Confidence lost: Understanding the risks of stablecoins. Part 2: Non-custodial stablecoins.
We have covered the risks of custodial stablecoins in Part 1. Click here if you’ve missed it. The focus of Part 2 is on non-custodial stablecoins.
Non-custodial Stablecoins
1. Characteristics
Non-custodial stablecoins aim to avoid the various risks of custodial designs by removing the need for a custodian or intermediary to achieve a more censorship-resistant and trustless architecture. However, new designs bring new risks and attack vectors which are not well understood: Deleveraging risks, price feed/oracle attacks, governance/miner extractable value, governance/DAO failures [1], regulatory issues or smart contract bugs.
In a nutshell, non-custodial stablecoin introduce different variations of collateral to establish a basis of primary value. Primary value is the concept of collateral serving as the basis for confidence in a stablecoin design.[2]
2. Types of collateral
Subsequently, we can distinguish between three types of collateral: Exogenous, endogenous, and implicit. The transition between the types can blur at times.
Exogenous collateral has a use case and value outside of the stablecoin design. Endogenous collateral is created to serve as collateral for the stablecoin (value is self-referential). Issuance can be agent/user-based or algorithmic (i.e., so called seigniorage shares)[3]. For exogenous collateral, the primary value derives from the confidence in the exogenous digital asset. For endogenous collateral on the other hand, the basis for confidence lies in the design itself.
Implicit/incentive-based collateral in a stablecoin design is created to adjust supply and relies on market mechanisms to achieve price stability: Speculators are incentivized to absorb losses in the event of an under peg (token oversupply) in return for rewards in the event of an over peg (token undersupply).[4]
3. Risks and available buffers
We can now explore if non-custodial stablecoin designs can be price-stable over time or if price instability is to be expected; and if the price instability is expected to be transitory only, due to available buffers. Adequate buffers in a stablecoin design make the long-term expected returns positive and help survive transitory events.
Most non-custodial stablecoins with exogenous collateral are based on leveraged lending markets, such as DAI by MakerDAO. A portion of a digital asset gets tied up in collateral and stablecoins are issued against this collateral based on a set of protocol rules dictating the minimum required overcollateralization. However, like leveraged lending in traditional capital markets, this design features deleveraging risks, which can cause a negative feedback loop on the design’s primary value. First, during liquidation events collateral value may be consumed too fast due to evaporation of stablecoin liquidity. Without strong price stabilization mechanisms, a limited token supply leads to price instability. Counterintuitively, a price drop of the underlying collateral leads to price appreciation during times of shock. This sort of deleveraging spiral resembles a short squeeze,[5] and the cost of deleveraging can go significantly higher[6] than the issuance price of the stablecoin.
A negative feedback loop can also affect designs with endogenous collateral. For both endogenous and implicit collateral, its perceived (self-fulfilling) value derives from high or low confidence of the users in the design and between the participants. Endogenous/implicit collateral amplifies negative feedback and liquidations can cause a fire sale effect of the collateral in parallel with confidence spiralling down, crashing the collateral value.[7]
Notoriously, algorithmic stablecoins with endogenous/implicit collateral designs[8] can suffer from great instability. The issue is not with transparency (all transactions are on the blockchain), but with complexity leading to fragility: Adjustments to the coin supply are not enough to permanently rule out a non-zero price as the design’s equilibrium (these designs are long-term unstable in a probabilistic sense: There always exist states where the peg can be broken). Once the threshold for a stable equilibrium in a tail event is reached, no more adjustments in quantity are possible[9] and the price of the stablecoin becomes a negative function of the stablecoin’s primary value. In other words, there’s a certain point where the deviation from the peg becomes too big, confidence is lost, a shortage of sufficient incentives makes risk absorbers exit the system[10] and the price of the stablecoin abruptly finds its new equilibrium at zero[11].
4. Custodial stablecoins as collateral for non-custodial stablecoins
Like custodial stablecoins, some variants of non-custodial stablecoin designs use other custodial stablecoins as a reserve asset to mitigate price risk of capital assets or fractional reserves. In turn, these designs not only add another layer of protocol risk and complexity, but more importantly these designs inherit all the risks from the underlying custodial stablecoin, i.e. risks they neither own, nor control, nor possess the ability to fully assess or understand or influence or mitigate in any relevant way. In other words, these designs adopt a black box of operational risks from an unrelated underlying custodial stablecoin issuer and extend them[12] on to the token holders.[13]
5. Overcollateralized non-custodial stablecoin designs
Some coins use overcollateralized designs with non-USD linked collateral (i.e., ETH). Historically, it hasn’t been possible to back stablecoins at scale in an overcollateralized manner. These designs are known to shrink over time. Since only the demand for the leveraged collateral asset makes the coin supply grow, the price of the asset must go up forever. However, crypto markets (like traditional markets) are subject to cycles and in a bear market/deleveraging cycle, the coin supply will shrink substantially. The demand for leverage goes away, followed hand in hand by no demand for the stablecoin, (DeFi) integrations diminish, the general utility of the stablecoin fades and people quickly start to favour other, more scalable stablecoins with more utility. By the time the market cycle turns to the upside the more scalable stablecoins designs will have leapfrogged in utility and the overcollateralized stablecoin never recovers or scales back to the size of the previous bull market. It’s an uphill battle the stablecoin design can’t win. Today, there is not a single overcollateralized stablecoin design known which has been able to scale over time successfully using non-USD linked collateral only.
6. Conclusion
As of this writing, there are no decentralized stablecoins of scale available whose collateral can be considered non-custodial. The only two relevant «DeFi stablecoins» which have been able to scale to a market capitalisation greater than USD $1b are DAI and FRAX. However, these designs rely heavily on USD linked collateral to establish primary value (approx. 80% for DAI and 90% for FRAX) in the form of USD Coin, with the intent to provide maximum price stability (we’ll discuss «the decentralized stablecoin trilemma» in another article). The risks of USDC have been analysed in Part 1. Their design choices make the largest decentralized stablecoins of today decentralized custodial stablecoins. An unsatisfactory outcome for crypto market participants.
The argument that USD linked custodial collateral of decentralized stablecoins can be/has been placed into smart contracts falls short. Pulling yourself out of the mire by your own hair does not work unfortunately, even if market participants act economically rational.
Consequently, large design gaps persist in the realm of decentralized stablecoins, specifically for decentralized non-custodial stablecoins.
It may be reasonable to expand designs beyond over- or undercollateralization to achieve risk segregation and add time as a variance factor for confidence to achieve a multidimensional design. The journey to go with decentralized stablecoins where no man has gone before is just about to begin.
[1] The shutdown of FEI triggered user losses https://coinnounce.com/fei-stablecoin-project-decided-to-close
[2] See Klages-Mundt, A., Harz, D., Gudgeon, L., Liu, J. Y., & Minca, A. (2020, October). Stablecoins 2.0: Economic foundations and risk-based models. In Proceedings of the 2nd ACM Conference on Advances in Financial Technologies (pp. 59–79). 62.
[3] Sams, R. (2014). A note on cryptocurrency stabilisation: Seigniorage shares. Brave New Coin, 1–8.
[4] For details refer to Klages-Mundt et al., 63.
[5] «Leverage effect». Klages-Mundt, A., & Minca, A. (2020). While stability lasts: A stochastic model of stablecoins. arXiv preprint arXiv:2004.01304. 1.
[6] MakerDAO introduced the PSM to mitigate this effect.
[7] See «leverage ratchet effect», Admati, A. R., DeMarzo, P. M., Hellwig, M. F., & Pfleiderer, P. (2018). The leverage ratchet effect. The Journal of Finance, 73(1), 145–198.
[8] NuBits, Empty Set Dollar, Basis Cash, Terra’s Luna/UST.
[9] Conversely, infinite quantity adjustments of the stablecoin’s implicit collateral are possible (depending on the design of the algorithmic stablecoin and the tail risk transfer mechanism) with the same result.
[10] Stability is never assured; it’s determined solely by market psychology. Benjamin Simon: Stability, Elasticity and Reflexivity: A Deep Dive into Algorithmic Stablecoins https://www.mechanism.capital/algorithmic-stablecoins
[11] d’Avernas, A., Bourany, T., & Vandeweyer, Q. (2021). Are Stablecoins Stable?. Working Paper. 23. Note that this result differs from the other designs, where the residual value at liquidation always pins down the value of the stablecoin at positive levels.
[12] For details revisit Part 1 of this essay.
[13] Essentially a wrapped version of the underlying token.
