Abstract

Chile has substantially expanded its renewable energy supply over the last 10 years. During this transition, Chileans faced a difficult decision about how to produce electricity, and where production should take place. Such decisions can be informed by discrete choice models, that characterize the welfare impacts of alternative electricity production scenarios on the public as a whole, as well as on subgroups of the population. However, it is difficult to determine subgroup impacts when there are many possible subgroups that could have heterogeneous preferences. We describe an approach that uses a hybrid adaptive sparse multinomial logit model to automatically determine subgroups that have heterogeneous preferences while also producing unbiased estimates. We apply the approach to survey data from 486 heads of household in Santiago, Chile, finding that the sparse multinomial approach fits the data better than the mixed logit model, and that it uncovered 16 interpretable two-way interactions between attributes of electricity production alternatives and respondent characteristics.