[ARC]: Implementing a GHO Backstop for Aave

Risk Liquidation
1

Short Summary

As GHO development is in progress, B.Protocol would like to propose Aave DAO add a user-based backstop for GHO that will enable safer and more robust liquidation mechanics that can ensure peg stability also for less liquid collaterals as well as during radical market conditions.

As the value and stability of GHO are dependent on the collateral assets that are backing it, the accrual of GHO’s bad debt is riskier than other, more “standard”, bad debt as it can cause GHO to lose its value - hence the proposal to set up a backstop as another line of defense to protect its stability.

Users will be able to deposit their GHO into a backstop pool, where the liquidity will be used to facilitate liquidations with a fixed slippage.

Motivation

B.Protocol provides B.AMM (Backstop AMM) integrations for lending markets and over-collateralized stables. It’s a novel DeFi primitive that uses provided liquidity to liquidate underwater positions without relying on flashloans, acting as a buffer that amplifies existing DEX liquidity. B.Protocol was launched in 2020 and had over $300m TVL at the bull market peak.

B.Protocol pools users’ funds into segregated Backstop pools. This liquidity is used for liquidations as they happen on the integrated platform, e.g. Aave. Once a liquidation takes place, the Backstop AMM (B.AMM) smart contract pulls the needed funds from the backstop to facilitate the liquidation and automatically puts the seized collateral for sale according to a market price feed. Once sold, the return is deposited back to the backstop pool, and profits are accrued.

The backstop enables GHO holders and community members to help protect Aave as a platform from bad debt while strengthening GHO stability and gaining from liquidations. You can read more about the B.AMM on our docs or directly on the (more technical) whitepaper.

Hi-level design of the Backstop -

A GHO backstop will ensure liquidations of even less liquid collateral assets on Aave can be made smoothly, or at least without exposing the Aave community to any bad debt from GHO liquidations, as the backstoppers who provide liquidity to the pool are willing to take the risk of potential losses, in return for the potential liquidation profits.

Currently, on Aave V2 there are a few assets with millions of USD value as collateral, but with very low DEX liquidity (see table below). Flashloan liquidation bots won’t be able to liquidate even medium size positions in one transaction and will have to break it into a few “chunks”, waiting for new liquidity to flow in by arbitragers. This process may take time and can expose the protocol to bad debt in case price keeps on dropping during this time period. Having a dedicated backstop offloads that risk from the protocol lenders and puts it on the backstop. You can read more about how a backstop improves illiquid assets’ liquidation process in this blog post.

Defending Aave from bad debt which is created by GHO (users’ borrowed GHO value is higher than the value of their deposited collateral in Aave) is more critical than other types of bad debt accrual in the Aave platform, as it can also affect the value of GHO as a stable. Assuming Aave wishes to maintain its competitive LTV values in the ecosystem, we believe a backstop for GHO will significantly improve the liquidation mechanics also during severe market conditions and will ensure GHO maintains its value while keeping the relatively high LTV values it offers to its users.

Asset TVL ($) LTV (%) Liquidation Penalty (%) Max liquidation size* ($)
BAT 4.85m 65 7.5 250k
ENJ 3.54m 60 6 150k
MANA 17.19m 61.5 7.5 450k
REN 4.38m 60 7.5 200k
xSUSHI 17.13m 50 8.5 100k
ZRX 14.03m 55 7.5 150k

Aave V2 Ethereum market assets with low DEX liquidity
*Max liquidation size possible with a price impact smaller than Liquidation Penalty (according to 1Inch, November 6th, 2022).


We believe that having a GHO backstop will not only enable the listing of less liquid collaterals but can also foster better trust in a new stable to hold during severe market conditions, as more robust liquidation mechanics will maintain a better GHO peg.

Integration Requirements

In order to kick-start the backstop pool, GHO holders will need to be incentivized to deposit their GHO in the backstop. We suggest 2 different options to incentivize the backstoppers:

  1. Allocate $3m as an incentive to the backstop pool depositors, calculated as 5% APY for a $60m pool size.
  2. Set the backstop as another eligible source for discounted interest rates for GHO. bGHO (GHO tokens in the B.AMM) holders can be set to be eligible for a discounted GHO interest rate similar to stkAAVE holders (with different parameters as the community will decide).

Next Steps

Upon a positive feedback signal from the community to support this proposal and the preferred method to incentivize GHO backstoppers (either one of the above-listed options or a third one preferred by the community), B.Protocol will make the development efforts for this integration, including any other requirements coming from the Aave community or the Aave Companies team who are leading the GHO development efforts.

The B.Protocol team is looking for any feedback on this proposal from the Aave community.

5 Likes
2

This seems more a topic to comment on by the Aave Companies developing GHO, but given the current pre-approval by the community, some questions, more generally about the backstop than its GHO-specific usage:

  1. Is the backstop supposed to enter into action on any type of liquidatable position below 1 HF? Or just on positions fully undercollateralized.
  2. How does the pricing of the risk taken by backstop suppliers work? This is important because if the positions they liquidate are fully undercollateralized, they are explicitly taken as a realized loss on each liquidation. This question origins from the following quote, as I depending on the answer, this looks too low, or too high.
  1. Could you elaborate on why the backstop depositors should be eligible for GHO discount?

Generally, I think the advantages of focusing more on pure secondary market liquidity of GHO are stronger but open to convincing on the model.

3

Thanks for the reply.

I would appreciate a comment from them as well, who is the best person to tag for that?

When the position is below 1 HF. Note that in practice, it would only catch a liquidation that MEV bot does not, so in normal course of events, liquidations would still be executed by MEV bots. and only when DEX liquidity is not sufficient, then liquidations will be done by the backstop.

The liquidation itself is profitable, however, because it will be done only when DEX liquidity is depleted, the liquidators bare a real risk. And hence, we ask to provide them with incentives. This can be seen as an insurance. Given Aave’s history of setting responsible risk params, we believe that 5% apy, in the current market, would be able to attract sufficient liquidity. But of course this might change if there is an alternative APY over the GHO that is less riskier. Though in such case, we may consider to deposit the backstop funds in that alternative, and compound the APY.

The backstop depositors can be seeing as providing an advance insurance mechanism for failed liquidations, and should be paid for that. One possible way to do it is with a discount on GHO borrowing. But a direct payment in the form of tokens or GHO liquidity mining could also work.

This is a great comment. But note that even if GHO has 1:1 liquidity with USDC, with $10B qty, it would still force Aave to be more conservative about GHO debt than on USDC debt.
The reason is that bad USDC debt is on the expense of the Aave suppliers, who are compensated with an interest rate on the risk they take. While bad GHO debt damage the trust (and potentially the pegging) of GHO holders. Who are not compensated for holding GHO, and in general, GHO trust level should be high in order to encourage mass adoption.

To give a concrete example, even if USDC/GHO liquidity is super high, there could be still risk when borrowing GHO with, e.g., CVX collateral.

Hence, a backstop could potentially increase the community trust in the stability of GHO, and mitigates concerns about how new collateral listing would affect the stability of GHO.

4

$3M allocated to backstop $GHO can be a big expense at low $GHO issuance levels, making it potentially not profitable, any studies on that?

5

I guess @AaveLabs

But if we assume an optimal ecosystem of bots liquidating whenever profit can be realized (including CEX depth + DEX aggregation), the liquidation for the backstop will for sure create a loss for them. And in that case, I feel it is more optimal to just have the Aave SM covering it, maybe potentially diversifying it a bit more, or using some AAVE collateralization.
In a lot of cases, and if not a really meaningful amount, small bad debt doesn’t create any systemic risk, as it can be covered progressively.

Regarding your final point, I would say it is fundamental to first understand the peg protection mechanisms in terms of the primary vs secondary market for GHO. At least for me, they are not clear yet, apart from the basic demand side for repayments on a down-trend market, together with burning pressure if the interest rate is moved up.
In addition, given the nature of GHO as an overcollateralized asset and the mechanism of borrow yield redirection to the protocol, it is possible that bad debt can be covered automatically with them. This then raises the question of which capital/technological advantage gives an external backstop.

6

I don’t assume that the bots ecosystem includes CEX depth. Especially not for big numbers. While surely such CEX/DEX arb bots exist for uniswap etc, there are no evidences there are ones for liquidations, and imo it is not likely for someone to hold a lot of CRV now both on-chain and off-chain just to wait for CRV to get liquidated on aave.
A backstop is similar in flavour to SM, however it is (under optimal setting) more efficient/cheaper as it is not a guaranteed lose, and it would periodically also make profits.
It should be noted, that the standard B.Protocol backstop also gets priority in liquidations, and hence, the insurance is even cheaper, as the backstop also has a profit revenue stream, but here, in order to simplify things on Aave’s side, we do not offer it.

Afaik, regardless to any PSM mechanism that might be introduced, the plan is to back GHO with a collateral in Aave’s market.

Our stand is that backstop is more efficient that an insurance fund (though a protocol should ideally have both), as it can potentially make profits.
Further, from optic point of view, Aave is currently setting it’s risk parameters in such way that bad debt is tried to be avoided, and as an evidence for that one can see that other than some dust accounts, no account in Aave ever accrued bad debt. It is true that the community could decide to be more aggressive about it. But it potentially it might be better optic to increase the liquidity side (thanks to the backstop buffer) than to promise that yield will be directed to cover the losses.