The objective of this work is to estimate, according to the methodology of Taylor (1993), the reaction function of the Bank of the Republic of Burundi. We first test the stationarity of the variables. The methodology used is inspired by Clarida, Galí and Gertler (1998), Mésonnier and Renne (2004) and De Lucia and Lucas (2007). In a second step, we estimate the reaction function of the BRB by the generalized method of moments. We use the Hodrick-Prescott filtering method as a mode of calculating the output gap used in the regression. The results of the stationarity test prove that all variables are stationary in first difference. The results obtained from the annual data (1980 to 2015), show that the estimated rule deviates from the Taylor rule. In addition, the exchange rate was not taken into account by the BRB in the implementation of its monetary policy.