Repository navigation
ESG Scoring System
This Python script establishes an ESG (Environmental, Social, and Governance) scoring framework designed to assign weighted scores to a company’s performance in three key areas: Environmental, Social, and Governance. The data extraction process involves several considerations:
Environmental Data: The Environmental score is derived from the latest available ESG report for each company. This report usually covers multiple years of environmental statistics, meaning that the most recent report provides a sufficient and comprehensive summary of a company’s environmental performance over several years. For this reason, we rely on the latest year’s ESG report to calculate the environmental score, as it consolidates multiple years of environmental data in a single report.
Social and Governance Data: In contrast, the Social and Governance metrics reported in an ESG report are specific to the year of the report. As these metrics are year-specific and do not usually aggregate multiple years of data, we collected data from multiple ESG reports across different years to build a comprehensive dataset for Social and Governance scores. This approach ensures that we capture the evolution of a company’s performance in these two areas, providing a more complete assessment.
The ESG rating system is tailored to the IT industry, referencing the Singapore Exchange (SGX) model, which incorporates the industry's best practices and guidelines. Based on SGX's ESG Ratings Model Hierarchy, these sub-indicators below are selected to measure environmental, social and governance categories. The allocation of weights for the Environmental, Social, and Governance categories, as well as the local weights for each individual metric, are referenced from established sustainability frameworks, including MSCI ESG Ratings, SASB, and other leading industry standards.
MSCI ESG Materiality Map for IT industry:
-
Indicators:
- Energy Consumption (40%)
- GHG Emissions (20%)
- Water Usage (20%)
- Waste Generated (20%)
- Metrics normalized per employee.
-
Indicators:
- Binary indicators: Data Security, Customer Privacy, Cybersecurity (20% each)
- Continuous indicators: Work-Related Injuries, Training Hours (10% each)
-
Indicators:
- Ethical Behavior (30%)
- Risk Management (30%)
- ESG-relevant Certifications (25%)
- Board Composition (15%)
- Points assigned based on the presence of governance practices and certifications.
This hierarchy and the allocation of different weights highlight the importance of privacy, ethical practices, and sustainability in IT governance.
For each sub-indicator, a metric score ranging from 0 to 10 is allocated. The final ESG score is calculated by multiplying the individual metric scores with the pre-defined local weights of the metrics (and the global weights of the E,S and G) categories.
In the Environmental (E) category, weighted at 40%, the scores for Energy Consumption, GHG Emissions, Water Usage, and Waste Generated, are calculated with respective weights of 0.4, 0.2, 0.2, and 0.2. To ensure fairness in company size, each metric is normalized by dividing by the company’s number of employees. The metrics are then standardized and percentile-ranked to create a relative yearly comparison of environmental performance across companies
For the Social (S) category, contributing 30% to the overall score, sub-indicators include Training Hours, Occupational Health and Safety, Gender and Age Diversity, Data Security, CyberSecurity and Customer Privacy. These indicators were chosen to reflect the IT industry’s emphasis on privacy, security, and responsible labor practices. Binary indicators like DataSecurity, CustomerPrivacy, Cybersecurity, GenderStats, and AgeStats, as well as standardized continuous indicators such as WorkRelatedInjuries and TrainingHours are selected. Data Security, Customer Privacy, and Cybersecurity indicators are each weighted at 0.2, while GenderStats, AgeStats, WorkRelatedInjuries and TrainingHours are weighted at 0.1 each.
In the Governance (G) category, weighted at 30%, governance practices are evaluated based on Board Composition, Ethical Behavior, and Risk Management as binary indicators, with 10 points awarded if present and 0 if absent. For CertificationList, each relevant certification mentioned in a company’s report contributes 2 points, capped at a maximum of 10. The weights for Ethical Behavior, Risk Management, ESG-relevant Certifications are set at 0.3, 0.3, 0.25 and 0.15 respectively. This structure underscores the importance of ethical practices, accountability, and adherence to international standards within IT governance.
The final ESG score is obtained by combining the weighted scores of the Environmental, Social, and Governance components according to their overall weights, and normalized between 0 and 10.
This comprehensive framework allows for a more tailored ESG score calculation for companies in the specific industry. The standardisation process eliminates bias and ensures that the scores allocated to the performance across companies are fair and comparable.
The percentile scoring used for quantitative metrics emphasises on relative performance rather than absolute values, meaning that ESG scores may fluctuate when the dataset is expanded. However, in this project, the interpretability of the scores is more crucial, as this makes it easier to identify where a company stands in comparison to its peers or competitors. Across multiple years, percentile scores provide a clearer understanding of a company's progress by highlighting the degree of improvement and competitiveness in a specific area relative to its peers.
For the qualitative indicators, the system uses a binary scoring to check for keywords present, which may overlook important contextual details. Future work can involve using NLP to help analyze the context and sentiment of the text.
The prediction system uses historical ESG scores to project future scores for companies. It is designed to help forecast Environmental, Social, and Governance (ESG) scores over the next three years while keeping predictions within reasonable bounds.
For companies with less than two years of historical data, the system forecasts future scores based on historical averages, providing a simple baseline for the next three years. For companies with sufficient historical data, the model uses Exponential Smoothing to capture trends and forecast each ESG component. To prevent unrealistic exponential growth, we apply constraints by setting confidence intervals based on historical data, ensuring predictions remain realistic. After separately predicting Environmental, Social, and Governance scores, the system combines these using a weighted formula (0.40 for Environmental, 0.30 for Social, and 0.30 for Governance) to produce a final ESG score. All scores are clipped between 0 and 10 to standardize results.
A subset of data consisting of 8 companies is used to explore the relationship between their Environmental, Social, and Governance scores and financial performance for the year 2023. This analysis aims to assess whether a company’s ESG performance correlates with its financial metrics, providing insights for investors and stakeholders interested in both sustainable and financial performance.
The analysis relies on financial data extracted from publicly available financial reports and stock price histories via the Yahoo Finance API (using the yfinance package in Python). For each of the 8 companies, relevant market indexes were identified to enable a comparison between individual company performance and broader market trends.
The following financial metrics are calculated based on the available data:
- Return on Equity (ROE): A measure of financial performance calculated by dividing net income by shareholders' equity.
- Return on Assets (ROA): Reflects how effectively a company is utilizing its assets to generate profit.
- Debt-to-Equity Ratio (D/E): A measure of a company's financial leverage calculated by dividing total liabilities by shareholders' equity.
- Total Assets (in thousands of USD): The sum of a company's assets, representing the total resources available to the company.
- Equity Beta (β): A measure of the company's stock price volatility in relation to the market, indicating the company’s risk level.
- Mean Stock Price: The average stock price over the course of the year.
- Year-end Stock Price: The stock price at the end of the year, used to assess the company’s performance relative to the market.
To perform statistical analysis, a correlation matrix is calculated using Pearson's coefficient and stored in the database. This matrix quantifies the strength and direction of the linear relationships between variables, which helps to identify whether higher ESG scores are associated with better or worse financial performance. Additionally, metrics that show at least a 90% significance level in their correlation with the ESG score were identified, providing statistical evidence of key relationships for further investigation.
This analysis is limited in its sample size, data availability and time period. However, it serves as an important starting point in understanding the relationship between ESG performance and financial outcomes, and this system and its methodology can be further refined to build more advanced statistical models to combine ESG and financial risk, and even provide sustainability insights and decarbonisation strategies for companies based on their strengths and weaknesses.