FRM Corner

Putting Stress Tests in the Blender

Amid the pandemic, the Fed recently ordered large banks to perform a second round of stress tests, leading to questions about whether CCAR should be applied more broadly and whether a hybrid approach to capital planning is needed. Given the "new normal," what tests are most appropriate for large and small banks, and how frequently should they run?

Friday, October 23, 2020

By Alla Gil


In September, to further test bank balance sheets for stress, the Federal Reserve took the unprecedented step of releasing a second set of 2020 scenarios for its Comprehensive Capital Analysis and Review (CCAR) assessment. Given that COVID-19 is still raging, this is totally justified. At the same time, it suggests a stress test's shelf-life is far shorter than previously thought, and lends further support to the idea of “always-on” stress testing.

CCAR, the Fed's benchmark for testing bank balance sheets for stress, is mandatory for the largest banks in the U.S. But what about the smaller ones? Should they be compelled to follow CCAR practice in full or in part? Alternatively, should they rely on a simplified economic capital approach that extends through the cycle?

To choose a strategy that best suits your firm's business objectives, one must first understand the structure of stress tests.

Alla Gil headshot
Alla Gil

CCAR consists of three main components: two or three scenarios (the baseline, and one or two stress scenarios); global market shock (GMS), which is applied to the largest trading operations; and largest counterparty default (LCPD), which is only relevant for institutions with substantial processing or custodial operations. In addition to CCAR, many institutions (especially those with international operations) still calculate economic capital.

The three components of CCAR are often treated separately, even by different departments - though they all are connected. There are benefits to carrying out a thorough risk analysis using all these methods (as well as economic capital), but there are also costs. For smaller institutions, a risk/reward balance must be struck.

To put it in human terms, it's the same kind of trade-off we make when we decide to drive to work versus taking public transportation. Driving is not risk-free, but the benefits (especially during COVID-19) outweigh the physical risks.

Likewise, longer-term stress scenarios can be compared to a regular exercise routine that improves your health. GMS is a “heart-attack” style stress test - it doesn't make you healthier, but if you don't pass, you'll need surgery. LCPD, on the other hand, represents a rare (but high impact) catastrophic event, like your house burning down.

Economic capital, meanwhile, is like a general health check: derived correctly (capturing true stress risk with dynamic correlations, over a longer-term horizon), it represents a real measure of risk that should be used in any calculation of risk-adjusted return.

While the use of each of these stress testing components has its own costs and benefits, they can be consistently connected within one comprehensive framework, with scenarios at their core. They must go through the cycle (a three-year horizon at minimum), incorporate shock events, and be severe, yet plausible.

Factoring in Shocks

Shocks should include not just geopolitical events and natural disasters, but company-specific scenarios too - things like cyberattacks, large counterparty defaults or a sudden change in the business environment. To measure the severity of such events and their impact, the marginal distribution of each variable within the scenario must be wide enough to include potential instantaneous shocks.

The GMS component can be used to verify the calibration for each variable. If you look at the distribution of every variable using standard historical variance and overlaying potential shocks, you see that - for many variables - the GMS impact falls within 2-3% of the worst-case tail. This means GMS impacts are 'severe but plausible' on the margin; they just don't happen together.

GMS assumes all these shocks happen at the same time, which means a correlation of 100%. Correlations do increase tremendously in crises, but only to the level of 80% -90%.

It is critically important to estimate correlations realistically for calculations of capital and liquidity needs. Underestimating the power of dynamic correlation can have dire consequences for the firm. But overestimating it can prevent institutions from finding the best use for their resources.

Capital Planning: Benefits of a Hybrid Approach

What is the optimal approach to capital planning for organizations with the luxury of selecting a strategy? The short answer is to take the best of each component to create a blended approach customized for your institution. This type of hybrid approach entails three steps.

First, generate internal scenarios. To be ready for any plausible outcome, make sure they are more severe than the Fed's.

Second, ensure that your scenarios are severe enough to incorporate shocks that are close to the GMS impacts. With consistency across variables (increased but still plausible correlations), assess the impact on your key performance indicators, measuring capital, liquidity and profitability.

Third, measure economic capital using the distribution directly from the scenarios that go back to the traditional definition of economic capital: the difference between expected and worst-case (with a certain probability) outcomes.

This will achieve the following important goals:

  • Distributions across the expected outcomes (used by financial planning) will be consistent: the worst-case outcomes will be measured by stress testing, and the true risk measure (economic capital) will be determined.

  • Mitigating actions and contingency plans can be overlaid, and their impacts can be seen simultaneously.

  • A culture of transparent communication - between senior management, board, financial, risk and business units, regulators, auditors, and investors - can be established.

Evolution of Stress Testing

Over the past 30 years, risk management frameworks and associated stress tests have gone through three major changes.

Value-at-risk (VAR) was the predominant methodology in the 1990s. It is a short-term, forward-looking risk measure (usually 10 days), calculated daily on all trading positions, with additional stress tests applied to each variable at a particular point in time.

In 1998, Basel II shifted the focus to economic capital; a longer-term (one year) horizon; and the integration of variables that drive market, credit and operational risk. Stress tests were a part (Pillar II) of this framework, but it was a point-in-time requirement only.

When the global financial crisis (GFC) struck, banks were still working on implementing Basel II. Would it have made a difference if the accord had been in full effect? Would the financial industry have been saved from the GFC - or suffered less - if economic capital had already been fully implemented?

I don't think so. This approach still had too short a time horizon: after five years of relatively low credit spreads, economic capital was at the lowest level right before the crisis - a potentially fatal shortcoming.

In the aftermath of the GFC, CCAR - with its longer-term scenario analysis - emerged. Intended to mitigate the deficiencies of Basel II, CCAR eventually became a universal benchmark. Requirements changed over the course of the following decade, but its main purpose - ensuring that a bank performs a long-enough scenario analysis to take it through the economic cycle - remains.

Those that applied Fed stress scenarios (whether they were obliged to or not) were better prepared for the current pandemic - even though these scenarios were released before the outbreak.

Parting Thoughts

As a complement to the formal CCAR process (on a less detailed level), the largest banks can and should employ a blended approach to capital planning. Given its affordability and flexibility, this hybrid strategy can benefit everyone - even the smallest community bank.

However, it must be performed more than twice a year. Indeed, it can be used as frequently as quarterly or even monthly, and surely must be applied every time there are major new shifts in the business, market or geopolitical environments.

Alla Gil is co-founder and CEO of Straterix, which provides unique scenario tools for strategic planning and risk management. Prior to forming Straterix, Gil was the global head of Strategic Advisory at Goldman Sachs, Citigroup and Nomura, where she advised financial institutions and corporations on stress testing, economic capital, ALM, long-term risk projections and optimal capital allocation.


BylawsCode of ConductPrivacy NoticeTerms of Use © 2023 Global Association of Risk Professionals