Estimating Interest Rate Setting Behavior in Brazil: A LSTR Model Approach

摘要

Given limited research on monetary policy rules in emerging markets, this paper challenges the applicability of a nonlinear Taylor rule in characterizing the monetary policy behavior of the Brazilian Central Bank. It also investigates whether and how the process of setting interest rates has been developed in response to contingencies and special events. We extend the linear Taylor rule to a regime-switching framework, where the transition from one regime to another occurs in a smooth way, using a Logistic Smooth Transition Regression (LSTR) approach. In this sense, we empirically analyze the movement of the nominal short term interest rate of the Brazilian Central Bank using quarterly data, covering the period 1994.Q4–2012.Q2. We find that the nonlinear Taylor rule provides a better description of the Brazilian interest rate setting and is consistent with historical macroeconomic events. In particular, our results show that adopting a nonlinear specification, instead of the linear, leads to a costs reduction in terms of fit: 190 basis points in 1995 and 140 basis points in the mid-2002 presidential election campaign in Brazil. Moreover, the Brazilian monetary policy exhibits nonlinear patterns that better captures special events and may contain relevant information rendering it applicable to unusual conditions, i.e., a financial crisis, which require disconnection from the automatic pilot rule and use of judgement to make decision.