desire to form opinions—would be negatively associated with deviations from ideological bipolarity, even after the effects of expertise were taken into account; and (2) that the need to evaluate would moderate the negative relationship between expertise and deviations from bipolarity.

Analyses using the 2000 and 2004 National Election Studies provided clear support for these hypotheses. Extending these basic results, other analyses provided support for a series of additional hypotheses. These analyses indicated that the need to evaluate was also negatively associated with deviations from bipolarity in evaluations of objects closely linked to ideology: namely, presidential candidates and the two political parties. Moreover, these analyses also indicated that the need to evaluate moderated the negative relationship between expertise and deviations from bipolarity in evaluations of each object. Finally, mediated-moderation analyses indicated that expertise and the need to evaluate interacted to predict decreased deviation from ideological bipolarity, which in turn led to decreased deviation from bipolarity in attitudes toward the other two sets of objects.

So, what implications do these results have? At the simplest level, they reinforce and extend earlier findings on the relationship between political expertise and the evaluative structure of citizens’ ideological commitments. Consistent with this earlier work (e.g., Sidanius & Duffy, 1988), the results presented here indicated that expertise was negatively related to deviations from ideological bipolarity in two different datasets, even after a variety of other factors were taken into account. Moreover, the results also suggest that expertise is negatively related not just to deviations from ideological bipolarity, but also to deviations from bipolarity in evaluations of objects closely linked to ideology—namely, attitudes toward presidential candidates and political parties.

More importantly, though, these results provide evidence for the notion that the evaluative organization of citizens’ ideological commitments is a function not just of ability but of motivation as well. Even after the effects of expertise and various controls were considered, a key motivational variable—the need to evaluate—was significantly related to the evaluative structure of respondents’ ideological commitments. This finding suggests that the desire to use information evaluatively may have an effect of its own on the extent to which individuals treat conservatism and liberalism as evaluative opposites.5 More broadly, these findings also add to a growing body of work on the nature and significance of the need to evaluate. Previous research on this construct clearly suggests that it may play a key role in the establishment and organization of attitudes across a variety of domains. Among other things, studies suggest that the need to evaluate is associated with a greater willingness to form and express attitudes toward various objects (Bizer et al., 2004; Jarvis & Petty, 1996), to express extreme attitudes (Federico, 2004), to display behaviors consistent with the possession of highly crystallized political preferences (Bizer et al., 2004; Federico, 2004), and to establish “precomputed” evaluations (rather than ones assembled on the spot; Tormala & Petty, 2001). Nevertheless, most of these findings relate to properties of attitudes toward specific objects. In contrast, the results reported here suggest that the need to evaluate may also have consequences for the structuring of a broader evaluative predisposition, namely, ideology. Thus, the impact of the need to evaluate may extend not just to the characteristics of attitudes toward specific objects, but also to the generalized predispositions believed to constrain these attitudes.

More broadly, our results suggest that researchers may want to focus more attention on the role of “motivated social cognition” in the construction and organization of political evaluations. In this vein, a number of perspectives in social and political psychology imply that judgments are a joint function of factors related to ability or the possession of information (such as expertise) and motivational factors that determine how skills or information are used to establish and organize attitudes (Kruglanski, 1996; Lavine, 2002). These perspectives have been applied in a variety of other literatures (e.g., persuasion; see Cacioppo et al., 1996; see also Kruglanski, 1996), but they have only recently been extended to analyses of how individuals structure their broader political commitments, such as those related to ideology. In general, work in this area has focused largely on the role of ability-related factors like expertise (Sidanius & Duffy, 1988; see also Judd & Krosnick, 1989). Moreover, in the present context, they have usually assumed that individuals with the expertise needed to understand the bipolar structuring of ideology in elite discourse are also motivated to make use of this expertise in organizing their own commitments.

Departing from this general focus, the present study takes a more interactive approach: It suggests that expertise is more likely to be associated with less deviation from ideological bipolarity in the presence of a motivation to use this expertise in a broadly evaluative fashion. As such, the present study adds to a body of recent work suggesting that the structuring and use of ideological predispositions is an interactive function of expertise and evaluation motivation. For example, Federico (2004) finds that expertise and the tendency to clearly identify oneself as liberal or conservative are more strongly related among those with a high need to evaluate. Interestingly, this study also found that expertise and the need to evaluate interacted to predict “evaluative integration,” a tendency to evaluate sets of politicians from the Democratic and Republican parties in an oppositely valenced fashion. Given the similarity between this measure and the present concept of “ideological bipolarity,”Federico's (2004) results can be looked at as early evidence for the interactive finding discussed here. However, the results of the present study go beyond these early findings in several ways. First, Federico (2004) did not focus explicitly on the evaluative structure of ideology per se; rather, the focus was on the mediating role of domain-specific evaluative consistency development of extreme attitudes. Second, Federico (2004)'s analyses did not focus directly on how individuals organize their evaluations of specifically ideological referents, i.e., “liberals” in general and “conservatives” in general. Thus, this previous study is only able to speak indirectly to the evaluative organization of citizens’ ideological commitments. Finally, the present study also looks at how the interactive effect of expertise and evaluative motivation fits into a broader set of relationships—namely, a pattern in which the interactive effect of expertise and the need to evaluate on deviations from ideological bipolarity also has important “downstream” effects on the structure of citizens’ evaluations of candidates and parties.

These, however, are not the only other results consistent with the general notion of a critical interaction between expertise and evaluative motivation. For instance, among those with a high need to evaluate, expertise more strongly predicts ideological conceptualizations of the differences between political parties (Federico & Schneider, in press). Perhaps most importantly, though, other analyses suggest that expertise is more likely to predict issue constraint (i.e., ideological agreement among one's issue positions; Converse, 1964) among those with a high need to evaluate (Federico & Schneider, in press). Like the results reported here, this last finding reinforces the notion that an “expert” understanding of how elites conceptualize ideology is more likely to be applied to the structuring of one's own commitments and attitudes in the presence of a strong motivation to evaluate various objects. Thus, under a high need to evaluate, experts’ broader belief systems are more likely to be marked both by an “appropriate” relationship between liberalism and conservatism themselves and by “appropriate” relationships between liberal and conservative issue positions.6

Thus, along with other findings, the results reported here make a notable contribution to our understanding of how citizens organize and make use of ideological commitments. Nevertheless, these data and analyses are not without their limitations. Perhaps most importantly, the correlational nature of the data and analyses makes it difficult to draw firm conclusions about underlying causal relations. While the logic of this paper implies that expertise and the need to evaluate should have a causal impact on the evaluative structure of individuals’ ideological commitments—and the structure of their evaluations of objects linked to ideology—the fact that these variables could not be easily manipulated in the context of the NES data precludes firm causal inference.

A related issue has to do with the present study's relatively novel methodological approach to the assessment of deviations from ideological bipolarity. Most studies in this area of inquiry have relied on factor-analytic methods where correlations between latent dimensions corresponding to conservatism and liberalism were examined at the sample or subsample level (Kerlinger, 1972, 1984; Sidanius & Duffy, 1988). However, the analyses reported here have attempted to measure the structure of ideological commitments at the individual level. While the latter approach avoids the repeated sample-splitting necessary for multigroup factor analyses, it also has a number of shortcomings. First, it makes the results of this study less comparable to the results of prior studies. Second, it precludes some of the adjustments for measurement error possible in confirmatory factor analysis. For example, the individual-level approach used here does not allow for the construction of a factor-analytic measurement model to account for random error in the operational definition of variables corresponding to “conservatism” and “liberalism” (Kerlinger, 1980; Sidanius & Duffy, 1988) or the inclusion of error covariances capable of accounting for nonrandom error due to method variance (Green, 1988).7 Where possible, future studies will need to consider these issues.

ACKNOWLEDGMENT

  1. Top of page
  2. Abstract
  3. Overview of the Analyses
  4. Results
  5. Discussion
  6. ACKNOWLEDGMENT
  7. REFERENCES

Correspondence concerning this article should be addressed to Christopher M. Federico, University of Minnesota, Twin Cities, Department of Psychology, N-218 EltH, 3281, 75 East River Road, Minneapolis, MN 55455. E-mail: federico@umn.edu

REFERENCES

  1. Top of page
  2. Abstract
  3. Overview of the Analyses
  4. Results
  5. Discussion
  6. ACKNOWLEDGMENT
  7. REFERENCES
  • Achen, C. H. (1975). Mass political attitudes and the survey response. American Political Science Review, 69, 12181223.
  • Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage Publications.
  • Alvarez, R. M., & Brehm, J. (2002). Hard choices, easy answers: Values, information, and American public opinion. Princeton: Princeton University Press.
  • Asher, H. B. (1980). Presidential elections and American politics, 2nd ed. Homewood, IL: Dorsey.
  • Baron, R. M., & Kenny, D. G. (1986). The moderator-mediator variable distinction in social-psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 11731182.
  • Bizer, G. Y., Krosnick, J. A., Holbrook, A. L., Wheeler, S. C, Rucker, D. E., & Petty, R. E. (2004). The impact of personality on cognitive, behavioral, and affective political processes: The effects of the need to evaluate. Journal of Personality, 72, 9951027.
  • Brown, S. R., & Taylor, R. W. (1973). Frames of reference and the observation of behavior. Social Science Quarterly, 54, 2940.
  • Cacioppo, J. T., Petty, R. E., Feinstein, J. A., & Jarvis, W. B. G. (1996). Dispositional differences in cognitive motivation: The life and times of individuals varying in need for cognition. Psychological Bulletin, 119, 197253.
  • Campbell, A., Converse, P., Miller, W., & Stokes, W. (1960). The American voter. New York: Wiley.
  • Conover, P. J., & Feldman, S. (1981). The origins and meaning of liberal-conservative self-identifications. American Journal of Political Science, 25, 617645.
  • Conover, P., & Feldman, S. (1984). How people organize the political world: A schematic model. American Journal of Political Science, 28, 95126.
  • Converse, P. (1964). The nature of belief systems in mass publics. In D.Apter (Ed.), Ideology and discontent. New York: Free Press.
  • Craig, S. C., Kane, J. G., & Martinez, M. D. (2002). Sometimes you feel like a nut, sometimes you don't: Citizens’ ambivalence about abortion. Political Psychology, 23, 285303.
  • Delli Carpini, M. X., & Keeter, S. (1996). What Americans know about politics and why it matters. New Haven: Yale University Press.
  • Duckitt, J. (2001). A dual process cognitive-motivational theory of ideology and prejudice. Advances in Experimental Social Psychology, 22, 41113.
  • Federico, C. M. (2004). Predicting attitude extremity: The interactive effects of expertise and the need to evaluate—and their mediation by evaluative integration. Personality and Social Psychology Bulletin, 30, 12811294.
  • Federico, C. M., & Schneider, M. (in press). Political expertise and the use of ideology: Moderating effects of evaluative motivation. Public Opinion Quarterly.
  • Fiorina, M. P. (2005). Culture war? The myth of a polarized America. New York: Pearson Longman.
  • Fiske, S. T., Lau, R. R., & Smith, R. A. (1990). On the varieties and utilities of political expertise. Social Cognition, 8, 3148.
  • Green, D. P. (1988). On the dimensionality of public sentiment toward partisan and ideological groups. American Journal of Political Science, 32, 758780.
  • Hamill, R., Lodge, M., & Blake, F. (1985). The breadth, depth, and utility of partisan, class, and ideological schemas. American Journal of Political Science, 29, 850870.
  • Hetherington, M. J. (2001). Resurgent mass partisanship: The role of elite polarization. American Political Science Review, 95, 619631.
  • Higgins, E. T., & Sorrentino, R. M. (eds.) (1990). The handbook of motivation and cognition: Foundations of social behavior (Vol. 2). New York: Guilford Press.
  • Hinich, M. J., & Munger, M. C. (1994). Ideology and the theory of political choice. Ann Arbor: University of Michigan Press.
  • Jarvis, W. B., & Petty, R. E. (1996). The need to evaluate. Journal of Personality and Social Psychology, 70, 172194.
  • Jennings, M. K. (1992). Ideological thinking among mass publics and political elites. Public Opinion Quarterly, 56, 419441.
  • Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. (2003). Political conservatism as motivated social cognition. Psychological Bulletin, 129, 339375.
  • Judd, C. M., & Krosnick, J. A. (1989). The structural bases of consistency among political attitudes: Effects of expertise and attitude importance. In A. R.Pratkanis, S. J.Breckler, & A. G.Greenwald (Eds.), Attitude structure and function (pp. 99128). Hillsdale, NJ: Erlbaum.
  • Judd, C. M., Krosnick, J. A., & Milburn, M. A. (1981). Political involvement and attitude structure in the general public. American Sociological Review, 46, 660669.
  • Kerlinger, F. N. (1967). Social attitudes and their criterial referents: A structural theory. Psychological Review, 74, 110122.
  • Kerlinger, F. N. (1972). The structure and content of social attitude referents: A preliminary study. Educational and Psychological Measurement, 32, 613630.
  • Kerlinger, F. N. (1980). Analysis of covariance structure tests of a criterial referents theory of attitudes. Multivariate Behavioral Research, 15, 403422.
  • Kerlinger, F. N. (1984). Liberalism and conservatism: The nature and structure of social attitudes. Hillsdale, NJ: Lawrence Erlbaum.
  • Kinder, D. R., & Sears, D. O. (1985). Public opinion and political action. In G.Lindzey & E.Aronson (Eds.), Handbook of social psychology (3rd ed., Vol. 2, pp. 659741). Reading, MA: Addison-Wesley.
  • Knoke, D. M. (1979). Stratification and the dimensions of American political orientations. American Journal of Political Science, 23, 772791.
  • Kruglanski, A. W. (1996). Motivated social cognition: Principles of the interface. In E. T.Higgins & A. W.Kruglanski (Eds.), Social psychology: A handbook of basic principles (pp. 493522).New York: Guilford Press.
  • Lane, R. E. (1962). Political ideology. New York: Free Press.
  • Lavine, H. (2002). Online versus memory-based process models of political evaluation. In K. R.Monroe (Ed.), Political psychology (pp. 225247). Mahwah, NJ: LEA.
  • Lavine, H., Thomsen, C. J., & Gonzales, M. H. (1997). The development of interattitudinal consistency: The shared consequences model. Journal of Personality and Social Psychology, 72, 735749.
  • Layman, G. C., & Carsey, T. M. (2002). Party polarization and “conflict extension” in the American electorate. American Journal of Political Science, 46, 786802.
  • Levitin, T. E., & Miller, W. E. (1979). Ideological interpretations of presidential elections. American Political Science Review, 73, 751771.
  • Long, J. S., & Ervin, L. H. (2000). Using heteroskedasticity consistent standard errors in the linear regression model. American Statistician, 54, 217234.
  • Luskin, R. (1987). Measuring political sophistication. American Journal of Political Science, 31, 856899.
  • Luttbeg, N. (1968). The structure of beliefs among leaders and the public. Public Opinion Quarterly, 32, 398409.
  • Marcus, G. E., Tabb, D., & Sullivan, J. L. (1974). The application of individual differences scaling to the measurement of political ideologies. American Journal of Political Science, 18, 405420.
  • McClosky, H., & Zaller, J. (1984). The American ethos. New York: Harvard University Press.
  • Millar, M. G., & Tesser, A. (1986). Thought-induced attitude change: The effects of schema structure and commitment. Journal of Personality and Social Psychology, 51, 259269.
  • Mulaik, S. A. (1972). Foundations of factor analysis. New York: McGraw-Hill.
  • Poole, K. T., & Rosenthal, H. (1997). Congress: A political-economic history of roll-call voting. New York: Oxford University Press.
  • Priester, J. R., & Petty, R. E. (1996). The gradual threshold model of ambivalence: Relating the positive and negative bases of attitudes to subjective ambivalence. Journal of Personality and Social Psychology, 71, 431449.
  • Schwartz, S. H. (1992). Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. Advances in Experimental Social Psychology, 25, 165.
  • Shafer, B. E., & Claggett, W. J. M. (1995). The two majorities: The issue context of American politics. Baltimore: Johns Hopkins.
  • Sidanius, J., & Duffy, G. (1988). The duality of attitude structure: A test of Kerlinger's criterial referents theory within samples of Swedish and American youth. Political Psychology, 9, 649670.
  • Sniderman, P. M., Brody, R. A., & Tetlock, P. E. (1991). Reasoning and choice. New York: Cambridge University Press.
  • Sniderman, P. M., & Bullock, J. (2004). A consistency theory of public opinion and political choice: The hypothesis of menu dependence. In W. E.Saris & P. M.Sniderman (Eds.), Studies in public opinion: Attitudes, nonattitudes, measurement error, and change (pp. 337357). Princeton, NJ: Princeton University Press.
  • Stimson, J. A. (1975). Belief systems: Constraint, complexity, and the 1972 election. American Journal of Political Science, 19, 393417.
  • Thompson, M. M., Zanna, M. P., & Griffin, D. W. (1995). Let's not be indifferent about (attitudinal) ambivalence. In R. E.Petty & J. A.Krosnick (Eds.), Attitude strength: Antecedents and consequences (pp. 361386). Mahwah, NJ: LEA.
  • Tormala, Z. L., & Petty, R. E. (2001). On-line versus memory-based processing: The role of “need to evaluate” in person perception. Personality and Social Psychology Bulletin, 27, 15991613.
  • Wegener, D. T., & Fabrigar, L. R. (2000). Analysis and design for non-experimental data: Addressing causal and non-causal hypotheses. In H. T.Reis & C. M.Judd (Eds.), Handbook of research methods in social and personality psychology (pp. 412450). New York: Cambridge University Press.
  • Weisberg, H. F. (1980). A multidimensional conceptualization of party identification. Political Behavior, 2, 3360.
  • Weisberg, H. F., & Rusk, J. G. (1970). Dimensions of candidate evaluation. American Political Science Review, 64, 11671185.
  • Wilson, G. D. (ed.) (1973). The psychology of conservatism. London: Academic Press.
  • Zaller, J. (1992). The nature and origins of mass opinion. New York: Cambridge University Press.
Footnotes
  • 1

    All response rates are in Response Rate 1 (RR1) format, as defined by the American Association for Public Opinion Research (see http://www.aapor.org/pdfs/standarddefs_4.pdf).

  • 2

    Previously published confirmatory factor analyses of the need to evaluate and need for cognition items in the 2000 NES indicated that a correlated two-factor model provides a better fit to the data than a one-factor model (Bizer et al., 2004). Similar analyses conducted on the two sets of items in the 2004 NES (using the WLS estimator, with polychoric correlations between the items as input) revealed that a correlated two-factor model fit the data well, χ2(1) = .19, p > .30; while a one-factor model did not, χ2(2) = 150.88, p < .001. This difference in fit was highly significant, Δχ2(1) = 150.69, p < .001. Thus, in both datasets, the need-to-evaluate and need-for-cognition items appear to measure distinct constructs, despite the correlation between the scales (see Table 1).

  • 3

    In contrast, simpler indices would capture only the first component, thereby underestimating the level of deviation from bipolarity in the data (Craig, Kane, & Martinez, 2002). For example, a simple absolute difference would treat someone who gave both conservatives and liberals a thermometer rating of 50 (at the midpoint) as showing the same deviation from bipolarity as someone who gave both objects the “warmest” rating of 100—even though there is presumably more deviation from bipolarity in an individual who is more attracted to both groups. In any case, the results of the key analyses were the same in both datasets when deviation from ideological bipolarity was operationalized solely in terms of the absolute difference between respondents’ thermometer ratings of conservatives and liberals.

  • 4

    All indirect effects are unstandardized.

  • 5

    An alternative explanation for the results is that deviation from bipolarity simply reflects increased error in the measurement of ideology and partisanship. In order to examine this possibility, heteroskedastic regressions were used to examine the relationship between the three deviation scores and the errors of prediction in models predicting respondents’ 7-point ideology and partisanship scores (see Alvarez & Brehm, 2002). However, in both datasets, the variance-equation estimates revealed that all three deviations were associated with lower levels of error in predicting ideology and partisanship. This suggests that the deviation scores do not reflect mere random error.

  • 6

    Federico and Schneider (in press) report results indicating that expertise and the need to evaluate interact to predict increased issue constraint in the 1998 NES Pilot Study and the 2000 NES. A similar analysis conducted in the other dataset used here—the 2004 NES—finds similar results. Expertise and the need to evaluate significantly interacted to predict constraint (b = .11, p < .05), and the relationship between expertise and constraint was stronger among those high in the need to evaluate (b = .07, p < .001) than it was among those low in the need to evaluate (b = .03, p < .10).

  • 7

    In the case of attitudes toward the 2000 and 2004 presidential candidates, the NES datasets included enough measures of each object for the estimation of a LISREL confirmatory-factor analytic model capable of accounting for both random and nonrandom measurement error. In these models, latent variables corresponding to evaluations of the Republican and Democratic presidential candidates were specified, with the four measures for each specified as the indicators for the latent variables. In turn, to account for the effects of nonrandom error on the correlations between these factors, error terms for measures using the same response format were allowed to covary across factors (see Green, 1988). Factor scores for the Republican and Democratic candidates were then generated from the resulting model estimates and plugged into the deviance equation described previously. The results for the candidate-evaluation models using these modified indices produced exactly the same pattern for results: The need to evaluate was negatively associated with deviations from bipolarity, and the negative relationship between expertise and deviations from bipolarity was significantly stronger among those with a high need to evaluate. Of course, given the indeterminacy of factor scores (Mulaik, 1972), these results should be interpreted cautiously.