Calendrier du 17 septembre 2020
TOM (Théorie, Organisation et Marchés) Lunch Seminar
Du 17/09/2020 de 12:30 à 13:30
salle R1-14, campus Jourdan - 75014 Paris
WEBER Giacomo (PSE)
Clustering and Analogy-based Expectation Equilibrium with an Application to Product-differentiated Oligopoly
écrit avec Philippe Jehiel
We combine together the notion of Analogy-based Expectation Equilibrium (ABEE) and the K-means clustering algorithm. ABEE is a bounded rationality equilibrium notion that tackles cognitive rationality: players form beliefs bundling games together, the collection of these bundles for a player is called analogy partition. Bundles are primitives in the ABEE original framework; behavior is endogenous. K-means is a machine learning problem of unsupervised learning. K-means algorithm is typically used to identify underlying structure in data, forming classes (clusters). Therefore, data is a primitive; while bundling is endogenous. Both concepts aggregate data. In ABEE, data is about players’ behavior. Our objective is to endogenize the way players bundle games in the ABEE in the spirit of K-means. Thus, both bundles and behavior are endogenous in our model.
We provide a mixed extension of a strategic environment to ensure existence of stable analogy partitions. We apply this concept of stability to oligopolistic markets with differentiated products. When firms compete à la Bertrand, we find multiple stable ABEEs. Instead, when firms compete à la Corunot, sometimes stable ABBE can be found only in the mixed extension.