Back to the future : economic rationality and maximum entropy prediction

An information-theoretic thought experiment is developed to clarify why the maximum entropy methodology is appropriate for predicting the equilibrium state of economic systems. As a first step, object allocation problems, modeled as knapsack problems, are shown to be equivalent to congestion games under weak assumptions. This proves the existence of finite improvement paths linking initial conditions and Nash equilibria. The existence of these improvement paths is precisely what enables the use of maximum entropy to make predictions concerning the equilibrium state. Finally an illustration of this predictive power is provided through an application to the Schelling model of segregation. -- Information entropy ; knapsack problem ; potential function ; Schelling segregation