Predicting Inflation in Ghana: A Comparison of Cointegration and Arima Models

Abstract

This study is a master thesis for a master’s program in Economics and Finance in the Department of Economics and Political Sciences at the University of Skovde. As the title indicates, the aim of the thesis is to use ARIMA and VAR models to predict inflation in Ghana. In order to fulfil this objective, economic variables such as the consumer price index, money supply, interest rate and exchange rate have been used to build the aforementioned models. In building the ARIMA model, the Box-Jenkins approach has been used and for the cointegration model, the Johansen’s approach has been utilised. The conclusion drawn from the study is that, the cointegrating VAR model is suspected to have lower Root Mean Squared Error (RMSE) than the ARIMA model and as such may be more efficient than the ARIMA model in forecasting inflation. Therefore, the VAR model is efficient in predicting which variables are likely to trigger off an inflationary process and the magnitude of the pressure that these variables exert on inflation. Thus it was found out that money supply, interest rate and exchange rate policies predict inflation in Ghana as per our VAR model. The models therefore predicted inflation for end of year 2003 as 22.65% and 21.66% for ARIMA and VAR models respectively. In view of this stringent contractionary monetary policy is recommended. Also measures should be adopted to increase money demand through increases in GDP. Export earnings should also be increased to bring down the exchange rate. These policy measures have the tendency to bring down inflation to a very desirably low level.