Evaluation of interest rate rules
Since we set the monetary policy instrument RSt in order to make a target variable xt stay close to its target level x*, it makes sense to evaluate the rules according to how well they achieve their objective. In the theoretical literature,
£(xt — x*)2 = J V [x] + (x — x* )2,
however, policy evaluation is often based on the unconditional variance of xt, denoted V [x]. An alternative measure which puts an equally large weight on the bias of the outcome, that is, on how close the expected value of xt is to the target x*, is the RMSTE. Since the bias could differ considerably between different monetary policy rules, it is of interest to investigate its effect in small samples. If we estimate the expected level E[x] by its sample mean x, the measure can be written as
which is the form we will adopt in the following sections.