Empirical Analysis and Forecasting of Gold Prices: Based on VAR Model

Abstract

To achieve accurate forecasting of future gold prices and thereby provide effective references for investors' decisions, this study selects time-series data of gold prices (gold), the U.S. Dollar Index (DXY), and WTI crude oil futures prices (WTI) from 2015 to 2020. Employing a comprehensive empirical framework that integrates stationarity tests, lag order selection, VAR model estimation, Wald tests, unit circle tests, Granger causality tests, impulse response analysis, and out-of-sample forecasting, this research systematically examines the dynamic interactive mechanisms among the three variables. The findings reveal a distinct asymmetric linkage effect: DXY exerts a more pronounced impact on gold prices than WTI, with a significant short-term negative correlation between first-order lagged DXY and gold. Granger causality tests confirm that DXY is a Granger cause of gold prices, while gold serves as a Granger cause for both DXY and WTI. Impulse response analysis further demonstrates that gold prices exhibit high short-term sensitivity to DXY shocks, with a negative inhibitory effect persisting for the first three periods before gradually weakening. Moreover, a 6-period ahead forecast of gold price first-order differences shows that the VAR model’s predicted values align closely with actual values in terms of trend and magnitude, verifying its reliability in forecasting. This study offers actionable insights for investors to mitigate risks and for policymakers to monitor the stability of the gold market.