Comparing Predictive Power of Exchange Rate Models in Emerging Asian Economies

In this paper, we have compared forecasting performance of three economic and two autoregressive models of exchange rate in five Asian economies; namely Pakistan, India, Indonesia, Korea and Sri Lanka. Models include purchasing power parity (PPP), interest rate parity (IRP), adhoc model, random walk model (RWM) and autoregressive integrated moving average (ARIMA) model. To compare the predictive capacity of these models, four statistical measures, namely Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Median of Absolute Deviation (MAD) and Success Ratio (SR) have been used. Our results document that models based on economic fundamentals do not outperform random walk and ARIMA in all the sample economies. If the effect of outliers is controlled, then adhoc model defeat single variable based parities i.e interest rate pairty and purchasing power parity. economic models have better predictive capacity except in case of Indonesian Rupiah, where random walk model defeat economic models even after controlling the effect of outliers. RMSE, MAE and MAD favor economic models. However, SR reports different results in different economies. In Pakistan, India and Korea, it gives vote to IRP while in Indonesia and Sri Lanka, it favors random walk model