Abstract

We analyse the portfolio of European funds, which hold the French SRI label using a Machine Learning approach for clustering given a set of environmental variables.
We show that classifying funds in four clusters allows to better capture their heterogeneous level of greenness compared to the three Sustainable Finance Disclosure Regulation (SFDR) categories. We find that the dark green cluster corresponding to SFDR Article 9 is homogenous while the light green cluster related to SFDR Article 8 should be divided into two grades owing to its heterogeneity. We estimate that 14% of funds declared as Articles 8 and 9 should belong to the brown cluster implying serious greenwashing concerns.
Finally, we report a positive relationship between the financial and environmental performance of those clustered funds, confirming the appropriateness to grade the French SRI label in four levels, which makes it more relevant for investors and fund managers.

Co-authored with with M. Mercadier and Y. Rannou, and M. Mercadier.