Fed’s outgoing CCAR chief defends stress tests

Timothy Clark rebuffs US Treasury recommendations; supports more transparency

Federal Reserve
Set in stone? Clark says the Fed does not target particular asset classes or business lines in stress tests

As the principal architect of the Federal Reserve’s stress tests, Timothy Clark deserves much of the credit for whipping the US banking sector into shape after the financial crisis. While it hasn’t always been popular with banks, the Fed’s Comprehensive Capital Analysis and Review has got results. For the first time since CCAR began in 2011, the Fed approved the capital plans of all 34 banks it stress tested in 2017 without objections – a testament to the industry’s newfound resilience and financial strength.

CCAR is not without controversy, however. To their many detractors, the stress tests are as unfair as they are arduous. Banks say they are overly burdensome, and accuse the Fed of being too secretive about its quantitative models and frustratingly vague in setting qualitative expectations.

Those concerns are shared by some in the US government. A Treasury Department report on regulatory reform published in June – prompted by the Trump administration’s pledge to review financial regulation – called on the Fed to disclose its CCAR models and to eliminate qualitative fails.

While the Fed exempted 21 smaller banks from the qualitative tests in this year’s CCAR, Clark, who is retiring from his post as deputy director of supervision and regulation at the Federal Reserve Board this month, disagrees with Treasury’s recommendation to do away with them entirely. “We don’t object to a firm on qualitative grounds because it forgot to dot an ‘i’ or cross a ‘t’,” he says. “This is fundamental stuff banks should have been able to do years ago.”

He also does not believe the Fed should reveal the inner workings of its models. Keeping the code secret prevents banks from defaulting to the Fed’s models, which by Clark’s own admission are less than perfect. “We don’t want the entire industry to just use our models,” he says. “We want them to use their own models and to innovate to create better models.”

For one thing, Clark says the Fed already provides a good deal of information on its models. Since 2016, an appendix describing the models used to project stressed capital and pre-tax income has been included in its annual CCAR disclosures.

Still, Clark says the Fed is considering disclosing the losses implied by its models for a set of hypothetical portfolios. This would permit a fairly accurate inference of the expected losses on any given set of assets, while ensuring banks are not able to game the models by scrutinising them for the precise points where they are weakest.

Clark says the idea, initially floated by former Fed governor Daniel Tarullo in an April 2017 speech, has some merit. “Showing the world indicative loss rates on these hypothetical portfolios would provide even more information on how different risk characteristics impact our model,” he says.

Nobody had a clue how far this knife was going to fall, but it was falling really, really fast. I’m not sure I got home in the summer of 2008. It was basically seven days a week for a few months

Clark also dismisses Treasury’s recommendation that the intermediate holding companies of foreign banks operating in the US be exempted from CCAR if their parent companies are stress tested by their national regulator. “In our IHC rule, what’s done in the home country is considered, but that does not negate the requirement that a local IHC do a stress test,” he says.

The IHCs of foreign banks with more than $50 billion of non-retail US assets were subject to CCAR for the first time in 2017. The results were not made public this year, but they will be disclosed from 2018 onwards.

While Clark defends the way CCAR has been conducted under his watch, he does see room for improvement. The Fed could do a better job of articulating what it expects from banks and explaining why firms are failed on qualitative grounds, he says.

“We do a little disclosure about each bank in our disclosure,” he says. “We can do a better job of linking deficiencies and weaknesses to potentially dire outcomes. We can do a better job of explaining why it’s important, but I would not get rid of the qualitative objection.”

CCAR, now in its seventh year, was born in the aftermath of the financial crisis, and Clark was there from its inception. He joined the Federal Reserve Board in Washington, DC as a senior adviser in March 2008, with a brief to develop analytical profiles of banks. Three weeks later, Bear Stearns collapsed and Clark was immediately seconded back to the Federal Reserve Bank of New York, where he had worked as a bank supervisor since 1995.


After Bear Stearns was sold to JP Morgan, Clark returned to Washington and worked on implementing the Housing and Recovery Act, which placed Fannie Mae and Freddie Mac in conservatorship. He spent most of the summer of 2008 consumed with trying to determine the extent of losses in the mortgage and structured credit markets.

“Nobody had a clue how far this knife was going to fall, but it was falling really, really fast,” says Clark. “I’m not sure I got home in the summer of 2008. It was basically seven days a week for a few months.”

At the time, regulatory minimum capital requirements were set at 4%, and while banks conducted internal stress tests of their trading and banking books, neither they nor the Fed knew if this would be enough to weather an extreme market shock. As the crisis unfolded, supervisors determined banks should be required to hold sufficient capital to withstand a severely adverse scenario.  

“During the crisis and well before Dodd-Frank, we were talking about this,” says Clark. “That meant the way to think about capital for these firms was on a post-stress basis, and not capitalising to withstand another hit was insufficient.”

If the total capital ratio requirement is going to be binding on the firm, they need to figure out how to allocate capital and maximise profits against those

As early as 2005, Clark had been developing desktop scenarios to determine the vulnerability of large banks if losses spiked in a crisis. As he worked on these rudimentary stress tests – the antecedents to CCAR – Clark encountered a problem that would recur later: the Fed lacked the data to conduct deep assessments of the financial condition of large banks, forcing him to make crude back-of-the-envelope calculations. He would later discover banks did not have the necessary data, either.

After the worst of the crisis had passed, Clark was part of a team that designed the Supervisory Capital Adequacy Program the direct precursor to CCAR – which was rolled out in early 2009 as part of the Obama administration’s effort to restore confidence in the US banking system.

The initial reaction to SCAP, which featured two scenarios – baseline and more adverse – was deeply negative. Banks complained the regime’s scenarios were absurdly conservative, arguing it was impossible to foresee any situation where losses could be that high. Ten of the 19 banks that participated in the first SCAP in 2009 failed the test and were required to raise additional equity capital, which did not sit well either.

US banks have come a long way since then. The last firm to flunk the stress test on quantitative grounds was Citi in 2012.

As banks have improved their capital planning, the stress tests have arguably become tougher. In developing its CCAR stress scenarios, the Fed relies on internal analysis and supervisory information to identify potential risks to the financial system. It then calibrates the macroeconomic and financial variables in the scenario to reflect these risks.

Some in banking circles have claimed the Fed’s severely adverse scenarios have become outlandishly bearish, while others complain the stress tests are being used as a policy tool to discourage certain activities and shape bank strategy. Critics have pointed to the sharp drop in commercial property prices in the 2017 severely adverse scenario – which was 5% deeper than in last year’s scenario – as evidence that policy concerns are creeping into the stress tests.

Clark denies this. The Fed makes no attempt to target particular asset classes or business lines in the stress tests, other than to highlight those that may present a salient risk, he says. While acknowledging CCAR has become a binding constraint on capital allocation decisions, Clark says it is no more binding than the Basel risk weightings.

We don’t want to shut down the evolution of frontier best practices by telling them they would all be OK if you just get by

“Whether you had stress testing or not, if you had a really strong Basel programme in place, it’s the same difference. If the total capital ratio requirement is going to be binding on the firm, they need to figure out how to allocate capital and maximise profits against those,” he says.

Bank risk managers have also long suspected the Fed is intentionally vague about its expectations for the qualitative test. Clark acknowledges the criticism, but insists the Fed is acting with good intentions – namely to incentivise banks to make continuous improvements. If the Fed was to clearly define a safe harbour, Clark worries banks might do the bare minimum to meet that standard.

“We don’t want to shut down the evolution of frontier best practices by telling them they would all be OK if you just get by,” he says.

Clark also bats away another common criticism of the qualitative tests: that the focus on governance, risk identification and data quality distracts from holistic risk management and does little to bolster the macro-prudential soundness of the banking system. “When we have all the firms focusing on identifying potential vulnerabilities in such a way they can continue to operate, that is macro-prudentially very sound, because it mitigates the impact weak institutions can have on the rest of the economy,” he says.

Data quality has been a huge concern for the Fed since the early days of stress testing. When SCAP was introduced, supervisors were shocked by banks’ inability to provide even the most basic information – such as whether a loan was issued by the bank or purchased from another bank – needed to perform a thorough assessment of their financial condition.

“We got some unbelievably bizarre answers,” says Clark. “We needed to go back and say, ‘we don’t understand how you could possibly be coming up with these answers’.” He says one of the primary benefits of stress testing is that it has spurred banks to improve data accuracy.

Another by-product is better governance around capital planning and the models banks use to make forecasts. Sources at the 13 large banks with assets of more than $250 billion that were subject to the qualitative component of this year’s CCAR tell Risk.net the Fed’s examiners exhibited a keen interest in how they validate and ensure consistency in their models.

When we have all the firms focusing on identifying potential vulnerabilities in such a way they can continue to operate, that is macro-prudentially very sound

Clark emphasises the Fed adheres to the same supervisory guidance on model risk, called SR 11-7, that it expects banks to follow. A central oversight group consisting of senior Fed officials scrutinises the assumptions and outputs of the models used in the supervisory stress test. Models are reviewed by an independent validation team, which looks at the design and implementations of the Fed’s internal models. The reviewers are not involved in model development and report to a separate oversight group, while the control procedures surrounding model design are reviewed by yet another team of experts.

In the feedback letters banks received after this year’s stress tests, the Fed emphasised it wants to see improvements in the governance processes by which CCAR models and scenarios are reviewed and challenged at the business and board levels. Banks have admitted governance has historically taken a back seat to other aspects of CCAR. However, the quality of governance has improved recently, particularly among boards, which are paying closer attention to the calibration of variables that drive the results of the stress tests.   

As Clark prepares to leave the Fed, he points to one area where there is still more work to be done: reconciling stress testing with resolution planning. A natural linkage exists between the two, in that understanding a bank’s vulnerabilities is a prerequisite to determining whether it may need to enter a resolution programme, and both require scenario-type forecasting.

“Trying to marry those directly can be massively complex,” he says. 

  • LinkedIn  
  • Save this article
  • Print this page  

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact [email protected] or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact [email protected] to find out more.

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here: