Decreasing Liquidation Ratios - Analysis

Hey all,

A while ago there was a discussion among mandated actors whether we should potentially lower liquidation ratios for Maker vaults in order to boost DAI supply and make Maker vaults more attractive and competitive to secondary lenders, which on average maintain lower liquidation ratios. In theory, lower LR should also lead to more DAI minting from existing vaults (i.e. if someone is actively protecting his vault versus x% price drop, he can maintain lower CR if LR is decreased, hence more DAI minting).

The Risk Core Unit prepared an analysis that aims to quantify how a lower liquidation ratio of a particular vault type affects “risk premium to debt exposure curve”. We used our risk model that is being actively used to quantify risk premiums of Maker’s vault types.

To give a bit of background to how the risk premium to debt exposure curve is affected, we must first understand that a higher debt exposure leads to a higher risk premium of a particular vault. This is mostly related to liquidity metrics of a particular asset: the higher the quantity of collateral asset sold in auction, the higher the slippage from keepers selling it, leading to higher potential loss for Maker. This is based on the fact that auction keepers rely on liquidity metrics of bought collateral when they “flip” it back to DAI or any other stablecoin. Therefore the risk premium to debt exposure curve has an increasing slope.

A lower liquidation ratio simply means that any debt size liquidated will have a higher probability of becoming a bad debt when certain auction slippage is applied. We simulated three types of decreases in liquidation ratio per vault type: a decrease by 5%, by 10% and by 15%.

The results suggest that the risk premium increases between 50% to 100% if LR is dropped by 15%. For instance, if the LR for ETH-A is lowered to 135%, the risk premium for current debt exposure increases from 2.6% to 4.8%. Lowering LR by 10% increases risk premium by less than 50% on average. Different collateral types have different responses to a decrease in LR, which mainly depends on liquidity metrics of a particular asset, as well as the current collateralization ratio distribution of a vault type. Also please note that we used pretty aggressive inputs when modeling risk premiums. For example, we assumed 50% asset price jumps twice per year for second tier assets.

The presentation with results can be found here.

The @Risk-Core-Unit suggests potentially decreasing LR only for second tier vault types by 10%, whereas a decrease of LR for leading vaults such as ETH-A, ETH-C, WBTC-A and also UNI LP vaults is at most suggested by 5%. We do not suggest a decrease of LR for ETH-B due the OSM delay risks.

Side note: Maker currently offers low rates and risk compensation is extremely low. Few vault types are not properly risk compensated right now. As risk premiums will increase by potentially implementing lower liquidation ratios, the spread between risk compensation from fees and expected losses will only increase and therefore caution is advised. This also means that we will not be able to increase debt exposure as much as we wanted, at least for second tier lower liquid vault types. Luckily the surplus buffer ratio is historically at an all time high, which offsets some of the extra risks taken. Currently, portfolio weighted risk premium measured as expected loss amounts to about 48m DAI, which is slightly below the current surplus buffer of 50m DAI. Portfolio is well hedged considering our aggressive price jump assumptions, but this can also suddenly change if a) debt levels are increased b) CR distribution worsens or c) liquidity metrics or asset specific risks worsen.

14 Likes

Good piece of work @Primoz. 10% it is then. Will you be putting this up in a signal request?

Yes, this was the plan, signal request unless there are some objections here. Didn’t want to have one right away, still an important change to the system.

Btw @Primoz - while we are at it - I assume that the by far biggest factor here is the fact that the liquidations come more or less all at once when liquidation threshold is hit?

Would it not be an improvement if Vault holders could set their own liquidation ratios (i.e. 165%, 185%, 225% etc) with an incentive? In this way liquidations would be a much more spread out affair…

3 Likes

Hm interesting proposal @Planet_X but I think this would have a lot of implications on auction parameters that are also dependant on LR. Also we’d need to start applying rates differently, formula based. I can see increased governance overhead.

2 Likes

I haven’t looked at the presentation yet but I think it’s good to keep in mind that conversely there are risks for Maker related to overreliance on the PSM(s) and to DAI undersupply which get reduced a bit by lower CRs (I am currently fairly biased towards being concerned about Maker failing at its mission to generate enough supply to maintain the peg and about Maker’s USDC exposure).

2 Likes