Forecasting carbon returns under structural breaks and model uncertainty: a time-weighted regularized combination approach
提出时间加权正则化组合方法，在碳收益预测中识别结构变化并降低模型不确定性，实证表明该方法优于其他组合，且具有经济显著性。
The carbon emission trading systems undergo a series of phases and changes in their associated mechanisms and policies. The movement of carbon price returns is accordingly affected by structural breaks and fundamental shifts, which are empirically confirmed by the statistical tests. In this paper, we propose a time-weighted regularized combination (TWRC), which identifies the structural changes and mitigates the uncertainty among extensive factor-based forecasts. Our empirical findings suggest that the TWRC method outperforms other competitive combinations in forecasting carbon price returns. The forecasts of TWRC are also economically significant, which generate high utility in a portfolio exercise with carbon futures. The TWRC improves the forecast accuracy by lowering both the variance and bias components of the forecast error. The mechanism of the TWRC reveals that when structural changes are more frequent, indicated by the heavier weights assigned to recent forecasts, stronger shrinkages are imposed on individual models. This implies a trade-off between the recency of information and model dimensions in predicting carbon returns under structural changes and model uncertainty.