Abstract:
We analyze the predictive value of climate risks for state-level realized stock market volatility, computed, along with other realized moments, based on high-frequency intra-day U.S. data (September, 2011 to October, 2021). A model-based bagging algorithm recovers that climate risks have predictive value for realized volatility at intermediate and long (one and two months) forecast horizons. This finding also holds for upside (“good”) and downside (“bad”) realized volatility. The benefits of using climate risks for predicting state-level realized stock market volatility depend on the shape and (as-)symmetry of a forecaster’s loss function.
