The Generalizability and Reliability of Scores on the Rosenberg Self-Esteem Scale Over Forty-Eight Years

Abstract

The objective of this study is to test the reliability of scores on the Rosenberg Self-Esteem scale (RSES; Rosenberg, 1965) from 1971 to 2019. Self-esteem is how highly one thinks of themselves and how much worth they feel they possess. The RSES is not the only measure of global self-esteem, but it is the most widely used (Whiteside-Mansell & Corwyn, 2003). In the late 1980’s and 1990’s, the self-esteem movement was enacted in the United States as an effort to improve the lives of adults and children, which may have changed the way self-esteem is interpreted (Humphrey, 2004). For example, items on the RSES may now be measuring narcissism or self-efficacy more so than self-esteem. Thirteen existing item-level datasets that used the RSES were obtained. Sample sizes varied, but all samples contained young adults in the United States between the ages of 15 and 26. Two Classical Test Theory (CTT; Meyer, 2010) coefficients were used, Cronbach’s α and Guttman’s λ_2. to test the reliability of scores on the RSES by finding the ratio of true score variance to observed score variance. Generalizability Theory (G-Theory; Shavelson & Webb, 1991) components, specifically G-studies and D-studies, were then used to identify the sources of variance present in scores on the RSES (i.e., variance from persons, variance from items, and remaining unexplained and error variance). The CTT coefficients and G-theory methods indicated scores on the RSES were reliable and, if anything, have increased slightly in reliability over the last 48 years. Despite the reliability of the scores on the RSES, the validity of the scores are still in question. It is important to periodically test the reliability of scores on widely used measures like the RSES to determine the extent to which they can be used for various forms of decision-making (Meyer, 2010) and for various research aims.