Climate calm? Long-run temperature signals and volatility in EU carbon futures

We test whether long-horizon climate information explains and forecasts volatility in European Union Allowance (EUA) futures. Daily Phase II–IV prices (2008–2024) are combined with monthly climate proxies in a two-step GARCH-MIDAS design. First, a state-space filter extracts the stochastic trend in global temperature anomalies; the trend’s first difference captures incremental shifts in the warming path. Second, an indicator flags months when the exponentiated trend exceeds its 88th percentile, isolating extreme shifts rather than ordinary weather noise. Both proxies load negatively in the MIDAS filter: faster warming and more frequent extreme shifts are associated with lower conditional variance once high-frequency weather shocks are removed. Out-of-sample, the climate-augmented model reduces one-day-ahead volatility errors by roughly 20% relative to the best conventional GARCH, a lead validated by Model Confidence Set and Diebold–Mariano tests. Feeding these forecasts into a mean–variance rule delivers higher excess and certainty-equivalent returns across risk-aversion levels. Incorporating slow moving climate metrics therefore sharpens volatility estimates and yields a clear volatility dampening effect, enhancing risk-management tools for investors and regulators.