Financial institutions face challenges in addressing sustainability from both regulatory and business perspectives. Larger financial corporations operate in many countries and face numerous local regulations in their sustainability reporting and assessments. These regulations relate to physical climate risk, transition climate risk, social impact measurement and the collection and evaluation of environmental, social and governance data.

These regulations also stem from the internalisation of the expected impact of climate change on the businesses and services housed within financial organisations and on the multitude of sectors catered to. Client interest and demand for ESG accountability, products and analysis weigh on financial institutions to demonstrate their sustainability credentials.

All these requirements pose data-related challenges. Collecting reliable and actionable sustainability data has proven to be difficult and, as a result, data collection garners particular attention from regulators.

The European Central Bank has released a publication on climate-related statistical indicators, which provides some guidance in terms of data metrics to use. The Monetary Authority of Singapore’s Project Greenprint seeks to facilitate efficient flows of trusted ESG data to help financial institutions and other businesses in sustainability assessments and investments.

While it is encouraging to see these initiatives, data disclosures suffer from another issue: staleness. These data are often published on an annual basis and can be as much as 18 months old at the time of publication.

Finding data solutions

To tackle these issues head-on, data need to be collected on a more timely basis. Where required, assessment methodologies need to be deployed to lay the foundation for estimate calculations. For example, to evaluate corporate emissions, it’s common to use industry averages as proxies. However, the approach tends to be quite rudimentary as it doesn’t account for the technological developments within the sector.

The Partnership for Carbon Accounting Financials addresses this through a specialist-driven method where emission factors and estimation approaches are based on historical regional data and available technologies. There are also academic studies that propose machine learning models could provide more accurate emission estimation using historical emissions and proxy financial data.

Artificial intelligence has been used to a large extent in climate monitoring. For example, corporates can use sensors to measure emissions in different sites they own and AI helps to optimise and alert any abnormal readings that would trigger timely responses. AI and machine learning models have also been deployed to analyse news sentiment in companies’ ESG performance measurements, using natural language processing techniques.

Another example where data and technology come together is assessing physical climate risk. Geographic attributes of locations and climate data are layered into climate effect models to evaluate financial and operational impact to locations and companies.

NLP can be also used in sustainable bond categorisation by analysing the use of proceeds in the offer document. By leveraging classification language models such as fastText, a deep learning algorithm, labelling sustainable bonds can be automated to a large extent.

Beyond the environment

Besides climate, more recent developments in sustainability have focused on biodiversity impact and the evaluation of the ‘S’ in ESG. For the former, remote sensing technology can be leveraged and layered with geo-location data. Machine learning is then deployed to process satellite imagery to measure the lushness of foliage, availability of infrastructure, nature reserves and change in land use, interacting with proximity of company facility locations. The layered approach can be used to evaluate the biodiversity impact of a single facility, which can then be aggregated to the firm level.

On the social side, assessing diversity, equality and inclusion performance in companies can be done by deploying NLP on worker profile data and employee reviews from websites to highlight areas of concerns and quantify the severity.

In the absence of timely and complete corporate disclosures, we believe that the data innovation assisted by technology in sustainability will evolve as the demand for more frequent assessments and evidencing sustainability performance continues to grow. An emerging example that is yet to be fully explored is the potential benefit generative AI might bring, especially on collecting sustainability information on private companies, as highlighted in our recent report AI in Investment Management - The Pursuit of a Competitive Edge.