Recursive equilibria in dynamic economies with bounded rationality

This paper provides a general framework to model bounded rationality in dynamic stochastic general equilibrium models with infinitely lived heterogeneous agents. A boundedly rational agent is associated with an information set $I$ and an extra parameter $\epsilon$, which can be interpreted as the ``level of irrationality''. To make decisions, the boundedly rational agent forms a belief of a stationary joint distribution of the exogenous and endogenous variables and uses the marginal distribution (conditional on $I$) to form forecasts. If the equilibrium distribution stays within $\epsilon$ of the forecasted next-period distribution, the agent would consider it as $\epsilon$-stationary. In equilibrium, each agent maximizes utility with an $\epsilon$-stationary belief and markets clear. The main theorem of this paper shows that for any strictly positive $\epsilon$, a recursive equilibrium exists. With a quantifiable ``level of irrationality'', the model incorporates many behavioral economics models as well as rational-expectations models with computational approximations into a unified framework. An illustration example is presented to show that the boundedly rational recursive equilibrium may substantially enrich the asset pricing dynamics of the rational-expectations model even for small $\epsilon$