Blockchain Confidential - 15 July 2022
The stablecoin risk condundrum
We need to talk about TINA.
This idea, that stablecoins are needed for DeFi and the broader digital asset trading ecosystem, and “There Is No Alternative” so we just have to accept the risks, deserves to be examined. Because if it’s true, we cannot afford to put our heads in the sand and ignore the risks in stablecoins, we have to think about how we assess the risks we accept and if possible mitigate them. And this is not just about UST and Terra: this is about the fundamental concept of a token pegged to a fiat currency and how this has been implemented in practice. This week’s news about Tether’s use of Celsius for its limited crypto holdings should not come as a surprise; it verges on inevitable if you follow the history in Zeke Faux’s Bloomberg piece last fall.
First, a level-set on classifications. Stablecoins fall broadly into three categories:
fiat-collateralized (e.g. USDC, USDT)
crypto-collateralized (e.g. DAI)
algostables (e.g. UST)
They fall into a broader category of synthetic exposures that include rarer tokens pegged to gold in vaults, diamonds, stocks and more. What they have in common is they aim to “peg” their prices to some other price, and track it as closely as possible. There are fiat currencies that have either historically tried to do the same (at least until George Soros rocked up to that particular party) or on a sustained basis more successfully, like the Hong Kong dollar, but the history here is generally not a pretty one. Nevertheless, absent tokenized fiat — CBDC is a whole separate story — they are useful instruments, serving much the same purpose for those unable to get direct fiat exposure as keeping a variable cash position in your portfolio: it gives you a place to hold roughly steady during a volatile period, and pays tribute to the fact that the decentralized world still has on-ramps and off-ramps to traditional finance and, in the case of USD pegs, that the U.S. dollar still dominates as a reserve currency and thus a point of reference for other prices as well as a haven.
So what’s the problem?
The problem actually depends on the mechanism. Fiat-collateralized stablecoins combine three big risks: regulatory risk, counterparty risk, and liquidity mis-matches. Stripping away all the tokenization tech, what you have is a broadly-unregulated, privately-issued, zero-interest bank account with no government backstop and no withdrawal limits. Park any Web3 politics for a moment, and let’s not debate whether regulation or government backstops are a good thing or not, and just acknowledge what we are dealing with here, because it’s a strange creature. You put a dollar in, and at any time you can get a dollar out. In return for this, you receive nothing other than a token and the utility you get from that.
This should ring alarm bells straight away, because what I just described is you became a creditor to the token issuer, which means you face the risk that your dollar might never come back. Normally, you get paid interest for that over and above the risk-free rate, and you get paid more the higher the credit risk. So the zero-interest component is immediately questionable. But what about that backstop, and the fact that you can get it back straight away? Well, even when you get paid for taking credit risk, with a bank things can still go horribly wrong if, as is normal for a bank, you don’t literally take that dollar and put it the safe. That means examination of what happens to that dollar next (bank supervision) and some protection in the event of a failure (FDIC insurance, in the U.S. case) typically comes with the package when you enter into such a transaction. That’s also not there with a stablecoin today.
And this, unfortunately, is where the liquidity mis-match comes in, because we have to look at the incentives of the token issuer sitting on top of all the reserves that back up the tokens. On one side of the ledger we have a perfectly liquid right to redeem exactly one dollar. On the other side of the ledger we have a pile of cash earning no interest, with an agent — the issuers — in control of it. If you are a creditor, what you want is for the issuer to be as conservative as possible: stuff it under a mattress, so at any point there is zero doubt that your dollar will come instantly back. But if you are the issuer, you could make money off investing those reserves, and even better, you don’t have to pay your creditor interest, so you get 100% of that yield. The issuer is thus incentivized to invest those reserves in riskier and less-liquid instruments like, say, crypto which you then lock into an over-collateralized DeFi protocol to boost yield, and — remember we have no bank supervision — it’s really up to you how much you disclose about how you manage those reserves. We can wave our hands at market mechanisms because nobody is, you know, putting a gun to your head to hold a stablecoin issued by a company whose leadership includes a former child actor from Disney’s Mighty Ducks, but … rational investor choice requires effective disclosure.
How much do you trust Tether or Circle or anyone else to do the right thing here in the face of that particular temptation?
Unfortunately, crypto collateralization does not substantially change the equation, and introduces new hazards. DeFi has in a number of places used over-collateralized positions in lieu of backstops, sometimes alongside automated margin call mechanisms to ensure positions get liquidated before they put the protocol at risk. But while clever, and potentially less damaging to the traditional financial system because the linkage between the cryptoverse and markets like, say, commercial paper are not coupled through reserve holdings, details matter here. What is the appropriate level of over-collateralization for a highly volatile asset? How reliable is the liquidation mechanism, and how much slippage can be expected when liquidating? Are the incentives for those performing liquidation — which often has an off-chain component — reliable enough to ensure the protocol always stays solvent? These are questions of risk management and incentive design, and not easy. I worry sometimes that such mechanisms get reflexively defended by pointing to nominally high levels of collateral, without asking how we know how high is high enough. Also, anything backed by collateral has the riverbank problem: collateral is always finite, and so just as there is always a flood that could overtop the sandbags or levees — a 100 year flood is okay, a 1000 year flood not so much — there is always some plausible scenario that would lead to a breakdown.
OK, so if we don’t like government backstops or fiat linkage and collateral is problematic as a peg defense whatever the reserve composition, can we create a mechanism that is not based on collateral that gets us the same effect? Enter algostables: through the magic of automatic, arbitrage-driven stabilization, we can preserve the peg. If the stablecoin goes above-peg, we incentivize arbitrageurs to sell it, and drive the price down. If it goes below-peg, we incentivize arbitrageurs to buy it, and drive the price up. Typically this is done with a secondary token. Very clever: all we need is a perfectly reliable incentive design that works under all market conditions, and we can sleep soundly.
Unfortunately. we just saw a $2T market crash triggered in part by an algostable with an incentive design problem. Terra’s UST never failed technically: the blockchain and all the protocols worked exactly as designed; it was the design itself that led to the problems. While I expect there will be entire PhD theses written on what happened with the UST de-peg in the years to come as part of the study of tokenomics, the problem appears to have been rooted in a flawed assumption about arbitrage-based mechanisms. For arbitrage to work, you need both buyers and sellers, i.e., when liquidity dries up, you may have profitable arbitrage opportunities where no one is willing to take the other side. UST-LUNA had effectively a non-linear response curve. For small variations around $1 there was an incentive and liquidity to use the LUNA mint/burn mechanism to steer UST back to its peg. But too far of a deviation led to increasing LUNA issuance and thus inflation, and eventually hyperinflation. For UST to be pushed back, people had to be willing to take the risk of doing those trades, and at a certain point liquidity evaporated as no one was willing to catch a falling knife except for Luna Foundation Guard, who engaged in increasingly desperate attempts to prop up their failing token pair.
So is it hopeless? Should we roll up our shutters on the dream of sustainably decentralized, private stablecoins and all enroll ourselves with eCNY wallets under the watchful eye of the Chinese Communist Party? Or should Circle just report in to the OCC in chains, collect its U.S. bank charter, and stablecoins slide into the bank regulatory regimes, fading away like the wildcat bankers of the 19th century? Is the only real alternative for blockchain that it becomes the means of a more efficient and effective tyranny, to birth a more perfect Leviathan?
Despite what happened with UST, right now I hold out the most hope for algostables, possibly with some means of stablecoin token holders not just governing the protocol but also getting some yield from it to compensate for the counterparty credit risk, eliminating all the hazards that come of building stability on a finite pool of reserves or a lender of last resort. And companies like Gauntlet have done great work with linking automated governance with risk management frameworks for DeFi that I hope to see extended to supporting a future algostable that finally cracks the problem of incentive design. Because this idea, the tokenization of a stable reference price, is valuable and worth saving, and even if there is no alternative to the many flawed implementations on offer, we can hope for a better alternative to come to life one day.
It is early 2024, just before midnight. Dr. Alexandra Hamilton is in her office at Columbia University, poring over the latest simulations of her model tokenomics. First thousands, then millions then billions of random paths in the cloud: the Terra collapse; Frax’s struggles; the Iron Finance collapse; a taper tantrum and a global recession; the Fed jacks up rates 100 basis points; a 60% drop in liquidity across all the major exchanges; a war breaks out. And as the results stream back and her models assess the outcomes the code streams back line after line:
Stable.
Stable.
Stable.
Stable.
Stable.
She sits back in her chair and says to herself, “now isn’t that something?”