Last time, we discussed the need for any L2 instance to be a separate deployment from L1 for security reasons, and which collateral assets are likely to be available in the protocol at launch. Today’s article is about the role of first loss capital in the Ethereum Credit Guild system and how it differs compared to other lending protocols onchain. We discussed GUILD minting previously here.
On the need for reserves
The venerable MakerDAO relies on minting and selling MKR to recapitalize itself against any loss greater than the surplus buffer (at time of writing, about 1.12% of the DAI supply). To quote from the excellent article by Hasu:
$6 million in credit losses from Black Thursday[March 12, 2020] erased $10 million in earnings accumulated over 3 years. The $4 million in additional losses could have been avoided if Maker held more treasury reserves in stable assets like DAI, as they could have used these funds to cover insolvent loans without needing to sell MKR at depressed prices. Or put another way, Maker could have seen up to $4 million in additional value accrual by holding a larger treasury.
In the case of Black Thursday, simply holding onto accrued profits in DAI as reserves instead of buying and burning MKR would have been sufficient to pay for the entire loss, and leave some reserves left over. The native debt asset is the best form for reserves, since it can directly cancel bad debt with no need for an auction, which induces delay and leaks value to validators.
On the necessary quantity
The right amount of first loss capital, whether in a shared system surplus or staked in something like the Aave safety module, clearly has some relationship to the amount of total deposits, and some relationship to the degree of risk in the collateral. Sometimes, it will be good for more capital to enter, and sometimes it will be good for it to leave.
The last part of the core CREDIT system we’ve been finalizing is a simple mechanism for this purpose. Our current approach is allowing GUILD to be minted against CREDIT, through a special holding contract that allows voting in gauges or proposing new lending terms, but not vetoing new terms or closing existing ones. In this way, we can avoid griefing risk of the protocol while allowing outside capital to enter and exit smoothly. The amount of GUILD mintable in total, and the interest rate (whether negative to reward participation in the early system, or positive as a revenue split between core holders and outside capital at maturity) can be adjusted optimistically through governance, subject to veto, just like adding a new lending term.
Initially, I thought of having various assets like stETH mint GUILD, but as noted above, the native debt asset is the most efficient way to respond to bad debt, since no auction or sale of reserve asset is required, reducing value leakage and minimizing periods of price uncertainty.
On rates and returns
The interest paid by borrowers is split between GUILD holders, the Credit Savings Rate, and the surplus buffer. Initially, GUILD holders will receive 0%, and the surplus buffer will receive 10% of the interest paid by borrowers, with the remainder streamed through the CREDIT savings rate. The GUILD minted against CREDIT will be incentivized through a negative rate. Changes to any of the above parameters are conducted through the same optimistic governance process as onboarding a new lending term. CREDIT holders and borrowers will also receive GUILD rewards in the early period. This is not a part of the core protocol and will be conducted Morpho-style with rewards calculated offchain and claimable onchain in an epochs system.
On redemption
In this model, GUILD tokens minted when users choose to provide additional first loss capital and earn yield, subject to the consent of the existing GUILD holders. They are burnt either when bad debt occurs, or when a user chooses to withdraw their CREDIT staked in the security module. The system is expected to absorb first loss capital when the protocol is growing, and release it when the reverse is true, while steadily accruing a surplus that will provide a sustainable first loss capital base once the protocol reaches maturity.
TLDR
Reserve in same denomination as debt is best. Anyone can mint GUILD against CREDIT to get started in governance. High performers may attract others to delegate to them. This system is quite different from what exists in DeFi today. Please enjoy responsibly, and ask questions on Discord or Twitter any time.
We’re now within a few weeks of code freeze prior to pre-audit review. All eyes on the repository are greatly appreciated.