Forecasting Foreign Exchange Rates in Kenya Using Time Series: A Case of Usd/Kes Exchange Rates

Abstract

In October 1993, Kenya adopted a floating exchange rate system where the exchange rates are determined by forces of demand and supply for the local currency. Exchange rate forecasts are necessary to evaluate the foreign denominated cash flows involved in international transactions. Therefore exchange rate forecasting is important to evaluate the benefits and risks attached to the international business environment. This study therefore sought to fit a Seasonal Autoregressive Intergrated Moving Average Model(SARIMA)(p, d, q)(P,D,Q)[12] to United States Dollar vs Kenya Shilling exchange rate since it is the most dominant exchange rate in Kenya. The secondary monthly data from January 1993 to March 2019 from Central Bank of Kenya official website was divided into two parts namely the in-sample data and the out-sample data. The in-sample data was used to fit the model while the out-sample was used to validate the model. Seasonal Mann-Kendall test established that there was seasonal trend. A first regular difference was used to stationarize the series since the ADF test established it was not stationary. Autoregressive Intergrated Moving Average (ARIMA) and Seasonal Autoregressive Intergrated Moving Average (SARIMA) models were fitted in the data. ARIMA(1, 1, 0) and SARIMA(1, 1, 0)(0, 0, 2)[12] were found to be the best models on the basis of Bayesian Information Criterion(BIC) and Akaike’s Information Criterion(AIC). In the short run i.e 3 months, the Seasonal Autoregressive Intergrated Moving Average had the least Mean Absolute Error(MAE), Mean Absolute Percentage Error(MAPE) and Root Mean Squared Error(RMSE) values of 0.1651, 0.1636 and 0.2037 respectively. This study therefore recommends the integration of the Seasonal Autoregressive Intergrated Moving Average Model in forecasting United States Dollar vs Kenya Shilling exchange rate in Kenya in the short run. v