How does liquidation work? How much of my collateral gets liquidated, and which collateral tokens get liquidated?

Hi! I just want to ask a few clarifying questions so I better understand the risks I’m taking.

Let’s say I have a portfolio with 5 ETH and 5000 USDC as collateral. Let’s say ETH is $5000 at T1 (Time 1) for the sake of these questions. I am able to borrow up to 80% of the value of my collateral before liquidation, so $28000. I borrow $20000.

It is now T2 and ETH has dropped to $4500. Now my collateral is worth $27500.

What gets liquidated at T2? $500 of my USDC, or $500 of my ETH? Or does more of my collateral than that get liquidated?

1 Like

@Emilio would you please answer the questions? Just trying to understand my risks better. Thank you!

Hi @robhaisfield.
First a remark, when having as collateral both ETH and USDC, your LTV and liquidation thresholds are actually a combination of the ones of ETH, USDC, weighted depending on the proportion of amounts of each forming your collateral. ETH and USDC happen to have the same parameters, but for the sake of understanding.
Concerning your question about liquidation, the rule is that up to 50% of the total value of your collateral gets liquidated, and it is the liquidator who chooses which asset to receive. So if you enter in liquidation (health factor below 1), taking into account your example, the liquidator could choose to take part of your ETH (as it is more than 50% of your collateral), or just the whole USDC. In this case it is probably smarter to choose the first, to liquidate and take bonus in as bigger amount as possible.

Thank you for the response. Is there a reason it was designed such that the liquidator chooses what gets liquidated? I would far prefer to rank the collateral myself for what gets collateralized first, and generally that seems more user oriented. I’d love to understand the rationale better.

The main reason is the robustness of the system. I can explain a bit exactly why.
Liquidation is probably the most important mechanism of Aave, because it enables everything else, for example, the 1:1 equivalence between underlying and aToken.
In technical terms, liquidation is also a pretty interesting dynamic, because it deals with non-fungible positions (each one of the user’s positions to be liquidated) compared with other components of the system which have more “fungible” or at least pool nature (e.g. aToken, model of rate based on utilisation, etc). Partially because of this:

  1. Liquidations are a decently expensive operation for the entity executing them, both on-chain and due to the infrastructure needed.
  2. They are influenced by the external market of assets. In the current crypto ecosystem, it is clear that is not the same in terms of slippage selling X amount of let’s say ETH compared with selling X amount of MKR.

As the second point is quite critical for liquidations to happen smoothly, having the liquidator choosing the collateral to liquidate for positions with multi-collateral incentivises them to do liquidations for example on situations where one of the collaterals is going down and the other not so.
As you point out @robhaisfield, this seems to go “against” the borrower, and partially it does, but let’s consider that actually a liquidation situation is from the start non-desirable for both borrower and the system. So if arriving at it, the risk over the system of having positions not-liquidated is way higher (affecting everybody) than the risk of the borrower losing exposure on a particular collateral. And so the decision.

Thank you for the detailed explanation! I suppose part of the benefit of using Aave vs. a bank is it’s programmatically (through code/game theory) solvent.

Correct, that is one of the benefits, the nature of its solvency. As the system itself is overcollateralised at any point in time, and the interest rates algorithm creates pressure on liquidity entry when there is not much available for withdrawing/borrowing.