[Discussion] Separate Centralization Risk Within Risk Premium

Setting stability fees have been made more intuitive by separating the Base Rate, which applies to all Vaults, from the Risk Premium, which is specific to each Vault Type (discussion, signal).

Discussion and signalling about the WBTC debt ceiling has highlighted that whereas ETH risks include market price but not centralization, and USDC risks include centralization but not market price, WBTC includes both.

It may therefore be prudent to begin consciously considering centralization risk on its own.

This could be done by breaking up the Risk Premium into two parts:

Risk Premium = Centralization Risk Premium + All Other Risks Premium

Then, a vault type’s Stability Fee would be:

Stability Fee = Base Rate + Centralization Risk Premium + All Other Risks Premium

Note that for simplicity I suggest a single Centralization Risk parameter common to all centralized collateral, but it could also be per centralized custodian if the community perceives them to have significantly different centralization risk.

The purpose of this topic is to gauge the level of support for updating the Fee Structure in this matter.


I like this idea and think we should open it back up for discussion. It seems to me the risk premium could be split into a few key components. Custodial/centralization risk, some sort of price risk that scales with the volatility of the asset, and then idiosyncratic risk per collateral type. It seems like splitting the risk premium up this way could give us more consistent stability fees across collateral types. As @Spidomo brought up above, the wbtc and usdc custodial risk probably should have the same premiums since that part of the risk profiles are approximately the same.

This kind of thinking is becoming more important now as we are adding more and more collateral types, and as we grow we risk having inconsistent stability fees for vaults with similar risk profiles. We have already seen this with KNC, ZRX, and BAT, with BAT having a similar risk profile, but a different premium. Likewise with ETH, which has non-zero risk associated with it, yet retains a 0% risk premium. With the base rate going negative today, we now have an opportunity to adjust these premiums and changing the Meta-Parameters without hurting the peg.


I think that this sort of framework should inform decisions about individual asset risk premiums. But making these factors meta-parameters for the weekly governance cycle like the base rate would probably create too much complication for voters. It’s definitely a great time to adjust risk premiums now that most rates are underwater.


I like the idea but it might be difficult to . For instance USDC has a centralization risk in Circle but also a regulatory risk (US administration can froze accounts on chain or off chain).

If you separate the risks (in two or three or more buckets) you need to have a diversification parameters.

Also you can argue that USDT shouldn’t have a full custody risk, they are only 74% asset-backed. 74% of the custody risk then?

Maybe having a framework but letting the community/risk team decide then end result ?

all centralization risk has regulatory risk baked in. USDT is actually more risky because it’s not fully backed, but I don’t think we plan on onboarding it because of that. In any case, there are unique differences between collateral types that can still be accounted for with non-global risk parameters. I just think we should be consistent with custodial risk for example applying the same premium for all collateral types it applies to.

1 Like

RP = custody_RP + regulatory_RP + volatility_RP + collateral_specific_RP (stealing the name from @befitsandpiper) ?

seems a good framework for me then. I see way less regulatory risk in WBTC than in USDC. USDC has also a tail risk of negative interest rates which would destroy the USDC business model. Again, EURS has more risk than USDC because of that (on top of the currency risk which can be put in the volatility bucket).

I’ll just jump in and ask if we would assign custody risk depending on the agent (is Circle equally as risky as Bitgo?) or if we would just treat all agents the same

I think most agents should have the same custody risk premium. Although, I guess if you wanted to really get into the weeds, it probably depends more on the jurisdiction than the agent, unless the company is doing something sketchy. Any sketchy things can also be accounted for in the collateral specific risk premium, but this is different from custodial risk in the abstract

I agree that looking at this framework and making them into meta-parameters seems complicated, but at the same time I think it is likely to reduced the difficulty in assigning risk parameters to individual collateral types.

Having these preset quantitative parameters for qualitative ideas makes it easier for us to have consistent stability fees. And if we accurately categorize types of risk, most collateral types should have 0% for unique risk premiums.

For example, our existing collateral types WBTC, ETH, and USDC all are likely to have no unique risk premiums. This is likely the case for KNC, BAT, and ZRX also, although they have more volatility risk. All of these collateral types could instead be summed up by a price risk weighted by average volatility, and a custodial risk. I think maybe the only collateral type we have with unique risks that aren’t a part of this framework yet is MANA.

In the future then we wouldn’t have to think quantitatively about every new collateral onboarded. The question would be a much more human question, “What kinds of risk does this collateral have?”. And we could also add new risk categories as we broke into new types of collateral.

Risk premium and base rate had an acutal technical function. BR is global, while RP is local to a single collateral. Introducing additional local rates has zero technical function, so comparing it to the introduction of RP is unfair.

When talk about how to set RP it might make sense to introduce categories but that methodology should and discussion should not take place via polls and smart contracts. There are infinity + 1 ways to produce a RP and we should not force governance into following 1 single methodology above all else.

Also, consider 10 days from now when we discover a new risk vector that we would need to track that doesn’t fall within these narrow categories.


Exactly, the Risk Premium is opening the floor to other collateral type of variants witch investors or users should not take part into that discussion. Hazardous opinions would lead to probably unwanted decision or non unified understanding in general.

like @monet-supply is saying, this could be sorted as multiple framework, but at what cost and once there, it’s individual thinking mostly, not an oasis anymore and the wanted main goal is going into a derivation area.