Exchange rate volatility is an important aspect of risk management. Volatility in the exchange rate environment can be accessed using a number of models. Each exchange volatility model takes into account a specific set of variables. Some of the most common methods of accessing volatility include the GARCH and the ARCH models. In an article by Rapach and Straus, the authors explore and test the GARCH(1,1) model of accessing exchange rate volatility (Rapach & Strauss 2008). The authors use the exchange rate of the dollar in the period beginning from 1980 to 2005. Nevertheless, some volatility models provide better results than others. This paper utilizes the theory and research from Rapach and Straus’ article to argue that among the GARCH models, GARCH(1,1) model provides the best forecast results.
Most economists are under the impression that a stable GARCH process addresses the instability of the exchange rate process by setting the variance levels at a constant level. Consequently, the variants that have low levels of accuracy have never been cataloged comprehensively and this explains why all the GARCH models have coexisted. Each variance of the GARCH model targets specific characteristics that are deemed to be most relevant to exchange rate volatility. Initially, the data sets used in Rapach and Straus’ article indicated that there might be a better model than the GARCH(1,1) however the results of both in-sample and out-of-sample tests indicated that the model performs optimally. Other models such as the IGARCH(1,1) indicate weaknesses in other pertinent outliers such as the ‘mean absolute deviation criterion’.
Financial markets are subject to internal and external factors that bring upsets to exchange rate mechanisms. These shocks can have detrimental effects on the unconditional variances that dominate the exchange rate markets. The structural breaks in the process of modulating volatility assessment model have been addressed by several aspects of the GARCH models. In this category, most models exhibit constant variants through their shape parameters. For example, most of the earliest models of exchange volatility worked under the assumption that the assumed density function was standard. However, the GARCH(1,1) model corrects this anomaly by doing away with restrictive assumptions and using an estimated time variance density for the consistent residuals via a series of semi-nonparametric methods (Rapach and Strauss 2008).
The article by Rapach and Straus suggests two methods of addressing the issue of structural breaks within the GARCH model. Their first method includes the investigation of the ‘out-of-sample’ volatility forecasting exercises to ensure that mistaken assumptions are not included in the process. For instance, the GARCH model uses an expanding or a fixed data window when accounting for the structural breaks that occur in the course of assessing exchange rate volatility. The structural breaks that feature within the GARCH (1,1) model of measuring volatility in the exchange rate market are structured around the consideration that ‘both in-sample and out-of-sample tests’ are dependent processes (Rapach & Strauss 2008). The variants that feature in the GARCH model can be compared with the Iterated Cumulative Sum of Squares (ICSS) but both methods are subject to unconditional variance in exchange rate. Nevertheless, the GARCH(1,1) model provides a unique mode of accounting for the conditional variance. For instance, the GARCH(1,1) model can produce a method of auto correcting exchange rate financial (Rapach & Strauss 2008). The superiority of the GARCH(1,1) model is also underlined by the fact that it utilizes a ‘rolling window model’ in its estimation parameters.
Reference
Rapach, D. E., & Strauss, J 2008, “Structural breaks and GARCH models of exchange rate volatility”, Journal of Applied Econometrics, vol. 23, no.1, pp 65-90.