This article estimates the monetary policy rule followed by the Brazilian Central Bank for setting its main policy instrument, the SELIC rate, for the period after the Real Plan. In order to overcome the uncertainty over the dates at which changes in parameters occurred, this paper uses regime-dependent-switching probabilities according to a hidden Markov chain to model possible deviations from a simple linear reaction function. From July 1996 to January 2006 the Brazilian monetary policy can be fully characterized by four policy regimes. The changes in monetary policy in this period are best described by recurring regime changes, instead of once-and-for-all shifts. We have identified substantial differences in the way monetary policy was conducted in the subperiods before and after 1999, when the Brazilian exchange rate policy regime changed from crawling peg to free-floating. At each of these subperiods there are two recurring regimes and the two regimes of one subperiod differ from the two regimes of the other.
