Quants' Conundrum: Funding Valuation Adjustment Comes Into Its Own

Sell-Side Technology's sibling publication Risk recently posed the question: Are there no more heroes in quantitative finance? Much of the debate surrounds a calculation for derivatives called credit valuation adjustment (CVA), which measures the market value of counterparty credit risk.

Many financial IT challenges involve doing a simple task faster, or more efficiently, or because the regulator mandates it. Rarely is it that a task genuinely tests the bounds of computing power, or the financial engineers behind it.

Valuation adjustments, which quants and analysts colorfully describe as financial "gymnastics," are one of those exceptions.

But CVA has an even more controversial-and potentially just as important-relative: funding valuation adjustment, or FVA. An internally derived reflection of a firm's creditworthiness, FVA measures the cost of funding-almost always borrowing-to operate in derivatives.

While the largest firms have built proprietary engines in-house for CVA, the addition of FVA adds a separate frontier, and a new technology wrinkle, which external specialists are helping firms tackle.

Rarely is it that a task genuinely tests the bounds of computing power, or the financial engineers behind it. Valuation adjustments, which quants and analysts colorfully describe as financial “gymnastics,” are one of those exceptions.

Precision
While funding calculations have been around for years and built into the price of plain vanilla products, it was not important to complex derivatives modeling pre-crisis, when funding options were low-cost and interest rates were aligned. "Banks could essentially disregard collateral and funding in assessing trading desk profit-and-loss (P&L)," explains Alexander Sokol, CEO at risk management software vendor CompatibL and a previous founder of analytics provider Numerix.

Post-crisis borrowing costs, combined with central counterparty (CCP) clearing, dual-curve discounting models, and Credit Support Annex (CSA) threshold mandates are set to bring higher stakes, and a greater squeeze, to funding.

"Profitability of individual transactions is lower as a result of the nature of more standardized structures, like swaps. When you're making 50 to 100 basis points on each transaction, you don't need to worry about three basis points. In the current world, FVA has come into focus as a specific item that needs to be measured and monitored, and then optimized," says Satyam Kancharla, head of the client solutions group at Numerix, explaining why highly precise methods are now required to track the adjustment.

That optimization, says Mario Onorato, senior director of balance sheet and capital management at IBM's Algorithmics company, revolves around an intricate "interconnection of market risk, credit risk, liquidity risk, and collateral." With these calculations being made at the portfolio-level-some firms are already operating fully dedicated desks to address CVA-highly robust analytics will prove crucial as the application of FVA further expands.

"Just look at pricing versus transferring risk and hedging. In the one case, the data is static and meant to be consumed, where in the other case the system has to be more action-oriented and be able to drill down deeply, roll back on transactions, amend them, or create offsetting transactions," Kancharla says.

Those actions are particularly crucial with "wrong way" liquidity risk, where disruptive market events cause a positive correlation between an institution's funding spread and the collateral it is required to post, rapidly limiting a firm's funding options. Lehman Brothers' collapse after massive defaults in the collateralized debt obligation (CDO) market is only the most famous example.

Addressing FVA from a technology perspective has to do, in part, with sheer data capacity: Modeling at a portfolio level can produce, for a single transaction, gigabytes of data. A daily snapshot could therefore easily represent hundreds of terabytes. But it also has to do with the type of modeling being done, according to Algorithmics' Onorato.

"You're looking at your own particular credit spread's future evolution, and need to compare the sum of these components to the expected cost of collateral based on overnight indexed swap (OIS) rate. On the surface, it appears simpler than CVA, but the modeling technique is different. You need to make assumptions, so it is more structural, or Merton approach modeling, rather than reduced-form stochastic-e.g., Monte Carlo or historical simulation-processes. When you're modeling a single firm in isolation versus as a participant in the broader market, there will be more particularities," he says.

We Could Be Heroes
FVA remains hotly contested and, as Onorato puts it, a measurement still looking for a "framework." Building out an effective engine therefore has everything to do with flexibility. Indeed, FVA remains-unlike CVA-unencumbered by regulatory mandate, but with banks continuing to argue over "double counting" dangers in new Basel III capital requirements rules, FVA could well see its way into "Basel 3.5," perhaps replacing debit valuation adjustment (DVA), an accounting procedure that measures the risk of default.

Whichever way regulators go, though, quants may have found in FVA something to study and manipulate for some time. Kancharla points out that a market with ever-wider choices for derivatives execution creates inefficiencies. With the help of customized implementations, those firms that can get funding right in the current environment will enjoy "significant advantage," he says. And indeed, some firms may well decide with that same FVA information to get out of the market for certain products, altogether, on a cost basis.

Today's heroes, after all, look more like gymnasts.

Bottom Line

• Funding valuation adjustment, like related measurements of credit and debit valuation adjustment, is a complex computation taking account of the compound risks associated with borrowing to fund derivatives trading.
• Firms, both in-house and with expert external providers, are developing valuation adjustment engines that can produce both structural and stochastic models, as well as process terrific amounts of data.
• FVA could well end up either mandated by regulators or as a broader industry standard, but already traders as well as quants are developing funding-aware strategies to closely monitor desks' P&L.