Cooperation and intentions in experimental bargaining games

Abstract

The Behavioral Bargaining Problem poses a trio of questions: (1) How do real economic agents behave in bargaining environments? (2) Why do they behave the way they do? and (3) What conditions/institutions sustain that behavior? This dissertation is about experimental results which suggest answers to each of these questions. Chapter 1 is a brief overview of early experiments which address how economic agents behave in bargaining environments. Under a wide variety of conditions a significant proportion of subjects' behavior deviates from that predicted by non-cooperative game theory. Chapter 2 tests several theories from behavioral game theory which aim at explaining why subjects cooperate in bargaining games. These models can be partitioned into two classes: outcome-based and intention-based. Outcome-based models treat the intentions that players attribute to one another as unnecessary for predicting behavior. Intention-based approaches, and in particular the trust and reciprocity hypothesis, rely on this attribution of intentions in an essential way. I report laboratory data from simple two-person trust games which is inconsistent with outcome-based models, but predicted by the trust and reciprocity hypothesis. Chapter 3 is devoted to exploring one way of sustaining cooperative behavior in a simple two-person trust environments. It is well-known in evolutionary game theory that "clustering" in a population can allow an evolutionary stable strategy to be invaded in a finitely repeated Prisoners' Dilemma game. This idea can be adapted to bargaining by noting that an agent's history of choices gives him a track record. Players can be typed based on their recent track record as whether or not they are trusting (for Players 1), and whether or not they are trustworthy (for Players 2). Once the players are typed, they can then be paired according to those types: trustors with trustworthy types, and similarly non-trustors with untrustworthy types. This sort of matching protocol induces clustering within the population. The empirical question is whether this adaptation of clustering to bargaining environments can sustain cooperative play analogous to the situation in finitely repeated Prisoners' Dilemma games. The results indicate that such a sorting mechanism allows cooperative play to emerge and, once it emerges, sustains it over time.