Modelling Taylor Rule Uncertainty

In recent years, one has witnessed a widespread attention on the way monetary policy is conducted and in particular on the role of the so-called monetary policy rules. The conventional approach in the literature consists in estimating reaction functions for a monetary authority (the Federal Reserve, in most cases) in which a nominal interest rate, directly or indirectly controlled by that monetary authority, is adjusted in response to deviations of inflation (current or expected) from target and of output from potential. These reaction functions, usually called Taylor rules, following John Taylor's seminal paper published in 1993, match a number of normative principles set forth in the literature for optimal monetary policy. This provides a good reason for the growing prominence of indications given by Taylor rule estimations in debates about current and prospective monetary policy stance. However, they are usually presented as point estimates for the interest rate, giving a sense of accuracy that can be misleading. Typically, no emphasis is placed on the risks of those estimates and, at least to a certain extent, the reader is encouraged to concentrate on an apparently precise central projection, ignoring the wide degree of uncertainty and operational difficulties surrounding the estimates. As in any forecasting exercise, there is uncertainty regarding both the estimated parameters and the way the explanatory variables evolve during the forecasting horizon. Our work presents a methodology to estimate a probability density function for the interest rate resulting from the application of a Taylor rule (the Taylor interest rate) which acknowledges that not only the explanatory variables but also the parameters of the rule are random variables.