Under the Basel III framework, stress tests are forward-looking exercises that aim to evaluate the impact of severe but plausible adverse scenarios on the resilience of financial institutions. Simply put, stress testing helps determine whether banks can sustain themselves and continue to lend, even in a recession. Although the practice first emerged in the 1990s, it took the devastation caused by the Global Financial Crisis (GFC) of 2008-09 for the world to fully appreciate the importance of the practice. Since then, every major financial regulator has advocated for embedding stress testing within every firm’s risk management framework. However, the ongoing challenge of COVID-19 has left the industry questioning the very assumptions it once held.
This paper explains how stress testing has evolved over time as a risk management practice and how COVID-19 stands to further challenge the status quo. We then delve deeper into what lenders can do to adapt to the situation, proposing a three-part approach: a generic stress testing framework, where the breadth of the exercise has been captured with details about each constituent step; the specific industry best practices which, if incorporated, can transform a generic stress testing suite into a best-in-class practice; and finally COVID-19 specific enhancements which can close all the gaps in the exercise left exposed by the recent pandemic.
What Happened in the Past?
Stress testing is not new; it was first used in a systematic way and with a financial sector-wide perspective by the International Monetary Fund (IMF) and the World Bank in 1999. It became very important as a practice for banks and financial institutions in the aftermath of Global Financial Crisis (GFC) of 2008-09, when many incurred heavy losses and others did not even survive.
Post-GFC, central banks, regulators and policy makers around the world implemented significant measures to push for tighter regulations and enforced stress testing practices to better prepare banks around the world for the next recession and future economic shocks. The Basel Committee on Banking Supervision (BCBS) also emphasized the importance of stress testing in 2009, opening up discussions on how it should be embedded in banking’s governance and risk management culture. Similarly, the US regulator Federal Reserve Board took strict steps after the GFC to safeguard U.S. financial institutions from recessions. It mandated yearly tests like the Dodd-Frank Act Stress Test (DFAST) and the Comprehensive Capital Analysis and Review (CCAR) for major authorized deposit-taking institutions (ADIs), depending on the amount of assets held by the financial institution. In Australia, on the other hand, the ramp up in stress testing requirements has been relatively slow. Australia was much less impacted by the GFC and has not been in a technical recession in the last 30 years. However, after a qualitative assessment of the stress testing capabilities of 28 ADIs in 2018-19, APRA is planning to transition to annual stress testing of large institutions in 2020.
How has Covid-19 Changed the Equation?
In the wake of the COVID-19 pandemic, another global economic crisis is upon us and while its impact is yet to be fully understood, indications suggest that it is more severe than its predecessors. There has already been a significant loss of both lives and livelihoods. In Australia, the unemployment rate jumped to 6.2% in April 2020 and then rose further to 7.1% in May 2020. In addition, the underemployment rate reached an unprecedented 13.7% in April 2020 before dropping marginally to 13.1% in May 2020.
This extraordinary loss of income and jobs will diminish the ability of households and firms to meet their financial obligations. The resulting wave of bankruptcies will have a lingering effect on aggregate demand while testing the resilience of financial systems everywhere.
No past stress test or financial supervision could foreshadow a shock of this magnitude. For instance, the ‘severely adverse’ scenario of the Federal Reserve’s 2020 stress test assumes that the unemployment rate rises from its February 2020 level of 3.5% to a peak of 10% six weeks later. However, the unemployment rate in U.S. had already breached the stress test severely adverse threshold of 10% to 14.7% in April 2020 confirming the economic pain due to COVID-19. But, we continue to see volatility in the U.S. unemployment rate (having dropped to 13.3% in May 2020 and then to 11.1% in June 2020)
With the number of new cases still increasing globally, no definitive breakthrough in vaccine development and growing geopolitical risks, existing scenario analysis capabilities must undergo change. The pandemic has exposed the gaps in the existing practice, especially with respect to the severity of scenarios to be considered. Another major gap highlighted by the pandemic is the need for greater flexibility, given the growing uncertainties about how the post-pandemic world will look.
What Can Lenders Do?
Given the unprecedented nature of COVID-19’s economic impact, lenders should evolve their risk management practices now. This section considers how lenders can effectively respond to the challenges they now face through a three-step solution which starts by building a robust stress testing framework, before implementing specific industry best practices and finally incorporating COVID-19 specific enhancements.
Building a Robust Framework
The COVID-19 crisis underlines the importance of a repeatable and robust stress testing framework. The institutionalization of the exercise is essential for early warning detection and proactive risk management. Figure 2 (below) presents EXL’s comprehensive stress testing framework, which draws on EXL’s rich experience in this domain and regulatory guidance on the subject.
Stage 1: Planning and Review
Stress testing exercises should ideally start with a thorough review of the existing practice and comparison with industry best practices. It is at this stage that financial institutions often weigh the two competing approaches - ‘bottom-up’ or ‘top-down’ and select the one which suits their requirements and capabilities. The exercise must be kept in sync with all regulatory updates, along with other material developments in evolving risks. We recommend that a periodic review of the exercise framework is undertaken regularly.
Stage 2: Stress Scenario Generation
In this step, lending institutions hypothesize a variety of possible scenarios with differing levels of severity. Each scenario is composed of important macro-economic drivers and the severity of each scenario is defined by stressing the level taken by these drivers. Regulators also provide guidance for possible stress scenarios and generally recommend defining three types of scenarios:
- Baseline: a normal economic scenario
- Adverse/downturn: in between normal economic scenario and recession
- Severely Adverse/ prolonged downturn: a global recession scenario
Stage 3: Model Development
This step connects the changes in macro-economic environment to the impact on losses experienced by the lending institution. This is achieved through statistical models which relate the impact on macro and micro drivers on the three loss parameters: PD, EAD and LGD. While banks generally have a suite of these models in place, they need to be calibrated if they are to be used in a stress testing exercise. Calibration generally includes changing Point-in-Time estimation to Through-the- Cycle estimation, which aims to avoid pro-cyclicality in calculations. Another important calibration is the introduction of important macro-economic drivers in the statistical models. While different models are used in the coverage of different risks and asset classes, a similar approach should also be followed in their calibration. The model building exercise is highly dependent on whether the bank chooses to adopt a bottom-up or a top-down approach. The figure below summarizes EXL’s strong capabilities in credit risk modelling and demonstrates how they are used for stress testing purposes.
Stage 4: Stressed Capital Calculation
This step uses the hypothesized scenarios to feed into the developed models for the quantitative estimation of the risk that different asset classes carry. This is represented numerically in the form of risk-weighted assets, which dictates the amount of capital an institution should hold to cover stressed losses. The final output in this step is the estimation of capital adequacy and liquidity ratios under stressed conditions. Weights are also added to balance the impact of scenarios of differing severity.
Stage 5: Management Actions & Documentation
The final step in the framework is thorough documentation of the methodology along with the results of the exercise. Results are internally audited and submitted to management for review and are then used to guide business decisions that may have a significant impact on capital. Regulators call for thorough governance of the entire exercise.
Adopting Industry Best Practices
While the previous section covers the stress testing process end to end, it is worth delving deeper into some of the best practices followed in the industry. Incorporating these practices can assist lenders in building a robust stress testing suite.
1. Robust Data Management
Robust data management is the basic requirement of an effective stress testing exercise. Maintaining correct data and storing it efficiently are important aspects of conducting reliable and repeatable stress tests. It is also important to capture data with sufficient history, so that the statistical models trained on rich data can achieve superior performance. There should be enough history available to cover information from multiple business cycles, especially up to the recessionary phase of the Global Financial Crisis. Another important dimension of sound data management is the ability to capture and maintain data feeds from sources external to the organization, including data from credit bureaus and macro-economic statistical repositories. The figure below depicts how a good data management approach can support the creation of better stress testing models.
2. Selection of Macro-Economic Indicators
Macro-economic indicators are central to the stress testing exercise and should be selected judiciously. Best practice is to start with an extensive set of indicators, including information like housing price index, unemployment rate, GDP growth rate, price indices, interest rate, dwelling approvals, credit growth etc, all of which should come from a reliable federal source which guarantees the information’s integrity and ensures replication. A crucial step in the selection process is checking the sensitivity of these indicators with respect to defined periods of economic stress. This ensures that only those indicators which are significantly impacted during an economic downturn are selected. Finally, qualitative factors like product concentration, industry-wide exposure or country-specific risks should also be considered in the selection process. For instance, major Australian banks have high concentrations in mortgage lending, which highlights the importance of the housing price index as a must-have indicator. The unemployment rate is another important driver, which is widely recommended as an essential indicator.
3. Exhaustive Scenario Generation
Since this step is the soul of the stress testing exercise, getting it right should be every lender’s top priority. A perfect balance of scenarios can determine whether or not banks are keeping excessive capital reserves, which restricts their ability to offer loans or fewer reserves, which put banks at risk in case of recession. To achieve this balance, different classes of scenarios, based on their severity, are recommended. Since APRA allows for flexibility, lenders are free to customize their approach and submit it for regulatory approval.
Careful selection of macro-economic drivers also lays the foundation for well calibrated scenario generation. Apart from applying regulatory guidance in this step, it is also important to incorporate the lessons learned from previous economic shocks—not only global recessions, but also individual firm-level shocks. In addition to the retrospective elements of the process, scenario generation should be forward-looking at the same time by considering emerging risks such as climate change and global pandemics, which have the potential for unprecedented consequences. Well-defined scenarios allow credit risk models to estimate losses more accurately and assess subsequent impact on capital adequacy.
4. Ending the Top-Down vs Bottom-Up Debate
Lenders generally follow one of the two approaches for stressed capital calculation. However, lenders should ideally perform calculations using both approaches. This will help them reap the benefits of the bottom-up approach, which is more granular and makes use of held data. At the same time, a top-down approach will enable lenders to look at their business from the regulator’s viewpoint and prepare well for regulatory assessments. Additionally, comparing the results across the two calculations will generate more insights.
5. Lessons from GFC
The global financial crisis provided a much-needed reality check to the risk management function. It exposed how the worst-case scenarios that banks were preparing for were not severe enough. The other major revelation was the prevalence of using short-term funds to invest in longer-term assets. This maturity mismatch squeezed liquidity for lenders during the recession, which is why Basel III brought liquidity ratios in the form of LCR (Liquidity Coverage Ratio) and NSFR (Net Stable Funding Ratio) in the fold. Another GFC lesson learned is the risk related to counterparties, which was not well understood before. Therefore, stress tests should also account for the impact on liquidity and counterparty credit risk in different scenarios.
Recommended COVID-19 Specific Enhancements
COVID-19 is already turning out to be a bigger crisis than the GFC itself, given the unprecedented losses in both public health and the economy. Any stress test conducted today is therefore incomplete without COVID-19 specific enhancements. This section explores some of these.
1. Top Down Analysis for Estimating Scenario Impact
Financial institutions are equally wary of the uncertainties associated with the pandemic’s impact. No one is certain how long it will take for life to return to normal. For this reason, it will be useful for lenders to build a Loss Simulator which can analyze the impact on capital requirements due to changes in macroeconomic conditions. Gamifying the process will allow risk managers to play around with scenarios that are forward-looking in nature. This tool can be designed to incorporate both bottom-up and top-down calculations and allows for the desired level of flexibility with a range of parameterization. It can also be used to back calculate the hurdle level of parameters with a target level of capitalization.
2. Consider a Prolonged Downturn
Australia went under lockdown in late March 2020 and within a month of restrictions, the spread of the pandemic was greatly curtailed. While Australia has, to date, better restricted the spread of COVID-19, the fear of a new wave of cases as restrictions are eventually lifted is very real. With so many unknowns and no vaccine available as yet, scenario development should capture the eventuality of a much-prolonged downturn.
3. New Normal and Industry-Specific Risks
Life post-pandemic is not going to be the same. Industries like aviation, public transportation and entertainment will suffer long-lasting impacts, yet there are some sectors that have managed to ride the wave. Stress testing should therefore incorporate the asymmetric economic impact of COVID-19 and make appropriate assumptions around how life would evolve, post-pandemic.
4. Assessment of Relief Packages
Federal government has come up with many measures to help re-start the economic engine. This includes the Jobkeeper program for employment protection, an SME loan guarantee scheme, payment holidays for distressed borrowers and a home builder scheme to stimulate construction and the housing market. At the same time, Reserve Bank of Australia has reduced interest rates to an all-time low, allowing for more lending. These positive aspects must also balance the scenarios.
5. Emerging Geopolitical Risks
The pandemic has also impacted geopolitical dynamics. For Australia, growing disputes with China are likely to turn into a trade dispute. Since China is Australia’s biggest trade partner, the dispute could have significant economic ramifications. Such developments should also be incorporated in the scenario development process.
While the hope is that COVID-19 will soon abate, the pandemic has highlighted that nothing is certain, and anything is possible. The big lesson for the banking industry, even as it makes every effort to ride out the current turmoil, is to prepare well and to take nothing for granted. Stress testing is now more important than ever—but this too must adapt for the new environment and quickly integrate the lessons being learned in real time.
- https://en.wikipedia.org/wiki/Basel_Committee_on_ Banking_Supervision
- https://www.garp.org/newmedia/gri/climate-riskmanagement- guide/Challenges_052919_PDF.pdf
- https://www.federalreserve.gov/publications/2019- february-supervisory-scenarios.htm
- https://www.oliverwyman.com/content/dam/oliverwyman/ global/en/2016/may/Australian-Stress-Testing.pdf
- https://www.moodysanalytics.com/regulatory-news/feb- 21-20-apra-to-transition-to-annual-stress-testing-of-largebanks- in-2020
Analytics Head, Australia & APAC
Amit Kumar Jain
Assistant Vice President, EXL Analytics
Senior Consultant, EXL Analytics