TEMPORARY EQ'UILIBRIUM DYNAMICS WITH BAYESIAN LEARNING

This paper examines the stability of deterministic steady-states in a class of economies where the state -variable is one dimensional and where agents use Bayesian techniques to form expectations. Thr dynamics with learning are locally convergent if the prior mean is close to a stable perfect foresight root having modulus less than 1 and if the prior beliefs are held with enough confidence. Thr dynamics are however divergent if the prior mean or the variance of the prior distribution is sufficiently large.