A Generalized Random Regret Minimization model

• Generalized model: nests random regret (RRM), random utility (RUM), hybrid models.
• Generalization holds in terms of choice probabilities, parameters, elasticities.
• Additional flexibility: G-RRM model includes attribute specific regret weights.
• Allows studying determinants of regret minimization behavior at the attribute level.
• Empirical application shows feasibility and interpretability of model estimation.

This paper presents, discusses and tests a Generalized Random Regret Minimization (G-RRM) model. The G-RRM model is created by recasting a fixed constant in the attribute-specific regret functions of the conventional RRM model, into an attribute-specific regret-weight. Given that regret-weights of different attributes can take on different values, the G-RRM model allows for additional flexibility when compared to the conventional RRM model, as it allows the researcher to capture choice behavior that equals that implied by, respectively, the canonical linear-in-parameters Random Utility Maximization (RUM) model, the conventional Random Regret Minimization (RRM) model, and hybrid RUM–RRM specifications. Furthermore, for particular values of the attribute-specific regret-weights, models are obtained where regret minimization (i.e., reference dependency and asymmetry of preferences) is present for the attribute, but in a less pronounced way than in a conventional RRM model. When regret-weights are written as binary logit functions, the G-RRM model can be estimated on choice data using conventional software packages. As an empirical proof of concept, the G-RRM model is estimated on a stated route choice dataset as well as on synthetic data, and its outcomes are compared with RUM, RRM, hybrid RUM–RRM and latent class counterparts.