Contingencies of Self-Worth in College Students: Theory ...

See discussions, stats, and author profiles for this publication at:

Contingencies of Self-Worth in College Students: Theory and Measurement

Article in Journal of Personality and Social Psychology ? December 2003

DOI: 10.1037/0022-3514.85.5.894 ? Source: PubMed

CITATIONS

560

4 authors, including:

Jennifer Crocker The Ohio State University 202 PUBLICATIONS 19,365 CITATIONS

SEE PROFILE

READS

13,138

Mary Lynne Cooper University of Missouri 112 PUBLICATIONS 16,143 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects: Relationships in the egosystem and ecosystem View project

All content following this page was uploaded by Mary Lynne Cooper on 06 September 2014.

The user has requested enhancement of the downloaded file.

PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Contingencies of Self-Worth in College Students: Theory and Measurement

Jennifer Crocker and Riia K. Luhtanen

University of Michigan

Alexandra Bouvrette

University of Michigan

M. Lynne Cooper

University of Missouri

The Contingencies of Self-Worth Scale assesses 7 sources of self-esteem in college students: academics, appearance, approval from others, competition, family support, God's love, and virtue. In confirmatory factor analyses on data from 1,418 college students, a 7-factor model fit to the data acceptably well and significantly better than several plausible alternative models. The subscales all have high internal consistency, test?retest reliability, are distinct from other personality measures, and have a simplex structure arrayed on a continuum from external to internal sources of self-esteem. Contingencies of self-worth assessed prior to college predicted how students spent their time during their 1st year of college.

It is almost axiomatic in social psychology that people seek to maintain, enhance, and protect their self-esteem (e.g., Allport, 1955; James, 1910; Rogers, 1961; Rosenberg, 1979). People are highly selective about the domains on which they stake their self-worth; only abilities or attributes on which people have "staked their salvation" are thought to contribute to global selfesteem (James, 1890). James concluded that "our self-feeling in this world depends entirely on what we back ourselves to be and do" (James, 1890, p. 45). For some people, self-esteem may depend on being attractive, loved, or competent. For others, selfesteem may depend on being virtuous, powerful, or self-reliant. In other words, people differ in the contingencies of self-worth they must satisfy to have high self-esteem (for a discussion, see Crocker & Wolfe, 2001).

James's (1890) hypothesis has profound implications for understanding the nature of self-esteem (Crocker & Wolfe, 2001). Contingencies of self-worth represent the domains in which selfesteem is threatened by setbacks and failures. Consequently, people may be most likely to show self-serving biases and defensive

Jennifer Crocker, Research Center for Group Dynamics, Institute for Social Research, and Department of Psychology, University of Michigan; Riia K. Luhtanen and Alexandra Bouvrette, Research Center for Group Dynamics, Institute for Social Research, University of Michigan; M. Lynne Cooper, Department of Psychology, University of Missouri.

The research reported in this article was supported by National Institute of Mental Health Grants R01 MH58869-01 and K02 MH01747-01. We are grateful to Phil Wood for his statistical advice on the higher order factor models.

Correspondence concerning this article should be addressed to Jennifer Crocker, Research Center for Group Dynamics, Institute for Social Research, 426 Thompson Street, Ann Arbor, Michigan 48106. E-mail: jcrocker@umich.edu

responses to negative outcomes in domains of contingency (Crocker & Wolfe, 2001). When threats in domains of contingency cannot be dismissed with defensive responses, negative events in these domains should result in drops in self-esteem, and positive events should result in increases. Thus, individual differences in stability of self-esteem across time (Kernis & Waschull, 1995) may be due, at least in part, to contingencies of self-worth interacting with relevant events (Crocker, Sommers, & Luhtanen, 2002). Because instability of self-esteem is associated with vulnerability to depression (Kernis et al., 1998; Roberts & Kassel, 1997), contingencies of self-worth in conjunction with relevant events may predict increases in depressive symptoms over time (Crocker, 2002; Crocker & Wolfe, 2001).

Contingencies of self-worth represent the domains in which goals are linked to self-worth (Crocker, 2002; Crocker & Wolfe, 2001; Wolfe & Crocker, 2002). Like all goals, these goals have self-regulatory consequences; people seek to achieve success and avoid failure in domains on which self-worth is contingent (e.g., Carver & Scheier, 1998). However, success and failure at goals linked to self-esteem result in intense affect and extreme fluctuations in self-esteem relative to equally important goals that are not linked to self-esteem (Crocker et al., 2002; Crocker, Karpinski, Quinn, & Chase, 2003). Goals linked to self-worth may be difficult to disengage from when success is slow in coming (Pyszczynski & Greenberg, 1987). Because these goals are linked to self-worth, when success is uncertain people may opt for protecting selfesteem rather than risking failure (Crocker & Park, 2003).

Internal and External Contingencies

Contingencies that are highly dependent on other people for their satisfaction, that represent relatively superficial aspects of the self, or that must be earned may be associated with lower levels of

Journal of Personality and Social Psychology, 2003, Vol. 85, No. 5, 894 ?908 Copyright 2003 by the American Psychological Association, Inc. 0022-3514/03/$12.00 DOI: 10.1037/0022-3514.85.5.894

894

CONTINGENCIES OF SELF-WORTH IN COLLEGE STUDENTS

895

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

psychological well-being than contingencies that reflect intrinsic aspects of the self or are relatively unconditional. Drawing on self-determination theory, Pyszczynski, Greenberg, and Goldenberg (2003) have argued that people are likely to function better when their self-esteem is based on core, abstract, unique features of self than when self-esteem is based on more superficial aspects of the self or on unstable aspects such as achievements or conditional approval from others. In a similar vein, Kernis (2003) has suggested that optimal self-esteem does not depend on the attainment of specific outcomes and does not require continual validation from others. From this vantage point, we would expect more internal contingencies of self-worth to be associated with higher levels of psychological well-being.

Measuring Contingencies of Self-Worth

Attempts to test the implications of James's (1890) hypotheses about individual differences in the domains on which self-esteem is staked have been hampered by problems with measurement. In some cases, contingency is inferred from the association between outcomes in the domain and self-esteem (Major, Spencer, Schmader, Wolfe, & Crocker, 1998; Osborne, 1995). For example, Osborne (1995) tested the hypothesis that Black students have disidentified with academics by examining the correlation between grades and global self-esteem in a longitudinal study of adolescents. However, this method is problematic. Some students might base their self-esteem on other domains in addition to school, leading to weaker relationships between school achievement and self-esteem even in highly identified students (Rosenberg, Schooler, Schoenbach, & Rosenberg, 1995), or they might discount the diagnosticity of their performance in a domain (Major & Schmader, 1998).

Most studies examining the consequences of individual differences in the domains on which self-worth is staked have relied on single-item ratings of the importance of the domain to assess contingencies of self-worth (e.g., Hoge & McCarthy, 1984; Marsh, 1986; Pelham, 1995; Pelham & Swann, 1989). Attributes may be personally important for a variety of reasons in addition to their role as a basis of self-esteem: They are valued in one's culture, they are a means to a desired end, they are important to one's friends or family, and so on (Deci & Ryan, 1991; Marsh, 1986, 1993). Although contingencies of self-worth are undoubtedly associated with the importance people place on domains, importance ratings only approximate James's (1890) construct of what "we back ourselves to be and do" (p. 45).

Occasionally, researchers have attempted to develop better measures of the construct of contingencies of self-worth, usually in a single domain (e.g., Major & Schmader, 1998; Schwalbe & Staples, 1991). For example, Schwalbe and Staples (1991) measured the extent to which self-esteem is based on others' approval, competence, and social comparison. However, their scales have just three items each with poor to acceptable reliability, and they are confounded with response set--for all of the items, agreement leads to a high score. More generally, researchers have yet to develop psychometrically sound measures of a variety of contingencies of self-worth.

Domains of Contingency in College Students

Our measure focuses on seven domains hypothesized to be important internal and external sources of self-esteem in previous

research and theory: others' approval, appearance, defeating others in competition, academic competence, family love and support, being a virtuous or moral person, and God's love.

Competencies

Self-esteem is derived in part from evaluations of one's specific competencies or abilities (e.g., Coopersmith, 1967; Harter, 1986; Rosenberg, 1979). Academic competence influences self-esteem in children from elementary school on (e.g., Coopersmith, 1967; Demo & Parker, 1987; Harter, 1986; Hoge, Smit, & Hanson, 1990; Rosenberg et al., 1995). Academic outcomes such as evaluations by teachers (Hoge et al., 1990), school grades (Rosenberg et al., 1995), and passing versus failing a minimum high school competence test (Richman, Brown, & Clark, 1987) are related to global self-esteem. Although college students may base their self-esteem on a wide variety of competencies, given their academic context, academic competence is likely to be an important source of self-worth.

Competition

For some people, self-esteem may be based less on competence per se than on being superior to others-- outdoing them in competition. Men, in particular, are assumed to derive self-esteem from being better than others (Cross & Madson, 1997; Josephs, Markus, & Tafarodi, 1992).

Approval From Generalized Others

Self-esteem is based in part on receiving approval and acceptance from others (e.g., Cooley, 1902; Coopersmith, 1967; Harter, 1986; Mead, 1934). Self-esteem is generally correlated with the positivity of what people believe others think of them more than how others actually see them (for reviews, see Shrauger & Schoeneman, 1979; Wylie, 1979). Others' views of the self are an important basis of self-esteem (Leary & Baumeister, 2000).

Family Support

The affection of close others may be particularly important to self-esteem (Bartholomew, 1990; Bowlby, 1982; Coopersmith, 1967). Attachment theory, for example, argues that mental models of the self as loveable and worthy of care and support are central to secure attachment (Bowlby, 1982; Griffin & Bartholomew, 1994). Perceived approval or love from family members is related to global feelings of self-worth (Harter, 1986). Our measure focuses specifically on love and support from family, because almost all college students could potentially derive self-esteem from this source.

Appearance

People in general, and females in particular, are evaluated in part on the basis of physical appearance (for a review, see Fredrickson & Roberts, 1997). One consequence of this objectification is that people learn to evaluate themselves in terms of their physical appearance. Among adolescents, for example, selfevaluation of one's physical appearance is the strongest predictor of global self-esteem for both boys and girls--stronger than self-

896

CROCKER, LUHTANEN, COOPER, AND BOUVRETTE

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

evaluations in other domains such as academics, athletics, or popularity (Harter, 1986).

God's Love

Religious faith is a powerful and important force in the lives of many Americans. National surveys of religious attitudes reveal that 90% of Americans say they believe in God or a supreme being, and two thirds say that religion plays an important role in their lives (Gallup & Bezilla, 1992). Intrinsic religiosity, which refers to the degree to which individuals internalize and express commitment to religious beliefs, is moderately and positively associated with self-esteem and other aspects of psychological well-being (e.g., Baker & Gorusch, 1984; Bergin, Masters, & Richards, 1987; Nelson, 1989). Religion may have positive effects on self-esteem through the belief that one is loved, valued, and unique in the eyes of God (P. Benson & Spilka, 1973; Blaine & Crocker, 1995; Spilka, Shaver, & Kirkpatrick, 1985).

Virtue

Another potential contingency of self-worth is one's moral adequacy or virtue (e.g., J. Benson & Lyons, 1991; Coopersmith, 1967). Adherence to a moral code may lead to the judgment that one is a good, moral, and worthwhile person (Solomon, Greenberg, & Pyszczynski, 1991).

Factor Structure of Contingencies of Self-Worth

How are these various sources of self-esteem related to each other? Some researchers have proposed that people differ not in what their self-esteem is based on but rather on whether their self-esteem is contingent or noncontingent (Deci & Ryan, 1995; Kernis, 2003). For example, Deci and Ryan (1995) have argued that self-esteem can be either contingent or "true," with true self-esteem developing naturally from autonomous, efficacious action in the context of supportive, authentic relationships. Kernis (2003) has suggested that noncontingent self-esteem is optimal and contingent self-esteem suboptimal. From this perspective, contingencies of self-worth should be captured by a single factor. In contrast, we hypothesize that distinct contingencies of self-worth have distinct correlates and distinct consequences, and therefore factor analyses of a measure that includes items assessing the seven domains we reviewed above will reveal seven distinct factors. If people tend to have relatively external contingencies or relatively internal contingencies, we would expect these factors to be correlated and have a simplex structure (Guttman, 1954) arrayed along a continuum from external to internal.

Another possibility for describing the relations among contingencies is that the seven distinct factors are organized within two higher order factors, one for external contingencies and one for internal contingencies. From this perspective, instead of being arrayed on a continuum from external to internal, contingencies of self-worth can be classified as either internal or external, and these two higher order factors capture the patterns of relations among contingencies. Finally, contingencies of self-worth may be captured best not by the specific domain on which self-esteem is based but rather by whether the items assessing these contingencies represent circumstances in which self-esteem increases, decreases, or simply depends, with the item wording ("self-esteem goes up,"

"self-esteem goes down," or "self-esteem depends") determining the pattern of shared variance among items. In other words, the pattern may be driven by method variance. Alternatively, selfregulation researchers have recently emphasized the distinction between approach and avoidance goals (Carver & Scheier, 1998; Carver, Sutton, & Scheier, 2000; Elliot, 1999). From this perspective, items worded in the direction of self-esteem increases (e.g., "My self-esteem goes up when I get a good grade") may capture approach self-esteem goals, whereas items worded in a down direction (e.g., "My self-esteem goes down when I get a bad grade") may capture avoidance self-esteem goals.

Gender and Ethnicity Differences

Gender scholars have frequently suggested that women and men differ in their contingencies of self-worth. Josephs et al. (1992) argued that the self-esteem of men "is derived, in part, from fulfilling the goals ascribed to their gender-- being independent, autonomous, separate, and better than others" (p. 392), whereas the self-esteem of women is derived, at least in part, from "being sensitive to, attuned to, connected to, and generally interdependent with others" (p. 392; see also Cross & Madson, 1997; Markus & Oyserman, 1989; Wood, Christensen, Hebl, & Rothgerber, 1997). The tendency to see oneself as superior to others is more strongly associated with self-esteem among men than among women (Josephs et al., 1992), and women are more likely to report that their self-esteem is based on others' approval whereas men are more likely to report that their self-esteem is based on social comparisons (Schwalbe & Staples, 1991). Studies 1 and 2 included both genders as participants and specifically tested for gender differences in contingencies.

Racial or ethnic differences in the bases of self-esteem have also been proposed. For example, Black Americans' self-esteem is more strongly correlated with religiosity than is the case among White Americans (Blaine & Crocker, 1995; St. George & McNamara, 1984). Compared with White Americans, the self-esteem of Black Americans may be based less on approval and regard from others (Crocker & Lawrence, 1999). Also, it has often been suggested that Black students are less likely than other students to base their self-esteem on their academic performance (e.g., Major & Schmader, 1998; Osborne, 1995; Steele, 1997). Studies 1 and 2 included an ethnically diverse sample of participants and tested for racial differences.

Study 1

Method

Participants and Procedure

Study 1 included 1,418 University of Michigan students (623 Introduction to Psychology students and 795 incoming freshmen; 510 men, 901 women, and 7 of unknown gender; ethnicity was reported as 787 White/ European American, 202 Black/African American, 207 Asian American, 154 Asian, and 68 other/of unknown ethnicity).1 Participants ranged in age from 16 to 27 years (M 18.22, SD 0.95).

1 Asians were considered to be those who indicated their ethnicity as Asian or Asian American and who also indicated that they were not born in the United States. For the sake of brevity, we refer to White Americans as Whites and African Americans as Blacks.

CONTINGENCIES OF SELF-WORTH IN COLLEGE STUDENTS

897

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

The Introductory Psychology students completed the measures voluntarily during class. The remaining participants were recruited during freshman orientation. All interested Blacks, Asians, and Asian Americans were invited to participate along with a random sample of interested Whites. Those who could participated through the World Wide Web; those without access to the Internet received a paper version of the same questions.

Materials

Ten items assessed whether self-esteem depends on outcomes in each of seven domains. Three types of items were included: (a) "up" items indicating that self-esteem increases in response to positive outcomes; (b) "down" items indicating that self-esteem decreases in response to negative outcomes; and (c) "depends" items indicating that self-esteem depends on outcomes in the domain without specifying whether the outcomes are positive or negative. Most items were worded so that "agree" responses indicated more contingent self-esteem, but some reverse-scored items were included on each subscale. The 70 items were randomly ordered within subscales, then ordered with Items 1?7 representing each of the subscales, and so on, for the resulting Contingencies of Self-Worth Scale (CSWS). Responses to each item were made on a scale from 1 (Strongly Disagree) to 7 (Strongly Agree), with the midpoint, 4, labeled Neutral.

Results and Discussion

Refinement and Validation of the Measure

The analyses proceeded in four phases. For the first and second phases, the sample was divided into two random halves (n 709 for both halves). The measure was refined (via exploratory factor analyses [EFAs]) in the first random half (Phase I), and crossvalidated (via confirmatory factor analyses [CFAs]) in the second random half (Phase II). In the third and fourth phases, the combined sample was used to compare the resultant best-fitting model to several plausible, alternative models (Phase III) and finally to test the invariance of the best-fitting model across gender and race subgroups (Phase IV).

Refinement of the Initial Model

EFAs on the first random half of the sample were used to select 35 items out of the original 70 items for the measure. Details of the procedure for selecting items are not included here because of space constraints but are available from the authors. In the end, 5 items loading on each of the seven factors representing the seven contingencies of self-worth were selected. Each of the scales included a mixture of all three item types (up, down, depends), with the exception of the Approval and Competition subscales. The former contained no items indicating that self-esteem increases in response to successes in the domain; only 2 of the original 10 items were worded in this manner, and both of these items failed to load above .40 in the initial default solution. The latter contained no items indicating that self-esteem decreases in response to failures in the domain; all 3 such items were dropped after the initial default solution because they failed to load above .40 with the remaining Competition items, they were factorially complex, or both.

Cross-Validation of the Final Model

In the second phase of analyses, we examined the generalizability of the final model by determining whether the factor structure derived in the first random half of the sample provided an adequate

fit to the data in the second random half, using CFAs. Following procedures recommended by Hayduk (1987), we specified two within-group models to test the fit of the seven-factor model to the data within each random half and then specified two betweengroup models to provide an explicit test of the equivalence of the fit across the two halves. In the first between-groups model, a common factor pattern was specified across samples, but the factor loadings were allowed to freely vary; in the final model, both the factor pattern and factor loadings were constrained to equivalence. Because the factor pattern invariant and factor loading invariant models are nested, the significance of the decrement in fit associated with the additional constraints on the factor loadings can be statistically evaluated via a chi-square difference test.

All CFAs were conducted with EQS (Bentler, 1989). Raw data matrices served as input, and maximum-likelihood estimation was used. Scaling metrics for the latent variables were fixed by setting factor variances to 1.0. Multiple fit indices were used to evaluate goodness of fit, including the nonnormed fit index (NNFI) and the comparative fit index (CFI; Bentler, 1989). Values .90 for the NNFI and CFI indicate that the specified model accounts for 90% or more of the covariance among the variables and indicate acceptable model fit (Bentler & Bonett, 1980). Chi-square statistics are also reported. A nonsignificant chi-square indicates a close fit between the variance? covariance matrix implied by the model and the observed data. However, even trivial differences between the reproduced and observed matrices can yield a significant chisquare in large samples such as the present one. The root-meansquare error of approximation (RMSEA; Steiger & Lind, 1980) is a combined measure of goodness of fit and parsimony based on the chi-square statistic. RMSEA corrects for the influence of degrees of freedom on the chi-square and may therefore provide a more reliable guide to adequacy of fit in large samples. According to Browne & Cudeck (1993), values under .08 represent reasonable errors of estimation in the population, and values less than .05 indicate a close fit. Finally, the squared mean square residual (SMSR) is reported; values less than .06 are generally thought to indicate good fit. Because each index has different limitations, consistency across indices provides the most reliable indication of goodness of fit (Cliff, 1983).

Summaries of the fit indices for the within- and between-group models are presented in Table 1. The correlated seven-factor model provided an acceptably good fit to the data in both random halves of the sample, as indicated by values above .90 for all fit indices and relatively small RMSEAs and SMSRs. All items loaded significantly on their intended factors within each sample, with standardized loadings ranging in magnitude from .43 to .95 in the first random half and from .44 to .95 in the second random half.

Examination of the modification indices revealed that the presence of significant cross-loadings depressed the overall fit of the model to the data. Because even small differences between the implied and observed covariance matrices can substantially depress overall model fit with large sample sizes, we conducted supplemental analyses in both random halves to evaluate the substantive importance of these cross-loadings. In both analyses, the five items in each random half with the largest modification indices were allowed to load on both their primary and secondary factors. Confidence in the proposed factor structure is enhanced to the extent that secondary loadings, when freely estimated, are relatively small compared with the same item's loading on its primary (or intended) factor.

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download