INTUITIVE AND RELIABLE ESTIMATES OF OUTPUT GAP AND REAL EXCHANGE RATE CYCLES FOR TURKEY

Abstract

Extracting cyclical components from time series data is among the priorities for policy maker institutions. Comparing with the unrestricted Beveridge-Nelson decomposition and Hodrick-Prescott filter, we implement a restricted Beveridge-Nelson filter developed by Kamber et. al. (2018) which limits the volatility of trend component. Utilizing the quarterly real GDP series, we find that restricted Beveridge-Nelson filter provides more persistent and larger cyclical values than Beveridge-Nelson decomposition. Taking the output gap estimates of Central Bank of Turkey as a benchmark, our results indicate that Beveridge-Nelson filter yields more sensible results. We also develop a measure to make an assessment on the end-point bias. Our results show that restricted Beveridge-Nelson filter performs better than Hodrick-Prescott filter regarding the magnitude of end point bias. When we compare the alternative methods by using the monthly real exchange rate series for Turkey, our findings are consistent with the results obtained with GDP data. Restricted Beveridge-Nelson filter produces persistent and large real exchange rate cycles than Beveridge-Nelson decomposition estimates. We find that restricted Beveridge-Nelson filter dominates Hodrick-Prescott filter in terms of the magnitude of end point bias.