Intergroup Attitudes 1

Ideologies that underlie concepts of ethnocentrism, authoritarianism, system justification, social dominance, and morality shape minds in sufficiently deep ways to bring about (a) congruence between implicit and explicit preferences, and (b) a consistently greater preference for socially advantaged groups among political conservatives than liberals on both explicit and implicit measures. Data from large web samples and representative samples from the American National Election Studies (ANES) provide support for these and two additional results: (a) liberals show greater mean dissociation between explicit and implicit attitudes than conservatives, reporting more favorable attitudes toward the underprivileged groups than they demonstrate on implicit measures; and (b) over time, conservatives’ racial preferences converge on those of liberals, suggesting that where liberals are today, conservatives will be tomorrow. Intergroup Attitudes 2 The Politics of Intergroup Attitudes Intergroup attitudes are made up of complex strands of social preferences. They are held together by political ideologies that serve as orienting systems guiding personality as well as responses to the environment such as decisions about the information one chooses to consume, the activities one pursues, and the policies one supports (Jost, 2006). They are sufficiently central to social cognition that they are visible in the the neural markers that distinguish a politically similar other from one who is dissimilar (Mitchell, Macrae & Banaji, 2006). In this chapter we rely on two large datasets that provide substantial evidence regarding attitudes toward multiple social groups (e.g., groups based on religion, sexuality, ethnicity/race, age, and gender). From these data we examine the role of political ideology as an organizing concept for the structure and function of social attitudes; simultaneously, we examine intergroup attitudes to understand more about the liberal-conservative (or left-right) political divide. In the last two decades, the idea that attitudes, like other mental processes, may reside in both conscious/explicit as well as less conscious/implicit form has come to be well-accepted (e.g., Bargh, 1997; Greenwald & Banaji, 1995). This distinction in attitudes may apply to philosophical and ideological belief systems as well (Jost, 2006; Jost, Banaji, & Nosek, 2004; Jost, Glaser, Kruglanski, & Sulloway, 2003). That is, political ideology – an interconnected set of beliefs and attitudes that shape judgment – may not exist solely as a reasoned or explicit collection of beliefs and attitudes. Ideology has unconscious as well as conscious determinants, and the latter is well explicated elsewhere (Cunningham, Nezlek & Banaji, 2004; see Ferguson, Carter, & Hassin, this volume). In this chapter, we examine the variation in ideological orientation in relation to implicit and explicit attitudes, with a specific focus on attitudes toward social groups. We start by revisiting Jost et al.’s (2003) theoretical argument (and supporting meta-analysis) that liberals and conservatives differ on two key dimensions: resistance to change and tolerance for inequality. We then focus on the key prediction that conservatives are more likely than liberals to have and express more positive attitudes toward high-status or advantaged groups and more negative attitudes toward low-status or disadvantaged groups. We will describe that this is true on both conscious and less conscious measures of intergroup attitudes. In this sense, the intergroup attitudes of conservatives tend to be more system-justifying than those of liberals insofar as they support and perpetuate the existing social hierarchy (see also Jost et al., 2004; Jost, Nosek, & Gosling, in press). We also find--using data from the American National Election Studies (ANES)--that liberals are at the forefront of the social movement toward racial egalitarianism, whereas conservatives’ attitudes were slower to change. Ideological Differences between Liberals and Conservatives The “classic” conception of the authoritarian personality (Adorno, FrenkelBrunswik, Levinson, & Sanford, 1950), its modern instantiation (Altemeyer, Intergroup Attitudes 3 1996), recent perspectives on ideology stressing motivated social cognition (Jost et al., 2003), system justification (Jost & Banaji, 1994; Jost et al., 2004, in press), social dominance (Sidanius & Pratto, 1999), and moral foundations (Haidt & Graham, 2007) provide accounts of differences between the political left and right, or, more commonly in the U.S., liberalism and conservatism. These perspectives converge on the expectation that, compared to liberals, conservatives are less concerned with equality, more comfortable maintaining the status quo, and more likely to show favoritism for high-status or advantaged groups over low-status or disadvantaged groups. To the extent that conservative, system-justifying attitudes are characterized by resistance to change and tolerance for inequality (Jost et al., 2003), their appeal should be maximized when stability and order are prioritized values. In the study of authoritarianism, psychologists have long observed that societal crises (e.g., economic upheavals, terrorist attacks) often precipitate rightward political shifts, presumably because conservative, right-wing opinions typically resonate with heightened needs to manage uncertainty and threat (Doty, Peterson, & Winter, 1991; Sales, 1972, 1973; McCann, 1997; Ullrich & Cohrs, 2007). As Huntington (1957) put it, “When the foundations of society are threatened, the conservative ideology reminds men of the necessity of some institutions and desirability of the existing ones.” In other words, system-level threats stimulate the motivation to justify the system. A meta-analytic review of the psychological antecedents of political conservatism by Jost et al. (2003) supports this view. Specifically, they found that situational as well as dispositional variables associated with the management of threat and uncertainty predicted various manifestations of political conservatism (including economic system justification). The original studies were conducted in 12 countries between 1958 and 2002 and employed 88 different research samples involving a total of 22,818 individual cases. Results indicated that the tendency to endorse conservative (rather than liberal or moderate) opinions is positively associated with threat variables such as mortality salience (or death anxiety), system instability, and fear of threat and loss, and it is negatively associated (albeit weakly) with self-esteem. Conservatism is also positively associated with uncertainty avoidance, intolerance of ambiguity, and needs for order, structure, and closure, and it is negatively associated with openness to experience and integrative complexity. Although the meta-analysis focused on explicit, self-reported attitudes and beliefs, recent research using implicit measures mirrors these ideological differences. For example, ideological differences in resistance to change were demonstrated by Jost, Nosek, and Gosling (in press), who found that implicit and explicit attitudes toward tradition, stability, and the status quo were predictors of political orientation. More specifically, conservatism was associated with greater implicit as well as explicit preferences for order compared to chaos, conformity compared to rebelliousness, stability compared to flexibility, tradition compared to progress, and traditional values compared to feminism. In simultaneous regressions, both implicit and explicit attitudes showed unique predictive validity of political orientation, suggesting that they are non-redundant indicators of Intergroup Attitudes 4 ideological proclivities. Ideological differences in implicit social cognition also relate to the tolerance of inequality and, specifically, favoritism for higher over lower status groups. Jost, Banaji, and Nosek (2004) found that individual differences in political orientation moderated implicit attitudes for social groups. Measured with the Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998), conservatives, compared to liberals, showed stronger preferences for White Americans over African Americans and for heterosexuals over homosexuals (see also Cunningham, Nezlek, & Banaji, 2004; see also Sidanius & Pratto, 1999 for similar results at the explicit level). The meta-analysis and subsequent investigations of implicit social cognition provide an initial basis for the notion that conservatives are more likely to show favoritism for higher than lower status groups than liberals, on both implicit and explicit measures. We examined this possibility across a variety of target groups with nationally representative samples of voters collected through the American National Election Studies (ANES) and with large datasets collected over the Internet. Measuring Ideology and Implicit Attitudes Validity of a single-item political ideology assessment Remarkably, even the simplest of questions – self-placement on a singleitem liberal to conservative dimension – appears to be an effective means of parsing individual differences in ideological orientations. Jost (2006) analyzed American National Election Studies data from 1972 to 2004 and found that a selfplacement on a 7-point single item of strongly liberal to strongly conservative explained 85% of the variance in voting behavior for Democratic and Republican candidates for president. Similar evidence is available from large datasets showing that selfplacement on a liberal-conservative dimension discriminates both explicit and implicit attitudes toward politicians. Nosek and colleagues (Nosek et al., 2007) summarized approximately six years’ worth of data collected at Project Implicit web sites (see https://implicit.harvard.edu/). The aggregated datasets included more than 2.5 million IATs and self-reported attitude assessments across more than a dozen topics, including social attitudes, stereotypes, and political attitudes. Three of the studies examined attitudes toward George Bush compared to (1) Al Gore (collected before and after the 2000 U.S. presidential election), (2) John Kerry (collected before and after the 2004 election), and (3) previous U.S. presidents (individual and aggregate comparisons). For each of these data collections, participants reported their explicit candidate preferences, completed an IAT contrasting Bush with another politician, and reported their political orientation on a 5-, 6-, or 7-point scale of liberalism-conservatism. Results for Project Implicit 2004 Election data using a 6-point liberal-conservative self-rating are presented in Figure 1 (N = 22,904), and all other datasets replicated this pattern (Nosek et al., 2007). Conservative participants favored Bush over Kerry both implicitly and explicitly, and liberal Intergroup Attitudes 5 participants favored Kerry over Bush. Also, the political preferences were strongest for more extreme conservatives and liberals. These effects replicate the political preferences expressed in the ANES, and show that self-reported political ideology is predictive of implicit as well as explicit political attitudes. These results are additional evidence that ideology is not incoherent and meaningless, as some have concluded (e.g., Bishop, 2005; Converse, 1964), even when measured with a “bare bones” single item (see also Jost, 2006). Measuring Implicit Attitudes In the 10 years since its initial publication (Greenwald et al., 1998), a sizable literature of over 500 papers has developed using and evaluating the Implicit Association Test (IAT; see Nosek, Greenwald, & Banaji, 2007 for a review; and, Greenwald, Poehlman, Uhlmann, & Banaji, in press for a metaanalysis of predictive validity evidence). The IAT specifically provides an estimate of the strength of association between concepts such as gay and good/bad compared with straight and good/bad. Participants sort exemplars representing those four categories in two key sorting conditions. In one condition, items representing gay people and bad things are categorized with one response (a key press), and items representing straight people and good things are categorized with an alternate response. In a second condition, the response configuration is switched so that gay people and good things are categorized with one response and straight people and bad things are categorized with the alternate response. Categorizing the items faster in the first compared to the second condition is interpreted as indicating an implicit preference for straight people relative to gay people. The accumulated evidence suggests that the constructs measured by the IAT and self-report measures are related but distinct (Nosek & Smyth, 2007), and the strength of their relation varies from weakly to strongly positive depending on the topic of assessment (Nosek, 2005). Notably, in contrast to the well-known result of dissociation between explicit and implicit attitudes, attitudes toward political candidates and some political issues (e.g., pro-choice versus pro-life) elicit some of the strongest implicit-explicit relations, with r’s sometimes above .70 (Nosek, 2007). The IAT also shows predictive validity of judgment and behavior in a variety of domains with explicit measures showing better predictive validity in some cases (e.g., consumer preferences), and the IAT showing better predictive validity in others (e.g., intergroup discrimination-related behaviors; Greenwald, et al., in press). Such congruence may stem from political attitudes being subserved by a general liberal-conservative ideology that binds preferences and provides psychological consistency. Current research efforts are investigating the underlying processes that contribute to IAT effects (Conrey, Sherman, Gawronski, Hugenberg, & Groom, 2005; Rothermund & Wentura, 2004), and refining the understanding of the IAT’s relation to self-report, behavior, and to other implicit measures such as the AMP (Payne et al., 2005), GNAT (Nosek & Banaji, 2001), evaluative priming (Fazio et al., 1986), and the SPF (Bar-Anan, Nosek, & Vianello, 2007). Intergroup Attitudes 6 Internet Data: Correlations between Political Orientation and Implicit and Explicit Attitudes A key topic investigated in this chapter is whether political ideology – measured by self-placement on a liberal-conservative dimension – is related to implicit and explicit attitudes toward social groups. In particular, following theories of ethnocentrism, authoritarianism, system justification, motivated social cognition, and moral foundations, are conservatives more likely than liberals to prefer higher status groups than lower status groups? And, is this pattern observed with both implicit and explicit measurement methods? In a review of the large datasets collected at Project Implicit (Nosek et al., 2007), eight of the investigated topics concerned preferences between social groups that differ in terms of social status. To examine the relationship between political orientation and social group preferences, Nosek and colleagues regressed the attitude measure (IAT or a single-item self-reported preference) on to self-reported political orientation and three demographic covariates (gender, age, and ethnicity). Across topics and measurement methods, conservatives consistently favored higher status groups to a greater degree than did liberals. Conservatives showed relatively stronger implicit preferences than liberals for White Americans compared to African Americans (ηP = .009), light-skinned people compared to dark-skinned people (ηP = .006), White children compared to Black children (ηP = .013), others compared to Arab-Muslims (ηP = .012), others compared to Jews (ηP = .008), abled people compared to disabled people (ηP = .005), straight people compared to gay people (ηP = .057), and thin people compared to overweight people (ηP = .004). Likewise, conservatives self-reported relatively stronger explicit preferences than liberals for White Americans compared to African Americans (ηP = .029), light-skinned people compared to dark-skinned people (ηP = .013), White children compared to Black children (ηP = .023), other people compared to Arab-Muslims (ηP = .039), other people compared to Jews (ηP = .035), and straight people compared to gay people (ηP = .126). There was minimal variation across political orientation in explicit preferences for the abled compared to the disabled (ηP = .001), thin people compared to fat people (ηP = .001), and there was a slight tendency for liberals to show a stronger proyoung/anti-old preference (ηP = -.002). An aggregated comparison of implicit and explicit preferences for all 8 topics is presented in Figure 2. The x-axis presents self-rated political orientation from strongly liberal to strongly conservative. Positive values on the y-axis indicate a preference for higher status groups. The displacement of effect sizes from zero indicates that, across the political spectrum, the sample as a whole (both liberals and conservatives) favored higher status groups. The positive slope shows that greater conservatism was associated with stronger preferences for higher compared to lower status groups. At the extremes, even strong liberals showed a preference for higher status groups both implicitly (d = 0.48) and explicitly (d = 0.28), but strong conservatives showed a 65% stronger implicit preference than liberals (d = 0.79), Intergroup Attitudes 7 and a 186% stronger explicit preference (d = 0.80). That is, all groups favored higher status groups on average, and conservatives did so to a greater degree than liberals did. In summary, liberals and conservatives differ somewhat in their implicit intergroup preferences, and differ more substantially in their explicit intergroup preferences. Liberals showed a larger discrepancy between their implicit higherstatus preferences and their comparatively weaker explicit higher-status preferences, whereas conservatives showed very little discrepancy between their strong implicit higher-status preferences and their strong explicit higher-status preferences (assuming the comparability of effect sizes). These findings suggest that liberals are more likely to have a complex or perhaps “conflicted” stance in their social evaluations insofar as they harbor some degree of automatic bias along with an explicit commitment to egalitarianism. In this sense, in everyday judgment and action, liberals may make greater efforts to override their automatic reaction in favor of a more egalitarian explicit response (e.g., Skitka et al., 2001). Conservatives, on the other hand, show greater consistency on average in their implicit and explicit social evaluations suggesting that conservatives may be more likely than liberals to justify and use their automatic reactions as a basis for explicit report and judgment (see also Jost et. al., 2003, 2004; Sniderman, Crosby, & Howell, 2000). Explicit attitudes toward African and White Americans by respondent ethnicity. The prior section examined relative preferences between higher and lower status groups because of the procedural constraint of relative comparison in the IAT and the use of relative explicit preference measures. For this chapter, we conducted additional analyses of Nosek et al.’s (2007) data and analyzed selfreported thermometer ratings for each group separately. Also, we compared racial attitudes of White American, African American, and other respondents. Social identity theory anticipates that group members will tend to like their own group more than others (Hewstone, Rubin, & Willis, 2002; Tajfel, 1978; Tajfel & Turner, 1986). But, the hypothesis that conservatives will be more favorable to higher status groups and more unfavorable to lower status groups than liberals will be does not depend on the individual’s own group membership. That is, to the extent that conservatism is linked to a desire to preserve the status quo and acceptance of inequalities among groups (Jost et al., 2003), conservatives should possess more negative attitudes than liberals toward lower status groups whatever their own group membership. For example, conservative Blacks would be expected to report more negative attitudes toward African Americans than would liberal Blacks (see also Jost et al., 2004). Figure 3 presents regression estimates of thermometer attitude ratings (0 cold to 10 warm) toward White Americans on the left and African Americans on the right separated by respondent ethnicity (White [n = 255,590], Black [n = 34,216], Other [n = 72,834]). The x-axis arrays respondents according to self1 Notably, the implicit-explicit consistency is greater at the mean level for conservatives than liberals, but this difference does not emerge at the level of individual differences. The correlation between implicit and explicit social attitudes seems to be equally strong for liberals and conservatives (Nosek et al., 2007). Intergroup Attitudes 8 reported political orientation. Considering attitudes toward White Americans first, there is a main effect of ethnicity in which White respondents, across the political spectrum, reported liking White Americans more than respondents of other racial groups did. At the same time, there was a main effect of conservatism such that, for all three racial groups, conservatives expressed more positive attitudes toward White Americans than did liberals in the same racial group. The strength of the conservatism-attitude relationship was strongest for White respondents compared to the other two groups (Whites β = .11; Blacks β = .06; Others β = .06). For Whites, this corresponded to an estimated attitude difference of .8 of a scale point (on a 0 to 10 scale) between attitude ratings of the strong liberals and the strong conservatives. A distinctly different pattern was observed for attitudes toward African Americans. As before, a main effect of respondent ethnicity was observed. This time, Black respondents across the political spectrum reported more positive attitudes toward African Americans than did White and other respondents. This in-group effect for both racial groups is consistent with social identity theory (Tajfel & Turner, 1986). Conservatism was not only positively associated with liking for White Americans, it was also negatively associated with liking for African Americans. For all three racial groups, liberals reported more positive attitudes toward African Americans than conservatives did. Again, the strength of the conservatism-attitude relationship was strongest for White respondents (Whites β = -.12; Blacks β = -.05; Others β = -.08). For Whites, this corresponded to a difference of .9 of a scale point between attitudes of strong liberals and strong conservatives. Strong liberals, who are White, are estimated to have slightly more positive attitudes toward White Americans (est M = 6.81) than African Americans (est M = 6.61) whereas strong conservatives, who are White, show a difference of almost two full scale points favoring White Americans (est M = 7.61) over African Americans (est M = 5.71). Black respondents showed a similar in-group preference in their White and Black warmth ratings, except that Black conservatives showed less in-group preference than Black liberals, whereas the opposite was observed with respect to White conservatives and liberals (see also Jost et al., 2004). This latter finding shows that ideology predicts variation in group attitudes beyond that accounted for by social identity – one’s group membership. Other high and low status groups. We reanalyzed the data for the other attitude domains from the Nosek et al. (2007) report that contained a 6-point single item liberal-conservative self-rating and separate thermometer ratings of 2 Because the samples were very large, all reported effects were estimated reliably with p < .0001. Figures and text report data analyses using a 6-point liberal to conservative response item. Data collected with a 7-point response item showed similar results for all reported effects. 3 Actual means were very similar to the regression estimates. Strong liberals (n = 32,880, Whites only) reported slightly more positive attitudes toward White Americans (M = 6.91) than African Americans (M = 6.66), and strong conservatives (n = 13,836, Whites only) reported substantially more positive attitudes toward White Americans (M = 7.78) than African Americans (M = 5.61). Intergroup Attitudes 9 the higher and lower status groups. These included the following with the higher status group listed second: dark skin/light skin (n = 67,561), old people/young people (n = 174,289), Arab-Muslims/other people (n = 34,520), Jews/other people (n = 39,021), disabled people/abled people (n = 20,729), gay people/straight people (n = 38,511), and fat people/thin people (n = 99,142). Regression estimates predicting warmth ratings by self-rated liberalismconservatism appear in Figure 4 with the higher status groups in the bottom panel and lower status groups in the top panel. For higher status groups, all of the regression lines show a positive slope suggesting that conservatives hold more favorable explicit attitudes toward these groups than liberals do. This effect was observed consistently and with varying magnitude for attitudes toward light-skinned people (β = .11), non-Arabs (β = .08), non-Jews (β = .27), abled people (β = .10), young people (β = .04), straight men (β = .34), and thin people (β = .11). Across all groups examined, attitudes toward higher status groups were again more positive among conservatives than among liberals (see also Jost et al., 2004). For lower status groups, greater conservatism was expected to be associated with more negativity toward one’s own group. As can be observed in Figure 4, there was variability across intergroup comparisons. Three comparisons revealed that conservatives had more unfavorable explicit attitudes than liberals did toward dark-skinned people (β = -.09), Arab-Muslims (β = -.34), and gay men (β = -.62). Four others had near zero or weakly positive attitudeideology relations: Jews (β = .01), thin people (β = .02), disabled people (β = .03), and old people (β = .05). This suggests that some comparisons do not elicit as ideologically differentiated attitudes toward the lower status group, as do other comparisons. Examining the high and low status groups separately with explicit, selfreport data provided some interesting insights into the differences in group attitudes by ideology. Explicitly, at least, the pattern is consistent, though with variable magnitude: Conservatives report stronger favoritism for high status groups than liberals do. The pattern is less consistent for attitudes toward lower status groups. When substantial differences are observed, the pattern was as expected – conservatives held less favorable attitudes toward low status groups than liberals did. In sum, these data suggest that the liberal-conservative differences are driven primarily by conservatives’ greater liking than liberals for higher-status groups, rather then consistently greater disliking than liberals for lower-status groups. This intriguing result deserves further investigation as the existing 4 Similar effects were observed with the portions of the data using a 7-point liberal-conservative item. Also, one task – attitudes toward Black and White children – only contained data using a 7point ideology item. Data for that task were similar to the other race tasks described. 5 Thermometer ratings were given separately for gay men, lesbians, straight men and straight women. Because they show the same pattern, only the male data are presented in Figure 4. Attitudes toward lesbians (β = -.52) showed a similar relation with ideology as attitudes toward gay men, and attitudes toward straight women (β = .29) were similar to attitudes toward straight men. Intergroup Attitudes 10 theoretical perspectives are mute with regard to whether the effect is primarily a low status disliking or a high status liking effect. The effects reported in this section have the advantage of being drawn from large and heterogeneous datasets allowing highly reliable estimation and confidence in the robustness of the effects in a very diverse sample. As an unselected dataset, the data are not, however, representative of the U.S. population. Next, we sought to replicate the explicit preference effects in a nationally representative data collection – the American National Election Studies (implicit measurement of nationally representative samples was not yet available at the time of writing this chapter). Data from the American National Election Studies Started in 1948, the American National Election Studies (ANES) has conducted studies of the American electorate every 2 or 4 years with a representative sample of 1,000 to 3,000 Americans on each occasion. The surveys were conducted in face-to-face or telephone interviews in a structured format. Besides surveying a representative sample of Americans, an important feature of the ANES is that similar or identical items were used on multiple occasions, allowing aggregation and cross-sectional comparisons by year. Two sets of items were of particular interest for the present purposes. For one, since 1972, ANES has included a 7-point “strongly liberal” to “strongly conservative” single-item measure. Also, ANES respondents provided warmth ratings on a thermometer scale toward a variety of different social groups. Attitudes toward some groups, such as Blacks and Whites, were measured on most occasions, and attitudes toward other groups were measured less frequently. From the available set, we selected feeling thermometer ratings toward a range of social groups and analyzed the available data from 1972 through 2004. This resulted in a sample of 14 social groups— each with 4 to 15 measurement occasions. The Table lists the social groups, the years that they were included in the feeling thermometer rating section of the ANES, the total sample size, the mean warmth rating (range 0 to 97), and the empirical relationships between ideology and intergroup attitudes, described next. We conducted a multilevel regression for each social group with year of data collection as the grouping variable, thermometer ratings as the dependent variable that was regressed onto political orientation and five covariates, namely gender, ethnicity, age, religion, and family income. The effects of political orientation on attitudes toward each of the social groups after adjusting for the other individual difference variables is presented in the last column of the Table as an effect size (Cohen’s d). Negative values indicate that more negative attitudes toward the group were associated with greater conservatism (vs. liberalism); positive values indicate that more positive attitudes toward the group were associated with greater conservatism (vs. liberalism). As can be observed in the Table, for seven social groups – gays/lesbians, 6 The dataset was the 1948-2004 ANES Cumulative Data File retrieved from http://electionstudies.org/ Intergroup Attitudes 11 people on welfare, illegal aliens, blacks, young people, Hispanics, and poor people – conservatism was associated with relatively more negative attitudes toward the group (average d = -.13). For four social groups – Catholics, middle class people, Whites, and southerners – liberalism was associated with relatively more negative attitudes toward the group compared to conservatism (average d = .07). There were no reliable relationships between political orientation and attitudes for three groups – Jews, Asian Americans, and Protestants (average d = .01). The social groups included in the ANES vary in terms of their size and social status. For domains in which the higher and lower status groups were included, conservatives tended to show greater liking for the higher status group and greater disliking of the lower status group, in comparison with liberals. For example, liking of Blacks and Hispanics was negatively associated with conservatism, whereas liking of Whites was positively associated with conservatism. Likewise, liking of poor people was negatively associated with conservatism, whereas liking of middle class people was positively associated (see also Jost et al., 2004). Across groups there was a strong association between the overall liking of the group (averaging across liberals and conservatives) and the ideology-attitude relationship. The correlation between mean warmth ratings and the ideologyattitude relationship was r = .76. In other words, the groups that were most disliked overall were the ones that differentiated liberals and conservatives the most, with conservatives being more negative. This association was driven by the three most disliked groups in the dataset – gays/lesbians (M warmth rating = 38.8), people on welfare (M = 50.9), and illegal aliens (M = 35.9). In short, ideological differences were most apparent for attitudes toward the most disliked groups. The results suggest that lower-status groups are viewed more unfavorably by conservatives than liberals. At the same time, liberals do not like everyone more than conservatives do. Conservatives reported more liking of some groups, especially Whites and the middle class, who possess higher social status than their racial and economic counterparts. In addition, conservatives reported greater favorability toward southerners than did liberals. Southerners are not easily recognized as high-status or dominant, but they are generally seen as more conservative than other regional groups. These data from nationally representative ANES data collections replicate the findings from the very large datasets reviewed earlier, at least with regard to explicit evaluations. In the next section, we narrow our focus to racial attitudes to take advantage of the fact that the ANES offers opportunities for cross-sectional comparisons over time. Following the Jost et al. (2003) review, conservatives’ comparatively greater resistance to change and tolerance of inequality should make liberals more likely to be at the forefront of social change movements aimed to increase egalitarianism between groups. Changes in explicit racial attitudes in the United States from the mid-20 to early 21 centuries provide an ideal circumstance to test this prediction. Intergroup Attitudes 12 Liberals as Social Change Agents Following the theoretical analysis comparing liberals’ and conservatives’ attitudes about social change and inequality (Jost et al., 2003, in press), liberals should be more likely than conservatives to instigate social change aimed at reducing social inequalities. Successful social movements, however, are those that ultimately convince liberals, moderates, and conservatives alike that the inequalities are both real and unjustified. Jost et al. (2003) did not suggest that conservatives are altogether unconcerned with inequality; rather their review suggested that conservatives are less concerned with inequality than are liberals (see also Graham, Haidt, & Nosek, 2007). They also observed that conservatives are more likely than liberals to believe that society is inevitably hierarchical. Further, because conservatives are relatively more resistant to change in general, they are likely more wary of corrective measures to redress inequality when they require substantive changes to the existing social order. The combination of differential concerns about inequality and resistance to change suggests that social movements for changing inequalities are more likely to be introduced and led by liberals and to be resisted, especially at first, by conservatives. Liberals, in prioritizing social equality over social order, may provide the culture with an early-warning signal for the presence of an excessive inequality among groups. Conversely, conservatives, in prioritizing social order over social equality, may caution against unnecessary changes to a societal status quo and warn against the uncertainties of change. Conservatives’ degree of sensitivity to order and their tolerance for inequality frequently puts them in the role of skeptic concerning social movements that seek to ameliorate social inequalities. We hypothesize that those social movements that produce change are ones in which liberals are ultimately able to convince conservatives that the inequalities are important to fix, and that social change can occur without cultural collapse. Eventually, the liberal-initiated changes become part of the status quo, and conservative resistance relents and embraces the new world order. In the context of social group prejudices, this account of liberal and conservative reactions to social movements leads to the following predictions: (a) Early in a social movement, liberals’ and conservatives’ attitudes toward a given stigmatized group will be most differentiated, with liberals being relatively more favorable toward the target group than conservatives; (b) Over time, for successful social movements, the attitude gap between liberals and conservatives will progressively narrow, as conservatives become more accepting of the social change and more supportive of equal rights and treatment for the stigmatized group. A Test of the Social Change Hypotheses: Attitudes toward Blacks in the United States From 1972 to 2004, the American National Election Studies included feeling thermometer ratings toward Blacks in every data collection except for 1978. These cross-sectional data provide an opportunity to test whether liberals and conservatives differ most in their attitudes toward Blacks early in the civil Intergroup Attitudes 13 rights movement (as early as is available in this dataset), and the extent to which the liberal-conservative gap closed in the intervening years. The ANES data provide clear support for both hypotheses (see Figure 5). We conducted a multilevel regression with attitudes toward Blacks as the dependent variable, year of data collection as the grouping variable, race, gender, family income, age, and religion as covariates, and political orientation entered as a predictor and random effects factor. The latter variable tests the hypothesis that the relationship between political orientation and attitudes toward Blacks changes over time. A main effect of year reveals that attitudes toward Blacks became more favorable over time (z = 2.54, p = .006). Likewise, a main effect of political orientation indicates that liberals tended to be more favorable toward Blacks than did conservatives over the 32-year period (B = -1.08, SE B = .18, t = -6.19, p < .0001). Critically, the random effect of political orientation by year was significant despite there being only 15 measurement occasions (z = 1.66, p = .048). The attitude gap between liberals and conservatives was strongest in the earliest years available, and the gap narrowed significantly in later years. The pattern shows that liberals’ and conservatives’ attitudes toward Blacks were most differentiated in the 1970’s and that this differentiation narrowed over time and was completely absent (even non-significantly reversed) in the latest two data collections (2002 and 2004). Figure 6 provides further illustration by showing the regression estimates for attitudes toward Blacks for the first (1972) and last (2004) data collection by ideology. The most extreme liberals were nearly as positive toward Blacks in 1972 as they were 32 years later, suggesting very little change among liberals over the time span. The most extreme conservatives, on the other hand, showed a substantial change (cross-sectionally) in attitudes over the same period. The regression estimates for extreme conservatives have them giving a 56 warmth rating for Blacks in 1972 and a 70 rating in 2004. In effect, these data suggest that conservatives’ explicit attitudes toward Blacks eventually “caught up” to liberals’ over this 32-year time period in American history. Cross-sectional analyses of attitudes toward Blacks in ANES data were consistent with our hypothesis that liberals are more likely to be the instigators of social change aimed at redressing social group inequalities, and conservatives initially resist. Whereas strong liberals were already quite positive toward Blacks in 1972, conservatism was then associated with markedly less favorable attitudes toward Blacks. Over time, attitudes toward Blacks became more favorable in general, with the rest of the political spectrum ultimately joining liberals in the expression of more positive attitudes. 7 An additional model that excluded Black participants from the dataset shows the same pattern of results as reported here. 8 Note that the large web datasets reviewed earlier found that a small political difference in racial attitudes persisted into the 21 century (data collected 2000-2006). 9 The main caution for this interpretation is the fact that the data is cross-sectional, not longitudinal. Different people participated in each year’s data collection. Intergroup Attitudes 14 Intergroup Attitudes and the Changing of Political Minds Across two large data sets, and using both implicit and explicit methods of measurement, we consistently observed that conservatives more than liberals tilt in their preferences toward higher status groups over lower status groups (Jost et al., 2004). Looking at explicit attitudes toward higher and lower status groups separately, we see that this effect in conservatives occurs at both ends: they prefer higher status groups more than liberals and dislike lower status groups more than liberals, with the former being the more consistent effect. There is variation in the extent to which liberals and conservatives differ. Some intergroup comparisons elicit strong polarization (e.g., gay men), whereas others elicit minimal variation by ideology (e.g., young people). An obvious next step for this area of research is to identify the factors that exaggerate or minimize ideological polarization with respect to specific target groups. An observation of interest from data from the Project Implicit website is that the ideological polarization for group preferences was substantially stronger for explicit reports than for implicit measurement. Nosek et al. (2007) reported that, after adjusting for variations in age, gender, and ethnicity, the average effect size across topics for political ideology predicting implicit preferences was ηP = .013 and the average effect size for explicit preferences was ηP = .030, more than twice the magnitude. Conservatives show robust preferences for high status groups, with implicit and explicit preferences having approximately the same effect sizes on average for moderate and strong conservatives (Figure 2). For liberals, however, a “conflicted stance” more accurately describes the data, with explicitly reported preferences being discrepant from implicit ones and consistently more egalitarian on average. Why might this be? Our prevailing hypotheses focus on change – both in the immediate situation and over time. At any given moment, liberals and conservatives may have fairly similar automatic reactions to social groups but deal with their reactions differently. Liberals may be more suspicious of their own automatic responses and seek to suppress, change or modify their influence on explicit report and judgment (Skitka et al., 2001). Conservatives, on the other hand, might be more likely to accept their automatic responses as valid and use them to guide explicit judgment. This hypothesis does not address the fact that there are indeed some implicit attitude differences across the ideological spectrum, suggesting that automatic reactions are not all the same for liberals and conservatives. This provides the basis for our hypothesis about change over time. Evidence for the automatization of cognitive processes points to practice and elaboration as key influences (Logan, 1988; Nosek, 2005). If liberals actively reject or modify their automatic responses and practice replacing those reactions with judgments that conform to their explicit values, then, over time, those alternative evaluations may become automatized themselves. This is consistent with evidence that people who are chronic egalitarians or high in the motivation to respond without prejudice show less implicit bias against African Americans than others do (e.g., Devine, Plant, Amodio, Harmon-Jones, & Vance, 2002; Moskowitz, Gollwitzer, Wasel, & Schaal, Intergroup Attitudes 15 1999). Liberals typically have stronger motivations to avoid prejudice than conservatives. As a consequence, they may practice egalitarian responses more frequently, accounting for their lower degree of implicit preference for higher status groups. From this perspective, a liberal’s effort to practice egalitarianism is first an explicit shift that, over time, fosters an implicit shift as well. It also highlights the possibility that, based on the present data, the change of mind is incomplete. If practicing egalitarian responses is the key to reducing implicit biases, then most liberals have more practicing to do before attaining a high degree of implicitexplicit consistency. Another possibility is that the type of change that leads to explicit egalitarianism is not sufficient for changing implicit evaluations. Deciding that one has changed one’s mind will likely alter what is said and endorsed, but it need not change much beyond that, and it certainly doesn’t have one-to-one impact on the local and global environment. If implicit evaluations are sensitive to stamp of culture, then a change of mind must be followed by a change of behavior and a change in the situation or else implicit-explicit discrepancies will persist. Conservatives, on the other hand, may be less likely to reject their automatic responses in the first place, less likely to practice egalitarian responses, and thus less likely to modify their automatic preferences for the higher status groups. The substantial change in conservatives’ racial attitudes in the ANES dataset suggests that ideological differences in explicit racial attitudes have dissipated substantially. If those indeed reflect genuine shifts, then they may be accompanied by increased motivations among conservatives to be racially egalitarian. This increase in motivation could lead to a more consistent practice of egalitarian responses and ultimately to a reduction or elimination of differences in implicit racial biases between liberals and conservatives. Future research will speak to the extent and quality of this social change. These hypotheses offer bi-directional perspectives on implicit-explicit influence: that changes in explicit preferences can, over time, lead to implicit changes as a cause of implicit-explicit consistency, and that, in the moment, an automatic reaction can either be used or rejected as a basis for generating an explicit report or judgment. At the same time, this chapter may reflect another directional assumption: that ideology shapes explicit orientations about social groups, and not the other way around. However, this need not be the case. Disliked or disadvantaged groups, notably Blacks, gays, and lesbians, are more likely to identify as liberal than conservative. This may be a sign that attitudes about some social groups shape ideological commitments as well. For members of disliked groups, maintaining a positive view of one’s group may favor an ideological belief that the status quo should be replaced with one in which “my group” is better valued (Jost & Thompson, 2000). If this is true, then as social acceptance of a group increases, then so should the conservatism of members of that group insofar as the group-serving motivation to take on a liberal position for change declines. This is another hypothesis for future study. Intergroup Attitudes 16 Concluding Remarks The fundamental ideological distinction of liberalism-conservatism shapes how individuals orient themselves toward the social world. Preferences for others as members of social groups are markers of such orientations, and both explicit and implicit measures provided the evidence from which we draw several conclusions. First, liberals both self-report and reveal on implicit measures greater favorability toward groups that are socially disadvantaged than do conservatives. Political ideology also affects the strength of the connection between implicit and explicit social attitudes, but the cause of this relationship is inconclusive. The data might have revealed a difference between liberals and conservatives on self-report measures but not on measures of automatic preference. That was not what we observed; liberals both deliberately report and automatically reveal less of a preference for the socially privileged than do conservatives. Those who maintain that political conservatism is not linked to differential preferences for advantaged versus disadvantaged groups are obliged to rethink their position based on such data. And those who believe that liberals are without social preferences or biases are also out of step with what the data show. The automatic preferences of liberals are discrepant from their selfreported attitudes. In this sense, liberals possess a more “conflicted stance.” It appears that their explicit egalitarian ideals have not been fully internalized or automatized. Finally, the most provocative comment we can offer from the evidence concerns the role that liberals play in shaping public opinion about social groups. Using race attitudes as the case in point, we found that the position arrived at by liberals in the 1970’s is the position of conservatives today. From this we put forward the hypothesis, to be assessed in future investigations of other social groups, that liberals lead the way in changing social beliefs and attitudes. In the 19 century, it was the liberals and progressives who first opposed slavery. Decades after the emancipation proclamation, conservatives agree. The 21 century may be witness to a similar social change in attitudes toward gays and lesbians. In our own data, liberals reveal much greater positivity than conservatives do toward gays and lesbians both implicit and explicitly. These ideological differences may not persist. Future generations of conservatives, we predict, will come to mirror today’s liberals in attitudes concerning sexual orientation. The larger point may be a simple one of the empirical evidence backing up a dictionary definition. In the area of attitudes toward social groups, conservatives embody the definition of the term conservative: “favoring traditional views and values; tending to oppose change.” And liberals, explicitly at least, embody liberal: “Not limited to or by established, traditional, orthodox, or authoritarian attitudes, views, or dogmas; free from bigotry; Favoring proposals for reform, open to new ideas for progress, and tolerant of the ideas and behavior of others; broad-minded” (The American Heritage Dictionary of the English Language, 2000). With respect to the preferences that we feel and reveal concerning members of advantaged and disadvantaged groups, liberals do what Intergroup Attitudes 17 by definition they are assumed to do. What we suggest is that conservativeseither do not or cannot remain where they are forever, but are bound to followliberals, however grudgingly, perhaps even without awareness of following at all,because liberals are agents of social change. ReferencesAdorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950).The authoritarian personality. New York: Harper. Altemeyer, B. (1996). The authoritarian specter. Cambridge, MA: HarvardUniversity Press. American Heritage Dictionary of the English Language (2000). Fourth Edition.Bar-Anan, Y., Nosek, B. A., & Vianello, M. (2007). The sorting paired featurestask: A measure of association strengths. Unpublished manuscript. Bargh, J. A. (1997). The automaticity of everyday life. In Wyer, R. S. Jr. (Ed).Advances in Social Cognition Vol.10, 1-61.Bishop, G. (2005). The illusion ofpublic opinion. Lanham, MD: Rowman & Littlefield. Conrey, F. R., Sherman, J. W., Gawronski, B., Hugenberg, K., & Groom, C. J.(2005). Separating multiple processes in implicit social cognition: The Quadmodel of implicit task performance. Journal of Personality and SocialPsychology, 89, 469-487. Converse, P. E. (1964). The nature of belief systems in mass publics. In D.Apter (Ed.), Ideology and discontent (pp. 206-226). New York: Free Press.Cunningham, W. A., Nezlek, J. B., & Banaji, M. R. (2004). Implicit and explicitethnocentrism: Revisiting the ideologies of prejudice. Personality and SocialPsychology Bulletin, 30, 1332-1346. Devine, P. G., Plant, E. A., Amodio, D. M., Harmon-Jones, E., & Vance, S.(2002). The regulation of explicit and implicit racial bias: The role ofmotivations to respond without prejudice. Journal of Personality and SocialPsychology, 82, 835-848. Doty, R.M., Peterson, B.E., & Winter, D.G. (1991). Threat and authoritarianismin the United States, 1978-1987. Journal of Personality and SocialPsychology, 61, 629-640.Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). On theautomatic activation of attitudes. Journal of Personality and SocialPsychology, 50, 229-238. Ferguson, M., Carter, T., & Hassin, R. (this volume). On the automaticity ofnationalist ideology: The case of the USA.Graham, J., Haidt, J., & Nosek, B. A. (2007). Understanding the partisan mind:Liberals and conservatives rely on different moral foundations. Unpublishedmanuscript. Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102, 4-27.Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring Intergroup Attitudes18 individual differences in implicit cognition: The Implicit Association Test.Journal of Personality and Social Psychology, 74, 1464-1480. Greenwald, A.G., Poehlman, T.A., Uhlmann, E.L., & Banaji, M.R. (in press).Understanding and using the Implicit Association Test: III. Meta-analysis ofpredictive validity. Journal of Personality and Social Psychology. Haidt, J. & Graham, J. (2007). When morality opposes justice: Conservativeshave moral intuitions that liberals may not recognize. Social JusticeResearch, 98-116. Hewstone, M., Rubin, M., & Willis, H. (2002). Intergroup bias. Annual Review ofPsychology, 53, 575-604. Huntington, S. (1957). Conservatism as an ideology. American PoliticalScience Review, 51, 454-473.Jost, J. T. (2006). The end of the end ofideology. American Psychologist, 61, 651-670. Jost, J. T., & Banaji, M. R. (1994). The role of stereotyping in system-justificationand the production of false consciousness. British Journal of SocialPsychology, 33, 1-27. Jost, J. T., Banaji, M.R., & Nosek, B.A. (2004). A decade of system justificationtheory: Accumulated evidence of conscious and unconscious bolstering of thestatus quo. Political Psychology, 25, 881-919. Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F.J. (2003). Politicalconservatism as motivated social cognition. Psychological Bulletin, 129, 339-375. Jost, J. T., Nosek, B. A., & Gosling, S. D. (in press). Ideology: Its resurgence insocial, personality, and political psychology. Perspectives on PsychologicalScience. Jost, J.T., & Thompson, E.P. (2000). Group-based dominance and opposition toequality as independent predictors of self-esteem, ethnocentrism, and socialpolicy attitudes among African Americans and European Americans. Journalof Experimental Social Psychology, 36, 209-232. Logan, G. D. (1988). Toward an instance theory of automization. PsychologicalReview, 95, 492-527. McCann, S.J.H. (1997). Threatening times, “strong” presidential popular votewinners, and the victory margin, 1824-1964. Journal of Personality andSocial Psychology, 73, 160-170.Mitchell, J. P., Macrae, C. N., & Banaji, M. R. (2006). Dissociable medialprefrontal contributions to judgments of similar and dissimilar others. Neuron50, 655-663. Moskowitz, G. B., Gollwitzer, P. M., Wasel, W., & Schaal, B. (1999).Preconscious control of stereotype activation through chronic egalitariangoals. Journal of Personality and Social Psychology, 77, 167-184. Nosek, B. A. (2005). Moderators of the relationship between implicit and explicitevaluation. Journal of Experimental Psychology: General, 134, 565-584. Nosek, B. A. (2007). Implicit-explicit relations. Current Directions inPsychological Science, 16, 65-69. Nosek, B. A., & Banaji, M. R. (2001). The go/no-go association task. SocialCognition, 19, 625-666. Intergroup Attitudes19 Nosek, B. A., Greenwald, A. G., & Banaji, M. R. (2007). The Implicit AssociationTest at age 7: A methodological and conceptual review. In J. A. Bargh (Ed.), Automatic processes in social thinking and behavior. Psychology Press.Nosek, B. A., & Smyth, F. L. (2007). A multitrait-multimethod validation of theImplicit Association Test: Implicit and explicit attitudes are related but distinctconstructs. Experimental Psychology, 54, 14-29. Nosek, B. A., Smyth, F. L., Hansen, J. J., Devos, T., Lindner, N. M., Ranganath,K. A., Smith, C. T., Olson, K. R., Chugh, D., Greenwald, A. G., & Banaji, M. R.(2007). Pervasiveness and correlates of implicit attitudes and stereotypes. European Review of Social Psychology, 18, 36-88.Payne, B.K., Cheng, C. M., Govorun, O., & Stewart, B. (2005). An inkblot forattitudes: Affect misattribution as implicit measurement. Journal of Personalityand Social Psychology, 89, 277-293. Rothermund, K., & Wentura, D. (2004). Underlying processes in the ImplicitAssociation Test: Dissociating salience from associations. Journal ofExperimental Psychology: General, 133, 139-165. Sales, S.M. (1972). Economic threat as a determinant of conversion rates inauthoritarian and nonauthoritarian churches. Journal of Personality and Social Psychology, 23, 420-428.Sales, S.M. (1973). Threat as a factor in authoritarianism: An analysis ofarchival data. Journal of Personality and Social Psychology, 28, 44-57.Sidanius, J. and Pratto, F. (1999). Social dominance. Cambridge: CambridgeUniversity Press. Sniderman, P.M., Crosby, G.C., & Howell, W.G. (2000). The politics of race. InD.O. Sears, J. Sidanius, & L. Bobo (Eds.), Racialized politics: The debateabout racism in America (pp. 236-279). Chicago: University of Chicago Press.Tajfel, H. (1978). Social categorization, social identity and social comparison. InH. Tajfel (Ed.), Differentiation between social groups: Studies in the social psychology of intergroup relations(pp. 61–76). London: Academic Press.Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroupbehavior. In S. Worchel & W. G. Austin (Eds.), The Psychology of IntergroupRelations (pp. 7-24). Chicago: Nelson-Hall. The American National Election Studies (www.electionstudies.org). The 1948-2004 ANES Cumulative Data File [dataset]. Stanford University and theUniversity of Michigan [producers and distributors], 2005. Ullrich, J. & Cohrs, J. C. (2007). Terrorism salience increases systemjustification: Experimental evidence. Social Justice Research, 20, 117-139. Intergroup Attitudes20 Table 1. Mean warmth rating (0 cold 97 warm) and ideology-attitude relationfrom ANES data for 14 social groups. Positive d’s indicate that conservatismwas positively associated with liking for the group; negative d’s indicate thatconservatism was negatively associated with liking for the group. Target GroupYears MeasuredTotal NMean