Stablecoins Need Collateral
The biggest news item in “cryptoland” this week is definitely the entire Terra ecosystem completely imploding. First, TerraUSD (UST) depegged, then its backing asset LUNA rapidly entered a hyperinflationary spiral and went down to zero.
Of course, everybody’s talking about it, but one thing I’ve noticed is that people usually focus on random “current events” like Anchor’s unsustainable yields, speculation on Do Kwon’s past history of failed stablecoins, how “lunatics” are going crazy, etc. Even when people try to draw some sort of general lesson out of the collapse, it oftens spirals into hilariously bad takes — “algorithmic stablecoins are doomed” “algorithmic stablecoins need strong leadership to do open market ops” “Do Kwon rugged us all” etc
“Algorithmic stablecoins” do use collateral!
Instead, the first lesson to learn is that all stablecoins need to be fully collateralized. This might seem obvious, but honestly it’s hard to count how many times I’ve heard of algorithmic stablecoins referred to as “unbacked”, or “backed by the algorithm”. But any stablecoin needs an answer to “if the stablecoin dumps, what store of value props it up”. If the answer is “lol nothing”, you just don’t have a stablecoin.
“Algorithmic stablecoins” are simply backed by collateral that’s a little more exotic and implicit than an inert stash of ETH, dollars, or whatever. That’s also one reason why I hate the term “algorithmic stablecoin” — it misleadingly suggests that all that matters is the magical pegging algorithm, even though all stablecoins’ safety relies on sufficient collateralization, whether it be stashed USD (USDC), expected capital inflows from interest rate increases (one part of MakerDAO), value from diluting and selling off another asset (TerraUSD), etc etc. The other reason is that the boundaries are really fuzzy: is MakerDAO, which uses a mix of implicit and explicit collateral, a “backed” stablecoin or an “algorithmic” one?
So for the purposes of this post I’m not going to talk about “algorithmic stablecoins” as a whole, but focus on what I’ll call equity-dilution stablecoins, a family that includes TerraUSD but also other systems like FRAX and in fact, Themelio’s native currency MEL. An equity-dilution stablecoin has two currencies and a very simple peg mechanism:
- A stablecoin that’s supposed to be always worth $1 (or whatever the peg target is)
- A separate volcoin (“volatile coin”) with a volatile price
- A dilution-based peg, where 1 stablecoin can always be destroyed to obtain $1 worth of freshly-minted volcoin, and vice versa. (“$1 worth of volcoin” would be determined by something like a price oracle)
For example, in TerraUSD’s case, the stablecoin is UST, the volcoin is LUNA, and the central mechanism guarantees that you can burn 1 UST for $1 worth of freshly minted LUNA and vice versa.
Here, the implicit collateral of the stablecoin is the total dilution-extractable value (let’s call it DEV) of the volcoin — how much money you can squeeze out of inflating more volcoins and diluting existing volcoin holders. If DEV exceeds total stablecoin issuance, you’re good, but if not, the whole system is insolvent and will collapse sooner or later.
A very important thing to note is that DEV is not the normal-case market cap of the volcoin. Instead, it’s actually significantly less than the volcoin market cap (and never more) for several reasons. Most obviously, diluting the volcoin itself will decrease the market cap since it “punishes” volcoin holders, so if all the stablecoins are getting dumped, you can’t extract the pre-dump volcoin market cap.
More insidiously, depending on the stablecoin design, demand for volcoins might be strongly correlated with demand for stablecoins. This means that exactly when you need DEV the most — a crash in stablecoin demand — you discover that there actually isn’t any DEV at all, since the volcoin crashed right with the stablecoin and nobody wants to buy newly inflated volcoin.
UST didn’t have (nearly) enough collateral
The problem with UST is that it didn’t have even close to enough DEV to support its massive issuance. In other words, UST has always been insolvent, and recent market conditions simply exposed just how unviable it was.
This is because UST had a particularly extreme version of the “volcoin demand correlated with stablecoin demand” problem. In fact, almost all LUNA demand was driven by UST demand. This demand did prop up LUNA’s raw marketcap to be higher than UST’s, perhaps giving an illusion of overcollateralization. But due to this correlation, none of this market cap actually counts towards DEV!
This is because despite Terra’s attempt to be an independent L1, all it really accomplished was hosting the UST/LUNA pegging mechanism. In fact, the biggest “value prop” of LUNA was essentially increasing UST demand causing LUNA to “moon”. Worse, UST demand was artificially increased through a number of mechanisms, the most infamous probably being the subsidized, unsustainable 20% interest rates on the Anchor platform.
On the surface, this massively increased LUNA marketcap and bolstered UST’s security. In reality, the opposite happened, since none of this UST-demand-source marketcap would count at all towards DEV when UST actually needs its collateral. All of the boosts to UST demand increased UST issuance without increasing its collateral.
So even before the crash, UST was already deeply insolvent, with issuance far outstripping collateral. Almost none of the LUNA marketcap actually contributed to UST collateral — UST was essentially backed by nothing but leveraged positions on itself, a classic example of “troll physics tokenomics”.
And indeed, once UST demand made the slightest of decreases, the “death spiral” began. The seemingly sky-high LUNA marketcap, supported nearly entirely by hopes of ever-increasing UST demand, started to plummet. Once it crossed the UST marketcap, which is certainly a powerful “meme” if not an accurate measure of DEV, everything utterly unraveled.
(As an aside, the granddaddy of equity-dilution stablecoins, “seigniorage shares”, has an even more extreme version of this problem. There, the only purpose of the volcoin is to “hold the bag” for the stablecoin, and there’s not even a pretense of any independent utility. Seigniorage shares has close to zero DEV and is about as insolvent as stablecoins get)
Overcollateralizing an equity-dilution stablecoin
So are equity-dilution stablecoins doomed? No, but they need to be carefully designed to ensure that DEV > issuance. This is especially important since there are many other stablecoins using equity dilution (e.g. FRAX). In fact, even Themelio’s native currency MEL is an equity-dilution stablecoin (albeit one that is not fiat-pegged).
Ensuring overcollateralized DEV requires designing the mechanism so that 1) most of the value that accrues to the volcoin don’t come from stablecoin demand 2) stablecoin issuance does not outstrip independent volcoin value. In particular, mechanisms that boost stablecoin demand yet don’t boost stablecoin-independent volcoin demand (like Anchor’s subsidies) actively hurt solvency.
In Terra’s case, for example, if Terra had succeeded as a strong general-purpose blockchain, this would give LUNA a strong and diversified demand beyond factors correlated to UST. Couple this with avoiding destructive mechanisms like Anchor that subsidize UST demand without increasing collateral, and the system would probably have been solvent and the whole collapse averted.
In the design of Themelio’s non-fiat “stablecoin” MEL, an overcollateralized DEV is a core design principle. In particular, the volcoin counterpart of MEL, Themelio’s PoS token SYM, is designed to capture more value from on-chain activity than usual PoS tokens. The MEL currency is also a rather small portion of Themelio’s value proposition, minimizing the MEL-correlated fraction of SYM demand.
This gives us a situation where DEV is some fraction of on-chain “GDP”, while volcoin issuance is closer to on-chain “M0” — and as illustrated in the original Melmint paper, GDP typically far outstrips M0 in real, diversified economies, so we are fairly confident that MEL will be overcollateralized.
“Anti-run” depegging to deal with illiquid collateral
One final point here is that even a well-collateralized stablecoin can run into issues if the collateral is illiquid. That is, the collateral (e.g. an asset pool, DEV for dilution stablecoins, etc) might be worth more than the stablecoin if sold gradually, but can only be sold quickly at a large discount.
The danger with an illiquidly collateralized stablecoin is that it’s vulnerable to runs. A mass “dump” of such a stablecoin can itself cause its collapse by discounting its collateral — and much worse, merely the fear of a dump will cause a dump, since everybody is incentivized to sell their stablecoins before the feared collapse, ensuring that the collapse does happen. In the case of equity-dilution stablecoins, runs are especially destructive since they also cause a hyperinflationary death spiral of the volcoin.
Preventing runs, though is actually quite easy, though a bit paradoxical: in crises where solvency is in doubt, whether due to bad collateral fundamentals or a temporary liquidity crisis, the stablecoin must rapidly depeg to fair market value rather than trying to defend an unsustainable peg. This is because, paraphrasing economist George Selgin, runs occur when three things are true at the same time:
- Everybody’s promised a fixed sum (e.g. $1)
- There are not enough assets to pay everybody off (e.g. stablecoin is undercollateralized/bank’s investments failed horribly)
- The first person in line to redeem his liabilities gets more money than the last.
Rapid depeg to a level where the stablecoin is no longer undercollateralized fixes 1 and 3, because when there are credible fears of stablecoin insolvency, the stablecoin would have already depegged. There’s no more incentive to dump the stablecoin before other people do: everybody is already too late.
Of course, a stablecoin should never get undercollateralized in the first place. But the really nice thing about anti-run depegging is that it no longer makes temporary liquidity problems destabilizing: they just cause a very short-run depeg. Absent a mechanism for a run, only those with strong short-term liquidity needs would dump the stablecoin, while others would be incentivized instead to “buy the dip” and profit from the depeg.
(As an aside, a familiar example of a run-proof, depeggable “stablecoin” is a money market ETF. Like a fractional-reserve bank account, in normal market conditions a money market ETF is “cashlike”, but unlike bank accounts, liquidity crises simply cause them to drop below the “peg” temporarily instead of leading to destructive runs)
Themelio’s Melmint has a very explicit anti-run depeg: how much MEL can be burnt for SYM and vice versa is directly proportional to actual liquidity available in the on-chain MEL/SYM pool, automatically turning the mechanism off in a liquidity crisis without the need for any kind of intervention. But many other stablecoins do have analogous mechanisms, often triggered through governance: MakerDAO has a run-proof “emergency shutdown” procedure, for example, that is designed to respond to a solvency problem by shutting down the whole system and paying back liabilities at current market value rather than trying to hold the peg.
Even TerraUSD had something resembling a depegging mechanism: a “spread fee” applied to the pegging mechanism that effectively rations attempts to convert 1 UST to $1 worth of LUNA, causing the peg to break if a run starts and hopefully slowing it down.
Unfortunately, after UST started depegging, the Terra community decided to weaken this depegging mechanism in what seems to have been a desperate attempt to save the peg. This, of course, supercharged the run on UST, sent LUNA quickly towards hyperinflation, and wiped out at least $30 billion in wealth.
Conclusion
In conclusion, the biggest takeaways from the whole TerraUSD fiasco are twofold:
- Fully collateralize your stablecoin, especially if it’s “algorithmic”. “Algorithmic” stablecoins don’t give you magical unbacked stability.
- Gracefully depeg to prevent runs rather than attempting to keep an unsustainable pegs.
Apparently, the Terra community is trying to resurrect the project in some form— if another stablecoin is the result, I do hope that this time it would be fully collateralized and run-proof.