Authors:
Dominic Strube
1
and
Christian Daase
2
Affiliations:
1
Hochschule Wismar, University of Applied Sciences, Technology, Business and Design, Wismar, Germany
;
2
Institute of Technical and Business Information Systems, Otto-von-Guericke University, Magdeburg, Germany
Keyword(s):
Environmental, Social, Governance (ESG), Impact Investing, Financial Performance, Sustainability, Machine Learning.
Abstract:
This short study uses machine learning (ML) to investigate whether the inclusion of sustainability ratings in the training data can improve the estimated accuracy of the prediction of a company’s abnormal returns. For this purpose, we examined 72 companies that are listed in the indices EURO STOXX 50 ® or/and EURO STOXX 50 ® ESG or/and EURO STOXX® ESG LEADERS 50. We found out that the mean-adjustment model used to estimate returns produces more accurate results than with adding MSCI’s sustainability ratings. The preliminary results suggest that sustainability ratings are currently inappropriate for estimating expected or abnormal returns and their inclusion in the training data interferes the algorithm behind the ML approach. By extension, this leads to the assumption that the relation between ESG ratings and a business’ success are suitably irregular to significantly decrease an ML models quality.