Fundamental Uncertainty as Model Uncertainty

Economic agents must form models of their environments in order to develop expectations and make decisions, yet these models are certain to be misspecified. An agent aware of their own inability to perfectly capture the structural relationships of their observed world will entertain model uncertainty. Under the plausible assumptions that the “true” model is not known to the decision-maker and the decision-maker knows this—known as the M-open case in Bayesian statistics—uncertainty over propositions becomes numerically irreducible. The notion of model uncertainty is developed with reference to Post Keynesian theories of fundamental uncertainty as well as relevant areas of study within decision theory, including the growing literature on unawareness. The model uncertainty view poses challenges for both literatures and provides a novel justification for the types of uncertainty associated with Knight and Keynes.