Dimensionality and measurement invariance in the Satisfaction with Life Scale in Norway

TLDR

Previous studies on the Satisfaction with Life Scale have shown inconsistent dimensionality and factorial invariance findings, often based on small samples, and comparisons across gender and age groups should be approached cautiously. The study investigates the factorial structure and measurement invariance of the Satisfaction with Life Scale in a large, nationally representative Norwegian sample. The authors used confirmatory factor analysis in AMOS on 4,984 participants to test dimensionality and invariance across gender and age, finding support for a modified unidimensional structure. The analysis demonstrated metric, strong, and strict invariance across gender and age groups, confirmed sensitivity to age, and indicated that a modified single‑factor model with correlated errors between items 4 and 5 best fits the data.

Abstract

Results from previous studies examining the dimensionality and factorial invariance of the Satisfaction with Life Scale (SWLS) are inconsistent and often based on small samples. This study examines the factorial structure and factorial invariance of the SWLS in a Norwegian sample.Confirmatory factor analysis (AMOS) was conducted to explore dimensionality and test for measurement invariance in factor structure, factor loadings, intercepts, and residual variance across gender and four age groups in a large (N = 4,984), nationally representative sample of Norwegian men and women (15-79 years).The data supported a modified unidimensional structure. Factor loadings could be constrained to equality between the sexes, indicating metric invariance between genders. Further testing indicated invariance also at the strong and strict levels, thus allowing analyses involving group means. The SWLS was shown to be sensitive to age, however, at the strong and strict levels of invariance testing.In conclusion, the results in this Norwegian study seem to confirm that a unidimensional structure is acceptable, but that a modified single-factor model with correlations between error terms of items 4 and 5 is preferred. Additionally, comparisons may be made between the genders. Caution must be exerted when comparing age groups.