Today saw the meltdown of Iron Finance, issuer of a supposed “stablecoin” called IRON that is now worth a lot less than the dollar it’s supposed to be pegged to. That’s obviously bad, since the entire point of a stablecoin is to be stable.

Stablecoins are important. They’re widely used by crypto speculators and day-traders as supposedly safe places to park their money between trades, without the hassle and expense of converting to dollars. That demand has made the category competitive, and Iron’s collapse is a big black eye, not just for its creators, but for other tokens that take a similar approach – what are broadly known as “algorithmic” stablecoins.

David Z. Morris is CoinDesk's chief insights columnist.

Algorithmic stablecoins are one of at least three varieties. Fiat-backed stablecoins such as Circle’s USDC are fully backed up by traditional currency. A second category is crypto-backed stablecoins, such as MakerDAO’s DAI. Dai is overcollateralized to increase stability in the event of a crypto crash (which creates some problems of its own).

(Tether, somehow the most widely-used stablecoin, is in a messy category by itself.)

Finally, algorithmic stablecoins use less than one-to-one backing and a complex web of automated transactions and issuances to maintain a price peg. The details of these systems are spectacularly arcane.

Iron Finance’s stablecoin (IRON) was 75% backed by USDC, which is fully backed by dollars. The other 25% of backing was provided by TITAN, a “share token” or “balancer token” also created by Iron Finance itself. The Titan token was burned or printed as needed to maintain Iron’s one-dollar peg. (The Daily Gwei has a more detailed breakdown of the IRON system.)

Other algorithmic stablecoins may be more sophisticated or robust than what Iron built, but most use a similar multi-token system. You may begin to see the intractable core problem: In very broad terms, these projects are fractional reserve banks. Bitcoiners and other “hard money” advocates loathe fractional reserve banking like vampires loathe garlic because of the risk of exactly the sort of catastrophe that hit Iron.

Things went south for Iron starting overnight Wednesday when the value of the Titan share token fell from $65 to $60 – and then quickly to what I’ll just call “effectively zero.” The drop was so sharp in part because, once Titan fell just slightly too low, the incentives behind the stablecoin became perverse, creating an arbitrage opportunity that was ruthlessly exploited. The role of an arbitrage parallels the fall of the English pound’s peg in the late Eighties, and it will be interesting to find out who made the George Soros money in this crypto-replay.

Iron’s creators attempted to not only justify, but seemingly to glorify their failure by claiming in a postmortem that “we just experienced the world’s first large-scale crypto bank run.” To which I would say two things: a) No, you didn’t, and b) those words don’t mean what you seem to think they mean.

First of all, the “world’s first large-scale crypto bank run” happened in 2014, when customers around the world struggled frantically to get a total of 850,000 BTC out of Mt. Gox, an early bitcoin exchange. Those were worth about $425 million at the time, or roughly $33 billion at today’s prices. And customers were right to scramble: despite calming assurances from those in charge, Mt. Gox really didn’t have their money. Most of it was lost in an apparent hack, and as of January 2021, only about 23% of depositors’ bitcoin has been recovered.

It’s not clear how much Iron holders have lost, but the protocol at its peak had $2 billion in collateral on its books. Even if all of that’s gone, compared to Gox, those are rookie numbers.

There is another difference here. Mt. Gox actually had the money at some point, then lost it. But the entire point of Iron and other algorithmic stablecoins is that the money isn’t really there in the first place: no matter the complex rules for burning or printing Titan, Iron was still a dollar-denominated token backed only 75% by anything resembling dollars. This was ultimately why a slight price fluctuation led to death-spiral panic-selling: Everyone wanted to get out before the bottom of the barrel was in sight.

So, just as with Mt. Gox – and despite what Iron Finance argues in its aw-shucks-non-apology – this bank run didn’t cause the collapse of the system. Here’s a fantastic quote on the subject from banking economist George Kaufman:

“The fable is that a run can bring down a solvent bank. What a run does is: It causes an insolvent bank to be recognized as insolvent.”

Kaufman wasn’t talking about Mt. Gox or Iron, but about Enron, a company that was essentially one very big leverage-driven scam from the bottom to the top. Through its masterful (and evil) manipulation of equity and accounting rules, Enron spent more than a decade propping up its stock price by essentially printing its own money, in the form of varieties of stock in itself and subsidiaries.

It used that money, in various arcane maneuvers, to buy parts of itself from other parts of itself, then book those sales as revenue reported to Wall Street. This is how Enron’s revenue and stock managed to grow at an impressive and steady pace. It was all fake, and eventually that became clear to the public.

Are all algorithmic stablecoins subject to the same risk? Can you print your own money, then move it around in a sophisticated enough way that it somehow becomes worth 25% of an actual dollar? I remain willing to be convinced. But until that happens, my personal assumption is that algorithmic stablecoins are basically magic beans. Mark Cuban now wants them regulated, but in the meantime, you can protect yourself from another “classic bank run” by simply steering clear.