Recent research suggests that commonly estimated dynamic Taylor rules augmented with a lagged interest rate imply too much predictability of interest rate changes compared with yield curve evidence. We show that this is not sufficient proof against the Taylor rule: the result could be driven by other equations of the model that the Taylor rule is embedded in. To disentangle the effects, we study the predictability of all variables in a simple model of monetary policy: inflation, the output gap, and the interest rate, and we compare with evidence from survey data and a VAR model. We find that the strongest evidence against the Taylor rule is that while it is easy to predict the variables that enter the rule, it is very hard to predict actual interest rate changes. This is consistent with usual Taylor-type rules if policy shocks are very large, but it is more likely that there are other aspects of monetary policy behaviour that are neglected by the Taylor rule.
Interest rate smoothing yield curve survey data VAR in-sample overfitting