Fair Value Hierarchy, Explained: How ASC 820 Ranks What Your Numbers Are Built On

Picture two businesses that each report a $500,000 investment on the balance sheet. The first number is the 4 p.m. closing price of a listed ETF, pulled from the exchange and verifiable by anyone in seconds. The second is the owner's own discounted-cash-flow spreadsheet for a minority stake in a private supplier — reasonable assumptions, honestly built, but uncheckable by anyone who was not in the room. Same line item, same dollar amount, completely different reliability. If you are the lender, the buyer, or the auditor reading those statements, the difference between the two numbers matters more than the numbers themselves.

ASC 820, the FASB's fair value measurement standard, exists to make that difference visible. At its center is the fair value hierarchy: a three-level ranking of the inputs behind every fair value number, from hard market quotes down to internal estimates. Here is how each level works, why one weak input drags the whole measurement down with it, and what to do the next time a number in your own books is carried at fair value.

What "fair value" actually means​

Before the levels, the definition they serve. Under ASC 820, fair value is 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. Four ideas in that sentence do most of the work:

• Exit price, not entry price. Fair value asks what you could sell for, not what you paid. The two often match on day one and drift apart after that.
• Orderly transaction. The sale is assumed to happen after normal marketing exposure — not a forced liquidation or a distress sale. A fire-sale price your neighbor accepted under duress is not your asset's fair value.
• Market participants. The hypothetical buyer or seller is independent, knowledgeable, and acting in its own interest. Your strategic reasons for overpaying, or a related party's sweetheart deal, do not set the number.
• Measurement date. Fair value is a snapshot. A quote from last quarter is history, not measurement, no matter how convenient it would be to keep using it.

Two supporting concepts show up constantly in practice. The measurement assumes the sale happens in the principal market — the market with the greatest volume and activity for that asset — or, if there is no principal market, the most advantageous market, the one that would maximize the sale price. And for nonfinancial assets, ASC 820 assumes the highest and best use: the use a market participant would put the asset to, even if that differs from how you use it today. Your warehouse district lot may be worth more as apartment land than as storage, and fair value follows the apartment math.

Why the hierarchy exists​

A standard that only defined fair value would leave every company free to defend any number with any method. The hierarchy adds comparability: it prioritizes the inputs to valuation techniques so that financial statements favor market evidence over private judgment. The governing instruction is one sentence — use the most observable inputs available, and use unobservable inputs only when observable ones cannot do the job.

That priority runs through three broad valuation approaches you will see named in footnotes. The market approach uses prices from comparable transactions. The income approach converts future amounts — cash flows, earnings — to a present value with a discount rate. The cost approach asks what it would cost to replace the asset's service capacity today. The hierarchy does not pick the approach for you; it grades the inputs whichever approach you feed them into.

Level 1: the market hands you the number​

Level 1 inputs are unadjusted quoted prices in active markets for identical assets or liabilities that you can access at the measurement date. Every qualifier matters:

• Quoted prices — a published trade price, not an indication, estimate, or model output.
• Unadjusted — you take the quote as printed. ASC 820 prohibits adjustments to Level 1 inputs, including the famous blockage discount: if you hold a million shares and the daily volume could not absorb them all at once, you still multiply price times quantity. No haircut for size.
• Active markets — transactions happen with enough frequency and volume to give pricing information on an ongoing basis. A quote from a market that has gone quiet may no longer qualify, which is how assets migrate down the hierarchy in a crisis.
• Identical — the quote must be for your exact instrument, not a close cousin. A quote for a similar bond is a Level 2 input, however tempting it is to call it Level 1.
• Accessible — your entity must be able to trade in that market at the measurement date. A quote from an exchange you cannot access does not help you.

Classic Level 1 residents: listed common stocks, exchange-traded funds, exchange-traded options and futures, mutual funds struck at a published daily net asset value, and on-the-run U.S. Treasury securities. Measurements here involve almost no judgment, which is why they carry the lightest disclosure burden — the market did the valuing, and anyone can check its work.

Level 2: observable, but one step removed​

Level 2 inputs are observable inputs other than Level 1 quotes — market evidence that still exists but needs a model, an adjustment, or an inference to reach your specific asset or liability. ASC 820 names four main kinds:

• Quoted prices for similar assets or liabilities in active markets. Your exact bond never trades, but comparable bonds from similar issuers do.
• Quoted prices for identical or similar items in markets that are not active. The quote exists but the market is thin — few transactions, wide spreads, stale prices.
• Observable inputs other than quotes, such as interest rates and yield curves, volatilities, credit spreads, and prepayment speeds.
• Inputs derived from or corroborated by observable market data — for example, a volatility inferred from traded options on a related security rather than guessed.

The workhorse Level 2 technique for debt securities is matrix pricing: pricing your bond from the observed yields of benchmark securities plus spreads for credit quality, maturity, and sector, rather than from a trade in your bond itself. Interest-rate swaps valued off observable yield curves, agency mortgage-backed securities, and over-the-counter derivatives with observable inputs all typically live here.

The boundary that trips people up is the broker quote. A quote from a dealer is not automatically Level 2. A binding quote for your exact instrument in an active market may support Level 1 or 2 treatment, but a single indicative, non-binding quote — one dealer's opinion with no market behind it — is often closer to an unobservable input, and the measurement may belong in Level 3. The label on the email matters less than what backs the number.

Level 3: your own assumptions, disciplined​

Level 3 inputs are unobservable — your entity's own assumptions about what market participants would assume, developed with the best information available. ASC 820 permits them only to the extent that relevant observable inputs are not available, and even then you must still aim at a market-participant perspective, not an entity-specific value in disguise. Your private synergy hopes do not count; a market participant's would.

Typical Level 3 residents: equity in private companies, complex or structured derivatives with no observable market, contingent consideration from an acquisition (earn-outs), mortgage servicing rights, and assets being measured for impairment with forecast cash flows. The technique is often a discounted cash flow or an option-pricing model whose key drivers — growth rates, discount rates, volatility, timing — come from management's estimates rather than market data.

Because Level 3 is where judgment concentrates, it draws the heaviest disclosure: the valuation techniques used, the significant unobservable inputs with their ranges, how the measurement would move if those inputs changed, and — for public companies — a rollforward reconciling opening to closing balances. That disclosure exists precisely because the reader cannot check your work any other way.

The rule that sets the level: the weakest significant link​

Here is the hierarchy's most tested rule. Real valuations mix inputs from multiple levels — a swap model fed by an observable yield curve plus an unobservable credit adjustment, a DCF anchored to observable rates with forecast cash flows. When that happens, the entire measurement is classified at the lowest level of any input that is significant to the measurement as a whole.

Think of it as a weakest-link rule. If 90 percent of your swap's value comes from Level 2 curve inputs but the Level 3 credit adjustment is significant to the total, the swap is Level 3 — all of it, not 10 percent of it. Significance is a judgment call made about the measurement in its entirety: how much would the answer move if that input moved, and would a reader be misled about reliability if the higher level were claimed? Auditors probe exactly this point, because classifying a Level 3 measurement as Level 2 understates how much of the number is estimate.

Two corollaries follow. First, classify by inputs, not by technique. A discounted cash flow is not automatically Level 3 — discount a contractually fixed payment stream at an observable market rate and the measurement may sit comfortably in Level 2. Second, levels can change. A market that goes inactive can push a measurement down; a new listing or newly observable input can pull it back up. Transfers between levels are themselves a disclosure item, because a migration from Level 2 to Level 3 quietly announces that the number got softer.

One family that sits outside the hierarchy entirely​

A common exam and interview trap: some fair-valued investments are never classified in the hierarchy at all. Certain funds and pooled vehicles — private equity funds, hedge funds, venture funds — may be measured using net asset value per share as a practical expedient, and since ASU 2015-07 those investments are shown as a separate reconciling line rather than forced into Level 1, 2, or 3. If your hierarchy table does not foot to total investments, the NAV line is usually why. Do not invent a level for it.

When a small business meets the hierarchy​

Most small companies rarely measure anything at fair value — until one of a short list of events lands on their desk:

• Marketable securities. A treasury reserve parked in ETFs or bonds is marked to market each period, usually Level 1 or 2.
• Crypto assets. Under ASU 2023-08, in-scope crypto is measured at fair value with changes flowing through net income. A Bitcoin holding priced off an active exchange is typically Level 1; a thinly traded token with no active market may fall to Level 2 or 3.
• Stock compensation. Valuing options for a private company — the 409A exercise — is a Level 3 exercise built on unobservable inputs, which is why an independent appraisal carries so much weight.
• Acquisitions. Assets acquired, liabilities assumed, and especially earn-outs owed to the seller are all measured at fair value, and earn-outs are textbook Level 3.
• Impairment testing. When goodwill or long-lived assets are tested, the fair-value side of the comparison often leans on forecast cash flows — Level 3 territory.

Three habits keep these encounters painless. First, track cost and fair-value adjustments in separate subaccounts so the unrealized gain or loss is always visible rather than buried in a blended balance. Second, date and source every quote — "Level 1 close, NYSE, October 31" — because a quote without a provenance is an audit finding waiting to happen. Third, write down Level 3 assumptions when you make them: the growth rate, the discount rate, the timing. Six months later, neither you nor your auditor will reconstruct them from memory.

The bookkeeping itself can stay simple. Suppose your business holds 100 shares of a listed ETF bought at $50.00 and marks it to the $54.25 Level 1 close at month-end, booking the unrealized gain where it belongs:

option "title" "Level 1 mark-to-market example"
option "operating_currency" "USD"
 
2026-01-02 open Assets:Cash:Checking USD
2026-01-02 open Assets:Brokerage:ETF XYZ
2026-01-02 open Income:Unrealized-Gains USD
2026-01-02 open Expenses:Unrealized-Losses USD
 
2026-09-30 * "Buy 100 shares of XYZ ETF at $50.00"
  Assets:Brokerage:ETF   100 XYZ {50.00 USD}
  Assets:Cash:Checking   -5000.00 USD
 
2026-10-31 * "Mark XYZ ETF to Level 1 closing price $54.25"
  Assets:Brokerage:ETF   -100 XYZ {50.00 USD} @ 54.25 USD
  Assets:Brokerage:ETF   100 XYZ {54.25 USD}
  Income:Unrealized-Gains   -425.00 USD

The cost basis steps up from $50.00 to $54.25 while the $425.00 unrealized gain lands in its own income account — cost history preserved, market adjustment isolated, and the Level 1 provenance recorded in the narration where the next person to read the ledger can see it.

Mistakes that misstate the level​

• Adjusting a Level 1 quote. Shaving the price for size, timing, or caution converts a Level 1 measurement into something lower — or into an error. Take the quote as printed or explain why it is not Level 1.
• Treating every broker quote as Level 2. An indicative quote with no market behind it is an opinion, and opinions are Level 3 until corroborated.
• Classifying by technique instead of inputs. Models do not have levels; inputs do. Ask what feeds the model, not what the model is called.
• Skipping the significance test. A small Level 3 input that barely moves the answer does not drag a Level 2 measurement down — but "barely moves" needs support, not assertion.
• Using stale prices. A quote from before the market moved is not a measurement-date input. Refresh it or move down a level.
• Forgetting the NAV line. Investments measured at net asset value per share sit outside the hierarchy. Filing them under Level 2 or 3 to make the table foot is a disclosure error.

Your next fair value number​

The hierarchy is ultimately a labeling discipline: every fair value number carries a grade for the evidence behind it, and the grade is set by the weakest evidence that matters. The next time a number in your books is carried at fair value — a security, a token, an earn-out, an appraisal — ask which level backs it, write the answer next to the number with its date and source, and check whether anything about the market has moved it since. That one-line habit is the difference between a balance sheet your lender trusts and one they discount.