Equilibrium Selection through Laboratory Experiments in a Complex OLG Economy

In this thesis I conduct a learning-to-forecast experiment to investigate which equilibria are selected empirically in a complex OLG economy. I use the overlapping generation model described in Araujo et al. (2000) with multiple perfect-foresight equilibria. Theory and theoretical selection criteria like learning cannot help in solving the problem of equilibrium selection. Laboratory experiments might offer a solution. A second question posed by this thesis is whether coordination on simple (steady state or 2-cycle) equilibria will arise, in line with Ariofovic et al. (2019), or whether more complex (chaotic) equilibria can arise. In the model of Araujo et al. (2000), dynamics depends on one parameter of the utility function. I conduct 5 treatments with different values of this parameter. The values were chosen based on the dynamics under naive expectations, and they correspond to: (i) a stable steady state, (ii) a stable 2-cycle, (iii) a stable 3-cycle, and (iv) chaos. In line with Arifovic et al. (2019), coordination on the simplest equilibria, like the steady state and 2-cycle, occurred. Coordination happened very quickly, but faster in the sessions which converged to the steady state. The dynamics was stable, and price converged to the steady state in the majority of groups. Surprisingly, highly nonlinear, chaotic dynamics in the model leads to coordination on the steady state in the lab.