Escaping Nash inflation

Mean dynamics govern convergence to rational expectations equilibria of self-referential systems under least squares learning. We highlight escape dynamics that propel away from a rational expectations equilibrium under fixed-gain recursive learning schemes. These learning schemes discount past observations. In a model with a unique self-confirming equilibrium, we show that the destination of the escape dynamics is an outcome associated with government discovery of too strong a version of the natural rate hypothesis. That destination is not sustainable as a self-confirming equilibrium but is visited recurrently. The escape route dynamics cause recurrent outcomes close to the Ramsey (commitment) inflation rate in a model with an adaptive government.