Simple Structure Assumptions in Bidirectional Constructs: Demonstrating ESEM as a Diagnostic Framework for the Work and Family Conflict Scale

Instruments that measure bidirectional constructs, where two directions of influence in the same domain are shown by different sets of items, may face misspecification in confirmatory factor analysis (CFA) if there are significant cross-loadings. Using the Work and Family Conflict Scale (WAFCS) as an illustrative case, this study examines how exploratory structural equation modeling (ESEM) can provide item-level information about dimensional separability beyond global model fit indices (e.g., χ2, CFI, TLI, RMSEA, SRMR). Data were collected from 1,173 working parents in Indonesia and randomly divided into calibration (n = 596) and validation (n = 577) subsamples for the analysis. None of the CFA specifications met the conventional fit criteria. The best-fitting two-factor correlated CFA model yielded χ2(34) = 262.11, CFI = .91, TLI = .88, RMSEA = .11, and SRMR = .05. ESEM demonstrated improved fit in both the calibration (χ2(26) = 130.27, CFI = .96, TLI = .93, RMSEA = .08, SRMR = .02) and validation subsamples (χ2(26) = 48.50, CFI = .99, TLI = .98, RMSEA = .04, SRMR = .01). Beyond improved global fit, ESEM revealed differences in item-level discriminability, including substantial cross-loadings, weak dimensional differentiation, and items whose classifications were not replicated across subsamples. These findings demonstrate that global model fit alone does not establish item-level separability and that combining ESEM with internal calibration–validation provides a more informative diagnostic framework for evaluating bidirectional constructs.