Convergence rate of agents' learning in macroeconomic models

This paper discusses agents7 learning on a market. The price level evolves through a multivariable autoregressive model, which the agents learn in a least-squares sense. A theorem is stated that shows how the agents7 learning might be divided into two classes with respect to the learning convergence rate. The results are exemplified by the well-known hyperinflation model. Further, for the hyperinflation model some interesting features concerning the "coupling" between the price and the learning dynamics are discussed. An explicit expression is derived for how the rate of the agents' learning depends upon this coupling