Linear Feedback Rules in Non-Linear Models with Rational Expectations

The adoption of inflation targets by a number of industrialised countries in the last decade has reawakened interest in the study of rules to characterise monetary policy. In the literature a clear distinction is drawn between instrument rules, such as that of Taylor, which are backward looking, and targeting rules, which are specifically forward looking. In this paper, we show that a properly specified control rule derived via dynamic programming encompasses both types of rule. So our approach addresses in a straightforward way the concerns of Svensson and others that monetary policy needs a forward-looking dimension. Moreover, we describe a computationally simple method to derive the control rule under rational expectations as well as a linearisation method when the model is non-linear.