Gold price modeling using system dynamics

Abstract

The global gold market has recently attracted a lot of attention and the price of gold is relatively higher than its historical trend. This paper constitutes the first exercise of system dynamics applied to predict gold price in monthly frequency from January 2010 to June 2011. Rather than static forecasting characteristics found in another quantitative method, time-series, system dynamics allows possibility for prediction based on capturing causal interactions and consequently the feedback loops usually found in a complex system behaviour such as the gold price system. Therefore, it was expected that moving toward forecasting method utilizing system dynamics model could result in better prediction of the gold price. Our paper supports such hypothesis. Having ability to take into account of qualitative factors particularly political chaos and economic crisis events, the model developed in this paper reduces the prediction error, mean absolute percent error (MAPE), to merely 2% compared to approximately 9% error found with Holt-Winter Exponential Smoothing and 11% error using Box-Jenkins Method.