Reference‐Dependent Choice Models Identifying Latent Heterogeneous Reference Points of Consumers

This study develops two reference‐dependent choice models that allow reference point selection to vary across individuals and attributes. Unlike conventional models that impose a common reference point, the proposed Heterogeneous Reference Point Selection (HRPS) model identifies whether each individual evaluates each reference‐dependent attribute relative to the status quo or the expected state through a latent stochastic selection mechanism. The covariate effect‐HRPS ( CE ‐HRPS) model further incorporates individual characteristics into reference point selection. Simulation results show that the two models perform similarly when reference point selection is random, whereas CE ‐HRPS substantially improves recovery when selection is systematically driven by covariates, achieving 94.67% complete reference‐configuration accuracy compared with 72.67% for HRPS. Empirical applications using stated‐preference data reveal substantial heterogeneity in reference point use. In the vehicle market, the status quo predominates for refueling time, accessibility, drivable range, and purchase price, while reference point selection for fuel cost is more balanced. In the telecommunication market, the expected state is the dominant reference point for information searching and purchasing time, although CE ‐HRPS identifies substantial heterogeneity in purchasing‐time evaluation. Overall, the proposed models provide a flexible framework for identifying latent attribute‐specific reference points and examining how individual characteristics shape reference‐dependent choice behavior.