Fair Value Hierarchy

The fair value hierarchy is a three-level framework that classifies fair value measurements according to how much they rely on observable market data. Level 1 uses quoted prices for identical items in active markets, Level 2 uses other observable inputs, and Level 3 relies on unobservable inputs such as a company's own assumptions. Banks, insurers, funds and any company holding derivatives or investments use it to measure and disclose those holdings.

The hierarchy is set out in IFRS 13 Fair Value Measurement, issued by the IASB in 2011, and in US GAAP's ASC 820, which grew out of FASB Statement 157 issued in 2006. Both define fair value as an exit price: the price that would be received to sell an asset, or paid to transfer a liability, in an orderly transaction between market participants at the measurement date. The hierarchy does not say when fair value must be used; other standards do that. It governs how the number is reached and reported.

Think of three ways to price a used car. If the exact same model, year and mileage is listed on a busy website with hundreds of sales a day, you just read off the price. If your car is a slightly rarer trim, you look at similar cars and adjust for the differences. If it is a one-off custom build that never trades, you have to estimate what someone might pay using your own judgement.

The fair value hierarchy gives those three situations names: Level 1, Level 2 and Level 3. The point is not that one car is worth more than another. It tells the reader how much they should trust the number, because the further down the levels you go, the more the figure depends on someone's opinion.

• Level 1. Unadjusted quoted prices in active markets for identical assets or liabilities, available at the measurement date. Shares listed on a stock exchange, exchange-traded futures and on-the-run Treasury securities are typical examples.
• Level 2. Inputs other than Level 1 quotes that are observable directly or indirectly: quoted prices for similar items, prices in markets that are not active, and market data such as yield curves, credit spreads and implied volatilities fed into a standard model. Most interest rate swaps, many corporate bonds and plain OTC derivatives fall here.
• Level 3. Unobservable inputs, used when little or no market activity exists. Private equity stakes, complex structured products, illiquid loans, some CDO tranches and contingent consideration in acquisitions are common Level 3 items.

A measurement is placed in the level of the lowest-level input that is significant to the whole value. A swap priced off observable rates but with a significant unobservable correlation input would be Level 3, even though most of its inputs are observable.

Valuation teams start from the most observable data available and are expected to maximise observable inputs and minimise unobservable ones. Exchange prices feed Level 1 directly. For Level 2, firms buy data from vendors and run discounted cash flow or option models. Level 3 valuations usually go through an independent price verification function and a valuation committee, and auditors test them in detail. The hierarchy feeds into several other areas: the impairment charge test under IAS 36 uses fair value less costs of disposal, fair value hedges rely on fair value changes in the hedged item, and the measurement of an embedded derivative that has been separated from its host contract also follows these rules.

Companies must show how much of their fair value balances sit in each level. Level 3 attracts the heaviest disclosure: a reconciliation of opening to closing balances (gains, losses, purchases, sales and transfers in or out of the level), the valuation techniques and key unobservable inputs, and, under IFRS 13, the sensitivity of the value to reasonably possible changes in those inputs. Transfers between levels must be explained. For banks, the size of Level 3 assets relative to equity is a closely watched indicator of how much reported capital depends on internal models, which links the hierarchy to prudential supervision of market risk.

During the 2008 financial crisis, trading in many mortgage-backed securities dried up. Banks argued that forced mark-to-market pricing in disorderly markets exaggerated losses, while others warned that moving assets into Level 3 let banks "mark to model" and delay bad news. Standard-setters issued guidance on measuring fair value when markets are inactive, stressing that distressed or forced sales are not orderly transactions. The episode showed both the hierarchy's value, since it flags where numbers are soft, and its limit: it reports how judgemental a value is, but cannot by itself make that value more reliable. The rise in unrealized losses on securities portfolios in 2022 and 2023 renewed the debate from a different angle, as many bank bonds held at amortized cost were outside fair value accounting altogether.

• IFRS Foundation
• Financial Accounting Standards Board