| This project analyzes the dynamics of the trading mechanism and the properties of the equilibrium outcomes observed in markets where sellers announce their demands, but do not have to commit, while buyers can move back and forth between the sellers and negotiate to get a better deal. Furthermore, it characterizes the optimal strategies of the market participants and identifies the conditions that the sellers prefer to compete with posted or bargain prices. The baseline game theoretic model used is the following. A buyer facing two spatially-separated sellers can negotiate with only one potential seller at a time. However, with some delay, the buyer can move back and forth between the two sellers. The introduction of behavioral types that are completely inflexible in their demands and offers, even with low probabilities, makes equilibrium essentially unique. This equilibrium has a war of attrition structure that engenders inefficiency due to possible delay in reaching an agreement. The maximum length of the negotiation that will take place before the buyer leaves to bargain with the other seller is an increasing function of the distance between the stores and the reputation of the second seller. If the sellers post the same initial price, the buyer will never visit one seller more than once in equilibrium. When initial posted prices are different, a given seller may be visited twice, and the buyer may choose to go first to the seller with the higher posted price. If the initial priors about the players' types are sufficiently small, in the unique equilibrium where the sellers are constrained to announce their prices before the buyer arrives at their stores, each seller posts the price of zero. However, this uniqueness result is not robust to perturbation on higher order beliefs about initial priors. If the sellers cannot announce their prices before the buyer's arrival, then a continuum of prices can be supported in equilibrium. But, independent of the sellers' announcements, the payoff to the seller (visited by the buyer first) is always the same, and the buyer's payoff approaches to a unique limit as the initial priors vanish. |