Partner choice and cooperation in networks. Theory and experimental evidence. (Q2468954): Difference between revisions

Let \(G\) be a network game played by a fixed set of players finitely many times. This monograph considers the theory of such a game, conducts a large controlled laboratory experiment of such a game, and computes simulations of play of such a game. The main result of the theory chapter, 3, is to show the existence of and characterize a nearly efficient subgame perfect equilibrium of \(G\) when players are free to choose the network. The logic of the proofs should be familiar to any reader familiar with \textit{J. P. Benoit} and \textit{V. Krishna} [Econometrica 53, 905--922 (1985; Zbl 0588.90095)]. The laboratory experiments were conducted at CREED, at the University of Amsterdam. The design of the experiment is clean, and the computer screen seen by the subjects is especially nicely done. The main results of the experiment are as follows. First, endogenizing the network increases observed cooperation (hence, efficiency), relative to a baseline of an exogenously fixed network. Second, with an endogenously chosen network, significant cooperation persists even in the end period. Indeed, the amount of end-period cooperation eludes explanation even by social utility models. This is a very challenging result. The simulation setup and runs are quite standard. Here we focus on treatment effects. A high outside option induces more cooperation than a low outside option. Forward-looking behavioral rules are sufficient for the existence of cooperation, while myopic best response converges to noncooperation. Social utility does not substantiallyi change the results. This was a solid doctoral dissertation at the University of Amsterdam, and the dissertation supervisor, A. Riedl, should be proud.