Accuracy of Green Bond Issuance Predictor

Abstract

Climate change is affecting the development of many industries in different aspects. These impacted enterprises transform into sustainable enterprises to avoid the risks, and by doing so they enter into the green bond market. The current literature provides effective reference indicators for participants in the green bond market. These indicators illustrate the funding size of the green bonds in different dimensions to the participants. As for the improvement of the policies about environmental protection there also emerge some new indicators such as ESG score. Besides, machine learning is an accurate and effective tool in many fields, and some researchers have established a model for predicting the issuance of green bonds but have not involved the new indicators in the past. In this paper, on the one hand, we discuss the new indicator, ESG scores, and how it affects the funding size of the green bonds, on the other hand, we add this new indicator and the old indicators into four machine learning models to compare the accuracy of predicting the issuance of green bonds of these four models. In these four models, the Random Forest Regressor and LGBM Regressor are the best models on average. The former has the best performance of accuracy but needs much more time than the latter. On the opposite, the latter is the most efficient model among all but is the second most accurate. Besides, other models have the best numerical measurements in different dimensions which means we could use different models depending on different situations. Choosing the proper model for the specific situation can optimize the benefit of the green bond market participant.