EWA learning in bilateral call markets

Abstract

This paper is about learning in bilateral call markets. In these markets, a buyer and a seller are privately informed of their values and submit their bids anonymously. If the buyer’s bid is (weakly) more than the seller’s ask, they trade at the midpoint of their bids. Understanding learning in bilateral call markets serves as a foundation for studying learning in more complex market institutions such as posted offers and double auctions. It also forces a generalization of learning models developed for simpler games to environments in which learning contingent on one realized random variable, such as a buyer’s valuation in one trial. Similarity-based generalization is a natural way to extend what is learned locally, which is undoubtedly important when people learn in very complex environments (which has not been thoroughly explored experimentally).