Ethereum, cryptoeconomics, governance
“First you must find… another shrubbery! Then, when you have found the shrubbery, you must place it here, beside this shrubbery, only slightly higher so you get a two layer effect with a little path running down the middle. (“A path! A path!”) Then, you must cut down the mightiest tree in the forrest… with… a herring!”
As MakerDAO moves closer to professional asset-liability management thanks in large part to the efforts of contributors like SteakFi, I am saddened to see blatant disregard for these principles in several recently published stablecoin mechanisms, including the new version proposed for Liquity and the “delta neutral” stablecoin backed by Arthur Hayes.
Asset-liability mismatch is the Achilles heel of banks and banklike systems. Fundamentally, there is no way to guarantee that any asset is worth exactly the same amount as another. Even fungible commodities like gold aren’t worth the same amount everywhere. This is an inconvenient truth for demand deposits or their more modern cousins, stablecoins, which intend to maintain a uniform price. Even a treasury bill that matures to $100 in a month is not worth $100 today. In other words, even if you back a debt asset exclusively with same-denomination collateral, there is still no way to avoid some amount of duration mismatch on a demand deposit, redeemable banknote, or stablecoin, and the problem gets worse the more divergence between the price behavior of the backing asset and the debt denomination.
Systems must be designed to handle the problem. In George Selgin’s “Theory of Free Banking”, we observe a few principles of healthy debt issuance with a long historical pedigree.
Flexible liabilities
Everyone knows that the value of money is not constant. Currencies fluctuate in value against each other and against commodities. A “controlled depeg” acknowledges this fact while still attempting to minimize deviations. Under Scottish free banking, one innovation along these lines was a convertible banknote. In principle a banknote was redeemable on demand for gold, but if the bank was unable to provide gold when the note was presented, it was permitted to replace the note with a zero coupon bond at a preset interest rate and maturity. So long as the underlying loan book of the bank is sound, the depositor could then sell this bond for a value close to or perhaps even higher than the face value of their deposit, depending on the bond’s interest rate vs market rates.
Slack in the system
For a bank or a stablecoin issuer, the spread between the interest rate currently paid out to holders and that charged to borrowers is “wiggle room” for the peg, since rates can be raised to compensate for redemption delays. Paying above market rates and taking in too much liquidity has the opposite effect, all but guaranteeing a depeg if external market rates increase further or there is any devaluation of collateral. A sufficient operating spread and a reserve to compensate for unexpected interest rate changes in the market allow a debt issuer to raise rates flexibly and credibly as needed, and survive to increase the spread again in the future.
Clearinghouses
While it is easy for a single bank or stablecoin issuer to experience excess withdrawals, the money has to go somewhere, and it generally does not go under mattresses as stacks of dollar bills, but only into a different bank, money market fund, etc. This means that credit issuers can benefit from mutually extending credit. If one bank experiences a liquidity crunch, so long as its backing assets are sound, another with excess liquidity can extend credit to its distressed peer more efficiently than the backing assets could be liquidated on the market. While this credit extension is not without risk, clearinghouse members are in the same business, and so well-positioned to understand the health of each others’ loan books. See this article by Sebastien Derivaux proposing a ClearingDAO. While I believe Seb overstates the extent to which total fungibility is desirable or possible (“$1 should equal $1 no matter its form”), it’s clear that forming mutual credit networks can greatly benefit the members and the stability of their pegs.
Setting aside its laughable claims to decentralization (“Collateral is distributed to secure onchain custodial wallets, and a corresponding short position is taken for delta-neutral collateralization… USDe is the first decentralized, scalable, and stable asset”), the new Arthur Hayes backed stablecoin makes a critical economic mistake.
We must first ask ourselves, if we start with ETH collateral, where does the short position come from? We must sell some of the underlying in order to pay for it. At times it’s possible to get paid to short on perpetuals platforms or to take the short side of options, but sometimes it is the other way around. There is no way to always harvest a positive funding rate, unless you can alternate between “hold spot ETH, short ETH perp” and “hold spot USD, long ETH perp”, and then you incur switching costs. There is no reason to fundamentally assume this behavior pattern will be profitable. It is speculative, and its success depends on how the market moves.
I’ll refrain from overmuch comment on Liquity 2.0 since the full specification has yet to be revealed, but what has been revealed suggests the same class of problem – offering “downside protection” to a class of borrowers and managing a “decentralized reserve”. Instead, let’s look closer to home. Fei Protocol launched with the intention of a pure-ETH backed stablecoin and lacked direct par redemptions. It didn’t go well. Over time, the protocol moved a large portion of its assets to DAI to reduce asset liability mismatch and ensure smooth redemptions at peg (though not as large as I would have liked), and ultimately switched all assets to DAI before closure.
A protocol that issues a USD denominated debt asset against ETH collateral is levered long ETH, no ifs, ands, or buts. Do not be fooled by complex mechanisms or convoluted incentive structures. Using a portion of the ETH collateral to hedge by shorting ETH can only temporarily alter this fact, and is an active risk management decision. Overcollateralization works only so long as the debt asset’s total circulation is small relative to the market cap of the underlying. What’s more, this is global, not local – all debt in denomination X using collateral Y has commingled market liquidity and thus shared risk. For a debt asset to remain stable at scale under all possible market conditions, there is no other way than like-kind backing.
Discretionary pegs and mismatched collateral backings have laid low greater powers than any stablecoin designer might possess. Instead of pursuing the philosopher’s stone of backing a USD denominated debt asset purely with a crypto-native collateral at scale, we should design lending markets that are aware of global leverage risks, such that lenders can respond and shift to safer assets at need.