CVA calculations see next evolution

Methodologies employed to calculate the credit value adjustment on derivatives portfolios have proved somewhat elastic, as banks continue to adapt models to reflect the changing nature of counterparty risk and ongoing technological developments.

Deutsche Bank is the latest to make changes to the way that it calculates the measure. Earlier this month, the bank confirmed a €186m negative CVA impact on its fourth-quarter 2012 results, driven by a methodology refinement.

According to sources familiar with the shift, similar adjustments are likely to be made by competitors in the coming months as banks adopt increasingly sophisticated methodologies.

It is an evolving process, but some are concerned that there may be too much divergence in calculations across counterparties and that some aspects of the measure could be being incorrectly valued.

“It’s not clear that everyone is calculating CVA correctly and it’s commonly said that if you ask five bankers what CVA is, you get 15 different answers,” said Bob Park, CEO of FINCAD, a risk management software firm offering CVA calculation tools through its F3 platform.

“It was relatively unheard of a few years ago and now most major banks have CVA desks, but as with all technical innovations, everyone is doing it differently and some parts aren’t being calculated accurately.”

One source of contention is the calculation for hybrid instruments and cross-asset portfolios due to the intricacies of the correlation involved, which aren’t always accounted for correctly.

“It’s a portfolio level calculation as opposed to a trade level calculation, so there are offsetting position and correlations involved. It is particularly complicated for netting sets with multiple asset classes in them as you need to consider the correlation between the market risk factors underlying those asset classes; also the correlation with counterparty default risk. It is very important to get those correlations correct and that’s not always happening,” said Tony Webb, director of analytics at FINCAD.

Banks have used CVA for years to measure and reserve against potential credit losses from derivatives counterparty risk. However, the regulatory focus has ramped up in recent years after it emerged that two-thirds of losses from counterparty credit risk in the 2008 financial crisis emanated from CVA losses and only a third from actual defaults.

The Basel Committee on Banking Supervision has since taken a more hands-on approach to CVA, most notably by requiring banks to hold capital against potential losses. Banks are still given some flexibility when developing their CVA methodologies, although these do require regulatory sign-off.

“According to a recent BIS report, banks applying their own internal models can come to very different conclusions regarding the risk of a portfolio. Both regulators and banks want to improve and benchmark existing models to get higher comparability over risk assessments,” said Per Sjoberg, chief executive of TriOptima.

The post-trade processing firm has just begun a pilot of triQuantify, a counterparty credit risk analytics service that looks to take advantage of technology pioneered in the computer gaming industry to produce more accurate portfolio simulations.

“People have previously used mathematical models like Black-Scholes that can produce unrealistic results when doing portfolio simulations across different asset classes. With these new numerical methods we can apply consistent measurements across all asset classes and develop much more accurate predictions of how portfolios are likely to perform in the future,” said Sjoberg.

But while regulators and many vendor solutions providers continue to push for greater convergence, some warn that some variation in methodologies employed reflects important differences between business models.

“With the kinds of granularity we’re looking at, we’ve moved away from a textbook theoretical view of the world. There are a lot of industry conversations going on right now on the variety of ways to calculate CVA. The methodology you use depends on your business model and the sophistication of your investor base – it’s not a one-size-fits-all approach,” said a credit valuation professional at a European house.

Even so, bankers believe there has been greater convergence in methodology in recent months, with a number of larger houses vying to take the lead in any move towards standard risk management practices.

“Technology tends to improve in jumps, rather than in a linear progression. Some banks are now calculating CVA in a similar way, but there is still a lot of variation in methodology. Smaller banks tend to use vendor solutions, but the big banks tend to implement their own models in-house,” said FINCAD’s Webb.

“It’s not necessarily a good thing for everyone to apply the same methodology or simplified approximate formula as this stifles technological progress and innovation. The fundamental definition of CVA is well understood and agreed; but the variations across the various methodologies are related to accuracy, speed of calculation, and the ability to correctly handle cross-asset portfolios and exotic instruments.”