The Impact of Global PV Solar Expansion on Gold and Silver Price Dynamics: Machine Learning Forecasts and Economic Implications

This study investigates how global photovoltaic (PV) solar expansion shapes gold and silver price dynamics using a novel annual dataset (2000–2024) and a hybrid empirical framework. The analysis integrates standard econometric tools (ADF, Johansen cointegration, panel OLS, Granger causality) with machine learning forecasts (LSTM and XGBoost), interpretability (SHAP), and stress testing to capture both causal channels and predictive performance. The results revealed three main findings. First, the Granger causality tests indicate that cumulative PV capacity significantly influences gold prices at short lags (1–2), while the effect on silver is statistically insignificant, pointing to indirect channels through macroeconomic factors and portfolio rebalancing. Second, panel regressions show that PV-driven silver demand exerts a modest impact on silver prices, whereas macroeconomic factors (the USD index, inflation, and geopolitical risk) remain the dominant drivers for both metals. Third, XGBoost provides superior out-of-sample accuracy compared to LSTM, and SHAP analysis highlights a multi-year lag structure (approximately four years) for the PV capacity effects. Stress testing further indicates that precious metals remain highly sensitive to macro shocks despite structural pressures from PV expansion. This study contributes a unified dataset, documents a causality paradox between PV expansion and precious metals, and demonstrates the practical value of combining machine learning with econometric methods for scenario forecasting. The policy implications include the need for strategic mineral planning, dynamic hedging strategies, and central bank awareness of energy-transition effects on inflation expectations.