Modelling and Forecasting Exchange Rate in Ghana: An Application of the Exponential Garch and Bayesian Vector Autoregression Models

Abstract

Ghana had faced a macroeconomic problem of exchange rate for a long period of time and this problem had somehow slowed the economic growth. The exchange rate has been one of the major economic challenges facing most countries in the world especially African countries including Ghana. Therefore, forecasting and modelling the nature of the exchange rate between the Cedi and the US dollar in Ghana is very important in order for the government to design economic strategies and effective monetary policies to combat any unexpected hike in the exchange rate. This research studied the variances in the exchange rate data in Ghana and found statistical models to represent and forecast the variance associated with it. Using monthly data on the exchange rates from January, 2008 to March, 2022, the Bayesian Vector Autoregression (BVAR) model gave good forecast of the rates with a BIC of-162.491 and an RMSE of 0.017. Based on the Bayesian VAR model, we also forecasted the rates outside the sample period (January, 2019 to December, 2021). The forecasted results showed consistency in the actual data. The study concluded that multivariate models such as the Bayesian VAR was good in forecasting the rates and was also able to assess the impact of other variables on the exchange rate data. The study recommends the use of the Bayesian VAR model in analyzing macroeconomic data. Also, the use of a complex model such as the GJR-GARCH could go a step further in dealing with the variances within the residuals of the exchange rate data, while assuming apriordistributionthatfits thedata for theparametersof theBayesianVAR could give better forecasts of the exchange rate.