Latent class structures: taste heterogeneity and beyond

Latent class models have a long tradition in choice modelling as a tool for capturing heterogeneity across individual decision-makers. This chapter discusses the key differences between latent class structures and their continuous analogues such as Mixed Logit. The chapter looks in detail at flexibility of distributional assumptions, correlation between different sensitivities, and the post estimation inference using posterior distributions. There is also detailed discussion of recent work that has made use of latent class models to go beyond taste heterogeneity alone, looking at accommodating heterogeneity in decision rules and processing strategies.