Intergroup Consensus/Disagreement in Support of Group Based Hierarchy: An Examination of Socio-Structural and Psycho-Cultural Factors

Abstract

A meta-analysis examined the extent to which socio-structural and psycho-cultural characteristics of societies correspond with how much gender and ethnic/racial groups differ on their support of group-based hierarchy. Robustly, women opposed group-based hierarchy more than men did and members of lower-power ethnic/racial groups opposed group-based hierarchy more than members of higher-power ethnic/racial groups. As predicted by social dominance theory, gender differences were larger, more stable, and less variable from sample to sample than differences between ethnic/racial groups. Subordinate gender and ethnic/racial group members disagreed more with dominants in their views of group-based hierarchy in societies that can be considered more liberal and modern (e.g., emphasizing individualism and change from traditions), as well as in societies that enjoyed greater gender equality. The relations between gender and ethnic/racial groups are discussed and implications are developed for social dominance theory, social role theory and biosocial theory, social identity theory, system justification theory, realistic group conflict theory and relative deprivation theory.

The biggest political struggles of the last several centuries have hinged on the question of whether social inequality occurs with the consent of the governed or in the face of their dissent. On one hand, group-based hierarchies are the most common form of stable society (Sidanius & Pratto, 1999). For example, certain nationalities, races, ethnicities, religious groups, tribes, or political parties have more prestige, more material resources, and more political influence—in a word, more power, across diverse societies (Flanagan, 1989). On the other hand, every movement led by oppressed groups against dominant groups, from the women’s suffrage movement worldwide, to the Russian revolution, to South Africa’s anti-apartheid movement to the Salvadoran Civil War, took place in societies in which the superiority of certain social groups was institutionalized and legitimized by potent cultural ideologies. This broad historical record leaves it unclear whether people generally accept group-based hierarchy or what social conditions lead them to reject it. Despite considerable social theory and empirical research, the social sciences have no general answer as to whether or when people in less powerful groups concur or disagree with people in powerful groups in their approval of group-based hierarchy. The answers to these questions are pertinent to the psychology and dynamics of intergroup relations, social justice, and equality, among other topics.

Across societies, members of more powerful groups (i.e., dominants), clearly have more interest in group-based hierarchy and in people’s psychological acceptance of such hierarchy than those in less powerful groups (i.e., subordinates). Thus, based on group interest alone, one might expect dominants to endorse group-based hierarchy more than subordinates. In contrast, a good number of psychological, social, and political theories argue that dominants use their power to convince subordinates of their views of group-based hierarchy, so it is possible that dominants and subordinates would share similar attitudes towards group-based hierarchy. In addition to these theoretical disagreements over whether subordinates should be expected to differ reliably from dominants in their attitudes about group-based hierarchy, another body of theories makes conditional predictions about the social, cultural, or political factors that should contribute to greater intergroup consensus or intergroup disagreement. The present study not only examined the extent to which dominants and subordinates concur or disagree in their endorsement or rejection of group-based hierarchy, but also examined what socio-structural and psycho-cultural conditions were associated with more intergroup consensus or disagreement. To that end, we conducted a meta-analysis of studies on a well-validated and culturally-general scale that assesses endorsement of group-based hierarchy. The social dominance orientation scale (Pratto, Sidanius, Stallworth, & Malle, 1994) is particularly useful because it does not refer to specific groups by name, is cross-culturally valid (e.g., Pratto et al., 2000), and is used in many nations and in varied languages (e.g., Roccato & Ricolfi, 2005; Sibley, Robertson, & Wilson, 2006).

We begin by summarizing cultural, political, and social psychological theories concerning processes that contribute to consensus between dominants and subordinates. Overall, these processes make group-based hierarchy seem culturally and psychologically normative. We then summarize theories of processes that predict that subordinates should reject group-based hierarchy more than dominants do and consider what particular forms of intergroup relations, such as gender and race, should exhibit these patterns. These reviews lead us to postulate that, when comparing different societies and types of groups, particular social-structural and psycho-cultural variables should moderate the size of group differences in support of group-based hierarchy. Although the archival data in our meta-analysis cannot show direct evidence of causal processes, they can test whether the data are consistent with several causal theories concerning when subordinates will differ from dominants. Our findings test alternative hypotheses garnered from several theoretical frameworks about the conditions that lead people to support group-based hierarchy, and point to new directions for research on intergroup relations.

Consensus or Disagreement between Dominants and Subordinates?

Theories in psychology, political science, and sociology concerning intergroup relations offer reasons for both why dominants and subordinates may concur in their level of support for group-based hierarchy, and why they may differ. We organize our review of such theories by summarizing three general processes that could lead either to intergroup consensus or to intergroup disagreement regarding the rightness of group-based hierarchy: (a) social and psychological adjustment to conditions of structural intergroup inequality, including social learning, (b) ways that people reason about and derive meaning about the social world, including their expectations and evaluation standards, and (c) how culture influences people’s conception of self in relation to others and adherence to history and tradition. As our theoretical review will show, several different theories can be subsumed under each process. In addition, this close comparison of theories shows that some theories predict both intergroup consensus and intergroup disagreement, under different conditions.

Dominants and Subordinates May Converge in Supporting Group-Based Hierarchy

The first process we examine concerns the idea that people generally must adjust to the social conditions of their cultures and societies. Learning the cultural worldviews and shared social beliefs of one’s local society is the primary aspect of psychological adjustment to social reality. The idea that dominants’ worldview becomes predominant in cultures such that even subordinates understand and are influenced by dominants’ worldviews is quite widespread across theories, but originated in Marxism. Marx and Engels’s (1846) treatise on ideology argued that dominants rule in part by bringing subordinates to share their own ideological worldviews. Addressing what would later be understood as the field of psychology, Marxist theory argues that such worldviews make hierarchical social arrangements such as class systems, sexism, racism, and colonization seem reasonable and necessary. The more general idea that dominants create ideological consensus in support of group-based hierarchy through cultural socialization is now widely held in the social sciences, including in sociological analyses of ideology and intergroup relations (e.g., Jackman, 1994), the concept of legitimizing myths in anthropology (e.g., W. R. Johnson, 1994; Sanday, 1981) and social psychology (e.g., Jost & Major, 2001; Sidanius & Pratto, 1999), the concept of elite beliefs influencing mass beliefs in political science (e.g., Converse, 1964; Gramsci, 1971; Jenkins-Smith, Mitchell, & Herron, 2004), and communication studies’ recognition of discourse and the media in creating dominant ideologies (e.g., Blackledge, 2002; Lewig & Dollard, 2001).

The second process we examine concerns people’s reasoning about the social world and how they derive meaning from it. Processes by which people understand and explain their social world may contribute to intergroup consensus on the rightness of group dominance. For example, social role theory argues that people generate gender stereotypes to explain why men and women obtain different roles (Eagly & Steffen, 1984). A number of other attribution approaches make the same argument more generally as one basis for stereotypes (e.g., Crandall, Silvia, N’Gbala, Tsang & Dawson, 2007; Pettigrew, 1979). Stereotypes, then, make sense of observed social patterns (e.g., Hilton & von Hippel, 1996; Lee, Jussim, & McCauley, 2001). Another process that can rely on stereotypes and prejudice to make sense of the social world is system-justification. To maintain the illusion that the social system is just (Lerner & Miller 1978), system-justification theory predicts that subordinates accept stereotypes (Jost & Banaji, 1994) that derogate their own groups (Jost, Pelham, & Carvallo, 2002) and also denigrate other subordinate groups (Jost & Burgess, 2000). The system-justification motivation may lead subordinates to endorse group-based hierarchy at least as much as dominants do.

The third general process that may produce intergroup consensus on group-based hierarchy concerns how culture influences people’s conceptions of self in relation to others. In “traditional” societies with small-scale agriculture and little industry and education, people acquiesce to authorities and feel bound to uphold traditions and prescribed roles than in more industrialized, “modern” societies (Schofer & Fourgade-Gourinchas, 2001). In particular, people in traditional societies tend to define themselves as extensions of other people, and view their interests as tied to those of others (Markus & Kitayama, 1991; Triandis, 1990). Thus, in traditional societies, people are expected to conform more to dominant norms than in societies with more education and more economic and social mobility. In fact, people in collectivist societies tend to uphold hierarchy (Smith, Peterson, & Schwartz, 2002) and to be higher on authoritarianism than people in individualistic societies (Kemmelmeier et al., 2003). Thus, dominants and subordinates may have more consensus on group-based hierarchy in more collectivistic societies.

Dominants’ and Subordinates’ Views May Diverge regarding Group-Based Hierarchy

The same three kinds of general processes, (a) learning about and adjusting to conditions of group-based hierarchy, (b) reasoning and making meaning about the social world, and (c) cultural influences on people’s conception of self in relation to others, can also be used to make predictions about when dominants and subordinates will differ in support of group-based hierarchy. We now delineate the relevant processes and how they contrast with those reviewed above.

In contrast to the idea that everyone in society learns and adopts the viewpoints of dominants, a number of theories focus on the idea that dominants and subordinates adopt different viewpoints regarding group-based hierarchy due to their own group position. The idea that people learn about their societies from the perspective of their own group’s position is highlighted by image theory (e.g., Herrmann, 1985), realistic group conflict theory (e.g., Campbell, 1965), and social dominance theory (e.g., Pratto et al., 1994; Sidanius & Pratto, 1999). Unlike models that hold that dominant groups’ stereotypes are normative (e.g., Cuddy et al., 2007; Devine, 1989), image theory assumes that groups each have their own perspective depending on their relative cultural standing, relative power, and goal compatibility. Indeed, Alexander, Brewer, and Livingston (2005) found that urban White Americans view Black Americans as barbarians or enemies, whereas Black Americans view other Black Americans as allies—a group with compatible goals and equal power and status. Like image theory, realistic group conflict theory presumes that people recognize their own group’s interests, and when those interests diverge from those of other groups, their social beliefs and attitudes will follow (e.g., Campbell, 1965). Indeed, when groups are in competition for desired scarce resources, they are more prejudiced against their competitors (e.g., Bobo, 1983; Brewer & Campbell, 1976; Sherif & Sherif, 1953). Incorporating this strain of realistic group conflict theory, social dominance theory posits that individuals’ psychological orientations toward group-based hierarchy develop from living in group-dominance societies (Pratto, 1999). The group interest and psychological agendas of dominants makes it relatively easy for dominants to endorse group-based hierarchy, whereas the conflict among group interest, psychological agendas, and social reality makes endorsing group-based hierarchy more difficult for subordinates (Pratto, Sidanius, & Levin, 2006). That is, in general, social dominance theory expects subordinates to be lower on social dominance orientation (more opposed to group–based hierarchy) than dominants. The group-interested assumption of image theory, realistic group conflict theory, and social dominance theory implies that dominants would tend to endorse group-based hierarchy and subordinates would oppose it.

The second process concerns how people make sense of their social world. Social identity theory (Tajfel & Turner, 1979, 1986) recognized that the motivations and psychological stances concerning the individual self, the collective self or ingroup, and one’s broader social reality (e.g., an unequal society) could create conflicts concerning its central motive, the desire for positive self-regard. For dominants, social identity theory posits that the stereotypes of their societies and identifying with those groups should provide them with positive self-regard (e.g., Houston & Andreopoulou, 2003; Verkuyten, 2007). In contrast, because accepting group inferiority does not fulfill the desire for positive identity for subordinates, social identity theory argues that the subordinates will be “socially creative,” for example, by changing the dimensions of social judgment, rejecting stereotypes and other dominant ideologies in favor of new ways of thinking. In fact, research has shown that subordinates reject their inferior status and concomitant stereotypes (Lalonde, 1992), particularly those subordinates with high self-esteem (Seta & Seta, 1996). Thus, social identity theory implies that subordinates oppose group-based hierarchy more than dominants do, particularly if group-based hierarchy seems unstable or illegitimate (e.g., Bettencourt, Dorr, Charlton, & Hume, 2001; Sidanius, Levin, Federico, & Pratto, 2001; Wright, Taylor, & Moghaddam, 1990).

Another way that cognition and reasoning about one’s social environment might be expected to produce intergroup disagreement is when dominants and subordinates have expectations or comparison standards that lead them to different comfort levels with group-based hierarchy (e.g., Leach, Snider & Iyer, 2002). In societies in which equality is strongly normed or a stated ideal, equality may serve as a comparison standard for subordinates. Using both surveys and experiments, Eibach and colleagues (Eibach & Ehrlinger, 2006; Eibach & Keegan, 2006) showed that U.S. dominants oppose subordinates gaining power because they compare their group’s position to a less-egalitarian past, but that subordinates are dissatisfied with the present level of inequality because they compare the present to an egalitarian ideal. Hence, the norms and ideals that dominants and subordinates hold may generally differ, leading to intergroup disagreement in approval for group-based hierarchy.

More broadly, societies vary in the degree of liberalism, which is used here in the historical sense to refer specifically to one’s freedom in relation to one’s society. Such freedom can occur in interpersonal relationships, in one’s adherence to tradition, and in one’s thoughts and feelings. We argue that more liberal societies should help people critically reflect on their traditions and practices, which should induce several cognitive processes (e.g., rejecting false consciousness or dominants’ worldviews) that will produce larger intergroup disagreement on support of group-based hierarchy. Across nations, Inglehart and Norris (2003) showed that gender differences on beliefs about gender roles are substantially larger in industrial or post-industrial societies than in agricultural societies. In addition, in “liberal” societies that are individualistic and in which equality is valued, research on social comparison has shown that there are often larger psychological differences between groups because in such societies, people compare themselves to other groups (Guimond et al., 2006; Guimond, 2008; Major, 1994). In individualistic societies that emphasize equality, subordinates compare themselves against dominants and feel relatively deprived; as a consequence, they may reject group-based hierarchy, producing large group differences on support for group-based hierarchy. Thus, we expect that in more liberal and individualistic societies, in which the autonomy and personal well-being of the individual are emphasized, subordinates would be more likely to realize that their own interests differ from those of dominants, and they should differ more from dominants regarding group-based hierarchy.

One important social condition that may enable psychological individualism and liberalism is increased societal wealth, especially as economies grow beyond subsistence levels with industrial development. In a survey around the world, Inglehart and Baker (2000) observed that increasing national wealth was associated with increasing endorsement of self-expressive values. In addition, societal wealth may allow for more individual mobility and the spread of information about other groups, which increases intergroup comparison (Guimond et al., 2006), thus increasing intergroup disagreement. The opposite prediction is made by realistic group conflict theory, which suggests that intergroup disagreement should be more pronounced in societies with fewer resources. If greater societal wealth promotes self-expression, and this encourages subordinates to recognize their position as separate from that of dominants, then group differences on support of group-based hierarchy should be larger in wealthier societies. Conversely, if fewer resources promote intergroup disagreement, then group differences on support of group-based hierarchy should be smaller in wealthier societies.

Alternatively, surplus societal wealth may enable the signs and power differentials that people use to create social stratification, and it may result in structural inequality. Structural inequality can influence people’s expectations and invokes certain self-other relations. Several theories make predictions about group differences that depend on the level of structural inequality. As relative deprivation theory originally predicted, greater structural equality and more integrated conditions make the contrast between subordinates and dominants more apparent. Thus, in more structurally equal societies, relative deprivation theory predicts that group differences in endorsement of group-based hierarchy should be more pronounced. The prediction is consistent with previous research evidence showing that subordinates are more likely to use dominants as the reference group or are more likely to reject prescriptive norms that disadvantage their groups in societies with greater structural equality (Guimond et al., 2006; Sequino, 2007). By treating attitudes toward equality differently from other psychological variables, social role theory recently also predicted that men and women would differ more in societies with greater equality. Eagly, Diekman, Johannesen-Schmidt, and Koenig (2004) suggested that women’s political empowerment, combined with their subordinate status to men, should allow women to endorse a more egalitarian view than men. Thus, we propose that greater structural equality will be associated with larger group differences on support of group-based hierarchy.

Special Forms of Intergroup Relations

Next we consider whether the social basis of group distinctions influences how much groups concur in support of group-based hierarchy. Some theories are designed to pertain to specific kinds of groups (e.g., for gender groups, ambivalent sexism theory, Glick & Fiske, 1996, and social role theory, Eagly, 1987; for ethnic/racial groups, ambivalent racism theory, Katz & Hass, 1988, and symbolic racism theory, Kinder & Sears, 1981). Other theories pertaining to a variety of types of social groups include realistic group conflict theory, system justification theory, and relative deprivation theory. In contrast both to theories about one type of group distinction (e.g., sexism) and to theories that pertain regardless of the social basis of groups (e.g., relative deprivation theory), social dominance theory focuses on the similarities and differences among three kinds of social groupings that are not functionally reducible to one another. Drawing on van den Berghe’s (1978) account of societal organization, social dominance theory recognizes that all complex societies contain (a) a gender system, in which men are observed to hold more power than women cross-culturally; (b) an age system, in which adults have more power than children; and (c) an arbitrary-set structure in which at least one socially constructed category such as a particular caste, class, ethnicity, race, religious group, nationality, wields more power than at least one other group (e.g., Sidanius & Pratto, 1999).

A particular point made by contrasting different groupings is that the relative status and power of men are consistently superior to women across cultures with economic surplus, whereas the relative status and power of a given arbitrary-set groups (e.g., Jews) may vary across cultures. As a consequence, gender differences in social dominance orientation should be particularly robust, a phenomenon Sidanius, Pratto, and Bobo (1994) called the invariance hypothesis, which states that, “higher average level of male [versus female] social dominance orientation should be found after cultural, situational, and environmental factors are considered” (p. 1000). The examination of the invariance hypothesis is important because it has implications for (a) the unique nature of gender relations from other group relations and (b) how intergroup consensus or disagreement is sustained. To provide evidence against the invariance hypothesis, researchers need to show a significant interaction between gender and another factor that influences social dominance orientation after controlling for both main effects (see Figure 1 of Sidanius, Pratto, & Bobo, 1994); the other factor should moderate the size of the gender differences and should reduce or reverse the direction of the gender difference. The strong version of the invariance hypothesis states that gender differences will not change in the face of potential moderators, whereas the weak version holds that the gender difference will not be eliminated by them. Rigorous empirical research to date supports the strong version, although some studies show evidence for the weak version. Based on a large random sample of Los Angeles county residents, Sidanius, Pratto and Bobo (1994) tested for but found no such moderation. Similar patterns appeared in Sidanius and Pratto’s (1999) 20 samples across 10 nations, and Sidanius and his colleagues’ (2000) samples from four nations.

Although some researchers (e.g., Caricati, 2007; Dambrun, Duarte, & Guimond, 2004; Foels & Pappas, 2004) have argued that their results refute the invariance hypothesis, their models differed in important respects from Sidanius, Pratto, and Bobo’s (1994) formulation. Specifically, rather than testing for moderation, these researchers examined whether the gender difference is completely mediated by socialization-cultural factors (e.g., Caricati, 2007; Dambrun et al., 2004; Foels & Pappas, 2004), which is not at issue in the invariance hypothesis. Nonetheless, two conditions have been identified to moderate gender difference on social dominance orientation, namely, gender identification and intra-group comparison. Wilson and Liu (2003) showed that by covarying out own-gender identification, gender differences on social dominance orientation disappeared in New Zealand samples. Yet, this result was not replicated by other researchers who tested both moderation and mediation effects using the same measures in samples from other countries (Snellman, Ekehammar, & Akrami, 2009). Snellman and his colleagues found significant mediation but no moderation of gender identification on social dominance orientation. Additionally, Guimond and his colleagues (2006) found a nonsignificant gender difference on social dominance orientation when asking participants to do intragender comparison. A similar finding was reported by Huang and Liu (2005) when participants responded in an ethnicity-salient condition. Interestingly, none of the above studies found a reversed gender difference with women higher on social dominance orientation than men, and such results support the weak version of the invariance hypothesis.

Hypotheses

Our review identifies several contradictory hypotheses concerning whether groups should differ on support of group-based hierarchy, the conditions under which they might differ more, and whether groups of different types, such as gender or arbitrary-set groups, should differ. Part of the utility of this paper is that, by considering a variety of theoretical implications broadly, and comparing data from a variety of societies, it can show where theoretical predictions depend on social conditions.

Hypothesis 1 concerns whether dominants and subordinates differ from one another in support of group-based hierarchy. As we showed, socialization and psychological adjustment can lead people to learn and sometimes adopt the prejudices, stereotypes, ideologies that promote group-based hierarchy, which could produce similarity between groups within the same societies. Yet, people in subordinate groups may experience a social reality that conflicts with such hierarchy-enhancing legitimizing myths, which should generally reduce their support of group-based hierarchy. Thus, there is strong theoretical reason both to predict group differences on social dominance orientation (Hypothesis 1) and to predict similarity between groups (no group difference). For this reason, Hypothesis 1 tests whether dominants are reliably higher on social dominance orientation than subordinates.

Another point on which different theoretical traditions diverge is whether gender is functionally different from other kinds of group distinctions, and what the comparison between gender and arbitrary group distinctions means theoretically. We address this question empirically in two ways. First, following from the invariance hypothesis, we tested whether gender differences are larger, more stable, and less likely to be moderated by other factors than arbitrary-set differences (Hypothesis 2). Second, for all the hypothesis tests concerning factors that could moderate the size of gender and arbitrary-set group differences, we examine whether the same pattern of results would obtain for gender and arbitrary-set groups. Our data and discussion will then inform questions about the particularism of theories regarding gender and arbitrary-set groups.

A number of researchers have hypothesized that group differences will be smaller in more collectivistic and hierarchical societies, whereas liberal societies, which emphasize individualism, independence, and rejection of obligations and traditions, will exhibit larger group differences (e.g., Guimond et al., 2007). Yet, some researchers found evidence that group differences of values and political views may be smaller in liberal societies (Caprara et al., 2006; Schwartz & Rubel, 2005, p. 1022). Here again, then, the literature makes competing predictions concerning whether group differences should be larger or smaller in more liberal societies. Hypothesis 3 states that the size of group differences would be increased by cultural factors that emphasize independence and individualism. We tested this hypothesis by examining the association between the group effect sizes on social dominance orientation with norms for cultural values for each nation gathered by Hofstede (2007) and Schwartz (personal communication).

The literature also offers competing hypotheses concerning the psychological effects of societal wealth. Some researchers postulate that increasing wealth is associated with greater self-expression and therefore, with larger gender differences in support for gender role differentiation (Inglehart & Norris, 2003). Realistic group conflict theory, in contrast, postulates that people should be especially group-biased when resources are scarce. That is, because low societal wealth (poverty) will make people feel threatened and blame other groups, group differences in support of group-based hierarchy should be greater in poorer societies. As such, Hypothesis 4 tests these competing ideas in stating that increasing societal wealth should be associated with larger group differences on social dominance orientation.

Several recent empirical comparisons of social and political attitudes across countries have shown that gender differences are larger in more structurally egalitarian societies (see Sequino, 2007 for a review). In addition, recent relative deprivation theory research suggests that in more egalitarian societies, people compare themselves with members of other groups and, if they are in subordinate groups, should reject group-based hierarchy more (Guimond et al., 2006). Hence, Hypothesis 5 states that greater gender equality, as evidenced by objective indices of women’s conditions, and greater social equality, as indicated by the Gini index, will be associated with larger group differences on social dominance orientation.

Finally, there is an issue concerning whether social dominance orientation is one construct or two. According to Pratto and her colleagues (1994; Sidanius & Pratto, 1999), social dominance orientation is a general orientation in support of group-based dominance and inequality. Initial factor analyses confirmed that its items measure one factor (Pratto et al. 1994), although the reverse-coded items sometimes form a strongly correlated second factor. Such illusory two-factor solutions are common for balanced scales and can be eliminated by correlating errors of items that are coded in the same direction (pro-trait or con-trait; Xin & Chi, 2010). Jost and Thompson (2000) argued that social dominance orientation contains two factors. They termed one factor with most of the con-trait items “opposition to equality” (OEQ), which they suggested operates at the societal level; they termed the remaining items group-based dominance (i.e., GBD), which they suggested operates at the group level. Jost and Thompson showed that the associations between GBD, OEQ, self-esteem, and ethnocentrism differed for U.S. Whites and Blacks. In one other study evaluating this prediction, Guimond, Dambrun, Michinov, and Duarte (2003) tested but did not find any difference of the relations between GBD, OEQ, and attitudes of prejudice and intergroup relations in France. In the current investigation, to evaluate whether social dominance orientation contains two factors operating at two levels, we tested group differences on GBD and OEQ, respectively. If social dominance orientation represents one factor, we expected converging results in group differences of GBD and OEQ. Because of the question Jost and Thompson (2000) raised, we first tested whether social dominance orientation is a unified construct, before testing the hypotheses stated above.

Overview of the Present Study

In this research synthesis, we investigated whether dominants and subordinates concur or disagree in their support of group-based hierarchy, and what factors are associated with more or less disagreement between groups. We did so by examining empirical studies on social dominance orientation conducted since the construct was first measured, 1979, through September, 2009. In particular, we investigated whether dominants tend to support group-based hierarchy, as measured by social dominance orientation, more strongly than subordinates in both gender and arbitrary-set groups (Hypothesis 1). We derived several predictions from the invariance hypothesis: We expected that the magnitude of gender effect sizes (i.e., d) on social dominance orientation would be larger compared to that of arbitrary group effect sizes, that the variation of gender effect sizes would be smaller than the variation of arbitrary group effect sizes, and that more stability over time in gender effect sizes would be observed than in arbitrary effect sizes (Hypothesis 2). Further, we examined whether psycho–cultural variables reflecting individualism and liberalism (Hypothesis 3), societal wealth (Hypothesis 4) and socio-economic equality (Hypothesis 5) would be associated with larger group differences.

Method

Sample of Studies

Computer-based information searches were conducted using keywords of “social dominance orientation” or “social dominance theory” or “SDO” or “SDT” in PsycINFO, Dissertation Abstracts, Web of Science, and Google Scholar.¹ Citations to Pratto et al. (1994), who first evaluated the psychometric properties of social dominance orientation scales, and Sidanius and Pratto (1999) were also searched and unpublished works that were referenced in other sources were sought from authors. We also made announcements and left fliers at several relevant conferences. The last searches were in September, 2009, and reports available as of that date were eligible for the meta-analysis. Studies were excluded if they (a) manipulated social dominance orientation (without first measuring it); (b) sampled clinical populations; (c) did not provide comparison results of gender or arbitrary-set groups; (d) only provided cross-national comparisons; and (e) reported insufficient information to calculate either the effect sizes or the direction of effect sizes. Authors of the reports were contacted for the missing information.

Variables Coded from Each Study

We coded not only variables of conceptual importance to our hypotheses (e.g., type of group comparison, time) but also a number of other study and sample features that may be alternative explanations to our theoretical predictions. For each study we recorded the following information: (a) year of publication and year of data collection, when year of data collection was not reported, published studies were estimated by the year difference between publication and data collection years in the reported studies (e.g., M Year Difference = 3.87 for the U.S. studies on gender comparisons, see the second row under study characteristics in Table 1); unpublished studies were estimated as being collected two years prior to the date of the manuscript; (b) publication form (journal article, other published document, dissertation or master’s thesis, unpublished data); (c) authors’ discipline (e.g., psychology, sociology, political science, etc.); (d) whether researchers were collaborators with the originators of the social dominance orientation scale, Pratto or Sidanius; (e) location of data collection (country); (f) types of group compared (gender or arbitrary group); (g) types of arbitrary-set groups in comparison (e.g., ethnic/racial, religion, school major, etc.); (h) sampling method (convenience, probability, or other); (i) the reliability, scaling, and number of items for the social dominance orientation scale; (j) the language of the social dominance orientation scale; (k) number of participants in each sample; (l) proportion of males in each sample; (m) participants’ mean age (estimated when not provided); (n) participant characteristics (children, adolescents, college students, adults, or general public); (o) percentage of samples with college education; and (p) racial or ethnic percentage for each sample (for Whites, Blacks, Asians, Latinos, and other, estimated by national racial proportions when not provided). Studies written in English and Chinese were coded twice by two of the six coders (three undergraduate students, two graduate students, and the first author) and inter-coder reliabilities were high, ranging from .64 to 1.00, M κ = .87. Disagreements were resolved by discussion between the coders. Studies written in other languages (Italian, Hebrew, Spanish, German, Dutch, and French) were coded by single coders who were social psychologists and fluent in those languages.

Table 1.

Summary of Characteristics of Studies that Compared Either Gender or Arbitrary-Set Groups.

	Gender groups	Arbitrary-set groups
Variables	k = 169	k = 80
Study characteristics
Mdn data collection year	1992/2001a	1994/2002a
Years between dates of data collection and publication	3.87/5.73b	4.78/3.75b
Published samples	82.2%	72.5%
Authors’ alliance with Sidanius or Pratto	29.0%	15.0%
All students samples	74.6%	77.5%
Number of studies in a report
Single studies	67.3%	52.5%
Two studies	20.8%	27.5%
Three studies or more	11.9%	20.0%
Country where study was conductedc
United States	49.7%	53.8%
Canada	7.1%	5.0%
France	5.9%	5.0%
Sweden	5.9%	1.3%
Taiwan	5.3%	6.3%
Israel	2.4%	5.0%
Italy	1.8%	5.0%
South Africa	1.2%	11.3%
Participants’ characteristics
M age	22.7	22.2
Total male participants	25,081	23,450d
Total female participants	27,745	27,985d
Race
Whites	77.0%	69.1%
Asian	9.5%	16.4%
Latino	3.1%	4.1%
Blacks	3.6%	5.7%
Education-College level	53.2%	46.0%
Scale characteristics
Length of social dominance orientation scale
14-items	27.8%	12.7%
16-items	52.1%	63.5%
Language of social dominance orientation scalee
English	66.3%	65.0%
French	7.1%	5.0%
Chinese	5.9%	6.3%
Swedish	4.7%	0.0%
German	3.6%	2.5%
Italian	1.8%	5.0%
Afrikaans	1.2%	8.8%
Spanish	0.6%	3.8%

^aThe former number is median publication year for studies of Sidanius and Pratto and their collaborators; the latter one is median publication year for studies of non-collaborators.

^bThe former number is reported mean year difference among U.S. studies and the latter one is reported mean year difference among no U.S. countries.

^cCountry names are reported when at least 5% of the samples for one comparison or the other are from those countries.

^dEstimated by a limited set of studies that reported exact numbers of male and female participants.

^eLanguages used for at least 3% of samples are shown in the table. In addition to those listed above, social dominance orientation scales were administered in Albanian, Arabic, Dutch, Flemish, Hebrew, Malaysian, Russian, and Polish.

To account for national differences, psycho-cultural and socioeconomic factors about each nation in the sample were collected and coded. As measures of psycho-cultural factors that could indicate the general liberalism of the socio-political climate, in which people think for themselves and have self-concern, we used Schwarz’s (personal communication) cultural values and Hofstede’s (1998, 2007; Hofstede & Hofstede, 2005) national norms from IBM employees because both sets of measures are diverse in scope, pertain to alternative forms of societal organization, include a large set of countries, and have been rigorously examined by other researchers. There were seven Schwarz’ cultural orientations at the national level: Intellectual autonomy, affective autonomy, embeddedness, egalitarianism, hierarchy, harmony, and mastery. Schwarz classifies the seven orientations with respect to three aspects of human society: (a) with the relations and boundaries between the person and the group (intellectual or affective autonomy vs. embeddedness), (b) emphasis of different types of social structure (egalitarian vs. hierarchical), and (c) the relation between humans and natural resources (harmony vs. mastery). There were five dimensions in Hofstede’s national norms. (a) Individualism-collectivism assesses emphasis on the separate self (individualism) versus on groups (collectivism). (b) The power distance index indicates individuals’ acceptance of unequal power distribution in organizations. (c) Societal masculinity-femininity describes salient values for men in societies. In “masculine” societies men emphasize values such as competitiveness and assertiveness, whereas in “feminine” societies, men value caring or modesty. (d) The uncertainty avoidance index indicates a society’s intolerance of uncertainty and ambiguity, traits often associated with prejudice. (e) Long-term orientation measures work ethics and how one views relationships. Societies with longer-term orientation emphasize persistence (e.g., not expecting immediate benefits) and having relationships as one’s role dictates (e.g., daughter, parents). Societies with shorter-term orientation emphasize immediate benefits and personal merit. To measure the general construct of liberalism, freedom for individuals and to consider themselves as separate from others and their culture’s traditions, we used Schwartz’s embeddedness, intellectual autonomy, affect autonomy, harmony, and mastery norms and Hofstede’s individualism-collectivism, uncertainty avoidance, and long-term orientation norms for each society in the study. Both these sources provide continuous indices on these several cultural parameters for each society. Schwarz and Hofstede only reported nations’ psycho-cultural scores when they remained stable over time.

The socioeconomic factors coded for each nation were of two types: Wealth of the society and particular indicators of gender equality and income equality. Societal wealth was measured by the gross domestic national product (GDP) per capita (Central Intelligence Agency, 2006). Regarding gender equality, we recorded a general gender empowerment measure (GEM), which measures gender equality in economic and political opportunities (United Nations Development Programme, 2002), and specific indicators of gender equality including the percentages of women in parliament and in professions, the proportion of women’s income compared to men’s, men’s and women’s average age at their first marriage, and fertility rate (Central Intelligence Agency, 2006; Jao, Lai, Tsai, & Wang, 2003; United Nations Development Programme, 2002; World Health Organization, 2005). Regarding income equality, we included the Gini coefficient, as well as the proportions of households in the nation that earned the highest and lowest 10% of income (Central Intelligence Agency, 2006). Socioeconomic factors were recorded in the year in which data for most nations were reported. Thus, the year in which socioeconomic factors were collected and the year in which a study was conducted may not match. We could locate no systematic data source that indexes arbitrary-set inequality systematically for nations (i.e., outcome differences for races or ethnicities or tribes or religious sects).

Computation and Analysis of Effect Sizes

Two kinds of effect sizes were computed from the available information in the studies. First, to examine convergent and divergent validity of the social dominance orientation scale, correlations between social dominance orientation and other scales were recorded and converted to Fisher’s Z_r. After computing the averaged Z_rs, they were converted back to correlations (rs) for ease of interpretation (Lipsey & Wilson, 2001). Analyses of r values were restricted to these overall results because exploring moderators of r values was outside the purview of the current study. Second, to provide direct comparisons of dominants to subordinates, the difference of M social dominance orientation scores between a dominant group and a subordinate group was divided by the pooled standard deviation, producing gs. These gs were converted to ds by correcting for sample size bias (i.e., gs overestimation of the population effect size, which occurs especially for small samples; Hedges & Olkin, 1985). Positive ds indicated that dominants (e.g., men, Whites) endorse social dominance orientation more strongly than subordinates (e.g., women, Blacks). When Ms and SDs were unavailable to calculate effect sizes, other information (e.g., t-tests, F-values) was used (e.g., B. T. Johnson, 1993; Lipsey & Wilson, 2001).

The homogeneity of each set of ds was examined to determine whether the studies in each set plausibly shared the same effect size following fixed-effects assumptions. Meta-analyses conventionally use Cochran’s (1954) Q, which has an approximate χ² distribution, but because the values that Q takes are typically highly correlated with the numbers of available studies, it is difficult to use it to compare homogeneity between different sets of studies. Therefore, we instead calculated I², which can range from 0 to 100, with the 0 value indicating no more variability than expected by sampling error, and higher values indicating greater variation than sampling error alone would predict. Confidence intervals for I² that do not include zero indicate rejection of the hypothesis of homogeneity and an inference of heterogeneity (Higgins & Thompson, 2002; Huedo-Medina, Sánchez-Meca, Botella, & Marín-Martínez, 2006), implying that the model evaluated by the mean is too simple to describe the effect sizes correctly. Because they lacked homogeneity, calculation of M d₊s followed conventional random-effects assumptions using the full-information, maximum-likelihood procedure (Lipsey & Wilson, 2001).

Because the ds did vary (lacked homogeneity), moderator analyses on categorical variables (similar to the analysis of variance) and continuous variables (similar to regression models) were conducted in order to account for variability in group difference effect sizes (d₊s), with fixed-effects slopes and random-effects intercepts, using macros for SPSS and Stata (Lipsey & Wilson, 2001). For continuous variables, to account for the varying magnitudes of effect sizes, we first examined each moderator in terms of how it modified effect sizes. Then, we fit models with the predictors entered simultaneously. The only exceptions were when moderators were highly correlated (rs ≥ |0.80|); when resulting in moderators having coefficients larger than 1 (an indication of multicollinearity); or when sensitivity analyses revealed unstable coefficients. In graphical depictions of national-level moderation patterns, we used the random-effects mean for nations with k > 1 and the observed effect size when k = 1.

Results

To provide an interpretative context for the hypothesis tests, we first describe the characteristics of the studies in the sample. Then we report the construct validity of the social dominance orientation scale, compare the degree of difference on social dominance orientation between dominants and subordinates, and finally examine the associations between the group effect sizes with hypothesized moderators.

Characteristics of the Studies

The final data set included 206 samples in 118 independent reports (see Appendix I). Characteristics of the studies are summarized in Table 1. Of the 118 independent reports, 25 reports (21.2%) were unpublished manuscripts; published studies appeared between 1992 and 2009. The majority of the samples were conducted in the discipline of psychology, recruiting convenience samples, and using paper-and-pencil measures. About half of the samples responded in groups. Among the samples, 169 (from 101 independent reports) provided gender comparisons and 80 (from 39 independent reports) provided arbitrary-set groups comparisons (see details regarding specific group comparisons in Table 1). Roughly half of the samples (22 out of 39 reports and 43 out of 80 samples) for the arbitrary-set groups were also included in samples of gender comparisons. Due to the overlapping samples and different classifications for the two major types of group comparisons, we examined them separately. There were 25,081 male and 27,745 female participants across 22 countries or geographic areas (e.g., Palestine) for the gender comparison studies, and 23,450 male and 27,985 female participants across 13 countries or areas for arbitrary-set group comparison studies, including 8 countries or areas for ethnic/racial comparison studies shown in Table 1. In addition to countries or areas shown in Table 1, gender comparisons were conducted in Australia (4.1% of samples), Germany (3.6%), Belgium (3.0%), New Zealand (2.4%), the Netherlands (1.8%), UK (1.2%), China (0.6%), Colombia (0.6%), Lebanon (0.6%), Palestine (0.6%), Russia (0.6%), Malaysia (0.6%), Albania (0.6%), and Tunisian Republic (0.6%), and arbitrary-set comparisons were conducted in Germany (2.5%), Poland (1.3%), UK (1.3%), Pakistan (1.3%), and India (1.3%). The majority of the participants were Whites, college students, and young adults. Our analyses focus on standardized group differences, which control for the pooled variance and N per sample. This information informs the reader about some of the variability due to group membership (for gender, male or female, and for arbitrary-set groups, dominant or subordinate groups) and due to different nations.

Appendix I.

Gender comparison studies

Study	Method	Country	% of female	Samples (%)	d+	SDO α	SDO item
Aiello, Chirmbolo, Leone, & Pratto (2004)	CS; FFS	Italy	49.5	ColS; MA: 23	0.38	0.88	16
Akrami & Ekehammar (2006)		Sweden	54.2	HS; ColS; MA: 23.2	0.41	0.86	16
Aquino, Stewart, & Reed II (2005) sample 1	CS; MG	US	54.1	ColS; 89 Wh; 8.4 Bl; 3 As	0.47	0.90	16
sample 2	CS; FFS; MG	US	45.0	ColS; 91 Wh; 6.9 Bl; 2 As	0.47	0.90	16
Bäckström, & Björklund (2007) sample 1	CS; MG	Sweden	59.2	HSS; MA: 18	0.63	0.87	16
Bern (n.d.)	CS; FFS	Sweden	59.2	HSS	0.63	0.87	16
Bern, & Heaven (2001)	SRS; FFS; MI	AUS	52.5	GP	0.21	0.80	14
Bruess (2004) sample 1	RS	German	49.8	HSS; MA: 17	0.26
Bruess (2004) sample 2	RS	German	50.0	HSS; MA: 17	0.39
Bruess (2004) sample 3	RS	German	50.0	HSS; MA: 17	0.31
Cameron, & Nickerson (2009)	CS; FFS; MI	Canada	42.6	GP; MA: 21	0.38	0.85	14
Capps (2002)	CS; FFS; MG	US	58.1	ColS; 78.3 Wh; 2.9 Bl; 5 As; 18.4 La; 2 OEG	0.39	0.89	14
Caricati (2007)	CS; FFS; MG	Italy	50.6	ColS; MA: 23	0.33	0.88	16
Chatard, Faniko, et al. (2006)	RS; FFS	Albania	60.6	ColS; MA: 20.8	0.38	0.57	44
Chatard, Lorenzi-Cioldi, et al. (2005) sample 1	CS; FFS	France	46.7	HSS; MA: 14.1	0.41	0.77	10
sample 2	CS; FFS	France	69.2	HSS; MA: 16.4	0.68	0.66	8
sample 3	CS; FFS	France	57.3	HSS; MA: 14.3	0.00	0.75	10
sample 4	CS; FFS	Tunisian	69.6	HSS	0.65	0.67	10
Chazal, & Guimond (2003)	CS; FFS; MG	France	49.6	HSS; MA: 15.5	0.71	0.80	10
Choudhury, & Pratto (1996)	RS; CS (p); MI	US	33.3	MA: 26.7; 64.8 Wh; 3.6 Bl; 19 As; 7 La; 5 OEG; 3 ColG; 95 GSG	0.52	0.80	8
Christopher, & Mull (2006)	RS; MS; MI	US	61.5	GP; MA: 54.8; 84.6 Wh; 6.1 Bl; 4 As	0.34	0.86	14
Christopher, & Wojda (2008)	CS; MS; MI	Canada	52.2	MA: 39.6	0.47	0.89	14
Christopher, Zabel, Jones, & Marek (2008)	CS; MS; MI	Canada	52.0	MA: 42.4	−0.02	0.76	16
Cohrs, Kielmann, & Maes (2005)	CS; MI	German	52.4	GP; MA: 30.7	0.39	0.90	16
Cohrs, Moschner, Maes, & Kielmann (2005)	CS; MI	German	46.9	MA: 30.5	0.28	0.89	13
Crownover (2007)	CS; CS (p); MS; MI	US	59.1	MA: 37.3; 66.9 Wh; 3.5 Bl; 4 As; 2.3 La; 12 OEG; 5 HSG; 69 ColG; 17 GSG; 7 OEdu	0.38	0.87	16
Crownson, DeBacker, & Thoma (2005)	CS; FFS; MG	US	80.2	ColS; MA: 25.6; 76 Wh; 6 Bl; 4 As; 4 La; 10 OEG	0.00	0.92	16
Crowson (2009) sample 1	CS; MS; MI	US	51.1	ColS; GP; MA: 35.9; 72 Wh; Bl: 9 ; 13 OEG; 9 HSG; 52 ColG; 31 GSG; 7 OEdu	0.47	0.92	16
sample 2	CS; MS; MI	US	50.3	ColS; GP; MA: 37.1; 72 Wh; 13 Bl; 10 OEG; 6 HSG; 52 ColG; 34 GSG; 5 OEdu	0.45	0.89	12
Dambrun, Duarte, & Guimond (2004)	CS	France	49.7	ColS; MA: 20	0.41	0.86	16
Dambrun, Guimond et al. (2002) sample 1	CS (p); FFS; MI	France	75.6	ColS; MA: 22.6	−0.18	0.81	10
sample 2	CS (p); FFS; MI	France	48.3	ColS; MA: 22.6	0.95	0.81	10
Dambrun, Kamiejski,	CS (p); FFS;	France	19.0	ColS; MA: 20.7	0.45	0.88	6
Haddadi, & Duarte (2009)	MG
Di Meo (2006)	CS; FFS; MI	US	63.8	56 Wh; 14 Bl; 7 As; 10 La; 12 OEG; 6 HSG; 45 ColG; 30 GSG; 14 OEdu	0.41	0.91	16
Dickins, & Sergeant (2008)	CS; MS; MI	UK	50.0	MA: 23.1	0.44	0.90	14
Duriez, Soenens, et al. (2008) sample 1	CS; FFS; MI	Belgium	53.2		0.12		14
sample 2	CS (p); MS; MI	Belgium	48.8	HSS; MA: 14.9	0.44	0.85	14
Duriez, Soenens, Vansteenkiste (2007)	CS; FFS	Belgium	58.7	HSS; MA: 17.6	0.49	0.86	14
Edwards (2008)	CS; MS; MI	US	41.0	ColS; 88.9 Wh; 8.3 Bl; 3 OEG	0.36	0.75	16
Feather, & McKee (2008)	CS; FFS; MI	AUS	71.4	ColS; MA: 22.1	0.73	0.85	14
Foels, & Pappas (2004) sample 1	CS; FFS; MG	US	50.0	ColS; 100 Wh	0.41	0.92	16
sample 2	CS; FFS; MG	US	39.8	ColS; 100 Wh	0.42	0.88	16
sample 3	CS; FFS; MG	US	34.6	ColS; 100 OEG	0.47	0.92	16
sample 4	CS; FFS; MG	US	51.1	ColS; 100 OEG	0.41	0.88	16
Freeman, Aquino, et al. (2008) sample 1	CS; FFS; MI	US	27.4	ColS; MA: 20.4; 100 Wh	0.45		16
sample 2	CS; MS; FFS MI;	US	55.8	ColS; MA: 20.2; 100 Wh	0.61		16
Guimond et al., in press sample 1	CS; FFS; MG; MI	UK	59.2	ColS; MA: 19.5	0.39	0.88	16
sample 2	CS; FFS; MG	France	60.1	HSS; MA: 15.8	0.35	0.75	10
Guimond, Branscombe, et al. (2007) sample 1	CS; FFS; MI	Belgium	58.1	ColS; MA: 20.5	0.00	0.74	10
sample 2	CS; FFS; MI	Dutch	61.4	ColS; MA: 20.5	0.16	0.85	10
sample 3	CS; FFS; MI	Malaysia	50.7	ColS; MA: 20.5	0.23	0.03	10
sample 4	CS; FFS; MI	France	49.4	ColS; MA: 20.5	0.55	0.88	10
sample 5	CS; FFS; MI	US	45.5	ColS; MA: 20.5	0.66	0.85	10
Heaven (1999)	CS; FFS; MG	AUS	70.2	ColS	0.63	0.83	14
Heaven, & Bucci (2001)	CS; FFS	AUS	67.4	ColS	0.39	0.90	14
Heaven, St-Quintin (2003)	CS; FFS MI	AUS	78.7	ColS	0.46	0.86	14
Howard (2003)	CS; MI	US	50.0	MA: 34.8; 61 Wh; 16.5 Bl; 8 As; 5.5 La; 9 OEG	0.61	0.94	16
Huang, & Liu (2005) sample 1	CS; FFS	Taiwan	38.2	ColS; MA: 29; 100 As	0.85	0.81	14
sample 2	CS; FFS	Taiwan	53.2	ColS; MA: 29; 100 As	0.20	0.81	14
sample 3	CS; FFS	Taiwan	52.6	ColS; MA: 29; 100 As	0.29	0.81	14
sample 4	CS; FFS	Taiwan	49.2	ColS; MA: 29; 100 As	0.20	0.86	14
sample 5	CS; FFS	Taiwan	48.5	ColS; MA: 29; 100 As	−0.16	0.86	14
sample 6	CS; FFS	Taiwan	54.1	ColS; MA: 29; 100 As	0.04	0.86	14
Islam (2005)	CS (p); FFS; MG	Colombia	24.3	MA: 30.6; 20 Wh; 4 Bl	0.28	0.73	13
Johnson (2007)	CS; CS (p); MS; FFS; MI	US	59.1	MA: 37.3; 5 HSG; 69 ColG; 17 GSG; 7 OEdu	0.38	0.87	16
Labuhn, Wagner, van Dick, & Christ (2004)	CS	German	51.0	HSS; MA: 15	0.26	0.74	4
Lalonde, Fontaine, & Smith (2007): Bls	CS; FFS; MI	Canada	55.8	ColS MA: 23	0.58	0.80	14
Lalonde, Fontaine, & Smith (2007): Whs	CS; FFS; MI	Canada	69.0	ColS MA: 19	0.48	0.89	14
Lambert, & Raichle (2000)	CS; FFS	US	66.2	ColS	0.43	0.92	16
Larsen (1997)	CS (p); FFS; MI	US	55.6	GP; 100 Wh	0.18	0.83	14
Lee (2006)	CS; FFS; MG	US	57.6	ColS; MA: 19.7; 74.2 Wh; 6.4 Bl; 13 As; 3.2 La; 3 OEG	0.80	0.90	16
Lee (2006)	CS; FFS; MG	US	68.8	ColS; MA: 20; 79.7 Wh; 4.2 Bl;4 As; 4 La; 8 OEG	0.36	0.91	16
Lee, Pratto, & Li (2007) sample 1	CS; FFS; MG	Taiwan	61.9	ColS; MA: 20; 100 As	0.13	0.85	16
sample 2	CS; FFS; MG	Taiwan	63.6	ColS; MA: 19; 100 As	0.39	0.71	16
sample 3	CS; FFS; MG	US	60.6	ColS; MA: 19.8; 81 Wh; 3.2 Bl; 9 As	0.61	0.90	16
Lehmiller, & Schmitt (2008)	CS; CS (p); FFS; MG	US	54.8	ColS; MA: 19.5	0.29	0.91	16
Lippa (1995)	CS; MI	US	64.4	ColS; MA: 18; 40 Wh; 13 Bl:; 25 As; 22 La	0.65	0.67	8
Lippa, & Arad (1999)	CS; FFS; MI	US	62.3	ColS; MA: 18; 42 Wh; 13 Bl; 22 As; 24 La; 12 OEG	0.47	0.79	8
Lyall, & Thorsteinsson (2007)	CS; MS; MI	AUS	60.9	GP; MA: 39.6; 45 ColG	0.43	0.88	14
McDermott, & Cowden (2001)	CS; FFS; MI	US	40.0	ColS	0.41	0.90	17
McGuigan, & Scholl (2007)	CS; MS	US	71.9	GP; MA: 25.8; 96 Wh	0.97	0.88	10
McIntyre, Barrett, McDermott, Johnson, Cowden, & Rosen (2007)	CS; FFS; MG	US	43.2	ColS; MA: 22; 60 Christian; Bl: 11 ; As: 20 ; La: 3 ; OEG: 6	0.40
Meyer (2004) sample 1	CS; FFS; MG	South Africa	60.2	HSS; GP; MA: 25.8; 32 Wh; 33 Bl; 35 OEG	0.30	0.83	16
Meyer (2004) sample 2	CS; FFS	South Africa	53.4	GP; MA: 24.6; 32.3 Wh; 35.1 Bl; 33 OEG	0.29	0.86	16
Nagoshi, Terrell, & Nagoshi (2007) sample 1	CS; CS (p); FFS; MG	US	62.6	ColS; MA: 19; 73 Wh; 3 Bl; 10 As; 8 La; 6 OEG	0.86	0.84	20
sample 2	CS; CS (p); FFS; MG	US	76.7	ColS; MA: 19; 73 Wh; 3 Bl; 10 As; 8 La; 6 OEG	1.24	0.84	20
Nelson, & Milburn (1999)	CS; FFS; MG	US	60.4	ColS; MA: 18.7; 84 Wh; 6 Bl; 10 As; 6 OEG	0.51	0.83	14
Nicol, Charbonneau, & Boies (2007) sample 1	CS (p); MS; FFS; MI or MG	Canada	29.0	Canadian Forces applicants & ColS; MA: 19.3	0.38	0.81	20
sample 2	CS(p); FFS; MI or MG	Canada	30.9	Officer cadets & ColS; MA: 20	0.57	0.88	20
O’Keefe (2006)	SRS (p); MS; MI	Canada	15.5	MA: 36.3; 53 HSG; 37 ColG	0.42		16
Pena, & Sidanius (2002)	SRS; PS; Int; MI	US	57.4	GP; 67.9 Wh; 32.1 La	0.16		6
Phelan, & Basow (2007)	CS; FFS; MI	US	57.1	ColS; MA: 18.8	0.24	0.91	16
Poteat, Espelage, & Green (2007)	CS; FFS; MG	US	50.7	HSS; MA: 14.4; 61 Wh; 1.9 Bl; 24 As; 1.9 La; 4 OEG	0.70	0.92	16
Pratto (1994)	CS; FFS; MG	US	58.1	ColS	0.67	0.89	16
Pratto (1995)	CS; FFS; MG	US	59.5	ColS	0.62	0.92	16
Pratto, & Hegarty (2000) sample 1	CS; FFS; MG	US	53.0	ColS; 38 Wh; 4 Bl; 40 As; 8 La; 1 OEG	0.36	0.83	14
sample 2	CS; FFS; MG	US	43.1	ColS; 17 Wh; 1 Bl; 62 As; 13 La; 7 OEG	0.52	0.82	14
sample 3	CS; FFS; MG	US	61.3	ColS; Wh: 56 ; Bl: 1 ; As: 10 ; La: 10 ; OEG: 23	0.51	0.92	16
sample 4	CS; FFS; MG	US	42.6	ColS; 17 Wh; 1 Bl; 62 As; 13 La; 7 OEG	0.00	0.78	14
Pratto, & Lemieux (2001)		US	48.3	GP; 85.2 Wh; 6 Bl; 9 As; 3.1 La; 2 OEG	0.48	0.72	6
Pratto, Lemieux, Glasford, & Henry (2003) sample 1	CS; FFS; MG	US	78.4	ColS; 83.4 Wh; 2.8 Bl; 5 As; 2.8 La:; 6 OEG	0.41	0.87	16
sample 2	RS; FFS	Lebanan	45.7	ColS; Wh: 4	0.35	0.83	16
Pratto, Liu, Levin, et al., (2000) sample 1	CS; FFS; MI	Canada	50.8	GP; 37.3 Wh; 1.5 Bl; 6 As; 1.5 La	0.47	0.85	14
sample 2	CS; FFS; MI	Canada	43.5	GP; 35.5 Wh; 1.1 Bl; 61 As; 2.2 La	0.47	0.91	16
sample 3	CS; FFS	Taiwan	58.7	GP; 100 As	−0.15	0.76	14
sample 4	CS; FFS	Israel	55.9	ColS; Other; 17 Wh	0.04	0.81	16
sample 5	CS; FFS; MG	Israel	59.1	ColS	0.41	0.86	16
sample 6	CS; FFS; MG	China	31.8	ColS; MA: 21; 100 As	0.39	0.66	16
Pratto, Sidanius, Stallworth, & Malle (1994) sample 1	CS; FFS	US	50.0	ColS; Wh: 48 ; 15 Bl; 23 As; La: 13 ; OEG: 1	0.60	0.85	14
sample 2	CS; FFS	US	45.9	ColS; 67 Wh; 4 Bl; 22 As; 4 La; 3 OEG	0.63	0.83	14
sample 3	CS; FFS; MG	US	40.7	ColS; 50 Wh; 4 Bl; 33 As; 1 La; 3 OEG	0.54	0.83	14
sample 4	CS; FFS	US	49.4	ColS; 38 Wh; 5 Bl; 40 As; 8 La; 9 OEG	0.67	0.83	14
sample 5	CS; FFS	US	53.2	ColS; 38 Wh; 5 Bl; 40 As; 8 La; 9 OEG	0.67	0.84	14
sample 6	CS; FFS	US	50.7	ColS; 53 Wh; 8 Bl; 24 As; 10 La; 5 OEG	0.22	0.81	14
sample 7	CS; FFS	US	30.6	ColS; 59 Wh; 2 Bl; 24 As; 15 La	0.76	0.84	14
sample 8	CS; FFS	US	50.0	ColS; 49 Wh; 6 Bl; 25 As; 10 La; 10 OEG	0.58	0.89	14
sample 9	CS; FFS	US	60.0	ColS; 29 Wh; 2 Bl; 51 As; 14 La; 4 OEG	0.56	0.82	14
sample 10	CS; FFS	US	67.0	ColS; 19 Wh; 10 Bl; 45 As; 17 La; 9 OEG	0.06	0.80	14
Pratto, Stallworth, & Sidanius (1997) sample 1	CS; FFS	US	53.2	ColS	0.40	0.61
sample 2	CS; FFS; MG	US	54.7	GP	0.19	0.72	8
Riley (2006)	CS; FFS; MI	US	76.8	ColS; 70.4 Wh; 20.4 Bl; 9.1 La	0.39	0.85	16
Rowatt, Franklin, & Cotton (2005)	CS; FFS; MI	US	84.2	ColS; MA: 19.6; 67 Wh; Bl: 12 ; 7 As; 10.5 La; 4 OEG	0.43	0.90	16
Russell, & Trigg (2004)	CS; FFS; MG	US	62.1	ColS; MA: 21; 77 Wh; 8 Bl; 6 As; 3.5 La; 5 OEG	0.84	0.90	16
Sabir (2006) sample 1	CS; CS (p); MS; MI	US	28.0	54.2 Wh; 21.5 Bl; 1 As; 21.5 La; 2 OEG	0.18	0.91	16
sample 2	CS; OM	Sweden	54.6	MA: 30.7; 27 HSG; 62 ColG	0.47	0.88	16
Schmitt, Banscombe, & Kappen (2003)	CS; FFS; MG	US	52.1	ColS; MA: 18.7; 90.1 Wh; 4 Bl; 5.9 La	0.52	0.87	10
Shorey, Cowan, & Sullivan (2002) sample 1	CS; FFS; MG	US	57.5	ColS; MA: 26; 100 Wh	0.27	0.91	16
sample 2	CS; FFS; MG	US	48.4	ColS; MA: 26; 100 La	0.56	0.91	8
Sidanius, & Jiu (1992)	CS; FFS; MI	US	39.7	GP; MA: 33	0.48	0.83	10
Sidanius, Levin, Liu, & Pratto (2000) sample 1	CS; FFS	Israel	61.7	ColS; MA: 23	0.35	0.86	16
sample 2	CS; FFS; MG	Israel	64.4	ColS; MA: 20	0.34	0.84	16
sample 3	CS; FFS	Palestine	46.5	ColS; MA: 21	0.35	0.66	8
sample 4	CS; FFS; MG	NZ	71.4	ColS	0.37	0.88	16
Sidanius, Pratto, & Mitchell (1994)	CS; FFS	US	70.8	ColS	0.60	0.84	20
Sidanius, Pratto, & Rabinowitz (1994) sample 1	CS; FFS	US	50.5	ColS; 100 Wh	1.70	0.79	8
sample 2	CS; FFS	US	50.8	ColS; 100 OEG	0.72	0.79	8
sample 3	CS; FFS	US	50.0	ColS; 100 Wh	1.09	0.89	29
sample 4	CS; FFS	US	50.5	ColS; 100 OEG	0.38	0.89	29
Sidanius, Pratto, Brief (1995), sample 1	CS; FFS; MG	AUS	61.4	HSS; MA: 16	0.39	0.80	5
sample 2	CS (p); FFS; MG	Sweden	54.7	HSS MA: 17	0.79	0.85	5
sample 3	CS; FFS; MG	US	44.7	ColS; MA: 19.5; 100 Wh	0.25	0.78	5
sample 4		Russia	58.7	MA: 19; 88 Wh; 34 HSG; 29 ColG; 37 GSG	0.68	0.83	5
Snellman, & Ekehammar (2005) sample 1	CS; FFS; MG	Sweden	48.9	Other	0.06	0.88	16
sample 2	CS; FFS; MG	Sweden	44.6	Other	−0.42	0.88	16
Snellman, Ekehammar et al. (2009) sample 1	CS; FFS; MI	Sweden	55.0	ColS; MA: 25	0.24	0.84	16
sample 2	CS; FFS; MI	Sweden	53.9	ColS; MA: 23.3	0.67	0.84	16
Thornhill, & Fincher (2007)	CS; FFS; MI	US	69.1	ColS; MA: 20.1	0.52	0.88	14
Turner (2003)	CS; Int; FFS; MG	US	50.0	GP; MA: 37.3; 50 Wh; 50 Bl	1.09	0.89	16
Umphress, Simmons, et al. (2008) sample 1	CS; MS; FFS; MI	US	59.5	ColS; MA: 22; 79.7 Wh; 10.1 Bl; 1 Asain; 6.3 La; 3 OEG	0.16	0.92	16
sample 2	CS; MS FFS; MI	US	51.5	ColS; MA: 21; 83.8 Wh; 2.9 Bl; 3 Asain; 8.8 La	0.42	0.89	16
Umphress, Smith-Crowe, Brief, Dietz, & Watkins (2007) sample 1	CS; FFS	US	51.9	ColS; MA: 20.5; 100 Wh	0.77	0.90	16
sample 2	CS; MS; MI	US	50.0	ColS; MA: 23; 70 Wh; 3 Bl; 11 As; 14 La; 2 OEG	0.51	0.89	16
van Hiel, & Duriez (2002) sample 1	CS; FFS	Dutch	50.0	GP; MA: 43;	0.10	0.83	14
sample 2	CS; FFS	Dutch	26.5	MA: 39	0.36	0.88	14
van Hiel, & Mervielde (2002)	CS; FFS; MI	Belgium	40.4	GP; MA: 44; 54 HSG & 41 ColG	0.22	0.83	16
Villano, & Zani (2007)	CS; FFS; MI	Italy	62.8	MA: 31	0.00	0.87	16
Walter, Thorpe, & Kingery (2001)	CS; FFS; MG	US	67.0	ColS; MA: 19.7	0.60	0.89	14
Wang (1999)	CS; FFS; MG	US	61.8	ColS; MA: 26.5	0.68	0.70	16
Werhun (2000)	CS; FFS; MG	Canada	50.0	ColS MA: 19; 78 Wh; 1 Bl; 14 As	0.62	0.87	16
Whitley Jr., Aegisdottir (2000)	CS; FFS; MG	US	51.9	ColS; 91 Wh; 4 Bl; 1 As; 2 La; 2 OEG:	0.63	0.85	16
Whitley, Jr. (1999) sample 1	CS; FFS; MG	US	51.4	ColS; 100 Wh	0.95	0.88	16
sample 2	CS; FFS; MG	US	48.4	ColS; 100 Wh	0.34	0.88	16
Whitley, Lee (2000)	CS; FFS; MG	US	51.3	ColS	0.70	0.86	14
Wilson (2003)	CS; FFS; MG	NZ	49.2	ColS; MA: 19; 100 Wh	0.57	0.88	16
Wilson, & Liu (2003) sample 1	RS; MS; MI	NZ	53.8	GP; 62.8 Wh; 37 OEG	0.44	0.84	16
sample 2	CS; FFS; MG	NZ	59.5	ColS; MA: 22; 93 Wh; OEG: 7	0.59	0.88	16
Worthington, Navarro, Loewy, & Hart (2008)	MS; MI	US	63.2	ColS; 65.3 Wh; 18.1 Bl; 15 As; 4.9 La; 4 OEG; 77 ColG; 8 GSG; 15 OEdu	0.60	0.93	14
Yeageley, Morling, & Nelson (2007)	CS; FFS; MI	Canada	50.0	ColS	1.11		4
Zakrisson (2008)	RS; MS; MI	Sweden	50.9	MA: 47	0.32	0.79	14

RS=Random sample; SRS=Stratified random sample; CS=Convenience sample; (p)=purposive; MG=Measured in group; MI=Measured individually; FFS=Face to face survey; PS=Phone survey; Int=Interview; MS=Mail survey; OSM=Other survey method; OM=Online measure; ColS=College students; HSS=High school students; GP=General public; MA=Mean age; OEG=Other ethnic group; ColG=College graduates; HSG=High school graduates; GSG=Graduate School graduates; OEdu=Other education; Wh=White; Bl=Black; As=Asian; La=Latino

About a half of studies used the 16-item social dominance orientation scale (Pratto et al., 1994), about a quarter of studies administered the 14-item social dominance orientation scale (Pratto et al., 1994), and the remaining studies used other versions of the social dominance orientation scale. Using Rodriguez and Maeda’s (2006) equations to calculate means, the social dominance orientation scale was found to enjoy very good reliability, mean α = .85 (95% CI = .84 to .86, k = 173), which was true for all versions of the social dominance orientation scale (mean α = .84 for 14-item version, mean α = .86 for 16-item version, and mean α = .83 for other versions).

Social Dominance Orientation as a Unified Factor

To evaluate whether social dominance orientation contains two separate factors, we tested for gender differences on the two factors identified by Jost and Thompson (2000). In studies that reported social dominance orientation as two factors (k = 15), the weighted mean gender effect sizes of GBD (mostly pro-trait items) was d₊ = 0.32 (CI 95% = 0.24, 0.41), whereas the averaged gender effect sizes of OEQ (mostly con-trait items) was d₊ = 0.31 (0.21, 0.42), with the confidence intervals indicating they are not reliably different. Two factors were correlated positively (r = 0.47). In arbitrary non-ethnic/racial groups, the averaged effect sizes of GBD and OEQ did not significantly differ from 0 (ps > .32, k = 4); the correlation between the two subscales was high (r = .93). In ethnic/racial groups, the averaged effect sizes of GBD and OEQ were −0.08 (CI 95% = −0.28, 0.12, k = 17) and 0.27 (CI 95% = 0.14, 0.40); the correlation between the two subscales was high (r = .83). According to the results of the three sets of comparisons (gender, arbitrary non-ethnic/racial groups, and ethnic/racial groups), the averaged effect sizes of GBD and OEQ differed only in comparison to ethnic/racial groups. Dominant ethnic/racial groups were more opposed to equality than subordinate ethnic/racial groups (the factor reflecting system justification according to Jost and Thompson, 2000), but they did not differ on group-based dominance (the factor reflecting group justification according to Jost &Thompson, 2000).

Clearly, these results indicate that subordinate ethnic/racial groups did not show signs of resolving dissonance between group justification and system justification by supporting the system to a larger degree than dominant ethnic/racial groups. The above findings, along with social dominance orientation’s high internal consistency, suggested that social dominance orientation should be considered a unified factor because (a) the results of gender differences and arbitrary non-ethnic/racial group differences on social dominance orientation were identical for GBD and OEQ, (b) the correlations between GBD and OEQ were high for arbitrary-set groups, and (c) the findings of group difference on GBD and OEQ were consistent with the semantic difference explanation offered by social dominance theorists (e.g., Pratto, Sidanius & Levin, 2006). That is, respondents may be more aware of equality norms when responding to the mostly pro-trait items, GBD, thus reducing intergroup disagreement. Finally, as a practical matter, relatively few studies in the literature presented results in a fashion that separated the GBD and OEQ subscales.

Construct Validity of the Social Dominance Orientation Scale

We examined the social dominance orientation scale’s predictive validity by testing whether social dominance orientation was correlated with constructs that social dominance theory hypothesizes should relate (Pratto et al., 1994). That is, we examined whether it correlated positively with endorsement of hierarchy-enhancing ideologies and negatively with endorsement of hierarchy-attenuating ideologies. For this analysis, we grouped correlations of measures conceptually. For example, multiple kinds of racism scales, usually tailored to the society in which they were administered, were grouped as racism measures. As Table 2 shows, although the number of studies reporting relevant measures was limited (k ranging from 12 to 63, depending on the type of ideology measure), social dominance orientation correlated positively as expected with nationalism, racism, heterosexism, sexism, and other types of hierarchy-enhancing ideologies (e.g., use of force, conservatism). Conversely and as expected, social dominance orientation correlated negatively with support for various types of progressive social policies (such as government social programs and civil rights). All of the I² values for the validation correlations suggested heterogeneity; some of the variation in these correlations may be due to the variety of ways the other constructs were measured. As social dominance theory predicts, regardless of how they were measured specifically, support for a variety of legitimizing ideologies corresponds with social dominance orientation. Finally, social dominance orientation was moderately correlated with right-wing authoritarianism. In addition, we should note that ingroup identification was unrelated to social dominance orientation (r = −0.01, k = 7). These results further demonstrate the robust construct validity of social dominance orientation scales.

Table 2.

Weighted Mean Correlations of Social Dominance Orientation with Other Scales, Ordered from Largest Positive to Largest Negative.

Scales	Number of studies (k)	Weighted M correlation (r+)	Homogeneity of rs: I2 (95% CI)a
Racism	35	.47	91.91 (89.75, 93.62)
Heterosexism	15	.43	79.44 (66.80, 87.26)
Sexism	35	.40	83.23 (77.53, 87.49)
Nationalism	12	.40	88.55 (81.90, 92.76)
Right-wing authoritarianism	48	.30	91.16 (89.13, 92.81)
Other beliefs in support of hierarchy (such as use of force or conservatism)	63	.32	86.58 (83.55, 89.06)
Support for progressive social policies	19	−.39	84.34 (76.80, 89.43)

Note. Higher social dominance orientation scores imply greater support of hierarchy; mean rs are random-effects means; each correlation differs significantly from zero, ps < .001. I² values vary between zero and 100, where lower values imply homogeneity.

^aWhen confidence intervals do not include zero, the hypothesis of homogeneity is rejected.

Group Comparisons on Social Dominance Orientation

Hypothesis 1 predicts that dominants have reliably higher social dominance orientation scores than subordinates, which would be shown if the mean effect sizes (d₊s) for gender and arbitrary-set groups were reliably greater than zero. Results confirmed this hypothesis. The top panel of Table 3 shows that men had reliably higher social dominance orientation scores than women both within and outside the U.S. and the bottom panel of Table 3 shows that arbitrary-set dominants had reliably higher social dominance orientation scores than subordinates.

Table 3.

Weighted Means and Mean Effect Sizes on Social Dominance Orientation for Gender and Arbitrary-Set Groups.

	All arbitrary-set groups	Non-racial groupsa	Gender or racial/ethnic groups
			All	U.S. studies	Non-U.S. studies
Gender Comparison
k of studies			169	84	85
Observed range (min, max)			−0.42, 1.70	0.00, 1.70	−0.42, 1.11
Mean d+ (95% CI)			0.43 (0.39, 0.47)	0.51 (0.46, 0.57)	0.36 (0.30, 0.41)
Homogeneity I2 (95% CI)			65.74 (59.74, 70.84)	58.01 (46.53, 67.03)	60.83 (50.40, 69.06)
Arbitrary Group Comparison
k of studies	80	24	56	37	19
Observed range (min, max)	−0.76, 1.06	−0.76, 1.06	−0.71, 0.95	−0.57, 0.95	−0.71, 0.45
Mean d+ (95% CI)	0.15 (0.06, 0.25)	0.15 (−0.02, 0.33)	0.15 (0.05, 0.26)	0.26 (0.13, 0.39)	−0.04 (−0.23, 0.15)
Homogeneity I2 (95% CI)	91.77 (90.37, 92.96)	93.03 (90.82, 94.70)	91.22 (89.37, 92.75)	89.75 (86.88, 91.99)	86.32 (80.03, 90.63)

Note. Tabled are weighted mean effect sizes using random-effects assumptions to compare dominant and subordinate groups’ social dominance orientation levels (d₊) and to generate weighted mean values on the social dominance orientation scale, which may range from 1 to 7. CI = Confidence interval. I² = Homogeneity statistic based on the fixed-effects model due to its unstable estimate in the random-effects model; the hypothesis of homogeneity is rejected when the 95% CI does not include zero.

^aNon-racial groups included groups classified based in academic major, educational level, job type, language, and political party affiliation or identification.

Hypothesis 2 concerned the invariance hypothesis and whether gender groups differ from arbitrary-set groups. Consistent with Hypothesis 2, the gender difference was reliably larger than that for the arbitrary-set groups. The weighted mean effect size, d₊ = 0.43, for the gender difference was more than twice as large as that observed for arbitrary-set groups, d₊ = 0.15. The frequency distributions of the two effect size groups appear in Figure 1. There were almost as many samples in which subordinate arbitrary-set groups (lower graphic in Figure 1) were higher on social dominance orientation (negative d) as samples in which subordinate arbitrary-set groups were lower on social dominance orientation (positive d), but the distribution centered near zero with a slight positive central tendency. In contrast, in nearly all of the samples, men were higher on social dominance orientation than women (upper graphic in Figure 1) with the mean substantially above zero. In only five out of 169 samples were the gender effect sizes negative. These results substantiate part of Hypothesis 2.

Figure 1.

Frequency distribution of group difference effect sizes (d₊) by whether group comparison was for gender or arbitrary set group.

Another part of Hypothesis 2 concerns the variability in effect sizes between samples for gender and arbitrary-set groups, which is not examined above. The invariance hypothesis implies that the variation in gender differences should be smaller than that for arbitrary-set group differences. Indeed, the confidence intervals show that the I² for the gender comparisons (upper panel of Table 3) was significantly smaller than that for arbitrary-set comparison among samples (lower panel of Table 3), and this difference remained whether the arbitrary-set group comparisons were ethnic/racial or not. These results support Hypothesis 2.

Gender Comparisons

In evaluating moderators of gender differences in social dominance orientation, we first entered each predictor alone and subsequently entered all significant predictors simultaneously in regression models. Moderators were considered significant when they were associated significantly with effect sizes alone and with methodological variables included as predictors (e.g., whether studies were conducted in the U.S.). We first evaluated the invariance hypothesis in gender ds with respect to stability over time considering U.S. studies only, a procedure that holds cultural variables relatively constant. Supporting Hypothesis 2, data collection year was not significantly related to ds in the U.S. studies (β = −0.13, p = .40 from the studies that reported data collection year and β = −0.16, p = .13 from all U.S. studies). Inspection of time plots also revealed no other systematic (i.e., non-linear) changes.

The magnitude of the mean gender effect sizes differed across the 22 nations, Q_B(21) = 54.75, p < .001, ranging from 0.68 in Russia to 0.16 in the Netherlands; thus, an examination of substantive variables that may underlie such differences can be informative. As described above, we used the continuous cultural indicators such as each nation’s cultural values and income inequality indices, to predict the effect size measures. Table 4 shows statistically reliable results, including the available numbers of studies and nations for each variable, which were sometimes small because of missing predictor values (e.g., Hofstede’s long-term orientation index was not available for 12 of the nations and was highly correlated with individualism-collectivism, r = −.81). Nations characterized by higher individualism, autonomy, and mastery were expected to have larger effect sizes (Hypothesis 3). Consistent with this hypothesis, gender differences were larger in studies from countries that were higher on individualism and affect autonomy and lower on long-term orientation and uncertainty avoidance (see regression weights and test statistics in Table 4). Figure 2, panel a, illustrates how the gender difference effect sizes increase with societal individualism. Matching the values of scores on the Individualism-Collectivism predictor scale with the effect size d₊ can help the reader interpret the column labeled Levels in Table 4, which shows estimated effect sizes at different levels of moderators. For example, a nation with a score of 13 on the collectivistic side of Hofstede’s Individualism-Collectivism scale is estimated to have a gender d₊ of 0.17, and a nation with a score of 91 at the individualistic side of the same scale, has an estimated gender d₊ of 0.48. The confidence intervals for values at these extremes, shown under the Weighted M effect size column, indicate that even in very collectivistic nations, the gender difference is significantly larger than 0, the value indicating exactly no difference. Table 4 shows that average gender differences across the extremes of the moderator variables never significantly reversed such that females supported structure more than males. Thus, the results are consistent with the weak version of the invariance hypothesis and provide supporting evidence for our reasoning that the greater freedom a society offers individuals to feel for themselves and consider themselves as separate from traditions, the more subordinates and dominants disagree on group-based hierarchy.

Table 4.

Moderator Analyses of Gender Differences on Social Dominance Orientation.

Moderator	Levelsa	Weighted M effect size, d+ (95% CI)b	Studies	Countries	Standardized regression weight (β)
					Bivariate	Combinedc
Hofstede’s nation-level cultural valuesd
Individualism-Collectivism	13 (More collectivist)	0.17 (0.06, 0.28)	169	22	0.34***	–
	91 (More individualist)	0.48 (0.44, 0.52)
Uncertainty avoidance	29 (More accepting)	0.52 (0.46, 0.58)	169	22	−0.25***	–
	95 (More avoidant)	0.28 (0.19, 0.37)
Long-term orientation	23 (More short-term)	0.50 (0.45, 0.54)	138	10	−0.29***	–
	118 (More long-term)	0.08 (−0.12, 0.28)
Schwartz’s cultural values
Affect autonomy	2.98 (lower)	0.30 (0.18, 0.42)	169	22	0.16*	–
	4.39 (higher)	0.50 (0.43, 0.58)
Equality and economic indices
GDP per capita	$1,100	0.19 (0.05, 0.32)	169	22	0.27***	–
	$46,000	0.48 (0.44, 0.52)
Gender Empowerment Measuree	.45	0.19 (0.03, 0.34)	169	22	0.23**	–
	.85	0.52 (0.45, 0.59)
Percentage women earn relative to men (relative income)	30%	0.10 (−0.04, 0.23)	169	22	0.35***	0.27***
	69%	0.54 (0.49, 0.60)
Percentage of women professionals	24%	0.15 (0.03, 0.27)	169	22	0.33***	0.25***f
	64%	0.58 (0.51, 0.65)
Fertility rate	1.1	0.38 (0.31, 0.44)	169	22	0.15*	–
	8.5	0.86 (0.43, 1.29)
Study features
USA × Measure in-group identificationg	U.S., 0 (No)	0.49 (0.43, 0.54)	74	1	0.50***h	0.40***h
	U.S., 1 (Yes)	0.69 (0.56, 0.82)	10	1	0.15*i	0.09 i
	Non-U.S., 0 (No)	0.36 (0.31, 0.41)	67	19	0.23*j	0.22*j
	Non-U.S., 1 (Yes)	0.34 (0.24, 0.44)	18	10
Measure other hierarchy beliefsk	0 (No)	0.46 (0.42, 0.51)	107	15	−0.17*	−0.17*
	1 (Yes)	0.38 (0.32, 0.44)	62	20

Note. Effect size values (d⁺) are larger to the extent that males support hierarchy to a greater extent than females. Models were estimated with the mixed-effects model.

^aEntries are predicted values for the extremes of each moderator.

^bEstimates are from the bivariate model, except where noted.

^cThe model excluding non-significant theoretical moderators: Y = −0.21 + 0.864 (proportion relative income) + 0.093 (US studies or not) −0.083 (measure other hierarchy-enhancing ideologies or not) + 0.028 (measure in-group identification or not) + 0.053 (US study × measure ingroup identification), where categorical variables are contrast coded (+1 vs. −1); none of the other moderators were significant when entered with the above moderators.

^dScores for Palestinians, Lebanon and Tunisia were estimated by scores for Arab World; scores for Albania were estimated by average scores for three countries in the Balkan region (Bulgaria, Greece, and Turkey).

^eMissing scores were estimated from scores in other years.

^fSignificant when entered with all methodological variables or with relative income, respectively, but not significant when entered with all methodological variables and relative income.

^gThe main effect terms for nation (USA vs. other) and measurement of ingroup are simultaneously entered in the equation.

^hThe main effect term for nation.

ⁱThe main effect term for measurement of ingroup identification.

^jThe interaction of nation and measurement of ingroup identification.

^kThe interaction of U.S. × Measure other hierarchy beliefs term was not significant.

⁺p < .08;

^*p < .05;

^**p < .01;

^***p < .001.

Figure 2.

Scatterplot of gender d (top panel) and arbitrary set d (bottom panel) versus nation’s score on Hofstede’s individualism-collectivism scale, with regression line indicated. Both regressions depicted differ reliably from 0 (see Tables 4, 5 for test statistics).

Greater national wealth was hypothesized to be associated with larger group differences (Hypothesis 4). Consistent with this hypothesis, gender effect sizes (ds) were larger in nations with higher GDP per capita (Table 4). Thus, national wealth was somewhat associated with how much men and women differed in their views of group-based hierarchy: The wealthier the society, the larger was the observed gender difference.

Greater structural equality was predicted to be associated with larger group differences (Hypothesis 5). As expected, in countries that had greater gender equality, as indicated by the gender empowerment measure, a larger ratio of women’s income to men’s, and a larger proportion of women professionals, larger gender effect sizes in support of group-based hierarchy appeared (Table 4). As indicated by the gender empowerment measure, the more women’s general political and economic opportunities were comparable to men’s, the larger gender effect sizes were observed, β = 0.23, p < .01. Similarly, the more closely women’s income matched men’s, the larger gender effect sizes grew, β = 0.35, p < .001 (see Figure 3, panels a and b). Gender effect sizes were also larger to the extent that countries had higher proportions of women professionals, β = 0.33, p < .001. That is, the smaller the gender gap between men and women, especially in economic opportunities and performance, the larger the gender difference on social dominance orientation. Women’s fertility rate was also associated with gender effect sizes (β = 0.15, p < .05), such that smaller differences emerged in low-birth-rate nations. The unexpected finding may be due to the oversampling of the U.S. studies. Once controlling for whether studies were conducted in the U.S., birthrate became non-significant (p = .49). In a combined model (Table 4, rightmost column), two moderators, the ratio of women’s income to men’s and percentage of women professionals, remained significant. This result indicates that among the correlated indicators of women’s empowerment, the ones most closely corresponding to women’s economic empowerment had the strongest or most direct effects on gender differences.

Figure 3.

Gender difference effect size (d₊) as a function of ratio of women’s income to men’s income by nation (top panel) and proportion of professionals who are women (bottom panel).

Sample characteristics and study features

Gender effect sizes were smaller when samples included other measures of hierarchy-enhancing ideologies beliefs (β = −0.17, p < .05). Larger gender effect sizes were observed in the U.S. studies that included measures of in-group identification than in the U.S. studies that did not include such measures; this pattern did not materialize in non-U.S. studies (β = 0.23 for interaction, p < .05). When all predictors were entered simultaneously, these two patterns remained significant (ps < .05; see the right-most column of Table 4).

Overall, the results showed that men supported social dominance orientation more strongly than women and that the gender difference on social dominance orientation was larger than the arbitrary group difference (both for non-ethnic/racial groups and for ethnic/racial groups). Moderator analyses found larger gender differences on social dominance orientation in societies that emphasize personal freedom and individualism or have greater gender equality. Conversely, smaller gender differences on social dominance orientation were found in societies that have larger tolerance of uncertainty or thinking of relationships in longer terms (e.g., in terms of one’s tradition and roles).

Arbitrary Group Differences: Ethnic/Racial Groups

We now present parallel tests of our hypotheses for arbitrary group differences. Recall that the weighted mean effect size for all such comparisons was d₊ = 0.15. Because there were significant differences among different types of arbitrary-set groups (e.g., ethnic/racial, school major) in terms of how much they supported group-based hierarchy, Q_B(6) = 27.60, p < .0001, we opted to focus on the most prevalent form of arbitrary-set comparisons, ethnic/racial groups (k = 56). As expected, when comparing ethnic/racial group differences to gender differences, the effect sizes for ethnic/racial groups changed over time. Among studies that reported data collection years, the differences between dominant and subordinate ethnic/racial groups decreased over time in the U.S. (β = −0.73, p < .01, k = 8; see Figure 4). Ethnic/racial group effect sizes also decreased over time among all U.S. studies (β = −0.45, p < .01, k = 37) and in all studies (β = −0.61, p < .0001, k = 56, see Table 5).These findings address Hypothesis 2.

Figure 4.

Arbitrary set group difference effect size (d₊) by data collection year.

Table 5.

Results of Moderator Analyses on Effect Sizes: Ethnic/Racial Group Comparison

Moderator	Levels	Weighted M effect size, d+ (95% CI)a	Studies	Countries	Standardized regression weight (β)
Data collection year	1992	0.58 (0.42, 0.75)	56	8	−0.61***
	2006	−0.27 (−0.43, −0.11)
Hofstede’s cultural values
Individualism-Collectivism	17 (More collectivist)	−0.21 (−0.49, 0.07)	56	8	0.34**
	91 (More individualist)	0.23 (0.12, 0.35)
Uncertainty avoidance	29 (More accepting)	0.33 (0.11, 0.55)	56	8	−0.23+
	81 (More avoidant)	−0.10 (−0.40, 0.20)
Masculinity-femininity	5 (More feminine)	−0.22 (−0.63, 0.20)	56	8	0.23+
	63 (More masculine)	0.20 (0.09, 0.31)
Long-term orientation	29 (Shorter-term)	0.24 (0.13, 0.36)	46	6	−0.30*
	87 (Longer-term)	−0.12 (−0.43, 0.19)
Schwartz’s cultural values
Mastery	3.81 (lower)	−0.16 (−0.49, 0.17)	56	8	0.25+
	4.12 (highest)	0.25 (0.11, 0.39)
Equality and economic indices
GDP per capita	$2,800 (lower)	−0.15 (−0.42, 0.12)	56	8	0.30*
	$46,000 (higher)	0.23 (0.11, 0.34)
Gender empowerment measure	0.60 (lower)	−0.03 (−0.24, 0.18)	54	7	0.27*
	0.82 (higher)	0.32 (0.14, 0.50)
Nature of racial comparison (U.S. studies only)	White vs. Asian	−0.12 (−0.33, 0.09)	9	1	−0.58***
	Other	0.40 (0.28, 0.52)	34	1

Note. Effect size values (d⁺) are larger to the extent that dominants support hierarchy to a greater extent than subordinates. Models follow mixed-effects assumptions.

^aEntries are predicted values for the extremes of each moderator.

⁺p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Similar to our gender findings, effect sizes that contrasted ethnic/racial groups were found to differ inside and outside the U.S., with U.S. studies having a larger mean effect size (d₊ = 0.26, 95% CI=0.14, 0.38, k = 37) than those in other parts of the world (d₊ = −0.04, 95% CI= −0.20, 0.12, k = 19) (according to a dummy-coded simple regression, β = 0.37, p = .003). To understand what accounts for differences among nations, one can examine each nation’s weighted mean effect sizes in relation to other national characteristics and test whether individualism and liberalism (Hypothesis 3), greater wealth (Hypothesis 4), or greater structural equality (Hypothesis 5) in societies are associated with larger group differences. Consistent with Hypothesis 3, that group differences will be larger in more individualistic and liberal nations, nations higher on Hofstede’s individualism-collectivism had larger group effect sizes (β = 0.34, p < .01, see Figure 2, panel b). Similarly, Schwartz’s cultural value of mastery (β = 0.25, p = .05) related positively to effect sizes and Hofstede’s masculinity-femininity dimension related marginally (β = 0.23, p = .07). Also consistent with Hypothesis 3, nations lower on Hoftede’s long-term orientation dimension (β = −0.30, p < .05) had larger group effect sizes and those lower on Hofstede’s uncertainty avoidance dimension had marginally smaller group effect sizes (β = −0.23, p = .07) (see Table 5 for all moderation results). Moreover, after entering indicators simultaneously, three indicators of societies’ liberalism were associated with larger ethnic/racial group differences, as predicted by Hypothesis 3. Greater individualism, higher mastery orientation and smaller long-term orientation were associated with larger differences between ethnic/racial groups.

Hypothesis 4 predicts larger group differences in wealthier societies. National wealth measured by GDP per capita related to the size of ethnic/racial group differences on social dominance orientation, β = 0.30, p < .05. As Figure 5 shows, racial dominants and subordinates disagreed about group-based hierarchy to a larger degree in wealthier than poorer societies. In addition, larger group differences were expected in societies with more structural equality (Hypothesis 5). We tested Hypothesis 5 by regressing group effect sizes on several indicators of structural equality. Except for the size of the population with lowest 10% income bracket, which approached significance (β = −0.22, p = .096), none of the other indicators of structural inequality was reliably associated, including the percentage of households in the highest 10% income bracket (p = .47) and Gini (p = .22). Yet, group effect sizes were related to gender equality as indicated by the gender empowerment measure, β = 0.27, p < .05 (see Table 5). These results show that structural gender equality was somewhat related to the size of ethnic/racial group differences.

Figure 5.

Arbitrary set group difference effect size (d₊) for each nation by GDP per capita.

Because there were so many U.S. studies, it was possible to examine whether the nature of the racial comparison related to arbitrary-set comparisons within that country. Indeed, within U.S. studies, comparisons between Whites and Asians yielded non-significant differences (d₊ = −0.12, 95% CI = −0.33, 0.09) whereas other ethnic/racial comparisons yielded significant differences favoring dominants (d₊ = 0.40, 95% CI = 0.28, 0.52). This difference remained significant when controlling for the year of data collection (β = −0.60, p < .001). We attempted to examine models with all of the variables in Table 5, but results were compromised by the relatively small numbers of studies available, coupled with the relatively small numbers of nations represented.

Overall, results of ethnic/racial group comparisons were both similar to and somewhat different from results of gender comparisons. The ethnic/racial group effect sizes were positive, indicating that dominants support group-based hierarchy more strongly than subordinates. Still, in most nations, there were no reliable differences between ethnic/racial groups. In general, as Figure 1 and Table 3 show, ethnic/racial group effect sizes were on average smaller but much more variable than gender effect sizes, and unlike gender effect sizes, they declined over time in the U.S. and elsewhere. One important consistency we found is that for both gender and ethnic/racial groups, effect sizes were larger in societies characterized more by structural gender equality and modern liberalism. That is, for both gender and ethnic/racial comparisons, nations higher on structural gender equality and Hofstede’s individualism vs. collectivism measure, and low on uncertainty avoidance and short-term orientation, had larger group effect sizes. Further, conceptually, other psycho-cultural factors associated with higher personal autonomy, including affect autonomy for gender comparisons, and including masculinity and mastery for ethnic/racial comparisons, were associated with larger group differences on support of group-based hierarchy. Moreover, some supporting evidence was found for larger group effect sizes in wealthier societies.

Discussion

Theorists in psychology (e.g., Sidanius & Pratto, 1999), international relations (e.g., Lebow, 2006), political science (e.g., Gurr, 1970), sociology (e.g., Jackman, 1994; Mills, 1959), and philosophy (e.g., Arendt, 1969) have considered whether inequality persists primarily because of coercion, or in part because of approval for domination among subordinates. This paper informs this question by examining whether group-based hierarchy is approved of equally by dominants and subordinates. Moreover, this paper highlights the important role of psychological processes in how people respond to their social, economic, and cultural conditions and to societal ideals to inform how much dominants and subordinates concur or disagree in their orientation toward group-based hierarchy. Our examination of the psycho-cultural and socio-structural factors thus has implications for societal stability, social inequality, and the psychological processes that influence both.

The first important psychological process we considered is adjusting to one’s social reality. Across societies, people learn that dominants have access to particular forms of power, privilege, jobs, and resources, whereas subordinates serve, do without, and are stereotyped as inappropriate to wield power or gain respect (e.g., Pratto, 1999). In many societies, people also learn from the major cultural worldviews and ideologies to which they are socialized that this kind of inequality is natural or at least justified (e.g., Pratto et al., 2000). Accordingly, people learn from the particular situation of their group. For subordinates, the contrast between their own life experiences and cultural apologies could plant the seed of skepticism about cultural ideologies that legitimize group-based hierarchy (Pratto, 1999).

The second main psychological process we described is that people not only learn, but also more actively think, reason, and make meaning about their social world, particularly regarding group inequality and favoritism. This process is partly influenced by the social conditions in which they live, for these conditions provide information and comparison points which can differ for different groups in the same society and between societies. In societies in which unequal groups are segregated into separate roles or living spaces, they may not compare their situations to those of other groups, and may be relatively satisfied. In such cases, we would expect dominants and subordinates to be more similar in their attitude towards group-based hierarchy. In contrast, in societies in which people purport to value equality, subordinates may come to expect and feel entitled to equality. The evidence and signs they observe of inequality would then mean that reality is falling short of their ideal standards. This condition may lead them to reassert their opposition to group-based hierarchy and to differentiate from dominants.

The third process concerns how people come to understand their own self in relation to others. In traditional societies, people differentiate themselves fairly little from others, as often their outcomes are interdependent with others. They assume that they should maintain long-term relationships with others, which often means subordinating their own personal desires. In such societies, subordinates may adopt similar orientations towards hierarchy as dominants. In contrast, in modern liberal societies, people are expected to differentiate themselves from others, although they may look to group memberships as one way of figuring out their identities (e.g., Markus & Kitayama, 1991). They also are expected to look out for themselves, which may make them especially attentive to how their conditions compare with those of others, and to when their conditions fall short of egalitarian ideals.

These psychological processes neither suggest that structural inequality determines whether people go along with inequality or do not, nor that social conditions are irrelevant to whether groups generally agree or whether groups disagree over how much social inequality is acceptable. That is, people actively respond to cultural norms and objective social conditions, such that structural variables alone do not completely account for whether subordinates differentiate from dominants or vice versa (B. T. Johnson et al., 2010). For this reason, examining the psychology of intergroup agreement about group-based hierarchy is especially important.

Intergroup Dissent Regarding Approval of Social Inequality

Our results show conclusively that people in less powerful gender and ethnic/racial groups do not concur with people in dominant groups concerning group-based hierarchy. Robustly, members of subordinate groups reject group-based hierarchy more than do members of dominant groups, in keeping with each group’s power position (see Figure 1). As Table 6 shows, this pattern confirms Hypothesis 1 and the predictions of social dominance theory. These results demonstrate that acquiescence to domination, or “false consciousness,” cannot be considered a default or “normal” state. Hence, structural inequality is not consensual. Rather, people in low-power social groups find group-based hierarchy more objectionable than those in high-power groups.

Table 6.

Hypotheses, Disposition, and Pertinent Results.

Hypothesis	Gender results	Arbitrary-set results (racial/ethnic)
1. Dominants are higher on social dominance orientation than subordinates.	Confirmed: d+ = 0.43 (Table 3)	Confirmed ; d+ = 0.15 (Table 3)
2. Larger, more stable, and less variable gender differences than arbitrary-set difference.	Confirmed: CI (0.39, 0.47); ds were unrelated to year of data collection (p = .40). (Figure 1)	Confirmed: CI (0.05,0.26); ds were related to year of data collection (β = −0.61, p < .01). (Figures 1 and 4)
3. Larger group differences in more individualistic and liberal societies.	Confirmed: dsrelatedto individualism (β = 0.34, p < .001), affect autonomy (β = 0.16, p < .05), uncertainty avoidance (β = −0.25, p < .001), and long-term orientation (β = −0.29, p < .001). (Figures 2 and 3)	Confirmed: ds related to individualism (β = 0.34, p < .01; Figure 2, panel b), long-term orientation (β = −0.30, p < .05); ds related marginally to mastery (β = 0.25, p = .07), masculinity (β = 0.23, p = .07), and uncertainty avoidance (β = −0.23, p = .07).
4. Larger group differences in wealthier societies.	Confirmed: ds related to GDP per capita (β = 0.27, p < .001), with larger effect sizes in wealthier societies.	Confirmed: ds related to GDP (β = 0.30, p < .05), with larger effect sizes in wealthier societies. (Figure 5)
5. Larger group differences in more equal societies.	Confirmed: ds related to gender empowerment measure (β = 0.23, p < .01), women’s relative income (β = 0.35, p < .001), and percent women professionals (β = 0.33, p < .001).	Partially confirmed: ds related to gender empowerment measure (β = 0.27, p < .05), but not to proportions of households with very low or very high income and not to the Gini coefficient.

The present results should be taken as informative precisely because they included samples from South America, North America, Southern Africa, the Middle East, Central Asia, East Asia, Eastern and Western Europe, and Britain. Our method of weighting nations using random-effects assumptions, coupled with sensitivity analyses, prevented over-represented societies such as the U.S. from having an undue influence on the results. Therefore, our results are far less culturally-particular than many other psychological studies. The fact that we found moderation by national variables helps to substantiate that the nations studied here did vary in important respects. Presumably, if we had been able to match the national moderating variables in time with when social dominance orientation was measured in the studies, the patterns would have been even more marked. Nonetheless, our samples include far more college-educated people, and more people from the Western nation-states, than can be said to represent the world. As is common in published psychological research, because the present study included no samples from central Africa and few from indigenous cultures, it is possible that other hypotheses would suggest themselves with samples of people living in different circumstances. In particular, the values and beliefs of people living in locations with high turmoil are rarely studied because of the difficulty and more important priorities that such conditions present.

Further, from its inception, social dominance theory has hypothesized that cultures invent and change legitimizing myths, or widely-known cultural ideologies, to maintain the particular kinds of group-based hierarchy in their local situation (e.g., Sidanius & Pratto, 1993), and that endorsement of such beliefs should correlate with social dominance orientation. For example, Pratto and her colleagues (2000) showed that locally-appropriate measures of sexism, ethnic prejudice, and other socio-political ideologies correlated with social dominance orientation in New Zealand, China, Israel, and Canada. Further, Sibley and Duckitt’s (2008) meta-analytic review found that social dominance orientation moderated how Openness to Experience and Agreeableness relate to prejudice. Yet, nearly all the 71 studies they examined were from Europe or the U.S. The present results show even more broadly that social dominance orientation has high internal reliability in a variety of languages and nations and that people’s levels of social dominance orientation robustly correspond to their endorsement of local social and political attitudes that legitimize group inequality. This strong evidence for construct validity increases confidence that the present results inform the study of intergroup relations and of support for group-based hierarchy, despite the fact that our meta-analysis only examined responses on the social dominance orientation scale.

Given that our study confirmed high construct validity for social dominance orientation across a variety of cultures and administered in a variety of languages, it would seem very important to use the present results to inform the study of intergroup relations. Contrary to Jost and Thompson’s (2000) suggestion that social dominance orientation contains two factors, our results demonstrate that social dominance orientation is a unified construct, because the scale enjoys high reliability, group differences on the two factors were highly correlated, and group differences on the two factors were of similar magnitude for gender and non-ethnic/racial arbitrary-set groups. Indeed, as Xin and Chi (2010) showed, much of the apparent two-factor structure is semantic, as Pratto et al. (2006) argued. We speculate that the group difference was not significant on “group-based dominance” among ethnic/racial groups because of strong opposition against overt dominance among women of dominant race/ethnicity. If women of dominant race/ethnicity are sensitive to equality norms, they may not endorse the mostly pro-trait items, “group-based dominance,” to the same degree as they may to the mostly anti-trait items, “opposition to equality.” Finally, as we demonstrated, these indexes are highly intercorrelated.

Gender in Intergroup Relations: The Invariance Hypothesis Reconsidered

Although a good deal of research and theorizing on group dissension concerns ethnic or racial groups (e.g., Bobo & Hutchings, 1996; Brewer & Campbell, 1976; Gurr, 1970), we found that the rejection of group-based hierarchy was even more common across samples by women compared to men. As social dominance theorists have argued for some time, gender is an integral aspect of intergroup relations, not a separate domain. Further, just as the present results question the concept of false consciousness as normal, they also imply that whatever the causes of male dominance are, they do not include women’s psychological acquiescence to group-based hierarchy.

Unlike many theories of gender or intergroup relations, social dominance theory argues that both gender inequality and arbitrary-set inequality (e.g., ethnic/racial inequality) make up group-based hierarchy, and that gender is not a special case of arbitrary-set relations or vice versa. Further, social dominance theory argues that societal hierarchy is partly maintained by different roles and differing psychological orientations toward hierarchy among men and women (e.g., Pratto, 1996; Pratto, Stallworth, Sidanius & Siers, 1997). One of the first documented clues to the idea that men and women play different parts in group-based hierarchy is the robust gender difference on social dominance orientation. In fact, because numerous studies in different nations have found no moderation or reversal of the gender difference, the prediction is described as the invariance hypothesis (Sidanius, Pratto, & Bobo, 1994; Sidanius & Pratto, 1999; Sidanius et al., 2000). Our results were consistent with the weak rather than strong version. As predicted by Hypothesis 2, ethnic/racial group differences varied more across time and across samples than gender differences did. In no cultures were women significantly higher on social dominance orientation than men. The finding that men score higher on social dominance orientation than women, then, is extremely robust. This psychological gender difference is consequential for group-based hierarchy because it accounts in part for why men and women select different occupational roles with respect to hierarchy (Pratto, Stallworth, Sidanius, & Siers, 1997), vote for different politicians, and prefer different social policies (e.g., Pratto, Stallworth, & Sidanius, 1997).

In addition, we were able to identify, for the first time, systematic socio-cultural moderators of the gender difference on social dominance orientation. By comparing at the sample level with different central tendencies, gender differences were larger in more individualistic societies (with high individualism and affect autonomy from others), and were smaller in societies higher in uncertainty avoidance and long-term orientation. Most importantly, and lending support to materialist arguments about feminism (e.g., Sequino, 2007), gender differences were larger when women’s income levels and professional standing had more parity with men’s. That is, the social development that allows women to enjoy more equality with men also helps to promote an appetite for more egalitarianism (Inglehart & Norris, 2003). This finding is theoretically important for several reasons.

First, the present results suggest that social scientists should reconsider how gender role norms may pertain to psychological traits, values, and worldviews. One fruitful change may be to recognize more active psychological processes rather than viewing role socialization as the passive adoption of gender role norms. Prescriptive norms may influence how people interpret their roles and social experiences. Diekman and her colleagues have shown that men and women expect to be more similar (e.g., Diekman & Eagly, 2000) and to have more similar levels of power (Diekman, Eagly, Mladinic & Ferriera, 2005) in societies with more gender equality. We have demonstrated that in such societies, these expectations may lead women to reject group-based hierarchy more than men. Our finding that more income parity and more women professionals are associated with greater gender differences suggests that women who assume that they can be men’s equals are more sensitized to social inequality. Further, it may be that when women enter domains such as professions that had been reserved to men, their experience confirms just how unequally they, and other groups, are treated. Women’s experience of becoming professionals may not simply be one of gaining power and status compared to if they had not become professionals, and hence becoming psychologically like men, but rather one of experiencing discrimination, double-standards, and double burdens compared with men professionals (e.g., Glass, 1990;J. A. Johnson & Johnson, 2008; Roth, 2004; Terbourg, 1977). In fact, experiments demonstrate that given close social comparison information (e.g., salaries of men and women in the same departments rather than in different departments), people are more likely to realize that sex discrimination is occurring (Rutte, Diekmann, Polzer, Crosby, & Messick, 1994). Another reason for the association between gender equality in societies and gender differences on social dominance orientation may be that both were brought about by feminist activism--where women (presumably especially those low on social dominance orientation) have fought for equality, they have created opportunities for higher education, equal pay, and professional roles for women as well as spreading and living out an anti-dominant ideology in their society. Such political work may change both orientations towards social dominance and the degree of inequality in a society.

Another major theoretical implication of our gender results concerns the invariance hypothesis itself. Sidanius (1993) argued that this stable and previously unmoderated psychological gender difference derives from selection for greater male dominance in humans. Although we found that the gender difference is relatively more stable than the difference between ethnic/racial groups, considerable variation in the size of the gender difference exists across cultures. In fact, we identified several moderators of this difference, although the fact that it was unreversed still requires an explanation. Instead of invoking an unknown evolutionary social past to account for the gender difference, we would point out that gender inequality is established and maintained at different but intersecting levels, from the more proximate family level to more distal society level. The reason that gender effect sizes on social dominance orientation are more stable than effect sizes for ethnic/racial groups is that women, even when they are of dominant race or higher social class, are subordinates in the family level in the sense that they serve others and are often more constrained in their life choices than are men. Conversely, men of subordinate races are still dominants in the family level (e.g., wives shouldering more domestic responsibilities even when they earn equal or more than husbands). If awareness of one’s subordinate group position allows one to reject group-based hierarchy, women should reject such hierarchy more consistently (i.e., across cultures and time) than ethnic/racial group subordinates. This reasoning is consistent with social dominance theory’s explanation of why subordinate arbitrary-set groups would often be less social dominance oriented than dominant arbitrary-set groups.

Our finding that there are larger variations among ethnic/racial group differences than on gender differences has implications for studying arbitrary-set group differences as well. To date, a relatively small number of nations have documented arbitrary-set differences on social dominance orientation, so one might expect to document more subtle patterns of moderation as greater numbers of studies accrue in the literature. To understand colonialism and post-colonialism as well as international relations, it is clear that research regarding arbitrary-set distinctions in societies that undergo societal changes is important. For example, comparing recently colonized people and their recent colonizers would be theoretically warranted. Comparison of mean scores across societies presents interpretational difficulties due in part to response styles and different norms regarding group-based hierarchy in different societies. For this reason, we recommend comparing effect size measures (e.g., mean effect sizes) of current and formerly dominant and subordinate groups within particular societies (e.g., Lee & Pratto, 2011). In addition to longitudinal studies, perhaps meta-analysis is most useful in revealing the interactions between arbitrary-set membership and other factors.

An important limitation of our study is that we could not evaluate whether gender differences are moderated by arbitrary-set differences (e.g., ethnic/racial differences) or vice versa, because very few sources reported statistical findings appropriate to a gender × arbitrary-set interaction. Examining gender and arbitrary-set differences conjointly would be more in keeping with social dominance theory and with the movements toward race/gender “intersectionalism” in sociology (e.g., Andersen, 2005; Brewer, 1993). Sidanius and colleagues (1991, 1994a, 1994b, 1995b, 1999b, 2000) did conduct arbitrary-set × gender comparisons on social dominance orientation within samples, and found no reliable interactions. Nonetheless, large numbers of samples appear necessary to find moderation of the gender difference on social dominance orientation. To enable future researchers to conduct a meta-analysis that would test whether arbitrary-set differences moderate gender differences, researchers are encouraged to report descriptive statistics for both participant genders within arbitrary-set status.

Although our documentation of the robust gender difference on social dominance orientation does not definitively identify the causes of this difference, it does help to emphasize that understanding gender is essential for theories of intergroup relations (e.g., Pratto, 1996; Sidanius, Cling, & Pratto, 1991; Sidanius et al., 1995; Sidanius & Pratto, 1999). The examination and discussion of similarities and differences between gender and arbitrary-set groups should be the subject of rigorous theory and research. For example, there are aspects of social life in which gender is organized differently than arbitrary-set groups (Pratto & Walker, 2004), most notably how families are formed. There is also variability in the structural nature of arbitrary-set groups – for example, there are quite different economic relations between dominants and subordinates under chattel slavery vs. apartheid vs. communism vs. feudalism vs. industrial capitalism, although economic (and other) power relations exist in all such economic systems. In some societies, subordinate arbitrary-set members are assigned to comparable work roles as are women. For example, in many societies, women (mostly) do laundry, but in different locations and periods within U.S. history, slaves, indentured servants, and Chinese immigrant men, have also done this work for others. In hiring experiments, both men and women from subordinate arbitrary-set groups are assigned to hierarchy-attenuating jobs over hierarchy-enhancing jobs in much the way that White women are (e.g., Pratto & Espinoza, 2001). This discussion points out that the functioning of gender and arbitrary-set groups can be described along several theoretical dimensions, such as psychological orientations towards group-based hierarchy, stereotypes, work roles, and kinds of power to which these groups have. The next step for theoretical and empirical work concerning how gender and arbitrary-set groups compare and intersect must delineate these aspects in detail so that overgeneralizations are not made.

The finding that more egalitarian contexts give rise to larger group differences highlights that the objective reality of inequality, and people’s subjective orientations toward inequality, are not the same thing. For example, one cannot assume that because African-Americans have higher self-esteem (reflecting subjective evaluation) than Euro-Americans, that Whites are subordinate to Blacks in the U.S. (reflecting objective reality). For theorizing about gender, this finding is important because it shows that more gender equality does not necessarily lead to gender sameness. Although we agree that there is no reason in principle to assume that gender differences should always be present (Constantinople, 1973; Hyde, 2005), such that, for example, women are inherently superior to men, the predominant social-situationist assumption that equality should be associated with smaller or no gender differences, may have led scholars to overlook the causes and meaning of such differences. In particular, if contemporary psychologists have been operating under the assumption that egalitarian contexts prevent rather than provoke gender differences (e.g., Hyde, 2005; Wood & Eagly, 2002) and find gender differences nonetheless, this assumption might lead them to look for “non-social” causes of psychological gender differences, such as hormones (e.g., Angold & Worthman, 1993), evolution (e.g., Wilson, 1997), or brain structure (e.g., Overman, Bachevalier, Schuhmann, & Ryan, 1996). Our point is not that such research or theories are wrong, but that the exclusive focus on biological factors may lead to a false sense that psychological gender differences in egalitarian social contexts reflect fixed and fundamental gender differences. If important cultural factors moderate the expression of gender differences, then surely a simple equation between “equality” and “sameness” oversimplifies reality. Moreover, this entire intellectual history suggests that theorists are still assuming a kind of nature/nurture dichotomy, in which differences that cannot be shown to be associated with cultural or social context factors must somehow be due to biology. This represents an inadequate understanding both of the nature of “culture” and the essential plasticity of biology.

Finally, we argue that “difference” should not be presumed to be the same as “social inequality.” Although some of the egregious ideological arguments equating difference with inferiority (e.g., eugenics, Nazi anti-Semitism; antebellum racism) may lead one to assume that difference must necessarily reinforce group-based hierarchy, it is important to recall that groups sometimes celebrate their “difference” as a way of rejecting social inequality and as an initial stage of achieving actual social equality. For example, a group of illiterate women in a marginalized region of ancient China invented Female Script and used the unique language to express their feelings, experiences, and hopes (Liu, 2005; Wang & Hu, 2005). Indigenization movements (e.g., Black is Beautiful) reject mainstream views of minority culture and in doing so, they assert honor and positive values of being a member of their race/ethnicity (Carmichael & Hamilton, 1967). Some people who were born with both male and female sexual organs assert their rights of not being forced surgically and socially to become male or female (e.g., Hegarty, 2000). Asserting and celebrating difference is sometimes an anti-dominance act. Given such examples, surely researchers can presume equal human rights and equal potentials across gender, race, nationality and other social categories without assuming that recognizing differences is necessarily regressive and essentialist.

The moderation effects we found allow us to reconsider theories concerning why subordinates reject group-based hierarchy more than dominants do. Next, we consider what the results imply with respect to social group identification, wealth, group inequality, and political liberalism in social dominance orientation.

Is Social Dominance Orientation Merely An Expression of Social Identity?

Our results can inform a controversy in the literature concerning whether group differences in social dominance orientation primarily reflect group differences in group identification. That is, the group differences in social dominance orientation may occur because dominants identify more with dominant ingroups than subordinates identify with subordinate ingroups (Sidanius, Pratto, & Rabinowitz, 1994; Wilson & Liu, 2003). Some have claimed that social dominance orientation is just a proxy for group identification (Schmitt, Branscombe, & Kappen, 2003; Reynolds & Turner, 2006). Our results provide two kinds of evidence against the idea that social dominance orientation represents dominant-group identification or in-group identification. First, Schmitt et al. (2003) observed that when responding to social dominance orientation scale items, American college students are more likely to think about race than about gender. If social dominance orientation levels are due to group identification with the most salient kind of group distinction, then Schmitt et al.’s findings imply that group differences on social dominance orientation should be larger between ethnic/racial groups than between gender groups, at least in the U.S. In contrast, we found larger gender differences on social dominance orientation than race differences both in the U.S. and worldwide. Second, group identification is increased by the motivation to avoid uncertainty (e.g., Hogg, 2000). For example, Mullin and Hogg (1999) showed that people who were experimentally induced to feel uncertain became more identified with minimal groups to which they were subsequently assigned and valued them more. Using this reasoning to understand group differences in social dominance orientation, and again assuming that social dominance orientation is a proxy for group identification, in nations with higher uncertainty avoidance, group identification and group differences in social dominance orientation should be larger. Yet our results showed just the opposite patterns: Nations with lower uncertainty avoidance had larger gender differences on social dominance orientation, and marginally larger ethnic/racial differences. In total, we have two kinds of evidence against the idea that social dominance orientation is a proxy for group identification or that the size of group differences on social dominance orientation is driven by identification with dominant groups.

Does Societal Wealth or Income Inequality Reduce or Increase Intergroup Dissent?

As societies become wealthier, some political actors and social scientists expect this wealth to increase structural equality such that societal wealth should reduce intergroup dissent (e.g., Campbell, 1965; Clinton, 1996). Similarly, Caprara et al. (2006) argued that increasing societal wealth will lead people to differentiate individually, effectively quelling group identity, group-interested politics, and group differences in political attitudes. These views suggest that group differences should be smaller in wealthier societies. Conversely, societal wealth may encourage individuals’ mobility and self-expression, which increase intergroup comparisons and result in larger intergroup disagreement (Hypothesis 4). The results supported the latter. Greater societal wealth was associated with larger gender differences and larger ethnic/racial group differences on social dominance orientation. This finding lends support to our thesis that psychological interpretation of structural conditions, such as feeling of relative deprivation or choices of reference groups, rather than just objective conditions, such as poverty, are essential to the social-psychological consequences.

According to relative deprivation theory, which implies that increasing equality can lead people in different groups to compare their situations to one another and develop different attitudes towards equality, this process should differentiate groups more in societies with more structural equality (Hypothesis 5). Indeed, our results showed that even in the century of globalization, people’s group positions matter to them because group membership is robustly related to social dominance orientation levels. Notwithstanding some political leaders’ and learned scholars’ views, our results showed support for greater intergroup dissension about group-based hierarchy in societies with more equality. For both gender and ethnic/racial groups, we found that greater intergroup dissension was related to higher structural equality indices (e.g., gender empowerment measure). We should also note that there are no objective indicators of levels of arbitrary-set inequality that are available across nations. Further research will be necessary to determine whether this pattern is actually due to gender equality producing arbitrary group differences, or whether the gender equality indicators correspond with equality between arbitrary-set groups that are unmeasured.

Without considering the subtleties of psychological adaptation levels and social comparison, it may seem counter-intuitive that we found that people in more unequal societies concur with others more than those in more equal societies. In other words, greater structural equality between groups appears to have the ironic consequence of increasing psychological value differences between groups. Experimental research that controls for expectancies and then changes the level of inequality would provide more definitive evidence about the psychological processes, such as social comparison, that contribute to the effects we found here. Nonetheless, our results show that the objective level of inequality between groups cannot be read from how much they subjectively disagree about inequality. As material wealth and structural equality alone do not paint the full picture for the psychology of inequality, let us consider more subjective psycho-cultural factors.

Liberal Social Environments Increase Group Dissension regarding Social Dominance

Based on research regarding culture and political psychology, we hypothesized that aspects of political culture would moderate the size of group differences on social dominance orientation. Supporting Hypothesis 3 and as shown in Figure 2, we found that subordinates differentiate themselves from dominants more in individualistic cultures than in collectivist cultures (see Table 6). Individualistic cultures offer individuals more freedom to think and feel for themselves, subordinates may use dominants as the reference groups and become more aware of how unequal their groups are, and reject group-based hierarchy. Conversely, in societies that value certainty and tradition, subordinates are more similar to dominants in their approval of group-based hierarchy. Our results emphasize that theorizing about how the cultural and political contexts in which people live may have a dramatic influence on their psychological orientations to the world. We concur with many prior authors in the position that, without seeking varied samples (e.g., not just those from individualistic societies), scientists may miss very important variability in social and psychological life (cf. Henrich, Heine, & Norenzayan, 2010).

An alternative explanation to the finding of individualism and intergroup dissension concerns cultural variability in the legitimacy of inequality. That is, it may simply be the case that collectivist societies promote power distance between individuals, whereas power distances are eschewed in individualistic, egalitarian societies. Guimond (2008) has argued that the power distance index corresponds to the size of other psychological gender differences. Yet, we found no evidence for cultural dimensions of power differential and intergroup dissension (i.e., Hofstede’s power distance index, Schwartz’s hierarchy index) relating to group differences. One plausible explanation for the difference between our results and Guimond’s (2008) concerns whether the legitimacy of inequality pertains to individual persons or to groups. In more collectivistic societies, power distance is legitimized both for dominants and subordinates in societies. In more individualistic societies, even small power differentials are scrutinized and objected to by subordinates. Thus, broader cultural values about the relation of self to others, such as individualism-collectivism and long-term orientation, were found to be associated with group difference on social dominance orientation, but not power distance or hierarchy. Nonetheless, our results and Guimond’s (2008) both demonstrate that relatively liberal, individualistic social environments have more intergroup dissension than more collectivistic societies or societies that value certainty and tradition.

A second alternative explanation for our finding of individualism and intergroup dissension is that subordinates in more collectivistic societies may fail to express their true feelings about group-based hierarchy in studies, perhaps due to their blind adoption of the society’s norm. Several results lead us to be skeptical of this interpretation. First, the group differences for gender and for arbitrary-set distinctions were reliable in collectivistic cultural settings; they were simply smaller in some settings than in others. Second, social dominance orientation scales had the same degree of internal reliability and convergent validity in countries that are collectivist as they do in Western nations. In the subset of our samples from Albania, China, Colombia, India, Israel, Lebanon, Pakistan, Palestine, Taiwan, the Tunisian Republic, and Russia, which are all scored fairly collectivistic (individualism-collectivism scores lower than 55), social dominance orientation has acceptable internal reliability (α = .78). Third, all the reported significant correlations between social dominance orientation and scales of hierarchy-enhancing ideology remained significant in these nations (rs > .20, p < .0001). The fact that social dominance orientation is not simply unreliable or predictively invalid in certain cultures suggests that the results are not due to an under-reporting bias associated with collectivist societies. Likewise, these results demonstrate that group differences in social dominance orientation are not an artifact of individualistic societies. Instead, our results show that social mobility and freedom are associated with greater intergroup dissent, whereas security concerns and stability lead to less intergroup differentiation.

In turn, these patterns suggest a new view of the major socio-political argument of the 20^th century, and indeed, of understanding societal modernity versus hierarchical tradition, namely, whether there is a relation or necessary trade-off between political freedom and social equality (Burke, 1790/1955; Rokeach, 1973). We found that in societies that value stability, people agree more with those in other groups. This pattern suggests that de-valuing freedom helps maintain or create political consensus by preventing calls from subordinates for greater equality, which helps preserve hierarchical social order. In other words, our results imply that by limiting freedom, societies may prevent egalitarian sentiments in subordinate groups. Conversely, our finding that freer societies that enable individualist, mobility societies have greater group dissension suggests that greater freedom may give rise to greater demands for equality, which eventually lead to society with greater equality. In other words, freedom may promote egalitarian expectations and demands, especially among subordinate groups. This conclusion implies that freedom and the aspiration to equality are not traded-off, but may be synergistic.

Lastly, we argue that psychological responses to liberal cultures may influence further social change and the stability of social systems differently from what psychological responses to traditional cultures may influence. In traditional cultures, hierarchy and constraint mutually reinforce one another, so that inequality and social stability are maintained almost through social inertia. That is, the existing hierarchical arrangements appear relatively beneficial, fixed, and inevitable, and because this situation does not invite strong opposition, the stability and cohesion of the society can be maintained. In contrast, in liberal cultures, subordinates adopt distinctively more egalitarian views than dominants. Their greater freedom allows them to agitate for equality and engage in other mobility strategies, which coupled with their opposition to group-based hierarchy, may bring about more equality even while threatening social cohesion. Such activities will produce more social change, which, it should be observed, has included revolutions that separated societies or threatened to (e.g., Czechoslovakia, Belgium), revolutions that inverted the dominant group (e.g., USSR, Zimbabwe), and revolutions that greatly increased equality for many members of the society (e.g., feminism). An important question for the future is to understand what cultural and psychological processes contribute to these different ways of effecting social change.

Appendix II.

Arbitrary groups comparison studies

Study	Method	Comparison group	Country	% of female	Samples (%)	d+	SDO α	SDO item
Nicol et al. (2007) sample 1		military vs. civilian	Canada	29.0	ColS; MA: 19.3	−0.41	0.81	20
sample 2		military vs. civilian	Canada	30.7	ColS; MA: 20	−0.20	0.88	20
sample 3		military vs. civilian	Canada	30.9	ColS; MA: 20	0.15	0.88	20
sample 4		military vs. civilian	Canada	33.3	ColS; MA: 22.3	0.22	0.92	20
Dambrun, Guimond, & Duarte (2002)	CS; CS (p); FFS; MI	Major	France	64.9	ColS; MA: 22.6	1.06	0.81	10
Dambrun, Kamiejski, Haddadi, & Duarte (2009) sample 1		Major	France	81.1	ColS; MA: 20.7	0.11	0.88	6
sample 2		Major	France	81.0	ColS; MA: 20.7	0.70	0.88	6
Guimond, Dambrun, Michinov, & Duarte (2003)	CS; FFS; MG	Major	France	81.1	ColS; MA: 19.8	0.39	0.86	16
Bruess (2004) sample 1	RS	R/E: Germans vs. Turks	German	49.9	HSS; MA: 17	0.15
sample 2	RS	R/E: Germans vs. Resettlers	German	49.9	HSS; MA: 17; 89 Wh; 11 As	0.03
Khan, & Jiu (2008) sample 1		R/E: Hindu vs. Muslim	India	33.3	ColS; MA: 23.2	−0.19	0.82	16
Levin, & Sidanius (1999) sample 2	CS; FFS	R/E: Ashkenazic vs. Mizrachim	Israel	58.8	ColS	0.30	0.72	8
sample 3	CS; FFS; MG	R/E: Israel Jew vs. Arab	Israel	63.6	ColS	0.42	0.78	8
Schleien (2009) sample 1		R/E: Israeli vs. Palestinian	Israel		MA: 15.4	−0.71	0.85	8
sample 2		R/E: Israeli vs. Arab	Israel		MA: 15.4	−0.64	0.85	8
Roccato (2008) sample 1		Major	Italy	76.9	ColS; MA: 20.4	−0.29	0.82	14
sample 2		Major	Italy	76.9	ColS; MA: 20.4	−0.09	0.82	14
sample 3		Major	Italy	76.7	ColS; MA: 20.4	0.20	0.82	14
Villano, & Zani (2007)		Education Level	Italy	62.8	GP; MA: 31	1.77	0.87	16
Khan, & Jiu (2008) sample 2		R/E: Hindus vs. Muslims	Pakistan	61.4	GP; MA: 29.2	−0.09	0.83	16
Golec, & Federico (2004)	CS; FFS; MI	Party: LPR vs. UW	Poland	24.2	GP; MA: 22	1.01	0.88	14
Heaven, Greene, Stone, & Caputi (2000)	CS; FFS; MG	R/E: Wh vs. Bl	South Africa		ColS; 67.9 Wh; 32.1 Bl	0.45	0.76	14
Meyer (2004) sample 1	CS; FFS; MG	Language	South Africa	60.2	HSS; GP; MA: 25.8; 32 Wh; 33 Bl	−0.75	0.83	16
sample 2	CS; FFS; MG	Language	South Africa	60.2	HSS; GP; MA: 25.8; 32 Wh; 33 Bl	−0.54	0.83	16
sample 3	CS; FFS; MG	R/E: Wh vs. Bl	South Africa	60.2	HSS; GP; MA: 25.8; 50 Wh; 50 Bl	−0.30	0.83	16
sample 4	CS; FFS; MG	Education Level	South Africa	60.2	HSS; GP; MA: 25.8; 32 Wh; 33 Bl	−0.27	0.83	16
sample 5	CS; FFS; MG	R/E: Wh vs. Colored	South Africa	60.2	HSS; GP; MA: 25.8; 49 Wh; 3 Bl	0.14	0.83	16
sample 6	CS; FFS	R/E: Wh vs. Bl	South Africa	63.0	ColS; MA: 19.5; 65.1 Wh	0.43	0.87	16
sample 7	CS; FFS; MG	R/E: Wh vs. Other	South Africa	53.4	GP; MA: 24.6; 48 Wh; 52 Bl	−0.24	0.86	16
sample 8	CS; FFS; MG	R/E: Wh vs. Colored	South Africa	53.4	GP; MA: 24.6; 50 Wh; 3 Bl	−0.12	0.86	16
Rowatt, Franklin, & Cotton (2005)	CS; FFS; MI	R/E: Swedish vs. Immigrants	Sweden	84.2	ColS; MA: 19.6; 63 Wh; 12 Bl; 7 As; 10.5 La	−0.27	0.90	16
Saucier, & Cawman (2004)	CS; FFS	Party Affiliation	Sweden	65.1	GP; MA: 25.8; 89.6 Wh		0.67	4
Huang, & Liu (2005) sample 1	CS; FFS	R/E: Mainlander vs. Hakka	Taiwan	51.4	ColS; MA: 29; 100 As	−0.07	0.81	14
sample 2	CS; FFS	R/E: Mainlander vs. Minnan	Taiwan	51.4	ColS; MA: 29; 100 As	−0.04	0.81	14
sample 3	CS; FFS	R/E: Mainlander vs. Minnan	Taiwan	51.5	ColS; MA: 29; 100 As	−0.28	0.86	14
sample 4	CS; FFS	R/E: Mainlander vs. Hakka	Taiwan	51.5	ColS; MA: 29; 100 As	0.15	0.86	14
Liu, Huang, & McFedries (2008)		Party ID (KMT vs DDP)	Taiwan	49.0	ColS; Other; MA: 33.3	0.15	0.87	12
Dickins, & Sergeant (2008)		Sexual Orientation	UK	25.8	MA: 22	0.99	0.90	14
Aquino, Stewart, &Reed II (2005) sample 1	CS; OSM; MG	R/E: Wh vs. Other	US	54.1	ColS; 89 Wh; 4 Bl; 7 La	0.16	0.90	16
sample 2	CS; FFS; MG	R/E: Wh vs. Other	US	45.0	ColS; 91 Wh; 4 Bl; 5 La	0.40	0.90	16
Caldwell (2007)		Education Level	US	64.6	MA: 39; 84.6 Wh; 9.2 Bl; 3.1 La	−0.22		16
Culhane, Hosch, & Heck (2008) sample 1		Law Enforcement: Current vs. Future	US	44.0	ColS; Other; MA: 26.6; 18.2 Wh; 4.5 Bl; 75.4 La	0.25		16
sample 2		Law Enforcement vs. Student	US	43.7	ColS; Other; MA: 26.6; 18.2 Wh; 4.5 Bl; 75.4 La	0.37		16
Fang, Sidanius, & Pratto (1998) sample 1	RS; PS; MI	R/E: Wh vs. Hispanic American	US	54.9	GP; MA: 41; 57 Wh; 43 La	0.95	0.65	4
sample 2	RS; PS; MI	R/E: Wh vs. As	US	54.9	GP; MA: 41; 69 Wh; 31 As	0.70	0.65	4
sample 3	RS; PS; MI	R/E: Wh vs. Bl American	US	54.9	GP; MA: 41; 57 Wh; 43 Bl	0.33	0.65	4
Federico (1999) sample 1	CS; FFS	R/E: Wh vs. Bl	US	59.8	ColS; MA: 21.5; 54.9 Wh; 45.1 Bl	0.77	0.87	16
sample 2	CS; FFS; MG	R/E: Wh vs. La	US	59.8	ColS; MA: 21.5; 54 Wh; 46 La	0.75	0.91	16
Hailey (2005) (unpublished)	CS; FFS; MI	R/E: Wh vs. Bl	US	33.0	GP; MA: 35.4; 58.4 Wh; 41.6 Bl	0.03	0.90	16
Hayward (2001) sample 1	CS; FFS; MG	R/E: Wh vs. Chinese American	US	63.5	ColS; MA: 20.2; 52 Wh; 48 As	−0.57	0.79	14
sample 2	CS; FFS; MG	R/E: Wh vs. Vietnamese American	US	65.5	ColS; MA: 20.5; 64 Wh; 36 As	−0.22	0.79	14
sample 3	CS; FFS; MG	R/E: Whs vs. Mexican American	US	68.0	ColS; MA: 20.6; 62 Wh; 38 La	0.35	0.79	14
sample 4	CS; FFS; MG	R/E: Wh vs. African American	US	73.6	ColS; MA: 21.9; 89 Wh; 11 Bl	0.72	0.81	11
sample 5	CS; FFS; MG	R/E: Wh vs. Chinese Americans	US	70.2	ColS; MA: 20.8; 45 Wh; 55 As	−0.43	0.81	11
sample 6	CS; FFS; MG	R/E: Wh vs. Korean American	US	71.4	ColS; MA: 21.2; 61 Wh; 39 As	−0.17	0.81	11
sample 7	CS; FFS; MG	R/E: Wh vs. Mexican American	US	70.3	ColS; MA: 21.3; 56 Wh; 44 La	0.27	0.81	11
sample 8	CS; FFS; MG	R/E: Wh vs. Vietnamese American	US	70.6	ColS; MA: 21.5; 65 Wh; 35 As	−0.28	0.81	11
Jost, &Thompson (2000) sample 1	CS; FFS; MG	R/E: Wh vs. Bl	US		ColS; 78 Wh; 22 Bl	0.48		16
sample 2	CS; FFS; MG	R/E: Wh vs. Bl	US		ColS; 79 Wh; 21 Bl	0.58		16
sample 3	CS; FFS; MG	R/E: Wh vs. Bl	US		ColS; 79 Wh; 21 Bl	0.62		16
sample 4	CS; FFS; MG	R/E: Wh vs. Bl	US		ColS; 77 Wh; 23 Bl	0.44		16
Levin, & Sidanius (1999) sample 1	CS; FFS	R/E: Wh vs. La	US	55.9	ColS; 55 Wh; 45 La	0.70	0.90	16
Lippa, & Arad (1999) sample 1	CS; FFS; MI	R/E: Wh vs. La	US	62.3	ColS; MA: 18; 64 Wh; 36 La	0.54	0.79	8
sample 2	CS; FFS; MI	R/E: Wh vs. As	US	62.3	ColS; MA: 18; 66 Wh; 34 As	−0.14	0.79	8
McIntyre, et al. (2007) sample 1		R/E: Wh vs. As	US	43.3	ColS; MA: 22; 74.5 Wh; 26 As	−0.14
sample 2		R/E: Wh vs. other	US	43.0	MA: 22; 87.5 Wh	0.24
sample 3		R/E: Wh vs. Bl	US	42.7	MA: 22; 84.7 Wh; 15.3 Bl	0.68
Nelson, & Milburn (1999)	CS; FFS; MG	Party Affiliation	US	60.4	ColS; MA: 18.7; 84 Wh; 10 As	0.50	0.83	14
Pena, & Sidanius (2002)	SRS; PS; Int; MI	R/E: Wh vs. La	US	57.4	GP; 67.9 Wh; 32.1 La			6
Pratto, & Lemieux (2001)		Party Affiliation	US	48.3	GP; 85.2 Wh; 6 Bl; 9 As; 3.1 La	1.32	0.72	6
Riley (2006)		R/E: Wh vs. Colored	US	76.4	ColS; 70.4 Wh; 20.4 Bl; 9.1 La	0.63	0.85	16
Sabir (2006) sample 1		R/E: Wh vs. Bl	US	28.4	54.2 Wh; 21.5 Bl; 1 As; 21.5 La	−0.33	0.91	16
Sabir (2006) sample 2		R/E: Wh vs. Hispanic	US	28.4	54.2 Wh; 21.5 Bl; 1 As; 21.5 La	−0.18	0.91	16
Schmitt, Banscombe, & Kappen (2003)	CS; FFS; MG	R/E: Wh vs. Other	US	52.1	ColS; MA: 18.7; 90.1 Wh; 4 Bl; 5.9 La	0.22	0.87	10
Shorey, Cowan, & Sullivan (2002)	CS; FFS; MG	R/E: Aanglo vs. Hispanic	US	53.4	ColS; MA: 26; 54.8 Wh; 45.2 La	0.17	0.91	16
Sinclair, Sidanius, & Levin (1998)	CS; FFS; MG	R/E: Wh vs. As	US	56.0	ColS; 43 Wh; 57 As	−0.09	0.77	4
Umphress et al. (2008) sample 1		R/E: Wh vs. Colored	US	51.5	ColS; MA: 21; 83.8 Wh; 2.9 Bl; 3 As; 8.8 La	0.12	0.89	16
sample 2		R/E: Wh vs. Colored	US	59.5	ColS; MA: 22; 79.7 Wh; 10.1 Bl; 1 As; 6.3 La	0.10	0.92	16
Umphress et al. (2007)		R/E: Wh vs. Colored	US		ColS; MA: 23; 70 Wh; 3 Bl; 11 As; 14 La	0.67	0.89	16
Worthington et al. (2008)		R/E: Wh vs. NonWh	US	63.2	ColS; Wh: 65.3; 18.1 Bl; 15 As; 4.9 La	0.33	0.93	14

RS=Random sample; SRS=Stratified random sample; CS=Convenience sample; (p)=purposive; MG=Measured in group; MI=Measured individually; FFS=Face to face survey; PS=Phone survey; Int=Interview; MS=Mail survey; OSM=Other survey method; OM=Online measure; ColS=College students; HSS=High school students; GP=General public; MA=Mean age; OEG=Other ethnic group; ColG=College graduates; HSG=High school graduates; GSG=Graduate School graduates; OEdu=Other education; Wh=White; Bl=Black; As=Asian; La=Latino