Climate Disaster, Investor Attention, and Tail Risk: Graph-based CoVaR

Abstract

We evaluate how climate disasters affect tail risk and conditional tail dependency in the energy and agricultural commodity markets. We employ an innovative Quantile LSTM-GNN method to capture the time-varying graph-based structure of tail-risk spillover networks. Using regression and event analyses, we show that climate disasters, especially droughts, significantly increase both tail risk and conditional tail dependence, emerging before disasters occur. Investor attention further amplifies the impact of climate disasters on tail risk. However, climate disasters do not alter the underlying structure of tail-risk networks.