Market selection and learning under model misspecification

This paper studies market selection in an Arrow-Debreu economy with complete markets where agents learn over misspecified models. Under model misspecification, standard Bayesian learning loses its formal justification and biased learning processes may provide a selection advantage. However, considering two cases of model misspecification and four learning processes, our analysis reveals a general difficulty in ranking learning behaviors with respect to their long-run performances and, hence, their survival in the market. For instance, prediction averaging stops being an advantageous strategy when the truth does not belongs to the same family of models agents use to learn. In general, learning rules that generically guarantee survival require an unreasonable amount of knowledge about the whole market ecology. Thus, the goal of a parsimonious long-run asset valuation model robust to model misspecification appears out of reach.