ABSTRACT

Volatility is integral characteristic of financial markets all over the world. Exchange rates of Pakistan have been highly volatile since last two quarters. The volatility of exchange market is more sensitive to bad news but resilient to good news, indicating that investors are more likely to be impacted by bad news than by good news. In this study, the data of Nominal exchange rate returns have been used to conduct analysis on the basis of GARCH-type models. To examine model effects under various distributions and orders for the sample series, we developed the autoregressive moving average (ARMA)-Generalized Autoregressive Conditional Heteroscedasticity model. We also choose a threshold-GARCH (T-GARCH) model, in contrast, to reflect the asymmetry characteristics of the financial-returns series. Furthermore, Markov switching GARCH model has been used to model volatility while being in two different states. Additionally, to analyze the error level and prediction results of various models, mean squared error (MSE), mean absolute error (MAE), and root-mean-square error (RMSE) are used. The results prove that the MS-GARCH model under Generalized-Error-Distribution outperforms other proposed models when predicting the Exchange rate returns series. According to our findings, switching models tend to hold good predictive powers in short term only. They may fail in long term i.e., more than a year.