Insights from Modeling Stablecoins
Deleveraging Risks, Integration with Custodial Assets, and an Alternative Non-Custodial Liquidity Mechanism for Crises
TL;DR A seeming contradiction arises in decentralized stablecoins: while the goal is to make non-custodial assets, these can only be fully stabilized by adding uncorrelated assets, which are currently centralized/custodial (e.g., Maker’s recent inclusion of USDC collateral). In fact, this issue applies more broadly to synthetic and cross-chain assets. We develop an alternative market-based mechanism to enhance stability in crises while remaining non-custodial. This creates a buffer that separates those who are willing to have stablecoins swapped to custodial assets (in return for ongoing yield from option buyers) in a crisis from those who require full decentralization.
Based on a new paper here and our previous 2019 paper here.
Background
On March 12, 2020, called ‘Black Thursday’ during the COVID-19 market panic, cryptocurrency prices dropped ~50% in the day. This was accompanied by cascading liquidations on cryptocurrency leverage platforms, including both centralized platforms like exchanges and new decentralized finance (DeFi) platforms that facilitate on-chain over-collateralized lending. Among many events from this day, the story of Maker’s stablecoin Dai stands out, which entered a deflationary deleveraging spiral. This triggered high volatility of the ‘stable’ asset and a breakdown of the collateral liquidation process. Due to market illiquidity exacerbated by network congestion, some collateral liquidations were performed at near-zero prices. As a result, the system developed a collateral shortfall, which prompted an emergency response and had to be made up by selling new equity-like tokens to recapitalize.
During this time, there was a huge demand for Dai. It became a much riskier and more volatile asset, yet traded at a high premium and fetched lending rates in the mid double digits. Leveraged speculators, who must repurchase Dai in order to deleverage their positions, were exhausting Dai liquidity, driving up the price of Dai and subsequently increasing the cost of future deleveraging (we discuss some further causes that led to market illiquidity in discussing the model in a later section). These speculators began to realize that, in these conditions, they face concrete risk that a debt reduction of $1 could cost a significant premium. Eventually, a new exogenously stable asset — the USD-backed custodial stablecoin USDC — had to be brought in as a new collateral type to stabilize the system.
In addition to the COVID-19 panic, the effects of these risks are also witnessed in bitUSD, bitBTC, Steem Dollars, and NuBits, which have suffered serious depegging events in 2018.
Stability Results
In the paper, we model incentives for speculators (CDP/vault holders) in a stablecoin like Dai to understand Black Thursday-like crises. These speculators decide the stablecoin supply by optimizing the profitability of leveraged positions while incorporating expectations about the cost of future collateral liquidations.
We formally characterize regions in which the stablecoin is stable and unstable. In stable regions, the stablecoin behaves very nicely: with high probability, the stablecoin price is constrained near to peg with minimal fluctuation. Outside of these regions, the stablecoin can be unstable with distinctly greater price fluctuation that can be on the level of underlying ETH volatility.
We characterize how this instability can be triggered by large underlying price swings, collapse of ETH expectations, and liquidity problems from deleveraging in a crisis (which can result in deflationary deleveraging spirals). To deleverage CDPs/vaults, speculators need to repurchase the stablecoin to pay back their debt. Intuitively, if speculators want to deleverage in large quantities and there is no counterbalance, either from marginal speculators wanting to increase leverage and issue new stablecoins or from stablecoin holders selling their stablecoins, then the speculators end up bidding up the stablecoin price. In essence, the speculators need to pay a premium to deleverage under these circumstances, as we saw on Black Thursday.
These triggers are a reality of working in the cryptocurrency setting — they turn out to be relatively likely events over long time periods. Thus we can’t dismiss them or expect them to change; we have to adapt our designs to account for them. We can focus on extending the breadth of the stable region. How large this region is depends on precise market structure. Under idealized (unrealistic) settings, the coin is stable over a large region. Importantly, however, even ideal settings deteriorate under collapsed expectations. As a result, there are not enough speculators who want to issue stablecoins as it involves entering risky positions that they don’t view as profitable.
Our analytical results may be adaptable into data-driven risk tools for stablecoins, for instance to estimate probabilities of peg deviations and to infer about when barriers are crossed exiting a stable region. However, we note an obvious caution: specific numerical results are likely highly model-specific and sensitive to market structure and distributions of underlying assets. Our main contribution is in a theoretical understanding of these systems and how to dampen crises.
One Solution: Integration with Custodial Assets
The presence of deleveraging effects poses fundamental trade-offs in decentralized design. One way to bring the stablecoin closer to the ‘perfect’ stability cases is to increase how adaptive demand is to supply (elasticity).
Medium-to-long-term elasticity relies on the presence of good uncorrelated alternatives to the stablecoin. As all non-custodial stablecoins likely face similar deleveraging risks, greater elasticity relies on custodial stablecoins or greater exchangeability to fiat currencies. Short-term can be tricky, however, even if demand is in principle elastic longer-term. A significant portion of stablecoin supplies are locked in other applications, like lending protocols and lotteries. These applications promise (in some sense) value safety in over-collateralization, but don’t guarantee liquidity to withdraw. Additionally, Ethereum transactions cannot be executed in parallel; during volatile times, transactions can be delayed due to congestion, causing timely trades (especially involving transfer to/from centralized exchanges) to fail. This occurs even if, in principle, there is liquidity in these markets.
Another way to bring the stablecoin closer to ‘perfect’ stability is to increase the supply of marginal speculators who are willing to leverage and issue more stablecoins. As there will not be unlimited supply of speculators with positive ETH expectations (especially during an extended bear market), this relies on having another uncorrelated collateral asset. As all decentralized assets are very correlated, this again largely relies on including custodial collateral assets, like Maker’s recent addition of USDC. Note, of course, that custodial assets face their own risks, which may not be uncorrelated in all extreme crises. This includes counterparty risk, bank run risks, asset seizure risk, and effects from negative interest rates. However, there is likely some substantial diversification potential.
While these measures to integrate with custodial assets strengthen the stability results, it’s at the expense of greater centralization and moves the system away from being ‘non-custodial’. This leads to a seeming contradiction: while the goal is to make non-custodial assets, these can only be fully stabilized by adding uncorrelated assets, which are currently centralized/custodial.
Alternative: Non-Custodial Vault Insurance Pools
We propose an alternative: a buffer to dampen deleveraging effects without directly incorporating custodial assets. This buffer works by separating those who are willing to have stablecoins swapped to custodial assets in a crisis (in return for ongoing yield from CDP/vault option buyers) from those who require full decentralization.
The Maker system charges fees to speculators, part of which it passes on to Dai holders as an interest rate if the holder locks the Dai into a savings pool. With modified mechanics, this savings pool can provide a buffer to deleveraging effects. For instance, if we allow Dai in the savings pool to be bought out at a reasonable premium by a speculator who uses it to deleverage, then deleveraging effects are bounded by the premium amount up to the size of the savings buffer. The Dai holders who participate in this savings pool are then compensated for providing a repurchase option to the speculator. The Dai holder could elect to have the repurchase fulfilled in the collateral asset, or something else, like a custodial stablecoin. In this way, this mechanism can provide some of the benefits of the ‘perfect’ stability settings while enabling Dai holders to choose how decentralized they want to be. A Dai holder who does not require high decentralization would elect to receive the compensation from the savings pool whereas a Dai holder who requires higher decentralization would choose not to use the savings pool. Our model is easily extended to consider mechanisms like this.
The mechanism could be implemented similar to the chai contracts. In the first step, speculators subscribe to the vault insurance pool subject to a fee. In return for their subscription, they receive an option to repurchase Dai. This option entitles them to receive Dai to repay the borrowed Dai of their vaults in a time of crisis. In the second step, users can provide Dai into this insurance pool. The Dai in the pool are forwarded to the Dai savings rate to earn interest (when it is above 0% again); the pool can also earn interest from offering flash loans. Further, the users earn a second interest through the fees provided by speculators. The asset for the fee consists of the pool of assets provided by speculators, e.g. ETH, BAT or USDC. The interest rate can be adapted to the market mechanics of the pool: in times of high insurance demand but low supply, fees will be high to motivate users to join. In times of low demand, fees will be adjusted accordingly. Notably, through the use of atomic function execution on the Maker contracts we can ensure that the provided Dai can only be used to repurchase Dai by vaults currently subscribed to the pool.
Extends to Cryptoassets Beyond Stablecoins
The model/results (and thus also proposed solution) apply more broadly to synthetic and cross-chain assets and over-collateralized lending protocols that allow borrowing of illiquid and/or inelastic assets — whenever the mechanism is based on leveraged positions and leads to an endogenous price of the created or borrowed asset. Synthetic assets generally use a similar mechanism just with a different target peg. Cross-chain assets that port an asset from a blockchain without smart contract capability (e.g., Bitcoin) to a blockchain with smart contracts (e.g., Ethereum) also tend to rely on a similar mechanism. In decentralized (or rather, custodial with on-chain recourse) constructions such as XClaim and tBTC, vault operators are required to lock ETH collateral in addition to the deliverable BTC asset. They bear a leveraged ETH/BTC exchange rate risk and face similar deleveraging risk. In particular, to reduce exposure, they need to repurchase the version of the cross-chain asset on Ethereum.
Recap
We construct new tools for modeling stablecoins and other collateralized cryptoassets and demonstrate their value in understanding and preventing crises. We develop an alternative market-based mechanism to enhance stability in crises while remaining non-custodial. This creates a buffer that separates those who are willing to have stablecoins swapped to custodial assets (in return for ongoing yield from option buyers) in a crisis from those who require full decentralization. If built, this should improve the long-term stability and survival of non-custodial stablecoins.
See the paper for full details.
Special thanks to Dominik Harz for valuable comments and input on implementation with the Chai contracts and Georgios Konstantopoulos for providing valuable feedback.
