全球智库信息集成服务系统

Discrete choice experiments are increasingly utilized to inform policy makers in various fields in energy on consumer preferences and willingness to pay values. When translating the results into policy recommendations, it is often difficult for non-experts to understand the underlying implications of different models and associated behavioral assumptions. In this paper, I review proposed methods to compare the two most frequently applied models, the random parameters logit model and the latent class logit model and investigate the challenges in and implications of model choice for policy makers and practitioners. As an example application, I use data from a discrete choice experiment on private households’ preferences for electricity supply quality in Hyderabad, India. The procedures used in the comparative analysis – measures of fit, tests for non-nested models, kernel density estimates of conditional willingness to pay values and choice probabilities – emphasize the difficulties in finding the ‘correct’ model. The methods presented here can be readily used by other researchers to better understand model performance which ultimately contributes to improving model choice in applied energy research.

More Information and download