Environmental Ethical and Sustainable investing (ESG) has AUM of more than USD 22.8 trillion. This paper provides a literature review on the concept of ESG alpha, the status quo and major drivers for future change (e.g. Green Covered Bonds). Although ESG alpha appears inconclusive under current market conditions, it is proposed that market innovation and regulatory pressure will force a near term ‘ESG tipping point’ where the inherent benefits of ESG will enable superior market returns. Drawing on conclusions from recent research in machine learning an alternative method to demonstrate such an ESG tipping point is proposed. A new ‘ESG Preference Model’ applies machine learning to force the use of ESG relevant Features for the prediction of market behaviour. A comparison of portfolio performance, one positioned in accordance with the ESG Preference Model and one via an ’Unconstrained’ benchmark model, trained on the same data and run in real time, could monitor the increasing ESG market impact and identify the ESG tipping point.
Paper
Full text
The Search for ESG Alpha by Means of Machine Learning - a Methodological Approach
Semantic Scholar · Environmental Science · 2020
Abstract
Environmental Ethical and Sustainable investing (ESG) has AUM of more than USD 22.8 trillion. This paper provides a literature review on the concept of ESG alpha, the status quo and major drivers for future change (e.g. Green Covered Bonds). Although ESG alpha appears inconclusive under current market conditions, it is proposed that market innovation and regulatory pressure will force a near term ‘ESG tipping point’ where the inherent benefits of ESG will enable superior market returns. Drawing on conclusions from recent research in machine learning an alternative method to demonstrate such an ESG tipping point is proposed. A new ‘ESG Preference Model’ applies machine learning to force the use of ESG relevant Features for the prediction of market behaviour. A comparison of portfolio performance, one positioned in accordance with the ESG Preference Model and one via an ’Unconstrained’ benchmark model, trained on the same data and run in real time, could monitor the increasing ESG market impact and identify the ESG tipping point.