Evolving external forces are creating many risks for corporate banks globally. Macro factors like inflationary pressures, pandemics, frequent climate change related events and heightened geopolitical risk have the potential to disrupt supply chains across industries and impact businesses in unexpected ways.
Corporate banks are now witnessing the end of an era. Following the 2008 financial crisis, banks rebuilt capital and invested in technology to strengthen their relationships with clients through improved efficiencies.
The banking industry now faces a new era which will challenge banks’ ability to future- proof their risks and exposures without having any historical evidence or data on how to predict or tackle the risk. Banks can no longer depend on their current Early Warning Systems (EWS) using backward looking indicators with high false positives, to protect them from future risks. Previous methods and data sources for assessing credit risk or identifying distress signals are quickly becoming out of date. Banks have a limited window of opportunity to adapt and respond to the emerging scenarios.
Although large systematic shocks are unpredictable, it is clear that the speed at which banks are able to react to these events can have a huge effect on the ability to actively manage the lending book and reduce any negative impact. The same can be said about the ability to have early warnings of more common “everyday” market movements and events too.
Banks are struggling to generate returns in their corporate lending books. Galytix estimates that ROEs on a standalone basis without the support of investment and transaction banking business are 3-6%.
This is in benign credit markets, where default rates are at near record lows. Taking a through the cycle or recession scenario default rates would reduce these ROEs and, in many cases, turn them negative without management actions. Credit risk costs for banks are likely to rise because of increased loan loss provisioning and a deterioration in credit portfolio quality.
External stakeholders - from investors and regulators to media and activists - increasingly will be seeking reassurances that banks have solid early warning capabilities ensuring robust credit monitoring and risk mitigation across the credit value chain. Credit risk of firms are being impacted by a whole host of new macro factors and stakeholder interests. The sanctions against Russia, the public pressure in the West to divest of any interests in the region, the sanctions against certain Russian individuals and the impact of higher commodity prices have altered the status quo overnight. Businesses that were seen as safe borrowers may look risky. Financial statements alone are no longer enough for risk managers to assess credit worthiness.
Banks that fail to improve their EWS will also face significant regulatory pressures. The European Central Bank (ECB) has highlighted the huge variation in the quality of early warning systems and how credit assessment at a micro as well as macro level is core to risk management and provisioning.
Furthermore, banks are also losing their competitive positioning to bonds and non-bank lenders, in many countries. Galytix estimates that in the US more than three-quarters of corporate financing comes from these sources. This trend is less pronounced in Continental European countries, where bank lending continues to dominate.
Risk aversion alone is not a strategy. Upgrading early warning systems (EWS) are crucial for corporate banks to drive their competitive advantage and improve returns. This needs to be embedded across the credit chain from loan origination to fulfilment to risk monitoring. The strategic case for developing and implementing an EWS is clear: effective risk monitoring will lower both credit losses and capital requirements – directly improving a bank’s ROE by over 20%. Experience shows that an effective EWS could help reduce loan loss provisions by 10-20 % and the required regulatory capital by up to 10 %. Moreover, an effective EWS will also maximise shareholder value by materially reducing the volatility of corporate bank earnings. This will underpin higher stock market valuation multiples.
The experience of both Galytix and PwC is that currently most early warning indicators that banks produce could be meaningless. Given the surge in false positives, several manual data checks are implemented by banks to ensure consistency, effectiveness, and accuracy of signals – which ultimately increases the cost and time of managing the early signal detection process. The ECB has also been critical of the use of adhoc manual triggers being implemented by banks and the need for a more systematic approach to alert monitoring.
For many banks, existing EWS frameworks are based on more easily available and traditionally used data sources including client financials and market data which is readily available. However, such indicators are usually backward looking and fail to predict corporate defaults well enough.
To generate signals, banks still use a traditional univariate modelling approach, but these have rather limited predictive powers. In addition, inflexible legacy data architecture – constrained by static, often hard-coded business rules – prevent banks from rapidly feature engineering their frameworks and adding / testing new relevant data as and when it becomes available.
Many banks lack mechanisms by which to seamlessly integrate non-traditional and unstructured data – data volume has been growing exponentially – into their downstream early warning analytical processes. Often, banks set up structured data warehouses or digital lakes, but these are costly to maintain and cannot handle the rising number of heterogeneous data formats that banks must use for effective early signal detection. As an example, banks often take six or more months to incorporate new and varied sources of data into their core early signal processes.
An effective EWS should identify borrowers at risk of non-performance (High Hit Ratio), distress or default sometime before an actual event (Time before Default). It must enable efficient and reliable assignment of borrowers to different watch-list categories and trigger other actions and escalation scenarios depending on the nature and severity of risk. The system must use indicators that are derived by combining both traditional and non-traditional data sources (internal and external) using a multivariate or decision tree modelling approach.
It must be capable of algorithmically streaming and engineering (continuous discovery, ingestion, transformation, and curation of data) both unstructured and structured data through orchestrated pipes – supervised by humans – into a common data ontology – not extracted ad hoc from a static datastore. Finally, the system should have rules based high integrity versioning and traceability of data processes running through the pipeline as part of scheduled batch processing.
A LEGO (Leverage, External Indicators, Governance and Ontology) framework implemented in an AI-driven pipeline architecture based solution can accelerate EWS quality and effectiveness. This framework involves streaming real-time data and analytics, making adjustments to current indicators, adding a few high impact ones, and providing a systematic and automated capability to manage the EWS without creating unnecessary burdens. These processes allow credit risk managers to quickly assess exposures of counterparties. For instance, it would allow for not only tracking of revenues and investments of Western firms in Russia, but also make sanctions related checks in Western countries through subsidiary filings and court documents. These could tie back to not just Russian companies but also individuals on the sanctions list.
Both PwC and Galytix believe that expanding the list of early warning indicators must be focused on the highest impact external data sources around equity market signals, governance, fraud, aggressive accounting, or cyber risks. There is a lot of interest in sentiment analysis amongst both financial investors and banks. However, we find that these are often too broad creating a large audit trail of false positives that could take risk teams a long time to parse through. There must also be processes in place to ensure these data points are accurate, specific enough and from high quality sources.
No single list of early warning indicators is a magic bullet. There is a need for a systematic front-to-back approach where data governance is prioritised to ensure accuracy of definitions, lineage, and policies. Banks must also be careful not to rely on mutualised data excessively. Loans tend to be relatively illiquid and following the herd doesn’t allow for credible risk management, in enough time. This includes which indicators are used, which data sources are seen as credible, what specific information or financial data percentage changes are seen as a threshold and how these are all combined in a fluid and evolving data ecosystem.
An EWS that generates real early warning indicators compared to late-emerging information such as credit ratings deterioration offers the potential to identify distress signals three to five months before an event. This enables banks to take risk-mitigating actions earlier and more effectively.
