Lund University Publications

Psychometric properties of Rosenberg’s self-esteem scale among adolescents : a Rasch model analysis

• Mark

Abstract

Background – The Rosenberg Self-Esteem Scale (RSES) is one of the most widely used measures of global self-esteem. While its psychometric properties have been extensively studied using classical test theory, limited evidence exists on its performance under modern test theory approaches. Aim – To evaluate the psychometric properties of the RSES using Rasch analysis in a sample of Swedish adolescents. Methods – A total of 403 adolescents (aged 9–19 years) from six schools completed the RSES. Rasch modelling was applied to assess item fit, person-item targeting, response category functioning, reliability, and differential item functioning (DIF) across sex, age, and BMI. Results – The RSES demonstrated good item fit, with most items... (More)

Background – The Rosenberg Self-Esteem Scale (RSES) is one of the most widely used measures of global self-esteem. While its psychometric properties have been extensively studied using classical test theory, limited evidence exists on its performance under modern test theory approaches. Aim – To evaluate the psychometric properties of the RSES using Rasch analysis in a sample of Swedish adolescents. Methods – A total of 403 adolescents (aged 9–19 years) from six schools completed the RSES. Rasch modelling was applied to assess item fit, person-item targeting, response category functioning, reliability, and differential item functioning (DIF) across sex, age, and BMI. Results – The RSES demonstrated good item fit, with most items aligning with the Rasch model assumptions. The reliability and internal consistency indicators were strong, reflecting the sound psychometric quality of the scale. The person reliability coefficient reached a high value (0.85), with PSI = 2.35, indicating the scale’s effective ability to discriminate among individuals across different levels of self-esteem. In addition, item reliability was high (0.99), accompanied by ISI = 8.38, reflecting the precision and stability of item difficulty estimates. Cronbach’s alpha was 0.87, with SEM = 2.11, all of which confirm a high level of reliability and internal consistency among the scale items. Regarding targeting, the range of participant abilities was broader than that of item difficulties, with adequate coverage of the central portion of the self-esteem continuum, though some challenges remain in representing extreme levels of the trait. DIF analyses revealed no significant bias across sex, age, or BMI, though several items displayed variation across self-esteem levels. Conclusion – The RSES showed acceptable Rasch model fit, structural integrity, and reliability among Swedish adolescents, with no evidence of DIF by sex, age, or BMI. However, suboptimal targeting and misfitting reverse-worded items limited measurement precision, particularly at very low and very high self-esteem levels. Adding items targeting the extremes and revising problematic items could enhance the RSES’s utility for adolescent self-esteem assessment in both research and clinical contexts.

(Less)