Staked ETH and AAVE Risk (June 11th Update)

Staked ETH and AAVE Risk (June 11th Update)

stETH has emerged over recent months as one of the most popular collateral types on AAVE. Many users have deposited stETH and borrowed ETH against it. This means that if those prices diverge or the asset “depegs”, there could be a lot of liquidations in AAVE. There has been much speculation on the risks posed by a stETH/ETH price depeg on the protocol and we wanted to follow up on Paul’s last post to provide an update.

Late yesterday we saw a large spike in AAVE VaR due to volatility in stETH price and immediately investigated it further. TL;DR: Current market liquidity is very likely sufficient to handle expected liquidations in the protocol and no immediate parameter changes or risk mitigations are necessary. The AAVE community may want to temporarily freeze the stETH market and pause ETH borrows as an extra precaution. These steps can ensure that a liquidation spiral doesn’t negatively impact AAVE tokenholders and users.

Current Risk Levels

We have done some quick analysis to size the estimated potential loss Aave could experience if these accounts get liquidated and kick-off a liquidation spiral.

At current parameters for stETH (81% Liquidation Threshold) we would see $250M liquidated at a 15% depeg. You can see the full liquidation breakdown below. We don’t see any major insolvency risk (at current parameters potentially ~$2M with a >20% depeg), though do note that this is with current liquidity conditions and could be higher if a more severe depeg occurs.

Right now there are 3 large users that are at a higher risk of being liquidated:

The first account is the closest to the margin for liquidation at a health factor of 1.12. If there is around an additional 10-12% depeg of stETH/ETH price, then we expect this account to get liquidated (all else being equal).

We are in contact with these users (to the best of our ability to contact them) to ensure that they are aware of the risks posed here and that they are actively managing their positions to avoid liquidation.

How we can further mitigate these risks

There is inherent uncertainty in making market estimations like this. While we try to be thoughtfully conservative in our assumptions here, conditions (e.g. liquidity) can worsen over time and the protocol should consider taking action now to limit the risk here. This plan would be threefold:

  1. Freeze the stETH market

    • This will prevent new positions from being opened. This will temporarily pause the stETH market except for people winding down their positions. This is essential as it allows us to take the next step, increasing the liquidation threshold (LT) for ETH. If we don’t pause the market, users could increase the risk by opening new positions that take advantage of the increased LT.
  2. Increase LT for stETH to 90%

    • This will give existing positions more breathing room with regards to stETH/ETH price deviation (you can the see effect of this in the heatmaps above) and will make the protocol much more resilient to a further depeg.
  3. Pause ETH borrowing

    • If LT is changed, then preventing existing positions from increasing their risk exposure is required. To do this, you would need to prevent new ETH borrows from being opened. Almost all ETH borrowing is recursive borrowing and so while there is limited demand for ETH borrow, it’s still something users want.

As the risk in the protocol declines, we’d want to re-enable these markets as soon as we can. We’ll reach out directly to key AAVE stakeholders to get feedback here, but wanted to post this to the forums ASAP so the broader community can comment.

Proactive Measures to Mitigate Risk Levels in AAVE

We want to continue to work with the community to understand how we can balance the tradeoffs here as this spike in risk is something that we investigated taking action to prevent. These changes were withdrawn after community feedback that lowering LTVs and LTs would have a negative user experience for a select few users with low health ratios. However doing so would also ensure that the other 99% of users can safely and efficiently use the protocol, as well as that AAVE tokenholders will not be left holding the bag in the event of a market downturn. We’re looking forward to continuing the conversation on this in the future but do want to note that the above changes will not have the same user experience impact.

Next Steps

Our risk monitoring system continues to measure risk in AAVE and you can track the current risk levels on our dashboard here. As governance proposals take a few days to pass and go into effect, we are encouraging the community to consider the impact of the above changes carefully. If market conditions worsen we will strongly recommend that these actions be taken immediately.

16 Likes

Personally being following the situation quite close, I support waiting without doing any changes, my rationale being:

  • sETH/ETH dynamics are way more complex than their more visible secondary markets (e.g. Curve). It is pretty clear that as the price of stETH in secondary markets decreases, there is hidden demand that will price it back higher.
  • Those 3 positions (mainly the riskier in terms of HF) are clearly platforms plugin into Aave. Their sophistication in protecting their position is higher than average users, given that they execute decently complex strategies.
  • Even with important price drops of multiple assets, the so-called “de-peg” is at the moment around 4%, and there is not really any fundamental aspect to believe a major correlation would happen (it can of course, but by understanding how stETH works).

Regarding the proposed measures, my opinion:

  1. Freeze the stETH market. AGAINST. The “risky” positions in the protocol are already open, freezing the asset will only affect new positions being open. A correction here, Aave v2 Ethereum has an emergency admin role, currently the Guardian, which can freeze (stop borrowing/deposit) instantaneously, for situations of extreme need (probably not this, but important to take into account).
  2. Increase LT for stETH to 90%. MAYBE. I would not support 90%, it is too high, but it is the factually effective measure, as will make currently open positions safe. But this is a major change, once higher, putting it down is a major endeavor, and there will be almost no other chances over 90% (or even 85%).
  3. Pause ETH borrowing. AGAINST. Freezing stETH is probably more effective, and this could potentially affect other parties borrowing against other collaterals.
3 Likes

I also do not support freezing the markets. As long as the potential damage is bearable, I think the markets should be kept up as I believe one of the primary purposes of AAVE is precisely to allow this kind of speculation and market activity.

Also, maybe I don’t fully understand the solution, but I think increasing LT could make liquidations more dangerous since the liquidation bonus is so high. A 90% LT with a 7.5% LB would mean that there’s really not much margin left, and basically all equity would be liquidated with these parameters. Would this not mean that a drastic change in price would make positions insolvent much quicker in this scenario?

After careful review and monitoring the stETH situation, I’ve noticed the “de-peg” hovering mostly in lines of 4-5% between ETH and stETH. Effectively once the efficient arbitrage is available, this de-peg most likely will decrease substantially (when ETH staking would be two-way stream - arbitrageurs would be able to buy stETH unlock the underlying ETH with profit when stETH is trading below the peg). I would not expect them to trade completely in pair as there is always some additional risk involved with derivative assets, which is acceptable.

The bigger positions concern me only to the extent that we don’t have much information on how these more sophisticated users would recapitalize their positions. We have seen many sophisticated users blow their positions up since they might use all their available liquidity or might not be willing to take the loss at this point.

Since Gauntlet is in contact with these users would be interesting to know @jmo of their risk mitigation plans which would help to assess the measures to be taken. Hopefully their measures are leaning towards providing additional collateral instead of selling stETH on the market to cover their ETH borrows since this would cause additional down pressure on stETH and increases the risk for the de-pegging.

Would not mind to increase the LT to 85% as it would bring wider breathing space to avoid liquidations together by decreasing the max LTV for new positions to avoid creating additional risk in case the appetite to borrow ETH against stETH comes back mid-term (and would help later to decrease it when the effective arbitrage is in place).

Additionally, would be interesting if Gauntlet could analyse the Curve liquidity on stETH and the sizes of those positions, given that any bigger position exit from stETH could actually further decrease the liquidity (increase the de-peg) and affect the ability to liquidate for these positions that are closest to potential liquidations.

6 Likes

In a protocol like Aave where bad debt effects all pools/users I’ve been surprised and extremely disappointed to watch this play out. As systemic risk builds you continue to loosen stEth’s risk parameters. It seems Aave has managed to recreate the system of socialized bailouts - increasing risk for all users to try to protect a few whales. Because of this around a week ago I closed all my positions from Aave for the time being. I am a hard NACK on changing anything at this point - I already think you’ve made multiple missteps and this is just another bad decision trying to play puppeteer. The terrible decisions IMO:

a) not enabling borrowing on stEth in the first place is very suspicious to me. If borrowing was enabled from the get go like all other assets we wouldn’t be here in the first place. Anchor drove a lot of demand for stEth in the first place bc eth needed to be staked as stEth to be used as collateral there. In fact we saw this new demand establish a tight 1:1 “peg” between eth and steth for the months of Anchors parabolic run. This probably gave a lot of people false confidence because prior to that it never traded 1:1 (as you would expect). Now let’s imagine Aave allowed borrowing: a lot of market participants likely would have borrowed stEth against Eth to farm in Anchor instead of having to mint new stEth. Meaning we wouldnt have as much new supply for the Anchor farming. We also would have had a free market where people took the other side of the trade shorting, maybe there wouldnt have been so much false confindce in a mythical 1:1 peg, and also as the peg broke we’d have lots of buying demand as shorts covered. Unfortunately bc of this suspect decision we find ourselves in a much worse place now

b) you continue to loosen risk parameters and liquidation thresholds for a risky asset. Currently Lido DAO has the ability to arbitrarily change stEth balances. So you have a very centralized Lido DAO (~$300M market cap?) w/ effectively full control of a $5B+ stash of Eth? Not to mention other smart contract, slashing, etc… risk in a very new and complicated system. Now you want the liquidation threshold of stEth to be more friendly than Eth itself??

We all know the rules when we open positions and none of us would expect Aave to change risk parameters to prevent us from being liquidated. It is pretty simple - post more collateral, unwind the position slowly, or risk getting liquidated. Don’t try to change the rules like this is Tradfi. Just putting this out there - I am totally out of Aave now so I’m interested to watch this play out. This is my 2 cents and I know the decisions makers here have more skin in the game and knowledge of the Aave protocol + long-term goals. I’m just left scratching my head on what I view as a terrible sequence of decisions when it comes to stEth

2 Likes

Increasing the LT is a terrible solution. It could put AAVE users in an even more dire situation in case of further depegging and would basically be letting large borrowers bully the protocol. I am surprised that this is considered and hope that it is not due to personal connections.

2 Likes

@bred has a solid point concerning the raise of LT, that is why I’m, if not completely against a small raise, but highly skeptical.
90% should be completely out of question, it’s 5% over ETH itself. 85% is still probably too high, being the same as ETH, and factually only done to have some more protection on the current opened position, which is not really a valid rationale if it doesn’t protect against an important system-wide risk.

In my opinion, the protocol/market should be able to bear a potential liquidation of the 2 most “risky” positions, accounting for ~200’000 stETH, but which probably would be liquidated progressively. That assuming those parties don’t find any solution to protect their position in that scenario (not so likely, hopefully for them).

On one thing I agree, the previous raise of LT stressed out already the margin of protecting positions, done exclusively because with the first “depeg”, it was important to activate some lever. But we should be mindful that all rest of the users of the protocol should not get affected by a small subset taking a bit more risk.

P.S. about not enabling borrowing, it was chosen by the proposer, both because they didn’t see an important use case and because of the required technical changes on the debtToken ARC: Add support for stETH (Lido)
I don’t think precisely right now is the appropriate moment, but I agree that should be enabled in the future, once the market is a bit more stable.

3 Likes

I’ve also been following the situation closely. All though the risk levels seem bearable, I support some precautionary measures to mitigate risks of deterioration

The StETH market is highly concentrated on weakened users:

  • The top 10 holders supply 70% of the liquidity
  • Celsius, currently under liquidity crush suspending activity, is Aave’s top stETH liquidity supplier with 28% of the pool; they user stETH as collateral with 20% WBTC
  • Aave’s risk mitigation cannot rely on trusting users

Screenshot 2022-06-13 at 11.32.05 Curve.fi ETH/sthETH shares

stETH exit liquidity on Curve.Fi is reducing :

  • Share of ETH dropped to 20%
  • lack of liquidity may prevent liquidations
  • lack of liquidity may further stress the peg
  • if this liquidity pool dries out, depositing on Aave and borrowing against would become one of the last options to get liquidity of StETH holders

I am in full support freezing the stETH market to contain Aave protocol’s exposures as stETH is already one of the most popular collaterals, with active borrowings against >$1B of stETH. Liquidations on this asset have become less easy and this may continue to deteriorate. Freezing deposits allows us to impose an exposure ceiling similar to v3 supply caps :white_check_mark:

Not sure about loosening stETH risks parameters to 90% LT which would leave little margin for liquidators to profit with the 7.5% LP increasing the risk of bad liquidations. It’s also difficult to unwind such a high LT without liquidations. If StETH risk increases, risk parameters should reflect this, one way could be to increase the liquidation penalty to incentivise smooth liquidations :x:

With regards to pausing ETH borrowing to prevent users from taking on more risks, as most of the ETH borrowing is recursive with StETH, it shouldn’t affect protocol revenue much. I support this temporarly to contain the risks while there isnt much borrowing activity to be affected :white_check_mark:

3 Likes

Regarding the proposed measures:

1- 3- I agree that we should freeze both deposits of stETH and borrow of ETH to slow the buildup of risk in the protocol and exposure to stETH in this current market.

2- I disagree with the increase of LT.

LT should remain the same as it is now, stETH is a derivative of ETH and the risks associated with it should be reflected in the LT, bringing it up to the same level as ETH makes no senses and it should not be adapted ad hoc for whales in periods of high volatility and risk.
Looking at @jmo 's graphs it seems that even with a depeg of 20+%, refinancing the potential insolvencies would cost ~5% of the budget dedicated to refinancing in the safety module which is bearable.
I think that as @Alex_BertoG mentioned the LB should actually be increased to make sure that in worst case scenarios the liquidation of these big positions is still profitable and can be done smoothly.
As it stands the two unhealthiest positions highlighted by @jmo : DeBank | Your DeFi wallet and
DeBank | Your DeFi wallet
are ~200k stETH which the curve pool could support if the position are liquidated by a sophisticated actor over time, which a better LB might incentivize.

2 Likes

I don’t think a higher LB is guaranteed to be beneficial for smoother liquidations. The LB can actually be thought of as a discount that liquidators get to buy the collateral asset from AAVE with the debt asset from the oracle price. A 7.5% premium is already quite a lot, and increasing this premium can actually be counterproductive to successful liquidations on a macro scale. Therefore, I think the premium should generally be tuned to account for how much we believe that oracle prices can be off by in a big market downturn, and to compensate for high gas costs. Any excess premium paid will typically just be sold by the liquidators, with no net positive for anyone in the protocol besides liquidators.

IMO what keeps the protocol safe is low LTs, as lower LTs increase the equity available before and after liquidations. More equity means more breathing room for liquidations before positions become insolvent.

Thank you to everyone for the suggestions and additional considerations. Below Gauntlet will address the concerns expressed above and outline our next steps.

The table below includes updated modeled liquidation and insolvencies using current parameters for stETH (81% Liquidation Threshold)

stETH/ETH Price deviation Liquidations (Millions, USD) Insolvencies (Millions, USD)
15% $144.14 $ 0.0
20% $ 251.78 $ 3.96
25% $ 244.76 $ 19.84
30% $ 271.78 $ 51.01

Updating the stETH Liquidation Threshold to 90% below:

stETH/ETH Price deviation Liquidations (Millions, USD) Insolvencies (Millions, USD)
15% $ 6.44 $ 0.0
20% $ 6.06 $ 0.20
25% $ 195.31 $ 12.75
30% $ 228.44 $ 42.34

Note: The liquidation amount is not monotonic. When there are cascading liquidations the liquidations are marked in USD terms. If there’s a large price deviation, all the liquidations happen at the same price X, but if there is a slightly smaller price deviation, some liquidations will happen at a higher price 1.2X, and more will happen at a lower price 0.9X, which could lead to a larger liquidation USD total (though the token total of stETH liquidated is less), even though the initial price deviation was smaller.

Follow-ups:

Increasing LT avoids “too big to fail” scenarios like significant liquidations due to stETH/ETH price deviation. The tradeoff, you have correctly identified, for this approach is that it leaves little margin for liquidators.

The major concern is stETH/ETH price deviation and the recursive positions. The price we are looking at here is the stETH/ETH price, which has lower volatility than the stETH/USD price.

In AIP-75, we lowered the stETH LTV from 73% to 69%, so “new” users would not be able to open more risky positions than the existing users.

Risk to stkAAVE:

As a refresher, if an Aave market reaches a state where its liabilities (insolvencies) are greater
than its assets (collateral), then the AAVE safety module covers the shortfall of all markets voted by the community. Currently, the safety module holds $118M is stkAAVE. Up to 30% ($35M currently) of the safety module can be slashed at once.

Next Steps:

With regards to the mitigation plan, suggestions have been made to implement some changes and not others. The risks we are observing are asymmetric for the protocol and users.

Assumptions:

  1. Freezing stETH and disabling ETH borrows limit further risk to the protocol.
  2. Gauntlet has taken reasonable steps to notify the relevant whale accounts at risk of liquidation.
  3. We are ignoring second order effects of our recommendations adding to FUD that might send the stETH/ETH price lower or interpretations that increasing LT means that these we endorse these accounts as “too big to fail”
  4. By increasing or maintaining the LT value we are not increasing the chances that the whale accounts will be liquidated.
  5. Thus, we should try to minimize the expected amount of insolvent debt for the protocol.

If liquidity decreases by 30% (based off the ADV change in the last four days against price deviation percentage) going from a 81% to a 77% price deviation, then the EV of insolvency for 90% LT and 77% price deviation is estimated as 100k (1.01MM VaR * 0.1), assuming a 10% chance of stETH/ETH reaching 0.77. The EV for 81% LT and 81% price deviation (VaR at 85% price deviation is negligible) is 321k (1.61MM VaR * 0.2), assuming a 20% chance of stETH/ETH reaching 0.85.

Gauntlet’s recommendation is to move to a governance vote with all three of the original risk parameter changes.

6 Likes

Due to the atypical risk stated above, Gauntlet has moved directly to AIP. We strongly encourage continued conversations. Our team remains on-call and closely monitoring market conditions.

3 Likes

Hi Aave community :wave:

This post has been written by @funkmasterflex and @MatthewGraham regarding Index Coop’s icETH product. The performance and leverage ratio for icETH can be found here. Do note, the leverage ratio has been reducing over time due to market conditions. Index Coop has been proactive in managing the holders of icETH exposure to the price of stETH on secondary markets, along with the cost of borrowing ETH on the Aave v2 market.

We are strongly in support of the Aave community being proactive in managing risk to its users. icETH makes up 3.43% of the ETH borrowing demand. With regards to icETH and stETH spot price, we would like to highlight the design of our product and the TokenSet infrastructure it is built on is capable of managing/handling the risk relative to the secondary market price of stETH. We have a lot of experience doing so via our FLI Flexible Leverage Index product range on mainnet and Polygon.

However, pausing stETH deposits, pausing borrowing of ETH and increasing the Liquidation Threshold (LT) for loans backed by stETH is not ideal for the icETH product. This creates a lot of operational challenges for us. It would lead to a supply cap being implemented which would then need to be revised lower as the units in circulating supply fall in time with users redeem icETH for the underlying constituents. We rely on the ability to mint and redeem icETH, which automates the depositing and borrowing of stETH and ETH respectively, to hold the product to Net Asset Value (NAV) on secondary markets. Arbitrage traders mint or redeem icETH when arbitraging the spot price on secondary markets to NAV. With Lending of stETH and Borrowing of ETH disabled, icETH’s spot price will move in line with market demand / supply which can lead to a really poor UX.

We are currently discussing how to vote on this proposal. icETH relies on the same Curve pool for liquidity to rebalance icETH and we are acutely aware of the challenges this pool is facing.

For icETH changing the LT up or down is something we can accommodate easily. However, not being able to use stETH as collateral or borrow ETH puts us in a less than ideal situation.

We would like to offer our services to anyone who is currently running this strategy to generate yield in migrating to a platform with an automated solution for managing this position. For further info on how we automate the rebalancing of the strategy, please do reach out via direct message. Naturally, when Aave v3 and E-Mode arrives, we will be able to accommodate this as well. We are here to be a part of the solution in making this strategy safe for all users.

5 Likes

A comment here on the decision to not allow stETH borrowing;
The decision was taken by the Lido team at the time the proposal was created to avoid introducing further risk of unforeseen attack surfaces. As seen with xSUSHI where the primary surface of attack was through leveraged borrowing, and given the novel implementation of astETH and the peculiar nature of stETH as asset, it was impossible for them at the time (and for us community developers that assisted in the listing process) to assess whether or not allowing borrowing would open up a similar attack surface or other unforeseen risks. Given the fact that demand for stETH borrowing would have been in any case limited (besides the anchor use case that you mentioned and honestly i dont know how that would have played out) decision was taken by the lido team to not allow borrowing, and the community voted for it. That’s pretty much all there is to it. I believe it was mentioned in the proposal text, but i may be wrong.
Regarding the proposal, you are describing it as something that has already been decided while this was, and is, just a proposal that needs to be voted and executed on chain first. Currently it seems the majority is voting against. So i very much disagree with the idea that the Aave community is trying to bail out whales and socialise losses. As a variegated community with different expertises and risk appetites, entities may have different opinions on how situations should be handled.

3 Likes

AFAIK, the V2 market which is the one in question does not actually have the capability to enforce LTV, as evidenced by the existence of some instadapp/defi saver like features. Can we really consider LTV to be something keeping AAVE safe?

When time permits, Gauntlet prefers to discuss, poll (Snapshot), and then move to AIP. We always want to facilitate a better understanding of market risk and work with the community to reach a consensus. In this instance, our analysis indicated it was prudent to start the AIP process while continuing the discussion in this community and others (Maker, Index).

Updated analysis with regards to the stETH/ETH price deviation and largest user positions no longer necessitates the AIP-83 changes. We request abstaining, rather than canceling/NAY. We will continue to monitor the situation and notify the community should things change. We are leaving the proposal live for the same reason it was put it so quickly—to have a faster mitigation path should conditions deteriorate.

Note the updated liquidations/insolvencies in the heat maps.

4 Likes

The heatmaps below are generated using market data from 30 minutes ago. Gauntlet continues to observe reductions in overall risk to the protocol.

2 Likes

The heatmaps below are generated using updated market data.

3 Likes

Can we access the data used to create these heat maps? If not, can you provide a table of liquidation prices and insolvencies at different % deviations for eth/steth like you had provided a few days ago?

1 Like

June 18th 23:00 UTC Update

Proposal AIP-83 has failed as per our updated recommendation above. Gauntlet continues to monitor stETH and the broader market.

The implied liquidation of WETH against 0x4093 has a negligible net impact on stETH/ETH price deviation—including liquidity implications for the Curve pool.

The table below includes updated modeled liquidation and insolvencies using current parameters for stETH (81% Liquidation Threshold)

stETH/ETH Price deviation Liquidations (Millions, USD) Insolvencies (Millions, USD)
15% $ 38.04 $ 0.18
20% $ 86.92 $ 0.37
25% $ 89.99 $ 24.24
30% $ 88.02 $ 67.11

stETH Liquidation Threshold at 90% below:

stETH/ETH Price deviation Liquidations (Millions, USD) Insolvencies (Millions, USD)
15% $ 0.17 $ 0.18
20% $ 31.38 $ 0.32
25% $ 53.09 $ 11.24
30% $ 78.61 $ 51.03

The heatmap below renders the risk of liquidations similar to the previous heatmap. The main difference, with the 81% LT for stETH unchanged, is a significant increase in the potential for insolvencies.

4 Likes