Analysis: What determines the mortgage coefficient of the multi-debt lending platform

The DeFi lending market sets different lending capabilities for different collaterals. Every collateralized asset has a collateralization factor (aka LTV), which determines how many dollars of debt can be borrowed for every $1 of collateralized assets.

The collateralization factor is always less than 1 to compensate for delays in the liquidation process and further price drops. For example, with a collateralization factor of 0.7, bad debts will only be created if the collateral price falls by 30% before the liquidation ends.

During a liquidation, the debt is paid by the liquidator, who in return receives the liquidated collateral at a discounted price. Almost all DeFi liquidators will sell seized collateral immediately after liquidation (in the same blockchain transaction), so if there is not enough DEX liquidity to sell it at a profit, there may be delays in liquidation. Furthermore, when mortgaged assets are sold, asset prices may fall further.

In a system with a single debt D and a single collateral C, the collateralization coefficient can be calculated based on the available DEX liquidity of the C/D trading pair and by simulating different price trajectories of the C/D trading pair. In this case, collateral C is "better" if it has higher DEX liquidity and higher price correlation compared to asset D, so it can be assigned a higher collateral coefficient.

Multiple Debt Risk Assessment

In a system compatible with Compound (v2) and Aave (v2), multiple debt assets can be borrowed against a single collateral. However, each asset is assigned a collateral factor that applies to all liquidations of that asset regardless of the debt asset.

Thus, one type of collateral may be "better" than another for a particular debt asset, but worse for a different debt.

For example, USDC collateral is easier to liquidate when the debt is in USDT or jEUR, while ETH collateral is easier to liquidate when the debt is in stETH or even WBTC. In general, some ERC20 asset prices may be less volatile when denominated in ETH and less volatile when denominated in USD.

Therefore, collateral risk parameters must be set according to their worst borrowable debt assets. Therefore, the introduction of a single “problem” debt asset could worsen the risk parameters and thus the capital efficiency of the entire lending platform.

Fortunately, other factors can be considered. For example. Assets like BAT and ZRX have historically rarely borrowed from each other, especially with Compound expecting fewer BAT/ZRX liquidations to occur.

In addition to historical user portfolios, the expected liquidation volume of the token0/token1 trading pair is also controlled by the following factors:

Minting cap for token0. This means how large the total amount of token0 collateral can be. A lower minting cap reduces the expected liquidation volume for all debt assets.

The borrowing limit of token1. Indicates how much token1 can be borrowed. A lower borrowing limit reduces the expected liquidation volume for all collateralized assets.

real world results

Aurigami Finance is one of the leading lending marketplaces on the Aurora blockchain. One of the collaterals they support is stNEAR, which is a bridge collateralizing NEAR. It’s less liquid on Aurora, but it’s highly correlated to NEAR and thus more correlated to ETH than to USD.

We apply our simulated environment to reason about recommended values for different asset collateralization factors.

By initial setup, ETH is marked as a "better" asset than USDC, with a higher collateralization factor.

However, when the cursor is placed on USDC, the system shows that if the borrowing limit of stNEAR is lowered, the collateral factor of USDC can be increased.

In fact, once we lowered stNEAR’s borrowing cap, USDC’s recommended collateral factor increased, and USDC is now the “best” asset.

Only when we lower both the ETH minting cap and the USDC borrowing cap will ETH become the best asset again.

in conclusion

The collateral factor in multi-debt lending platforms stems from the complex dependencies between minting and lending caps, and manual optimization is not an easy task.

It's worth noting that in more modern architectures, such as Aave v3 and Eular, more granular configuration is possible. In Compound v3, a design choice was made to only support a single debt asset lending market. However, to date, most DeFi lending platforms still stipulate that each asset has only one collateral factor.