Using Experimental Data to Model Bargaining Behavior in Ultimatum Games

Subgame perfect equilibrium predictions of ultimatum bargaining games correspond poorly to the data gathered from human subjects in laboratory environments. Attempts to reconcile this discrepancy have taken one or more of three routes: (1) expanding the agent foresight and scope of decisions, (2) explicit modeling of agents' initial beliefs and their dynamics, and (3) adding social arguments to agent preferences. We take the first two routes by including the probability of rejection by the responder in proposer's decision, and using experimental data to estimate a static model of agent beliefs. Data from previously reported experiments is compared to the predictions of the optimal decision rule to validate the proposer model. Models in which the probability of acceptance of a proposal declines with the amount offered to the responder are better able to organize the data about the behavior of both players. Explanation of responders' behavior remains weak