Searching for Arms

DEPARTEMENT ECONOMIE ET SCIENCES DE LA DECISION

Intervenant :  Alessandro PAVAN (Northwestern Univ.)

HEC Campus- Bâtiment T - Salle T015

Abstract :

We study a model of experimentation or sequential learning where the set of al-ternatives is endogenously shaped by past search decisions. The environment reflectsa tradeoff between exploration of existing alternatives and search for additional ones,faced with uncertainty about the set of outside alternatives that may be discovered. De-spite the effect of search on the future set of alternatives and the potential correlationbetween the types of alternatives found over time, we obtain a simple characterizationof the optimal policy. We extend our analysis to environments in which an irreversiblechoice may be made among the endogenous set of alternatives, at any time and givenany amount of potentially inconclusive information. As a special case, the analysisyields a generalization of Weitzman’s (1979) canonical Pandora’s boxes problem, aswell as its solution, to an environment where the set of boxes is endogenous.