The estimation of central bank interest decisions as a dichotomous variable

Economic stability during the last 15 years drastically reduced changes in interest rates determined by central banks. The small fluctuations in interest rates reduced the possibility of using quantitative econometric models to estimate the reaction function which central banks use when determining the interest rate. We regard interest decisions as a dichotomous variable and use a Multinomial Logit model to estimate the probability that a central bank would decrease, increase or leave interest the rate unchanged. The model was used to estimate the American Fed's interest decisions. The main results are that real economic activity, as measured by product gap, has a strong effect over the Fed's interest decisions, while inflation deviating from its target does not affect the Fed's decisions. We find that the probability of raising the interest rate increases at a larger magnitude than the probability of lowering interest rates following a relevant symmetric change of a feeding variable.