British Journal of Social Psychology (2003), 42, 357–3702003 The British Psychological Society www.bps.org.uk
Tests of both corollaries of social identitytheory’s self-esteem hypothesis inreal group settingsDiane M. Houston1* and Alexia Andreopoulou21Department of Psychology, University of Kent, UK2University of Crete, Greece
The present study tests both corollaries of the self-esteem hypothesis from socialidentity theory derived by Abrams and Hogg (1988). Greek students completed apre-test collective self-esteem (CSE) measure and then rated Greek students and eitherAmerican or Turkish students. They then completed a post-test collective self-esteemmeasure. Inconsistent with Corollary 2, higher prior CSE was associated both withhigher in-group ratings and higher out-group ratings, but not with bias. Consistent withCorollary 1, change in CSE was positively associated with bias, positively associatedwith in-group ratings and negatively associated with out-group ratings. There were nodifferences due to type of out-group. The Ž ndings are consistent with the idea that, forspeciŽ c comparisons between real groups, individual differences in self-esteem mayaffect the zone within which bias occurs rather than the amount of bias. However,achieving increases in self-esteem depends on establishing positive distinctiveness for thein-group in the context of a speciŽ c intergroup comparison.
According to social identity theory (Tajfel & Turner, 1979), one important reason whypeople display intergroup discrimination is that it enhances social identity and therebyelevates the self-esteem of group members. This prediction of social identity theory hasbecome known as the ’ self-esteem hypothesis’. Abrams and Hogg (1988) explicatedtwo corollaries of the self-esteem hypothesis. Corollary 1 is that successful intergroupdiscrimination enhances social identity and thus elevates self-esteem. The more thatthe in-group is positively differentiated from the out-group, the higher self-esteem as agroup member should become. Corollary 2 is that, because of a motivational need forpositive self-esteem, low or threatened self-esteem will motivate intergroup discrimi-nation. Thus there should be a negative association between prior self-esteem andsubsequent positive differentiation in favour of the in-group. Despite much researchinto self-esteem and intergroup behaviour, there have been few studies which havetested the self-esteem hypothesis systematically using a real intergroup context. Thepresent article describes a study designed to provide such a test.Abrams and Hogg (e.g. Abrams, 1992; Abrams & Hogg, 1988, 2001; Hogg &Abrams, 1990) argue that the evidence in support of the self-esteem hypothesis is not
*Requests for reprints should be addressed to Dr Diane M. Houston, Department of Psychology, University of Kent,Canterbury, Kent CT2 7NP, UK (e-mail: D.M.Houston@kent.ac.uk).
357

particularly overwhelming, and that this may reflect shortcomings in the way thecorollaries have been understood and operationalized. Social identity theory distin-guishes between social identity, which involves relevant group memberships, andpersonal identity, which involves aspects that distinguish a person as an individual.According to Abrams and Hogg, and later reviewers (e.g. Rubin & Hewstone, 1998),intergroup discrimination should be most associated with collective esteem relating toa specific group. Unfortunately, much of the available research is based on measures ofpersonal self-esteem. Moreover, as observed by Hunter, Platow, Howard, and Stringer(1996), amongst others, studies measuring collective self-esteem have most oftenexamined global collective self-esteem (the self-esteem derived from all the groups onebelongs to) rather than the specific self-esteem derived from self-categorization interms of a particular group. For the present article we focus primarily on evidenceregarding collective self-esteem.Corollary 2Aberson, Healy, and Romero (2000) conducted a meta-analysis of 34 studies thatprovided evidence of a moderate positive relationship between prior global self-esteemand in-group bias. In their review ’ self-esteem’ was operationalized as includingmeasures of group identification, global personal self-esteem and global collectiveself-esteem. They found no difference in the effect sizes associated with personal andcollective self-esteem measures, presumably because both reflect a general level ofself-evaluation that have equal relevance to any specific evaluation. Overall, theiranalysis finds against Corollary 2 and in favour of a consistency bias (higher self-esteempredicts greater bias). However, their analysis only focused on bias scores, not separatejudgments about the in-group and out-group. Evidence for some kind of consistencybias is compatible with Branscombe and Wann’s (1994) results, described below, andis supported by Verkuyten’s (1997) finding of a positive relationship between priorspecific collective self-esteem and subsequent bias by Dutch students againstMoroccan students. The consistency bias prediction, which was also posed as analternative possibility by Abram and Hogg (1988), runs directly counter to Corollary 2from the self-esteem hypothesis.A further refinement of the consistency bias idea is that high and low self-esteemindividuals display bias in different ways. Crocker, Blaine, and Luhtanen (1993) haveargued that high self-esteem individuals believe that they have positive attributes andexpect positive outcomes for the future; they are oriented towards self-enhancement.Low self-esteem individuals are uncertain about their attributes and are not veryoptimistic about the future; they are primarily concerned with self-protection. Crockeret al. (1993) suggested that in-group ratings reveal enhancement motives, whereasout-group ratings reveal protection motives. Thus, Crocker et al. (1993) predict thatpeople with high self-esteem should rate the in-group positively rather than derogatingthe out-group. People with low self-esteem should derogate the out-group rather thanenhancing the in-group (cf. Brown, Collins, & Schmidt, 1988). Crocker et al.’s analysistreats high and low self-esteem as qualitatively distinct. However, when self-esteem istreated (and measured) as a continuous quantitative variable, the statistical predictionremains that there should be a positive correlation between prior self-esteem andin-group bias (see also DeCremer & Oosterwegel, 2000). Finally, another possibility isthat self-esteem simply confers an optimistic or positivity bias, such that those withhigher self-esteem are simply more positive about any evaluative target. In summary, it
358 Diane M. Houston and Alexia Andreopoulou