Measuring forecasting performance of vector autoregressive and time series regression models

Abstract

Correlation and Regression are the traditional approach of determining relationship between two or more variables. When the variables are multiple and the dependent variable is considered having an explanatory variable, then a Vector Autoregressive model is used to determine the structural relationship between the variables. If these variables are co-integrated, VAR model is not appropriate, but our focus is on the structural relationship and measuring forecast performance of a VAR and Time series regression with Lagged Explanatory Variables. Some Nigerian economic series (Government Revenue and Expenditure, Inflation Rates and Investment) data were analysed and the Root mean Square forecast Error (RMSFE) and Mean Absolute Percentage Forecast Error (MAPFE) are used as measurement criteria. The VAR model was found to be better than Time series regression with Lagged Explanatory Variables model as indicated by Meta diagnostic tools. The forecast values from the VAR model is more realistic and closely reflect the current economic reality in Nigeria indicated by the forecast evaluation tools.