When global standard-setters first revealed their planned revamp of trading book capital rules, bankers responded with furrowed brows. The new regime was dizzyingly complex, costly to implement and, worse, packed a huge capital hit.
Several consultation periods later, the Basel Committee on Banking Supervision has released its latest – and supposedly final – version of the Fundamental Review of the Trading Book. The good news for banks? The capital impact is lower. The bad news? The rules are as complex and costly to implement as ever.
“I don’t think the FRTB framework has been made any easier. There is still the burden of running all of the machinery for internal models, which is disproportionate compared to what we are used to,” says a senior risk manager at a large European bank.
The promise of softened capital treatment is clouded by uncertainty over exactly how large the reprieve will be. The Basel Committee estimates that capital will increase on average by 22% – a big difference from the mooted 60% hike in the 2016 version of the rules. But the effect on individual banks could vary wildly, with some firms expected to see their capital shrink by around 19% and others swell by as much as 58% (see table A).
Bankers also complain that the rules target certain asset classes more than others. Market risk accounts for a small proportion – around 5% on average – of banks’ overall risk-weighted assets, so any uplift in capital is only a blip for the institution as a whole. However, banks often consider the profitability of trading desks on an individual basis, leaving firms in a quandary over whether to shutter business lines that are too much of a drain on capital.
“A 20% uplift would be under 1% effect on total capital for banks in Europe so not difficult to absorb,” says a capital manager at a second European bank. “However, this is not how banks work. In general, banks apply the use test to businesses and actively consider the consumption [of capital] at lower business levels. The extra 20% is a very significant uplift to accommodate.”
Banks have a choice of two methods for calculating capital under FRTB: a regulator-set sensitivities-based approach (SBA) or, if trading desks pass certain tests, an internal models approach (IMA). One of the internal tests measures how accurately a trading desk is able to model its profit and loss. This P&L attribution test has gone through significant revisions between its first design in 2016 and the final version of the rules published on January 14 this year.
The changes have made it easier for banks to pass the test, without reducing the burden or costs of implementing the infrastructure needed to run it, bankers warn.
“The rules have been modified to be a little bit more sensible but we will still have to spend roughly the same amount in implementing this as we would have done in 2016,” says a modelling expert at a third European bank. “I wouldn’t underestimate the operational effort to get these things up and running.”
“The P&L attribution test will still be the most difficult piece of FRTB to implement,” agrees Eduardo Epperlein, global head of risk methodology at Nomura. “There is still a lot of engineering work that needs to be done.”
Under the previous version of the rules, a desk that fails the test tumbles from IMA to SBA. The industry complained this would cause a cliff effect on capital, so the Basel Committee proposed an “amber zone” to act as an intermediate phase. Banks pointed out the green and amber zones were too narrow to be effective, so rulemakers widened these zones in their latest version of the regime. However, banks are still not confident of passing the new test as it has only been tried with hypothetical portfolios.
“We don’t yet know how much the figures show an improvement because we haven’t tested it on real portfolios,” says the modelling expert at the third European bank. “It is a good starting point but it might be something that needs to change [in local legislation].”
Desks that slide into the red zone also face a struggle to haul themselves back to internal modelling. Similar to traffic lights in continental Europe, desks can transition from green, to amber, to red. But there is no amber stage in the opposite direction. Once a desk is stuck in red, the only way of advancing is to jump straight to green.
“There is an asymmetry between going from green to amber to red and not being able to go from red to amber the other way,” says a market risk expert at a European bank. “That has to evolve and be further calibrated because you could be waiting a long time before you’re back on internal models.”
There are so many variables that will need to be aligned
Azar Khurshid, Mizuho International
The P&L attribution test requires desks to compare the “hypothetical P&L” generated by front-office pricing models with a “risk theoretical P&L” based on inputs for the risk management model. If the results of these two models are too dissimilar, the desk fails the test. Hence, banks will have to begin to align their back-office risk models with pricing models used on the trading desk.
Azar Khurshid, a director in global risk management at Mizuho International, says: “Throughout the years banks’ trading desks and back-office risk management have developed independently. Even if they are using the same models to price risk they could be using different data. If you are using the same data and models you might be validating them differently. There are so many variables that will need to be aligned.”
The Basel Committee has made concessions in allowing banks to match up data inputs between pricing and risk model P&Ls subject to supervisory approval. The trading desks of large, multinational banks are often not in the same time zone as back-office risk management. Predicting future values of an instrument based on observed prices from different time series can cause breaks between the pricing and risk P&L.
Epperlein of Nomura says: “If we had four time zones for a trading desk, the moment you reach close of day it causes disconnections between the risk engine and front office. We can now align those inputs to avoid disconnections.”
Away from the demands of internal models, firms using the sensitivities-based approach can expect lower capital under the latest framework compared with the 2016 version. Bankers have welcomed a relaxation of risk-weights for interest rate risk, down 30%, and foreign exchange risk, down 50%. Covered bond risk-weights have also been reduced from 4% in the 2016 FRTB to 2.5% and 1.5% for bonds rated AA– or higher. This change will particularly help European banks that sell mortgages to investors in covered bonds rather than securitisations.
European banks also gain a capital reprieve in the form of extra flexibility in calculating capital for forex risk. In particular, the change helps banks affected by movements in exchange rates between forex swaps relative to the currencies that banks use to report their capital ratios.
In the 2016 version of FRTB, banks had to capitalise forex exposures arising from movements in exchange rates between their reporting currency and swap currencies. This forced many non-US banks to perform the calculation twice as they often trade currency pairs (especially involving US dollars) where both currencies are different from their reporting currency – for example, eurozone banks reporting in euros. US banks on the other hand primarily trade and report their capital in US dollars.
Now, banks can nominate a base currency and calculate the exposure relative to that currency, subject to approval. They would then convert the capital charge in the nominated currency into their reporting currency by using the spot exchange rate. The Basel Committee has not included the effect of this relief in its estimates for the capital impact of the latest FRTB.
“For some banks, a big problem with the former forex curvature formula was the double-counting of capital on cross-currency pair options not involving the domestic currency, and the resulting non-level playing field with banks trading options mainly against their domestic currency,” says a consultant. “It is a good development that the regulators have been more flexible and recognise banks were facing different situations and client needs.”
The latest rules also alter the calculation used to capture the risk of changes in correlation between instruments in a portfolio during bouts of market stress. Banks have to calculate capital under three correlation scenarios: low, medium and high, where the medium scenario is the current assumed correlation of the portfolio. The scenario producing the highest capital number has to be used as the minimum capital requirement.
Under the 2016 FRTB, the low correlation scenario imposed a 25 percentage point reduction in correlation compared with the portfolio assumptions – so for a portfolio that is currently 99% correlated, the bank would have to calculate its behaviour if this correlation dropped to 74%. Hjalmar Schröder, head of market risk at Swiss regional bank, Zürcher Kantonalbank, says this would send capital charges “through the roof” for highly correlated but low risk portfolios.
Banks are now able to reduce correlations by a sliding scalar in exact proportion to the gap between the existing correlation level and 100% correlation. Hence if a portfolio is 99% correlated, then the low scenario means reducing this to 98%. The maximum possible reduction in correlation is 25 percentage points, for portfolios with a current correlation of 75% or lower. This leads to lower capital outputs for hedged trades using instruments that are relatively stable and highly correlated, such as interest rate swaps.
“We were happy to see the change to the low correlation scenario confirmed, as this assures an appropriate recognition of hedges, especially for interest rate derivatives,” Schröder says. “For example, the new formula doesn’t penalise interest rate risks that you have against swaps with three-month Libor versus a six-month Libor hedge, which is a low risk strategy and shouldn’t have had the capital charges the low correlation scenario used to produce.”
The promise of lower-than-expected capital on the regulator-set standardised approach will not necessarily result in banks flocking to use this method. Using internal models is still preferable for banks that are able to: Basel estimates that these banks will see a 20% average increase in capital versus a 30% increase for the standardised approach.
A key part of the IMA is a test of whether trading desks have enough observations to prove risk factors are modellable. Risk factors deemed non-modellable, or NMRFs, must be separately capitalised with a stress capital surcharge. In the most recent iteration of FRTB, the test and capital calculation for NMRFs has been significantly relaxed.
“The NMRF charge used to be many multiples of the expected shortfall number,” says Epperlein. “It could have been as many as five times bigger, which is just too high of a penalty for liquidity. Unless it is a fraction of the expected shortfall generated on internal models, it is not a credible number.”
Epperlein sees 30% as a credible add-on number for NMRFs but it is not yet clear whether the changes by the Basel Committee will match that number.
For banks to model a risk factor they need to have either 24 real price observations of a risk factor within 12 months and no gap between three observations spanning longer than 90 days, or have 100 real price observations.
The changes will increase the number of risk factors banks are able to model by widening the maximum possible gap between observations as well as allowing banks to use proxy data to infer risk factors and their own quotes as observations.
I would be very suspicious of any reliable number on NMRF capital charges
Eduardo Epperlein, Nomura
Allowing banks to use proxy data could be beneficial for instruments that can’t meet the criteria for modellability. In order to be able to use proxy data banks must demonstrate the proxies don’t have significant idiosyncratic risk compared with the real price for observation.
Khurshid of Mizuho says this means they should be able to use indexes to prove observability of, for example, Japanese municipal bonds, which are mostly traded in primary markets.
“It means we can now pass the modellability test for local markets that trade mostly on primary issuances,” Khurshid says.
It is not clear how far the changes to NMRFs will lower the capital impact to desks using internal models. The Basel Committee has assumed the changes will result in a 60% drop in capital from the previous FRTB.
“I would be very suspicious of any reliable number on NMRF capital charges because there has been a lot of changes and there was also the problem that until recently many firms hadn’t calculated a reliable NMRF because they made a lot of simplifying assumptions,” says Epperlein. “It was only right at the end that many firms did the more realistic NMRF calculation which then led to a lot of changes to the framework.”
An end to the vendor?
The relaxation of the observability test could decrease the need for vendor solutions and presents a way for banks to cut costs of implementation.
Vendors had planned to either offer banks their own real price data or pool price submissions from banks and sell the aggregated data back to banks.
“We will have less reliance on data pooling services, especially for markets where we are the market-maker now,” Khurshid says. “There was a significant amount of work to even participate in these services.”
Desks where banks aren’t market-makers and so have fewer observations are also less likely to be using the IMA, meaning they wouldn’t face the NMRF charge anyway.
The Basel Committee has clarified that banks can use their own and other banks’ quotes if they are validated by a third-party vendor. But some banks are deterred by the governance processes they would need to adopt to ensure quotes are genuine. Firms leave themselves open to operational risk if their own or other bank traders are accused of offering quotes purely for creating observations to use in FRTB or using quotes that aren’t for trading purposes.
Any regulatory penalties imposed for governance failures are an added worry for banks that are watching nervously as the costs of FRTB mount up.
Weighting game: equity trading under FRTB
For equity desks that are unable or unwilling to use internal models, the standardised approach comes with unwelcome costs. To the surprise of bankers, the Basel Committee reverted back to higher charges for equity risk factors set out in 2016, rather than adopting the lower scalar proposed in the March consultation.
For portfolios of options on single-name equities, the reversed risk-weights may cause unpredictable swings in capital requirements.
“For equities the 2016 risk-weights were causing somewhat strange fluctuations in capital,” says Hjalmar Schröder at Zürcher Kantonalbank. “If you had a couple of options on financial stocks that were 30% out-of-the-money and had three weeks to expire, you don’t really focus on them in risk management as they represent extreme tail risk, well beyond the kind of shocks that drive the stressed value at risk under Basel 2.5. If you then have to apply a 50% shock under the 2016 FRTB those positions suddenly drive your capital charge on the standardised approach. Then they’d expire and your capital requirements would fall back down again.”
The Basel Committee has, however, introduced two new risk-weights intended to lower the capital impact on equity index derivatives. It is unclear whether this will benefit trading desks because banks often use an alternative treatment for indexes that lowers the amount of capital compared with the assigned risk-weights.
Under this approach, known as look-through, desks use the risk-weights of individual components of an index to determine capital requirements for the instrument. If equities within a single index are sufficiently uncorrelated, banks are able to apply a lower risk-weight.
The rule change could reduce the operational costs of using the look-through approach on index derivatives used purely for hedging as banks will not need to monitor the underlying components.
No fun for funds
Investments in equity funds could be heavily penalised in the final framework as they will have to use the higher risk-weights for equities. A senior risk manager at a European bank argues banks are far more likely to use the look-through approach for indexes rather than funds because it is easier to acquire information needed to use the approach.
“Investments in funds are penalised because they don’t have their own bucket like there is for indexes,” the risk manager says. “I find this peculiar because if you’re going to be able to look through anything it is indexes but not funds – because asset and fund managers do not disclose much more information than their mandates.”
If the fund only tracks an index then the bank can treat it as an index and use the lower risk-weights. However, if the fund doesn’t track an index it must either be capitalised using the highest risk-weight of the sector described within the fund’s mandate or the highest risk-weight assigned for a single equity class.
Lack of clarity surrounding the treatment of funds in the calculation for estimating the value of a fund after a default could also pack a further capital hit.
Single-name equities have to be treated as if they have no value once the issuer defaults, which produces higher capital charges.
The final FRTB doesn’t specify how to estimate the value of investment in funds once a default occurs. As the rest of the framework requires funds to be treated as the single highest risk-weight, the senior risk manager worries they might also be required to treat it as a single name for the default risk charge, meaning all the holdings of the fund would suddenly be assigned a value of zero if just one company held by the fund were to go into default.
“A fund seems to be treated in the text similarly to single equity,” says the senior risk manager. “It is not stated how it should be treated in the default calculation but implicitly if you’re treating it as a single name elsewhere you’d also treat it as a single name here. That would mean it would have a value of zero but a fund is a portfolio of potentially several thousand equities and not all issuers are going to default together. Attributing a zero value seems harsh.”
Correlation trading portfolios and FRTB
The sensitivities-based approach is the only option for calculating capital for correlation trading portfolios – bank-issued portfolios containing securitisations and credit default swaps referencing baskets of underlying single-name credit default swaps – meaning banks are barred from using internal models.
A modelling expert at a large European bank points out correlation trading portfolios could be unduly hit by FRTB as the risk charge does not allow banks to recognise all single-name hedges made on exotic securitisations.
“The way the charges are described in the text contradicts the risk management practices of the correlation trading business,” says the modelling expert. “You can be hedged from your mark-to-market perspective but not hedged on capital. That is not a good outcome of any regulation.”
Editing by Alex Krohn