Publicly available, consistent, and comparable asset-level datasets are urgently needed for actors across the financial system. To manage risks and opportunities facing assets and portfolios, as well as positive and negative externalities, we need to know where assets are located, and who owns them, as well as other sector-specific information on how they operate, and their impacts on local and global environments.
This allows us to develop applications relating to climate risk, ESG transparency and verification, inequality, human and ecosystem health, disaster response, and infrastructure development.
As the example set by the Human Genome Project in the 1990s and early 2000 shows, the timely reconciliation of public and private efforts to produce universally trusted, transparent and verifiable datasets is instrumental for reducing duplication, ensuring consistency and dramatically accelerating progress, while also providing the basis for countless potential innovations with both private and social benefits across multiple fields.
Existing asset level datasets, including those produced on a commercial basis, often suffer from inconsistency in measurements and definitions, significant time lags, financial and practical barriers to access, a lack of transparency of verification techniques, and major data gaps. Even more importantly, there are few examples of assets being linked to ownership, and eventually to financial securities, in part due to the difficulty of matching financial and asset-level datasets.
Technological advances made in recent years present opportunities to address these challenges. Rapid progress in distributed cloud computing has dramatically increased the processing capacity available to implement machine learning and artificial intelligence techniques, while geospatial data (including but not limited to Earth Observation, remote sensing, payments systems, and telecommunications) is now available in unprecedented volume and granularity. In combination, these advances have the potential to address the challenges of incomplete or inconsistent data and employ advanced matching techniques to tie multiple datasets together.
One of the key applications of asset-level data relates to environmental change and the transition to a net zero carbon economy. Without significantly better access to asset-level data tied to ownership, there will be very significant challenges in understanding the physical, transition, and liability risks and opportunities faced by individual companies and across investors’ portfolios. Aspects of interest to financial institutions and supervisors include measurement of the impacts of, and social and political responses to, environmental change, migration and conflict, among others.
As the importance of granular data on companies’ underlying assets and investor portfolios has increased, the use of sustainability data has become more sophisticated, but remains concentrated in limited areas of the financial system. As one illustrative example, voluntary disclosure to climate risk, driven by the recommendations of the Task Force on Climate-related Financial Disclosures (TCFD) is gathering pace, but is still limited and does not offer either the granularity or objectivity of asset-level datasets.
Information about physical and non-physical assets tied to ownership information allows analysts to understand who owns what, where, and what the consequences are for individual companies, countries, or other constituencies. Different financial and non-financial actors can use this data for different purposes and in different ways:
Key aspects of a robust asset-level dataset include:
Where incomplete or insufficient, the data can be complemented by peer-reviewed estimation and matching techniques that improve over time.
Iron and steel production and cement production are two of the most carbon-intensive industries, jointly accounting for about 15% of global CO2 emissions. The available proprietary datasets for these sectors cover 70-75% of all assets, with significant gaps in certain regions. These datasets are infrequently updated, and often do not contain the information needed for comprehensive analysis – including exact location and plant capacity.
The project will apply machine learning techniques to earth observation data in conjunction with existing asset-level datasets, to expand coverage and comprehensiveness of existing data. Expected use cases range across the information value chain but will include the use of the expanded dataset to conduct climate risk analysis on assets and the portfolios in which they are held.
If you would like to support the GeoAsset project, have suggestions to guide its work, or would otherwise like to get involved, please complete the contact form below.
The GeoAsset Project is supported by EIT Climate-KIC