Medium-term and long-term volatility forecasts for EUA futures with country-specific economic policy uncertainty indices

This research uses country-specific economic policy uncertainty (EPU) indices to predict the volatility of European Union Allowance (EUA) futures and compares dimension reduction techniques and combination forecasting methods to improve the accuracy of EUA futures volatility. We use the classic autoregressive (AR) model as a benchmark and construct some extensions of this model. The empirical results show that the economic policy uncertainty indices mainly have long-term predictive ability for the volatility of carbon futures. Both dimension reduction techniques and combination forecasting methods can produce robust long-term out-of-sample forecasting effects. However, these effects depend on the overall forecasting ability of the variables participating in dimension reduction and combination to a certain extent. We have also made a model extension to consider the out-of-sample forecast performance during periods of high volatility and low volatility. We find that country-specific EPU indices are more effective in predicting low volatility of the EUA futures. Finally, we confirm that our results are robust based on the recursive forecasting method.