Game theory has had a deep impact on the theory of industrial organization, in a similar (but less controversial) way as the rational expectations revolution in macroeconomics. The reason it has been embraced by a majority of researchers in the field is that it imposes some discipline on theoretical thinking. It forces economists to clearly specify the strategic variables, their timing, and the information structure faced by firms. As is often the case in economics, the researcher learns as much from constructing the model (the extensive form) as from

Drew Fudenberg, and Jean Tirole, “The Fat-Cat Effect, the Puppy-Dog Ploy, and the Lean and Hungry Look”, in Industrial Organization, Oliver E. Williamson (ed.), Edward Elgar Publishing, 1990, pp. 450–455.

Justified communication equilibrium (JCE) is an equilibrium refinement for signaling games with cheap-talk communication. A strategy profile must be a JCE to be a stable outcome of nonequilibrium learning when receivers are initially trusting and senders play many more times than receivers. In the learning model, the counterfactual “speeches” that have been informally used to motivate past refinements are messages that are actually sent. Stable profiles need not be perfect Bayesian equilibria, so JCE sometimes preserves equilibria that existing refinements eliminate. Despite this, it resembles the earlier refinements D1 and NWBR, and it coincides with them in co-monotonic signaling games.

We use machine learning to uncover regularities in the initial play of matrix games. We first train a prediction algorithm on data from past experiments. Examining the games where our algorithm predicts correctly, but existing economic models don’t, leads us to add a parameter to the best performing model that improves predictive accuracy. We then observe play in a collection of new “ algorithmically generated” games, and learn that we can obtain even better predictions with a hybrid model that uses a decision tree to decide game-by-game which of two economic models to use for prediction.

A principal specifies time paths of effort provision, task allocation, and knowledge transfer for a cash-constrained apprentice, who is free to walk away at any time. In the optimal contract the apprentice pays for training by working for low or no wages and by working inefficiently hard. The apprentice can work on both knowledge-complementary and knowledge-independent tasks. We study the optimal time path of effort distortions and their impact on the knowledge transfer, and analyze the effect of regulatory limits on the length of apprenticeships and on how much effort apprentices are allowed to provide.

We propose that a simple "dual-self" model gives a unified explanation for several empirical regularities, including the apparent time inconsistency that has motivated models of quasi-hyperbolic discounting and Rabin’s paradox of risk aversion in the large and small. The model also implies that self-control costs imply excess delay, as in the O'Donoghue and Rabin models of quasi-hyperbolic utility, and it explains experimental evidence that increased cognitive load makes temptations harder to resist. The base version of our model is consistent with the Gul-Pesendorfer axioms, but we argue that these axioms must be relaxed to account for the effect of cognitive load.

We argue that some, but not all, superstitions can persist when learning is rational and players are patient, and illustrate our argument with an example inspired by the Code of Hammurabi. The code specified an “appeal by surviving in the river” as a way of deciding whether an accusation was true. According to our theory, a mechanism that uses superstitions two or more steps off the equilibrium path, such as “appeal by surviving in the river,” is more likely to persist than a superstition where the false beliefs are only one step off the equilibrium path.