Applied Forecasting on the Forex Market: A Macroeconomic and Cointegration Approach

Abstract

The thesis specifies and studies different models intended to forecast the exchange rates for the currency pairs, CADUSD, EURUSD, and GBPUSD. The results show that it is possible to forecast exchange rates to some extent when applying a model that only relies on lagged values of itself and relevant macroeconomic variables. The forecasts are only rather consistent for two of the three variations of this model when the forecasts return a positive growth rate. When the expected growth rate is negative the model is less precise and applying the model as an investment strategy will provide much less consistent results. The other two models are based on cointegration and despite showing definite cointegration relationships between the variables, the forecasts are very imprecise. The results for all model variations are delivered through graphical depictions of level data indicating the exchange rate dynamics, first differences of the data to achieve the growth rates of the exchange rates, and lastly through computing the Mean Absolute Error and Root Mean Square Error. By comparing the results for each model variation the models can be compared against each other to find the most precise model. The results provide a renewed insight into the literature on cointegration between currency pairs and present the effectiveness of applying cointegration to forecast exchange rates. Lastly, relevant future research on the area is proposed thereby allowing the subject to expand even further.