Academic journal article
By Moore, Sarah; Laflin, Molly T.; Weis, David L.
Adolescence , Vol. 31, No. 123
A commonly held belief by many researchers, teachers, school administrators, and health professionals is that low self-esteem (SE) is associated with drug use and/or abuse (Dielman, Campanelli, Shope, & Butchart, 1987; Dielman, Leech, Lorenger, & Horvath, 1984; Steffenhagen & Steffenhagen, 1985; Wright & Moore, 1982; Young, Werch, & Bakema, 1989). That is, a poor self-concept and low evaluations of one's self worth are posited to be related to, and, moreover, cause tobacco, alcohol, marijuana, and/or other drug use. Proponents of this belief hold that because tobacco, alcohol, and other types of drugs are harmful, only those with low self-worth would be inclined to ingest such substances. Accordingly, many drug prevention and treatment efforts have been directed at enhancing the SE of youth (Dielman et al., 1984; Schroeder, Laflin, & Weis, in press).
Empirical research examining the SE-drug and alcohol use/abuse relationship has been mixed, with some studies reporting a statistically significant relationship between these two variables, and others failing to find such a relationship. For example, Steffenhagen and Steffenhagen (1985) found no significant correlation between SE and alcoholism, but did find a significant correlation between depression and alcoholism. Wright and Moore (1982) found a statistically significant relationship between SE and drug use, although they found a stronger relationship between drug use variables and general life satisfaction than the drug use measures and SE. Dielman et al. (1984) likewise concluded that the relationship between SE and drug use, although statistically significant, was not large enough to suggest that intervention programs be directed toward the enhancement of SE. Similarly. Dielman et al. (1987) reported significant bivariate correlations between drug use variables and "family," "peer," and "school adjustment" indices of SE. For any of the correlations, however, SE accounted for no more than 4% of the variance in drug use. Thus although statistically significant, the small effect size raises questions about the practical importance of the association.
Researchers in other studies, however, have concluded that SE is not only statistically significant, but meaningful and useful in the prediction of drug use. For example, Young et al. (1989) measured three types of SE in over 2,000 4th to 9th graders: school SE, home SE, and peer SE. They reported significant differences in 18 of 19 drug use variables for home and school SE, but not peer SE. These researchers concluded that examination of specific facets of SE would be a useful avenue for future research.
In a recent review of the literature that examined the relationship between SE and drug use, Schroeder et al. (in press) concluded that much of the research was fraught with conceptual and methodological shortcomings such as poor measurement of both SE and drug use, failure to consider confounding variables, and the inappropriate inference of causality from correlational data. On the basis of their review, these researchers concluded that the effects of SE on drug use/abuse is inconsistent, and when the relationship is statistically significant, the effects are minimal. Thus, drug prevention efforts directed at SE enhancement are unlikely to prove fruitful.
Nevertheless, the association between SE and drug use is intuitively appealing and persists in the literature. Moreover, given the conceptual and methodological shortcomings outlined by Schroeder et al. (in press), and noted by others (e.g., Bingham, 1983; Gilberts, 1983), failure to find significant relationships may be due to poor methodology. That is, true absence of a relationship between two variables is but one reason for failure to find a relationship between SE and drug use. Poor measurement, treatment contamination, low power, lack of variance in the variables, and poor modeling of the relationship between the variables are but a few reasons for failure to find statistical significance between two or more variables. …