A cross-cultural investigation into a reconceptualization of ethnocentrism

European Journal of Social Psychology Eur. J. Soc. Psychol. 39, 871–899 (2009) Published online 31 December 2008 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/ejsp.589

A cross-cultural investigation into a reconceptualization of ethnocentrismy BORIS BIZUMIC1*, JOHN DUCKITT2, DRAGAN POPADIC3, VINCENT DRU4 AND STEPHEN KRAUSS5 1

The Australian National University, Australia University of Auckland, New Zealand 3 University of Belgrade, Serbia 4 University of Paris X, France 5 University of Illinois at Chicago, USA 2

Abstract This investigation tests a reconceptualization of ethnocentrism based primarily on Sumner’s definitions. Ethnocentrism is reconceptualized as ethnic group self-centeredness, with four intergroup expressions of ingroup preference, superiority, purity, and exploitativeness, and two intragroup expressions of group cohesion and devotion. The reconceptualization was supported in Study 1 among 350 New Zealand participants and in Study 2 among 212 US, 208 Serbian, and 279 French participants. Ethnocentrism in each country consisted of two correlated second-order factors representing intergroup and intragroup ethnocentrism and six first-order factors representing the six primary expressions. Analyses in Study 2 supported the measurement invariance of the scale and a third-order factor model, with one ethnocentrism factor at the broadest level of generalization. Ethnocentrism was empirically distinct from outgroup negativity and mere ingroup positivity. Intragroup ethnocentrism appeared primarily based on ethnic insecurity, personal self-transcendence, and ethnic identification, whereas intergroup ethnocentrism appeared primarily based on self-aggrandizement, warlikeness, and generally chauvinistic attitudes. Accordingly, although related, the two kinds of ethnocentrism tend to have quite differential implications for group attitudes and behaviors. Copyright # 2008 John Wiley & Sons, Ltd. ‘‘The sentiment of cohesion, internal comradeship, and devotion to the in-group, which carries with it a sense of superiority to any out-group and readiness to defend the interests of the in-group against the out-group, is technically known as ethnocentrism.’’ (Sumner, 1911, p. 11). Ancient Greeks called all Non-Greeks ‘‘barbarians,’’ because, they said, barbarians speak unintelligibly—‘‘bar–bar–bar’’ (Jahoda & Krewer, 1997). A Brazilian tribe has the term ‘‘kura’’ to signify ‘‘we’’ and ‘‘good,’’ and the term ‘‘kurapa’’ to signify ‘‘not we,’’ and ‘‘bad.’’ The tribe believes that everything that is bad in the world comes from the outside and is sent by aliens (Von den Steinen, cited in Keller, 1931/1973). Most of the names given by preindustrial societies to their own tribe are translated in English as ‘‘people,’’ ‘‘the only people,’’ or ‘‘people of people’’, and often these tribes see themselves as the origin of all humans (Sumner, 1906). It has been recognized for many centuries that people tend to be ethnocentric. For example, the ancient Greek historian Herodotus (trans. 1947) wrote: ‘‘For if one were to offer men to choose out of all the customs in the world such as seemed to them the best, they would examine the whole number, and end by preferring their own; so convinced are they that their own usages far surpass those of all others’’ (p. 160). Ethnocentrism is still quite prominent in the world, in which many groups and cultures assume superiority and tend to reject and exploit those belonging to other groups and cultures. *Correspondence to: Boris Bizumic, Department of Psychology, The Australian National University, Canberra, ACT 0200, Australia. E-mail: [email protected] y This article was published online on 31 December 2008. Errors were subsequently identified. This notice is included in the online and print versions to indicate that both have been corrected [date to be added later].

Copyright # 2008 John Wiley & Sons, Ltd.

Received 5 June 2007 Accepted 20 October 2008

872

Boris Bizumic et al.

Ethnocentrism is a major component of the modern-day nationalism, and has been implicated in the development of prejudice (see Brewer, 1999). Accordingly, ethnocentrism has been of particular interest to social psychologists because of its adverse effects on harmonious intergroup relations, and many theories have been developed in attempt to explain origins of ethnocentrism. Some of the well-known theories are authoritarian personality theory (Adorno, FrenkelBrunswik, Levinson, & Sanford, 1950), belief congruence theory (Rokeach, 1960), realistic conflict theory (Sherif, 1966), and social identity theory (Tajfel & Turner, 1986). The anthropologist McGee (1900) appears to be the first to use the term ethnocentric to describe this phenomenon and Sumner (1906) appears to be the first to define it. Although both saw ethnocentrism as ethnic group self-centeredness, Sumner also assumed that it involves outgroup negativity. Sumner’s writings on ethnocentrism were highly influential and ethnocentrism is today seen as one of the fundamental concepts in the social sciences. Although widely influential, Sumner’s descriptions of ethnocentrism involved considerable ambiguity and were open to differing interpretations (LeVine & Campbell, 1972). As a result, many studies have operationalized and measured ethnocentrism in rather different ways. Ethnocentrism has been operationalized as social distance (Thomas, 1975), the extent to which one’s own ethnic group is central in one’s life (Tzuriel & Klein, 1977), or preference for the ingroup over outgroups (e.g., Crocker & Schwartz, 1985; Grant, 1993). It has often been studied as the association of positive ingroup attitudes and negative outgroup attitudes (e.g., Adorno et al., 1950; Chang & Ritter, 1976). Others, such as Altemeyer (1998) and Beswick and Hills (1969), have measured it primarily as generalized outgroup negativity. Given these very different ways of measuring or operationalizing ethnocentrism, Raden (2003) recently concluded that it is not clear what, if any, general conclusions can be drawn from the body of research on ethnocentrism. Some have even argued that the concept of ethnocentrism is useless (Heaven, Rajab, & Ray, 1985). These tend to be serious problems for the concept, which is widely used and is considered to be a fundamental social scientific concept. Accordingly, to ensure meaningful research on ethnocentrism, it is vital to conceptually clarify what ethnocentrism is. In order to do this, we reviewed numerous definitions of ethnocentrism, which were also very diverse and covered various ingroup, intergroup, and outgroup attitudes (cf., LeVine & Campbell’s 23 facets of ethnocentrism). We focused primarily on Sumner’s writings (1906, 1911), but also covered many other definitions in the literature. We identified three broad themes in the definitions: ethnic group self-centeredness, mere ingroup positivity, and outgroup negativity. We argue that ethnocentrism should be seen as ethnic group self-centeredness and that mere ingroup positivity and outgroup negativity should be seen as conceptually distinct potential correlates of ethnic group self-centeredness. It should be noted that Sumner used the term group in his definitions, but the context of his writing and the term ethnos in ethnocentrism imply that he was writing primarily about ethnic groups.

ETHNIC GROUP SELF-CENTEREDNESS Definitions of ethnocentrism such as ‘‘regarding one’s own race or ethnic group as of supreme importance’’ (Oxford English Dictionary, 1989, p. 424) or ‘‘this view of things in which one’s own group is the center of everything’’ (Sumner, 1906, p. 13) indicate general ethnic group self-centeredness. However, most definitions deal with some specific expressions of this general group self-centeredness. These could be classified into those dealing with intergroup and intragroup expressions of ethnocentrism. Those dealing with intergroup expressions assumed that ethnocentrism involves a belief that the ingroup is more important than other groups and involve preference, superiority, purity, and exploitativeness. In contrast, those dealing with intragroup facets assumed that the ingroup is more important than individual ingroup members and involve group cohesion and devotion.

Preference Many authors (e.g., Berry & Kalin, 1995; Herskovits, 1948; Taylor & Jaggi, 1974) assumed that ethnocentrism involves ingroup preference or a tendency to favor or like one’s own ethnic group and its members over others. Preference as an expression of group self-centeredness should be seen as seeing the ingroup as more important for oneself than outgroups, Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

873

but not as necessarily superior to outgroups. Whereas Sumner did not use the term preference in his definitions, his writings emphasized ingroup favoritism as a facet of ethnocentrism.

Superiority Ethnocentrism has been most widely defined as the belief that one’s own ethnic group is better than or superior to others (e.g., Adorno et al., 1950; Brewer & Campbell, 1976; Keller 1931/1973; LeVine & Campbell, 1972; Sumner, 1906, 1911; Sumner, Keller, & Davie, 1928). For example, according to Brewer and Campbell (1976), ‘‘self-regard or hyperevaluation of the ingroup’’ (p. 74) is one of the four central components of ethnocentrism.

Purity Sumner et al. (1928) asserted that ‘‘ethnocentrism is concerned to keep the blood pure’’ (p. 1897). Other authors also assumed that rejection of outgroups is part of ethnocentrism (e.g., Berry & Kalin, 1995; LeVine & Campbell, 1972). Group self-centeredness in this facet is expressed in the sense that one should associate primarily or even exclusively with ingroup members, whereas outgroup members should be kept at a distance or even completely shunned.

Exploitativeness The exploitativeness facet was included in the definitions of ethnocentrism offered by Adorno et al. (1950), LeVine and Campbell (1972), and Sumner (1911). It could be seen as the belief that one’s own ethnic group interests are of foremost importance and that in pursuing them little or no consideration should be given to outgroups.

Group Cohesion Sumner (1911) explicitly included ‘‘the sentiment of cohesion’’ (p. 11) as a facet of ethnocentrism, and a number of other definitions also included it (e.g., Brewer & Campbell, 1976; LeVine & Campbell, 1972). This facet involves the belief that one’s own ethnic group should be highly integrated, cooperative, and unified. Given that the group is seen as more important than any of its individual members, the needs of the group are more important than those of individual members. Accordingly, ingroup members should reject individuality for the sake of group unity.

Devotion Sumner (1911) also explicitly included ‘‘the sentiment of . . . devotion’’ (p. 11) and the American Psychological Association (2001) included ‘‘exaggerated tendency to identify with one’s own ethnic group’’ (p. 92) in their definitions of ethnocentrism. Accordingly, ethnocentrism appears to involve loyalty, dedication to one’s own ethnic group and its interests, and even readiness to sacrifice for the ingroup, stemming from the group’s central significance to its members.

OUTGROUP NEGATIVITY Outgroup negativity has been often included in definitions of ethnocentrism (e.g., Adorno et al., 1950; Brewer & Campbell, 1976; LeVine & Campbell, 1972; Sumner, 1906). Although this view has been widely emphasized, it is problematic and can be criticized on both conceptual and empirical ground. Outgroup negativity or prejudice is concerned with negative attitudes toward a particular outgroup or outgroups (Adorno et al., 1950; Crocker & Schwartz, 1985), Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

874

Boris Bizumic et al.

whereas ethnocentrism involves giving high importance to one’s own ethnic group. In this sense, ethnocentrism can be seen as the group level analogue to narcissism (LeVine & Campbell, 1972). Accordingly, like narcissism, which is conceptually distinct from dislike of other people (misanthropy), ethnocentrism appears conceptually distinct from dislike of other groups (prejudice). For example, Allport (1954) assumed that people tend to cohere together, prefer their ingroups, and segregate from outgroups, without any necessary hostility to outgroups, simply because it is more convenient and less stressful to associate with familiar and similar ingroup members than with unfamiliar and different outgroups. Research has also suggested that group self-centeredness and its expressions, such as preference, superiority, or devotion, tend to be distinct from outgroup negativity and can be unrelated to it (see Brewer, 1979, 1999, 2007; Brewer & Campbell, 1976; Cashdan, 2001; Turner, 1975, 1978). For example, Brewer and Campbell’s (1976) research with 30 ethnic groups in East Africa showed that preference for the ethnic ingroup was independent of outgroup attitudes (see also Brewer, 1981). Thus, ethnocentrism as group self-centeredness appears distinct from outgroup prejudice.

MERE INGROUP POSITIVITY Some definitions of ethnocentrism (e.g., Phillips, Penn, & Gaines, 1993; Sumner, 1906) included mere ingroup positivity or a general positive attitude toward the ingroup. This is also problematic because people may evaluate their groups positively, without necessarily giving more importance to their groups in relation to others. They may have attitudes toward outgroups that are as positive or even more positive than toward their own group. In some circumstances, ethnocentric people may even evaluate their ingroups negatively (e.g., the Nazis in the Weimer Republic). Furthermore, research on group identification, which is based on positive ingroup evaluation (cf. Brown, 2000), indicates weak positive relationships between ingroup identification and preference (Hinkle & Brown, 1990). Similarly, nationalism, which is similar to ethnocentrism with its focus on national superiority, and patriotism, which can be seen as a mere ingroup positivity, were found to be correlated, but conceptually distinct and differentially related to other variables (Kosterman & Feshbach, 1989). Accordingly, it seems that mere ingroup positivity is conceptually distinct from group self-centeredness.

RECONCEPTUALIZATION OF ETHNOCENTRISM We argue that ethnocentrism should be seen primarily as ethnic group self-centeredness, which is conceptually distinct from outgroup negativity and mere ingroup positivity. More specifically, ethnocentrism seems to consist of the four intergroup and two intragroup expressions. Positive correlations between the six facets of ethnocentrism can be expected because they are all expressions of ethnic group self-centeredness and appear to be mutually reinforcing. Additionally, stronger relationships are to be expected within the cluster of intergroup facets and within the cluster of intragroup facets than between the two clusters. Accordingly, there seem to be two main kinds of ethnocentrism: Intragroup and intergroup ethnocentrism. We define ethnocentrism as follows: Ethnocentrism is an attitudinal construct that involves a strong sense of ethnic group self-centeredness and selfimportance. This sense has intergroup and intragroup expressions. Intergroup expressions involve the central belief or sentiment that one’s own ethnic group is more important than other ethnic groups, whereas intragroup expressions involve the central belief or sentiment that one’s own ethnic group is more important than its individual members. Intergroup expressions involve preferring ethnic ingroups over outgroups, a belief in the superiority of one’s ethnic group over outgroups, the wish for ethnic purity within the ethnic ingroup, and acceptance of exploitation of outgroups when this is in ingroup’s interests. Intragroup expressions, on the other hand, involve a need for strong group cohesion and a sentiment of strong devotion to one’s own ethnic group. These six different expressions of ethnocentrism should be mutually interacting and reinforcing. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

875

We place this reconceptualization within a theory of intragroup processes and psychological group memberships, that is, self-categorization theory (Turner, Hogg, Oakes, Reicher, & Wetherell, 1987), and two theories of intergroup processes, that is, social identity theory (Tajfel & Turner, 1986) and realistic conflict theory (Sherif, 1966). Using self-categorization theory, we see intragroup ethnocentrism as an outcome of psychological group formation. When people categorize themselves as group members, their social identity is made psychologically operative. This leads to depersonalization, in which each group member perceives themselves as relatively interchangeable with other group members. Depersonalization transforms the ingroup members’ personal self-interests into group self-interest and causes them to perceive decreased discrepancies and increased similarities between themselves and other ingroup members. These processes form the basis of group cohesion and devotion, which cause each ingroup member to give more importance to the group than to its individual members, including themselves. So, intragroup ethnocentrism appears to be primary and is a natural consequence of psychological group formation. It should be related to other variables broadly concerned with ingroup positivity, such as ingroup identification, ingroup pride, positive ingroup attitudes, and personal selftranscendence. However, it is conceptually distinct from them, because intragroup ethnocentrism is specifically concerned with group self-centeredness, that is, with the idea of giving more importance to the ingroup than to its individual members. Groups are formed in the world in which there are other groups. Thus, given that group members tend to give a lot of importance to the ingroup as a result of intragroup ethnocentrism, and given that they want to see themselves in a positive light, when they engage in intergroup comparisons, they will also want to see their ingroup in a more positive light than outgroups (primarily on the dimensions that are relevant to the ingroup or that are group defining). As a result, they will favor the ingroup over outgroups and enhance any discrepancies between the ingroup and outgroups. This gives rise to the four specific intergroup facets of ethnocentrism: Preference, superiority, purity, and exploitativeness. In contrast to the two intragroup expressions, which are concerned with the submission of the individuals and their freedoms to the group, the four intergroup expressions are concerned with the relationship of the ingroup to outgroups: the ingroup is liked more than outgroups (preference), the ingroup is seen as superior (superiority), the outgroups are rejected from the ingroup (purity), and the ingroup interests take precedence over outgroup interests (exploitativeness). Intergroup ethnocentrism should be ideologically linked to other variables concerned with intergroup attitudes, such as those dealing with ingroup superiority and exploitativeness (e.g., social dominance orientation (SDO) and ingroup status), and with outgroup negativity, because often other ethnic groups, which usually have different norms and values, can be seen as dissimilar, and sometimes of a lower status, and may therefore be disliked. However, intergroup ethnocentrism, although potentially related to these constructs, should be distinct from them. According to the realistic conflict theory, both kinds of ethnocentrism would be related to threat perception because they should help in defending the ingroup against threat, and would increase under times of threat. They both should also be more prominent among people who are more sensitive to threat perception, such as authoritarians, who are more likely to perceive the world as a threatening and dangerous place (Duckitt, 2001). Additionally, given that relationships of the ingroup with outgroups can be conflictual or harmonious, it is expected that intragroup and intergroup ethnocentrism would be particularly strongly correlated under the condition of strong intergroup conflict and would be relatively weakly correlated under the condition of relative intergroup harmony.

RESEARCH PROJECT We investigated the utility of the reconceptualization in two studies. The objectives of the investigation were, first, to develop a valid and reliable measure of ethnocentrism conceptualized as group self-centeredness, which would be applicable across cultures and ethnic groups; second, to investigate the structure of ethnocentrism across cultures; and third, to investigate the relationship between ethnocentrism and external variables. The first study was conducted in New Zealand, while the second cross-national study was conducted in the United States, Serbia and Montenegro, and France. Based on the literature and theory, we formulated the following predictions: 1. There would be two distinct and correlated higher order dimensions of ethnocentrism, which would correspond to intergroup ethnocentrism (consisting of the lower order dimensions of preference, superiority, purity, and exploitativeness) and intragroup ethnocentrism (consisting of the lower order dimensions of group cohesion and devotion). Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

876

Boris Bizumic et al.

2. Intragroup and intergroup ethnocentrism would not be equivalent to ingroup positivity and outgroup negativity. 3. Intragroup expressions would correlate with constructs that are broadly concerned with intragroup positivity, such as ingroup identification, ingroup pride, positive attitudes to the ingroup, and self-transcendence (e.g., religiosity). 4. Intergroup expressions of ethnocentrism would correlate primarily with constructs concerned with ingroup superiority and exploitativeness, such as SDO and positive attitudes to war, and higher group status. Additionally, intergroup expressions would be associated with being male, because males tend to be more status-oriented and generally more intolerant of ethnic outgroups than females (see Johnson & Marini, 1998). 5. Both intergroup and intragroup expressions would correlate with threat perception and authoritarianism.

STUDY 1: NEW ZEALAND Method Participants The sample consisted of 350 undergraduates at the University of Auckland, New Zealand, who volunteered to participate in the study. Their average age was 20.46 (SD ¼ 5.84). Two hundred and forty-nine were females and 101 were males. There were 203 NZ Europeans, 49 Asians, 9 Maori, 12 Pacific Islanders, and 77 participants classified themselves as ‘‘Other.’’

Materials and Procedure The aim of the scale construction procedure was to have at least four protrait and four contrait items in each potential subscale measuring each facet of ethnocentrism. The final item pool of 174 items included items that were selected or modified from numerous measures of ethnocentrism and similar phenomena (e.g., Adorno et al., 1950; Altemeyer, 1996; Beswick & Hills, 1969; Bizumic, 2000; Chang & Ritter, 1976; Duckitt, 1992, 2000; Kosterman & Feshbach, 1989; Neuliep & McCrosky, 1997; Prothro, 1952; Selznick & Steinberg, 1969) and items written specifically to measure each facet of ethnocentrism according to explicit construct definitions. Because it was not practical to administer all 174 items on one occasion, it was decided to eliminate a number of items on the basis of content validation. Therefore, the items were given to four experts in the field of attitude measurement. The raters were presented with positive and negative definitions of the six ethnocentrism facets and were instructed to rate the items according to the following scale: 1—not relevant, 2—good, and 3—excellent. They were also allowed to use halfpoints and they were also asked if any aspects of the content were overlooked. They mentioned that all the relevant aspects of the content were included and that the items represented the construct’s content domain correctly. Only the items that were rated with the highest marks (i.e., 2.5 or 3) by the raters were selected for inclusion in the final scale. Additionally, item similarity and the number of positive and negative items were also taken into account in selecting the final 60 items, with 30 positive and 30 negative items (each of the six facets of ethnocentrism was measured by five positive and five negative items). These items were also given to seven psychology students in order to ensure that the items would be understandable and clear to potential participants. The students did not indicate any significant problems with the items. The scale was included in a questionnaire, together with questions about participants’ background and other measures. The students who volunteered to participate in the study completed the questionnaire at the end of laboratory classes. The questionnaire asked participants to rate their agreement or disagreement with the items of most scales (unless noted otherwise) on a nine-point Likert scale (
4—very strongly disagree, 0—unsure/neutral, þ4—very strongly agree). The description of the other measures, some of them shortened because of time constraints, and their as follow. All the analyses were based on the dataset in which isolated missing values (.98%) for all the variables in the questionnaire, except ethnicity, religion, age, and gender, were estimated using expectation maximization. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

877

1. A shortened SDO Scale (a ¼ .71), consisting of three protrait and three contrait items from the SDO6 Scale (Sidanius & Pratto, 1999). 2. A shortened RWA Scale (a ¼ .63) consisting of four protrait and four contrait items (Altemeyer, 1996). 3. An Ethnocultural Identification Scale (a ¼ .65), with two slightly modified items taken from Bizumic (2000): ‘‘I have a sense of identification with my culture or ethnic group’’ (protrait) and ‘‘I feel NO attachment to my culture or ethnic group’’ (contrait). 4. An Ethnic Pride Scale (a ¼ .70), which was an adapted version of a nationalism scale by Mummendey, Klink, & Brown (2001). The five items of the scale asked participants to rate to what extent they were proud of their ethnic group’s ‘‘history,’’ ‘‘political power,’’ ‘‘economic success,’’ ‘‘technical and scientific success,’’ and ‘‘cultural achievements’’ (1—not at all, 9—very proud). 5. A perception of Ethnocultural Threat Scale (a ¼ .52), consisted of two items (taken from Bizumic, 2000): ‘‘I fear that my culture and its most important values are in danger’’ (protrait) and ‘‘I don’t believe anything is likely to weaken or harm my culture or the identity of my people’’ (contrait). 6. An Ethnic Power/Status Scale (a ¼ .90), constructed to assess participants’ perception of ethnic group power/status, asked participants to rate the extent to which others see their ethnic group as being ‘‘economically strong,’’ ‘‘politically strong,’’ and ‘‘high in status’’ (1—negligible, 9—extreme). Past research has shown that people tend to rate their group’s status accurately (e.g., Sidanius & Pratto, 1993). 7. A shortened Attitude Toward War Scale (a ¼ .63) by Stagner (1942) with two protrait items (‘‘war is the only way to right tremendous wrongs’’ and ‘‘the desirable results of war have not received the attention they deserve’’) and two contrait items (‘‘war is a futile struggle resulting in self-destruction’’ and ‘‘there is no conceivable justification for war’’). 8. A single item asked participants to rate their level of religiosity (0—not at all, 8—very religious). 9. A feeling thermometer measure asked participants to indicate the warmth of their feelings toward New Zealanders and several outgroups: Arabs, Chinese, Romanians, Americans, Africans, and French. Similar feeling thermometer scales have been successfully used to study intergroup relations (e.g., Stangor, Sullivan, & Ford, 1991). 10. Four shortened and balanced scales by Duckitt (2001) were included to measure attitudes toward New Zealand ethnic majority group, NZ Europeans (four items, a ¼ .57), and minority groups, that is, Maori (two items, a ¼ .75), Pacific Islanders (two items, a ¼ .59), and Asians (two items, a ¼ .68). Some of the illustrative items are: ‘‘Maori are still being very unfairly treated in this country,’’ and ‘‘Asian culture has helped to make New Zealand a far more interesting place.’’ Before completing the scales in the study, participants were asked to indicate their ethnic group membership (e.g., NZ European/Pakeha, Asian, and Maori). We expected that this would prime their own specific ethnic group membership while completing the measures of ethnocentrism and ethnic identification, threat, power, and pride, given that these measures were developed to be culture-general.

Results and Discussion The measure of ethnocentrism had a high a (.91), but its mean inter-item correlation of .16 was rather low. This, together with the existence of stronger correlations between the items devised to measure each component construct and weaker correlations between the items devised to measure different component constructs, suggested multidimensionality. Therefore, the following analyses focused on the six hypothesized subscales. Items that had corrected item-to-subscale correlations lower than .30 (Kline, 2000) or correlations with other scales higher than their corrected item-to-subscale correlations were discarded. In this way, four items from the Preference Subscale, one from the Devotion Subscale, and one from the Group Cohesion Subscale were discarded. The resulting subscales had good as: preference (.73), superiority (.81), purity (.89), exploitativeness (.84), group cohesion (.79), and devotion (.85). The correlations between the subscales (see Table 1) gave support to the proposed structure, indicating the expected subscale clusters of intragroup and intergroup ethnocentrism expressions. Although the correlations were positive, group cohesion did not correlate significantly with preference, purity, or exploitativeness. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

878

Boris Bizumic et al.

Table 1. Intercorrelations of the Ethnocentrism Subscales in the New Zealand sample Variable 1. 2. 3. 4. 5. 6.

Preference Superiority Purity Exploitativeness Group cohesion Devotion

1

2

3

— .44 .55 .34 .07 .20

— .61 .54 .15 .30

— .55 .08 .16

4

— .02 .23

5

6

— .47

—

Note: N ¼ 350.  p < .01; p < .001.

To investigate the structure of ethnocentrism further, a confirmatory factor analysis (CFA), using maximum likelihood procedure with a covariance matrix of item parcels, was performed. Item parcels are recommended because it is quite difficult to confirm a factor structure when more than 5–8 items load on one latent variable (Frank & Widaman, 1995), and because item parcels tend to be more reliable indicators (Marsh, 1994) and have a distribution that is more likely to be approximately normal (Loo, 2001). In addition, item parcels can help control for acquiescence. The item parcels were developed using Brooke, Russell, and Price’s (1988) approach of successively combining items with highest and lowest loadings in separate monofactorial models, with the added proviso that each parcel contained a protrait and a contrait item to control for acquiescence. However, three item parcels were not fully balanced because of the discarded items. There were at least three item parcels per factor. Theoretical reasoning suggested five plausible models for testing:

Model A A one-factor model, with all item parcels modeled to load on one factor. This model is based on the assumption that ethnocentrism is a unitary construct and that the six facets of ethnocentrism are not distinct. This is the most parsimonious model of ethnocentrism.

Model B A two-correlated-factor model, with item parcels measuring intergroup expressions of ethnocentrism modeled to load on one factor, and those measuring intragroup expressions modeled to load on the other factor. This is the second most parsimonious model of ethnocentrism. This model is based on the assumption that intragroup and intergroup ethnocentrism are conceptually distinct, but that the facets comprising intragroup ethnocentrism and those comprising intergroup ethnocentrism are equivalent.

Model C A six-correlated-factor model, with item parcels from each subscale modeled to load on their respective six factors. This is the least parsimonious model of the five tested models. However, this model was tested because the following two hierarchical models could be tested only if this model fitted the data well.

Model D A one second-order-factor model that accounts for the correlations between the six first-order factors by one second-order factor. This model assumes that at a lower level of generalization there are six dimensions of ethnocentrism, but at a higher level, ethnocentrism forms one factor. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

879

Model E A two-correlated-second-order-factor model, with one second-order factor accounting for the correlations between the two first-order factors indicating group cohesion and devotion, and the other accounting for the correlations between the four first-order factors representing preference, superiority, purity, and exploitativeness. This model assumes that at the lower level there are six dimensions of ethnocentrism, which can be organized, at a higher level, into two general kinds of ethnocentrism: Intragroup ethnocentrism and intergroup ethnocentrism. This is the model that corresponds to the definition that we proposed. However, if any of the more parsimonious models fits the data well, then this model should be rejected in favor of the more parsimonious model. The x2, df, and significance levels were applied to estimate objective model fit. However, given that the x2-test can be aversely affected by sample size, a number of more robust subjective fit indices were used. Those were Bentler’s (1990) CFI, Bentler and Bonett’s (1980) NNFI, Bentler’s (1995) SRMR, the RMSEA (Steiger, 1990), and the x2/df ratio. Criterion values for satisfactory fit were: CFI and NNFI should be greater than .90 (Bentler, 1992; Bentler & Bonnet, 1980), SRMR values (Hu & Bentler, 1999) and RMSEA values (Browne & Cudeck, 1993) should be generally less than .10 (smaller values indicate a better fit), and a x2/df ratio value should be less than 3 (Carmines & McIver, 1981). Higher values of CFI, NNFI, and the x2/df ratio, and lower values of SRMR and RMSEA indicate a better model fit. If CFI and NNFI values are less than .90, RMSEA and SRMR are higher than .10, and the x2/df ratio is higher than 3, then the fit of the model is generally seen to be poor. The best model is the one that has the best fit indices, which also need to be acceptable. However, if two models have very similar fit indices, the one that is more parsimonious should be chosen as a preferable model. The tested solutions reached convergence, but the residual variance of one first-order factor, representing devotion, had to be fixed to a small value of .05 (see Gerbing & Anderson, 1987) because the initial two-correlated-second-order-factor solution was improper. Table 2 shows that the six-factor model had the most satisfactory fit to the data having overall best fit indices. However, it was closely followed by the two-correlated-second-order-factor model, which had all the acceptable fit indices, which were either equal to (as was the case with RMSEA or NNFI) or close to the six-correlated factor model’s fit indices. Although the one second-order factor model had acceptable fit indices, all of them, apart from CFI were worse than the two second-order factor model. Marsh and Hocover’s (1985) target coefficient (T), which is the ratio of the first-order model’s x2 value to the second-order model’s x2 value, was calculated. Its upper limit is 1 and higher values indicate an increased efficiency of the second-order models. Its value was high (T ¼.95) indicating that the twocorrelated-second-order-factor model should be accepted as the more parsimonious and theoretically meaningful model. The loadings of the first-order factors on their second-order factors were strong (see Figure 1), with devotion seeming to be an almost perfect expression of intragroup ethnocentrism. The correlation between the two higher-order factors was moderate. These analyses tested the structure of ethnocentrism, but they did not test whether ethnocentrism was distinct from ingroup positivity and outgroup negativity, or whether it was equivalent to them. To investigate that, one needs to compare models in which items measuring ingroup positivity, outgroup negativity, intragroup ethnocentrism, and intergroup ethnocentrism load on the same factors to those in which they load on different factors. Accordingly, three additional models were tested. A CFA, using maximum likelihood procedure with a covariance matrix of item parcels, was used in the ethnic majority subsample, NZ Europeans (N ¼ 203), which was the only ethnic group with a sufficient number of participants for the analysis. Given the reduced number of participants and the introduction of more variables to the models, multi-item parcels were used in order to compensate for the decreased number of participants to the number of parameter estimates ratio (which Table 2. Goodness of fit indices for five models of the Ethnocentrism Scale structure in the New Zealand sample x2

df

x2/df

CFI

NNFI

SRMR

RMSEA

3407.82 1917.70 548.050 664.119 573.995

299 298 284 293 293

11.40 6.44 1.93 2.27 1.96

.82 .89 .98 .97 .97

.80 .88 .97 .96 .97

.137 .088 .049 .082 .057

.173 .125 .052 .060 .052

Model A. One factor B. Two factors C. Six factors D. One second-order factor E. Two second-order-factors

Note: N ¼ 350. The residual variance of one first-order factor, devotion, was fixed to .05. All x2 statistics are significant at p < .001. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

880

Boris Bizumic et al.

Figure 1. The two-correlated-second-order-factor model of ethnocentrism, with intergroup ethnocentrism (E1) and intragroup ethnocentrism (E2) as second-order factors, in the New Zealand sample (N ¼ 350). Manifest indicators and their loadings are not shown. p < .001

can produce a poorer fit of the models). The previous CFAs indicated that intergroup and intragroup ethnocentrism form two higher order dimensions. Accordingly, four multi-item parcels, which included items measuring intergroup ethnocentrism, were formed to represent intergroup ethnocentrism, and other four multi-item parcels, including items measuring intragroup ethnocentrism, were formed to represent intragroup ethnocentrism. There were four multi-item parcels that included items measuring outgroup negativity (using the Likert scale items measuring negative attitudes to Maori, Asians, and Pacific Islanders, and the feeling thermometer measures of negative attitudes to Arabs, Chinese, Romanians, Americans, Africans, and French), and given that there were four items measuring attitudes to NZ Europeans, two item parcels were included to represent ingroup positivity. All item parcels were created on the basis of Brooke et al.’s (1988) method. Three relevant models were tested:

Model A A one-factor model, with item parcels modeled to load on one factor. This model rests on the theoretical assumption that ethnocentrism, ingroup positivity, and outgroup negativity are equivalent and not conceptually distinct from each other.

Model B A two-correlated-factor model, with item parcels measuring intergroup ethnocentrism and outgroup negativity modeled to load on one factor, and those measuring intragroup ethnocentrism and ingroup positivity modeled to load on the other factor. This model rests on the theoretical assumption that intragroup ethnocentrism and ingroup positivity are equivalent to each other (i.e., they represent one dimension), that intergroup ethnocentrism and outgroup negativity are equivalent to each other (i.e., they represent one dimension), but that these two dimensions are different from each other.

Model C A four-correlated-factor model, with item parcels measuring intergroup ethnocentrism, intragroup ethnocentrism, outgroup negativity, and ingroup positivity modeled to load on their respective four factors. This model rests on the Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

881

Table 3. Goodness of fit indices for three models testing the conceptual distinctiveness of ingroup positivity, outgroup negativity, intragroup ethnocentrism, and intergroup ethnocentrism in the NZ European sample Model A. One factor B. Two factors C. Four factors

x2

df

x2/df

CFI

NNFI

SRMR

RMSEA

852.09 546.08 98.09

77 76 71

11.07 7.19 1.38

.74 .84 .99

.69 .81 .99

.158 .122 .045

.223 .175 .043

Note: N ¼ 203. All x2 statistics are significant at p < .001.

theoretical assumption that although the four constructs are correlated with each other, they are not equivalent and would form four distinct dimensions. For example, this model include items measuring ingroup positivity loading only on its own factor, but not on other factors, such as intragroup ethnocentrism or intergroup ethnocentrism. The same fit indices and criterion values for satisfactory fit that were used to test the structure of ethnocentrism were used to test the fit of these models. All tested solutions reached convergence. Table 3 shows that the one factor and two factors models did not have acceptable fit indices, whereas the fit indices were all acceptable for the four factors model, suggesting that this model is the best fit for the data. All the correlations between the four latent variables were substantially below unity (the mean correlation between the factors was .34, and between the corresponding manifest variables .28). This indicates that given that the latent variables were measured without any error, the four variables were not equivalent. Kline (2004) argues that correlations between factors that are higher than .85 indicate that only one concept underlies these factors; otherwise, the concepts are not equivalent. The correlations between the latent variables were much lower than .85. These CFAs suggest that the two kinds of ethnocentrism, conceptualized as ethnic group selfcenteredness, with intragroup and intergroup expressions, seem to be related, but are not equivalent, to ingroup positivity and outgroup negativity. To simplify analyses of the relationships between ethnocentrism and the other variables, composite measures of intergroup ethnocentrism (E1), as the sum of the items measuring the four intergroup ethnocentrism expressions, and intragroup ethnocentrism (E2), as the sum of the items measuring the two intragroup ethnocentrism expressions, were created. Both higher order kinds of ethnocentrism had good psychometric properties in the total sample (E1: a ¼ .92; E2: a ¼ .83) and in the ethnic majority subsample (E1: a ¼ .93; E2: a ¼ .83). In addition, two other variables were created because the attitudes toward ethnic outgroups on the three Likert scales and on the feeling thermometer measure were strongly intercorrelated for NZ Europeans, the ethnic majority group. One represented the sum of attitudes toward Maori, Pacific Islanders, and Asians on Likert scales (a ¼ .76). The other represented the sum of attitudes toward Arabs, Chinese, Romanians, Americans, Africans and French on the feeling thermometer measure (a ¼ .87). These variables were coded so that high scores indicated more negative attitudes. We used the composite measures because they summarized the relationships between ethnocentrism and outgroup attitudes, given that the relationships of the two kinds of ethnocentrism with specific outgroup attitudes and with the composite measures were very similar. Although Likert and feeling thermometer measures did intercorrelate, we formed two separate measures because the feeling thermometer measures were content free. The results of multiple regression analyses for the two higher order dimensions of ethnocentrism will be reported, because they summarize the relationships of the six ethnocentrism subscales with the external criteria and are similar to the findings of the correlational analyses. They also help disentangle the unique influences of E1 and E2 on the external criteria given that E1 and E2 were correlated. First, each external criterion was simultaneously regressed on both E1 and E2 in the total sample. It was expected that E1 would be primarily associated with SDO, positive attitudes to war, higher group power, and being male. As seen in Table 4, in the total sample, E1 was indeed positively associated with three of these constructs, but it was not related to power, and unexpectedly it was related to ethnic pride. E2 was expected to relate to identification, ingroup pride and religiosity. Table 4 shows that it was the case. Finally, although it was expected that E1 and E2 would be both related to threat perception and RWA, they were both related to RWA, but only E2 was related to threat. Accordingly, most associations were expected, but those regarding ethnic pride, threat, and power, were not.1 Table 4 also indicates that ethnocentrism among NZ Europeans tended to predict most of the criterion variables even when controlled for ingroup positivity (which included the sum of all the measures of positive ingroup attitudes) and 1 It should be noted that tests for the interactions between E1 and E2 showed that no consistent and meaningful interactions emerged, with almost every analysis giving non-significant interaction terms.

Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

882

Boris Bizumic et al.

Table 4. b coefficients of the simultaneous multiple regression of the external criteria on E1 and E2 in the New Zealand total sample (N ¼ 350) and step 2 of the sequential multiple regression of the external criteria on anti-outgroup attitudes, pro-ingroup attitudes, E1, and E2 in NZ European subsample (n ¼ 203) RWA Total sample E1 E2 NZ Europeans Anti-outgroup Pro-ingroup E1 E2 DF

Ethnic pride

SDO

War

Gender

.26 .41

.19 .17

.52
.03

.25
.07

.29
.08

.13
.07 .27 .34 25.99

.32
.20 .19 .11 5.71

.18
.10 .55 .07 33.31

.13
.13 .26
.10 5.12

.02
.08 .34
.11 8.75

Ethnic identification .01 .32 .02
.05 .06 .20 4.37

Ethnic threat

Religiosity

Ethnic power


.13 .22


.05 .29

.06
.25


.02 .07
.05 .12 1.20


.02
.07 .09 .21 5.74

.01
.09 .03
.13 1.51

Note: Positive bs in the gender column indicate that males were higher. For NZ Europeans, anti-outgroup and pro-ingroup attitudes were entered in the first step, and E1 and E2 in the second. DF ¼ F-value change.; p < .05; p < .01; p < .001.

outgroup negativity (which included the sum of all measures of negative outgroup attitudes). Outgroup negativity and ingroup positivity were entered in step 1 and the two ethnocentrism variables in step 2 of the regression analyses. These analyses were included to give additional support to the claim that ethnocentrism is distinct from ingroup positivity and outgroup negativity. As can be seen in Table 4, ethnocentrism contributed to significant F-value changes in most analyses, suggesting that it was able to predict the criterion variables over and above ingroup positivity and outgroup negativity. The general pattern of the findings was largely similar to those for the total sample, but given a smaller subsample size, and the fact that the four constructs tended to have moderate to strong correlations (perhaps affecting a certain amount of multicollinearity), the findings were largely, but not completely, consistent. The relationship of ethnocentrism with pro-ingroup and anti-outgroup attitudes was investigated for NZ Europeans using multiple regression analyses. The relationships of ethnocentrism with the measures of generalized outgroup negativity are presented here because they summarize the relationships of the six dimensions of ethnocentrism with specific outgroup attitudes. It was expected that E1 would be associated with outgroup negativity, and E2 with ingroup positivity. Table 5 shows that indeed E1 was associated with outgroup attitudes and E2 with pro-ingroup attitudes. Unexpectedly, E1 was also related to pro-ingroup attitudes, perhaps because of the content of some items (e.g., ‘‘the most important reason why Pakeha are better off today is because of the way in which they stole the land and oppressed the indigenous people in this country’’), which involved both ingroup and intergroup attitudes. Table 5 also reports squared semipartial correlations, which present the proportion of total variation in the response variable that is uniquely explained by one predictor variable (e.g., E1) controlling for the impact of another (E2). The mean squared-semipartial correlation for the relationship of E1 and outgroup attitudes was .22, and for the relationship of E2 and ingroup attitudes was .01. To sum up, most expectations about the structure of ethnocentrism and its relationships with the external criteria were confirmed in the study. Even the deviation from the expectations appeared comprehensible. The positive association of E2, but not E1, with threat perception and lower ethnic power/status perception (which is threatening) could be explained by

Table 5. b coefficients and squared-semipartial correlations of the simultaneous multiple regression of positive ingroup and negative outgroup attitudes on E1 and E2 in the NZ European subsample Variable E1 (b) (DR2) E2 (b) (DR2)

Pro-ingroup (L) 

.26 .06 .21 .04

Anti-outgroup (L) 

.52 .24 .16 .02

Anti-outgroup (FT) .47 .20
.05 .00

Note: n ¼ 203. L ¼ attitudes on Likert scales. FT ¼attitudes on the Feeling Thermometer Scale. Outgroups: Maori, Asians, and Pacific Islanders (L); Arabs, Chinese, Romanians, Americans, Africans, and French (FT). DR2 ¼ squared semi-partial correlation.; p < .01; p < .001. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

883

the defensive functions of group cohesion and devotion, whereas people’s pride in their ethnic group may be based on both intragroup and intergroup appraisal. The next step was to test the validity of the proposed reconceptualization across different cultural contexts.

STUDY 2: THE UNITED STATES, SERBIA, AND FRANCE The cross-cultural investigation included several new or modified items to replace the items discarded in the first study and two items that did not seem culturally generalizable. The original hypotheses again guided the investigation, but those relating to the associations of ethnocentrism with ethnic group power/status perception, threat perception, and ethnic pride were modified. The modifications were based on the findings of the first study and their theoretical interpretation. Thus, it was expected that both E2 and E1 would relate to ethnic pride, whereas E2, but not E1, would relate to the perception of threat to one’s ethnic group and the perception of one’s ethnic group’s (lower) power/status.

Method Participants Participants in the US were 212 undergraduates at the University of Tennessee at Chattanooga. The average age was 19.65 (SD ¼ 4.18). There were 145 females and 67 males in the sample, and 152 of them were members of the ethnic majority group, Whites/Anglos. There were also 43 Blacks/African Americans, 3 were Hispanics/Latinos, 5 Asians, and 9 others. Two hundred and eight undergraduates at the University of Belgrade participated in the study in Serbia. There were 140 females and 68 males, their average age was 20.12 (SD ¼ 1.92), and 198 participants were members of the ethnic majority group, Serbs, and 9 were members of other ethnic groups, whilst one participant did not indicate ethnic membership. Participants in France were 279 students at the University of Paris X, Nanterre. There were 141 males and 138 females and the average age of participants was 19.81 (SD ¼ 2.96). The sample consisted of 226 members of the French ethnic majority group, 11 Maghrebians, 5 Africans, 7 West Indians, 19 of mixed background, 9 others, and 2 students who did not indicate their ethnicity.

Materials and procedure The questionnaire consisted of 64 items measuring ethnocentrism, external criteria, and questions asking for background information, with the same rating scales used in the scale construction study. The questionnaire was translated into Serbian and French using the back-translation procedure (see Brislin, 1980). US participants completed the questionnaire outside their classes in exchange for course credit, and Serbian and French participants voluntarily completed the questionnaire during their classes. The external criteria measures included in the questionnaire and their as are reported below. The calculation of these as as well as other analyses were based on the datasets with isolated missing values (.73% in the US, .45% in Serbia, and .15% in France) estimated using expectation maximization. 1. A shortened, eight-item balanced version of the SDO Scale (Sidanius & Pratto, 1999). Its a coefficients were satisfactory (US: .75, Serbia: .74; France: .65). 2. A shortened, 10-item balanced version of the RWA Scale (Altemeyer, 1996). The as were satisfactory in the US (.75), Serbia (.77), and France (.73), where one item was discarded because its corrected item-total correlation was nonsignificant. 3. A shortened, two-item balanced Ethnocultural Identification Scale (Bizumic, 2000), consisting of the same two items used in Study 1. Given the scale’s length, it performed well in the US (.61), Serbia (.79), and France (.65). 4. An Ethnic Pride Scale, based on Mummendey et al.’s (2001) scale, with the same items used in Study 1. The as were generally satisfactory (US: .84, Serbia: .61, France: .68). Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

884

Boris Bizumic et al.

5. Because the a for the perception of Ethnocultural Threat Scale in Study 1 had been low, two protrait items (‘‘I fear that my ethnic or cultural group is in danger,’’ ‘‘I think that some important aspects of our culture are under threat’’) and two contrait items (‘‘at this time there is nothing that poses a serious threat to our people,’’ ‘‘I don’t think that anything bad can happen to my cultural or ethnic group’’) were devised to assess this construct. The as were satisfactory in the US (.67), Serbia (.71), and France (.79). 6. An Ethnic Power/Status Scale was included to measure participants’ perception of their ethnic group power/status. The same items were used in Study 1. The as were high in each country (US: .94, Serbia: .87, France: .82). 7. A shortened Attitudes Toward War Scale (Stagner, 1942), with the same four items used in Study 1. The scale’s a was satisfactory given the scale’s length in the US (.68), Serbia (.63), and France (.58). 8. A single item rating of religiosity. 9. A Feeling Thermometer Scale was used to measure attitudes toward the ethnic ingroup and various outgroups. The scale assessed participants’ attitudes toward Africans, French, Romanians, New Zealanders, Serbs, Arabs, Chinese, and Americans in Serbia and in France, and toward US Blacks, US Asians, US Whites, US Latinos, French, Romanians, Serbs, Arabs, Chinese, and New Zealanders in the US. 10. Attitudes toward the ingroup and three outgroups were assessed with multi-item Likert scales adapted from existing measures. The items used in Serbia and France were adapted from Duckitt’s (2001) generalized measure of ethnic attitudes previously used in South Africa (e.g., ‘‘I have a very positive attitude to the (ethnic group’s) people’’ and ‘‘I can understand people having a negative attitude to the (ethnic group’s) people’’—contrait). One additional item was devised for the present investigation (‘‘I think that the (ethnic group’s) people are capable of doing some very negative things’’). The items used in the US were adapted from Duckitt’s (2001) measure of ethnic attitudes in New Zealand, such as ‘‘In general, Whites are treating other ethnic groups very fairly’’ and ‘‘Latinos just don’t try hard enough to get ahead in this country’’ (contrait). Unless indicated all these scales consisted of three contrait and three protrait items. The as are included in parentheses. The scales assessed the attitudes toward French (.69), Maghrebians (.78), Asians (.75), and Germans (.81) in France, and toward Serbs (.71), Croats (.69), Hungarians (.70), and Albanians (.81) in Serbia. They assessed the attitudes toward Whites/Anglos, measured by four protrait and four contrait items (.79), Asians (.78), and Latinos (.69, with one item having to be excluded because of its non-significant item-total correlation) in the US. The Modern Racism Scale by McConahay, Hardee, and Batts (1981) was used to assess attitudes toward Blacks in the US, with one positive and six negative sentiment items (.85). Participants were asked, at the beginning of the questionnaire, to indicate their ethnic group membership (e.g., Serbian, Croatian, or Montenegrin in Serbia, White/Anglo American, African American, or Latino American in the US, French, Arab, or Asian in France). This was done to make participants’ ethnic group membership salient while answering culturegeneral items, included in the measures of ethnocentrism and ethnic pride, threat, power/status, and identification.

Results and Discussion The as for the Ethnocentrism Scale were high in the US (.94), Serbia (.96), and France (.93), but the mean inter-item correlations were relatively low in the US (.19) and France (.19), though less so in Serbia (.26). The correlation matrix of the ethnocentrism items indicated larger correlations between the items of the same subscale than those of different subscales. This seemed to indicate that the measure is multidimensional. Accordingly, the following analyses focused on the subscales. Items for which corrected item-to-subscale correlations were lower than .30 or for which the items’ correlation with any other subscale was higher than its own corrected item-to-subscale correlation were discarded as weak items, but special care was taken to ensure that the subscales remained balanced. Three items of the Preference Subscale, two of the Exploitativeness Subscale, and one of the Group Cohesion Subscale were excluded. The resulting scale had 58 items, with all subscales balanced and with at least 8 items per subscale (see Appendix A for the items). Their as were satisfactory in each country: Preference (US: 81, Serbia: .82, France: .80), superiority (US: .87, Serbia: .88, France: .88), purity (US: .91, Serbia: .92, France: .88), exploitativeness (US: .84, Serbia: .86, France: .74), group cohesion (US: .77, Serbia: .88, France: .80), and devotion (US: .84, Serbia: .89, France: .75). Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

885

The subscales’ intercorrelations were positive and significant, with only two nonsignificant correlations in the US sample for group cohesion (see Table 6). The intercorrelations seemed to support the hypothesized structure and CFA was conducted to investigate this further. Two kinds of CFA were performed: A single-sample CFA, which tested the structure of ethnocentrism in each sample, and a multi-sample CFA, which tested the measurement invariance across the three samples. The same procedure, models, fit indices, and criteria for good fit that were used in Study 1 were used again for the single-sample CFAs. Their findings (see Table 7) were also consistent with Study 1. Both one factor and two factors models had poor fit to the data, but the fit of the six factors model, the one second-order factor model, and the two secondorder factors model were generally acceptable. Overall, the six factors model had the best fit to the data as it had the lowest x2/df ratio, SRMR, and RMSEA values in each country, and the lowest NNFI value in the US. This model was closely followed by the two second-order factors model, which had higher x2/df ratio, SRMR, and RMSEA values in each country, and equally good NNFI and CFI values in each country (except the NNFI value, which was slightly smaller in the US). Although the more parsimonious one second-order factor model had generally good fit indices (apart from the SRMR value in France, which indicated a poor fit), the two second-order factor model had the lower x2/df ratio, SRMR, and RMSEA values in each country than the one second-order factor model, and a higher CFI value in the US, and a higher NNFI value in Serbia, suggesting that the two second-order was superior. In addition, the differences between the fit indices of the two second-order factor model and the six factor model were much smaller than the differences between the fit indices of the one second-order factor model and the six factor model. So, the two-second-order factor model was accepted as the most parsimonious and theoretically meaningful. Additionally, Marsh and Hocover’s (1985) T coefficient for the two second-order model was quite high in all three samples (US: .94, Serbia: .97, France: .94), suggesting that this hierarchical model was appropriate. Figures 2–4 show that the loadings of the first-order factors on their second-order factors were generally strong. The correlations between the two second-order factors were moderate in the US, very powerful in Serbia, and strong in France. Devotion loaded more strongly on its second-order factor than did group cohesion in each sample, suggesting that it was a clearer expression of intragroup ethnocentrism.

Table 6. Intercorrelations of the Ethnocentrism Subscales in the US, Serbian (SE), and French (FR) samples Variable 1. Preference US SE FR 2. Superiority US SE FR 3. Purity US SE FR 4. Exploitativeness US SE FR 5. Group Cohesion US SE FR 6. Devotion US SE FR

1

2

3

4

5

6

— — — .53 .61 .61

— — —

.53 .62 .65

.65 .57 .58

— — —

.45 .42 .55

.64 .33 .54

.67 .50 .65

— — —

.33 .53 .32

.25 .51 .30

.06 .40 .23

.11 .21 .24

— — —

.47 .59 .39

.27 .56 .40

.15 .50 .27

.29 .26 .31

.51 .66 .38

Note: US: N ¼ 212. SE: N ¼ 208. FR: N ¼ 279.; p < .05; Copyright # 2008 John Wiley & Sons, Ltd.



p < .01;

— — —



p < .001. Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

886

Boris Bizumic et al.

Table 7. Goodness of fit indices for five models of the Ethnocentrism Scale structure in the US, Serbian (SE), and French (FR) samples Model

x2

A. One factor US 3286.06 SE 3219.90 FR 2466.97 B. Two factors US 1780.21 SE 2129.08 FR 1687.91 C. Six factors US 695.69 SE 534.95 FR 456.50 D. One second-order factor US 809.82 SE 596.24 FR 513.00 E. Two second-order factors US 737.43 SE 551.08 FR 484.13

df

x2/df

CFI

NNFI

SRMR

RMSEA

377 377 377

8.72 8.54 6.54

.85 .89 .90

.84 .88 .89

.147 .116 .106

.191 .191 .141

376 376 376

4.73 5.66 4.49

.90 .92 .93

.90 .92 .93

.107 .097 .086

.133 .150 .112

362 362 362

1.92 1.48 1.26

.97 .99 .99

.97 .99 .99

.070 .048 .048

.066 .048 .031

371 371 371

2.18 1.61 1.38

.96 .99 .99

.96 .98 .99

.107 .068 .062

.075 .054 .037

370 370 370

1.99 1.49 1.31

.97 .99 .99

.96 .99 .99

.087 .058 .054

.069 .049 .033

Note. US: N ¼ 212. SE: N ¼ 208. FR: N ¼ 279.; All x2 statistics are significant at p < .001.

Figure 2. The two-correlated-second-order-factor model of ethnocentrism, with intergroup ethnocentrism (E1) and intragroup ethnocentrism (E2) as second-order factors, in the US sample (N ¼ 212). Manifest indicators and their loadings are not shown.  p < .001. [Correction made here after initial online publication.] Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

887

Figure 3. The two-correlated-second-order-factor model of ethnocentrism, with intergroup ethnocentrism (E1) and intragroup ethnocentrism (E2) as second-order factors, in the Serbian sample (N ¼ 208). Manifest indicators and their loadings are not shown.  p < .001. [Correction made here after initial online publication.]

Figure 4. The two-correlated-second-order-factor model of ethnocentrism, with intergroup ethnocentrism (E1) and intragroup ethnocentrism (E2) as second-order factors, in the French sample (N ¼ 279). Manifest indicators and their loadings are not shown.  p < .001. [Correction made here after initial online publication.] Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

888

Boris Bizumic et al.

The extent to which ethnocentrism, conceptualized as ethnic group self-centeredness, was distinct from ingroup positivity and outgroup negativity was tested in each country. Similarly to Study 1, given a smaller number of participants and an increasing complexity of the models a smaller number of item parcels was used for each measure. They were based on items measuring ethnocentrism and items of the Likert scales and feeling thermometer scales measuring intragroup positivity and outgroup negativity. Four item parcels corresponding to the measures of outgroup negativity, intergroup ethnocentrism, and intragroup ethnocentrism were used. For ingroup positivity, there were four item parcels for US Whites, and given a smaller number of items, three item parcels were included for Serbs and French. The same three models that were tested in Study 1 were tested. The four-factor model had all the acceptable fit indices, and was always superior to the two-factor and one-factor models, which had very poor fit indices (see Table 8). The mean correlation between the four factors was .37 (.42 for US Whites, .47 for Serbs, and .21 for French). This represents a moderate relationship given that these correlations are between latent variables and corrected for measurement error. The mean correlation between the equivalent measured variables across the three countries was .27 (.28 for US Whites, .36 for Serbs, and .18 for French). These findings also suggest that the four constructs, although correlated, are not equivalent to each other. Next, multisample CFA was used to assess measurement invariance. This procedure tests if a scale performs equivalently and demonstrates the same structure across multiple groups (e.g., Marsh, 1994). The widely used procedure for testing measurement invariance progresses from the least constrained to the most constrained models (e.g., Marsh, 1994; Marsh & Hocover, 1985). Although the x2 difference test is usually used to establish the difference between the models in hierarchy, this test is biased against more constrained models when dealing with larger samples, as was the case here (N ¼ 699). For this reason, we focused primarily on comparing subjective fit indices and followed Marsh’s (1994) guidelines. He argued that a reasonable level of measurement invariance across groups is established when the fit indices are acceptable and the differences in the fit indices between the two consecutive models in hierarchy are very small. The same fit indices and criterion values as used in the other CFAs in this research were used here, with the exception of SRMR, which was replaced with CAIC (see Steenkamp & Baumgartner, 1998). CAIC, like NNFI and RMSEA, takes into account the lack of model parsimony, with smaller values indicating better fit. Preliminary analyses indicated a reasonable level of measurement invariance for the six-factor measurement model. Thus, the following two-correlated-second-orderfactor models were tested:

Model A Baseline model, in which all the parameters were freely estimated.

Table 8. Goodness of fit indices for three models testing the conceptual distinctiveness of ingroup positivity, outgroup negativity, intragroup ethnocentrism, and intergroup ethnocentrism in the US White, Serbian, and French ethnic majority group subsamples Model A. One factor US SE FR B. Two factors US SE FR C. Four Factors US SE FR

x2

df

x2/df

CFI

NNFI

SRMR

RMSEA

1128.86 1308.19 1632.54

104 90 90

10.85 14.54 18.14

.73 .80 .72

.69 .77 .67

.183 .149 .185

.255 .262 .276

768.88 677.36 1316.05

103 89 89

7.46 7.61 14.79

.81 .89 .78

.78 .87 .74

.154 .120 .165

.207 .183 .248

225.98 226.94 240.41

98 84 84

2.31 2.70 2.86

.95 .96 .96

.94 .95 .95

.089 .071 .062

.093 .093 .091

Note: US: N ¼ 152. SE: N ¼ 198. FR: N ¼ 226.; All x2 statistics are significant at p < .001. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

889

Model B The factor loading invariance model, in which the loadings of all the manifest indicators and first-order factors were constrained to be equal across groups.

Model C The factor loading invariance and factor covariance invariance model, in which both factor loadings and the covariance between the second-order factors was constrained to be equal across groups.

Model D The factor loading invariance, factor covariance invariance, and factor variance invariance model, in which in addition to the constraints imposed in Model C, the variance of the second-order factors were constrained to be equal across groups.

Model E The factor loading invariance, factor covariance invariance, factor variance invariance, and error variance invariance model, in which all the parameters were constrained to be equal across groups. Unless constrained in the model, factor variances for the second-order factors in multisample CFA were freely estimated and varied across groups. Thus, the loadings of the devotion and purity first-order factors were set to one. Although the x2 difference test showed significant differences between the consecutive models, all subjective fit indices, however, were acceptable (see Table 9) and the differences in fit indices between the consecutive models were small. The difference between the CFI values between two consecutive models was never larger than .01 (see Cheung & Rensvold, 2002). According to the CAIC value, the most constrained model was the second best model. Thus, these analyses suggested that there was a reasonable level of measurement invariance for the hierarchical model of ethnocentrism. Given measurement invariance, we combined the data from the three samples and found good fit (x2 ¼ 1220.51, df ¼ 370, x2/df ¼ 3.30, CFI ¼ .98, NNFI ¼ .98, SRMR ¼ .052, RMSEA ¼ .057) for a model in which ethnocentrism formed a clear third-order factor at the highest level of analysis (see Figure 5). This suggests that although ethnocentrism consists of two distinct dimensions of intergroup and intragroup ethnocentrism, it also comprises a unitary construct at the highest level of generalization. Finally, the associations of ethnocentrism with the external criteria were investigated. Once again four new variables were formed to summarize the results more concisely: Intergroup ethnocentrism (E1; US: a ¼ .94, Serbia: a ¼ .94, France: a ¼ .93), intragroup ethnocentrism (E2: US: a ¼ .86, Serbia: a ¼ .92, France: a ¼ .82), a composite Likert outgroup attitude measure (US Whites: a ¼ .88, Serbs: a ¼ .86, French: a ¼ .89) and a composite feeling thermometer outgroup attitude measure (US Whites: a ¼ .92, Serbs: a ¼ .84, French: a ¼ .90). Once again all outgroup attitudes were strongly Table 9. Goodness of fit indices for measurement invariance of the two-correlated-second-order-factor model of ethnocentrism in the US, Serbian, and French samples x2

df

x2/df

CFI

NNFI

RMSEA

CAIC

1950.55 2117.86 2168.36 2206.64 2713.13

1110 1164 1166 1170 1240

1.76 1.82 1.86 1.89 2.19

.98 .97 .97 .97 .96

.97 .97 .97 .97 .96

.057 .059 .061 .062 .072

3422.73 3182.36 3217.76 3225.84 3203.86

Model A. Baseline B. FL invariance C. FL, FC invariance D. FL, FC, FV invariance E. FL, FC, FV, EV invariance

Note: N ¼ 699. FL ¼ factor loadings. FC ¼ factor covariances. FV ¼ factor variances.; EV ¼error variances.; All x2 statistics are significant at p < .001. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

890

Boris Bizumic et al.

Figure 5. The model of ethnocentrism consisting of total ethnocentrism (E) as a third-order factor, intergroup (E1) and intragroup ethnocentrism (E2) as two-order factors, the six facets of ethnocentrism as first-order factors in the sample of US, Serbian, and French participants (N ¼ 699). Manifest indicators and their loadings are not shown. p < .001

intercorrelated in each country and the relationships of the two kinds of ethnocentrism with specific outgroup attitudes and with the two composite measures of outgroup attitudes were very similar. As seen in Table 10, the associations of E1 and E2 with the external criteria were generally consistent across samples. Both E1 and E2 were positively related to ethnic pride (though not in France) and RWA. E1, but not E2, was positively associated with SDO, pro-war attitudes, and being male (though not in Serbia). Finally, E2, but not E1, was positively associated with ethnic identification, threat perception, lower status perception (though not in Serbia), and religiosity. Thus, despite a few minor differences, the overall results supported our expectations. Table 10 also shows that among the ethnic majority groups, ethnocentrism (entered in step 2) most often predicted the criterion variables over and above ingroup positivity (the sum of all the measures of positive ingroup attitudes) and outgroup negativity (the sum of all measures of negative outgroup attitudes), which were entered in step 1. These findings give further support to the view that ethnocentrism is conceptually distinct from ingroup positivity and outgroup negativity. Although the pattern of the findings is similar to those of the total samples, there were some differences, probably caused by somewhat smaller ethnic majority subsample sizes and sometimes strong correlations between the four predictors (which somewhat influenced multicollinearity). This was particularly the case among US Whites, possibly because this subsample had the smallest subsample size (152 participants), and certain items of the Likert scales measures of ingroup and outgroup attitudes included sentiments pertaining to intergroup preference and exploitativeness. Nevertheless, the findings indicate that ethnocentrism most often can explain the variance in the criterion variables, over and above ingroup and outgroup attitudes, suggesting that it is indeed conceptually distinct from ingroup positivity and outgroup negativity. The associations of the ethnocentrism dimensions with ingroup and outgroup attitudes were examined for the ethnic majority subsamples. Only the relationships of ethnocentrism with the composite measures of outgroup negativity are reported because the effects tended to be very similar across specific outgroups. As expected, E1, but not E2, was generally associated with anti-outgroup attitudes for both the Likert scales and the feeling thermometer measure, whereas E2, but not E1, was associated with pro-ingroup attitudes on the feeling thermometer measure (see Table 11). Contrary to expectations, among US Whites and Serbs both E1 and E2 were associated with pro-ingroup attitudes on the Likert scales, Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

891

Table 10. Beta coefficients of the simultaneous multiple regression of the external criteria on E1 and E2 in the US (N ¼ 212), Serbian (N ¼ 208), and French (N ¼ 279) total samples, and step 2 of the sequential multiple regression of the external criteria on anti-outgroup attitudes, pro-ingroup attitudes, E1, and E2 in the US White (n ¼ 152), Serbian (n ¼ 198), and French (n ¼ 226) ethnic majority group subsamples RWA US total sample E1 .40 E2 .29 US Whites Anti-outgroup .00 Pro-ingroup .11 E1 .36 E2 .27 DF 19.92 Serbian total sample E1 .16 E2 .60 Serbs Anti-outgroup .11 Pro-ingroup
.03 E1 .14 E2 .56 DF 48.74 French total sample E1 .30 E2 .36 French Anti-outgroup .04 Pro-ingroup .10 E1 .32 E2 .26 DF 25.50

Ethnic pride

SDO

War

Gender

Ethnic identification

Ethnic threat

Religiosity

Ethnic power

.22 .27

.55
.12

.38
.07

.27
.17


.06 .42


.08 .25

.04 .18

.10
.24

.06 .43 .10 .16y 3.41

.12
.16y .50 .09 19.31

.09 .08 .35
.02 6.37

.09
.12 .28
.10 3.63

.01 .25
.07 .22 2.94y

.08 .10
.10 .06 .42


.23 .14 .17 .04 1.71

.00 .35
.14
.08 1.60

.18 .27

.64
.31

.37
.06

.02 .07

.06 .57


.14 .18 .20 .24 10.58

.02
.16 .65
.24 26.99

.06
.13 .39
.04 9.28


.03 .04 .00 .07 .26

.05 .28
.01 .46 21.30

.04 .18
.15 .22 .05 .26 9.46

.53 .06
.07
.12 .60 .08 39.47

.20 .04
.18y
.23 .33 .06 9.27

.24 .06
.05
.03 .27 .04 5.63

.05 .36
.01 .19 .09 .25 8.96

.15 .30 .16y .00 .03 .29 6.66 .04 .23 .04
.12 .08 .18 4.44

.08 .33

.12 .08

.02 .05 .08 .30 8.42


.20
.06 .23 .11 5.55

.01 .27


.01
.15

.12 .15
.02 .14y 1.97


.07 .11
.06 .01 .17

Note: Positive bs in the gender column indicate that males were higher. For the ethnic majority subsamples, anti-outgroup, and pro-ingroup attitudes were entered in the first step, and E1, and E2 in the second. DF ¼ F-value change.; p < .05; p < .01; p < .001; yp ¼ .06.

and among the French neither E1 nor E2 was associated with pro-ingroup attitudes on the Likert scales. Similarly to Study 1, this might have been influenced by the fact that the Likert scale items involved some aspects of intergroup attitudes. Table 11 presents squared semi-partial correlations, which represent the proportion of total variation in the response variable (outgroup negativity or ingroup positivity) that is uniquely explained by one predictor variable (e.g., E1) controlling for the impact of another (e.g., E2). Its mean value for the relationship of E1 and outgroup attitudes was .21 (.26 for US Whites, .12 for Serbs, and .24 for French), and for the relationship of E2 with ingroup attitudes its mean value was .05 (.05 for US Whites, .09 for Serbs, and .01 for the French). To sum up, the cross-cultural investigation gave further support to the proposed reconceptualization of ethnocentrism. The measure of ethnocentrism and its subscales were fully balanced and reliable in each culture. Its associations with external criteria showed how the two kinds of ethnocentrism differ from each other and have differential implications for group attitudes and behaviors. Given that both the structure and the associations with external measures were expected gave further support to the construct and predictive validity of our measure.

GENERAL DISCUSSION The findings of this investigation supported the proposed reconceptualization of ethnocentrism in four countries. They indicated that ethnocentrism can be located at three levels of concept breadth: (a) six lower order expressions of Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

892

Boris Bizumic et al.

Table 11. b coefficients and squared-semipartial correlations of the simultaneous multiple regression of pro-ingroup and antioutgroup attitudes on feeling thermometer (FT) and Likert scales (L) on E1 and E2 in the US White, Serbian, and French ethnic majority group subsamples Variable E1 US Whites (b) (DR2) Serbs (b) (DR2) French (b) (DR2) E2 US Whites (b) (DR2) Serbs (b) (DR2) French (b) (DR2)

Pro-ingroup (L)

Pro-ingroup (FT)

.42 .14 .25 .04 .06 .00

.03 .00 .04 .00 .04 .00

.26 .05 .46 .13 .03 .00

.25 .05 .29 .05 .17 .02

Anti-outgroup (L) .71 .39 .40 .10 .57 .26 .15 .02 .04 .00
.01 .00

Anti-outgroup (FT) .41 .13 .46 .13 .53 .22
.06 .00 .00 .00
.09 .01

Note: US Whites: n ¼ 152. Serbs: n ¼ 198. French: n ¼ 226. Outgroups for US Whites: Latinos, Asians, and Blacks (L); Blacks, Asians, Latinos, French, Romanians, Serbs, Arabs, Chinese, and New Zealanders (FT). Outgroups for Serbs: Croats, Hungarians, and Albanians (L); Africans, French, Romanians, New Zealanders, Arabs, Chinese, and Americans (FT). Outgroups for French: Maghrebians, Asians, and Germans (L); Africans, Romanians, New Zealanders, Serbs, Arabs, Chinese, and Americans (FT). DR2 ¼ squared-semi-partial correlation.; p < .05; p < .01; p < .001.

ethnocentrism, (b) two second-order expressions, and (c) at the highest level of generalization, a unitary third-order construct of ethnocentrism. The research also indicated that ethnocentrism is related to ingroup positivity and outgroup negativity, but that it appears to be distinct to it, and also that only intergroup, not intragroup, ethnocentrism is related to outgroup negativity. The results showed that intergroup and intragroup ethnocentrism are correlated, seem to reinforce each other, and that at the highest level of generalization form a unitary concept of ethnocentrism. Preference, superiority, purity, and exploitativeness loaded on the intergroup ethnocentrism factor, confirming our expectation that the common factor underlying these four dimensions is intergroup and comparative. Both devotion and group cohesion loaded intragroup ethnocentrism, confirming that their common factor is primarily of an intragroup nature. Devotion, however, had consistently higher loadings on intragroup ethnocentrism than group cohesion in all samples, indicating that devotion seems to be a purer expression of intragroup ethnocentrism perhaps because it is more directly related with the view that the group is more important than its individual members. However, group cohesion always had strong loadings on the intragroup ethnocentrism factor suggesting that it is indeed an important facet of intragroup ethnocentrism. The finding that the two kinds of ethnocentrism formed a higher order factor of ethnocentrism suggests that they are both the expressions of one general factor of group self-centeredness. Thus, ethnocentrism, like many other constructs in psychology, appears to have a hierarchical structure. The hierarchical constructs have been widely used in psychology (e.g., the five-factor model of personality), particularly because they allow both highly accurate prediction at the narrower band level and considerable generality at the broader level. As Harding, Proshansky, Kutner, & Chein (1969) put it, ‘‘at times we need to focus on the fine details, and at other times, we need to stand back to see the big picture’’ (p. 17). Thus, depending on the research question, at times it may be particularly useful to focus on the total concept of ethnocentrism, at other times on intragroup and intergroup ethnocentrism, and sometimes on the six primary dimensions of ethnocentrism. Although it is important to cross-validate factor solutions across different samples (Frank & Widaman, 1995), replication is very difficult when dealing with hierarchical solutions (Gorsuch, 1983). Thus, the finding that a hierarchical structure was supported in four different cultures, and that the measurement invariance was obtained for the scale, contributes significantly to confidence in the robustness of the factor structure and validity of the reconceptualization. This is in line with the view by influential psychometricians, such as Cattell, Guilford, and Eysenck, who argued that the most important evidence of validity is the replication of factor structure across different groups and that this is much more Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

893

important than how the factors relate to external criteria (Hogan, 2005). We have, however, also investigated the relationship of ethnocentrism with external criteria in order to investigate if the two higher order constructs have differential implications for group attitudes and behaviors, and to gain additional evidence for the validity of the measure. Our findings show that intergroup ethnocentrism related primarily to SDO, pro-war attitudes, being male, and outgroup negativity, whereas intragroup ethnocentrism related primarily to ethnic identification, group threat perception, lower group status perception, religiosity, and mere ingroup positivity. Both kinds of ethnocentrism were generally related to RWA and ethnic pride. The findings, therefore, suggest that intergroup and intragroup ethnocentrism are related, but are also distinct and have differential implications for group attitudes and behaviors. Intragroup ethnocentrism appears to be primarily concerned with ethnic and status insecurity, self-transcendence, mere ingroup positivity, and ethnic identification. It deals with the relationship of individuals with their own groups and appears to be a consequence of group formation. On the other hand, intergroup ethnocentrism appears primarily based on selfaggrandizing attitudes, warlikeness, and a general chauvinistic outlook toward intergroup relations. It deals with the relationship of the ingroup to outgroups and helps instill the belief that other groups are relatively unimportant. This should facilitate the willingness to sacrifice any outgroups when required. The two kinds of ethnocentrism were intercorrelated in each sample, suggesting that they are both concerned with the same idea of giving great importance to the ingroup. Although to a large extent, the findings confirm the main theoretical framework, which was presented in the introduction, there were several findings that contradicted. The expected association of ethnocentrism with threat perception, power/status, and ethnic pride were not found. First, participants’ level of threat perception was associated only with intragroup, but not intergroup, ethnocentrism in each culture. These findings might have been influenced by the measure of threat, which dealt with a very broad and unspecific perception of group threat, whereas most studies demonstrating the relationship between threat and intergroup facets of ethnocentrism focus on outgroup threat. Therefore, the measure might have been associated only with intragroup ethnocentrism because of its clear defensive functions. Second, the results showed that intragroup, but not intergroup, ethnocentrism, was related to lower power/status perception of one’s ethnic group. This again might have been a problem of the measurement of the status. It seems plausible that individual differences in the status perception were related only to intragroup ethnocentrism, probably because the perception of lower group status is threatening (e.g., Lewin, 1948) and intragroup ethnocentrism may act as a kind of defense against the threat. Finally, although ethnic pride was originally expected to relate only to intragroup ethnocentrism, it related to both intergroup and intragroup ethnocentrism in all the countries except France, where the results were consistent with the initial hypothesis. Thus, people seem to generally take pride in the group in itself, but also in comparing it to other groups. Accordingly, ethnic pride may have both intragroup and intergroup aspects. The present research indicates that ethnocentrism is differentially related to outgroup negativity and mere ingroup positivity, and that it is not equivalent to them. Intergroup, but not intragroup, ethnocentrism was associated with outgroup negativity. Thus, in contrast to Sumner (1906) and others, who have argued that outgroup negativity is related to ingroup attitudes, this was not the case even for strong ingroup attitudes, such as devotion and group cohesion. This finding gives further support to the view that group formation and categorization, as represented in intragroup ethnocentrism, do not necessarily and directly cause outgroup negativity and prejudice (see Turner & Reynolds, 2001). In this study then, it seemed that only comparative or relational positivity (cf. Mummendey et al., 2001), exemplified in intergroup ethnocentrism, relates to outgroup negativity. However, the CFAs testing the conceptual distinction suggested that intergroup ethnocentrism although related to outgroup negativity is not strongly enough related to indicate unidimensionality (cf. Allport, 1954; Brewer, 1979, 1999, 2007; Brewer & Campbell, 1976; Cashdan, 2001; Turner, 1975, 1978). Nevertheless, there was a fairly substantial relationship suggesting that intergroup ethnocentrism does seem to contribute significantly to outgroup negativity. However, it appears that the constructs are not equivalent, supporting the view that ‘‘it is wrong to equate in-group favoritism and out-group hostility’’ (Brewer, 2007, p. 7). Accordingly, the argument of conceptual distinction does not mean that ethnocentrism is not related to outgroup negativity and ingroup positivity. These concepts are indeed related, and in some instances strongly, but they are not equivalent, in the same way that E1 and E2 are not equivalent, even though their relationship is often strong (e.g., in Serbia their correlation is .80, but the model with two factors is still superior to the model with one factor). Kline (2004) argues that if the correlation between two latent variables is higher than .85, we are dealing with only one underlying concept; otherwise, the two concepts are not equivalent. The correlations between ethnocentrism, ingroup positivity, and outgroup negativity were much lower than .85. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

894

Boris Bizumic et al.

The relationship of ethnocentrism with mere ingroup positivity seems more clearly explained by focusing on proingroup attitudes as measured by the content-free Feeling Thermometer Scales. Mere ingroup positivity was generally positively related to intragroup, but not intergroup, ethnocentrism, and in general, much more weakly than what was the relationship between intergroup ethnocentrism and outgroup negativity, suggesting that intragroup ethnocentrism and ingroup positivity are much more independent from each other than intergroup ethnocentrism and outgroup negativity. As theorized, the strength of the relationship between the two kinds of ethnocentrism differed across cultures and it appears that it was affected by the conflictual nature of the relationships of the ingroups with outgroups. This explains why the two kinds of ethnocentrism were so powerfully related in Serbia, where intergroup relationships are highly conflictual, as Serbs has recently engaged in a series of wars with the different ethnic groups and still experience major tensions in their relationships with several ethnic groups in the region, as well as with a number of Western countries. The relationship between the two kinds of ethnocentrism was also strong in France. This is a possible outcome of the fact that the French experience increasing conflicts with Arab and West African Muslim immigrant groups, which are not being successfully integrated within French society. Additionally, the French also fear that the status and eminence of their culture and language have been threatened by globalization and particularly by Anglo culture. On the other hand, the levels of interethnic conflict in both New Zealand and the US are much smaller. Although there are some ethnic tensions in both societies, ethnic relationships do tend to be relatively harmonious. Additionally, the dominant culture in both societies is Anglo culture, which is hegemonic and is not strongly threatened by other cultures. Accordingly, the two kinds of ethnocentrism were weakly-to-moderately correlated in both of these two countries. This explanation, although plausible, is, however, still speculative, and future, primarily experimental, research can show the extent to which the relationship between the two kinds of ethnocentrism is indeed affected by presence or absence of intergroup conflict. The objective of this research was to test a reconceptualization of ethnocentrism. To achieve this, we developed a new measure of ethnocentrism. The development of the measure has several implications. With its ability to measure ethnocentrism across cultures and languages it clearly improves upon previous measures, which were most often constrained to measure ethnocentrism in a particular culture and time period (e.g., Adorno et al., 1950; Beswick & Hills, 1969). In addition, the measure enables researchers to study ethnocentrism multidimensionally and at different levels of concept breadth. As such, the measure might enable a more systematic and comprehensive approach to the study of ethnocentrism than has been possible thus far. The present research had several limitations. First, the findings were not obtained from representative samples but from students. In addition, the US sample was from a small city in the South, while the other samples were from the countries’ largest cities. The samples from New Zealand and Serbia were from the countries’ premier universities, while this was not the case for France and the US. This precluded meaningful cross-national comparisons of ethnocentrism. A study that would use a number of ethnic groups could study the structure of ethnocentrism, group differences in its levels, and its relationships with other constructs at the group or national level of analysis. Furthermore, the reported studies primarily investigated ethnocentrism in four ethnic majority groups. The proposed structure of ethnocentrism might be characteristic of ethnic minority groups as well, but this needs to be investigated because there are different kinds of ethnic minorities and they can have different patterns of relationships with majority groups. Perhaps, intragroup ethnocentrism is more pronounced in ethnic minorities than intergroup ethnocentrism because they are generally more vulnerable and culturally insecure than ethnic majorities. In addition, the extent to which ethnocentrism is distinct from ingroup positivity and outgroup negativity in ethnic minority groups needs to be also tested. The present research did not directly investigate why ethnocentric people dislike ethnic outgroups. This can be investigated using correlational and experimental research. If similarity is a factor, then perceived outgroup similarity should mediate outgroup dislike. If threat or self-aggrandizement are factors, then perceived threat or status/power differences should be mediators. Experimental studies could manipulate the salience of outgroup dissimilarity, threat, or status/power differences to investigate this. Do intergroup ethnocentrism and outgroup negativity impact causally on each other? If so, which influence is more important and might this depend on situational factors? Systematic longitudinal studies of ethnocentrism and outgroup negativity in ethnic groups might answer this. To conclude, this research supported the proposed reconceptualization of ethnocentrism in four countries. Although intergroup and intragroup kinds of ethnocentrism appeared to have quite distinct consequences for group behaviors and attitudes, they were intercorrelated, suggesting that both involve the main belief that that one’s ingroup is of central importance. Therefore, the lesson that ethnocentrism seeks to inculcate is, ‘‘the individual, even when from our own group, is not very important; other ethnic groups are very not important; what is really of central and critical importance is our group.’’ Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

895

ACKNOWLEDGEMENTS We thank John C. Turner, Marilynn Brewer, and Katherine Reynolds for their very useful and constructive comments on earlier drafts of this manuscript. This research was supported by the Bright Future Scholarship (Top Achiever Doctoral) of the Tertiary Education Commission in New Zealand to the first author. REFERENCES Adorno, T. W., Frenkel-Brunswik, E., Levinson, P. J., & Sanford, R. N. (1950). The authoritarian personality. New York: Harper & Brothers. Altemeyer, B. (1996). The authoritarian specter. Cambridge, MA: Harvard University Press. Altemeyer, B. (1998). The other ‘‘authoritarian personality’’. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 30, pp. 47–92). San Diego: Academic Press. American Psychological Association. (2001). Thesaurus of psychological index terms (9th ed). Washington, DC: American Psychological Association. Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107, 238–246. Bentler, P. M. (1992). On the fit of models to covariances and methodology to the bulletin. Psychological Bulletin, 112, 400–404. Bentler, P. M. (1995). EQS structural equations program manual. Encino, CA. Multivariate Software. Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88, 588–606. Berry, J. W., & Kalin, R. (1995). Multicultural and ethnic attitudes in Canada: An overview of the 1991 national survey. Canadian Journal of Behavioural Science, 27, 301–320. Beswick, D. G., & Hills, M. D. (1969). An Australian ethnocentrism scale. Australian Journal of Psychology, 21, 211–225. Bizumic, B. (2000). Ethnocentric ideology. Unpublished master’s thesis, University of Auckland, Auckland, New Zealand. Brewer, M. B. (1979). In-group bias in the minimal intergroup situation: A cognitive motivational analysis. Psychological Bulletin, 86, 307–324. Brewer, M. B., & Campbell, D. T. (1976). Ethnocentrism and intergroup attitudes: East African evidence. New York: Sage. Brewer, M. B. (1981). Ethnocentrism and its role in interpersonal trust. In M. B. Brewer, & B. C. Collins (Eds.), Scientific inquiry and the social sciences (pp. 345–360). San Francisco: Jossey-Bass. Brewer, M. B. (1999). The psychology of prejudice: Ingroup love or outgroup hate. Journal of Social Issues, 55, 429–444. Brewer, M. B. (2007) The importance of being ‘we.’ Human nature and intergroup relations. American Psychologist, 62, 728–738. Brislin, R. (1980). Translation and content analysis of oral and written materials. In H. Triandis, & J. Berry (Eds.), Handbook of crosscultural psychology (Vol. 2 pp. 389–444). Boston: Allyn & Bacon. Brooke, P. P., Russell, D. W., & Price, J. L. (1988). Discriminant validation of measures of job satisfaction, job involvement, and organizational commitment. Journal of Applied Psychology, 73, 139–145. Brown, R. (2000). Social identity theory: Past achievements, current problems and future challenges. European Journal of Social Psychology, 30, 745–778. Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen, & K. Long (Eds.), Testing structural equation models (pp. 136–162). Newbury Park, CA: Sage. Carmines, E. G., & McIver, J. P. (1981). Analyzing models with unobserved variables: Analysis of covariance structures. In G. W. Bohmstedt, & E. F. Borgatta (Eds.), Social measurement (pp. 65–115). Thousand Oaks, CA: Sage. Cashdan, E. (2001). Ethnocentrism and xenophobia: A cross-cultural study. Current Anthropology, 42, 760–765. Chang, E. C., & Ritter, E. H. (1976). Ethnocentrism in Black college students. Journal of Social Psychology, 100, 89–98. Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233–255. Crocker, J., & Schwartz, I. (1985). Prejudice and ingroup favoritism in a minimal intergroup situation: Effects of self-esteem. Personality and Social Psychology Bulletin, 11, 379–386. Duckitt, J. (1992). The social psychology of prejudice. New York: Praeger. Duckitt, J. (2000). Ethnocentrism items. Unpublished manuscript. Duckitt, J. (2001). A dual-process cognitive-motivational model of prejudice. In M. Zanna (Ed.), Advances in experimental social psychology (Vol. 33, pp. 41–113). San Diego, CA: Academic Press. Frank, F. J., & Widaman, K. F. (1995). Factor analysis in the development and refinement of clinical assessment instruments. Psychological Assessment, 7, 286–299. Gerbing, D. W., & Anderson, J. C. (1987). Improper solutions in the analysis of covariance structures: Their interpretability and a comparison of alternate respecifications. Psychometrika, 52, 99–111. Gorsuch, R. L. (1983). Factor analysis (2nd ed). Hillsdate, NJ: Lawrence Earlbaum Associates. Harding, J., Proshansky, H., Kutner, B., & Chein, I. (1969). Prejudice and ethnic relations. In G. Lindzey, & E. Aronson (Eds.), The handbook of social psychology (2nd ed, Vol. 5, pp. 1–76). Reading, MA: Addison Wesley. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

896

Boris Bizumic et al.

Heaven, P. C. L., Rajab, D., & Ray, J. J. (1985). Patriotism, racism, and the disutility of the ethnocentrism concept. Journal of Social Psychology, 125, 181–185. Herodotus. (1947). The history of Herodotus (G. Rawlinson, Trans.). New York: Tudor. Herskovits, M. J. (1948). Man and his works. New York: Knopf. Hinkle, S., & Brown, R. (1990). Intergroup comparisons and social identity: Some links and lacunae. In D. Abrams, & M. A. Hogg (Eds.), Social identity theory: Construction and critical advances (pp. 48–70). New York: Springer-Verlag. Hogan, R. (2005). In defense of personality measurement: New wine for old whiners. Human Performance, 18, 331–341. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. Jahoda, G., & Krewer, B. (1997). History of cross-cultural and cultural psychology. In J. W. Berry, & Y. H. Poortinga, & J. Pandy (Eds.), Handbook of cross-cultural psychology (2nd ed, Vol. 1, pp. 1–42). Boston: Allyn and Bacon. Johnson, M. K., & Marini, M. M. (1998). Bridging the racial divide in the United States: The effect of gender. Social Psychology Quarterly, 61, 247–258. Keller, A. G. (1973). Societal evolution: A study of the evolutionary basis of the science of society. Port Washington, NY: Kennikat Press. (Original work published in 1931). Kline, P. (2000). The handbook of psychological testing (2nd ed). London: Routledge. Kline, R. B. (2004). Principles and practice of structural equation modeling (2nd ed). New York: The Guilford Press. Kosterman, R., & Feshbach, S. (1989). Toward a measure of patriotic and nationalistic attitudes. Political Psychology, 10, 257–274. LeVine, R. A., & Campbell, D. T. (1972). Ethnocentrism: Theories of conflict, ethnic attitudes, and group behavior. New York: John Wiley & Sons. Lewin, K. (1948). Resolving social conflicts: Selected papers on group dynamics. New York: Harper and Brothers. Loo, R. (2001). Motivational orientations toward work: An evaluation of the work preference inventory (student form). Measurement and Evaluation in Counseling and Development, 33, 222–233. Marsh, H. W. (1994). Confirmatory factor analysis of factorial invariance: A multifaceted approach. Structural Equation Modeling, 1, 5– 34. Marsh, H. W., & Hocover, D. (1985). Application of confirmatory factor analysis to the study of self-concept: First and higher order factor models and their invariance across groups. Psychological Bulletin, 95, 562–582. McConahay, J. B., Hardee, B. B., & Batts, V. (1981). Has racism declined in America? It depends on who is asking and what is asked. Journal of Conflict Resolution, 25, 563–579. McGee, W. J. (1900). Primitive numbers. Annual Report of the Bureau of American Ethnology (1897–1898), 19, 825–851. Mummendey, A., Klink, A., & Brown, R. (2001). Nationalism and patriotism: National identification and out-group rejection. British Journal of Social Psychology, 40, 159–172. Neuliep, J. W., & McCrosky, J. C. (1997). The development of a U.S. and generalized ethnocentrism scale. Communication Research Reports, 14, 385–398. Oxford English Dictionary. (1989). (2nd ed.). Oxford: Clarendon Press. Phillips, L., Penn, M. L., & Gaines, S. O. Jr., (1993). A hermeneutic rejoinder to ourselves and our critics. Journal of Black Psychology, 19, 350–357. Prothro, E. T. (1952). Ethnocentrism and anti-Negro attitudes in the Deep South. Journal of Abnormal and Social Psychology, 47, 105– 108. Rokeach, M. (1960). The open and closed mind. New York: Basic Books. Raden, D. (2003). Ingroup bias, classic ethnocentrism, and non-ethnocentrism among American Whites. Political Psychology, 24, 803– 828. Selznick, G. J., & Steinberg, S. (1969). The tenacity of prejudice: Anti-Semitism in contemporary America. New York: Harper & Row. Sherif, M. (1966). In common predicament: Social psychology of intergroup conflict and cooperation. Boston: Haughton Mifflin. Sidanius, J., & Pratto, F. (1993). The inevitability of oppression and the dynamics of social dominance. In P. Sniderman, P. E. Tetlock, & E. G. Carmines (Eds.), Prejudice, politics, and the American dilemma (pp. 173–211). Stanford, CA: Stanford University Press. Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup theory of social hierarchy and oppression. Cambridge: Cambridge University Press. Stagner, R. (1942). Some factors related to attitude toward war, 1938. Journal of Social Psychology, 16, 131–142. Stangor, C., Sullivan, L. A., & Ford, T. E. (1991). Affective and cognitive determinants of prejudice. Social Cognition, 9, 359–380. Steenkamp, J. E. M., & Baumgartner, H. (1998). Assessing measurement invariance in cross-national consumer research. Journal of Consumer Research, 25, 78–90. Steiger, J. H. (1990). Structural model evaluation and modification: An interval estimation approach. Multivariate Behavioral Research, 25, 173–180. Sumner, W. G. (1906). Folkways. Boston: Ginn and Company. Sumner, W. G. (1911). War and other essays. Freeport: Yale University Press. Sumner, W. G., Keller, A. G., & Davie, M. R. (1928). The science of society. New Haven: Yale University Press. Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup behavior. In S. Worchel, & W. G. Austin (Eds.), Psychology of intergroup relations (2nd ed, pp. 7–24). Chicago: Nelson-Hall. Taylor, D. M., & Jaggi, V. (1974). Ethnocentrism and causal attribution in a South Indian context. Journal of Cross-Cultural Psychology, 5, 162–171.

Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

897

Thomas, D. R. (1975). Authoritarianism, child rearing practices and ethnocentrism in seven Pacific Islands groups. International Journal of Psychology, 10, 235–246. Tzuriel, D., & Klein, M. M. (1977). Ego identity: Effects of ethnocentrism, ethnic identification, and cognitive complexity in Israeli, Oriental, and Western ethnic groups. Psychological Reports, 40, 1099–1110. Turner, J. C. (1975). Social comparison and social identity: Some prospects for intergroup behaviour. European Journal of Social Psychology, 5, 5–34. Turner, J. C. (1978). Social comparison, similarity and ingroup favouritism. In H. Tajfel (Ed.), Differentiation between social groups: Studies in the social psychology of intergroup relations (pp. 235–250). London: Academic Press. Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987). Rediscovering the social group: A self-categorization theory. Oxford: Basil Blackwell. Turner, J. C., & Reynolds, K. J. (2001). The social identity perspective in intergroup relations: Theories, themes, and controversies. In R. Brown, & S. L. Gaertner (Eds.), Blackwell handbook of social psychology: Intergroup processes. Malden, MA: Blackwell.

APPENDIX A. ITEMS OF THE FINAL ETHNOCENTRISM SCALE (R—CONTRAIT ITEMS) Preference In general, I prefer doing things with people from my own culture than with people from different cultures. In most circumstances it is right AND natural to favor members from one’s own cultural or ethnic group over strangers or foreigners. I would probably be quite content living in a cultural or ethnic group that is very different to mine. (R) I don’t think I have any particular preference for my own cultural or ethnic group over others. (R) I feel much more relaxed and comfortable in the company of people from my cultural or ethnic group than I feel in the company of others. In most cases, I like people from my culture more than I like others. I do NOT prefer members of my own cultural or ethnic group to others. (R) If I could be born again, it would be fine for me to be born into a different cultural or ethnic group to my own. (R) Superiority The world would be a much better place if all other cultures and ethnic groups modeled themselves on my culture. There are many cultures in which life is as good as in ours. (R) I don’t believe that my cultural or ethnic group is any better than any other. (R) I feel that, in general, the basic values of my cultural or ethnic group are fairer, more decent, and more reasonable than those of many others. Our cultural or ethnic group is NOT more deserving and valuable than others. (R) The values, way of life, and customs of most other cultures are probably just as good as those of my own. (R) I suspect that the lifestyle of my cultural or ethnic group probably gives more opportunity for people to live happy and fulfilling lives than the lifestyle of most other groups would. The idea that my culture is superior to others is completely ridiculous and unreasonable. (R) If people from other cultures understood the true value of our culture, I think that most of them would prefer to live in our culture. In general, other cultures do not have the inner strength and resilience of our culture. It is simply NOT true that our culture and our customs are any better than other cultures and other customs. (R) On the whole, people from my culture tend to be better people than people from other cultures. Purity I don’t think it is to our advantage to mix with people from other cultural or ethnic groups. Our culture would be much better off if we could keep people from different cultures out. Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

898

Boris Bizumic et al.

It is better for people from different ethnic and cultural groups not to marry. I’d like to live in a neighborhood where there are many people from all sorts of quite different cultural and ethnic groups to mine. (R) It is NOT wise to allow people from completely different cultures to join our culture. I like the idea of a society in which people from completely different cultures, ethnic groups, and backgrounds mix together freely. (R) I prefer not to be around people from very different cultures. I’d really enjoy working and being with people from completely different cultures and ethnic groups. (R) I would be very happy for a member of my family marrying a person from a different cultural or ethnic group. (R) I would really like to have many friends from completely different cultural or ethnic groups. (R)

Exploitativeness We should always put our interests first and not be oversensitive about the interests of other cultures or ethnic groups. In dealing with other cultures we should always be honest with them and respect their rights and feelings. (R) There’s nothing wrong with us taking advantage of some other cultural or ethnic group. Every group should look out for itself. We should always show consideration for the welfare of people from other cultural or ethnic groups even if, by doing this, we may lose some advantage over them. (R) It would upset me if my cultural group were becoming wealthier at the expense of other cultures. (R) We need to do what’s best for our own people, and stop worrying so much about what the effect might be on other peoples. In dealing with other ethnic and cultural groups our first priority should be that we make sure that we are the ones who end up gaining and not the ones who end up losing. I would be extremely unhappy if our actions had negative effects on other cultures, no matter how much advantage we might be gaining. (R)

Group Cohesion For the sake of our future, all the people of my cultural or ethnic group need to pull together, bury our disagreements, and unite in a common group identity. I believe it is far more important that our people have independence and individual freedoms than cohesion and unity. (R) It is absolutely vital that all true members of my ethnic or cultural group forget their differences and strive for greater unity and cohesion. We should focus all our energy on trying to develop a greater sense of unity, community, and solidarity in our cultural group. The spirit of cooperation, unity, and cohesion should be our foremost goal, even if it endangers some of our personal freedoms. Those people who keep saying we need a more cohesive and united cultural or ethnic group are threatening our individuality and right to disagree. (R) Instead of greater unity and more cohesion, our people need more change, innovation, and freedom for individuals to express themselves however they want to. (R) We, as a cultural group, should be more integrated and cohesive, even if it reduces our individual freedoms. We don’t need more unity and cohesion in our cultural group; we should rather encourage people to be more ready to think for themselves and express themselves and their individuality in whatever way they wish. (R) Personal freedoms and allowing people from our cultural group to do exactly what they want to do are more important than achieving unity and cohesion. (R) Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp

A reconceptualization of ethnocentrism

899

Devotion I am totally dedicated to the defense of the interests of my ethnic or cultural group and our way of life. I cannot imagine myself ever developing an intense, passionate, total devotion, and commitment to my ethnic or cultural group. (R) I just DON’T have the kind of strong and passionate attachment to my people and our culture that would make me make serious sacrifices for their interests. (R) No matter what happens, I will ALWAYS support my cultural or ethnic group and never let it down. The values, way of life and beliefs of my culture or ethnic group must be preserved whatever the sacrifices. If I am doubtful about actions of my cultural or ethnic group, I am prepared to be disloyal to it. (R) We must not betray our cultural values and traditions no matter how much energy, sacrifice, and dedication it takes. I think it is foolish to be completely and unconditionally devoted to one’s cultural or ethnic group. (R) I have a total loyalty to our people and our way of life. A strong, intense, and unreserved dedication to one’s own ethnic or cultural group can have some very negative consequences. (R)

Copyright # 2008 John Wiley & Sons, Ltd.

Eur. J. Soc. Psychol. 39, 871–899 (2009) DOI: 10.1002/ejsp