Climate policy uncertainty and the U.S. stock markets: The predictability using particle swarm optimization with eXtreme gradient boosting

Abstract

Climate policy uncertainty is assuming an increasingly pivotal role in the global financial markets, presenting systemic risks to the financial sectors. The relationship between the uncertainty surrounding climate policies in the United States (U.S.) and its repercussions on the U.S. financial markets clearly play a role towards a more ecologically sustainable economy. This study aims to examine the intricate relationship between the U.S. climate policy uncertainty and stock markets, concurrently identifying the more robust machine learning model for predicting the U.S. stock market grounded in climate policy uncertainty. The results obtained from the quantile regression method show that the U.S. stock indexes are affected by the climate policy uncertainty more pronounced at the right tails which means the climate policy uncertainty is more likely to affect the boom stock markets. The XGBoost and PSO-XGBoost models outperform other machine learning models in terms of predictability for the U.S. stock markets using the US climate policy uncertainty while the PSO-XGBoost outperforms XGBoost. The findings shed light on the relationship between the U.S. climate policy uncertainty and stock markets. It also provides valuable insights to policymakers and investors when dealing with the effects of climate policies on the stock markets and making investment decisions using different machine learning algorithms, respectively.