Impact Measurement: Aave
Gauntlet has executed 17 proposals during our current Aave engagement, the first of which was executed on Sept 4, 2021 and the most recent of which was executed on Oct 1, 2022.
The chart above shows the time series of Liquidation Threshold (LT) changes for each asset during this period. Since most parameter changes in this period were LT increases, the impact study will focus on the increase in borrows and borrow interest as a result of these. Our changes also helped mitigate liquidation losses, which we will touch on later, and reduced the protocol VaR. Overall our recommendations resulted in net LT increases for the majority of assets, and no net decreases in LT for any assets, as seen in the chart below.
Since we seek to improve capital efficiency without adding risk to the protocol, it is important to highlight that all these changes occurred with no insolvencies through significant market volatility. A full recap of our risk mitigation during the downturn can be found in our January and May Market Downturn Reviews.
Expected Impact
When LT increases, we expect users to take advantage of some of their additional borrowing power. Thus, we can think of the expected impact as the increased usage that we expect when the change was made. As changes go live, we can then calculate how much of this additional usage is realized, which depends on market conditions and user behavior.
The assumption that borrowers will increase their borrows to take full advantage of their increased borrowing power as a result of our parameter recommendations is what we refer to as elasticity. A fully elastic borrower in our framework maintains utilization at a constant percentage of their limit, adjusting their position frequently to keep up with parameter changes and market moves. If we can confirm this holds in the user data, the additional borrow due to a recommendation can be calculated as below:
Where the utilization in this formula is a stable percentage of the available borrowing power, the supply is the supply of assets affected, and both are multiplied by the net change in LT. Since the interest rates on any given day are known, it is then straightforward to convert this borrow impact into the additional interest generated for protocol suppliers and reserves. To present our realized impact, we will first show the results assuming elasticity, then proceed to validate that real users adjusted their positions frequently and were indeed elastic.
User Analysis
Assuming full elasticity as described above, we find that our recommendations generated an average of $124M of additional borrows, which resulted in $3.70M of additional realized borrow interest between Sept 4, 2021 and Oct 1, 2022. The chart below shows the additional interest generated, with the contributions of some of the most impactful users broken out. The dark blue (0x8ac) segment corresponds to the impact from the highest impact user, the orange segment (0x3dd) corresponds to the impact from the second highest impact user, and so on, with the light blue (remainder) piece at the bottom aggregating the impact across the remaining smaller users.
Since the top users accounted for a material share of the total impact, let’s zoom in on those to validate their elasticity.
User 1: 0x8ac
The charts below show the highest impact user’s behavior over time. The first two series are the supply and borrow composition, respectively, and the third series is the Utilization (i.e. the fraction of the borrowing limit used). The vertical green lines indicate the dates of relevant parameter changes where LTs of symbols this user supplied were increased. As we can see, the user tends to hover around a utilization of 50% throughout the course of their interaction with the protocol. Sometimes utilization increased to about 70% or decreased to about 30%, but these levels were not sustained due to the user frequently adjusting their position, indicating this user was indeed elastic.
Similarly we can see that the next two largest accounts shown below are also elastic. Though the total supply and utilization vary over time, all of these users adjust their positions frequently and are clearly adapting to parameter changes.
User 2: 0x3dd
User 3: 0x409
Mitigating Liquidation Losses
Separate from our impact on borrows and borrow interest, our reduction of Liquidation Bonuses resulted in significant savings of collateral awarded to liquidators during our engagement. Based on liquidations that did occur, we observed realized borrower savings of $7.36M which would have been rewarded to liquidators. Liquidation bonuses were reduced across multiple assets, sometimes more than once, without sacrificing protocol risk.
Closing Thoughts
There are several other ways we impacted the protocol, which we can quantify to varying extents. Notably, our analysis has only discussed the impact realized to date. Assuming our recommendations remain in place, they will continue to generate impact into the future, which we do not attempt to measure here.
Finally, we did not try to measure the effects of capital efficiency improvements on aggregate TVL. Since it is difficult to know to what extent this occurred, our methodology takes total supply as an input and does not assign any impact to probable user growth or retention. Similarly, we did not attribute any impact to new listings (such as FEI) attracting users to join Aave. While these forms of impact are harder to put a number to, it is important to note that they exist even if beyond the scope of our quantitative framework.
Looking ahead, Gauntlet continues to refine simulation quality metrics, including weighted liquidation recall and precision. After every set of parameter changes, we run retrospective models to fine-tune our simulated environment against actual observed user behavior and liquidations. By checking that our models accurately replicate historical activity, we can further increase our impact and ability to measure it.
