Valueco's paper on Bias assessment and mitigation for ESG scoring models investigates how to assess and quantify bias when constructing environmental, social and governance (ESG) ratings.
The paper caught the judges' attention for attempting to address the challenge of a lack of standardised ESG rating assessments.
This makes comparability an issue and methodologies must be assessed at the individual rating level to understand their relative value. Such variability can leave ESG ratings and impact reporting claims open to criticisms of ESG-washing.
"The bias assessment could be an entry point for the world to agree upon a standard ESG rating system that will be beneficial for all stakeholders," said one IMPACT awards judge.
Valueco, which is an ESG benchmark platform, separated internal and external bias factors and proposed a methodology to mitigate these biases to compute an ESG score that is independent of such factors.
The model also proposes a framework to detect the companies whose intrinsic performances outperform their peers independently of external biases, such as company size, activity sector, and main geographical area.
According to Valueco, the originality of the approach lies in the combined use of data science and game theory techniques to analyse the data.
"Bias analysis can be used to control the positioning of a scoring model on the market and assess its consistency with the investment strategy of an asset manager. The practical uses of these analyses are obvious for many players in the sustainable finance field, from supporting extra-financial communication with quantitative insights to making more informed investment decisions," said Valueco chief technology officer Mathieu Joubrel.
The source of divergence in ESG ratings was studied by Florian Berg, researcher at Massachusetts Institute of Technology, Julian Kölbel, assistant professor for sustainable finance, University of St. Gallen, and Roberto Rigobon a Professor of Applied Economics at the MIT Sloan School of Management.
They identified three main sources:
- Measurement: the indicators used to measure a specific attribute (accounts for 56% of the divergence);
- Scope: the set of attributes used in the scoring model (accounts for 38% of the divergence);
- Weights: the relative importance of the attributes used in the model (accounts for 6% of the divergence).
Bias analysis can bring a clearer structure to the ESG market and make it possible for all stakeholders to identify relevant signals and levers to improve their ESG performances, be they corporates or financial institutions, they concluded.
"The divergence in ESG ratings is a major obstacle for finance professionals who want to set up sustainable investment policies. I believe it can become an asset for responsible investors who are able to disentangle the reasons behind this diversity of opinions. I hope this award drives more attention to collaborative projects trying to tackle this issue," commented Joubrel.