Decoding the Spill Over and Predictive Power of Global Markets on Green Bonds: Deep Learning Perspective

In this study, I examine whether global financial markets can predict the global green bond prices, and how they are connecting with green bonds. The study uses, daily time series data from 04:2013 to 03:2025, and applies DCC-GARCH, Diebold and Yilmaz (2012) spillover framework, and deep learning-based Recurrent Neural Network-Long Short-Term Memory (RNN-LSTM), and Graph Neural Network – Graph Attention Network (GNN-GAT) based models. The combined approach of traditional and deep learning-based models, found the financial markets to be significantly connected with the green bonds. Highest linkage of green bonds is found with the US Dollar Index, crude oil prices and global equities index. However, global commodity index and crypto markets are relatively weakly connected. However, as per the deep learning-based models, US dollar index strongly overshadows the other variables in impacting the green bonds. The novel findings have significant implications for the investors in the sustainable finance, as study finds diversification opportunities and supports portfolio optimisation.