Energy transition metals (ETMs) have become increasingly critical to the global energy transition, yet their market dynamics are exposed to a rapidly evolving landscape of various types of crises. Understanding how these crises shape the behaviour of ETM markets and affect the return spillover dynamics is an important area of research that is untapped. This paper develops a two-dimensional crisis typology that classifies the impact of major crisis episodes on return spillover connectedness across ETM markets. Using daily data for 17 metals, spanning base metals, precious metals, battery metals, and technology metals, and the rolling-window Diebold–Yılmaz (2012) connectedness framework, the paper classifies 11 crisis events between 2017 and 2025 along two orthogonal axes: the direction of the connectedness shift (integration vs. fragmentation) and the stability of the response (volatile vs. stable), with the latter benchmarked against a bootstrap-derived threshold from pre-crisis markets. The typology reveals a fragmentation paradox: supply-side shocks (armed conflicts and bilateral trade restrictions) consistently decrease market connectedness, while demand-side shocks (pandemics, monetary tightening, and financial crises) consistently increase it. This dichotomy, which is robustly preserved across all 25 event-study window specifications, challenges the prevailing assumption in the connectedness literature that crises uniformly amplify systemic risk through heightened spillovers. The mechanism is grounded in the distinction between common-factor and idiosyncratic transmission channels: supply disruptions affect individual metals asymmetrically based on their geographic exposure, fragmenting the co-movement structure, while demand shocks operate through common macroeconomic channels that drive all metals together. The typology has implications for strategic reserves policy, portfolio hedging, and energy transition risk assessment, suggesting that geopolitical risk to critical minerals supply chains is more modular than the contagion hypothesis predicts.
