By Julia Chong

Banks are no stranger to risk, but some risks are more equal than others.

In Banking Insight’s July/August issue, our article, Simulating Mega Risks of Tomorrow, covered much of the response to the pandemic, including adjustments and alternatives which banks must consider in their risk modelling as historical analysis breaks down under the weight of scrutiny.

We continue our focus on credit risk, this time on banks’ obligations and experimentations in improving the quality and efficiency of credit risk management.

But first, some necessary updates on the regulatory front.

Goal Post in Sight

Referred to by the market as ‘Basel IV CVA’, on 8 July 2020, the Basel Committee on Banking Supervision (BCBS) issued its final revisions to the credit valuation adjustment (CVA) risk framework, which is applicable to all derivatives (except those cleared through a qualifying central counterparty) and the fair value of securities financing transactions (SFTs). Its policy paper, Targeted Revisions to the Credit Valuation Adjustment Risk Framework, is based on consultations with and feedback from market participants in November 2019 and the proposed amendments align parts of the CVA risk framework with the final market risk standard and capital requirements for bank exposures to central counterparties.

This latest and final round of credit risk adjustments by the supervisory body is aimed at relieving banks’ operational burden, incentivising central clearing, and promoting consistency in implementation. With this, the goal post is now in sight. Postponing the deadline for implementation by 12 months, the BCBS has now confirmed go-live on 1 January 2023.

The targeted revisions are summarised below:

• Reduced risk weights. This applies to the standardised CVA (SA-CVA) and basic CVA (BA-CVA) calculation approaches. Bear in mind that under CVA, banks must use the basic approach unless they receive approval from their relevant supervisory authority to use the alternative standard approach.

• Introduction of new index buckets where banks could, under certain conditions, calculate capital requirements using credit and equity indices directly instead of looking to the underlying constituents. The Committee has agreed to introduce the same new buckets in the:

  • counterparty credit spread risk class;
  • reference credit spread risk class; and
  • equity risk class of the SA-CVA.

• Revised formula for aggregation of capital requirements across buckets in the CVA risk framework for better alignment to the market risk framework.

• Alterations to the scope of CVA risk capital requirements. Measures include preferential treatment by exempting certain client-cleared derivatives for client exposures and reducing the floor for the margin period of risk for some centrally cleared client derivatives. This brings the CVA more in line with the counterparty credit risk framework.

• Overall recalibration of the CVA risk framework. A reduced aggregate multipliers under SA-CVA and a scalar for banks using BA-CVA. The effect is a recalibration of capital requirements for banks.

Risk Cover at Ground Zero

With over nine months of the pandemic behind us, the report card for the financial sector is overall positive. A plus point has been banks’ rapid response in speeding up credit decisioning to ensure uninterrupted delivery of essential financial services. Moving forward, what’s required is now to test out new approaches to more accurately evaluate or predict creditworthiness (or riskiness) of borrowers.

For this purpose, it helps to have an overview of the current landscape. McKinsey & Co’s July 2020 report, Managing and Monitoring Credit Risk After the Covid-19 Pandemic, found five major changes to the credit-risk environment for which financial institutions must anticipate in order to minimise further shocks to the ecosystem. We summarise the findings below:

+ Changes in creditworthiness at the sector and subsector level.

Leading financial institutions are beginning to approach underwriting and monitoring with a new configuration of sector analysis, borrower resilience, and high-frequency analytics. A key trend is that leaders are moving relatively quickly from a sector view to a subsector view and finally an obligor view, using real-time data and analytics, which then supports decision-making.

Most banks have developed refined hypotheses about specific subsectors and are approaching (or have already arrived at) an obligor view of risk assessment. The analysis of sectors and subsectors translates into a probability-of-default (PD) shock. One UK bank quantitatively analysed the PD change for each sector by stress-testing the profit and loss of the counterparties on the basis of the expected shock and recovery trajectories for each sector, reassessing the debt repayment ability accordingly. The results proved that the PD shock can vary three or four times in magnitude.

+ Hard to differentiate between borrowers in the same sector or subsector.

The distinctly different profiles recognised by banks within subsectors depend on varying demand patterns, supply chain factors, and market organisation. Much attention has focused on reopening the economy but banks and businesses should also think about horizons: different regions and countries are at different stages of the pandemic and thus reopening at different speeds. Economies that are now mostly open are experiencing trade and supply-chain distortions from lagging former partner economies.

Lenders will need to think through these eventualities and codify perspectives in their analyses. The public-health dimensions of the present crisis led one US bank to develop composite risk scores at the intersection of geography and industry sector, which helped the bank differentiate more clearly among borrowers.

+ Pertinent data on crisis conditions are scarce, lagging, and not fed automatically into decision-making.

Beyond this horizon are approaches using real-time business data in decision-making and advanced analytics to review credit underwriting processes. The transition to these new methods will help banks cope with the present crisis but also serve as a rehearsal for the step change that, in our view, credit risk management will have to make in the coming months and years.

+ Socially responsible collections needed to meet changing customer preferences.

Banks must shift to a customer-assistance interaction model and make it a priority in their digital transformation.

+ A large wave of non-performing exposures is beginning and must be addressed in new ways.

Financial resilience will be determined less by pre-Covid-19 profitability than by indebtedness and liquidity, attributes that will establish a borrower’s ability to weather the crisis. Operational flexibility, including the soundness and adaptability of a business model in the new environment, is determined by the cost base and the possibility that it can shrink in line with demand. These factors can be evaluated through transaction data: current account inflows, credit line utilisation, and the evolution of point-of-sale transactions.

The management consultancy advises: “The best banks will keep and expand these practices even after the crisis, to manage credit risk more effectively while better serving clients and helping them return to growth more quickly.” The differentiator will be how quick banks can incorporate new and emerging data into its risk evaluation framework. This requires a shift from old style ‘forecasting’ to dynamic ‘nowcasting’.

Nowcasting

The need to populate credit risk models with current data is acute in this unprecedented crisis as pure reliance on historical data is certainly misleading. This envisions a fundamental and immediate shift from what banks are accustomed to in their credit risk, a move from ‘forecasting’ to ‘nowcasting’.

The term, which has its origins in meteorological science, describes the economic technique for very short-range forecasting or prediction of the present. Author Alexander Ineichen, in his paper titled Nowcasting and Financial Wizardry, defines nowcasting as “the economic discipline of determining a trend or a trend reversal objectively in real time. Nowcasting is fact-based, focuses on the known and knowable, and therefore avoids forecasting. [It] is the basis of a robust decision-making process.

“A ‘nowcaster’ does not try to predict the future but focuses what is known today, i.e. know now in real-time. Forecasts are an integral part of orthodox asset allocation and are essentially guesswork. In other words, guessing is an integral part of how assets are allocated and risk is taken. There is an alternative; a focus on facts rather than forecasts.”

To correctly apply nowcasting, non-traditional or alternative sources of data (‘alt data’) must be sourced and used by financial institutions to innovate and remodel within a period of weeks instead of the customary months or years.

What is alt data and how does it enhance credit risk modelling? Traditional data is broadly understood as information assembled and managed by official sources such as credit reporting or rating agencies, government sources, and internal reports. Information originating outside of this sphere is considered to be alt data.

But do not be misled. Incorporating alternative sources of information in credit risk models isn’t a radical, new idea.

For more than a decade, international lenders and regulators have advocated that banks use alt data to serve the ‘unbanked’ or ‘underbanked’ in our midst. The World Bank, the International Monetary Fund, and national regulators have encouraged banks to refine their credit assessments and scoring methodology to ensure the uninterrupted provision of credit to the ‘unbanked’ and ‘underbanked’, i.e. people and businesses who have no history with credit reporting companies (they’re called ‘credit invisibles’) or carry ‘thin’ credit files (known as ‘credit unscorables’, i.e. individuals with less than three sources of payment information or trade lines).

As some of today’s corporate behemoths – airlines, tourism, luxury retail – devolve to similar standing as ‘undesirables’ in our economy and are faced with the real possibility of being labelled ‘unbankable’, typical risk models using conventional data sources have become obsolete.

The ability to speedily determine creditworthiness, PD, and expected credit losses – reasons for the growing adoption of nowcasting – is contingent upon access to and incorporation of alt data such as:

> Resilience scores: In addition to conventional credit scoring/rating/ranking – which impacts a loan’s structure, drawdown conditions, interest rate, tenure, margin of financing – lenders are today also relying on resilience scores. This is an index that ranks the ability to withstand economic disruption like Covid-19 and reflects tightening credit standards. A lower resilience score is better and is achieved by evaluating using longer-term metrics than credit scoring or rating, such as less frequent credit inquiries or applications, more experience managing credit, lower total revolving balances, and fewer active accounts.

> Behavioural attributes: For starters, look out for data beyond account balances and credit scores/ratings. Creative new sources of data include timeliness of bill payments (rent, utilities); volume of e-money transfers; smartphone activity; number of packages shipped a month; online reviews on Yelp, TripAdvisor; and social media feeds. Remediation clients will hopefully eventually return to the black soon, but in the meantime, lending decisions based on alt data can aid market liquidity whilst encouraging clients to cultivate good credit habits, which lead to more favourable loan terms down the road. Note that accumulation or aggregation of such data in the risk model must adhere strictly to privacy law such as General Data Protection Regulation and equivalents.

> Machine learning predictive models. These embed multiple forms of analytics into the credit evaluation process and automate the decision logic to deliver near-real-time, highly relevant assessments, especially in high-volume environments. Algorithms and technologies such as distributed ledgers sift through high-frequency data to uncover anomalies and flag potential risk of money laundering. Furthermore, investing in a self-learning model can take fraud detection to the next level by minimising the number of ‘false positives’. Additionally, new solutions could also offer integration with other areas of risk and finance, such as capital calculation and IFRS 9, creating more integrated and seamless reporting.

Ultimately, the best banks will leverage on the lessons endured throughout this crisis and constantly expand and reinvent practices. This requires courage as well as creativity. Those which do so will reap the rewards of cheaper credit acquisitions, better risk management, happier customers, and faster growth trajectories.

Julia Chong is a Singapore-based writer with Akasaa. She specialises in compliance and risk management issues in finance.