In recent years, environmental, social, and governance (ESG) considerations are increasingly important in the financial sector due to improvements in the regulatory framework including key pieces such as the European Union’s Corporate Sustainability Reporting Directive (CSRD) and the EU’s Sustainable Finance Disclosure Regulation (SFDR). This is where the accountability and transparency aspect comes into play, which of course drives businesses to integrate them in their decision-making. Of course, you can suffer the consequences for noncompliance. My thesis applies machine learning techniques to them in order to quantify the relationships between ESG and financial performance. Using ESG and financial databases from Refinitiv, this study reported from the fiscal years of 2020 through 2023 and analyzed models of Linear Regression, Random Forest, XGBoost, and Support Vector Regression (SVR) to predict Earnings Per Share (EPS), Return on Assets (ROA), and Revenue. Feature engineering techniques such as standardization, temporal means & interaction terms were employed to boost model performance. The results presented in this table show that ensemble-based approaches, and notably XGBoost, regularly outperform simpler models with high R-squared values and low root mean square Error (RMSE). The result confirms that governance and environmental features do matter, with environmental pillar scores appearing as consistent predictors of revenue and EBITDA. Sectoral and geographic analyses provide additional evidence of divergent results: Strong ESG performance correlates with better financial results across both countries such as Finland and Sweden and industries such as utilities. This research highlights the significant role of ESG data in modeling financial performance and the need for more sophisticated machine learning techniques to identify complex and nonlinear relationships.
Negli ultimi anni, le considerazioni ambientali, sociali e di governance (ESG) stanno diventando sempre più importanti nel settore finanziario grazie ai miglioramenti nel quadro normativo, tra cui elementi chiave come la Direttiva sulla rendicontazione di sostenibilità delle imprese (CSRD) dell'Unione Europea e il Regolamento sulla informativa sulla finanza sostenibile (SFDR). È proprio qui che entra in gioco l'aspetto della responsabilità e della trasparenza, che ovviamente spinge le aziende a integrare tali considerazioni nei loro processi decisionali. Naturalmente, la mancata conformità può comportare conseguenze negative. La mia tesi applica tecniche di machine learning per quantificare le relazioni tra ESG e performance finanziaria. Utilizzando database ESG e finanziari di Refinitiv, questo studio, riferito agli anni fiscali dal 2020 al 2023, ha analizzato modelli di Regressione Lineare, Random Forest, XGBoost e Regressione con Support Vector (SVR) per prevedere l’Utile per Azione (EPS), il Rendimento delle Attività (ROA) e i Ricavi. Sono state applicate tecniche di feature engineering come la standardizzazione, le medie temporali e i termini di interazione per migliorare le prestazioni dei modelli. I risultati presentati in questa tabella dimostrano che gli approcci basati su ensemble, in particolare XGBoost, superano regolarmente i modelli più semplici, mostrando valori di R-quadrato elevati e bassi errori quadratici medi (RMSE). Il risultato conferma che le caratteristiche legate alla governance e all’ambiente sono significative, con i punteggi del pilastro ambientale che si rivelano predittori costanti di ricavi e EBITDA. Analisi settoriali e geografiche forniscono ulteriori evidenze di risultati divergenti: una forte performance ESG correla con risultati finanziari migliori in paesi come Finlandia e Svezia e in settori come le utility. Questa ricerca evidenzia il ruolo significativo dei dati ESG nella modellazione della performance finanziaria e la necessità di tecniche di machine learning più sofisticate per identificare relazioni complesse e non lineari.
GESTIONE DEI RISCHI AMBIENTALI, SOCIALI E DI GOVERNANCE E PERFORMANCE FINANZIARIA: UN APPROCCIO DI MACHINE LEARNING PER QUANTIFICARE LA RAPPORTO
CEESAY, MUSA
2023/2024
Abstract
In recent years, environmental, social, and governance (ESG) considerations are increasingly important in the financial sector due to improvements in the regulatory framework including key pieces such as the European Union’s Corporate Sustainability Reporting Directive (CSRD) and the EU’s Sustainable Finance Disclosure Regulation (SFDR). This is where the accountability and transparency aspect comes into play, which of course drives businesses to integrate them in their decision-making. Of course, you can suffer the consequences for noncompliance. My thesis applies machine learning techniques to them in order to quantify the relationships between ESG and financial performance. Using ESG and financial databases from Refinitiv, this study reported from the fiscal years of 2020 through 2023 and analyzed models of Linear Regression, Random Forest, XGBoost, and Support Vector Regression (SVR) to predict Earnings Per Share (EPS), Return on Assets (ROA), and Revenue. Feature engineering techniques such as standardization, temporal means & interaction terms were employed to boost model performance. The results presented in this table show that ensemble-based approaches, and notably XGBoost, regularly outperform simpler models with high R-squared values and low root mean square Error (RMSE). The result confirms that governance and environmental features do matter, with environmental pillar scores appearing as consistent predictors of revenue and EBITDA. Sectoral and geographic analyses provide additional evidence of divergent results: Strong ESG performance correlates with better financial results across both countries such as Finland and Sweden and industries such as utilities. This research highlights the significant role of ESG data in modeling financial performance and the need for more sophisticated machine learning techniques to identify complex and nonlinear relationships.File | Dimensione | Formato | |
---|---|---|---|
THESIS ESG_pdfA.pdf
accesso aperto
Dimensione
4.43 MB
Formato
Adobe PDF
|
4.43 MB | Adobe PDF | Visualizza/Apri |
È consentito all'utente scaricare e condividere i documenti disponibili a testo pieno in UNITESI UNIPV nel rispetto della licenza Creative Commons del tipo CC BY NC ND.
Per maggiori informazioni e per verifiche sull'eventuale disponibilità del file scrivere a: unitesi@unipv.it.
https://hdl.handle.net/20.500.14239/29121