Forecasting the Direction of Policy Rate Changes: The Importance of ECB Words

This paper evaluates the predictive power of different information sets for the European Central Bank (ECB) interest‐rate‐setting behaviour. We employ an ordered probit model, i.e. a limited dependent variable framework, to take into account the discreteness displayed by policy rate changes. The results show that the forecasting ability of standard Taylor‐type variables, such as inflation and output gap, is fairly low both in‐sample and out‐of‐sample, and is comparable to the performance of the random walk model. Instead by using broader information sets that include measures of core inflation, exchange rates, monetary aggregates and financial conditions, the accuracy of the forecasts about ECB future actions substantially improves. Moreover, ECB rhetoric considerably contributes to a better understanding of its policy reaction function. Finally, we find that that the ECB has been fairly successful in educating the public to anticipate the overall future direction of its monetary policy, but has been less successful in signalling the exact timing of rate changes.