The Role of Self-identity in the Theory of Planned Behavior: A Meta-Analysis Jostein Rise1

Paschal Sheeran

Norwegian Institute for Alcohol and Drug Research Oslo, Norway

University of Sheffield Sheffield, UK

Silje Hukkelberg Department of Mental Health, Norwegian Institute of Public Health Oslo, Norway The present study used meta-analysis to evaluate the role of self-identity in the theory of planned behavior (TPB). Altogether, 40 independent tests (N = 11607) could be included in the review. A large, sample-weighted average correlation between self-identity and behavioral intention was observed (r+ = .47). Multiple regression analyses showed that self-identity explained an increment of 6% of the variance in intention after controlling for the TPB components, and explained an increment of 9% of the variance when past behavior and the TPB components were controlled. The influence of self-identity on behavior was largely mediated by the strength of behavioral intentions. Theoretical implications of the findings are discussed. jasp_611

1085..1105

If a person sees himself or herself as concerned about the environment, does this mean that the person is likely to intend to recycle? And does this “environmental identity” directly influence recycling intentions, or does holding the identity mean that the person is more likely to see recycling as advantageous (i.e., hold a positive recycling attitude), perceive social pressure to recycle (i.e., believe there is a supportive subjective norm), or believe that recycling is easy (i.e., see the behavior as under one’s personal control)? The present research is concerned with these questions. In particular, we meta-analyzed the findings of studies that have measured self-identity, attitude, subjective norm, perceived behavioral control, and intention to assess the nature and strength of relations between self-identity and behavioral intentions.

1 Correspondence concerning this article should be addressed to Jostein Rise, Norwegian Institute for Alcohol and Drug Research, P.O. Box 565, Sentrum 0105, Oslo, Norway. E-mail: [email protected]

1085 Journal of Applied Social Psychology, 2010, 40, 5, pp. 1085–1105. © 2010 Copyright the Authors Journal compilation © 2010 Wiley Periodicals, Inc.

1086 RISE ET AL. Theories of Reasoned Action and Planned Behavior The relation between attitudes and behavior has been an area of major concern in social psychology ever since the seminal review by Wicker (1969), indicating that the ability of attitudes to predict behavior is actually quite poor (cf. Eagly & Chaiken, 1993). Perhaps the most important attempts to remedy this problem have been the introduction of the theory of reasoned action (Fishbein & Ajzen, 1975) and its successor, the theory of planned behavior (TPB; Ajzen, 1988). The theory of reasoned action (TRA) proposed that the concept of behavioral intention (e.g., “I intend to buy organic produce”) mediates the relationship between attitude and behavior, and put forward the concept of subjective norm as a second predictor of intention. Whereas attitude refers to the person’s overall evaluation of performing the behavior (e.g., “For me, buying organic produce would be good/bad”), subjective norm refers to perceived social pressure from important others to perform, or not to perform, the behavior (e.g., “Most people who are important to me think that I should buy organic produce”). The theory of planned behavior (TPB) added the concept of perceived behavioral control to the TRA as a third predictor of intention. Perceived behavioral control (PBC) refers to the perceived ease or difficulty of performing a behavior (e.g., “For me, buying organic produce would be easy/ difficult”). Thus, according to the TPB, the more positive the person’s attitude, the stronger the subjective norms and the greater the perceived control over the behavior, the more likely it is that the person will intend to perform the behavior. Correspondingly, the stronger the intention to perform the behavior, the more likely it is that the person will perform the behavior, assuming, of course, that the person possesses “actual control” over the performance (see Ajzen & Madden, 1986; Sheeran, Trafimow, & Armitage, 2003). Several meta-analyses have shown that behavioral intention is predictable from the three components of the TPB (e.g., Armitage & Conner, 2001a; Godin & Kok, 1996; Sheeran & Taylor, 1999). However, the level of prediction is far from perfect; the variance explained in intention ranges from just 28% to 40%, on average. This consideration has led several researchers to question the sufficiency assumption of the TPB; that is, the assumption that the theory adequately captures all theoretical determinants of intention. In fact, Ajzen (1991) relaxed this assumption when he developed the TPB, stating that the TPB is, in principle, open to inclusion of additional predictors so long as they increase the explained variance in behavioral intentions. Accordingly, several researchers have proposed additional predictors that might be used to augment the model’s predictive validity (for reviews, see

SELF-IDENTITY META-ANALYSIS

1087

Abraham, Sheeran, & Johnson, 1998; Conner & Armitage, 1998). The aim of the present research is to provide the first comprehensive quantitative assessment of one important additional variable; namely, self-identity.

Role of Self-Identity in the Theories of Reasoned Action and Planned Behavior The case for including self-identity as an additional predictor in the TPB derives from theorists who have argued that identity processes should be taken into account in the prediction of specific behaviors, and from empirical evidence that self-identity predicts behavioral intentions after attitudes and norms have been taken into account (e.g., Biddle, Bank, & Slavings, 1987; Charng, Piliavin, & Callero, 1988; Sparks & Shepherd, 1992). Self-identity refers to salient and enduring aspects of one’s self-perception (e.g., “I think of myself as a ‘green consumer’”; cf. Sparks, 2000). According to identity theory (e.g., Thoits & Virshup, 1997), people apply socially meaningful categories to describe themselves when answering the question “Who am I?” in terms of, for example, sociodemographic characteristics (e.g., gender), social roles (e.g., mother, father), social types (e.g., smoker, exerciser, healthy eater, blood donor), and even personality traits (e.g., honest, optimist). Thus, self-identities (or “me” identifications) are the perspective one takes toward oneself when taking the role of specific or generalized others, implying that one incorporates the meanings and expectations associated with a relevant categorization into the self, thus forming a set of identity standards that guide identity-relevant behaviors (Stets & Burke, 2000). However, from a reasoned action perspective, self-identity constitutes an external variable that is assumed to exert its effect through the components of the model and should, accordingly, have no independent value in the prediction of behavioral intentions. Sparks (2000) reviewed two theoretical grounds for assuming that components of the TPB mediate the self-identity/intention relation. The first line of argument assumes that self-identity exhibits conceptual overlap with attitudes because self-identity is likely to represent a class of behavioral outcomes that are on a par with utilitarian and affective outcomes expected to flow from behavioral performances (cf. Eagly & Chaiken, 1993). According to this idea, the concept of attitude should capture whatever influence selfidentity has on intention. When empirical studies do not support this prediction, it has been argued that this is because self-identity concerns may not have been especially salient when people responded to the evaluative scales typically used to tap attitude toward behaviors (Eagly & Chaiken, 1993).

1088 RISE ET AL. However, identity theorists (e.g., Biddle et al., 1985) have argued that attitudes, norms, and self-identity have different motivational roots. Individuals conform to attitudes for instrumental reasons and to norms for fear of being rejected by significant others (i.e., external sanctions), whereas one acts in accordance with one’s self-identity for self-verification reasons. That is, people are motivated to retain and affirm the sense of self and identity (cf. Stets & Burke, 2000): People act to be consistent in their identity standard. By this account, when the social categorization including the identity is activated, the person behaves so as to maintain consistency with the meanings held in the identity standards. Accordingly, self-identity will tend to predict intentions above the components of the TPB. The second reason why self-identity may not predict intention after TPB components have been taken into account relates to the possibility that self-identity may simply reflect past performance of a behavior. The argument is that people understand what kind of persons they are by making inferences based on their past behavior (i.e., through a self-perception process; Bem, 1972). This idea suggests that self-identity should have no direct effect on behavioral intentions once the effect of past behavior has been controlled. Relatively few empirical studies have addressed this issue, and mixed findings have been obtained. Some studies have observed that self-identity retains a unique effect on behavioral intentions after TPB components and past behavior have been taken into account (Conner, Warren, Close, & Sparks, 1999; Hildonen, 2001; Thompson & Rise, 2002); some studies have not shown an independent effect of self-identity on behavioral intentions (e.g., Fekadu & Kraft, 2001); whereas in other studies it has not been possible to separate the effect of self-identity on intention from that of past behavior because other predictors were included in the same step (Conner & Flesch, 2001; Conner & McMillan, 1999; Terry, Hogg, & White, 1999). In sum, it remains unclear whether the association between self-identity and behavioral intention merely reflects experience with the focal behavior. A further conceptual difficulty associated with evaluating the strength of the self-identity/intention relation relates to a general problem with what can be termed the additional-variables paradigm in TPB research. In this paradigm, researchers identify a variable that is not specified in the TPB, measure that variable in a TPB study of a particular behavior, and then assume that if the variable captures unique variance in intention (after TPB predictors are controlled), then their variable constitutes an additional predictor in the TPB. As O’Keefe (2002) pointed out, this practice undermines the principle of parsimony and is likely to lead to the development of a plethora of behavioral intention models whose validity and generalizability are indeterminate (also see Trafimow, 2004).

SELF-IDENTITY META-ANALYSIS

1089

To guard against this problem, O’Keefe (2002) proposed two criteria that should be used to evaluate additional predictors in the TPB. First, a given conceptual candidate should provide a large additional contribution to the prediction of intention (after controlling for components of the TPB), which reaches well beyond statistical significance. Second, the proposed concept needs to demonstrate its utility in predicting behavioral intentions across a wide range of behavioral domains. However, it is clear that primary research studies are rarely in a position to satisfy O’Keefe’s (2002) criteria. What is needed is a meta-analytic strategy that accumulates effect sizes across studies in a manner that permits general conclusions. Only one meta-analysis of the self-identity/intention relation has been conducted to date (Conner & Armitage, 1998), and this review deserves updating, for two reasons. First, the meta-analysis examined only six studies; and second, past behavior was not taken into account in the analysis. The conclusion we draw is that a new meta-analysis of the self-identity/intention relation—one that includes recent research and that permits statistical control of both TPB components and past behavior—is overdue.

The Present Study Based on the foregoing discussion, the aim of the present study is to provide a meta-analytic integration of research on self-identity and the TPB. In particular, the review aims to (a) quantify the strength of the relationship between self-identity and behavioral intentions; (b) estimate the increment in the variance in intentions that is attributable to self-identity after TPB variables have been taken into account; (c) estimate the increment in variance attributable to self-identity after both TPB variables and past behavior have been taken into account; and (d) assess whether intention mediates the selfidentity/behavior relationship.

Method Selection of Studies Several procedures were used to collect the samples of studies: (a) social scientific databases (e.g., BIDS, Conference Papers Index, PsychLit) were searched; (b) reference lists of identified papers were evaluated for inclusion; and (c) authors of published papers were contacted for potential unpublished studies and studies that were in press. In order to be included in the review, a bivariate statistical association between self-identity and behavioral

1090

RISE ET AL.

intention had to be retrievable from the studies. We also coded correlations for future behavior, past behavior, and TPB variables whenever available. In total, 40 independent tests of the self-identity/intention relation were identified from 33 papers. Table 1 presents the characteristics and effect sizes of the studies that were included in the review.

Meta-Analytic Strategy To provide an estimate of the effect size, the weighted average of the sample correlations (r+) was used. This coefficient describes the direction and strength of the relationship between two variables ranging from -1.00 to +1.00. We assumed that studies in the meta-analysis were sampled from populations with mean effect sizes that vary (i.e., random-effects model). Therefore, we used the Hunter–Schmidt method (Hunter & Schmidt, 1990; Hunter, Schmidt, & Jackson, 1982; for a discussion, see Field, 2001, 2005). Homogeneity analyses were conducted using the chi-square statistic (Hunter et al., 1982) to determine whether variation among the correlations was greater than by chance. The formula for degrees of freedom for this test is k - 1, where k is the number of independent correlations. If chi square is nonsignificant, then the correlations are homogeneous and the average weighted effect size (r+) can be said to represent the population effect size. Computation of weighted average effect size and homogeneity statistics were conducted using Schwarzer’s (1988) Meta computer program.

Multiple Regression Analyses All of the studies included in the review reported intercorrelations between self-identity and TPB variables. Correlations among self-identity, past behavior, and all TPB predictors were available in 11 cases. We used computations of the sample-weighted average correlations among selfidentity, TPB variables, and past behavior as the input matrix for multiple regression in order to determine the increment in variance attributable to self-identity after controlling for relevant predictors.

Results Sample-Weighted Average Correlations The guidelines provided by Cohen (1992) are useful for interpreting the magnitude of the sample-weighted average correlations (r+). Cohen

Give money to charity Exercise Casual sex Cannabis use Alcohol consumption Alcohol consumption

Students

Austin & Sheeran (2001) Campbell & Sheeran (2001) Conner & Flesch (2001) Conner & McMillan (1999) Conner, Warren, Close, & Sparks (1999, Study 1) Conner, Warren, Close, & Sparks (1999, Study 2)

Young adults, general population Students Students Students Students Students

Åstrøm & Rise (2001)

Eat a low-fat diet Eat a low-fat diet Eat a low-fat diet Donate blood Donate blood Intentions to work for National Health Services Eat healthy food

Behavior

Students Hospital workers General population Prospective students Students General population

Sample

Armitage & Conner (1999a) Armitage & Conner (1999b) Armitage & Conner (1999c) Armitage & Conner (2001b, Study 1) Armitage & Conner (2001b, Study 2) Arnold et al. (2006)

Authors

Studies of the Relation Between Self-Identity and Behavioral Intention

Table 1

175

251 181 384 249 176

735

221 413 110 134 172 978

N

.57

.17 .56 .29 .81 .48

.65

.57 .56 .54 .69 .44 .19

r

SELF-IDENTITY META-ANALYSIS

1091

Moan & Rise (2006) Moan, Rise, & Andersen (2004) Ouellette & Wood (1998) Rapaport & Orbell (2000)

Conner, Warren, Close, & Sparks (1999, Study 3) de Pelsmacker & Janssens (2007) Evans & Norman (2003) Fekadu & Kraft (2001) Giles, McClenahan, Cairns, & Mallet (2004) Hagger & Chatzisarantis (2006, Study 1) Hagger & Chatzisarantis (2006, Study 2) Hildonen (2001) Jackson, Smith, & Conner (2003) Mannetti, Pierro, & Livi (2004)

Authors

Table 1 Continued

Various mundane behaviors Dieting behavior Buy ecological products Physical activity Recycling

University students

Students University employees Students and young workers Adolescents Parents Students University students

Reduce smoking Smoking Various behaviors Provision of emotional support and practical assistance

Speeding behavior Road crossing Use contraceptives Donate blood

General population Adolescents Female adolescents University students

University students

Alcohol consumption

Behavior

Students

Sample

145 159 71 195

206 85 230

250

241

334 1833 354 100

159

N

.31a .25 .69 .23

.69 .50 .41

.80

.83

.23 .42 .31 .59

.65

r

1092 RISE ET AL.

Community residents Females University students

Terry, Hogg, & White (1999) Theodorakis (1994)

Theodorakis, Bagiatis, & Goudas (1995) Thompson & Rise (2002)

Eat organic vegetables Attitudes toward genetically modified foods Household recycling Participate in a physical fitness program Teach individuals with disabilities Exercise and recycle drinking cartons

Eat a low-fat diet

Eat a low-fat diet

Quit smoking Quit smoking Buy sustainable-produced foods Purchase British beef Eat a low-fat diet

232

99

143 395

261 99

239

216

182 242

204 204 550

.61

.71

.56 .31

.37 .74

.46

.70

.44 .64

.17 .34 .30

Because self-identity was measured with respect to smoker identity, whereas intention was measured with respect to smoking reduction, this correlation has been recoded.

a

Sparks & Shepherd (1992) Spence & Townsend (2006)

Sparks & Guthrie (1998, Study 3)

College students

Students UK sample, general population Danish sample, general population Finnish sample, general population General population General population

Sheeran (1998) Sparks & Guthrie (1998, Study 1)

Sparks & Guthrie (1998, Study 2)

Spanish students Norwegian students General population

Rise & Ommundsen (2010, Study 1) Rise & Ommundsen (2010, Study 2) Robinson & Smith (2002)

SELF-IDENTITY META-ANALYSIS

1093

1094 RISE ET AL. Table 2 Meta-Analysis of Self-Identity and Theory of Planned Behavior Variables Relationship

r+

95% confidence interval

c2

Intention/self-identity Intention/attitude Intention/subjective norm Intention/PBC Attitude/self-identity Attitude/subjective norm Attitude/PBC Subjective norm/self-identity Subjective norm/PBC PBC/self-identity

.47 .50 .39 .35 .37 .36 .25 .29 .14 .25

.46–.49 .49–.51 .37–.40 .33–.36 .36–.39 .35–.38 .23–.27 .28–.31 .12–.16 .24–.27

761.83*** 604.19*** 203.06*** 1280.11*** 438.46*** 258.46*** 749.62*** 263.69*** 434.04*** 1185.81***

Note. N = 11607. k = 40. N = sample size on which sample-weighted average correlation is based; k = number of correlations; r+ = sample-weighted average correlation; c2 = chi-square test for homogeneity of sample correlations; PBC = perceived behavioral control. ***p < .001.

proposed that a correlation of .10 is small, .30 is medium, and .50 is strong. Table 2 presents the average correlations obtained among self-identity and TPB variables (k = 40). The sample-weighted average correlation between self-identity and behavioral intentions was of medium magnitude (r+ = .47), according to Cohen’s (1992) criteria (95% confidence interval = .46–.49). The robustness of this correlation can be determined by estimating the number of unpublished studies with null findings that would be required to invalidate the conclusion that self-identity and intentions are significantly associated at the 5% alpha level. The fail-safe N (Rosenthal, 1984) was 35485, which greatly exceeds the recommended tolerance level of 5k + 10. Because it is extremely unlikely that there are so many studies with null results that we were unable to locate, the average correlation between self-identity and intention should be considered robust. The reliability of measures of self-identity included in the meta-analysis was generally high (mean a = .78). Nevertheless, we elected to account for measurement error using the formulas described by Hunter et al. (1982). The sample-weighted average correlation corrected for unreliability was .52.

SELF-IDENTITY META-ANALYSIS

1095

Thus, 1.3% of the variance was a result of unreliability, and 4.1% of the variance was a result of sampling error. The discriminant validity of self-identity, compared to the other components of the TPB, was also supported by the findings (Table 2). The highest average correlation was between self-identity and attitude (r+ = .37). The magnitude of this association indicates that there is only modest conceptual overlap between the two concepts. Similarly, small to medium correlations were obtained between self-identity and both subjective norm and PBC (r+s = .29 and .25, respectively). Self-Identity as an Additional Predictor of Intention in the TPB A two-step hierarchical regression analysis was conducted to determine whether self-identity enhances the prediction of behavioral intentions beyond that engendered by the TPB on its own. The components of the TPB were included in the first step, and self-identity was entered in the second step (see Table 3). Attitude was the strongest determinant in the first step (beta = .36, p < .001), although subjective norm and PBC were also significant predictors of intention. These three predictors accounted for 35% of the variance in intentions. Table 3 Hierarchical Regression of Intention on Theory of Planned Behavior Variables and Self-Identity Variable Step 1 Attitude Subjective norm Perceived behavioral control Step 2 Self-identity 2 R Model F DR2 Fchange Note. N = 11607. k = 40. ***p < .001.

Beta .36*** .23*** .23***

.35 2075.09***

Beta .29*** .18*** .18*** .28*** .41 651.50*** .06 2021.14***

1096 RISE ET AL. Notwithstanding this level of prediction by TPB variables, the inclusion of self-identity in the second step significantly enhanced the prediction of behavioral intentions (DR2 = .06, p < .001). Together, the four predictors accounted for 41% of the variance in intention. Attitudes and self-identity exhibited the highest beta weights (betas = .29 and .28, respectively), as compared to .18 for both subjective norms and PBC.

Predictive Validity of Self-Identity Controlling for TPB Variables and Past Behavior In the next analysis, TPB variables were entered on the first step, past behavior was entered on the second step, and self-identity was entered on the third step of a hierarchical regression (k = 16; N = 3488). The TPB components accounted for 31% of the variance (see Table 4), and past behavior explained an additional proportion of the variance (5%). In the final step, self-identity was able to account for a highly reliable increment of 9% in the variance explained in behavioral intentions. In this subgroup of studies, Table 4 Hierarchical Regression of Intention on Theory of Planned Behavior Variables, Self-Identity, and Past Behavior Variable Step 1 Attitude Subjective norm Perceived behavioral control Step 2 Past behavior Step 3 Self-identity 2 R Model F DR2 Fchange Note. N = 3488. k = 16. ***p < .001.

Beta .36*** .25*** .20***

.31 525.38***

Beta

Beta

.31*** .20*** .18***

.22*** .16*** .14***

.23***

.16***

.36 485.92*** .05 253.37***

.34*** .45 560.75*** .09 552.38***

SELF-IDENTITY META-ANALYSIS

1097

self-identity turned out to be the strongest determinant of intentions, along with attitude (betas = .34 and .22, respectively), whereas the betas for TPB predictors were clearly lower. These findings suggest that self-identity cannot be construed as simply a reflection of past behavior, but captures a separate and distinct psychological process in the formation of behavioral intentions.

Does Intention Mediate the Self-Identity/Behavior Relation? There were 13 studies (N = 2141) that included a prospective measure of behavior in which intercorrelations with self-identity, intention, and PBC could be retrieved. Using the sample-weighted average correlations as the input matrix, regression of behavior on intention and PBC showed significant beta coefficients for both predictors (betas = .47 and .29, respectively; ps < .001), and 36% of the variance in behavior was explained. The bivariate association between self-identity and behavior was significant (beta = .43, p < .001). Including self-identity on the second step of the regression equation engendered a significant increment in the variance accounted for (Fchange = 82.93, DR2 = .02, p < .001). Self-identity, intention, and PBC each had significant beta coefficients (betas = .20, .30, and .35, respectively; ps < .001). Although these findings appear to suggest that self-identity has a direct influence on behavior, even after intention and PBC have been taken into account, two considerations speak against this interpretation. First, correlations between PBC and both intention and self-identity were unusually small in this subset of studies (r+s = .22 and .08, respectively), so findings are likely to have been different if the values were similar to those obtained in the larger sample of studies (see Table 2). Second, self-identity was significantly associated with behavior after intention and PBC were controlled in none of the 13 primary studies, which suggests that the significant association may simply be an artifact of the large sample included in the meta-analysis. Finally, it is worth noting that a Sobel’s test indicated that intention was a highly reliable mediator of the self-identity/behavior relation (Z = 22.76, p < .001).

Discussion The present study examined 40 tests of the predictive validity of selfidentity using meta-analytic procedures and, therefore, constitutes the most comprehensive and systematic analysis of the self-identity/intention relation to date. The key findings from the review can be summarized as follows: Self-identity had a medium-sized average correlation with behavioral

1098

RISE ET AL.

intention, according to Cohen’s (1992) criteria. The association between self-identity and intention was similar in magnitude to the attitude–intention relationship (r+s = .48 and .50, respectively) and was larger than the average subjective-norm/intention, and PBC–intention correlations (r+s = .39 and .35, respectively). In addition, fail-safe N analyses indicate that the self-identity/ intention association was robust (i.e., resistant to future null results). Multiple regression analyses show that self-identity enhanced the prediction of intention after components of the TPB—and components of the TPB plus past behavior—were taken into account. Findings show that selfidentity captured 6% additional variance in intention above and beyond that afforded by attitude, subjective norm, and PBC, whereas a 9% increase in explained variance in intention was observed when past behavior was also controlled. Finally, mediation analysis suggests that the influence of selfidentity on behavior was largely, and perhaps entirely, mediated by the strength of behavioral intention. Although the results of the meta-analysis seem to support the inclusion of self-identity as an additional predictor in the TPB, O’Keefe’s (2002) criteria must be considered before firm conclusions can be drawn. O’Keefe’s first criterion is that any potential additional variable in the TPB should contribute a large additional contribution to the prediction of intention, and not simply a statistically reliable increment. In our view, values of 6% and 9% additional variance satisfy this criterion. Self-identity explains substantial additional variance in intention beyond that engendered by TPB variables and past behavior. O’Keefe’s (2002) second criterion is that the efficacy of a candidate variable must be demonstrated across a wide range of behaviors. Canary and Seibold’s (1984) comprehensive categorization of behaviors, examined in attitude–behavior theories (also see Kim & Hunter, 1993), indicates that research in the following categories were included in the review: health behavior (e.g., Sparks & Guthrie, 1998), consumer behavior (e.g., Hildonen, 2001), contraceptive behavior (Fekadu & Kraft, 2001), academic behavior (Theodorakis, Bagiatis, & Goudas, 1995), altruistic behavior (e.g., Rapaport & Orbell, 2000), and environmental behavior (e.g., Terry et al., 1999). The major categories in which data were not available for inclusion in the meta-analysis concern the domains of group participation, voting, race relations, religiosity, and deviance. Although it would have been desirable to include behaviors from these categories, we do not believe that their absence seriously undermines the meta-analysis, not least because identity concerns seem highly relevant for behaviors in the absent categories. Overall, it is fair to suggest that self-identity exhibited good predictive validity across a wide range of behaviors. Thus, the findings from the present review would seem to

SELF-IDENTITY META-ANALYSIS

1099

satisfy O’Keefe’s (2002) criteria and warrant the conclusion that self-identity constitutes an important additional predictor in the TPB. Theoretical implications of this conclusion also deserve attention. Attitude theorists (e.g., Eagly & Chaiken, 1993) have argued that self-identity refers to a particular class of behavioral outcomes and, therefore, should overlap with standard attitude measures that (according to both the TRA and TPB) capture these outcomes. However, two findings in particular seem to contradict this analysis. First, the variance shared by self-identity and attitude was quite modest (R2 = .14); and second, self-identity had a significant association with intention, even after the association between attitude and intention was statistically controlled. Both of these findings should have been impossible if attitude and self-identity referred to the same concept. Instead, our findings seem to be more consistent with Biddle et al.’s (1985) perspective that self-identity has different motivational origins, compared to attitude and subjective norm. According to this view, a key component of people’s motivation to formulate behavioral intentions (and, subsequently, to enact those intentions) is to reinforce, support, and confirm their sense of self (see Stets & Burke, 2000). In this context, it is timely to reinforce the idea that the role of self-identity should be interpreted in terms of socially defined influences, distinct from normative influences (Åstrøm & Rise, 2001; also see Hagger & Chatzisarantis, 2006). Thus, self-identities derive from socially constructed categories or types of person, which are accepted by individuals as descriptive of themselves (see Thoits & Virshup, 1997). In this capacity, the role of self-identity in the TPB provides an account of the failure of subjective norm (i.e., the social influence component of the TPB) as a predictor of intentions, as compared with attitudes and PBC (see Ajzen, 1991). The findings are also contrary to a self-perception theory perspective on the nature of self-identity, for similar reasons. Although there was a reliable association between past behavior and self-identity, the variance shared by the two concepts was quite modest (R2 = .11), and self-identity predicted intention, even after past behavior was controlled. Thus, although the acquisition of particular self-identities may be derived, at least in part, from experience of particular behaviors, self-identity is clearly not simply reducible to such experience. Finally, it may be worthwhile to note that self-identity is distinct from group identity, although both of them are social identities (Stets & Burke, 2000; Thoits & Virshup, 1997). While self-identity constitutes me-identification—that is, identification of the self as, say, a smoker—and includes the meanings, expectations, and activities related to being a smoker, group identity constitutes we-identification of the self with, say, other smokers, which implies acting on behalf of the group of smokers (cf. Thoits & Virshup, 1997). This distinction tends to be blurred in conceptualizations

1100 RISE ET AL. (Hagger & Chatzisarantis, 2006) and measurement of self-identity (Falomir & Invernizzi, 1999). Thus, Falomir and Invernizzi (1999) found that a measure of smoker identity predicted intention to quit smoking above the TPB components in a sample of Spanish adolescents. A close inspection of the measure of smoker identity in Falomir and Invernizzi’s study reveals that it was a mixture of self-identity as a smoker (“To what extent do you feel as a smoker?”) and identification with the group of smokers (“To what extent do you identify with the group of smokers?”). Hence, this study was not included in the present study. The basic idea is that smokers in many contexts view themselves in terms of what it means to be a smoker as a certain type of social person (“I am a smoker”), but may in another context shift to a group identity (“we smokers”) to unite in opposition to a common identity threat; for example, when health authorities restrict smoking in public places (e.g., restaurants). Evidence for this distinction was provided by Rise and Ommundsen (2010), who found that the two concepts were only weakly related and that they predicted intentions to quit smoking independently. In conclusion, the present meta-analysis provides the strongest evidence to date that the concept of self-identity is conceptually and empirically distinct from attitude, subjective norm, PBC, and past behavior. Across a wide variety of behavioral domains, the self-identity/intention relation rivaled the strength of the attitude–intention relation. Moreover, self-identity was responsible for a substantial increment in variance explained in behavioral intentions, even after the components of the TPB and past behavior had been taken into account. In our view, these findings warrant the conclusion that self-identity is a vital predictor of intentions and behavior and should be incorporated into the dominant model of attitude–behavior relations; that is, the theory of planned behavior. This suggests that self-identity may be a distinctive target for persuasive strategies, but, as noted by O’Keefe (2002) in his review of the literature, there appears to be little systematic research on identity-based influence strategies. Nevertheless, he proceeded to suggest two possible labeling strategies: to make an existing identity more readily activated, and to provide people with alternative identities. In this context, it may be important to reiterate the implications of the theoretical underpinnings of the concept; namely, that identity change is a long-term and reciprocal process. For example, behavior adjusts to conform to the meanings of the identity standard; while at the same time, the identity standard changes to adjust to the meaning of the behavior (cf. Stets & Burke, 2003). These ideas imply that when health authorities implement intervention programs directed at breaking unhealthy behavioral patterns, they must consider that they are, in effect, trying to construct new identities in the sense that the meanings, expectations, and activities

SELF-IDENTITY META-ANALYSIS

1101

associated with becoming a certain type of person or belonging to a novel social category must be incorporated into the self so as to complete the behavioral change process. To some extent, this may explain why it is so difficult to break unhealthy behavioral patterns (e.g., quitting smoking).

References (Note. Studies that were included in the meta-analysis are preceded by an asterisk.) Abraham, C., Sheeran, P., & Johnston, M. (1998). From health beliefs to self-regulation: Theoretical advances in the psychology of action control. Psychology and Health, 13, 569–591. Ajzen, I. (1988). Attitudes, personality, and behavior. Buckingham, UK: Open University Press. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179–211. Ajzen, I., & Madden, T. J. (1986). Prediction of goal-directed behavior: Attitudes, intentions, and perceived behavioral control. Journal of Experimental Social Psychology, 22, 453–474. *Armitage, C. J., & Conner, M. (1999a). Distinguishing perceptions of control from self-efficacy: Predicting consumption of a low-fat diet using the theory of planned behavior. Journal of Applied Social Psychology, 29, 72–90. *Armitage, C. J., & Conner, M. (1999b). Predictive validity of the theory of planned behavior: The role of questionnaire format and social desirability. Journal of Community and Applied Social Psychology, 9, 261–272. *Armitage, C. J., & Conner, M. (1999c). The theory of planned behaviour: Assessment of predictive validity and “perceived control.” British Journal of Social Psychology, 38, 35–54. Armitage, C. J., & Conner, M. (2001a). Efficacy of the theory of planned behaviour: A meta-analytic review. British Journal of Social Psychology, 40, 471–499. *Armitage, C. J., & Conner, M. (2001b). Social cognitive determinants of blood donation. Journal of Applied Social Psychology, 31, 1431–1457. *Arnold, J., Loan-Clarke, J., Coombs, C., Wilkinson, A., Park, J., & Preston, D. (2006). How well can the theory of planned behavior account for occupational intentions? Journal of Vocational Behavior, 69, 374– 390. *Åstrøm, A. N., & Rise, J. (2001). Young adults’ intention to eat healthy food: Extending the theory of planned behavior. Psychology and Health, 16, 223–237.

1102

RISE ET AL.

*Austin, M., & Sheeran, P. (2001). Incorporation of charity donation behavior into the self-concept. Unpublished raw data, University of Sheffield, UK. Bem, D. J. (1972). Self-perception theory. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 6, pp. 1–62). New York: Academic Press. Biddle, B. J., Bank, B. J., Anderson, D. S., Hauge, R., Keats, D. M., Keats, J. A., et al. (1985). Social influence, self-referent identity labels, and behavior. Sociological Quarterly, 26, 159–185. Biddle, B. J., Bank, B. J., & Slavings, R. (1987). Norms, preferences, identities, and retention decisions. Social Psychology Quarterly, 50, 322–337. *Campbell, S., & Sheeran, P. (2001). Self-identity and exercise. Unpublished raw data, University of Sheffield, UK. Canary, D. J., & Seibold, D. R. (1984). Attitudes and behavior: A comprehensive bibliography. New York: Praeger. Charng, H. W., Piliavin, J. A., & Callero, P. L. (1988). Role identity and reasoned action in the prediction of repeated behavior. Social Psychology Quarterly, 51, 81–105. Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159. Conner, M., & Armitage, C. J. (1998). Extending the theory of planned behavior: A review and avenues for future research. Journal of Applied Social Psychology, 28, 1429–1464. *Conner, M., & Flesch, D. (2001). Having casual sex: Additive and interactive effects of alcohol and condom availability on the determinants of intentions. Journal of Applied Social Psychology, 31, 89–112. *Conner, M., & McMillan, B. (1999). Interaction effects in the theory of planned behaviour: Studying cannabis use. British Journal of Social Psychology, 38, 195–222. *Conner, M., Warren, R., Close, S., & Sparks, P. (1999). Alcohol consumption and the theory of planned behavior: An examination of the cognitive mediation of past behavior. Journal of Applied Social Psychology, 29, 1676–1704. *de Pelsmacker, P., & Janssens, W. (2007). The effects of norms, attitudes, and habits on speeding behavior: Scale development and model building and estimation. Accident Analysis and Prevention, 39, 6–15. Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth, TX: Harcourt Brace Jovanovich. *Evans, D., & Norman, P. (2003). Predicting adolescent pedestrians’ roadcrossing intentions: An application and extension of the theory of planned behavior. Health Education Research, 18, 267–277. Falomir, J. M., & Invernizzi, F. (1999). The role of social influence and smoker identity in resistance to smoking cessation. Swiss Journal of Psychology, 58, 73–84.

SELF-IDENTITY META-ANALYSIS

1103

*Fekadu, Z., & Kraft, P. (2001). Augmenting planned behavior with selfidentity theory: Self-identity, past behavior, and its moderating effects in predicting intention. Social Behavior and Personality, 29, 671–685. Field, A. P. (2001). Meta-analysis of correlation coefficients: A Monte Carlo comparison of fixed- and random-effects methods. Psychological Methods, 6, 161–180. Field, A. P. (2005). Is the meta-analysis of correlation coefficients accurate when population correlations vary? Psychological Methods, 10, 444–467. Fishbein, M., & Ajzen, I. (1975). Beliefs, attitudes, intention, and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley. *Giles, M., McClenahan, C., Cairns, E., & Mallet, J. (2004). An application of the theory of planned behavior to blood donation: The importance of self-efficacy. Health Education Research, 19, 380–391. Godin, G., & Kok, G. (1996).The theory of planned behavior: A review of its applications to health-related behaviors. American Journal of Health Promotion, 11, 87–97. *Hagger, M. S., & Chatzisarantis, N. (2006). Self-identity and the theory of planned behaviour: Between- and within-participants analyses. British Journal of Social Psychology, 45, 731–757. *Hildonen, C. (2001). Teorien om planlagt atferd og kjøp av miljømerkede produkter [The theory of planned behavior and buying of environmental products]. Master’s thesis in Psychology, Norwegian University of Science and Technology, Trondheim, Norway. Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Correcting error and bias in research findings. Newbury Park, CA: Sage. Hunter, J. E., Schmidt, F. L., & Jackson, G. B. (1982). Meta-analysis: Cumulating research findings across studies. Beverly Hills. CA: Sage. *Jackson, C., Smith, A., & Conner, M. (2003). Applying an extended version of the theory of planned behavior to physical activity. Journal of Sports Sciences, 21, 119–133. Kim, M. S., & Hunter, J. E. (1993). Attitude–behavior relations: A metaanalysis of attitudinal relevance and topic. Journal of Communication, 43, 101–142. *Mannetti, L., Pierro, A., & Livi, S. (2004). Recycling: Planned and selfexpressive behavior. Journal of Environmental Psychology, 24, 227–236. *Moan, I. S., & Rise, J. (2006). Predicting smoking reduction among adolescents using an extended version of the theory of planned behavior. Psychology and Health, 21, 717–738. *Moan, I. S., Rise, J., & Andersen, M. (2004). Predicting parents’ intentions not to smoke indoors in the presence of their children using an extended version of the theory of planned behavior. Psychology and Health, 20, 353–371.

1104

RISE ET AL.

O’Keefe, D.J. (2002). Persuasion: Theory and research. Beverly Hills, CA: Sage. *Ouellette, J. A., & Wood, W. (1998). Habit and intention in everyday life: The multiple processes by which past behavior predicts future behavior. Psychological Bulletin, 124, 54–74. *Rapaport, P., & Orbell, S. (2000). Augmenting the theory of planned behavior: Motivation to provide practical assistance and emotional support to parents. Psychology and Health, 15, 309–324. *Rise, J., & Ommundsen, R. (2010). Predicting the intention to quit smoking using an extended theory of planned behavior: A comparative study among Spanish and Norwegian students. Manuscript submitted for publication. *Robinson, R., & Smith, C. (2002). Psychosocial and demographic variables associated with consumer intention to purchase sustainable produced foods, as defined by the Midwest Food Alliance. Journal of Nutrition Education and Behavior, 34, 316–325. Rosenthal, R. (1984). Meta-analytic procedures for social research. Beverly Hills, CA: Sage. Schwarzer, R. (1988). Meta: Programs for secondary data analysis. Berlin, Germany: Free University of Berlin. *Sheeran, P. (1998). Relevance of identity to the purchase of British beef. Unpublished raw data, University of Sheffield, UK. Sheeran, P., & Taylor, S. (1999). Predicting intentions to use condoms: Meta-analysis and comparison of the theories of reasoned action and planned behavior. Journal of Applied Social Psychology, 29, 1624–1675. Sheeran, P., Trafimow, D., & Armitage, C. J. (2003). Predicting behaviour from perceived behavioural control: Tests of the accuracy assumption of the theory of planned behaviour. British Journal of Social Psychology, 42, 393–410. Sparks, P. (2000). Subjective expected utility-based attitude–behavior models: The utility of self-identity. In D. J. Terry & M. A. Hogg (Eds.), Attitudes, behavior, and social context: The role of norms and group membership (pp. 31–46). Hillsdale, NJ: Lawrence Erlbaum. *Sparks, P., & Guthrie, C. (1998). Self-identity and the theory of planned behavior: Useful addition or unhelpful artifice? Journal of Applied Social Psychology, 28, 1393–1410. *Sparks, P., & Shepherd, R. (1992). Self-identity and the theory of planned behavior: Assessing the role of identification with green consumerism. Social Psychology Quarterly, 55, 388–399. *Spence, A., & Townsend, E. (2006). Examining consumer behavior toward genetically modified (GM) food in Britain. Risk Analysis, 26, 657–670. Stets, J. E., & Burke, P. J. (2000). Identity theory and social identity theory. Social Psychology Quarterly, 63, 224–237.

SELF-IDENTITY META-ANALYSIS

1105

Stets, J. E., & Burke, P. J. (2003). A sociological approach to self and identity. In M. R. Leary & J. P. Tangney (Eds.), Handbook of self and identity (pp. 128–152). New York: Guilford. *Terry, D. J., Hogg, M. A., & White, K. M. (1999). The theory of planned behavior: Self-identity, social identity, and group norms. British Journal of Social Psychology, 38, 225–244. *Theodorakis, Y. (1994). Planned behavior, attitude strength, role identity, and the prediction of exercise behavior. Sport Psychologist, 8, 149–165. *Theodorakis, Y., Bagiatis, K., & Goudas, M. (1995). Attitudes toward teaching individuals with disabilities: Application of the theory of planned behavior. Adapted Physical Activity Quarterly, 12, 151–160. Thoits, P. A., & Virshup, L. K. (1997). Me’s and we’s: Forms and functions of social identity. In R. Ashmore & L. Jussim (Eds.), Self and identity: Fundamental issues (Vol. 1, pp. 106–133). New York: Oxford University Press. *Thompson, M., & Rise, J. (2002). The theory of planned behavior extended: The role of past behavior and social influence. Unpublished manuscript. Trafimow, D. (2004). Problems with change in R2 as applied to theory of reasoned action research. British Journal of Social Psychology, 43, 515– 530. Wicker, A.W. (1969). Attitudes versus actions: the relationship of verbal and overt behavioral responses to attitude objects. Journal of Social Issues, 25, 41–78.