Wednesday, 26th February 2020
14:00 - 15:00
Clay Room - Nuffield College
History, Efficiency and Implementation: An Experiment
(with James Best and Daniel Quigley)
In standard repeated games settings predictions are difficult due to the existence of folk-theorems. At the same time repeated interactions frequently take place in environments where a third party controls what historical data can be observed about past behaviour. Theoretical results in Best and Quigley (2020) suggest that control over historical data allows for both the sustaining of Pareto superior equilibria and the partial selection of equilibrium outcomes. This latter result is of particular value for making testable predictions about equilibrium outcomes in repeated games. We experimentally study these theoretical results in an indefinitely repeated game setting, in which a long-run Firm plays against a sequence of short-run Investors. We introduce a simple Badge system, in which Good or Bad badges are granted to the Firm and are visible to the Investors, which represent a summary of the Firm’s past behaviour. Importantly, these badges restrict what other historical data is available to the Investors. We compare this with a full history control condition, in terms of two aspects. First, we want test whether the Badge system can provide efficiency gains by supporting Pareto superior equilibria when the Firm is relatively patient. Second, we test whether the badge system can help select particular equilibrium outcomes.