We examine the relationship between ESG Scores and corporate financial risk, using credit ratings as a measure of financial stability. We rely on ESG Scores and the Rating SP Equivalent Rank indicator, that standardizes credit ratings, both indexes are downloaded from Refinitiv Provider. A key focus is on reconstructing missing credit rating data through shadow rating methodology, concerning both Altman Z Score and machine learning models. We apply ML techniques such as Linear Regression (LR), Neural Networks (NN), Random Forest (RF) and Support Vector Regression (SVR), with random forest identified as the most reliable method. Moreover, we refine this model to extend historical data and explore correlations with ESG Scores using various analytical techniques. In addiction, the analysis is conducted across focusing on different firm clusters, one based on firms credit rating (investment grade vs. high yield) and the other based on the geographic region in which the company operates. Finally, we examine the relation over time to assess consistency.
In questa tesi analizziamo la relazione tra i punteggi ESG e il rischio finanziario delle imprese, utilizzando i rating di credito come misura di stabilità finanziaria. I dati su ESG Score e sull’indicatore Rating SP Equivalent Rank, che standardizza i rating di credito, sono estratti da Refinitiv provider. Un aspetto centrale dell’analisi è la predizione e la ricostruzione dei rating mancanti tramite la metodologia dello shadow rating, che racchiude sia l’Altman Z-Score che modelli di machine learning. Le tecniche di machine learning applicate includono la Regressione Lineare (LR), le Reti Neurali (NN), la Random Forest (RF) e la Regressione Vettoriale di Supporto (SVR). Tra questi modelli, la Random Forest si dimostra il metodo più affidabile e performante. Questo modello viene quindi utilizzato per estendere l’analisi a dati storici ed esplorare le correlazioni con i punteggi ESG, impiegando diverse metodologie statistiche. L’analisi è condotta considerando diversi cluster di aziende: il primo basato sulla qualità del rating di credito (investment grade vs. high yield), il secondo sulla regione geografica in cui opera l’impresa. Inoltre, viene esaminata l’evoluzione della relazione tra ESG e rischio finanziario nel tempo, per valutarne lo sviluppo.
Exploring the relationship between credit ratings and ESG Scores using machine learning techniques
PAGANI, DAVIDE;PADELLI, CARLOTTA
2024/2025
Abstract
We examine the relationship between ESG Scores and corporate financial risk, using credit ratings as a measure of financial stability. We rely on ESG Scores and the Rating SP Equivalent Rank indicator, that standardizes credit ratings, both indexes are downloaded from Refinitiv Provider. A key focus is on reconstructing missing credit rating data through shadow rating methodology, concerning both Altman Z Score and machine learning models. We apply ML techniques such as Linear Regression (LR), Neural Networks (NN), Random Forest (RF) and Support Vector Regression (SVR), with random forest identified as the most reliable method. Moreover, we refine this model to extend historical data and explore correlations with ESG Scores using various analytical techniques. In addiction, the analysis is conducted across focusing on different firm clusters, one based on firms credit rating (investment grade vs. high yield) and the other based on the geographic region in which the company operates. Finally, we examine the relation over time to assess consistency.File | Dimensione | Formato | |
---|---|---|---|
2025_04_Padelli_Pagani_Thesis.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Corpo tesi
Dimensione
4.87 MB
Formato
Adobe PDF
|
4.87 MB | Adobe PDF | Visualizza/Apri |
2025_04_Padelli_Pagani_Executive_Summary.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Executive Summary
Dimensione
395.81 kB
Formato
Adobe PDF
|
395.81 kB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/235803