Analyzing Tunisia's exchange rate fluctuations: Fuzzy filter approach

The standard approach analyzing the business cycle (RBC) is based on Hodrick-Prescott (HP) and Band- Pass (BP) filters to extract the cyclical component of the aggregate time series. By Monte Carlo experiments using Matlab program, we compared these traditional filters to a more recent Fuzzy Regression Filter. Our aim is to help the user to choose the appropriate filter in the appropriate time series. Through our application on the exchange rate (Dinar/U.S.$), the results show an improvement of the Fuzzy filter when the time series is non-stationary. In fact, through only the Fuzzy filter, (i) there no imposed restrictions on time series distribution; and (ii) the use of traditional filters lead to an evaluation bias of the exchange rate cyclical components. Additionally, when the implemented exchange policy is based on HP and/or BP filter, it then could be misled. Would it be necessary in a further research to review the stylized facts established by the standard RBC model by applying the Fuzzy filter as an alternative?