In-person interviews were carried out with Mexican millers who were administered a conjoint-type survey designed to incorporate uncertainty in attribute levels. Two methods were used to model millers’ risk preferences: a modified mean-variance approach and an explicit expected utility...