Global stability in spite of ''local instability'' with learning

We analyse learning dynamics in deterministic economies when agents generat e their forecasts by performing least squares regressions on lagged values of the endogenous state variable. We establish conditions for the global co nvergence to the deterministic steady state of the dynamics under learning. Our global stability result is particularly relevant since the learning ru les we consider are already known to produce locally divergent behaviour ar ound the steady state. (C) 2000 Elsevier Science B.V. All rights reserved.