Random walk or arima? exploring the power of predictability of rupee-dollar exchange rate

Abstract

This paper examines the randomness and predictability of the rupee-dollar exchange rate. We developed a working hypothesis to compare the out-of-sample Jorecasting abilities of the ARIMA model to those of a pure random walk (RW). The results indicate that the rupee-dollar exchange rate not only exhibits a trend component but also a statistically significant one. Hypothesis testing reveals that the ARIMA model outperforms the pure RW model as a predictor. The error measures, using RMSE and MAE, demonstrate that the ARIMA model possesses more than twice the predictive power compared 1o the pure random walk. Therefore, it can be concluded that the rupee<dollar exchange rate is anything but a random walk.