COVID-19 exposes flaws in bank lending models

Posted on December 9, 2020

Bank loan

Sean Hunter

By Sean Hunter, CIO, OakNorth

When it comes to commercial lending, banks rely on risk models to make decisions.

These models have been built up internally over decades of lending across thousands, if not tens of thousands of loans, but COVID-19 has exposed unexpected flaws in them.

The first issue is that these models are based on historic data which doesn’t adequately reflect the unique situation we now find ourselves in or take into account the future challenges that the world will be facing as it enters into the worst recession in three centuries.

The second issue is that they make broad assumptions about entire sectors rather than developing an understanding of the portfolio at the granular loan level and taking into account the individuality of each business.

Look at the retail clothing sector, for example: a luxury boutique which specialises in made-to-measure gowns will likely see its revenues obliterated as occasions where a gown would be needed – such as weddings and red-carpet events – will have been cancelled or postponed.

Contrast that with the experience of an eCommerce business that specialises in yoga wear, which may see a substantial increase as customers order more online and take up yoga under lockdown.

Making assumptions about entire sectors ignores the diverse business models and cash flows of individual sub-sectors, which are all impacted differently by lockdown.

The third and final issue is that these models don’t take into account how quickly the situation is changing.

Take a residential property developer for example – during the first UK lockdown in March, construction sites were forced to close which means entire projects will have been delayed. If you look at the second lockdown, it was a very different picture as construction sites – along with schools and universities – were allowed to stay open.

So even making credit assumptions based on what we saw in the previous lockdown doesn’t work.

The only way to course-correct for “the new normal” is to take a fundamentally different approach to commercial lending than what’s been done for decades.

In the future, banks are going to have to combine backward-looking data with a forward-look view, as well as take a granular, loan-by-loan approach rather than an overall portfolio or sector-level approach to credit analysis.

They are also going to have to conduct reviews on an ongoing basis, rather than annually; update parameters to reflect the ever-changing situation; and use alternative data such as foot traffic to inform their models.

In order to manage credit risk within our own UK bank, OakNorth, and support our various bank partners around the world, we combine dynamic data sets, auto-analysis capabilities, cloud-computing and state of the art machine learning to undertake portfolio diagnostics.

This enables us to rate loans from 1-5 based on their vulnerability to the new economic environment, with 1 being least vulnerable, and 5 being most vulnerable. The ratings are based on multiple factors including liquidity, debt capacity, funding gap and profitability, and can be dynamically customised to reflect the lender’s credit risk criteria and appetite.

There are several components to this:

  • Instant Credit Analysis enables faster decision-making and consistent analysis across new credits, periodic reviews of existing cases, or detailed re-underwrites. For new potential borrowers OakNorth Credit Science Suite analyses the business’ financial data, as well as dynamic data sets for sectors, geographies and macroeconomic trends, instantly delivering basic credit analysis on that business
  • Instant Financial Analysis provides standardised models for each of the sub-sectors in our taxonomy, pre-configured stress-scenarios, automated vulnerability scoring and presentation of capital structure
  • Real-time Sector Insights allows you to see industry insights that drive borrower performance tailored for over 100 sub-sectors
  • Automated Peer Comparison uses machine learning to surface up to 20 peer companies to consider in your credit analysis, allowing you to make comparisons with the borrower
  • Continuous Monitoring of Active Credits helps you turn monitoring into a real time process and lets you focus on relationships not admin. The platform monitors billions of data points to detect anomalies in your loan book. This gives relationship managers all the information they need and more for check-ins or annual reviews with their customers
  • Portfolio Diagnostics generates a forward-looking view with sub-ratings on Liquidity, Debt Capacity and Business Profitability using scenarios specific to the sector and country of the borrower

Podcasts

View All

Insights

View All