Derivatives traders have been shocked by a $712 million revaluation loss at Standard Chartered after the bank changed the way it calculates counterparty risk – a switch that lagged the rest of the industry by as much as a decade.
A spokesperson for the bank says Standard Chartered's predominantly emerging markets portfolio lacked the data that would have allowed it to change its approach to credit valuation adjustment (CVA) earlier. Insiders claim the bank's traders have long supported a change in policy.
Rivals say there was nothing preventing the bank moving sooner. Three sources note Standard Chartered would have been able to price certain trades more cheaply with its old methodology, which used credit ratings to work out probabilities of default (PDs) – a so-called historical approach – rather than market prices.
"I remember having debates with other banks many, many years ago and being astounded people were still arguing that you could use historical spreads – it never came into anything we did. The idea that Standard Chartered was still doing that as recently as last year is pretty surprising," says one CVA expert.
A trader at one European bank says it has used market-implied PDs since introducing CVA in the mid-2000s. His counterpart at a second bank says it had switched from historical PDs by 2010. An industry survey run by consulting firm Solum Financial suggests the market approach was generally regarded as standard by the start of this decade.
The gap between the two approaches can be significant. According to Moody's Investors Service, any company rated Aaa has a 0.0029% chance of defaulting in the course of the next five years. But the five-year PD for Aaa-rated pharmaceuticals giant Johnson & Johnson was 1.77% as of March 10, according to analysis of its credit default swap (CDS) spread by data vendor Markit (see table A). As a rule of thumb, traders and pricing experts say the market-implied approach – when used across a diverse business, rather than for individual trades or portfolios – will generate CVA numbers that are as much as double those produced by historical PDs.
"The PDs you get from CDSs are typically far larger than those from historical. Especially for high-quality names – there the difference can really be extreme," says Claudio Albanese, chief executive of advisory firm Global Valuation in London.
Market-implied CVA is also more volatile, which is why dealers hedge it systematically, often using CDSs – the same instrument on which PDs are usually based.
This is said by one insider to have played a part in the size of the revaluation charge. Spreads for Markit's Asia CDS index – focusing on non-Japan investment-grade names – widened 48 percentage points between the start of November last year and a February 11 peak (figure 1), as concerns about China's economic growth hit the region.
Standard Chartered's methodology change was accompanied by the creation of a CVA desk to risk-manage the greater dynamism of the numbers.
The disclosures appear in the bank's annual report, published on February 23. On page 231, it states that it has changed its CVA approach "to be in line with market practice".
It continues: "The revised methodology for credit valuation adjustments uses market-implied spreads for estimating loss rates, rather than historical or internally modelled loss rates which were used in the past." The resulting loss, which is close to the full-year revenues of $793 million earned by the rates business at Standard Chartered, is first broken out on page 270.
A spokesperson for the bank says it has always recognised CVA and moved to the market-implied approach as soon as it was able: "Prior to now, we have not had an observable market for the majority of our derivatives positions, given the markets we operate in and the counterparties we deal with."
Speaking to Risk.net in 2012, the bank's head of markets at the time, Lenny Feder, took a different line, arguing it was "virtually impossible" to hedge market-implied CVA for a largely Asian client base, given the CDS market's limited coverage of the region.
Feder (pictured) also opposed market-implied CVA on philosophical grounds: "Looking at CDS to determine the creditworthiness of a client doesn't seem appropriate to me," he said. "I believe most banks in Asia determine their credit appetite within their credit department by undertaking fundamental analysis of their client's financial health."
Feder left the bank in mid-2014.
Insiders say traders at Standard Chartered were well aware of the difference between historical and market-implied CVA valuations, but lacked support for the change.
Whatever the reason, the delay may have allowed – or even encouraged – the gulf between the book's historical and market-implied CVA to widen, by letting Standard Chartered price more aggressively for business with higher levels of counterparty exposure.
"If you're pricing CVA or capital lower than someone else, you're going to end up with a lot of spicy trades – a lot of long-dated, uncollateralised stuff. So your CVA is going to be big when you calculate it," says a London-based derivatives consultant.
Other banks recognise it is harder to calculate and hedge CVA for Asia portfolios but say they have not made an exception for the region. Where there is no CDS trading on the client in question, banks find other ways to estimate default risk. This typically means using an available price for a comparable name – a similarly rated entity in the same industry sector – or another proxy.
The head of CVA at a large European bank says his employer generates market-implied PDs by cross-referring to Markit's CDS indexes, such as the iTraxx family, which are made up of well-traded single-name default swaps.
"What we'll do is take the name – it has an internal rating – and take how the names in the index with that same rating regress against the iTraxx index. They all regress to a certain beta. Then we apply that beta to the name in question, so that name becomes beta times iTraxx," he says. "Once you have those betas you're done because then you just hedge with the iTraxx, so your accounting CVA is almost perfectly hedged."
If a counterparty doesn't have a relevant iTraxx index, he says, the bank would use the relevant sovereign's CDS.
A quant at one UK bank says it is possible to use equity or balance sheet data to estimate distance to default of a name without a CDS, and use that in choosing a single-name CDS that has the same distance to default. Others say this is probably too complicated to deploy on a wide scale.
Fair valuation accounting says all parameters should be calibrated to traded instruments. This implies one must use CDS spreads to find accounting CVA. Everyone knows that in Europe and North America
Claudio Albanese, Global Valuation
Sticking to historical PDs could give a bank an advantage over its peers, allowing it to charge less, and there have long been suspicions that some banks active in the Asia market do not apply CVA fully. As a group, Japanese banks are often said to be using a more flexible approach, and one London-based trader says his bank often felt it was competing against a rival that was "using a completely different pricing paradigm".
It isn't possible to conclude that Standard Chartered's CVA methodology change will also produce a change in its pricing, though. Its traders may have been in the habit of adding a buffer to the official number. The bank's spokesperson declined to comment, and interview requests were ignored.
The make-up of Standard Chartered's credit risk portfolio shows a steady migration towards better-rated counterparties, which would be consistent with keener pricing for those customers. Regulatory reports show that since 2010, when most dealers moved to market-implied PDs, the bank's exposure-at-default (EAD) numbers – which capture trading and non-trading book positions – have gradually improved. In 2015, 57% of its non-financial corporate exposures were with companies rated BBB– and above on the Standard & Poor's (S&P) rating scale, up from 49% in 2010. This effect could also reflect a general improvement in creditworthiness during the period, or a shift in risk appetite by the bank.
The Standard Chartered regulatory disclosures also show the internal rating scale used by the bank when modelling the credit risk of its loan portfolio, alongside the estimated PD range and the appropriate S&P rating grade.
It's not known whether the bank had been using the same internal ratings and PDs in its historical CVA methodology, but if so, the scale would assign a one-year PD of 0.015% to an AAA-rated borrower. This compares to CDS-derived five-year PDs of between 0.35% and 1.77% for three randomly selected corporate names, according to Markit. Companies rated BBB by S&P would fall into a Standard Chartered internal ratings bucket with a one-year PD range of 0.11–0.17%; three corporates currently rated BBB have CDS-implied five-year PDs of between 2.3% and 6.95%.
Given the size of the revaluation, some observers ask why Standard Chartered's auditor, KPMG, didn't press the bank to adopt market-implied PDs earlier. International and US accounting standards require fair value for financial instruments – which includes CVA – to be calculated on the basis of exit prices, using market data where possible.
"Fair valuation accounting says all parameters should be calibrated to traded instruments. This implies one must use CDS spreads to find accounting CVA. Everyone knows that in Europe and North America. I've only heard about some lower-tier Asian banks that are still using historical PDs for the CVA," says Global Valuation's Albanese.
Auditors do consider PD inputs used for accounting CVA models and would need to validate any change. KPMG declined to comment. A source within the firm says it generally encourages banks to move to market-implied PDs, but considers the liquidity of the relevant inputs as a factor.
The UK Prudential Regulation Authority (PRA) does not have formal oversight of accounting CVA models but is known to discuss issues around this as part of its day-to-day supervisory relationship. The PRA declined to comment.
Additional reporting by Viren Vaghela, Aaron Woolner and Duncan Wood
The week in Risk.net, May 19-25 2017Receive this by email