How Well the Ringgit-Yen Rate Fits the Non-linear Smooth Transition Autoregressive and Linear Autoregressive Models

This study compares the forecasting performance between Smooth Transition Autoregressive (STAR) non-linear model and the conventional linear Autoregressive (AR) time series model using the simple random walk (SRW) model as the standard reference model. To accomplish this objective, quarterly frequency exchange rate data, which is well known for its non-linear adjustment towards purchasing power parity equilibrium path is employed. The empirical results suggest that both the STAR and AR models exceed or match the performance of SRW model based mean absolute forecast error (MAFE) mean absolute percentage forecast error (MAPFE) and mean square forecast error (RMSFE). The results also show that the STAR model outperform the AR model, its linear competitor. This is consistent with the emerging line of research that emphasised the importance of allowing non-linearity in the adjustment of exchange rate toward its long run equilibrium.