Do Carbon Price Forecasts Improve Compliance Procurement? Evidence from European Union Allowances
Firms covered by emissions trading systems need forecasts not only to value allowances, but also to decide when to buy them. This paper asks whether European Union Allowance (EUA) prices contain short-horizon predictability that survives a forecast-origin information design and improves simulated compliance procurement. Using daily data from 2019 to 2025, we produce direct forecasts for one to five trading days ahead. All predictors are observable at the forecast origin, and calibration and model-selection rules are fixed before the final holdout. The released forecast has the lowest point-estimate RMSE at every horizon among fourteen benchmarks, with the strongest loss-difference evidence at horizons three and four. Relative to a random walk, out-of-sample R^2 rises from 1.2% at one day to 15.5% at five days. We then use the forecast path in a constrained procurement problem with execution costs, market impact, capacity limits, and tail risk; sensitivity exercises add demand uncertainty. For a fixed 100,000-EUA order, optimized schedules lower average realized costs by 8.5 to 38.5 basis points relative to uniform execution across horizons h=2 to h=5. The gains come from reallocating purchases within a fixed window, not from reliable next-day directional timing.