What is a stablecoin actually pegged to?
A stablecoin is a token that aims to hold a constant value against something else — almost always the US dollar. The word "pegged" does a lot of quiet work in that sentence, because a peg is not a property of a token. It is the output of a mechanism, and there are three fundamentally different mechanisms with three fundamentally different failure modes.
Type 1: fiat-backed
An issuer holds off-chain reserves — bank deposits, Treasury bills, repo — and issues one token per dollar received. The peg is held by redemption arbitrage: if the token trades below $1 and you can redeem at par, buying the discount is free money until the discount closes.
The arithmetic. With the token at $0.995 and redemption at $1.00, $10,000,000 buys 10,050,251 tokens, redeemed for $10,050,251 — a gross profit of $50,251, or 0.5025%. The peg holds because that trade exists and is cheap.
So the real questions are never about the token. They are: who can redeem (many issuers restrict redemption to vetted institutional clients above a minimum size), how fast, what are the reserves actually made of, and is the attestation an audit or a management report. The dependency chain runs through commercial banks, custodians and payment rails — none of which are on-chain.
Documented case. On 10–11 March 2023 the issuer of USDC disclosed that roughly $3.3 billion of its approximately $40 billion of reserves were held at Silicon Valley Bank, which had just failed. USDC traded as low as about $0.87 on 11 March. It returned close to $1 after the 12 March announcement that SVB depositors would be made whole. Note what moved the peg: not a bug, not the blockchain, not the token contract, but a bank.
Type 2: crypto-collateralised and over-collateralised
Here the backing is on-chain and volatile, so the system demands more collateral than it issues — commonly 130% to 200% or more. Users lock collateral, mint stablecoin against it, and the peg is defended by liquidation (undercollateralised positions are sold), by adjustable stability fees, and by the same par arbitrage.
Everything is verifiable on-chain, which is the honest advantage. The cost is that the peg now depends on liquidations completing at reasonable prices — which requires liquidators to be present, an oracle to be right, and the network to be usable.
Documented case. On 12 March 2020 — "Black Thursday" — ETH fell roughly 50% in a day and Ethereum gas prices spiked. In the resulting congestion, some of MakerDAO's collateral auctions were won with zero-value bids: roughly $8 million of collateral taken essentially for free, leaving about 5.67 million DAI of uncollateralised debt, which the system covered by auctioning newly minted governance tokens. The design worked as written. The environment it assumed did not exist that day.
Type 3: algorithmic
No external collateral at all. The peg is maintained by a mint-and-burn arbitrage against a companion volatile token issued by the same system: burn $1 of the volatile token to mint one stablecoin, or burn one stablecoin to mint $1 of the volatile token.
Read that carefully and the fragility is visible in the design rather than the code. The stablecoin's value rests on the market value of the companion token, whose value rests on demand for the system, which rests on confidence in the peg. It is circular. In calm conditions the loop is self-reinforcing. Under stress it runs in reverse at the same speed, and the next lesson is the full case study.
Type 4: hedged, or synthetic
The issuer holds a volatile asset and sells an equal notional of perpetual futures against it, so the position's dollar value is fixed while the hedge holds. The yield is the funding rate the short collects plus any staking reward on the collateral. The peg therefore depends on three things the other types do not: funding staying non-negative for long enough, the exchanges carrying the short remaining solvent and open, and the custodian holding the collateral off-exchange honouring redemptions. Ethena's USDe (launched February 2024) is the documented example, and its own risk documentation lists the venues used. The funding lesson in the markets course is the mechanism; this is that mechanism sold as a dollar.
The question to ask of any peg
What must remain true for this to hold, and who is obliged to make it true?
- Fiat-backed → a bank, a custodian and an issuer's willingness and ability to redeem at par.
- Crypto-collateralised → collateral value, working liquidations, honest oracles, a usable network.
- Algorithmic → confidence in a token whose value depends on confidence in the peg.
Three answers of visibly different quality. Names appear here only as documented cases; nothing in this lesson is a recommendation to hold, use or avoid any of them, and "stable" is a design objective, not an observed guarantee.
In the data
A peg is observable as a price series, and the number carrying the break is not the one most people plot. Here is USDC across the weekend in March 2023 when a bank holding part of its reserves failed:
On 11 March the low was 0.8774 against a close of 0.9715. The depeg sits in the intraday extreme, and a close-to-close chart of that week understates it by about nine cents. The baseline exists too: tether, the largest stablecoin, closes within single-digit basis points of 1.0000 on an ordinary day, a median of about 6 over the five weeks to 6 September 2026 and above par about one day in nine, which is what any deviation has to be measured against.
Try it now
- A major fiat-backed stablecoin is below. Drop a Level at exactly 1.00 and look at how long the line sits on it. That flatness is the product working, and it is worth seeing before you look at the exception.
- Now navigate to March 2023. Measure the maximum deviation from your 1.00 line and count the bars until it returns. Read the low as well as the close — the two give very different answers, and naming the one you used is part of the answer.
- Find one issuer's most recent reserve attestation. What instruments are the reserves in, who signed the document, and is it an audit or an agreed-upon-procedures report? Then for any crypto-collateralised stablecoin, find its current collateralisation ratio on a public analytics site and state in one sentence what would have to happen for that ratio to fall below 100%.