A Framework for empirical modelling of consumer demand with latent quality attributes

This paper discusses a particular approach to empirical consumer demand modelling when products are differentiated and the product attributes are unobservable. In contrast to the traditional approach to this problem, see e.g. Epple (1987) and Deaton (1987, 1988), where the product variants are treated as infinitely divisible goods, the present approach assumes that the consumer is making his choice of variant from a set of discrete "packages" of attribute combinations. Subsequently, given the (discrete) choice of variants the corresponding quantities are treated as continuous choices. Thus in this approach the consumer's decision process is formulated as a discrete/continuous choice problem. The empirical analysis is based on microdata from the Sichuan province in China. We show that in this case the estimation methods work well and yield reasonable results.