VECTOR AUTOREGRESSION AND MULTIRESPONSE REGRESSION APPROACHES FOR MODELING GOLD, TIN, AND NICKEL PRICES

Abstract

The mining sector plays a strategic role in the global economy, especially during periods of high economic uncertainty. Between 2020 and 2024, global markets experienced severe volatility due to the COVID-19 pandemic, geopolitical tensions, energy price shocks, and monetary policy tightening. These conditions intensified price fluctuations in major mining commodities such as nickel, gold, and tin. However, limited empirical research has compared different multivariate modeling approaches for analyzing commodity price dynamics during this volatile period. This study examines the dynamic relationships between nickel, gold, and tin prices and key global factors, namely crude oil prices, the USD exchange rate, and silver prices, using monthly data from January 2020 to December 2024. The VAR model captures temporal interdependencies among variables, while the MRR model examines simultaneous relationships among multivariate response variables. The stationarity and cointegration tests show that all variables become stationary after first differencing and exhibit no long-term equilibrium relationship. The Impulse Response Function (IRF) and Variance Decomposition (VDC) analyses reveal that fluctuations in nickel, gold, and tin prices are primarily driven by their own past values, with minor cross-commodity effects. The MRR results indicate that crude oil and silver prices significantly influence metal price variations, while the USD exchange rate has the strongest overall effect. The comparison across three evaluation metrics shows that the MRR model provides better predictive performance than the VAR model. The MRR model yields higher R² than VAR, which records R² of 0.339, 0.584, and 0.529, with MAPE up to 22.42%. The results demonstrate that the MRR model consistently outperforms the VAR model, providing stronger explanatory power and higher predictive accuracy. These findings highlight the added methodological value of comparing VAR and MRR models and offer practical insights for investors, industry stakeholders, and policymakers in managing commodity price risk under volatile economic conditions.