Banks get stressed

The UK's Financial Services Authority has called on banks to improve stress testing of possible disruptive events, while the Basel Committee declared last year that firms must supplement their value-at-risk models with stress tests. But some bankers question whether stress testing should be prescribed by regulators at all. John Ferry investigates

Financial markets have had several years of relative stability, characterised by low volatility, tight credit spreads and falling default rates. What would happen, though, if credit spreads were to quickly widen following a cluster of high-profile defaults, if a terrorist attack were to cripple payment and settlement systems, or a global flu pandemic meant banks were drastically understaffed?

Any one of these events could disrupt markets and have knock-on effects across the financial system, causing instability, a drying up of liquidity and established correlations between asset classes to shift. But are banks aware of how they would be affected by these events?

In its Financial Risk Outlook 2006 report, published in January, the UK's Financial Services Authority (FSA) recommends that banks do more to evaluate how to respond to extreme risk scenarios such as these. The regulator says banks should improve their use of stress testing, and eventually aim to stress aggregated risks across multiple business units. While it acknowledges that fully embedding stress testing into risk management processes remains a challenge, the FSA says that without proper systems in place to aggregate risks in their stress tests, "there is an increased risk that financial conglomerates may not be fully aware of hidden correlations across portfolios".

This comes on the back of the Switzerland-based Basel Committee on Banking Supervision's announcement last July that it intends to change the market risk section of Basel II to make stress testing part of the regular process for generating market risk information (see box overleaf). Under the revised rules, banks must supplement their value-at-risk models with stress tests, the results of which should then be "reconciled back to a clear statement setting out the premise upon which the bank's internal capital assessment is based".

It sounds sensible. However, the call by regulators for banks to use and expand stress testing raises a number of issues for risk managers. How should they expand their range of assumptions and scenarios? Which 'what-if' scenarios should they be using that they are not at the moment?

Stress testing is already used by most banks to supplement VAR calculations. Value-at-risk, for example, does not address unexpected changes in historical correlations. Take the example of a currency that is pegged to the dollar. Such a currency will exhibit strong historical correlation to the dollar, but what if a risk manager wants to analyse the impact of the peg collapsing? In this case, because the historical relationship is so strong, VAR would not address the risk of a revaluation caused, for instance, by sustained pressure from currency speculators. However, a stress test would give risk managers an idea of the effect on their portfolios if the relationship broke down.

Given that stress testing can act as a useful conduit to VAR analysis, it should come as no surprise that regulators are looking to formalise its use. But some risk managers believe regulators are acting rashly by calling for an expansion of stress testing. "Basically, the regulators have realised that trading book assets can default, and that VAR does a very poor job of capturing default, so they're trying to put a patch on it," says a market risk manager at a large investment bank.

In particular, risk managers point to the FSA's recommendation that banks should aim to stress aggregated risks. While the regulator notes that using aggregated risks in stress testing may reveal hidden correlations across portfolios, this is far from a simple matter, say practitioners.

"The problem with stress tests is that, although it is possible to identify several portfolios and carry out stresses for those portfolios, it is very difficult to aggregate the hypothetical losses you get from the stressed portfolios, because you have to guess how many catastrophic losses could occur at the same time," says Riccardo Rebonato, global head of market risk and quantitative research at Royal Bank of Scotland in London.

Ton Vorst, Amsterdam-based corporate executive vice-president of group risk management at ABN Amro, agrees it is very difficult to get a meaningful statistical result from stress testing across portfolios. "You can look at what equity market indexes are doing along with credit market indexes over a period, for example. But it's very difficult to get an exact number. It's always an approximation," he says.

One of the difficulties is in collecting data across various business lines, as well as building models sophisticated enough to stress test across multiple asset classes. And stress testing becomes increasingly complex and unreliable the greater number of portfolios you try to aggregate, adds Rebonato. "There is an enormous combination of possible stresses that could give rise to the same loss. If you simply look at a high percentile of a profit and loss distribution, you lose the ability to identify which events give rise to which losses."

Another difficult factor to contend with is different time horizons across portfolios. While market risks tend to materialise quickly - a currency peg could come under pressure from speculators and be revalued within days - credit risks tends to materialise over a longer period. "If it's a trading book you're looking at, then your time frame of relevance is days or weeks. If it's overnight liquidity, the time frame is even shorter, and if it's the credit book, then it's months or years," says Rebonato.

Clearly, any risk manager hoping for a meaningful result from stress testing has to think about the extent to which it is used - testing for too many scenarios and aggregating risks could make the results meaningless.

The UK regulator has not given specific guidance on the scenarios it believes banks should be testing for. Its report is not prescriptive, but merely encourages risk managers to move in a certain direction. "We continue to encourage firms to develop stress-testing scenario analysis that includes risks not currently captured in VAR models," the FSA report states.

Stress testing has, however, been worked into a number of parts of the new Basel capital Accord. The updated requirements stipulate that stress-testing factors should include market risks that the Basel Committee believe are not adequately captured in VAR models, such as jumps-to-default, significant shifts in correlation and gapping of prices.

Ultimately, though, the issue of where risk managers should draw the line with stress testing and where they should expand its use by testing more scenarios has yet to be answered. David Rowe, executive vice-president for risk management at SunGard-Adaptiv, suggested in his regular Risk column last October (page 65) that certain fixed scenarios should be analysed repeatedly, such as the oil embargoes in the 1970s and the Asian crisis of 1997/98. But he also believes these should be supplemented by bespoke scenarios tailored to existing positions. These could be identified by analysing the results of Monte Carlo simulations that fall outside the 99% threshold and applying stress tests to a limited number of these exposures.

Others agree, noting that while banks can't possibly stress test all possible scenarios, risk managers do need to constantly review the inputs they use and ensure they reflect the current composition of the bank's portfolio, as well as market, political and social trends. "Banks can't just define their scenarios once then walk away from them," says Josie Palazzolo, New York-based senior vice-president at SunGard and market risk product manager for the technology company's Adaptiv enterprise-wide risk management system. "Institutions need to have a more fluid scenario generation and risk process."

This is something banks say they currently do, to some degree. "Every now and then we will look at our scenarios and adjust them, but we won't for example do it on a fixed annual basis," says ABN Amro's Vorst.

Certainly, the consensus is that using VAR by itself is not enough, and that there has been far too much regulatory attention on publishing a VAR figure. "The appeal of VAR seems to be its ability to distil a complex world into a single number from which decisions can be made, but like all single numbers it can be abused as well as used," says Diane Reynolds, Toronto-based director of economic capital solutions at risk technology company Algorithmics.

However, risk managers warn against taking a similar course with stress testing and enforcing it as a regulatory best practice measure. "VAR was best industry practice 10 years ago, and because it was called best at the time it became cast in stone," says Rebonato. "In reality, it's better to have a more fluid regulatory and industry environment, without this obsession with best practice."

In reality, stress testing is more intuitive, and less fitting for use as a formal process to aid the assessment of regulatory capital, say practitioners. "Constructing a coherent stress test and applying it appropriately is still somewhat more art than science," says Reynolds. "The art remains in attempting to suitably quantify the stress event in terms that can be translated into an impact on the entire enterprise."

As with many parts of the new Basel Accord, it seems at least as much effort will go into debating how banks should apply stress tests and VAR analysis as to actually getting down to applying them.

THE BASEL REQUIREMENTS

In July, the Basel Committee released a paper, The application of Basel II to trading activities and the treatment of double default effects, in which it states that a bank must have in place sound stress-testing processes for use in the assessment of capital adequacy. Stress measures must be compared against the measure of expected positive exposure and be considered part of the bank's internal capital adequacy assessment process.

The paper also states that banks must identify possible events or future changes in economic conditions that could have unfavourable effects on their credit exposures, and assess their ability to withstand such changes. The Basel Committee highlights economic or industry downturns, market events and increased illiquidity as examples of scenarios to be used. Where stress tests reveal a particular vulnerability to a given set of circumstances, management should consider appropriate risk management strategies, such as hedging against the outcome or reducing the size of the firm's exposures.

Under Basel II, banks can only use value-at-risk if their risk management framework incorporates the results of a 'routine and rigorous programme of stress testing'. The proposed Pillar II changes require banks to demonstrate that they hold enough internal capital to withstand a 'range of severe but plausible' market shocks, while internal market assessments will include an evaluation of market concentration and liquidity risks under stressed market conditions. Banks should be able to demonstrate to their supervisor that they combine their different risk measurement techniques in an appropriate manner to arrive at the overall internal capital assessment for market risk.

Risk