Construct validity, dimensionality and factorial invariance of the Rosenberg Self-Esteem Scale: a bifactor modelling approach among children of prisoners

Background
The Rosenberg Self-Esteem Scale (RSES) has traditionally been conceptualised as a  unidimensional measure of self-esteem, but empirical evidence is equivocal, with some studies supporting a one-factor solution and others
favouring multidimensional models.

Participants and Procedure
The aim of this study was to examine the factor structure, factorial invariance and composite reliability of the RSES within a European sample of children affected by parental imprisonment (N = 724). The study specified and tested six
alternative factor models using conventional confirmatory factor analytic (CFA) techniques and a confirmatory bifactor modelling approach.

Results
The RSES was most effectively represented by a bifactor model including a  general self-esteem factor comprising all ten scale items and separate method effects for the positively and negatively phrased items. This model was
found to be factorially invariant among boys and girls.
Composite reliability indicated good internal consistency for the general self-esteem dimension but slightly less so for the positive and negative method effects.

Conclusions
It follows that the calculation of a total RSES score is appropriate for children of prisoners, providing that the presence of method effects is taken into consideration to avoid giving rise to false interpretations. This study demonstrated the application of a bifactorial modeling approach as a potential solution.