Definition and Background

Self-esteem is a widely used concept both in popular language and in psychology. It refers to an individual's sense of his or her value or worth, or the extent to which a person values, approves of, appreciates, prizes, or likes him or herself (Blascovich & Tomaka, 1991). The most broad and frequently cited definition of self-esteem within psychology is Rosenberg's (1965), who described it as a favorable or unfavorable attitude toward the self (p. 15).

Self-esteem is generally considered the evaluative component of the self-concept, a broader representation of the self that includes cognitive and behavioral aspects as well as evaluative or affective ones (Blascovich & Tomaka, 1991). While the construct is most often used to refer to a global sense of self-worth, narrower concepts such as self-confidence or body-esteem are used to imply a sense of self-esteem in more specific domains. It is also widely assumed that self-esteem functions as a trait, that is, it is stable across time within individuals. Self-esteem is an extremely popular construct within psychology, and has been related to virtually every other psychological concept or domain, including personality (e.g., shyness), behavioral (e.g., task performance), cognitive (e.g., attributional bias), and clinical concepts (e.g., anxiety and depression). While some researchers have been particularly concerned with understanding the nuances of the self-esteem construct, others have focussed on the adaptive and self-protective functions of self-esteem (see Blascovich & Tomaka, 1991, for a review of conceptual and methodological issues).

Self-esteem has been related both to socioeconomic status and to various aspects of health and health-related behavior, as has a related construct, self-efficacy. Self-efficacy, a term associated with the work of Bandura, refers to an individual's sense of competence or ability in general or in particular domains. Research on both constructs as they relate to SES and to health will be reviewed below.

Measurement

Self-esteem. Among the most popular and well-utilized measures of self-esteem are the Rosenberg Self-Esteem Scale (1965) and the Coopersmith Self-Esteem Inventory (I 967/1981). Rosenberg's scale was originally developed to measure adolescents' global feelings of self-worth or self-acceptance, and is generally considered the standard against which other measures of self-esteem are compared. It includes 10 items that are usually scored using a four-point response ranging from strongly disagree to strongly agree. The items are face valid, and the scale is short and easy and fast to administer. Extensive and acceptable reliability (internal consistency and test-retest) and validity (convergent and discriminant) information exists for the Rosenberg Self-Esteem Scale (see Blascovich & Tomaka, 1991).

The Coopersmith Self-Esteem Inventory was developed through research to assess attitude toward oneself in general, and in specific contexts: peers, parents, school, and personal interests. It was originally designed for use with children, drawing on items from scales that were previously used by Carl Rogers. Respondents state whether a set of 50 generally favorable or unfavorable aspects of a person are "like me" or "not like me." There are two forms, a School Form (ages 8-15) and an Adult form (ages 16 and older) (Anastasi, 1988; Blascovich & Tomaka, 1991; Pervin, 1993). Acceptable reliability (internal consistency and test-retest) and validity (convergent and discriminant) information exists for the Self-Esteem Inventory (see Blascovich & Tomaka, 1991).

Self-efficacy. Virtually all measures of self-efficacy, by virtue of the nature of the construct, are domain specific, assessing individuals' sense of competence in particular areas. Rodin and McAvay (1992; see also Seeman, Rodin, & Albert, 1993) developed and validated a self-efficacy measure designed to be particularly relevant to older adults that includes the domain of health. It taps both interpersonal efficacy (dealing with friends and family) and instrumental efficacy (finances, safety, productivity) and has a total of eight items.

Froman and Owen (1991) published a health self-efficacy measure intended for use with high school students. The 43 item scale has two subscales, Physical Health and Mental Health, and has acceptable reliability and validity (Froman & Owen, 1991). Respondents are asked to indicate their confidence in their ability to perform 43 behaviors, such things as "eating a balanced diet", "maintaining friendships", and "telling the truth".

Relationship to SES

Self-esteem. Perhaps the most famous investigation into the relationship of self-esteem to SES is Rosenberg and Pearlin's (1978) assessment of social class and self-esteem among children and adults. In an effort to clarify decades of inconclusive work on what many thought would be an obvious connection between one's social status or prestige and one's personal sense of worth, Rosenberg and Pearlin suggested that age was a critical factor in teasing apart this relationship. Indeed, they found virtually no association between social class of parents (measured by the Hollingshead Index of Social Position) and self-esteem among younger children, a modest association among adolescents, and a moderate association among adults based on their own social class. They rely on theories about social comparison processes, reflected self-appraisals, self-perception theory, and psychological centrality to explain the age graded relationship. Because the salience of class in the interpersonal context differs for children and adults, and because the social class of children is ascribed while that of adults is generally considered achieved, Rosenberg and Pearlin argue, the extent to which the sense of inequality inherent in the meaning of social class is mirrored within individuals is not the same for children as it is for adults.

Coopersmith's (1967) original work was designed to assess the origins of self-esteem in children. The results of this work in which children filled out the Self-Esteem Inventory and provided ratings of their parents, staff members interviewed mothers, and mothers filled out questionnaires, indicated that "external indicators of prestige [of the parents] such as wealth, amount of education, and job title did not have as overwhelming and as significant an effect on self-esteem as is often assumed" (Pervin, 1993, P. 189). Parental attitudes and behaviors -acceptance of their children, clear and well-enforced demands, and respect for actions within well-defined limits -- were the primary antecedents of children's sense of self-worth (Pervin, 1993).

Since the work by Rosenberg and Pearlin (1978) and Coopersmith (1967), others have explored the relationship of self-esteem to SES, especially among adolescents. With some exceptions, Rosenberg and Pearlin's results have been replicated (though it appears that more people have studied adolescents than adults). Filsinger and Anderson (1982) found no relationship between own SES (Duncan SES Index) and self-esteem (Rosenberg Self-Esteem Scale) among adolescents, but a significant relationship between the SES of the person's best friend and self-esteem. They attribute this to a heightened sense of self-efficacy among those who interact with friends who are of a higher social status than themselves, as it may be the social status of significant others from which adolescents derive their own sense of social status (p. 383). Demo and Savin-Williams (1983) replicated and extended Rosenberg and Pearlin's findings, and demonstrated that the relationship between SES (father's occupation) and self-esteem (Coopersmith Self-Esteem Inventory, plus two others to assess reflected appraisals and academic self-esteem) was greater among eighth-graders than among fifth-graders.

Richman, Clark, and Brown (1985) found a main effect for the relationship between self-esteem and SES among adolescents, but demonstrate complicated interactions of gender, race, and social class: white females (including high SES individuals) were significantly lower in general self-esteem than white males and black males and females. There has been considerable research on the relationship between race and self-esteem. As for social class, in which the expectation is that the social order will be reflected in individual self-assessments, people of color are hypothesized to have lower self-esteem than are white people. In research comparing whites and blacks, blacks often have equal or higher self-esteem than whites, and a number of theories, including those related to self-protection and disidentification, have been offered to explain these findings (see Crocker, Voelkl, Testa, & Major, 1991; Steele, 1992).

Using both traditional and non-traditional measures of social class (including father's unemployment status, neighborhood unemployment, family welfare status, and neighborhood evaluation), Wiltfang and Scarbecz (1990) found that father's education had a small positive relationship with adolescents' self-esteem and non-traditional measures had moderate to strong (neighborhood unemployment) associations with self-esteem (items from both Rosenberg and Coopersmith), all in the expected direction; they also found, however, that adolescent achievement variables (school grades, group leadership, report of many close friends) contributed significantly more to their self-esteem than did parental social class variables (P. 180).

In a study of 711 sixteen-year-olds in England, Francis and Jones (1995) found that the relationship of SES and self-esteem varied with the measure of self-esteem. There was a significant relationship between SES and the Coopersmith Self-Esteem Inventory (r = -.122, p <.001) and a moderate relationship with the Rosenberg (r =.063, p <.05).

Considerably less attention appears to have been paid to the self-esteem-SES relationship among adults. In their study of 228 employed men, Gecas and Seff (1990) were interested in the role of psychological centrality and compensation in maintaining self-esteem. Simple bivariate correlations between self-esteem (measured by a 14-item semantic differential scale) and SES were as follows: with occupational prestige, r = .21; with education, r = .16; with income, r = .08 (significance level unavailable, N = 228). There were, however, mediating effects of the centrality of particular contexts to the self. They found that when work was a central aspect of men's self-concept, occupational variables (occupational prestige, control at work) were more strongly related to self-esteem than when they were not; similarly, when home was important, home variables (control and satisfaction at home) were strongly related to self-esteem.

Self-efficacy. Clark (1996) suggested that resources, assessments of ability, and expectations about the environment all make up a sense of control, which combines with outcome expectations, physiologic states, primary and secondary experiences, and verbal persuasion, to affect self-efficacy. More specifically, he noted that individual components of SES may influence efficacy through a sense of control and active problem solving (associated with higher levels of education and occupation) and that sense of control is affected by income through material resources. In an empirical investigation of the effect of SES on exercise self-efficacy, Clark, Patrick, Grembowski, and Durham (1995) found direct effects of age and education on exercise self-efficacy, and indirect effects of age, education, income, and occupation that generally operated through previous exercise experience, satisfaction with amount of walking, depression, and outcome expectations.

Relationship to Health

Much of the research about the relationship between self-esteem and health appears to have been done in terms of the influence of self-esteem on health-related behaviors. Self-esteem has been related to such health practices as the use of birth control (Herold, Goodwin, & Lero, 1979), doing breast self-exam (Hallal, 1982), and exercise (e.g., Lih-Mei Liao, Hunter, & Weinman, 1995; Vingerhoets, Croon, Jeninga, & Menges, 1990). Self-efficacy has been related to smoking cessation, pain management, weight control, and adherence to health prevention programs (Pervin,1993). Rodin and McAvay (1992) found that older adults' decline in perceived health was associated with decreased self-efficacy. At least one study did not find a linear relationship between self-esteem and health behaviors. Hollar and Snizek (1996) found that young adults with high self-esteem and high levels of knowledge about AIDS employed safer practices for non-conventional sexual practices than those with lower self-esteem, but were riskier than those with lower self-esteem for more conventional sexual practices.

Abood and Conway (1992) found a relationship between self-esteem and health values, and between self-esteem and general wellness behavior, but not between self-esteem and tobacco or alcohol use. The relationship between self-esteem and general wellness behavior remained significant even when health values were controlled for. Rivas Torres and colleagues (Rivas Torres & Fernandez Fernandez, 1995; Rivas Torres, Fernandez Fernandez, & Maceira, 1995) examined the relationship among self-esteem, health values, and health behaviors among adolescents. They found a significant relationship between self-esteem and general health behavior for both younger and older adolescents, and that self-esteem accounted for a significant percent of the variance in mental health behavior, social health behavior, and total health behavior.