Predicting Changes in the Interest Rate: The Performance of Taylor Rules Versus Alternatives for the United Kingdom

We consider an experiment where we use the Taylor rule information set, inflation and the output gap, to predict the next change in monetary policy for the United Kingdom 1992 - 2000. To do this we use a limited dependent variable approach, where the next rate change could be `upwards', `downwards' or `no change'. A Multinomial Logit model is used to predict the next most likely change using monthly data, and these predictions are compared to the actual outturn. Against this hypothesis we compare a wider information set including more than just inflation and output gap variables. The in-sample and out-of-sample prediction tests are evaluated using forecast performance tests.