Blurred Boundaries: A Flexible Approach for Segmentation Applied to the Car Market

Abstract

Prominent features of di¤erentiated product markets are segmentation and product proliferation that blurs the boundaries between segments. I develop a tractable demand model, the Ordered Nested Logit, which allows for overlap between neighboring segments. I apply the model to the automobile market where segments are ordered from small to luxury. I …nd that consumers, when substituting outside their vehicle segment, are more likely to switch to a neighboring segment. Accounting for such asymmetric substitution matters when evaluating the impact of new product introduction or studying the e¤ect of subsidies on fuel-e¢ cient cars.

Demand Models for Differentiated Goods with Complementarity and Substitutability

“…This is true, for example, for hotels that can be ordered according to their number of stars and for breakfast cereals according to sugar content. The example below provides GEM that is similar to the GEV ordered models of Small (1987) and Grigolon (2017).…”

“…This allows us to build GEM that are similar in the spirit to existing generalized extreme value (GEV) models that have already proved useful for demand estimation purposes. We show how to build ordered models describing markets having a natural ordering of alternatives (see Small, 1987;Grigolon, 2017) and nested models that generalize multi-level nested logit models in the spirit of Bresnahan et al (1997).…”

“…See e.g., the ordered logit models(Small, 1987;Grigolon, 2017); the PDL model(Bresnahan et al, 1997); and the flexible coefficient MNL model(Davis and Schiraldi, 2014).…”

An Empirical Model of Quantity Discounts with Large Choice Sets

“…The PONL model is a specialization of Generalized Nested Logit (GNL) model (Abbe et al, 2007;Ben-Akiva and Bierlaire, 1999;Bierlaire, 2006;Wen and Koppelman, 2001), the most general example of GEV model, to the case of demand for bundles. Our econometric treatment of the PONL model contributes to a small but growing literature on the use of GEV models for the convenient estimation of demand from aggregate market-level data (Bresnahan et al, 1997;Davis and Schiraldi, 2014;Fosgerau et al, 2021;Grigolon, 2021), as initiated by Berry (1994) for the multinomial logit and the nested logit. In particular, our constrained 2SLS estimator and iterative procedure can be helpful also for the estimation of other GNL models with large choice sets.…”