Thesis

Corporate sustainability performance : analysing sustainability initiatives in multinational corporations through a hybrid meta-analysis and predictive modelling approach

Downloadable Content

Download PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2026
Thesis identifier
  • T18151
Person Identifier (Local)
  • 202554249
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • In recent years, the imperative for corporate sustainability has gained significant traction, with Multinational Companies (MNCs) increasingly acknowledging their responsibility to support global sustainability agendas, particularly the United Nations Sustainable Development Goals (UNSDGs). Prior studies largely focus on descriptive assessments of sustainability performance, meanwhile there is limited empirical evidence exists on the predictive capability of company-level ESG disclosures in explaining and forecasting corporate sustainability performance (CSP). Therefore, this thesis addresses this gap by assessing the relationship between sustainability initiatives and CSP within MNCs and evaluates the usefulness of predictive modelling in projecting future sustainability outcomes. Subsequently, Environment, Social, & Governance Key Performance Indicators (ESG KPIs) that were manually taken from corporate sustainability reports between 2017 and 2023 are synthesised by the meta-regression component. This method aggregates quantitative variables, such as environmental, social, and governance factors, to assess their empirical links with CSP across several sectors instead of examining secondary studies. This offers a methodical, cross-company review of how sustainability initiatives impact external ESG assessments and reflect operational performance. Moreover, machine learning techniques include Random Forest, Extreme Gradient Boosting (XGBoost), Least Absolute Shrinkage and Selection Operator (LASSO regression), Decision Tree, and Support Vector Regression (SVR), are employed to generate predicted CSP scores and compare against actual outcomes. By capturing both linear and non-linear patterns in ESG datasets, these models analyse complicated relationships that are frequently missed by conventional statistical techniques. This study evaluates the accuracy and robustness of predictive methods for projecting future CSP outcomes by comparing model performance using assessment criteria like Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination or R-squared (R²). Finally, the study shows how predictive analytics can help with strategic planning and sustainability evaluation through hybrid approach. It also offers new insights into the relationship between Corporate Sustainability Performance (CSP) in Multinational Corporations (MNCs) and ESG-related key performance indicators (KPIs) reported in sustainability reports. The results demonstrate that prediction accuracy differs among machine learning models and indicate quantifiable relationships between specific ESG KPIs and CSP. In comparison to linear methods, ensemble-based and non-linear approaches produce predicted CSP scores with greater accuracy among the five models used. This study provides a forward looking analytical viewpoint that helps MNCs, investors, regulators, and other stakeholders make more informed sustainability-related decisions in line with global sustainability agendas by contrasting actual and anticipated CSP outcomes.
Advisor / supervisor
  • Quigley, John, Dr.
  • El Raou, Hanane
Resource Type
DOI

Relations

Items

Thumbnail Title Date Uploaded Visibility Actions
file details PDF of thesis T18151 2026-09-17 Public Download