Sensitivity Analysis of Change of Currency Exchange ARIMA Modeling Parameters between (0,1,1) and (1,1,0) Depending on Economic Policies

This thesis contains a literature survey of exchange rates, and forecasting exchange rates by using ARIMA models. The thesis shows that the best forecasting ARIMA parameters are changing in time, looking like the characteristics of the time series is changing randomly. A search of forecasting parameters within AR and MA orders 1 to 5 indicated that (1,1,1) is successful in forecasting the five-day ahead values. This study tested the sensitivity of the change of best-forecasting model parameters between AR (p=1, d=1, q=0) and MA (p=0, d=1, q=1) characters by using the financial policy decision dates as a predictor. The test is evaluated on GBP=X time series using the Federal Bank Federal Fund Rate decision dates. The test results indicate that the statistical value of sensitivity for 26 FFR decision dates is almost 5% shifted on the decision dates, indicating that the FFR decisions had a considerable structural effect on the dynamics of the market.