The significant drop in Ethereum prices, beginning at the end of July and culminating in a crash on August 5th, led to substantial liquidations on major borrowing and lending platforms such as Aave and Compound. On the Ethereum V3 Aave protocol alone, liquidations amounted to $200 million, with additional liquidations occurring on other networks like Polygon and Arbitrum etc., and other protocols such as Compound.
Introduction to DeFi Borrowing and Lending
DeFi borrowing and lending platforms allow users to borrow assets (typically stablecoins) against collateral (usually volatile cryptocurrencies) in an over-collateralized manner. When the price of the collateral drops, bringing the collateralization ratio close to or below the required threshold, a liquidation mechanism is triggered. This mechanism allows other users to purchase and liquidate a portion of the collateral at a discounted price, thereby restoring the collateralization ratio to a safe level. While this process incentivizes liquidators through a liquidation fee, it is detrimental to borrowers who lose part of their collateral.
Borrower Behavior Analysis
Borrowers can be divided into two categories:
- Active Borrowers: These borrowers consistently monitor and maintain their over-collateralization ratio (collateral value to debt) above the safe level to avoid liquidation.
- Inactive Borrowers: These borrowers do not actively manage their positions, making them more susceptible to liquidation.
Study of Liquidation Events
To understand the behavior of borrowers, I analyzed a set of liquidation transactions from August 1st to August 6th, with a total liquidation value of $116 million. By tracking the activity of liquidated users five days prior to the liquidation, I categorized them based on whether they actively managed their collateral ratios.
Out of the total liquidations, $53 million were incurred by active borrowers who had maintained their over-collateralization ratio above the protocol’s threshold in their last interaction before liquidation. This distinction among the borrowers is crucial because it shows that protocols could have prevented these liquidations by setting higher over-collateralization ratios.
Potential for Reducing Liquidations
To evaluate how protocols could prevent liquidations, I simulated scenarios with increased over-collateralization ratios over the five-day period and measured the impact on liquidations among active borrowers. The weighted average over-collateralization ratio for active borrowers across all Aave pools during this period was 1.33. According to my analysis, if the ratio had been increased to 1.52 (15% increase), more than half of the liquidations affecting active borrowers could have been avoided, preventing approximately $26 million of liquidations.
Figure 1: Collateral to Debt vs. Avoidable Liquidation.
To effectively address the issue, borrowing, and lending protocols should frequently adjust their parameters to adapt to market changes. Observing the 30-day volatility of Ethereum (ETH) prices over the past month indicates that volatility began increasing significantly from July 26th, suggesting a need for higher over-collateralization requirements.
Ideally, these protocols should automatically measure the risk of near-future liquidations based on price data and adjust parameters accordingly. For further insights, refer to our recent paper, “Thinking Fast and Slow: Data-Driven Adaptive DeFi Borrow-Lending Protocol”, which is an early attempt toward more adaptivity in Defi borrow lending. In this paper, we provide a closed-form expression for expected liquidation and we address the crucial issue of incentive-compatible, adaptive interest rate setting.
Conclusion
This analysis indicates that 46% of the liquidated users during the timeframe of August 1-5 were active borrowers. By adopting a 15% higher over-collateralization requirement, Aave could have significantly reduced liquidation costs for active borrowers, making the platform more favorable and attractive to them. This highlights the need for greater adaptability and data-driven approaches in DeFi borrowing and lending protocols.