Benchmark selection for monthly real EU ETS carbon price forecasting

Determine the appropriate benchmark model for evaluating point and direction-of-change forecasts of the monthly real price of EU ETS allowances at the monthly real frequency, given that most prior carbon price forecasting studies use daily or weekly nominal data and do not establish a benchmark suitable for monthly real prices.

Background

The paper focuses on forecasting the monthly real price of carbon in the EU Emissions Trading Scheme and contrasts this objective with a literature that predominantly relies on daily or weekly nominal data. This difference in frequency and deflation raises the question of what constitutes an appropriate benchmark when assessing forecast accuracy for monthly real prices.

In their empirical work, the authors compare several univariate and multivariate models, including Random Walk, ARIMA, and Bayesian VAR specifications, but they explicitly note the initial uncertainty surrounding benchmark choice at the monthly real frequency. Establishing a canonical benchmark is important for consistent evaluation across studies and for informing policy-relevant forecasting exercises.