Forecasting time series using Vector Autoregressive Model

Abstract

In this study, a vector Autoregressive model was used to analysis the relationship between two time series as well as forecasting. Two financial time series have been used, which are a series of global monthly oil price and global monthly gold price in dollars for a period from January 2015 to Jun 2019. It has 54 monthly values, where the data has been transferred to get the Stationarity, Diekey Fuller test for the Stationarity was conducted. The best three order for model was determined through a standard Akaike information AIC, it is VAR(7) , VAR(8) and VAR(10) respectively. The comparison was made between selected orders by AIC based on the accuracy measure and mean square error (MSE). It turns out that less MSE value of the VAR(10) model. Some tests were conducted like Lagrange-multiplier, Portmanteau, Jarque - Bera to residuals for the selected model, with forecasting for the VAR(10) model for the period from Jun 2019 to Jun 2021 , It is 24 monthly value. It turns out that less MSE for forecasting value for oil price series is to VAR(7) model and less MSE for forecasting value for gold price series is VAR(10) model. The results have been computed through the Stata program.