Not a trading signal or financial advice.Disclaimer
EvergreenNovember 10, 2026

Copper Price Volatility: What Drives the Big Swings and How to Measure Them

Copper
Copper realized vol: 18-25% annualized, with event jumps dominating variance

Copper trades differently from most exchange-traded metals. Gold reprices on real rates and currency narratives; copper reprices on the marginal tonne. That shapes both its factor structure and its measurement problem: implied volatility prices the continuous component of variance well and the jump component badly.

Market microstructure: copper clears at the edge of the supply curve

The starting point is thin inventory. Copper's visible exchange inventory has hovered near one week of global refined consumption, with LME warrants falling to a small fraction of that in 2022 and 2023. Prices therefore set on marginal supply interruption risk, not average cost: when a mid-sized operation stops, the market reprices the supply curve rather than the cost curve, fast and convex.

The second-order effect sits in the spread. LME copper cash-to-3M backwardation reached roughly $200 to $400 per tonne in March 2022 and March 2023, against a structurally contangoed curve. Backwardation is both a physical signal and a carry cost that penalizes long inventory. Third, value density limits arbitrage: storage, financing and freight keep regional prices from converging quickly, and the 2022 LME rule changes on price limits and warehouse rent obligations altered that elasticity directly.

Demand: one credit cycle, one grid

Chinese demand accounts for roughly half of global refined copper consumption, concentrating the demand driver in a single property cycle and a single industrial policy regime. Property starts drove the 2015-16 base, the 2021-22 unwind and the post-2023 plateau. Grid capex and EV and renewables demand have partially decoupled grid copper from housing since 2021, but not fully. Transmission is fast: Chinese import arbitrage is tight, so a shift in import appetite surfaces in the LME price within days, which makes the copper-to-China-credit spread a cleaner systematic exposure than the outright level.

Supply: the fat left tail and the TC/RC tell

Mine disruptions define the fat left tail of copper supply risk. Escondida in 2017, Grasberg in 2018, and the Cobre Panamá shutdown, which removed roughly 1.5% of global mined copper supply from a market with thin visible stocks, are the reference events. Each landed on a market where the marginal tonne was already priced tight.

The second channel runs through smelters, not mines. Spot copper treatment and refining charges turned negative during 2024 and 2025 as Chinese smelter capacity outpaced concentrate availability. Negative TC/RC means smelters pay miners for concentrate, and the benchmark is a clean, high-frequency read on concentrate tightness and on forward expectations for refined surplus. The third channel is scrap, the marginal release valve in a stress event: price-elastic, but slow.

Measuring the swings: realized, implied, and jump decomposition

Three estimator families apply to copper futures. Close-to-close standard deviation of log returns is the baseline and undersamples intraday. Range-based estimators such as Parkinson and Garman-Klass are more efficient but assume continuous diffusion, which fails on news days. For rigorous work the intraday estimators lead: Yang-Zhang and bipower variation estimators outperform close-to-close realized volatility by separating continuous movement from jumps.

That separation is the point. A meaningful share of copper's realized variance arrives as jumps around policy, mine and exchange headlines rather than continuous diffusion, so a single 30-day close-to-close number cannot distinguish an elevated but stable regime from a compressed range with jump premium already loaded into the front contract.

On the options side, CME copper options are quoted against the front futures contract. CME copper options typically carry implied volatility in the 18% to 28% range, above subsequent realized volatility in most metals regimes, and repricing arrives ahead of the front-month roll and around Chinese policy windows. The tradable instruments are term structure, call skew and the variance risk premium. Copper vol is not homogeneous across tenors, and the front contract carries the largest event premium.

Positioning, risk, and the Volterra signal

For options desks, treat copper volatility as a term structure rather than a flat level. An elevated 7-day probability with a flat 30-day signal means the event premium is front-loaded: sell front-month richness and hold longer-dated exposure closer to model. HIGH and EXTREME readings justify carrying front-month put spreads and reducing delta, because the risk is gap risk rather than continuous drift. The mechanics of mapping probability signals onto a surface are set out in how options desks use volatility probability signals.

For risk and procurement teams, the concentration behind these tails is measurable ahead of the event. The Herfindahl-Hirschman index in commodity markets maps jurisdiction concentration onto the same left tail, and copper contracts reprice on interruption rather than demand collapse, which is the asymmetry to price into supplier terms.

How Volterra treats copper

Volterra's XGBoost model is walk-forward cross-validated with a mean AUC of 0.815 and processes 96 GDELT GKG news files daily. The Volterra dataset combines that news flow with supply chain, geographic concentration and market context features, then emits 7-day, 14-day and 30-day volatility probability forecasts across five risk levels from LOW to EXTREME.

The Volterra model measures the probability that copper's volatility regime changes within a horizon rather than forecasting the copper price, which is the correct target for hedging, inventory risk and supplier terms. The walk-forward validation note explains why out-of-sample discipline, rather than a single backtest, is what makes a signal usable at desk level.

Figures from the Volterra daily pipeline. Full historical backfill available on AWS Data Exchange.

Get daily volatility research data

10 minerals. 2 horizons. Delivered daily by 08:00 UTC.