Identifying the presence of heterogeneous discrete choice heuristics at an individual level

Abstract

Discrete choice models that allow heterogeneous choice heuristics have been recently proposed and applied in different contexts. These models are traditionally based on latent classes, where each class represents a choice heuristic. Different specification challenges arise from applying the traditional approaches and identifiability issues are not unusual. We propose a Mixed Heuristic Model (MHM) to identify the presence of different choice heuristics through a latent class approach, giving higher flexibility to the class membership function through a random variable and analyse it at an individual level. The MHM identifies individuals who are likely to have followed the choice heuristic without explicitly specifying the class membership function. The MHM also avoids specifying the class membership and choice levels simultaneously. The MHM is tested on simulated data and applied to model an air travel survey. Results show that the MHM is able to identify the presence and structure of the heuristics with high accuracy within the simulated data. On the real data, the MHM identifies Random Utility Maximization and Stochastic Satisficing behaviour within the individuals.