Reaction Function Estimation when Central Banks Face Adjustment Costs

The main instrument of monetary policy in industrialized countries is currently a short-term interest rate. It typically remains unchanged during long spans of time. This paper tries to answer three questions. Why do Central Banks change targeted interest rates so seldom? How should we estimate Central banks' reaction functions? And what are the driving forces behind rate changes? This paper takes the point of view that Central Banks face a fixed cost when adjusting the targeted interest rate and therefore smoothe the targeted interest rate by using a discrete policy rule. In the estimation of the reaction function this discrete nature is taken into account by applying a grouped data model to a Swedish data set. It is found that the reaction function is best represented in terms of changes in growth rates of macro variables and changes in levels of financial variables. Probabilities of the target rate being raised, lowered or kept constant are computed and compared with actual interest rate behavior. The model has a prediction rate of 88% versus 78% for the best naive estimator.