Banks fear outcome as Basel Committee wraps up RWA review

Regulators have spent much of the past year trying to work out why risk-weighted asset numbers are so varied. With the results due soon, bank participants say the study should paint a kinder picture than other recent analyses, but they fear the policy changes that may follow. By Michael Watt

In the next few months, regulators will emerge, oily and dishevelled, from a year spent under the bonnet of the bank capital system. They are trying to find out whether huge bank-to-bank differences in risk-weighted assets (RWAs) – the modelled numbers on which capital ratios are based – are legitimate. The stakes are high. Banks have an obvious incentive to play tricks with their RWAs as a way of reducing the impact of Basel III’s tough new capital requirements, and if the market loses faith in those numbers, it could wreck the whole post-crisis regulatory project.

“There’s obviously no right way of calculating such a subjective thing as risk. But whatever calculation process is chosen has to demonstrate clarity and integrity,” says Mike Harrison, European banking analyst at Barclays in London. “The capital increases in Basel III make it very important for banks to reduce RWAs to remain competitive, but it will be very difficult for banks to convince the world they are doing this properly if there is no trust in the RWA modelling process. Right now, there’s a big credibility gap, and that may start to influence the industry’s ability to raise capital and other sources of funding.”

Banks expect the Basel Committee on Banking Supervision’s RWA study to tell a kinder story than much of the analysis conducted over the past couple of years, which shows big divides – up to 60 percentage points – in the proportion of RWAs to assets at large, international banks. Much of these differences are the result of legitimate modelling choices or guidance provided by national regulators, banks argue (Risk April 2012, pages 35–39). With these factors stripped away, some bank participants estimate the biggest residual gaps will be something like 15 to 20 percentage points.

The question is what regulators choose to do about this remaining, unexplained difference – and although banks are largely complimentary about the study itself, they worry about the possible responses, which range from extra disclosures to the imposition of cruder, standardised models.

And while the committee has been carrying out its work, the policy debate has moved on. Andrew Haldane, the Bank of England’s executive director for financial stability, spoke for an increasing portion of the supervisory community when, in a speech last August, he took aim at the “tower of Basel”, and called for a simpler approach to prudential regulation. Individual jurisdictions, notably the UK and Sweden, have already removed the freedom of their banks to model certain risks. In this context, a complex analysis of a complex capital engine may fall on deaf ears – whatever its findings.

If we’re going to use this model as an extra bit of data for investors to look at, then let’s at least make it credible

The work is being carried out by the Basel Committee’s standards implementation group (SIG), chaired by Charles Taylor, deputy comptroller for capital and regulatory policy at the Office of the Comptroller of the Currency (Risk April 2012, pages 30–33). Within the SIG, one group is looking into RWA calculation for the banking book, while a second is looking at the trading book, and both are expected to report their findings early this year – the latter could be released this month. In both cases, the hope is that by giving banks a number of hypothetical portfolios to model, regulators will be able to work out how big a gulf exists between institutions and can then isolate the sources of that divergence (see box).

But that only holds if every bank approaches the test portfolios in the same way – and ensuring a level playing field was an iterative process, according to participating banks.

“The hypothetical portfolio work done by the SIG’s market risk team was very, very thorough,” says Eduardo Epperlein, global head of risk methodology at Nomura in London. “There were obviously some significant discrepancies in the way banks had interpreted the sample portfolios, because the Basel Committee spent a lot of time drilling into the fine details of the way we had set up the hypothetical trades. This kind of back-and-forth exchange presumably helped it root out the reasons for outlying results.”

A similar thing happened with the benchmarking exercise for the banking book, although a senior regulatory policy manager at one large European bank says at least part of this was because the test had not been properly defined.

“It was a cumbersome process. On the banking book side, the qualitative and quantitative questions obviously came from people without any real-world modelling experience. For the test portfolio, they started at a very conceptual level, rather than focusing on the detailed, technical issues that influence modelling. For instance, they gave no indication whether we should model RWAs on a portfolio with risk mitigation in place – like credit default swaps. It took time to clear these issues up,” he says.

That is one reason why the banking book review is lagging behind its counterpart – the benchmarking exercise was only completed in early November, say some participants, whereas the market risk review was largely wrapped up by the end of September.

Banks agree both streams of work, in the end, were pretty robust. The hope among participants is that the study will show RWA differences to be primarily the result of modelling choices allowed by the capital regime, rather than anything sinister.

“Hopefully this intensive, detailed process has demonstrated to the committee that the differing methodological choices made by banks are made in good faith, and that disparities are mostly driven by numerous detailed but well-founded factors,” says Harry Stordel, head of regulatory co-ordination, policy and controls at Credit Suisse in Zurich.

Even if that’s true, though, regulators may decide banks currently have too much freedom to make these choices (Risk March 2012, pages 20–23). One response could be to fall back on the standardised models that appear in both the trading book and banking book capital rules. These are cruder than banks’ internal models and the result is generally thought to be a higher capital requirement that is less sensitive to risk.

Perhaps the simplest proposal is to continue allowing banks to model their own regulatory capital but to require publication of a standardised RWA number in parallel – theoretically letting investors make an apples-to-apples comparison between banks and enabling them to ask questions when the modelled number deviates dramatically from the standardised one.

Supervisors are already thinking along these lines – the suggestion appeared in the first, conceptual draft of the Basel Committee’s fundamental review of trading book rules last May – and some bankers are not opposed to the idea (Risk July 2012, pages 16–20).

“This type of disclosure will be useful. Its purpose is not to embarrass a bank. It provides a common yardstick and without that there’s no way to compare RWA numbers across banks. If a bank’s internally modelled capital is 80% of the standardised approach, or 110% of the standardised approach, then banks should have to explain to investors why they made those internal choices. It’s a way of restoring some objectivity into the RWA debate without throwing away the whole concept of internal risk assessment,” says Mark White, head of capital management and optimisation at Bank of Montreal and former head of the Basel Committee’s risk measurement group.

But for certain business lines, the standardised benchmark may simply be wrong, critics argue – leaving it some distance from banks’ modelled RWAs and potentially resulting in unfounded scepticism.

“The standardised approach is not risk-sensitive. If you have a derivatives portfolio, then it mostly produces a totally non-risk-sensitive number,” says Manoj Bhaskar, global head of regulatory and risk analytics at HSBC in London. “Depending on the make-up of your portfolios, you can end up looking a lot more risky than you actually are. Because of this risk insensitivity, it can also vary from one day to the next by quite a lot depending on position moves, often even more than internal models. This makes it pretty unsuitable as a number that everyone can trust to reflect a bank’s level of risk.”

Epperlein at Nomura agrees. “For a parallel run of the standardised model to have any useful impact, there needs to be a recalibration of the standard rules to make them more risk-sensitive, especially on the market risk side. If we’re going to use this model as an extra bit of data for investors to look at, then let’s at least make it credible,” he says.

Even with a recalibration, other risk modelling experts say publishing the standardised RWA number would put too great a strain on banks’ resources. “Forcing a bank to publish its standardised RWA number adds more data, but doesn’t provide much more information – banks just don’t risk-manage to those numbers anymore. But running the standardised number is still very complicated, so banks would have to massively expand their modelling capabilities just to publish a measurement that no-one uses,” says the regulatory policy expert at a large European bank.

A tougher response, also floated in the trading book review, would be to start approving capital models at a more granular level. Rather than signing off on a model for the whole group, as they do now, regulators would approve individual models for each trading desk. If a desk fails to get the green light, then it would have to use the standardised approach instead. This would give regulators more flexibility when dealing with RWA oddities or errors – currently, they only have the nuclear option of bumping the entire institution down to the standardised approach.

Epperlein at Nomura describes this as a more “credible threat” to the current status quo than a parallel run of the standardised model, but warns of difficulties. “Creating a framework for more granular model approval processes without also creating a logistical nightmare will be tough. Do regulators really have the time and the manpower to approve a model for every single trading desk? It would certainly be another big strain on a bank’s resources. Regulators might be tempted to do it on a product-by-product basis. But if banks responded by taking some products out of action altogether, then it would leave behind big hedging imbalances,” he says.

But it could get even worse. In current trading book capital rules, known as Basel 2.5, banks are allowed to model RWAs for correlation trading positions under the comprehensive risk measure (CRM), but also have to calculate the capital required under the standardised approach as a floor – they can only use the modelled number if it is higher than 8% of the standardised charge. Dealers found that it usually isn’t, sparking bitter criticism of the CRM during the drafting process – one frustrated banker described it to Risk as “this really, really, stupid method” (Risk February 2010, pages 19–21). Despite that, Basel 2.5 is already in place in the European Union, Canada, Japan and other jurisdictions.

US regulators have taken a slightly different tack in their Basel 2.5-equivalent rules, introduced from the start of this month. As with the standard applied elsewhere, US banks are able to use the CRM, but they have to take 8% of the standardised charge and add it to their internally modelled total. This 8% surcharge will be kept in place by regulators for at least a year, after which it can be removed with their approval.

The same tools – floors or surcharges – could be used more broadly to ensure a minimum common capital level. Again, it rears its head in the trading book review, where the Basel Committee suggests RWAs would not be allowed to drop below some percentage of the standardised approach. It notes that the floor would be set below 100% – meaning there would still be some incentive to use internal models – but the idea has few fans in the industry.

“The idea of using a percentage of the standardised number as a backstop is inappropriate,” says Bhaskar at HSBC. “Every time you have an RWA floor, you add another level of constraint on how you run your bank. If hedging via an option means the standardised charge will jump massively compared with using a future, then you might be tempted to use the future even if it’s not the best solution from an economic risk management perspective. In a fast-moving rates or foreign exchange market, that is not a very sensible thing to do.”

JP Morgan became the poster child for that particular criticism last year, when its chief investment office sought to reduce the burden of the incoming CRM surcharge – which was set to add $40 billion worth of RWAs – by abandoning a tried-and-tested hedging strategy for a more complex programme that massively increased its economic risk, resulting in a loss that stood at $5.8 billion by June 30 (Risk October 2012, pages 23–26).

But whatever the problems associated with the various regulatory responses, the status quo is generating some odd results. Research published by Barclays in October last year looked at banking book risk-weights – the multiplier applied to an underlying position in order to generate RWAs – and found a wide range being used for portfolios that were similarly rated. Across a sample of 20 banks, highly rated corporate portfolios received risk weightings of between 6% and 27%. Perhaps more disquieting from a regulatory point of view is the extent to which risk weights vary over time at the same bank. For example, risk weights on corporate loans rated AAA to AA– increased by almost 70% at Commerzbank and UBS between 2010 and 2011, but decreased by more than 30% at Crédit Agricole. According to Barclays, a similar story is found for all ratings buckets extending down to BB– (see figure 1).

“When a bank’s RWAs move around so much within the same portfolio and relative to similar portfolios at other banks, it makes a mockery of all the negotiations around the required capital-to-RWAs ratio. Can investors be sure that a 10% capital ratio is enough when the RWA number on which that is based jumps up and down every year?” asks Barclays’ Harrison.

The Basel Committee’s RWA study is a first attempt to answer that question, but fixing it may take longer – and the danger is that, given time, scepticism about RWAs will eat away at post-crisis attempts to restore confidence in the industry. In essence, Harrison says, regulators have to choose whether to bolster trust in the system now, albeit with an imperfect solution, or to craft a more thoughtful response that arrives too late to help.

“Everyone agrees that we want the problems with RWA models to be properly settled sooner rather than later. The question is, does the industry want a six out of 10 answer now, an eight out of 10 answer in three years’ time or a 10 out of 10 answer in five years’ time? I’m not sure there’s any consensus on that,” he says.

 

BOX: The answers

An investor looking at two banks with similar asset bases might expect them to have similar risk-weighted asset (RWA) numbers and, therefore, similar amounts of capital. Currently, they don’t, which raises an obvious question – is it the result of differing risk assessments, of inconsistencies in national regulation, or evidence of attempts to cheat? The Basel Committee on Banking Supervision hopes to provide the answer.

The banking book and trading book working groups are using a top-down approach – sending questionnaires to participating banks to gather information on their methodologies – as well as a bottom-up approach, in which banks are asked to work out the RWA numbers generated by identical test portfolios.

The banking book group sent questionnaires to 56 large, internationally active banks and 44 non-internationally active banks in 15 jurisdictions, while the portfolio benchmarking process has involved 33 banks from 13 countries. On the trading book side, at least 15 banks in nine jurisdictions have been involved in both the questionnaire and benchmarking processes.

For banks, the review should provide a great opportunity to do some benchmarking of their own. “Although the Basel Committee hasn’t yet shared with us the RWA numbers it received from other banks, through our conversations with regulators we expect to get a sense of where we are in relation to our peers,” says Harry Stordel, head of regulatory co-ordination, policy and controls at Credit Suisse in Zurich. “This benchmarking process will be very useful in showing where we are more or less conservative in our calculations than our competitors.”

The trading book work stream got off the ground first, with questionnaires sent out last May – and banks were not given much time to respond. “Initially, the questionnaire caught us by surprise. We only had a few weeks to complete it. The document ran to well over a dozen pages, and mainly dealt with our internal modelling assumptions. But it’s good for them to ask these sorts of questions. It was very similar to what we’d be asked in a model approval application,” says one senior risk modeller at a large, international bank.

The questions covered the whole gamut of market risk modelling under Basel 2.5, he says, and provides examples – how is the bank’s value-at-risk model designed? Is it based on variance or co-variance, or does it use a Monte Carlo simulation? What percentage of the bank’s total RWA number comes from VAR? When calculating stressed VAR, does the bank use one stressed period for all exposures or does the period vary by portfolio? Are equity positions included in the incremental risk charge (IRC)? How many liquidity horizons are used in the bank’s IRC model? Where does the bank get the default data for the IRC? Is it point-in-time or through-the-cycle? He estimates the questionnaire took around 80 hours to complete.

The portfolio benchmarking exercise was more onerous. Banks were given a couple of small, hypothetical portfolios for each asset class to run through their internal RWA models. After submitting their numbers, most banks received a day-long, on-site visit from their national regulator in September to work through any outstanding issues.

“It was obviously very tricky for the regulators to get everyone to consistently set up even a simple portfolio like this,” says Eduardo Epperlein, global head of risk methodology at Nomura in London. “They seem to have found big discrepancies between banks during the portfolio benchmarking phase, but most of this was due to the way firms had set up the test portfolios, rather than differences in their risk models. During the review, the committee clarified how it expected the portfolio to be set up, and the final result will probably be a fairly narrow RWA dispersion, at least compared with that on display in banks’ annual reports, which is likely to be driven by differences in portfolio composition, rather than different modelling assumptions. Hopefully, it will show that some of the drastic policy solutions that regulators have suggested to fix the RWA calculation system are not necessary.”

It’s a similar story for the banking book work stream. A questionnaire was sent out, followed by a benchmarking exercise in which banks were asked to submit RWA calculations for different hypothetical portfolios.

However, some bankers complain the review lacked depth. “I’m not convinced the amount of work put into the banking book by the Basel Committee is enough to formulate proper, large-scale policy changes,” says Manoj Bhaskar, global head of regulatory and risk analytics at HSBC in London.

The banking book RWA review is lagging behind its market risk counterpart. Banks were approached by the committee in the third quarter, and only finished answering queries on the benchmarking exercise in early November. The conclusions of the market risk work are due to be finalised early in the first quarter of this year, but the banking book study may not see the light of day for some time after that. It’s not known whether the Basel Committee plans to release both together – committee staff declined to comment for this article.