Reflections on stress testing

Reflections on stress testing

By Tally Ferguson, SVP, Director of Market Risk Management at the Bank of Oklahoma.

Ahead of the Stress Testing USA: CCAR & DFAST Congress 2017, Tally shared his insight with us on stress testing.

Stress testing rose in profile after the Dodd Frank Act required the Federal Reserve to impose stress testing on banks over $10 billion. The resulting lucrative markets for consultants and financial service companies developed reasonably advanced stress testing techniques. It also drew a significant amount of high quality, well trained quants to the validation space. Sadly, banks had to devote almost all this talent and product to projecting capital and pre-tax, pre-provision income from regulatory prescribed economic scenarios. Sceptics suggest this merely served to calibrate the Federal Reserve’s capital-intensive model and to train technical writers. At this point, you might expect me to go into a diatribe about how a significant amount of human, systems and financial resources was wasted in a narrow replication exercise that did no more than keep excess capital in the system. Shouldn’t these resources have been doing something useful like building rockets that can go to Mars or replicating the smell of honeysuckle in a lab?

Well, tempting though that is, Risk Insights is not a political publication, and, in fact, I have a friendlier interpretation of what regulatory-forced stress testing wrought. I maintain that, like Kennedy’s bully pulpit got us to the moon, Dodd Frank’s insistence on stress testing brought practical resources and techniques to the financial industry. With these, we can do something useful.

Consider what financial institutions built because of regulatory imposed stress tests. We created conceptually sound models that predict useful things like credit performance, funding needs, and fee income based on economic indicators. Regulatory imposed stress testing forced many financial institutions to integrate budgeting, accounting, risk and credit systems. Also, this mandate reinvigorated financial institutions’ boards’ interest in where weaknesses are in their portfolios of business and where these weaknesses would show up. I elaborate on these three below.

Conceptually sound predictive models. CCAR and DFAST tests stress pre-tax pre-provision income and hence capital under three prescribed economic scenarios. All financial institutions must run the same economic scenarios. CCAR banks must do so twice a year. This makes sense from a regulatory perspective because they can compare financial institutions to one another. The problem with this is that the scenarios apply to national statistics, and banks have to defend the use of regional differences. Beyond that, the scenarios reflect macro situations that interest the Federal Reserve – inflation, stagflation, recession, credit deterioration. These might not reflect worst cases for reporting financial institutions.

Nevertheless, we now can take economic scenarios and predict with sound mathematical theory how such events impact core elements of financial institutions health; revenue growth, capital adequacy, asset quality and funding needs. We can use these skills to produce idiosyncratic stress tests. An informative idiosyncratic stress test finds the regional economic scenario to which the institution is most exposed. There are quantitative methods for finding this scenario. Use the coefficients² of your regressions and apply differential equations to solve for the level of independent variable that causes most harm. Alternatively, if your quant team is more horsepower oriented and less math-phyllic, simulation provides a brute force, but equally effective solution.

I should note that mathematical equations and computer simulations tend to have less experience in the financial markets than seasoned executive and senior managers. Equations and simulations could give an unrealistic worst-case scenario something like high interest rates, zero growth, a flat yield curve and a strong dollar – not something you’d make strategic plans around. To keep the worst case scenario reasonable, the quants need input from executive and senior management. We should not overplay the role of experienced management, though. People have a habit of selective memory. Many a time, events transpire that experts say were impossible. Remember the mortgage crisis? No one³ believed all the subprime mortgage could erode at once. So we also shouldn’t rely exclusively on professional judgement to come up with stress scenarios. In my presentation at CefFPro’s Stress Testing USA conference in November in New York, I will offer some specific techniques that I think can lead to finding meaningful stress test scenarios and how to react to results from them.

Systems Integration; One of the financial crisis learnings was that targeted risk management can miss the boat. Focusing on the trading desk while the funding desk issues credit derivatives is less effective than forcing the results of both through an enterprise-wide stress test. Now that DFAST and CCAR make us integrate our budgeting, accounting, risk and credit systems into an enterprise-wide stress test, pockets of risk that could otherwise be hidden show up. With our ability to pull in relevant data from disparate systems, we can start stress scenarios with exposures from the same set of assets, liabilities and off-balance sheet items that we report in publicly disclosed financial statements. Further, tying to our budget systems keeps stress scenarios consistent with projections that management uses to forecast and plan for resources and growth. This shows management the impact of their budget in stressed conditions.

Board Involvement; Dodd Frank forced the Fed to codify enhanced prudential standards. Some of these are overreaching. The reader of the enhanced standards could mistakenly deduce that the financial crisis could have been avoided if only boards spent more time approving budgets and polices. However, amidst the overreach is a new recognition by board members that they do have the clout and responsibility to credibly challenge management. This is not an easy task when overseeing professionals who do this work day in and day out and who have done so for years. Equally important, with heightened standards, second lines of defense have the boards’ ears and boards have second lines’ backs.

Reviewing stress test results is one way to offer a credible challenge. Three years into the DFAST exercise – four years into CCAR – boards are now used to seeing and reviewing stress test results. I expect a common discussion in board review of stress testing is “how likely are the scenarios.” With Regulatorily prescribed scenarios boards and management could be understandably skeptical. In contrast, they might react more seriously to stresses from home-grown scenarios supported by an independent second line function.

Putting it together; Now imagine a financial institution that has wrestled with my three points. Namely: (1) they apply conceptually sound models to economic scenarios which highlight the banks vulnerabilities;(2) they integrate their source systems to start stress scenarios from an actual exposure starting point and project growth consistent with corporate budget projections; (3) they leverage the attention of the board and the gravitas of their second line risk management experts to present vulnerabilities in a clear enough fashion to allow the board to make an informed decision whether to accept the estimated risks.

From this article, the casual reader could conclude I advocate a single, universal stress test to rule them all. That is not the case. In fact, while I do feel one consistent source of truth is vital and that stress scenarios should be commonly applied, I am a fan of letting disparate stress models do what they do best. Capital stress models to predict capital adequacy, liquidity stresses for funding strains, credit models for asset quality and so on. In my presentation in November at the Stress Testing USA Congress, I will offer an exploratory method for aggregating three models into one vulnerability story.

Regulatory imposed stress testing created a great deal of expense, angst, resource redeployment and a pause in growth. However, it also laid the groundwork for the financial industry to identify vulnerabilities and potential for intolerable exposures better than we have before and in time to avoid them in the future. We can do so through idiosyncratic stress tests.


[1] These are reflections of Tally Ferguson and should not be construed as endorsements, attitudes or strategies adopted b BOK Financial. 

[2] or betas depending upon where and when you learned econometrics,

[3] Actually, the Michael Burys and Steve Eisman’s of the world did, but not the ‘experts.’