Being stressed is good for you

Increased regulatory focus means stress testing can no longer play a minor role in banks’ strategic thinking and capital considerations. Many institutions require cultural and procedural change to make this happen, but are they capable of bringing it about? By Jonathan Hawkins

Maarten Gelderman, De Nederlandsche Bank

The financial crisis highlighted numerous failings in risk management. Regulators argue that senior management was over-reliant on certain metrics such as value-at-risk, and failed to appreciate the limitations of the approach - specifically, that it did not capture low probability but high-impact events at the tail of the distribution. Banks are now being pushed to address this blind spot by making better use of qualitative measures such as stress tests and scenario analysis - techniques that were part of the standard risk management toolkit prior to the crisis but were inexpertly used or went largely ignored by boards at many institutions.

Beyond calling for change, regulators haven't yet tried to assess whether industry practices have improved - although a Basel Committee on Banking Supervision official says there are plans to do so next year. When that happens, it's likely to uncover significant differences in how scenarios are selected and applied from one institution to the next, and may result in some friction between regulators and the industry.

A likely flashpoint is the way stress-test results are fed into capital planning and management. Maarten Gelderman, head of quantitative risk management at De Nederlandsche Bank (DNB), the Dutch central bank and financial regulator, believes banks always had the capability to incorporate stress testing into their capital planning but were reluctant to do so. "The main challenge for us was in convincing the banks it was important to feed it into their capital planning. I'm afraid it took the crisis to convince most institutions it is important to do so," he says.

Not everyone is convinced, however. "It strikes me as a very bad idea on a number of fronts," says David Rowe, London-based executive vice-president for risk management at SunGard. "The most obvious - especially if we're talking regulatory capital - is that the point of doing effective stress tests is to let your imagination run wild, to think the unthinkable. In fact, if there's a price tag attached to the worst kind of result you can come up with, I have a feeling it is going to constrain your imagination quite severely. You simply cannot hold enough capital to protect against systemic shocks - you have to limit your vulnerability in the first place."

A half-way house would be to use stress testing to inform economic - rather than regulatory - capital calculations. This approach is advocated by Fernando De la Mora, lead partner of PricewaterhouseCoopers' (PwC) risk and finance practice in New York. But he envisages a world in which banks would give themselves a fair amount of flexibility by also taking into account their ability to generate resources to cover any losses: "If the results identify a capital shortfall in the very short term, then you probably need to consider increasing your capital. But if your capital shortfall will take place in year two or three, there are capital preservation or remediation options other than to just raise equity straight away. You could reduce balance sheet growth, cut dividends or sell businesses, so there is a need to integrate stress-testing processes with capital management contingency plans."

These differing philosophies rest uneasily on a foundation of practical problems: if stress tests are to be used to inform firm-wide capital in some way, then the testing regime has to be all-encompassing and it needs to be fairly dynamic. That's not easy - even something as apparently simple as adding up and comparing results across business lines may be beyond many institutions, says S Ramakrishnan, chief executive of Reveleus and Mantas products at Oracle Financial Services Software.

"Different asset classes and lines of business have built their own models, and they've built them on their own historical islands of data. So when you want to do something in an integrated, consistent manner, you're dealing with departmental capabilities and knowledge now suddenly needing to co-ordinate for an enterprise-level outcome," he says.

On top of that, much of the industry has some way to go in order to carry out consolidated, firm-wide stress testing frequently enough to meet regulatory requirements and make adequate capital provisions, says PwC's De la Mora. He estimates around 30-40% of institutions had consolidated stress-test processes in place before the crisis and so were able to react to the regulators' requirements - but others struggled. A number of firms are undergoing redesigns of their governance models and infrastructure, but carrying out consolidated stress testing on a regular basis remains difficult, he says.
"Most of the large firms can run consolidated stress tests on a twice-a-year frequency. A few can do it quarterly and are trying to look at a monthly run, but it's a challenging exercise to do it that frequently," says De la Mora. There is also a push from regulators to view risk by legal entity, an aim that very few institutions will be equipped to meet, he adds.

There are more interesting challenges, too - for example, using stress-testing results to inform a firm's incentive mechanism. This is something most firms have yet to work out, says Patricia Jackson, Europe, Middle East and Africa head of prudential advisory for Ernst & Young in London. "If a business is running big, concentrated risks it will turn up a greater need for capital in the stress testing, and I think that's an area that needs to be developed further: how to get the stress test output back into the incentive structure of the business," she says.

Faced with these problems, it might be tempting to sweep stress testing under the carpet and hope no-one notices. That's not an option. In May this year, the Basel Committee published its Principles for sound stress testing practices and supervision, a set of guidelines designed to push stress testing higher up the risk management agendas of both banks and regulators. The principles require banks to integrate stress-testing results into their overall governance and risk management culture in a way that affects strategic decision-making at the highest level, necessitating the aggregation of results across asset classes and business lines.

While stopping short of specifying particular scenarios, the guidelines highlight the need for firms to identify those areas and events that would damage them most - so-called ‘reverse stress testing'. Regulators are called on to assess the adequacy of banks' stress-testing programmes and, under Pillar II of the Basel II framework, to ensure their capital resources and liquidity risk management are sufficiently robust to survive the most adverse scenarios.

So, what adverse scenarios should be used? Before the crisis, many banks tended to base stress tests on major historical events such as the 1987 stock market crash, the 1997/98 Asian financial crisis, the bursting of the dotcom bubble and the September 11 terrorist attacks. However, in some cases, banks displayed a lack of imagination and insufficient pessimism in considering the breadth, duration and nature of future events. In a December 2008 consultation paper on stress and scenario testing, the UK Financial Services Authority (FSA) highlighted "an over-reliance on historical data in designing stress tests in the past which, amongst other failures, led to an underestimation in the magnitude of market moves, did not adequately capture tail events and failed to capture the systemic nature of the crisis". Furthermore, according to the Basel Committee paper, many risk managers who did consider more extreme scenarios were faced with a lack of buy-in from senior management.

Three-pronged approach

SunGard's Rowe has long advocated a three-pronged approach to selecting appropriate scenarios, in which historical events, specific portfolio and business model vulnerabilities and subjective judgement all contribute to the origination of scenarios (Risk March 2007, page 71). Of the three, Rowe considers the second, which he terms ‘pessimisation' and is a form of reverse stress testing, to be of primary importance. "I think it makes sense to begin with your own institution's business strategy and positioning in the market," he explains. "You can then consider how these things might make you vulnerable and think of extreme scenarios in this range. It's the Achilles' heel approach."

Prior to the crisis, banks frequently ran scenarios that were not severe enough and were implausible given the nature of their business, says Ernst & Young's Jackson. "If you've got a short equity book, then you have to think about the effect of increases in equity prices, not falls in equity prices. If you build a scenario that has a big fall in the FTSE 100 index in it, then you've given yourself an area where you've actually got profit coming through," she says.

One particular lesson learned was that when liquidity froze up, the assumptions firms had made about how quickly they could react to adverse conditions often proved to be off the mark. Elliot Noma, managing director at Garrett Asset Management in New York, argues that stress testing should evolve to take in how a drop in liquidity will affect markets, and to reflect the fact that crises are often formed of sequences of events in combination rather than single shocks.

"As an example, in the current financial crisis, many decision-makers did not consider the possibility and implications of home prices going down worldwide - such an event has a chain of implications in terms of how we model the cashflows, what that means for pricing, how it will affect the overall markets, and the liquidity of these markets," Noma says. "In other words, the testing has to go beyond scenarios that consider what happens if equities drop 20% or if interest rates go up 500 basis points."

As both the Basel Committee's principles and the FSA's consultation paper pointed out, stress testing in many institutions historically lacked not only an appreciation of the correlated, firm-wide impact of events, but also a sense of their severity. According to PwC's De la Mora, this need not only refer to the magnitude of a given scenario: "What really defines the severity of a stress is not so much the size, but the time horizon or duration of the shock. One could ask ‘what happens if the credit crisis takes place again?' If you just take two weeks that were very bad, the impact would not be as dramatic as if you assume the crisis lasts two years." The time horizon to which a stress can be applied should vary depending on the liquidity of the underlying asset, he adds.

It is also vital banks recognise not only the severity of a scenario or its duration, but the incremental progression of its magnitude - an important component when considering what strategies could be put in place to mitigate a given risk.

"Quite often, stresses - especially on the regulatory side - are focused simply on the profit-and-loss effect of going from a certain state today, to a very extreme stressed state tomorrow," says Volker Wellmann, head of interest rate and foreign exchange global trading risk management at BNP Paribas in London. "In practice, it doesn't happen like that - there will be stages in between. There may be pullbacks, there might be overshoots. Since every bank has particular risks on their books, they shouldn't only be concerned about the scenario profit and loss, but also with a proposed hedging strategy to deal with the market evolution leading to this scenario."

Heightened awareness

The industry has clearly woken up to the importance of testing the effect of extreme scenarios, but has this increased focus galvanised financial institutions into action across the board? According to the official at the Basel Committee, perhaps not: "I think we can say anecdotally that there is a heightened attention and awareness of supervisors' views on the importance of stress testing, but I think it's too early to say that practices have uniformly improved."

Next year's vetting process will assess the degree to which the principles have been implemented, he adds: "For us, it's really a two-step process: to make sure we produce sensible rigorous standards; but then take that all-important second step, which is to make sure they're actually being used, not only by banks, but also by the supervisors."

Naturally, many eyes will be trained on supervisors to both monitor and assist banks' efforts to prioritise stress tests, and the Basel official sees it as the role of regional regulators to enforce guidelines specific to the institutions under their jurisdiction. "The process has worked in a lot of countries in that we've set out less than granular guidance in a lot of cases, just because the situation is so different from country to country, region to region. I think what works very well is when a national supervisor - the FSA, for example - prescribes more granular scenarios, to tailor them to a particular country or to a particular region. So there is certainly an onus on supervisors to take the next step," he says.

DNB's Gelderman agrees with this approach, adding that having flexibility around the particular scenarios prescribed is also important. "In the Netherlands, we introduced more granular guidelines with respect to our evaluations for specific portfolios, such as mortgages, but banks could overrule that if they could demonstrate to us their own analysis was more convincing," he says. "The remaining weakness lies in how banks come up with their evaluation of a global economic impact without us as supervisors giving them exact instructions to develop the scenarios themselves." To remedy this, he suggests, banks must take a more proactive approach to monitoring risk concentrations and conducting reverse stress testing.

Although regulators and banks appear to be fostering an industry-wide focus on stress testing, the danger remains that in laying out more stringent guidelines, the imagination so vital to designing scenarios may itself be stifled.

"I worry we will simply go gung-ho down the stress-testing road in the sense that we'll use the same system as VAR and just do a bunch of shocks," says SunGard's Rowe. "These are important, don't get me wrong, but that's not the full story. You need to think in less technical, quantitative terms and begin to think in more qualitative, structural terms with an explicit recognition that real crises unfold through a sequence of loosely connected events over time."