FVA has been the center of intense debate between quantitative analysts, traders, treasurers and risk managers. Famous financial theorists, Hull and White, recently published a hotly debated paper in which they present arguments that funding costs have no place in risk-neutral valuation and thus that derivative pricing should remain independent of funding cost.

Fair value is the only relevant accounting measure for a derivative. Quoted values can be used for derivatives that trade in organized markets. In the accounting literature, these are called “level one” inputs to valuation models.

Over-the-counter derivatives must be valued using models that ensure that no one can make an “arbitrage profit” by, say, selling a comparable quoted derivative and buying the OTC derivative. Alternatively, often you can construct a replicating portfolio of bonds and equity instruments that would give the same payoffs as the derivative whatever happens. Then the fair value of the derivative is the sum of the value of the items in the replicating portfolio. These are called “level two” inputs in the accounting literature. If constructing an actual replicating portfolio is impossible, you must use “level three” inputs, which are subjectively derived to approximate inputs that market participants “would” use if they were available.

A shortcut in the process is to derive “Martingale probabilities” for the derivative’s payoffs; this gives the same result as deriving the replicating portfolio (real level two or approximated level three). To understand this, one really needs to read a textbook, such as Hull, on derivative pricing; or leave the valuation exercise to professionals.

More than 80% of the derivatives carried by banks do NOT have quoted values, so valuation models are required more often than not.