An assessment of the current use of hybrid choice models in environmental economics, and considerations for future applications

This study examines the use of hybrid choice models (HCM), also referred to as integrated choice and latent variable (ICLV) models, within environmental valuation studies. The investigation is motivated by the fact that stated preference surveys in this field increasingly incorporate additional data into their modelling, particularly respondents' attitudes towards the environment in a broader context or specifically towards the environmental changes under evaluation. Key findings include the fact that sample sizes are usually too small for such complex models, that many studies use ad hoc scales as indicators of latent variables without first testing the validity and reliability of the scales, and that model results are often not compared with a benchmark model. One particularly notable finding of the simulation study is that excluding a latent variable, such as estimating Random Parameter Logit (RPL) instead of HCM, does not necessarily lead to biased willingness to pay (WTP) estimates. Therefore, if the inclusion of a latent construct is not critical to the study, we suggest opting for more traditional and robust models such as RPL or Latent Class Models (LCM). The perceived benefits of gaining a better understanding of how latent factors influence decisions come with risks associated with defining and estimating an HCM. To improve the quality of research, we provide recommendations for future applications of HCM in environmental economics. • HCMs are complex and resource-demanding models. • Sample sizes for complex HCMs are often limited, ad hoc scales lack validation, and benchmark model comparisons are missing. • HCMs help understand latent variable impacts, while RPL models or LCMs are better suited for accurate WTP estimations. • Caution is needed in the selection and representation of indicators.