Art and Science: Credit Valuation Challenge II

In part one of this two-part feature, published last month, several challenges were identified with the calculation of credit valuation adjustment (CVA), the numerical value extrapolated to convey the likelihood of counterparty default risk with respect to derivatives trades. Large computer grids, Monte Carlo simulations, and vast amounts of data are required to calculate CVA on an entire portfolio, even for overnight batch runs. For incremental, pre-deal CVA formulated as close to real time as possible, that challenge is exacerbated.

There are further complexities to consider, however. As stated in the previous feature, during the height of the financial crisis, two-thirds of counterparty losses originated from the CVA charge rather than actual default on the other side of the trade, according to the Bank for International Settlements. Hedging CVA, a natural byproduct of the process, has become a continuous issue and places further strain on the already large technical requirements in place.

Beware of Greeks Bearing Gifts
If a typical CVA calculation considers trade data, collateral agreements, and information on the whole portfolio of trades with a counterparty, then hedging the risk of a trade, calculated using CVA, requires the use of Greeks. These are measures of price sensitivities within a derivative, so named for the Greek letters identifying which aspect they quantify, with delta being the change in option value due to underlying price, vega for sensitivity to volatility, and so on.

“I came from JPMorgan, where they have a centralized CVA group, and the reason behind that was to better hedge the risk,” says Dmitry Pugachevsky, director of research at Quantifi and the former head of counterparty credit modeling at JPMorgan. “Otherwise, the interest-rate desk, which has counterparty risk, is exposed to credit risk, so they have to buy credit default swaps (CDSs) or other credit derivatives that they don’t know a great deal about. The idea was to create a specialized desk that basically aggregates all of the CVA and counterparty risk from all of the desks, nets it, and then hedges it as one unit. Basically, whatever your portfolio has exposure to will create market risk and credit risk. CVA is proportional to the probability of default in the counterparty spread. You hedge this by buying and selling, in the simplest case, CDS, or maybe index protection, but you need all of the data that sets the exposure, and you need the credit data.”

The hedging process in itself is difficult, though, as counterparties will often not have a single-name CDS for their transacting partner to purchase protection on. When a CDS that directly references the counterparty is unavailable, higher collateral requirements or costs are inevitable. Also, multi-asset platform capabilities will be required, along with the data requirements.

CVA is proportional to the probability of default in the counterparty spread. You hedge this by buying and selling, in the simplest case, CDS, or maybe index protection, but you need all of the data which sets the exposure and you need the credit data.

“All market data for calibrations, including volatility surfaces and credit curves, portfolio trade data, netting, and credit support annex information is required,” says Jon Gregory, former global head of credit analytics at Barclays Capital and the author of Counterparty Credit Risk: The Next Challenge for the Global Financial Market.

The sheer complexity of hedging CVA causes problems, Gregory says, with serious consequences. “There are lots of imperfect or impossible hedges across volatility, correlation and credit, along with cross-gamma issues.”

In most cases, “subbing in” for counterparties that don’t have their own CDS takes a bit of educated alchemy. It’s a complex task, with whole groups in large banks now dedicated to creating credit spreads for illiquid partners. “With newer accounting rules, they insist on completing CVA using credit spreads rather than default probability, and for some local banks, which have plenty of counterparties like small corporates or even farmers for commodities, they won’t have a CDS,” Quantifi’s Pugachevsky explains. “You have to be creative and create proxies—basically a map with a credit spread for the names which don’t have truly liquid, traded CDSs, either single-name or indices. That’s partly art and partly science. You have to do this based on geography or sector, or some kind of ratings idea, assigned internally or through an external source.”