Forecasting long memory time series when occasional break occur

Recent research has found that processes with occasional structural breaks could be hardly distinguished from long memory ones because the memory properties may spuriously arise in model with level shifts. In this paper, we investigate if a long memory model will provide good forecasts even if the real DGP is affected by level shifts (as suggested by Diebold and Inoue, 2001). We compare via simulations the forecasting performance of long memory and occasional breaks processes