A Confirmatory Factor Analysis of the Rosenberg Self-Esteem Scale Using a Sample of Adolescents of Caribbean Descent

Background/ Purpose: The Rosenberg Self-esteem Scale (RSE) is a widely used measure to assess self-esteem in children and adolescents. Despite this, the factor structure of the measure is the subject of much debate. The developer of this measure conceptualized it to be a unidimensional measure; however, researchers have found that it is multidimensional, with two factors --one factor consisting of the negative items and one consisting of the positive items. The two factors can be attributed to a method effect. Method effects are a source of measurement error that have implications for the interpretation of scores derived from a measure. The method effect that affects the RSE is due to the negatively worded items. To address the method effect on the RSE, three approaches have been used. The correlated trait correlated method model (CTCM), the correlated trait correlated method minus one model (CTC(M-1) model, and the correlated trait uncorrelated method (CTUM) model.

The purpose of this study was to examine the factor structure of the RSE using a sample of adolescents of Caribbean descent to determine if it has one or two factors. Assessing the factor structure of this measure is an important step in determining its appropriateness for use with this population. Moreover, assessing the measure’s factor structure with this population takes into consideration that Black adolescents are not uniform in their experiences.

Methods: A secondary data analysis was conducted using the National Survey of American Life –Adolescent Supplement (NSAL-A; 2001-2003), a probability sample of African American and Caribbean Black youth. The dataset is publicly available. To adjust for variation in probabilities of selection within households and nonresponse rates for adolescents and households, the NSAL-A sample was weighted. For the purposes of this study, only data from the Caribbean Black youth sample (n=360; ages 13-17) was used. All youth resided in the United States.

Measures: All 10 items from the RSE measure were used in the analysis.

Analysis Strategy: A confirmatory factor analysis (CFA) using structural equation modeling (SEM) was conducted in Mplus8.2. Sampling weights were not used in analyses, as inferences to the population were not made. Eight models were examined; each were evaluated using the following fit statistics: Chi-square, CFI, TLI, RMSEA, and SRMR. Previously described approaches were used to assess for the method effect.

Results: Consistent with previous research, we found a method effect. That is, the RSE is a multidimensional scale, consisting of two factors--one consisting of positively worded items and the other consisting of negatively word items.

Implications: When assessing the self-esteem of adolescents from Caribbean descent, it is important that the total score (i.e., summing the positively worded and negatively worded items together) not be used to determine their level of self-esteem. Research has demonstrated that treating a measure as unidimensional when it is not could lead to biased decision-making. Hence, it important that the scores be used from both factors for clinical decision-making.