Vault User Research: Intra-Protocol Exposure Migration
Studying exposure movement within MakerDAO vault types
This report was originally published on Maker Forum on October 20th, 2022
Introduction
In our previous report we analyzed vault exposure migration from stETH-A to stETH-B vault type because of a decrease in stETH-B’s stability fee. We also looked at its potential impact on revenue cannibalization. Our key goals were to (i) create an initial methodology for quantifying vault exposure migration, (ii) bring up revenue cannibalization across different vault types in the same collateral symbols as a topic, and (iii) surface its importance to the community given that it’s likely to recur in the future when deciding on parameter setting.
While the above results already showed interesting insights, we decided to look further into vault exposure migration through a more generalized, high-level overview.
In this analysis, we focus only on intra-protocol migrations (exposure movement within Maker system). This is aligned with the long-term approach to scientific governance at MakerDAO. This means making systemic behavioral assumptions explicit by formulating them as research questions and validating them with empirical data.
More specifically, it was previously unknown how common was vault exposure migration throughout the system and a deep dive into that aspect of the system dynamics can enable better informed governance decisions. This report aims to quantify historically migrated exposure across different vault types of major collateral symbols (ETH, stETH, WBTC).
Analysis
In the analysis, we assume that one proxy wallet address equals one user. While blockchain data is open, there is no reliable way for us to estimate how accurate this assumption is. There are exceptions with wallets with large exposure (whales) which we actively monitor for which we merge multiple proxy wallets into a single entity (such as 7 Siblings).
We begin by analyzing directly migrated exposure which is detected when a vault owner moves their exposure atomically (within the same transaction). This most often means repaying debt and withdrawing collateral in one vault type and then opening/boosting/depositing in the second.
Below we can see total monthly migrated amounts aggregated based on migration direction (from which vault type to which vault type the exposure was moved), split into separate charts based on their collateral symbol. The amounts are relatively low compared to overall risky exposure which currently stands at around $880 million. For ETH-collateralized vaults we see up to $25 million of monthly migrated exposure, for WBTC-collateralized vaults up to $7 million and for stETH-collateralized vaults up to $3.5 million. That’s most likely because these kinds of migrations are usually executed by vaults that are protected with services such as DeFi Saver which (i) are usually not used by whales which could contribute to larger amounts and (ii) a small portion (5%) of current risky exposure is using their services.
For ETH-collateralized vault exposure we can see that ETH-A to ETH-B and the inverse seem to be commonly executed migrations.
We continue by combining both direct (atomic) and indirect (non-atomic) exposure migrations. To build an intuition on how we detect those, we provide an example of an individual proxy wallet (DSProxy address) which managed multiple ETH-collateralized vault types in the past. While the chosen proxy wallet had ETH-A, ETH-B and ETH-C exposure historically, it currently has exposure only in ETH-C.
The usage of the three vault types is evident in the below chart which shows exposure over time. We get a glimpse that there could be some exposure migration in the period between May 2022 and August 2022.
That becomes evident when we aggregate monthly exposure difference per vault type. There are two monthly aggregation instances when the vault owner migrated their exposure, in June and July 2022. In June, they moved their exposure from ETH-C to ETH-B and in July they moved it back, from ETH-B to ETH-C. This period didn’t include any changes to liquidation ratios or stability fee changes in these vault types that could influence this so there must be some other exogenous factor that we don’t have data on. This shows how nuanced measuring causal impact of different parameter changes can be when we investigate individual data points.
In general, the proxy wallet owner closed their entire ETH-A position in January 2022, returned in May 2022 by minting ETH-C exposure, migrated their exposure in two consecutive months and after that slightly increased ETH-C exposure in September 2022 and decreased it in October 2022.
We quantify the migrated exposure by taking the monthly difference in the principal amount of the vault type to which exposure was migrated. The results of migrated exposure amount and migration direction can be seen below.
We next turn to aggregating migrated exposure across proxy wallets with more than $50,000 in current total exposure which opened a ETH/stETH/WBTC-collateralized vault in the past. This filtering is made because it’s unlikely to be economical for small vault owners to manage their vaults as actively during times of high gas cost. Currently, the chosen sample contributes to more than 90% of current risky exposure collateralized by these three collateral symbols.
To give some context on cross collateral symbol migration, around 10% of proxy wallets that open a vault in one of the major collateral symbols also open more than 1 vault type. This shows that most proxy wallets tend to not open multiple vault types collateralized by a single symbol which would enable them to migrate their exposure based on the change in stability fee or change in their individual risk profile.
As mentioned above, it’s also possible (or likely) that highly skilled vault users change their wallets / proxy wallets regularly. It’s a potential future research topic to bundle multiple wallets together to get a better approximation of 1 address = 1 user assumption.
Additionally, 15% of the chosen proxy wallets (85) performed an exposure migration this year. This share drops rapidly with a decrease of total proxy wallet’s exposure, as expected given that it is more economically feasible for well capitalized vault owners to move exposure due to changing conditions.
For ETH-collateralized vaults we can see that migrations this year are highly dominated by a single whale. This entity’s migrated exposure comprises a $222 million ETH-A to ETH-C migration in May 2022, a $183 million ETH-C to ETH-B migration in June 2022 and a $171 million ETH-B to ETH-C migration in July 2022. Since then they’ve mostly been exposed to ETH-C vault type (they contribute to about ⅓ of ETH-C’s total exposure).
In total, 57 proxy wallets migrated their ETH-collateralized exposure (10% of total in the sample), around 5–15 uniques on a monthly basis.
The pattern is similar for stETH-collateralized vaults where the vast majority of migrated exposure was done in July 2022, from stETH-A to stETH-B (aligned with our previous research).
In total, 13 proxy wallets migrated their stETH-collateralized exposure (2% of total in the sample), 2–4 uniques on a monthly basis.
Given that the above mentioned whale does not have an active exposure to WBTC-collateralized vaults we don’t see the same pattern in the below chart. Meanwhile, out of 32 migrations detected in the chosen period, 8 of them migrated exposure above $1 million. In total, 21 proxy wallets migrated their WBTC-collateralized exposure (4% of total in the sample), 3–7 uniques on a monthly basis.
We can also remove the dominant whale as we look at the migrated amounts in a filtered view. It’s immediately clear that the amounts vary a lot across different months, directions and also collateral types. As expected, ETH-collateralized vault types show the largest amounts of migrated exposure, followed by WBTC and stETH.
For stETH, stETH-A to stETH-B migration tends to be much more common than the inverse. That’s aligned with the intuition given that stETH-B’s risk premium was recently decreased to 0%.
Migrated exposure amounts also decreased significantly in the last few months which is expected given the system-wide increase in CR buffer (collateralization ratio over liquidation ratio). This shows more focus on risk mitigation by lowering capital efficiency to avoid liquidations caused by potentially incoming further market shocks. This showed to have a stronger effect compared to recently more affordable transactions across the Ethereum ecosystem which could encourage certain users to more actively manage their vaults.
We continue by moving away from quantifying migrations explicitly. Instead, we look at the dynamics of total debt per vault type across the three major collateral symbols. In the charts below it is clear that there is an increasing dominance of high collateralization ratio, low stability fee vault types. In the last 6 months, ETH-C’s share of total debt has increased from 14% to 57%. WBTC-C’s share of total debt has increased from 10% to 51%. And stETH-B’s share of total debt has increased from 0% (wasn’t launched 6 months ago) to 70% today.
So why is there a clear change in exposure per vault type while there was limited detection in migrations? The chart below shows a distribution of current exposure for these low fee vault types, split by the cohort date (month of their first interaction with Maker). We can see that apart from 7 Siblings (the vast majority of the proxy wallet exposure that started using Maker between November 2019 and January 2020), most of the exposure is relatively new. This means besides the dominance of this individual whale, most of the exposure that was brought into low fee vault types came from new user wallets, also when we’re looking at only unique wallets.
Conclusion
In this report, we show that intra-protocol migrations are not that common. Most vault owners tend to stick with their initial vault type without moving their exposure to another vault type because of better borrowing conditions. Meanwhile, the migrated exposure amount can be significant because of the dominance of large users.
While previously the reason for less vault migration could be large gas costs, we are currently in a mostly negative (bear market) sentiment which caused vault owners to be more cautious with optimizing their capital efficiency at the potential expense of liquidations. This is also reflected in the currently low Capital at Risk compared to Surplus Buffer. While this is not ideal for Maker’s bottom line, it can definitely contribute to its long-term resilience. We also showed that in the last 6 months, a large share of exposure moved to low fee, high liquidation ratio vault types, dominated again by the 7 Siblings whale and new users (proxy wallets). This indicates a strong sensitivity to stability fee setting which caused the large users to migrate and for new users to open vault types with lower stability fees.
Our potential future research direction includes looking at cross-protocol exposure migration, taking into account flows across other lending protocols such as Aave and Compound. Especially interesting research direction could be to investigate how much of the new low fee exposure migrated from other lending protocols.
