The Internal Structure of Responses to the Trait Emotional Intelligence Questionnaire–Short Form: An Exploratory Structural Equation Modeling Approach

Abstract

Notwithstanding the wide use of the Trait Emotional Intelligence Questionnaire–Short Form (TEIQue–SF) as a brief assessment of trait emotional intelligence (TEI), the psychometric properties of this measure have not been systematically examined. This article reports on research conducted to evaluate the latent structure underlying TEIQue–SF item data and test the gender invariance of scores as critical initial steps in determining the psychometric robustness of the inventory. In doing so, the article demonstrates an application of exploratory structural equation modeling as an alternative to the more restrictive independent clusters model of confirmatory factor analysis for examining factorially complex personality data. On the basis of 476 responses to the TEIQue–SF, evidence was obtained for the multidimensionality of the inventory reflected in a retained correlated traits solution. Tests of gender invariance revealed equivalence of item factor loadings, intercepts, uniquenesses, correlated uniquenesses, and the factor variance–covariance matrix, but not latent means. Men were found to be moderately higher on self-control and sociability than women, whereas women scored marginally higher on emotionality than men. No significant gender differences were found on mean levels of well-being. The benefits of the multidimensionality of the TEIQue–SF, limitations of the study, and directions for future research are discussed.

Acknowledgment

Thanks are extended to Peter McIlveen for comments on an earlier draft of this article.

Notes

¹ Although the correlated-traits ESEM model was retained in this study and subjected to tests of invariance, because the higher order model is the most prevalent structure of TEI in the literature, it is informative to summarize results of invariance tests of the higher order ESEM structure. These equality analyses were conducted in accordance with the taxonomy of invariance tests for second-order factor models proposed by Chen, Sousa, and West (2005), adapted for ESEM in an EwC framework. Marginal support was found for the configurally invariant higher order model (Model 1). In addition, evidence was obtained for the invariance of first- and second-order factor loadings (Models 2 and 3, respectively), intercepts of the observed indicators (Model 4), and intercepts of first-order factors (Model 5), although, for the latter model, the decrement in the CFI approached one (ΔCFI = –.009). Support was also found for the equality of item uniquenesses (Model 6), first-order factor disturbances (Model 7), and item correlated uniquenesses (Model 8). Finally, evidence was obtained for the invariance of the second-order factor variance (Model 9) and mean (Model 10). Although the absence of appreciable changes in fit between Models 9 and 10, with and without latent means constrained to equality, respectively, is indicative of the invariance of the higher order factor mean, it is instructive to note that the inspection of Model 9 revealed that men had trivially higher levels of GTEI than women (d = .175). However, caution is urged in the interpretation of these invariance tests, particularly those concerning Model 5, as the CFI, used in this study to detect appreciable differences between nested models, has been reported to be insensitive to mean structures, such that differences in the intercepts of observed and latent variables might not be detected (Chen et al., 2005). This might, in turn, obfuscate conclusions regarding the invariance of the higher order latent mean. Complete results of the higher order invariance tests may be obtained from the author by request.