Open Source Intelligence for Sustainability Reporting
Summary
From next month, 50,000 European and 10,000 non-EU companies will be subject to the new Corporate Sustainability Reporting Directive (CSRD). The purpose is to improve the reliability and utility of corporate sustainability information for investors in order to price in upside and downside risks. In other words, CSRD will attempt to build sustainability into share price.
Companies will start reporting FY2024 and the estimated costs to do so are significant: €1.7bn in first year one off costs and €1.9bn in ongoing annual costs to prepare the information and initially €2.6-3.9bn in annual “limited” audit costs expanding to €6.0-9.7bn as “reasonable” audit requirements come in over time. (EFRAG 2022). The next few years will see concerted efforts to use technology to meet reporting requirements without additional armies of auditors.
To be successfully adopted, any new technology here will need a number of characteristics:
This is a stringent list – too stringent for the current means of reporting. The Big Four auditors (PwC, EY, Deloitte, KPMG) have the manpower to address materiality but need to innovate their processes to cope with the information overload (Hartmann & Weissenberger, 2023). The large ESG ratings agencies (MSCI, Sustainalytics, ISS) are opaque in their methodologies, may have conflicts of interest in their business models, and investors only have "moderate confidence" in the ratings. And the academic research community continues to struggle to find a correlation between ESG ratings and share price.
Open Source Intelligence
An old technology is being repurposed to better meet these stringent needs. Open Source Intelligence (OSINT) is intelligence produced by collecting, evaluating and analysing publicly available information to answer specific questions. There are six sources of OSINT: traditional media (newspapers, magazines, etc); trade press (industry journals, academic publications, etc); social media (forums, blogs, Reddit, Twitter, etc); public government data (especially public hearings); companies’ own data (annual reports, quarterly earnings calls, etc); and grey literature (technical reports, working papers, newsletters).
The first OSINT examples date to mid-19th century in the United States and early 20th century in the UK (Block, 2023). OSINT was used in World War II by the Foreign Broadcast Monitoring Service (FBMS). A famous example of their research was spotting the correlation between the price of oranges in Paris with the successful bombings of railway bridges (Williams & Blum, 2018). More recently, the intelligence failures identified post-9/11 led to the establishment of Open Source Enterprise by the CIA. Similar ventures exist across the ‘Five Eyes’ network. There are also innovative not-for-profit investigative groups such as?Bellingcat.
While specific questions being researched have historically been related to national security, cyber security and war crimes, the same technology is now being used to answer other specific questions. For example, sustainability materiality. By gathering all the publicly available information about a company’s sustainability performance, it is possible to assess the materiality of individual aspects by the volume of conversation on that topic. Are Environmental issues more material than Governance issues according to all the public conversation in the market? How does that materiality profile evolve over time? How does it compare against peers? And against benchmarks?
OSINT materiality is different to traditional materiality assessments in that it is a posteriori rather than a priori. That is, by collecting all the relevant information in the public domain, OSINT materiality is calculated as the size of the conversation on the individual topic (driver) as a percentage of the total conversation (screen) every day. For example, greenhouse gas emissions as a driver is 8% material as a percentage of the total ESG screen on 1 January. Traditional materiality, by contrast, is estimated by expertise and experience once a year, and reported six months later. Given how central materiality is to the CSRD framework and its use in pricing sustainability risk, an annual analysis published 6 months after the fact with little comparable analysis is of very limited use to investors. Having identified material issues, further automation combined with human intelligence can give an assessment of the impact materiality (actual/potential, positive/negative, short/medium/long-term time horizons) or financial materiality (such as a company’s current or future cash flows, development, performance, position, cost of capital, or access to finance).
2. Models
Now that we are getting clarity and convergence around models of sustainability, it is possible to build open source data pipelines to populate those models. This involves transposing key drivers from a name and definition to sophisticated search Booleans in local language. This has the advantage of overcoming the well-known “information explosion” problem of OSINT. By using models, an OSINT pipeline need ingest ‘only’ millions of articles daily rather than billions. As ever, the “Garbage In Garbage Out” principle applies. The quality of the OSINT analysis is built on the robustness of the data sourcing, the cleanliness of the data gathered, and accuracy of the machine reading into the model. Once this data pipeline is built, however, it is relatively trivial to add ever more sources. For example, comparing the materiality of drivers in a company’s annual report with the conversation around that company in the market is a useful analysis for greenwashing risk. ?Which key topics are being glossed over and which are being over-focused on as distraction?
领英推荐
3. Correlation to share price
The clear purpose of CSRD is to facilitate better reflection of sustainability risks into market value. As such, it is imperative that the outputs of any new technology show a correlation to share price. A great deal of academic research is ongoing on whether, to what extent, and under what conditions this correlation exists. This research is hampered by the nature of the independent variable. Annual ratings with weak correlation between agencies raise considerable statistical challenges. OSINT has an advantage here in being available at the same cadence (daily) as share price close. OSINT delivers daily materiality and sentiment scores. By incorporating learnings from the academic research, OSINT ESG scores have strong and reliable correlation to share price both for quantitative trading firms with a 48-hour holding window and portfolio managers reviewing quarterly (Mettle Capital and Brunel University London, forthcoming 2024).
4. Transparent methodology
By its nature, an OSINT approach to sustainable investing has two clear advantages over traditional financial reporting. First, machine reading allows every available data point to be captured in the dataset allowing them to be surfaced to provide the source of the information. Traditional auditing only works by analysing a “representative” sample. Second, the ESG ratings agencies rely on issuers (companies) to fill in long questionnaires each year to build their datasets. This self-reported data is partial, inevitably biased and voluntary. To fill the holes, agencies use elaborately-constructed synthetic data. In both cases, a technology solution provides the data trustworthiness and scale required for the CSRD regime.
5. Affordable
Technology solutions will be sought by the 60,000 companies looking at the extra compliance costs. OSINT is an attractive option because, while the data collection and sorting isn’t free [it is not “open source” in the software sense of the phrase], it is considerably cheaper than a manual process. Humans are still required to interpret the output data but the data gathering cost efficiencies are considerable.
6. Avoid conflict of interest
Key to the CSRD regime is the independence of the data. A number of current players have clear conflicts of interest in their business models. The ratings agencies consult to the issuers on how to improve their scores. The auditors consult to the issuers on risks discovered in their audits. By contrast, OSINT is at arms’ length from issuers by design. A reasonable challenge is whether the OSINT data can be gamed by malicious actors, as has been noted in the national security space, not least by the St Petersburg troll farms. If designed well, an OSINT data pipeline will spot unusual activity as a deviation from historic norms. These deviations could be signal but need to be first checked for any mass scale malicious activity. Fortunately, this kind of interference is obvious and can be excluded from the database.
7. Identify greenwashing
A well-designed OSINT data pipeline can identify potential greenwashing from a gap analysis between materiality according to a company and materiality according to the market. Most often, these anomalies are ignorance rather than conspiracy where companies don’t realise what their investors care about most. However, the likelihood is that companies under investor pressure will try to present a stronger sustainability performance than is accurate. Traditional auditing does not have the subject matter expertise to spot these instances – they are most often caught through manual research by sector-expert NGOs. OSINT offers the opportunity to scan for greenwashing risk at scale, and so act as a deterrent to the temptation.
Into 2024
The European Financial Reporting Advisory Group (EFRAG) has recently published the European Sustainability Reporting Standards (ESRS). These standards include general, cross-cutting requirements applicable to all in-scope companies and topical disclosures that may or may not be material to a company. Reporting against ESRS is a huge challenge for companies, their advisors and the investment community. Using an OSINT approach can help meet this imminent challenge.
Mettle Capital is an OSINT data provider for ESG, Reputation, Trust and Thematic models. Established in 2019, its dashboard product went live in October 2023.
Andrew Tucker, Ph.D. is co-Founder and Chief Data Officer of Mettle Capital.
Founder | Veteran | Trustee | Championing AI-Driven Safeguarding | Cybersecurity Thought Leader in Education
1 年Andrew Chattaway MBE