Does digital trust discriminate? Testing construct validity and mediation among Indonesian Generation Z

Type of the article: Research ArticleAbstractDigital trust is widely modeled as the mechanism through which influencer credibility, e-WoM, and social media interaction shape the decisions of Generation Z consumers, yet studies rarely verify that trust is empirically distinct from the outcome it mediates. Using survey data from 255 Indonesian Generation Z respondents, this study tests whether digital trust and purchase decision-making are empirically distinct constructs, and whether indirect effects are correctly derived from the model’s own structural coefficients rather than a differently specified equation. Influencer credibility, e-WoM, and social media interaction all positively predicted digital trust (β = 0.307, 0.323, 0.291), and trust strongly predicted purchase decision-making (β = 0.908, R² = 0.825). Applying the Fornell-Larcker criterion, HTMT, and a nested chi-square difference test, trust and the outcome show weak discriminant validity (HTMT = 0.904, above the conventional ceiling) despite a Chi-square test rejecting their full merger, a disagreement that a single criterion would not reveal. Correctly specified mediation finds all three sources produce significant, similarly sized indirect effects (0.264–0.293), preserving their direct-effect ranking; a separately computed reduced-form regression bypassing trust produces a starkly different pattern (0.503, 0.302, 0.077) that is not a trust-mediated effect but an artifact of multicollinearity among the three sources. These results show that overall model fit and an unverified validity claim are insufficient for mediation claims, and that Generation Z trust-formation studies require multi-criterion construct verification before accepting indirect-effect claims.AcknowledgmentsNo external funding was received for this study. The authors declare no acknowledgements.