Ambiguous State Dynamics, Learning, and Endogenous Long-Run Risk

Abstract

This paper considers learning about unobservable state variables when their dynamics are ambiguous. Ambiguity in dynamics, differently from that in a parameter, is never fully resolved; and, since data are to be filtered through the state equation, learning about the state becomes more difficult, permanently. This endogenously amplifies the long-run risk in estimation. An application to the long-run risks model shows that some of the large long-run risk (and high risk aversion) required by returns data can be attributed to this lack of confidence.