Unravelling the Cipher of Indian Rupee’s Volatility: Testing the Forecasting Efficacy of the Rolling Symmetric and Asymmetric GARCH Models

Abstract

Modelling exchange rate volatility is crucially important because of its diverse implications on the profitability of corporations and decisions of policy makers. This paper empirically investigates exchange rate volatility of India's currency by applying rolling symmetric and asymmetric GARCH models to the USDINR and EURINR daily exchange rates for a period spanning April 1, 2006 through January 31, 2018, resulting in total observations of 2861. To estimate GARCH (1,1) and EGARCH (1,1) models, the data window is rolled over five years with nearly 1200 observations and one month is used as forecast period for each window. Both, in-sample criteria like the log likelihood criteria, Akaike information criterion (AIC), the Bayesian information criterion (SIC) and Hannan Quinn criterion (HQC) as well as the out-of-sample criteria like Mean Squared Error (MSE) and Mean Absolute Error (MAE) have been used to test model fit and forecast accuracy of the models. To test the robustness of the findings, Diebold-Mariano test is used to compare the predictive accuracy of both the models. Further, the forecasting accuracy of the two models has also been tested by splitting the sample period into periods of tranquility and volatility in Indian exchange rate. Results show that GARCH (1,1) model with generalized error distribution is adequate to capture the mean and volatility process of USDINR and EURINR exchange rate returns.