Fiat-backed stablecoins use reserves such as bank deposits and short-term government assets; crypto-backed stablecoins use digital collateral; pure algorithmic designs rely on supply incentives without an external collateral pool. Hybrids combine these approaches. To classify a stablecoin, record its reference asset, actual backing, redemption output and stabilization mechanism separately, then identify which dependency can fail.
Start with the reference asset
The peg describes what a token aims to track. The backing describes what supports that aim. A token can target one dollar while holding something other than dollar cash, so “dollar stablecoin” is not a reserve description.
The reference unit also matters outside dollar tokens. Paxos describes PAXG as representing one fine troy ounce of physical gold. That is a gold reference, whose dollar price can move; a changing USD quote is not evidence that the token has missed a one-dollar target. Paxos's PAXG description.
Before assigning a type, complete four fields: what one token targets, what assets actually support it, what a holder receives through redemption, and what process is intended to pull the market price toward the target. A name or a flat chart supplies none of the missing evidence.
What the familiar stablecoin types mean
Fiat-backed: external reserves and an issuer route
Fiat-backed usually describes reserves in the traditional financial system. It need not mean every token has a physical dollar bill behind it. Circle's reserve disclosure lists bank deposits, short-dated US Treasuries and overnight reverse Treasury repos for USDC. The issuer describes weekly reserve disclosure and monthly third-party assurance. Those are evidence about a specified reserve arrangement, rather than a definition that every fiat-backed issuer must satisfy. Circle's reserve composition and disclosures.
Redemption is a separate field. Under Circle's USDC terms dated 15 October 2026, which apply outside the EEA, direct redemption with Circle requires a Circle Mint account in good standing. The reserve label does not grant every wallet holder the same operational access. Circle's non-EEA USDC terms.
The price incentive is easier to understand once those fields are separate. An eligible participant may buy below the reference value and redeem, or create tokens and sell above it, when the difference covers costs and uncertainty. Federal Reserve research describes this issuance-and-redemption arbitrage and its dependence on confidence in conversion. It is a mechanism intended to narrow deviations, not a guarantee of an executable price. Federal Reserve analysis of stabilization mechanisms.
Crypto-backed: collateral value and protocol rules
A crypto-backed design can lock digital assets and issue stablecoin debt against them. For a documented example, Liquity v2 lets users deposit ETH or liquid staking tokens and mint BOLD. Liquity v2's general documentation.
Its redemption output differs from a bank payout: the documented process exchanges BOLD at face value for a mix of WETH, wstETH and rETH, less a redemption fee. The redeemer receives collateral whose subsequent dollar value can change. This is one protocol's mechanism; other collateralized systems need their own rules checked. Liquity v2's redemption documentation.
Overcollateralization means the measured collateral value exceeds the debt at the observation time. It provides a buffer, but the buffer can shrink before collateral is converted. Liquity's risk disclosure identifies collateral-price, oracle and liquidation dependencies, including scenarios that can leave bad debt. On-chain visibility helps identify balances; it does not establish that liquidation will realize the last oracle valuation. Liquity v2's risk disclosure.
Algorithmic: inspect what the contraction delivers
“Algorithmic” needs a narrower explanation than “uses software.” Collateralized protocols also run algorithms. In a pure non-collateralized design, the price mechanism instead depends on changing supply or exchanging tokens within the system, without an external collateral pool available for redemption.
NIST distinguishes rebasing, which changes token balances, from seigniorage designs involving a paired volatile token or other incentives. It also separates pure algorithmic arrangements from hybrids that hold some collateral. A rule that changes supply cannot manufacture an external dollar claim. NIST's technical taxonomy, sections 4.4–4.5.
For classification, ask what leaves the system when demand contracts. Is it an independently held asset, newly created auxiliary tokens, or a claim on future expansion? These are different outputs even when each is described as worth one dollar at an input price. A fixed conversion rule still depends on someone accepting what it delivers.
Hybrids require looking through each layer
The three labels are useful starting points, but actual designs can combine mechanisms. A protocol may hold another fiat-backed stablecoin alongside volatile crypto collateral, or maintain reserves while also relying on an auxiliary token. NIST explicitly recognizes combined arrangements as hybrids. The useful record lists each component instead of choosing whichever single label sounds most reassuring.
Hedged synthetic dollars add another distinction. Ethena's current overview describes USDe using crypto assets and corresponding short futures positions. It also lists lending, tokenized real-world assets and liquid-stable holdings among backing strategies, and describes a portfolio that can change. “Crypto-backed” alone would omit the offsetting-position mechanism and the wider asset mix. These are the provider's documented design statements, not a claim that hedging eliminates loss. Ethena's current overview.
A look-through record follows each claim to its next dependency. If collateral is another issuer's token, identify that issuer's reserves and access terms. If a position is intended to offset another position, identify where both are held and settled. Stopping at the first on-chain contract can hide the institution or market needed to realize value.
Trace a hypothetical hybrid's backing
Consider a fictional dollar-targeting token, H, with 1,000,000 tokens outstanding and a $1 reference value. Its treasury holds the three assets below. The auxiliary token belongs to H's own system and has no independent redemption claim on external assets. Every quantity and price is invented.
The comparison denominator is target-valued supply: token count multiplied by the $1 target. It does not assert a legal right to a dollar payout. This exercise freezes token supply and asset quantities. It assumes no interest, liquidation, replenishment, fees, taxes or other debts. It marks inventory at assumed prices; it does not estimate actual recovery proceeds or model a named stablecoin.
| Inventory component | Quantity | Starting price | Starting value | Stress price | Stress value |
|---|---|---|---|---|---|
| Another issuer's reserve-backed dollar token | 600,000 | $1.00 | $600,000 | $0.98 | $588,000 |
| Volatile crypto collateral | 300 | $2,000 | $600,000 | $1,300 | $390,000 |
| H system's auxiliary token | 20,000 | $5.00 | $100,000 | $1.00 | $20,000 |
| Total quoted inventory | — | — | $1,300,000 | — | $998,000 |
The starting quoted inventory is 130% of the $1,000,000 target-valued supply. Excluding the endogenous auxiliary token leaves $1,200,000, or 120%, in external assets. “External” here means outside H's own token system; it does not mean risk-free, independent of the crypto market, or immediately redeemable.
The stress marks are a 2% decline in the reserve-backed token, a 35% decline in the crypto asset and an 80% decline in the auxiliary token. The new total is $998,000, or 99.8% of target-valued supply. External assets alone total $978,000, or 97.8%. Treating the remaining $20,000 auxiliary-token quote as equivalent to bank cash would conceal a different support mechanism.
These percentages are an inventory diagnostic. They do not imply that H must trade at $0.998 or $0.978. Actual realizable value depends on asset rights, conversion costs, market depth and the order in which claims are satisfied. The inputs also illustrate different stresses rather than a standardized ranking of stablecoin types.
Now suppose H has a separate fictional contraction rule: burn H and receive newly minted auxiliary tokens valued at one dollar per H using the rule's price input. Burning 10,000 H would produce 2,000 auxiliary tokens at $5 each, but 10,000 at $1 each. Five times as many units are needed to express the same nominal $10,000 conversion.
That creates no additional bank reserves. Nor does it prove that the recipient can sell those units for the input price. This is the distinction between inventory backing and an endogenous stabilization promise: the ledger should show both rather than add them into one undifferentiated “dollars available” balance.
Separate a backing problem from an access problem
A reserve can retain its measured value while a holder's route becomes unavailable. Equally, a functioning transfer or redemption interface does not establish that the supporting assets retain their value. The stablecoin redemption versus market-liquidity guide follows access and cash-arrival timing after the mechanism has been identified.
Authority is another field. A reserve category alone does not say who can restrict transfers or alter a contract. Use the stablecoin freeze-authority review to examine those powers for the actual token and network.
For a trader, exchange collateral treatment is a further step. A venue can recognize a different value from the reserve estimate or market quote. The stablecoin collateral-risk worksheet addresses that account-level valuation; it should not be substituted for the issuer's balance sheet.
Token units also need a separate record. A rebase mechanism changes balances as part of its rules; a fixed redenomination changes the unit of account. When reviewing such changes, use the token redenomination and adjusted-chart method to distinguish unit conversions from changes in economic value.
Build a mechanism record you can update
Save a dated record for the exact contract and network. Use “unknown” where evidence is missing instead of filling the field with the token's target price.
| Field | Evidence to capture | Failure question |
|---|---|---|
| Reference | Currency, commodity or index and units per token | Are you measuring deviation against the correct target? |
| Backing | Assets, quantities, valuations, custody and report date | Can value fall or be counted twice across layers? |
| Redemption | Eligible holder, output asset, price rule and costs | Does the route return cash, collateral or a system token? |
| Stabilization | Creation, contraction, liquidation or hedging process | Which transaction must participants be willing and able to execute? |
| Dependencies | Issuer, custodian, oracle, market, contracts and control powers | What fails if one required service or asset becomes unavailable? |
Check the record again when reserves, accepted collateral, redemption rules or control powers change. The useful classification is a sentence you can substantiate: this token targets this unit, is supported by these assets, returns this output through this route, and relies on these mechanisms to keep its market price near the target.