Frequently, real economic agents do not follow purely rational strategies.  These individual non-rational behaviors (due to errors in judgment, incomplete information, emotional bias, etc.) can result in some fascinating organized large-scale structures, which depend on the degree of non-rational behavior.

We look at two such models for Potential Games [Shapley and Monderer]: a dynamical drift-diffusion model, and a static large deviation theory model based on Shannon information entropy and arbitrage.  The equilibrium measure in both cases is the Gibbs measure found in statistical mechanics.  We show that the variables that gauge non-rational behavior in both models are related to “temperature” by a fluctuation-dissipation relation.

A type of localized discrete Cournot oligopoly has a rich phase diagram with an "antiferromagnetic" checkerboard state, striped states and maze-like states with varying widths, and finally a "paramagnetic" unordered state. Such phases have economic implications as to how agents compete given various restrictions on how goods are distributed. 

The theory is also applied to a Speculative and Hedging Model in Oil and U.S. Dollar Markets [Carfi and Musolino] for a single multinational “airline” and many “bank” players.  Based on results for the Nash equilibrium (zero temperature) and preliminary results, there is a phase transition for which a single equilibrium exists at higher non-rational behavior (high temperature), and two equilibria at lower non-rational behavior (low temperature), when the “airline” makes no purchase of oil.  The low temperature phase is in the spirit of the Sonnenschein–Mantel–Debreu theorem, with the extra insight of symmetry-breaking to explain multiple equilibria.  Likewise, Huw Dixon’s result on the “inevitability of collusion” is shown to hold for a Cournot Oligopoly with a Veblen good.  Purely rational neoclassical theory (i.e., Nash equilibrium analysis) alone does not predict this, even though it is observed to occur in more general cases.