Comparing CB-SEM, PLS-SEM, and GSCA on Trust and Reuse Intention of OTA PayLater Users

Structural equation modeling (SEM) offers several estimation approaches, yet applied studies rarely show how the choice of method changes substantive conclusions. This study compares covariance-based SEM (CB-SEM), partial least squares SEM (PLS-SEM), and generalized structured component analysis (GSCA) using one dataset from 120 Millennial and Generation Z users of OTA PayLater. Six constructs (service quality, perceived risk, promotion, user satisfaction, trust, and reuse intention) and nine hypotheses were tested. The methods were compared on the measurement model, structural model, simultaneous equations, R-squared, estimated versus actual values, and the hypothesis test recap. All three methods produced valid and reliable measurement models, but CB-SEM required model modification, PLS-SEM removed two trust indicators, and GSCA showed two HTMT values above 0.90. Eight of nine hypotheses received identical decisions; only the effect of promotion on reuse intention was significant in GSCA alone. CB-SEM produced the highest R² (0.889 and 0.911) and the closest estimate of reuse intention (difference 0.207), whereas PLS-SEM gave the closest estimate of trust (difference 0.002). No method dominated every criterion, so the method should follow the research purpose