Essays on structural estimation of demand.

Summary Structural estimation of demand models in differentiated product markets plays an important role in economics. It allows us to better understand consumer choices and, among other things, to measure the effects of a company merger, the introduction of a new product on the market or a new regulation. The traditional approach is to specify a utility model, typically an additive random utility model, calculate its demands and invert them to obtain inverse demand equations that will serve as a basis for estimation. However, in general, these inverse demands do not have an analytical form. Estimation therefore requires numerical inversion and the use of nonlinear estimation procedures, which can be difficult and time consuming.This thesis takes a different approach, developing new inverse demand models that are consistent with a heterogeneous consumer utility model. This approach allows for more flexible capture of substitutions between products through simple linear regressions based on data including market shares, prices and product characteristics. The first chapter of this thesis develops the inverse product differentiation logit (IPDL) model, which generalizes nested logit models, allowing for flexible capture of substitutions between products, including complementarity. It shows that the IPDL model belongs to a class of inverse demand models, called generalized inverse logit (GIL), which includes a large majority of additive random utility models that have been used for demand estimation purposes. The second chapter develops the flexible inverse logit (FIL) model, a GIL model that uses a flexible nesting structure with one nest for each pair of products. It shows that the FIL model, projected into the product characteristic space, yields price elasticities that depend directly on product characteristics and, using Monte Carlo simulations, is able to reproduce those of the "flexible" logit model with random coefficients. The third chapter studies the microfoundation of the GIL model. It shows that the restrictions that the GIL model imposes on the inverse demand function are necessary and sufficient conditions for consistency with a model of heterogeneous consumers maximizing their utility function, known as the perturbed utility model (PUM). He also shows that any GIL model generates a demand function that satisfies a slight variant of the Daly-Zachary conditions, which allows the combination of substitutability and complementarity in demand.