Black Swans, Trusted Assets and Multi-Collateral DAI
There’s some strong discussion brewing on which asset types should be added to multi-collateral DAI (MCD) when it launches. In particular, the introduction of tokens being pegged to real-world assets that may involve KYC and/or the risk of government seizure.
These types of discussions are very challenging because humans are notoriously bad at conceptualising extreme events. We tend to rely heavily on intuition built from experience and, by definition, we don’t experience extreme events very often.
Insurance is all about managing extreme events, and I’ve been doing the insurance gig for a while now, so here are my thoughts on the discussion framed using widely used insurance techniques.
One way of thinking about taking on a new risk is understanding the likely worst case, or probable maximum loss (PML). If you are comfortable with the outcome in the PML scenario then the question becomes more about what is the right price for the risk, rather than should it be taken on at all.
I’d advocate thinking about introducing new assets to MCD in the same way. First establish what a PML scenario would look like, then decide if this is acceptable to the system, and if so then start thinking about risk parameters. Noting there is a feedback loop between these items as they aren’t entirely independent.
What could cause DAI to fail?
Firstly, it’s worth understanding the current failure modes for Single Collateral DAI (SCD) as a reference point. The most obvious one is if the ETH price drops very far, very quickly or “gaps”. This could cause the DAI peg to break because DAI would become under-collateralised and MKR minting may not prove sufficient enough to stabilise the system. 2018 proved that substantial decreases in ETH price can be handled adequately by the SCD system, sometimes with large gaps of 20%-30%. But we still don’t know at what point the system is really tested. 50%? 70%? Who knows.
So what scenarios could cause a massive price gap in the ETH price or otherwise place DAI in a failure mode.
- Rune Christensen (MakerDAO’s Founder) has suggested a coordinated action against exchanges could cause ETH price to gap substantially. I’m not as convinced, as even coordinated action by governments would not be entirely coordinated timing wise and while large gaps could occur I’m not sure they would be enough to break the peg. Of course, I don’t actually know and if MakerDAO can minimise this risk without opening up others then why not.
- Contentious fork leading to ETH price on both chains being substantially lower. I’d argue this risk is actually more likely to occur than #1 and is probably the key reason to add other assets. This would likely amount to MakerDAO picking one side of any fork (as some assets would be unlikely to co-exist on two chains).
- A technical failure in Ethereum is another possibility. This risk will exist regardless of MCD and asset types so it is worth noting but doesn’t really impact this discussion.
- Price feed oracles could be hacked. This risk will also exist regardless of MCD, but increases with each new asset as more oracles are required.
- MakerDAO governance could be co-opted. Again, this risk will exist regardless of MCD and asset types so I will ignore it for this discussion.
- MCD technical risk of smart contracts failing. SCD is fairly well batted-tested at this point but moving to MCD regardless of asset types increases the risk of failure.
- There may be others. Cue Donald Rumsfeld quote.
Why move to MCD?
Adding any asset to MCD, has to benefit the system as whole. So it needs to either:
- Reduce the likelihood of entering one of the failure modes above; and / or
- Benefit the ecosystem more generally, for example allowing DAI to scale more.
I’d argue that ETH alone has more than enough “economic bandwidth” to significantly grow DAI from where it is, and the current bottleneck is genuine DAI demand. This may change in the future but I believe MakerDAO should currently be prioritising reducing the risk of failure in their move to MCD. I’d also argue more generally that preventing failure modes should take much higher priority when introducing system changes than scaling DAI and any trade-offs that increase the risk of failure should be viewed with skepticism.
Specifically, moving to MCD with new assets has to:
- Reduce the risk of DAI becoming under-collateralised as a whole; AND
- Be comfortable that increased risk from extra oracles and smart contract technical risk is low enough to still be better off.
Including Trusted Assets
Let’s get back to introducing “trusted” assets and what the impact could be on the DAI peg. Intuitively, the probable maximum loss scenario is where asset values of the trusted asset go to zero quite quickly. Likely not from the actual underlying asset value changing, but because government or regulatory action forces the token issuer to freeze or confiscate assets.
So what happens in this scenario?
New MKR is minted and then sold to effectively re-collateralise the now under-collateralised assets. This is not necessarily a failure mode for DAI, it depends on how large a portion of the collateral needs to be re-collaterised.
This highlights a key point:
In MCD, the likelihood of any asset failing is increased, but the consequence is less severe and the entire system can actually be more resilient to overall failure.
This is because the more non-ETH assets are added, the lower the consequence if ETH were to suffer a large enough price gap.
Risk is a combination of likelihood and consequence and “trusted” assets would be expected to actually reduce risk at an overall level. Of course, this requires that the trusted assets don’t all fail at once and aren’t a big portion of the collateral base. But we’re now into how to price the risk, rather than whether it should be taken on at all.
Should MakerDAO add trusted assets to MCD?
Adding new assets, even trusted ones, to MCD can actually reduce risk at an overall level if governed and priced appropriately. This is perhaps counter-intuitive, but it also means there is only a real benefit to MCD if the proportion of ETH held as collateral is meaningfully reduced, as this is where the benefit lies. If ETH is still 90% of the collateral the risk position hasn’t materially changed.
All things considered, I prefer SCD right now. Until there is a concrete proposal to add multiple non-ETH correlated assets to MCD that substantially reduces the proportion of ETH as collateral, I don’t see much point in MCD. So perhaps this is a timing issue, as the ecosystem needs other assets to be established first.
And one last point, if you can have MCD without “trusted” assets that are still uncorrelated to ETH, then that’s obviously better. One less risk to worry about. Will that be an option in the future? I’m not sure, but I hope so.
I’d just wait and see what options eventuate.