Predictability of climate policy uncertainty index

Since Gavriilidis (2021) constructed the monthly climate policy uncertainty (CPU) index, researchers have used the index to investigate the impact of CPU on economic, financial, and energy markets at the level of both mean and volatility. Focusing on the predictability of CPU on energy returns using linear predictive regression, we apply the well-known Hill tail index estimation to the product of the residual and predictor and find that the least squares estimate has a nonnormal limit, implying that Newey-West t-tests are problematic due to the lack of finite moments. Therefore, we propose employing the weighted least squares estimate and random weighted bootstrap method, as presented in Jiang et al. (2025), to test the predictability of the CPU. Data analysis yields different results when considering and ignoring the lack of finite moments. It also indicates that the CPU index, regardless of using the US CPU index, the Global CPU index, or the China CPU index, exhibits weak predictive power over daily returns of solar, wind, clean energy, coal, and crude oil markets. Limited predictability is observed only in the natural gas market and only under a subset of model specifications. The CPU mostly influences the natural gas market in over 10 out of 30 considered cases. The results suggest that climate policy uncertainty affects energy markets primarily through volatility and risk channels rather than through systematic return predictability, thereby providing empirical support for the widespread use of GARCH-MIDAS-type volatility models in the existing literature.