Compound Finance: Asset Risk
Arguably the most important variable in Compound Finance is an asset’s collateral factor. The collateral factor dictates how much of an asset’s value is borrowable. For example, the collateral factor for ETH is 75%. For every ETH supplied as collateral, 75% of the value can be borrowed in any asset. Accordingly, with 1 ETH anyone can borrow up to 285 USDC with ETH’s price at $380. Now, if one borrows the 285 USDC and ETH price declines, they will be liquidated. As a result, the1 ETH collateral supplied to the platform will be sold for USDC to cover the platform's exposure and a portion (8%) goes towards paying the liquidator, and if there is a remainder, it gets returned to the borrower.
The goal of the collateral factor is to make sure accounts supply more assets than they borrow from the protocol. Setting a low collateral factor of 0% makes it impossible for the asset to be borrowed against. Setting a collateral factor of 75% allows the protocol a safety net to perform liquidations. With the 25% remaining, 8% is paid to the liquidator. The remaining 17% exists to absorb slippage and declining values. Liquidations generally occur at inconvenient times. The market will likely be very volatile and liquidity will be less than normal. Liquidations are triggered by Compound’s oracle system receiving a price update. Those updates are often after a market move has already occurred, so the needed liquidation will likely occur after the price has declined further. Once an account is available for liquidation, a liquidator will likely perform the liquidation very quickly thanks to the large 8% incentive. Quick tangent, a liquidator wants to collect the 8% incentive, but they have to pay a gas-fee for the transaction and make sure they can hedge themselves on the liquidation they are performing. Back on topic, the take away here is the collateral factor/75% is NOT the number to focus on. Instead, focus on the collateral factor minus the liquidation incentive because that is the margin of error.
Now that I have covered what collateral factors are and what they are meant to do, let's dive into why the current system needs to change. If you’re reading this, you have an idea of how volatile ETH/USD is and how volatile altcoin/USD can be. If I am lending 1 ETH and borrow 285 USDC, the protocol has ETH/USD risk. If I lend 1 ETH and borrow $285 of UNI ( or any altcoin listed), the protocol has UNI/ETH risk. Currently, the protocol doesn’t consider that UNI/ETH is more volatile and has less liquidity than ETH/USD, so I’m able to put on a risker position with the only difference being the borrowing cost. Similarly, I could lend some UNI to the platform with a 60% collateral factor and borrow ZRX or anything else under the same terms giving the protocol altcoin/altcoin risk. The last scenario I want to highlight is Tether’s 0% collateral factor. No one can lend USDT to the platform and borrow against it. This protects the platform from needing to sell USDT and buy some other coin in a liquidation. However, users can still borrow USDT while supplying other collateral. For example, if I supply 1 ETH, I can borrow 285 USDT. If my ETH goes down in value, the protocol needs to sell my ETH for USDT to remain solvent. So despite the collateral being 0%, it is still possible for the protocol to take on a large amount of USDT risk.
Problem: the borrowed asset is not a component in the protocol’s assessed risk.
Solution: Safety Scores
Similar to collateral factors, but operating on both sides instead is safety scores (open to suggestions on the name). The safety score of an asset acts as the new collateral factor. The idea is to have a more holistic approach to managing account risk to the protocol.
The supply side looks at each coin supplied and the dollar amount supplied. For each asset, it divides the dollar value supplied by the account’s total dollar value supplied to get the asset’s pro rata share and multiply it by the asset’s safety score (determined by governance); repeating this process for each asset and then summing the products returns the supply safety score for the account.
To determine the borrow safety score look at each coin borrowed and the dollar amount borrowed. For each asset, divide the dollar value borrowed by the account’s total dollar value borrowed to get the asset’s pro rata share and multiply it by the asset’s safety score, repeating this process for each asset and then summing the products, then take the inverse by taking 1 minus the sum to get account’s borrow risk, and add 1 to get the borrow safety score.
The supply safety score will never be greater than 1, and the borrow safety score will never be less than 1. With the account’s supply safety score anyone can calculate the risk-adjust supply value. Multiplying the value of the supplied assets by a number less than one will reduce the assets borrowing power. Similarly, with the account’s borrow safety score one can multiply it by the value of the asset borrowed to find the risk-adjust borrow value. Since the borrow safety score is always 1 or greater it will be boosting the value of borrowed assets. The idea is to discount the value of assets supplied to the protocol creating the risk-adjusted supply value and boost the value of assets borrowed to create the risk-adjusted borrow value.
With the account’s risk-adjusted supply and borrow values, we can find the remaining liquidity/cushion before liquidation occurs. Substracting the risk-adjusted borrow value from the risk-adjusted supply value returns the remaining liquidity. Once the liquidity goes negative, the account is available for liquidation.
The liquidation incentive for the liquidator comes from the difference in the risk-adjusted supply value and the real supply value. Governance would vote on a percentage of the difference that the liquidator will collect, similar to the current system. Based on the current 8% liquidator incentive on Compound and a 75% collateral factor loan, the 8% incentive is 32% of the 25% margin remaining; 32% would act as a conservative starting point. The liquidator incentive needs to be adjusted to the risk-adjusted liquidator incentive to account for the risk of the assets borrowed and supplied. To calculate the risk-adjusted liquidator incentive, find the average safety score by adding the supply safety score to the borrow safety score, divide by two, and then use the average safety score for the account multiplied by the starting liquidator incentive. Since the incentive is factored in after the loan is valued, the liquidators’ real incentive will differ on an account by account basis. Accounts with ETH/USDC risk will have the lowest real incentive (under the example safety scores) and accounts with altcoin/altcoin risk have the highest real incentive. By incentivizing the liquidations based on asset risk, the protocol will attract liquidators and incentivize account owners to manage their risk.
The main advantage of safety scores over collateral factors is the ability to get a more holistic approach to managing asset risk. By including the borrowed asset into the equation the protocol has a new tool to handle risk. With a tool to control asset risk on both sides, the protocol is more robust and has more utility because it can list more assets while continuing to manage risk.
To see a model of this check out the idea here. If you want to play around with it go to File > Make Copy.
If the community expresses interest in the idea, I will explore developing it further. Please feel free to comment on the idea on Compound’s Forum: here.
