Abstract from O. Ashton Morgan, John C. Whitehead & William L. Huth, Journal of Environmental Economics and Policy, Vol 5, No 3, Pages: 283-297:
Traditional revealed and stated preference models consider a typical individual's behavioural responses to various policy-based information treatments. For some cost–benefit applications in which resource managers are concerned with responses from a representative individual, this is sufficient. However, as behavioural responses to information treatments can vary across respondents, we develop a latent class analysis with covariates to examine unobserved heterogeneity responses to health-risk information treatments. Results from a probabilistic model indicate that classes of consumers respond differently to the health-risk information treatments. Principally, we find that the media form of the information treatment is important, with raw consumer groups typically more responsive to a brochure information treatment, while cooked oyster consumers are more responsive to the same information in a video format. We also find that a proposed US Food and Drug Administration policy on processing all raw oysters before market has a greater effect on reducing demand for consumers of cooked oysters. However, with an associated price premium, all consumer classes reduce demand. Overall, the results suggest that future policy-based research could benefit from examining potential heterogeneity in individuals’ responses to risk information treatments in order to fully understand the efficacy of treatments on behaviour.
Have I told this story before? If so, stop now.
These data were collected before the Deepwater Horizon Oil Spill. We estimated a latent class oyster demand model to account for preference heterogeneity with variables that modeled the class structure. We sent the paper out and a referee said the latent class model was nonsense, just split the sample based on those variables and estimate separate demand models. We did as told and the paper was published [Morgan et al. (EARE 2013)]. We decided the latent class model wasn't total nonsense and shipped a paper out with a model that, for some reason, did not model the class structure. The referee said why don't you model the class structure? We thought back and remembered, oh, we did.
And if you are wondering ... Morgan et al. (Land Econ 2016) uses data from a post-spill survey and compares how consumption changes.