Consensys / defi-score

DeFi Score: An open framework for evaluating DeFi protocols
https://defiscore.io
Other
279 stars 79 forks source link

Asset-based LTVs and adjusting the LTV levels based on volatility mitigates risk #29

Open EthWarrior opened 4 years ago

EthWarrior commented 4 years ago

Current DeFi score takes collateral and Loan-to-Value (LTV) ratios into account as a risk within the overcollateralized lending protocols. These lending protocols might apply either asset-based LTVs such as Aave and Compound or apply same LTV across all listed assets (the so called MakerDAO model).

Firstly, applying same LTV across all assets increases the risk of the protocol as the peculiarities of different assets in not taken into account. More importantly, different assets have different liquidity risk on the secondary market which affects the ability to resell the collateral in the secondary market due to a liquidation event. The level of LTV exposes directly to the liquidity risk of an asset since higher LTV triggers quicker liquidation event in case of volatility. Cross asset-based LTV protocols are subject to higher risk and therefore should decrease the overall score.

Secondly, the volatility of an asset affects the liquidation event as stated in the DeFi score white paper. Being able to react on the volatility of an asset overtime should be included into the risk assessment. Lending protocols that are monitoring assets and their volatility and adjusting their LTV ratios bears less risk compared to lending protocols that do not adjust LTV ratios in case the volatility of an asset changes radically.

For example, an asset might gain 40-60% increase in value on day one (the REP example on 16th January 2020). Borrower pledges the asset as a collateral with 66% LTV (which might be max for that particular asset). If the price rise is momentary and bounces back down 40-60%, there would be more liquidation events for new borrowers (who are pledging collateral on the increased market price) as the collateral value decreases back to normal, which increases the sell pressure of an asset upon a chain of liquidation events. Therefore, lending protocols that do not react on the increased volatility and adjust the LTV ratios are subject to higher risk and the overall score should decrease.

Analogy form property price shocks and boom-bust scenarios and how adjusting LTVs is connected to the risk: https://www.ey.com/Publication/vwLUAssets/ey-effectiveness-of-loan-to-value-ratio-policy-and-its-transmission-mechanism-empirical-evidence-from-hong-kong/$FILE/ey-effectiveness-of-loan-to-value-ratio-policy-and-its-transmission-mechanism-empirical-evidence-from-hong-kong.pdf

jclancy93 commented 4 years ago

Hey @EthWarrior,

Thanks for bringing this up. I agree that the ability to change LTV/Collateral ratio parameters for volatile/illiquid assets are important.

I don't think Maker is committed to keeping the same LTV for all assets, but could be wrong there. I think LTV is particularly important when a protocol supports less liquid assets, like REP. For a platform like dydx that only supports USDC, Dai and ETH, I think a single LTV is acceptable.

Are there any platforms that support a large variety of assets that only use a single LTV value?