Consumer demand with unobservable product attributes

In this paper we discuss statistical inference associated with the theoretical model developed in Part I. Specifically, we demonstrate how the relationship between the distribution of prices and unit values can be exploited to estimate some of the structural parameters. These estimates are essential for constructing price indexes that account for unobservable taste-shifters and quality/location attributes. Finally, the remaining structural parameters can be estimated from data on demand by inserting the price indexes into the corresponding demand system. Finally, we discuss the estimation procedure in the discrete case when consumers choose one unit of a variant at a time.