A Three-Factor Model of Social Identity

Abstract Despite the importance of the social identification construct in research and theory on group processes and intergroup relations, the issue of its dimensionality remains unresolved. It is proposed that social identity can be represented in terms of three factors: centrality; ingroup affect; and ingroup ties. I examined the efficacy of this model in five studies involving a total of 1078 respondents, one nonstudent sample, and three group memberships (university, gender, and nationality). Results of confirmatory factor analyses support the acceptability of the tripartite model, which fits the data significantly better than one- or two-dimensional (cognition/affect) alternatives. Correlations with theoretically relevant variables provide support for the convergent and discriminant validity of the three factors. Advantages and implications of the three-factor model are considered, with particular reference to social identity theory. Acknowledgments A University of Queensland Postdoctoral Research Fellowship and a University of Queensland New Staff Research Grant supported this work. Portions of the data were presented in April 1999, at the meeting of the Society of Australasian Social Psychologists. The data used in Studies 1, 3, and 4 derive from larger data sets analyzed (respectively) by Cameron (Citation1999), Cameron et al. (Citation2002), and Cameron (Citation2001). I thank Scott Reid, Mala McHale, and Daniel Johnson for their assistance with the data, and Jolanda Jetten and Russell Spears for their helpful suggestions. Notes Preliminary examination of item statistics indicated that some responses, most associated with the hypothesized affective component, were negatively skewed. Considering that this is typical in the measurement of self-esteem (e.g., Gray-Little, Williams, & Hancock, Citation1997), no attempt was made to normalize the items for the primary analyses. To ensure the robustness of the results, however, transformations were performed on five items with significant skews in Study 3 (where nonnormality was particularly marked); following Tabachnick and Fidell (Citation1996), log transformations were applied after “reflecting” the distributions (reflection changes a negative skew to a positive skew). A confirmatory factor analysis of the transformed data yielded a somewhat greater χ2 value than the comparable statistic for the three-factor model in Table 1, but the remaining (rounded) goodness-of-fit indexes were unchanged. In Study 2, there was unexpected evidence of a negative, though nonsignificant, relationship between centrality and affect in the case of gender. Correlations computed for each sex showed that scores on these subscales were negatively related for men, r(66) = −.29, p< .05, but not for women, r(78) = .13, ns. In Study 4, however, centrality and affect were positively and nonsignificantly associated for both men, r(106) = .14, and women, r(192) = .12. Following Tajfel (Citation1978), a distinction between “affective” and “evaluative” aspects of identification is often made (e.g., Brown et al., Citation1986, p. 275; Ellemers et al., Citation1999, p. 373; Hinkle et al., Citation1989, pp. 306 – 307). Whereas there is general agreement that the evaluation of group membership constitutes group-specific self-esteem (or collective self-esteem; Luhtanen & Crocker, Citation1992), the “emotional” aspects of social identity have been less consistently defined. Indeed, some conceptualizations (or operationalizations) of the emotional or affective aspects of group membership include reference to (or items that reflect) group-specific self-esteem (e.g., Deaux, Citation1996; Hinkle et al., Citation1989; Jackson & Smith, Citation1999). The present model similarly embodies the assumption that the positive or negative evaluation of social identity has emotional consequences (thus the label “ingroup affect”). A second way in which the emotional significance of group membership has been considered refers to interpersonal ties within the group and/or perceptions of common fate. Jackson (Citation2002), for example, referred to the relevant factor in his measure as “affective ties,” which corresponds most closely to “ingroup ties” in the present model. Such apparent ambiguity need not be problematic, however, considering its consistency with empirical relationships between indexes of psychological well-being (including self-esteem) and perceptions of ties to, or belonging with, the group (e.g., Baumeister & Leary, Citation1995; Hoyle & Crawford, Citation1994; Cameron, Citation1999), as well as evidence, deriving from the present data and elsewhere (Bollen & Hoyle, Citation1990) that factors reflecting group-derived esteem and perceived cohesiveness are empirically related. Nevertheless, researchers should be aware of various conceptual and semantic variations that are potential sources of confusion. That ingroup ties were unrelated to the interdependent facet of self-construal might seem contradictory. However, the personality dimension of interdependence, at least as operationalized by Singeles (Citation1994), appears to hold a somewhat different psychological meaning than interdependence as articulated, for example, in terms of interpersonal bonds with ingroup members, or the perception of common fate (Jackson, Citation2002). This suggests that interdependence can occur, and can be defined, in several ways, only some of which are captured by the present model (see also Deaux, Citation1996).