Forecasting foreign exchange rates in Kenya using stochastic models

Abstract

The shift of exchange rate regime from fixed to floating has prompted the need for forecasting exchange rates in Kenya. Portfolio managers and corporate finance managers whose forms' and clients' cash flows are affected by exchange rate movements would benefit from an estimate as this would provide a basis for decision making. The paper examines the accuracy of the two stochastic models, Geometric Brownian motion and Merton-diffusion jump using an error statistic; Mean Absolute Percentage Error (MAPE) of the USD/KES forecasts under consideration. A lower value of the MAPE indicated the inclusion of the Merton jump process in the Geometric Brownian motion was superior as a smaller error was obtained.