We tackle the issue of producing point, sign, and density forecasts for the monthly real price of carbon in the European Union Emissions Trading System (EU ETS). We show that Bayesian Vector Autoregressive (BVAR) models augmented with factors based on macroeconomic and financial variables yield improvements in point and density forecasting performance, with the largest point-forecast gains emerging at intermediate and longer horizons. In particular, the one-factor BVAR model provides lowest values at six and nine months ahead, while the baseline BVAR performs best at the one-year horizon. By contrast, a large BVAR including all predictors individually does not improve point forecast accuracy, supporting the use of parsimonious model specifications. Simple forecast pooling delivers modest gains from medium horizons onward, but does not mitigate the time variation in relative forecast performance. We also provide a qualitative comparison of model-based forecasts with survey expectations and forecasts released by data providers, which further highlights that relative forecast performance varies over time. Moreover, we consider verified emissions and show that stochastic volatility improves point forecasts, with reductions of about 6.5 to 7.1% at the one-year horizon. Lastly, we use model-based forecasts to construct market monitoring tools that track demand and price pressure in the EU ETS.

翻译：暂无翻译