Vol. 6, Issue 6, Part E (2020)

Abstract
This study considers the generalized autoregressive conditional heteroskedastic approach in modeling exchange rate volatility of the five major currencies of Sri Lanka using daily observations over the period of 7th of May 2010 to 31st December 2019. The currencies considered are United States Dollar (USD), Euro (EUR), British Pound (GBP), Japanese Yen (JPY) and Indian Rupees (INR), all against Sri Lanka Rupee (LKR). The study applied symmetric ARCH (q) and GARCH (p, q) models that estimate exchange rate volatility with the normal distribution. The result of the study shows that the three currencies fulfill the conditions of volatility models and these currencies modeled by GARCH. ARIMA (2, 1, 2,) -GARCH (1, 1) specification is proven to be the best model to estimate GBP exchange rate volatility. ARMA (1, 1) –GARCH (1, 1) is more appreciate for JPY and USD exchange rate volatility and it is required to have a mean reverting variance process for JPY exchange rate. ARMA (2, 2) -GARCH (2, 1) is a best fit model for INR exchange rate volatility Finally the study concluded that the exchange rate volatility can be adequately modeled by GARCH model.