Fractional Reserves & Defending against Death Spirals

This has been a tumultuous week for cryptocurrencies in general, and stablecoins in particular. In the following article, we would like to share some of our thoughts about the current landscape of algorithmic stablecoins and how Lyrebird intends to move forward.

First, we reaffirm our positive outlook for algorithmic stablecoins. The inherent advantages of algorithmic stablecoins are decentralization and capital efficiency, although we have seen that this comes at a cost to peg stability. We believe that these distinctions give algorithmic stablecoins a permanent place in any blockchain ecosystem alongside collateralized stablecoins.

Second, we acknowledge that Lyrebird’s current model is vulnerable to the exact same death spiral that brought down Terra.

Third, we believe that it would be irresponsible for any algorithmic stablecoin protocol to ignore this vulnerability and continue with an unmodified model hoping to prevent it from happening.

Given these thoughts, we have been hard at work surveying different protocols and developing a mechanism to defend against death spirals.

What is a Death Spiral?

Let us first recap the mechanism implemented by LUNA/UST, LRB/USDL, and others. These protocols incentivize arbitrageurs by providing a swap at the target price ($1.00) between the reserve token (LUNA/LRB) and the stablecoin token (UST/USDL), regardless of the exchange traded price.

This is an elegant model that works very well during times of stability and expansion, as the arbitrage opportunity is reasonably low risk. In essence, the model attempts to match the stablecoin supply to the stablecoin demand by incentivizing market participants to take specific actions.

However, this model has a demonstrated weakness when the demand for both the reserve token and stablecoin token drops simultaneously as in the following scenario:

• Changes in market sentiment incentivize market participants to sell the stablecoin token.
• The reserve token is minted to cover the stablecoin supply that wants to exit.
• The reserve token price drops due to the increase in supply.
• Remaining stablecoin and reserve token holders panic after seeing the reserve token’s price drop and sell more of both tokens.
• The reserve token supply can no longer cover the stablecoin supply that wants to exit.

The key insight here is the following:

Because the reserve token is holding up the stablecoin peg, if confidence in the system is lost, stablecoin holders feel that they would incur a smaller loss by selling now versus selling later when the reserve token has depreciated even more in price. This leads to stablecoin holders selling as quickly as possible, causing further stress on the peg.

Next, let’s take a look at some measures that other protocols have taken to address this issue.

Existing Defense Mechanisms

Full Reserves

Lyrebird is far from being the only algorithmic stablecoin protocol that follows Terra’s model. In particular, USDN has been having continuous difficulties holding up its peg, whereas the recently released USDD and USN currently have backing USDT reserves that are larger than their floating supply.

We believe that USDD and USN should currently be well-positioned to withstand market downturns because they can back their entire stablecoin supply with fiat-backed stablecoin reserves. However, this isn’t very scalable. At some point, USDD and USN will have to have their floating supply greater than their fiat-backed reserves, because otherwise they lose the benefits of algorithmic stablecoins altogether. We will be watching closely to see how these protocols develop in the future.

It is interesting to note that in essence, Terra attempted this exact same strategy, albeit with BTC reserves that were smaller than the UST stablecoin supply. It is possible that the peg could have been upheld if the reserves were larger, kept in fiat-backed stablecoins, or both.

Fractional Reserves

This is an interesting space, one that can be viewed as a hybrid between algorithmic and fiat-collateralized stablecoins.

Fractional reserve stablecoins are partially backed by collateral, with a collateral ratio (CR) between 0 and 1. A CR of 0.5 would mean that $1 of a fractional reserve stablecoin is backed by $0.50 of collateral and $0.50 of an algorithmic mechanism.

These stablecoins provide a more measured approach to algorithmic stablecoins, since they can partially provide both the decentralization and capital efficiency aspects of algorithmic stablecoins while adding partial protections in the form of “hard” collateral.

FRAX is the most notable protocol in this space, being backed mostly by USDC. In many ways, FRAX dampens the death spiral by allowing FRAX holders to redeem their tokens for CR USDC and (1-CR) FXS, where CR is the collateral ratio. If CR is high, the stablecoin supply that has to be absorbed by FXS during a market downturn is low, reducing the probability of a death spiral.

Fractional Reserve Redemption Strategies

We saw above that FRAX allows its collateral reserve to be redeemed at a constant ratio — that is, every redemption of FRAX will yield CR USDC and (1-CR) FXS, regardless of when it is performed. But what other strategies exist for redemption if the collateral reserve is smaller than the outstanding stablecoin supply?

If we borrow a page from Terra’s playbook, we can try use the collateral reserves to uphold the peg right away. This works as long as the stablecoin supply that wants to exist is less than CR, since it will be fully absorbed without any de-pegging. Once the reserves are depleted though, there is no defense mechanism other than the algorithmic stability, likely causing a precipitous drop in confidence in the protocol.

The opposite strategy is to reserve the collateral until the very end. This would mean that we allow the peg to compress below $1.00 if the algorithmic stability mechanism is stressed until enough stablecoin supply has exited, and then allow gradual or full redemptions of the remaining stablecoin supply for collateral. As far as we know, there is no protocol that currently implements this strategy.

One obvious downside of this last-redemption strategy is that de-pegging is almost expected during market downturns. However, this strategy has the advantage that the collateral reserves will not be touched until enough of the stablecoin supply has already been burned, making recovery more likely as more and more of the algorithmic stablecoin supply exits.

Lyrebird’s Defense

Before we discuss Lyrebird’s proposed defense against death spirals, let us first re-examine Lyrebird’s value proposition for the Neo N3 ecosystem.

Lyrebird’s primary value proposition is to provide a decentralized and capital efficient stablecoin. We believe that absolute peg stability is of tertiary importance because that is the domain of fiat-collateralized stablecoins.

Given this value proposition, we have decided to move forward with the fractional reserve model described below — one that doesn’t guarantee that the peg will be held at all times, but one that gives the protocol a reasonable chance at recovering from a severe market downturn.

Lyrebird’s Fractional Reserve Model

• Lyrebird will shift to a fractional reserve model that keeps CR fiat-backed stablecoin in reserve for every 1 outstanding USDL.
• The fiat-backed stablecoin will be purchased through open market operations by a smart contract when USDL is minted.
• When USDL supply ≤ fiat-backed stablecoin reserve, every outstanding USDL token will be redeemable 1–1 for the fiat-backed stablecoin.
• Initially, CR will be set to 0.5. This will later be changed to a governance parameter.

Fractional Reserve Swaps

The swap mechanism between LRB and USDL will remain the same as before. Anyone who holds LRB will be able to swap for an equivalent value of USDL and vice versa.

When USDL is converted into LRB, the USDL offered will simply be burned as before.

When LRB is swapped into USDL, a smart contract will check to see if (current collateral / current USDL supply) < CR, where CR is the collateral ratio between 0 and 1. If this condition holds true, the contract will swap some of the incoming LRB for a fiat-backed collateral token to try to keep this ratio at or above CR.

Reserve Redemption

There will be a Reserves page implemented in the Lyrebird front-end application. Here, users will be able to check the collateral ratio and reserves of the protocol.

When current USDL supply ≤ fiat-backed stablecoin reserve, there will be an option for users to redeem any amount of their USDL 1–1 for the fiat-backed stablecoin.

What does this mean for market downturns?

The first thing we must acknowledge is that this mechanism does not prevent de-pegging, and is not intended to. If there is a severe market downturn, we expect that USDL can deviate significantly from the peg for a period of time.

However, we believe that USDL will not permanently be priced below CR. This is because we have CR fiat-backed stablecoin supply backing every USDL supply, giving each USDL a value of ≥ CR in expectation.

Next, we hope that our reserve mechanism can reverse the psychology of the death spiral. Death spirals occur because token holders believe that while they may be selling at a discount now, they will get a worse deal if they act later.

Lyrebird’s redemption model attempts to reverse this incentive structure. The first (1-CR) supply of USDL will need to be absorbed by LRB at a potential loss, but every USDL token after that will be redeemable 1–1 for a fiat-backed collateral token, which means that USDL holders benefit from acting later as opposed to now.

Theoretically, this means that the value of USDL should reach an inflection point and start increasing back toward $1.00 as more of the supply is burned for LRB, as the remaining supply is now closer to being redeemable for fiat-backed collateral. Once USDL supply ≤ fiat-backed stablecoin reserve, we expect arbitrageurs to bring USDL back to $1.00 since they would otherwise be able to perform low-risk arbitrage.

What’s Next?

Concretely, we plan to take the following steps:

• Burn the remaining 9.95M genesis USDL in reserve. We do not have any immediate use for this reserve and it will take too long to build up collateral.
• Implement a mechanism to purchase CR fUSDT on the open market whenever 1 USDL is minted.
• Implement a redemption mechanism to exchange USDL for fUSDT whenever USDL supply ≤ fUSDT reserves.
• Add support for more fiat-backed collateral reserves as they are introduced to the Neo N3 ecosystem.