Comparison between Causal Model and Time Series Model to Forecast Gold Prices

Abstract

Gold is the most precious and valuable commodity in the world. It is not only used to make jeweleries but gold also plays an important role in world monetary system. Thus, it is important to predict or to forecast gold prices due to the highly demand of gold in order to plan the right time to buy and invest gold. The current study proposed a univariate model that is Box-Jenkins method and time series regression method as a causal model to forecast gold prices. The data used in this study are daily prices from the 1^st of March 2011 to 23^th May 2011 and was analyzed using the Statistical Software Minitab Version 15.1.2. A comparison between these two models was made by evaluating the mean absolute deviation (MAD) and mean absolute percentage error (MAPE). The finding suggests that time series regression method produced a smaller value of MAD and MAPE compared to Box-Jenkins ARIMA (0,2,1). This means that in forecasting gold price data, another independent variable that influences them can provide extra information to forecast gold prices more precisely.