Depegging

Why it matters

The term is used for two things that have almost nothing in common. In the first, a token's quoted price on a trading venue falls below the unit it references while the issuer continues to honour redemptions at that unit. In the second, the reference itself fails: there is no portfolio, or not enough of one, and no price at which the promise can be met. The first is a liquidity and microstructure event. The second is insolvency. Treating them as the same phenomenon has produced a great deal of confused analysis and at least one badly calibrated risk framework.

The distinction matters because market participants price collateral off venue quotations. A discount that is harmless to a holder who can redeem can still liquidate a borrower marked to that quotation, which is how a price gap becomes a credit loss with the backing intact.

How it works

Two prices exist at once. The primary price is fixed by the issuer's redemption terms. The secondary price is discovered on order books and automated market makers, and it can be anything the flow makes it.

What links them is arbitrage, and what bounds the gap is the cost of exercising primary access: the redemption fee, the minimum ticket, the onboarding requirement, and the settlement window during which the arbitrageur carries risk. Where redemption is cheap, fast and open to many, the band is a basis point or two. Where it is priced, minimum-sized or restricted, the band widens by exactly the amount of that friction.

A third mechanism sits on top for tokens used as collateral. Lending markets and perpetual futures venues mark positions off an oracle, and the choice between a venue price and a composite or redemption-anchored price determines whether a dislocation on one book propagates into liquidations across the market.

Economic mechanism

Start with the arbitrage that is supposed to close a discount. A counterparty with primary access buys at 99.8 cents, redeems at a dollar, and keeps the difference less costs. Tether's published schedule prices that trade precisely: the redemption fee is the greater of $1,000 or 0.1%, so at the $100,000 minimum ticket it is 100 basis points and at $1m it is 10 basis points. A 20 basis point discount is therefore unarbitrageable in small size and profitable in large size. Shallow, persistent deviations are usually a statement about ticket economics, not about backing.

Now the run. Where redemption is rationed by queue, by fee or by eligibility, holders face a first-mover advantage, and the incentive to sell into the secondary market rather than wait for the primary channel is strongest for those who cannot reach it. The IMF's 2025 paper notes that major issuers do not guarantee redemption rights for all holders and that reliance on secondary markets creates first-mover advantages during a loss of confidence. It also records the empirical shape of the distribution: about 99% of intraday deviations from target have been within 1%, and USDC traded 12% below parity in March 2023.

Deviation is also relative rather than absolute, which the best empirical work on the subject makes clear. In HKMA Research Memorandum 09/2022, Ronald Yip found that in May 2022 the stablecoins backed by more reserve assets suffered smaller declines in market capitalisation, with USDC the least affected by contagion from the Terra collapse. Stress does not lift or sink all tokens together; it reallocates balances toward the ones with better-verified backing. A discount on one token is frequently the mirror image of a premium in another.

Finally, the amplification channel. When a token is collateral, a venue discount reduces the marked value of collateral, triggers liquidations, and the liquidations sell into the same order book that produced the discount. More than $19bn was liquidated across the market in the 24 hours around 11 October 2025. That loop is why the identity of the price feed matters more than the size of the initial gap.

Participants

Market makers and authorised redeemers close the gap and set its width. Trading venues determine where the quoted price forms and how thin it can get. Oracle providers decide which price the rest of the market believes, and lending protocols and derivatives venues transmit the consequences. Issuers control the primary channel and, in a stress, the narrative. Central banks and the IMF supply the framing under which supervisors decide whether an episode was a market event or a monetary one.

Examples

TerraUSD in May 2022 is the reference case for the second kind of event. The HKMA memorandum describes an unsustainable deposit rate drawing in balances, tightening global liquidity, mass redemptions, and a stabilisation mechanism that minted Luna until Luna's price reached zero, leaving the token entirely unbacked. There was no external reserve to run on, so the arbitrage that is supposed to restore par instead destroyed the asset backing it.

USDC in March 2023 is the reference case for the first. Backing was disclosed and independently examined; the discount arose because $3.3bn of reserve cash sat in deposits at a bank that failed going into a weekend, and the question was access, not existence. The token traded 12% below parity and returned to par once the deposit was made whole.

USDe on 11 October 2025 is the cleanest recent separation of the two prices. The token printed 65 cents on one venue during the liquidation cascade while Ethena stated that mint and redeem functionality had remained operational throughout with no downtime and that the token remained over-collateralised. Supply nonetheless fell, from more than $14bn during 2025 to roughly $5.9bn in 2026, which is the more meaningful measure of what the episode cost.

Risks and limitations

There is no agreed definition, and the disagreement is not pedantic. A 1% deviation for ten seconds on a thin venue, a 1% deviation for a weekend across the deepest pools, and a 35% print during a liquidation cascade are different objects, and any threshold-based monitoring rule will classify at least one of them wrongly. Practitioners who build collateral systems tend to favour composite or redemption-anchored marks, which protect borrowers from venue dislocations at the cost of exposing lenders to stale prices; practitioners who run venues favour the traded price, on the grounds that it is the only price at which anyone can actually transact. Both positions are coherent and they cannot both be satisfied.

Measurement is unsettled. Issuers publish par figures for tokens issued and not redeemed; trackers publish market capitalisations at traded prices. Circle's most recent examined figure available on 25 July 2026 was $75.89bn of USDC in circulation for 29 May 2026, while trackers put USDC at $74.33bn that day. Those are two measures of different things, and reading the gap as a deviation is a common error.

The tail is unobserved. The finding that 99% of deviations fall within 1% says nothing useful about the remaining 1%, and both the 2022 and 2023 episodes sit in it. Synthetic designs add a further complication: their return comes from perpetual funding and basis, so the yield that supports demand turns negative in precisely the conditions that widen the discount.

Finally, holding par is not the same as being money. The BIS Annual Economic Report 2026 argues that stablecoins fail the test of singleness: redeemability at par with central bank money, with finality. A token can trade at a dollar every day and still not satisfy it.

Key metrics

Track the maximum and the duration of deviation separately, and always with the venue and the depth attached. Track net redemptions through the primary channel during and after an episode, which is the measure of whether holders actually lost confidence, and the uptime of that channel. Track the discount at which the redemption fee starts to bite, about 10 basis points at scale on Tether's schedule, since that is the theoretical width of the no-arbitrage band. And track supply: aggregate stablecoin supply was $312.1bn on DefiLlama's measure on 25 July 2026, and it is contraction rather than price that records a durable loss of confidence.

Related research

The IMF's Understanding Stablecoins is the best institutional treatment of run dynamics and the limits of redemption rights. HKMA Research Memorandum 09/2022 remains the most careful event study of a multi-token stress and the only one to quantify the flight to better-backed instruments. Chapter III of the BIS Annual Economic Report 2026 supplies the monetary standard against which any of this should be judged.

The data on this site

The series below are the ones that bear directly on this concept. Each carries its own source, cadence and coverage.