The ESG Risk Insight Tool is based on an in-depth content analysis of annual reports published by a selection of Eurostoxx companies between 2016 and 2023.

1. Data collection
   Annual reports were examined to identify all sentences that:

• Contained the word “risk” or any derivative (e.g., risks, risky, risk-taking).
• Related specifically to non-financial risk categories defined by the EU Non-Financial Reporting Directive (NFRD):
  • Environmental matters
  • Social and employee-related matters
  • Human rights
  • Anti-corruption and bribery matters

Sentences concerning financial risks or risks outside the NFRD scope were excluded.

2. Assessing disclosure quality
   Each risk-related sentence was manually coded by an independent coder according to three analytical dimensions:

• Risk type – Following NFRD categories: environmental, social/employee, human rights, anti-corruption/bribery.
• Tone – Classified as:
  • Positive: portraying risks in a favorable light
  • Negative: portraying risks unfavorably
  • Neutral: factual or without clear sentiment
• Temporal perspective – Classified as:
  • Past: referring to previous years
  • Present: referring to the current year
  • Future: referring to upcoming years
• Completeness – Reflecting depth and specificity:
  • Mention: vague, unspecific references
  • Description: detailed and explanatory
  • Evaluation: supported by quantitative or numerical evidence

This multi-dimensional approach enables the ESG Risk Insight Tool to capture both the extent and quality of ESG risk reporting, facilitating meaningful benchmarking across companies, sectors, countries, and years.

3. Risk type metrics
   When users select a specific industry (e.g., Industry X), the tool calculates the share of reports in that sector that contain at least one sentence related to the selected risk type.
   Example for environmental risk:
   
   

If both industry and country are selected (e.g., Manufacturing and Italy), the formula becomes:

• Since each company publishes one report per year, the calculation refers to the share of reports containing a given type of information, not the share of companies.
• When data are shown by year, the interpretation can also be expressed in terms of share of companies.
• The database will be updated over time with new years and companies, and the algorithm is designed to adapt to these changes.