Bassel Tarbush

We bring the decision theoretic notion of satisficing, according to which agents play good but not necessarily their best action, to games. We provide static and dynamic foundations for the concept, and consider applications.

We characterize the outcomes of a canonical theoretical model of the intergenerational transmission of capital that features differential fertility. We establish easy-to-verify conditions on the primitives to guarantee global convergence to degenerate or atomless steady state distributions of capital.

We show that for any ϵ>0, as the number of agents gets large, the share of games that admit a pure ϵ-equilibrium converges to 1. The asymptotic share of games that admit a pure Nash equilibrium (i.e. ϵ=0) is approximately 1-1/e. Conclusion: very small deviations from perfect rationality suffice to ensure the general existence of stable outcomes.

Among generic games that have a pure Nash equilibrium, all but a small fraction are connected, i.e. every non-Nash profile can reach every Nash via directed best-response paths. Implication: there are simple adaptive dynamics that converge almost surely to a pure Nash equilibrium in all but a small fraction of generic games that have one. This limits the scope of known impossibility results about learning pure Nash.

When played according to a fixed cyclic order, the best-response dynamic converges to a pure Nash equilibrium in a vanishingly small proportion of all large generic games.

We evaluate the general equilibrium impact on aggregate emissions of taxing the emissions of any set of sectors in an intersectoral production network.

In any symmetric equilibrium of the pure location game, as the number of firms becomes large, the market share distribution converges to a Gamma(2,2) distribution. The result holds regardless of the distribution of consumers.

Dynamic network formation model on a fixed number of nodes and overlapping social groups. Homophily can be non-monotonic over time.

Interpreting the Sure-Thing Principle as a counterfactual proposition resolves issues with the original “agreeing to disagree” result.