Fractional Reserve, Fractional Stability



  • Stablecoins that are backed by onchain assets and can be managed by smart contracts are designed to remove the need for trust in a centralised organisation.
  • This desire is becoming more popular as it aligns with the “decentralised” ethos that attracts many people to digital assets in the first place.
  • Many users would much rather have the rug pulled from beneath their feet by a smart contract than a firm incorporated in the Cayman Islands (although it has repeatedly been made clear that they can have both).

Decentralised finance

The arguments in favour of this are obvious. Smart contracts on the blockchain are open source, persistent, and deterministic. This is in contrast to the leading stablecoin in the DeFi ecosystem, USDT, whose distributor repeatedly refuses to disclose the makeup of its reserves, faces no checks or balances on the collateral that they accept for new tokens, and admit that they “could abscond with the reserve funds” at any time (to abscond: leave hurriedly and secretly, typically to escape from custody or avoid arrest).

On the other hand, many attempts to implement a decentralised system have failed, and Terra UST is just the most recent example. As covered in a previous report, a fatal vulnerability of the project was the lack of backing by assets that derive their value independently from the protocol. Instead, UST merely transferred its price volatility to the project’s equity token, LUNA. As soon as LUNA crashed, the volatility transfer mechanism wore too thin to support the peg and the stablecoin trended towards $0. A similar issue brought a similar stablecoin mechanism (the original implementation of IRON) to a similar price in late 2020.


The smart contracts underpinning the most popular decentralised stablecoin, DAI, manage reserves of cryptoassets onchain without the need for trust in a custodian. To mint new DAI, users can over-collateralise the minting of new DAI by depositing some onchain asset (commonly ETH) into a smart contract (called a vault) and receiving less than the dollar value of the DAI tokens in return. For illustration, let’s say that depositing $150 of ETH allows a user to mint 100 in return. In practice, this value depends on the asset used as collateral, and is subject to changes by holders of DAI’s own “governance” token. 

Users can close this position by returning the 100 DAI tokens to the smart contract, plus some extra accrued as interest, and receive their ETH collateral back. If the ETH backing the 100 DAI tokens falls below a specified dollar value (makerDAO’s documentation gives $150 as an example, but this is again dependent on the asset deposited) the vault facilitates an auction of the collateral to the highest bidder. This bidder pays for the collateral in DAI, with the system aiming for the loan to be repaid before it becomes under-collateralised. This, coupled with some other mechanisms, is how the protocol incentivises the market to value the stablecoin near to $1.

FRAXional reserves ...

This system has proven to be more robust than that of Terra, but at the cost of capital efficiency. Minting new tokens requires users to lock up more value than they receive in DAI, which is unattractive to many users given the huge yields offered by other savings protocols. FRAX attempts to improve this efficiency by recreating a “fractional reserve” onchain – holding a lower value of assets as collateral than there are tokens in circulation.

It does so by blending DAI’s over-collateralisation with Terra’s volatility transfer: users must still provide collateral in order to mint new FRAX tokens, but the collateral deposited is made up of some cryptoasset (like the USDC stablecoin) and some of the protocol’s equity token, FXS. The same is true of the collateral returned by redemptions of FRAX: users that close positions receive the same dollar value of collateral that they deposited (minus fees), but in the prevailing ratio of USDC and FXS.

This prevailing ratio is called the “collateralisation ratio”, and is adjusted once every hour by in steps of .25%. When FRAX is above its peg, the collateralisation ratio decreases, when it is above, it increases. This ratio is intended to represent the trust in the protocol, as it reflects the aggregate reserve amount that is instantly available for redemption for non-FXS assets. The current collateralisation ratio is 89%, meaning that 89% of the total circulating balance of FRAX tokens is backed by USDC or other collateral, and 11% is backed by the protocol’s equity token. It also determines the proportion of USDC and FXS collateral that is required to mint new tokens and that is returned on redemption of FRAX.

... FRAXional stability

However, this still suffers from the same vulnerability as UST and IRON. If FXS drops to $0 (as did LUNA and IRON’s equity token TITAN), then arbitrageurs will only receive $0.89 of total value on redemption of each FRAX token. If the system experiences a run on redemptions whilst this collateralisation ratio is significantly below 100%, then arbitrageurs will only be able to redeem FRAX tokens for $0.89, meaning there is no incentive to value the token any higher.

In this case the protocol would begin to increase the collateralisation ratio which, at the current rate of 0.25% an hour, would take 44 hours. The losses of confidence in Terra and IRON happened much faster, with a 50% drop in just two hours in IRON’s case. The peg would likely remain at the lower bound set by the collateralisation ratio until either all collateral has been redeemed, or confidence in the protocol is restored.

Synthetic shorts

These risks are enough to depeg the stablecoin (and have previously depegged similar stablecoins) whilst only making up 11% of the collateral that backs it. This is not to mention the risks taken on by the protocol by virtue of its reliance on USDC, which makes up the other 89% of the reserves. Whilst the current iteration of FRAX “mostly” takes USDC as collateral, further iterations may be expanded to include assets such as wrapped BTC and ETH.

This would open FRAX up to a similar synthetic short that incentivised large market players to knock UST off its peg. If FRAX is backed 89% by ETH, then a crash in FRAX would cause a similar selloff in ETH as users redeem their FRAX. This would provide the necessary motivation for anyone sufficiently short ETH to exploit the vulnerability in the algorithmic fraction of FRAX’s peg for huge profit- an historically straightforward task as evidenced by the depegging of both IRON and UST.