Learning About Ambiguous Long-Term Prospects

Abstract

This paper investigates whether and when ambiguity afflicting the long-term prospects of a market fades away in a nonexchangeable environment (time-varying short-term prospects). Two types of ambiguity are considered: static (multiple priors) and dynamic (multiple laws of motion). In the absence of dynamic ambiguity, likelihood-based learning resolves the static ambiguity. In the presence of dynamic ambiguity, on the other hand, likelihood-based learning fails. In this case, the static ambiguity fades away if the agent incorporates into the objective criteria (likelihood) her subjective criteria (penalty proportional to the Kullback-Leibler divergence).