Tail-risk and sustainability

: can ESG scores accurately predict value at risk? : a machine learning based approach

Abstract

I, Ulrich Mohme, find in the scope of the master thesis: “Tail-Risk and Sustainability: Can ESG scores accurately predict Value at Risk? – A machine learning based approach“ that predicting Value at Risk at the 1% and 5% confidence level by applying various machine learning algorithms onto ESG scores show low degrees of accuracy. Random Forest Regressors show the highest degree of accuracy from the algorithms used and the ESG as well as Environmental Score correlate most strongly with Value at Risk indicating the most significant predictive power. Data from companies listed in the S&P500 are used from the year 2000 to 2024. The findings imply ESG scores alone not to be a reliable predictor of Value at Risk at various significance levels. Yet a slightly linear correlation is detected and machine learning algorithms outperform benchmark linear regression models.

Date of Award	25 Jun 2024
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Paul Karehnke (Supervisor)