Perfect predictions in economic dynamical systems with random perturbations

The paper studies multivariate non linear economic dynamical systems with an expectations feedback subjected to exogenous perturbations. In these systems, agents form expectations on future variables based on subjective transition probabilities given by a Markov kernel. The notion of a perfect Markov kernel that generates rational expectations along all orbits of the system is proposed. Conditions are provided under which perfect Markov kernels exist. Applications are given to models of the Cobweb type, to multivariate affine-linear systems, and to the stochastic OLG model of economic growth. For the latter two models, it is shown when a globally attracting random fixed point with rational expectations exists.