Not all algorithmic stabilization mechanisms are the same

We (the ‘royal we’) have previously discussed various flavors of pegged coins, “stablecoins,” as well as CBDC proposals. This short, non-comprehensive post will look into the rise and rapid fall of the Luna and UST, two cryptocurrencies native to the Terra blockchain.1

What are the separate categories that the “stablecoin” idea can be bucketed into?

Figure 1 Source: Robert Sams based on the model by Klages-Mundt et al.

Above is a helpful taxonomy created by Klages-Mundt et al. and adapted by Robert Sams.2 One of the commonalities among all of these efforts above is that they are intended to administer an elastic money supply (as opposed to fixed, deterministic, or inelastic supplies used in many cryptocurrencies such as bitcoin).

Most analysis on this topic lacks the important nuances separating custodial and non-custodial “stablecoins” as well as those that depend on exogenous collateral versus endogenous collateral.

We are not going to dive into each one of the projects above. Furthermore, the usage of a name or logo is not an endorsement of a particular company or project.

So what happened to Luna and the UST this past week?

To answer that we need to quickly explain what the Terra blockchain is and how and why there are more than one layer 1 token such as Luna, UST, KRT (Korean Won) and SDT (an SDR token). 3

Brief history

Launched just over two years ago – in April 2019 – the Terra blockchain incorporated elements of the “Seigniorage Shares” idea with a couple of twists. Whereas several other projects attempted to collateralize (back stop) a single stabilized asset through a mint/burn mechanism, Terra enabled arbitrageurs to burn Luna (the volatile, staking token) and mint one of several different pegged coins, the most prominent of which is UST. UST was marketed as being stable relative to the USD. That is to say, through an automated on-chain program, a trader could burn $1 worth of Luna (at Luna’s prevailing market price) and receive 1 unit of UST (irrespective of the prevailing market price of UST), and vice versa: a trader could exchange 1 unit of UST and receive $1 worth of Luna.4 In theory.

You might be asking yourself, what guarantees that traders will be able to redeem $1 of either at any point in time? Terraform Labs (TFL) is the main developer behind the the Terra blockchain. One of the ways TFL attempted to architect guaranteed redemption and simultaneously mitigate a “death spiral” (an existential crisis that multiple “algo stablecoins” have crashed into), was by capping the daily minting of UST.5 The exact amount has changed over time but the goal was to help throttle the unbounded risk of an oversupply of UST (or some other pegged coin).

Why is this important?

Because as mentioned above: UST (and the other pegged coins that can be minted) were explicitly uncollateralized — although there has been an implicit acknowledgement that the aggregate UST (and other minted currencies) needs to remain below the marketcap of Luna which is the key conduit for redemptions. An imbalance, or “flippening,” could (and did) result in a crisis of confidence and collapse.

Figure 2 Source: Coinmarketcap

The chart (above) shows the aggregate market caps of both UST and Luna over the past 12 months. At their height last month, they together represented almost $60 billion in (paper) value. Today that has dropped to just over $1 billion.6

Why did things go wrong?

Before we answer that, let us look at when “the flippening” occurred.

The chart (above) shows the aggregate UST marketcap relative to Luna’s marketcap over the course of a single day. At around 1am SGT on May 10th, UST’s marketcap overtook Luna for good.

What is another way to visualize this?

The chart (above) shows the same aggregates but over the course of the past 6 weeks.

What does this mean? Due to the “macro” bear market in cryptocurrencies (the aggregate coin market is more than 50% off its all-time high from last year), Luna’s market cap saw a rapid decline that quickly became a vicious cycle due to the Why.

While there are a bunch of mostly cliché conspiracy theories as to which traders took advantage of the knowledge and conditions to short Luna (and UST), the conditions that led to UST’s rapid ascent (relative to Luna) seen in Figure 2, are pretty pedestrian.

What was the key reason for this ascent starting in November? The popular Anchor dapp on the Terra blockchain. What is Anchor?

Launched in March 2021, Anchor is an all-in-one asset management dapp that allowed traders to deposit their Luna as collateral and borrow UST against it. Often traders would go to an exchange and convert the UST into Luna, depositing the Luna into Anchor and lever up several more times. Its ease of use led to rapid growth, with total-value-locked (TVL) growing from zero to $6 billion within six months. The loan-to-value (LTV) ratio shifted over time but was northward of 70% when UST overtook Luna this past week.

Why did TVL grow so fast on Anchor?

The main reason was the dapp subsidized both lenders and borrowers through the emission of a governance token called ANC. For over 6 months, Anchor marketed itself as being able to provide 19.5% APY on all UST deposits via a blended combination of Luna staking emission and reoccurring ANC airdrops.7 Both sophisticated and unsophisticated investors, believing that $1 UST was redeemable at par with $1 USD, deposited large quantities of $UST (which others could then borrow as well). Anecdotally we have heard of startups at incubators and seasoned fintechs in emerging markets offering retail users access to this high yield product. The yield was unsustainable and developers knew it so various interest groups (including several high profile investors) proposed ways to reduce the ANC yield each month depending on economic indicators.

But by the time the downward adjustment was implemented it was too late. This relatively high yield had turned UST into a “hodl asset,” a “store of value” — something that the uncollateralized system was not properly designed to absorb.8

Prior to the collapse of Luna and UST, the development teams behind Terra and Anchor recognized this shortcoming and this past February announced the Luna Foundation Guard (LFG) and organization that would accumulate exogenous collateral to defend the $1 USD peg.

Recall that at the very top in Figure 1, Terra was categorized as using endogenous collateral, that is capital native to the protocol itself (e.g., Luna, UST). As part of the initial LFG announcement, the organization aspired to accumulate large quantities of exogenous coins starting with bitcoin and later others (such as AVAX, and even both USDC and USDT). At its height, LFG’s reserves tallied over $3.5 billion and as of this writing it has shrunk to around $80 million (sans some squirrelly BTC).

Anchor aweigh

Even without Anchor the fundamental problem is that the underlying collateral is volatile, so what is over-collateralized can become under-collateralized very quickly (whether it is endogenous or exogenous).9

Those who argue that the solution for decentralized stablecoins is to be “fully backed” are still kind of missing the point. If these protocols are all using the same 3-5 major coins as collateral, you can get the same ‘death spiral’ scenario materializing if the stablecoin supply grows large vis-à-vis the collateral marketcap. After all, even LFG’s liquidation of $1.6 billion BTC moved the largest coin cap.

So who is the buyer-of-last resort? If it is actually decentralized, it can only be the parties who can liquidate or redeem the collateral. CDP systems like Maker have the incentives for this behavior, but suffer from the coin supply side being driven by lending and no mechanism to equilibrate that supply to the demand side (the mechanism is the stability fee and savings rate, but that is set by governance, not the market)

The root problem for UST and Luna, (as Kevin Zhou, Matt Levine, and others have mentioned), was that neither had any source of value independent of the other. If the market decided to sell both, there was nothing to give you confidence that they would recover. UST was built on Luna and for the past 6 months Luna was built on essentially Anchor yield savings. Even a large “stabilization fund” – with a transparent and automated mechanism for how it would be deployed – would not prevent the Luna/UST market cap from growing to dwarf the LFG backstop, thus a sequence like this past week was always a risk.

We could spend pages describing alternate plans and paths the development teams, users, traders, and other interest groups could have taken to stymie the collapse. Instead we wanted to highlight one final chart that we found interesting.

Divergence

The chart (above) shows the intraday prices each day over the course of a week between Luna (in dark pink) and bLuna (in blue).

What is bLuna? bLuna is a liquid staking mechanism managed by Lido in a partnership with Anchor.10 Liquid staking is actually an interesting concept. Most readers are probably vaguely familiar with staking on a proof-of-stake network: users deposit their coins to an address on-chain and receive some form of remuneration (emission) for helping to secure the network (and process transactions, if they are a validator).

But the coins used in staking are effectively frozen and cannot be easily used elsewhere as collateral. Enter liquid staking. As the name suggests, liquid staking is a concept that has been implemented in two different ways: at the dapp layer (via Lido, Marinade, and a few others) or at the native L1 layer (Osmosis in the Cosmos ecosystem is about to be the first to do so).

Liquid staking is neat because it allows all of the locked up (“frozen”) capital to be used as collateral for lending. An imperfect example: Bob purchases $200,000 of Apple stock. He wants to buy a new home and instead of selling the stock he finds a bank willing to use his Apple stock as collateral for the down payment on the house. Similarly, liquid staking is not rehypothecation as no new asset is created.11

The reason a lot of brain cycles have been spent on creating liquid staking dapps (like Lido) is that the vast majority (>95%) of all staked assets on proof-of-stake networks is illiquid. If they can become liquid that would enable more capital to be used for endogenous lending — instead of having to rely on exogenous capital like wrapped assets (WETH, WBTC) or real world assets (USDC, USDT).

In theory, when an asset transforms from a staked asset into a liquid staked asset, the market prices of the two should be very similar. In some cases, such as stETH (ether deposited in Lido on Ethereum) or mSOL (sol deposited in Marinade on Solana), the liquid asset accrues the emission reward therefore becoming slightly more valuable over time (in proportion to the emission rate).

In the case above, bLuna and Luna were tightly coupled but clearly broke down between May 9th-11th due to the massive selling pressure and unstaking that took place (more than 95% of all Luna has been unstaked down considerably over the past month). This brings us to the final section.

Proposed category

Surprise! I have a couple ideas on how to evolve the “algo stabilization” world, including adding a (possible) new category to the four incumbents above: a demurrage-based settlement asset.

But first, let’s take a step back and ask the question what amount of UST could Luna have absorbed?

Even the most hardened maximalist or anti-coiner would concede that a single solitary 1 UST could probably be absorbed by Luna’s market cap.

So where is the limit? Where do the wheels fall off? When do things become unwieldy?

It was not the UST borrow side that was a priori the fundamental culprit. Amplifying the problem was goosing the UST demand side with 19.5% “risk free” returns on Anchor. For instance, if the arbitrage mechanism only allowed the creation of UST (or other pegged assets) based on a small single digit percentage of Luna’s marketcap, it is likely this collapse might not have happened in such a dramatic fashion.

Yet as mentioned above, this approach alone still would not have staved off simultaneous sell-offs of both UST and Luna and/or hyperlunaflation.

Future developers looking to enter this arena could construct an asset with a stabilized unit-of-account that maintains a diminutive aggregate relative to the staked asset being burned. E.g., depending on the use case, an aggregate the size of $100,000 could conceivably power a small on-chain economy much like in traditional markets rely on a high velocity of money to grease the economy (where money is circulates among participants like a hot potato).12

That is to say, a high velocity stabilized unit-of-account, one that is used as a medium-of-exchange and not as a store-of-value or hodl asset, probably has a lot more longevity so as long as its creation (or borrowing) is not heavily subsidized. Sprinkle in some demurrage – or negative interest rates – to further disincentivize hodling and focus on a handful of uses (n.b. “hodling” is not using).1314

Final remarks

It is pretty easy to dance on the grave of another dead / dying cryptocurrency, there have been a few dozen marathon’s worth of victory laps on social media this past week. Despite autopsies and red flags, it is likely that some folks will attempt to emulate the heavily subsidized borrowing model too.

Apart from designing a purposefully limited high velocity, stabilized unit-of-account, what can non-developers do?

Arguably, the most accurate commentator on this topic is a friend, Kevin Zhou (founder of Galois Capital), who publicly predicted what would occur months ago. But unlike the maximalists and anti-coiners who stridently label everything a scam and a fraud, Zhou actually modeled out several scenarios in detail. Give him a follow.

Future analysis could look into the on-chain contagion such as dapps that were impacted including Mirror protocol (did the yield at Anchor cannibalize the other use cases by acting as a liquidity gravity well?). As of this writing it is unclear what direction a “LunaV2” will take but worth pointing out that key stake holders in the ecosystem agreed to shut down the network twice and switched to PoA.

Endnotes

  1. There are oodles of news articles exploring how the “death spiral” took place, this is not really one of them. []
  2. In 2015 Sams created the USC consortium (which has evolved into Fnality) as well as proposed the original “Seiniorage Shares” concept in 2014. []
  3. Note: according to the Terra Token Cash Flow chart, Terra was actually generating more in KRW fees (primarily via Chai) than it was earning in UST fees. The KRT ecosystem had more velocity: KRT turning over ~500 per month versus UST at a mere 1.5 times with the caveat that the KRT ecosystem is very small. []
  4. The actual arbitrage opportunity would be if UST is trading for $1.10, a trader could exchange $1 of Luna for 1 UST, therein arbing a profit while increase UST supply and bring price down. Conversely, if UST is $.90, a trader could exchange 1 UST for $1 of Luna. []
  5. There are some similarities with the collapse of Titan / Iron bank last year, although part of that involved a discrepancy with the oracle feed. []
  6. A simple way to observe the troubling trend early on was the UST / Luna marketcap ratio (based on circulating supply). Below are specific numbers that appeared in a chatroom I was in:

    April 17 7:30pm EST — 63%
    April 30 11:30am EST — 66%
    May 7 5:00pm EST — 78%
    May 7 6:30pm EST — 81%
    May 8 9:00am EST — 90%
    May 9 1pm EST — 95%
    May 9 3pm EST — 113%
    May 9 11pm EST –125%
    May 10 7:30am EST –149%
    May 10 3:45pm EST– 166%
    May 10 5:30pm EST — 207%
    May 10 6pm EST — 211%
    May 11 5:30am EST — 291%

    In April, the ratio flirted with and fell below the 2/3rd mark. But due to the persistent bear market coupled by sell side pressure of both UST and Luna, by the morning of May 10th, ‘hyper hyperinflation’ was well underway with a massive expansion of Luna’s total supply. []

  7. As mentioned in the bLuna section: users can mint a bAsset called bLuna by depositing Luna into Anchor. Staked funds are effectively pooled together by a white list of validators (users collectively share emission rewards as well as slashing events). These staked funds are used as collateral for borrowers who are subsidized through what is now a money-market. Thus there are three different tokens active in the dapp and the “19.5%” headline figure largely consists of a recurring airdrop of the ANC governance token.  E.g., if Bob deposited bLuna as collateral, he is paid out in ANC (and UST fees) in lieu of his regular staking rewards (or at least pre-crisis that was the case).  And borrowers were subsidized in the form of ANC as well.  Those who deposited UST (not Luna) received 19.5% APY up until this month (where it dropped to 18%).  This came from ANC rewards as well as a reserve fund that TFL topped up on occasion. []
  8. Some analysts think that Anchor was not that big of a deal yet at a minimum it was important as a supply sink. It is not as important in terms of how the system got insolvent; that’s more because of the underlying mint / redeem mechanism. Or as Kevin Zhou concisely explained on Odd Lots: “And they [TFL] would also use that to keep basically topping up the Anchor protocol on their yield reserve. Because they were paying more interest to depositors than they were collecting from borrowers. And, you know, I think in the end stages of Luna in its final days, you could see that the, you know, the deposit amount was way, way higher than the borrowed amount. So, you know, they, they were bleeding.” […] “I think the system was way in the past, it was already insolvent, you know, it’s just that nobody realized because they had created such a strong supply sink in Anchor for this UST, you know, if that disappeared overnight, or even gradually, the entire system was insolvent.” []
  9. Several commentators have attempted to downplay Anchor as little more than a user acquisition strategy, stating “There was nothing wrong with Anchor, they just paid more yield than what was sustainable as a growth strategy. Tons of businesses operate at a loss as a customer acquisition growth strategy.” But we can clearly see, what works for tech platform business development does not apply generally. You probably cannot integrate a heavily subsidized GTM strategy into the incentive mechanisms of your dapp or L1 without contorting the financial system you are building. As one reviewer noted: “sustainable mechanism design needs to make pessimistic assumptions (where assumptions must be made) with respect to the behavior of actors. That means minimizing mercenary behavior (e.g., “I’ll come for the subsidy and immediately depart when the freebie is removed.”). []
  10. Lido is the largest and most popular liquid staking dapp for Ethereum, Terra, and several other blockchains. []
  11. A Luna holder can pledge their Luna as collateral and receive bLuna which pay out rewards in Terra-related tokens such as UST and ANC. []
  12. In this strawman example: a stabilized unit-of-account would not need expand much so as long as its usage is high velocity. “Velocity” is an economics term used to describe how quickly the average unit of money (e.g., dollars) turns over in a given year. If this stabilized unit-of-account is only used to top up loans or fulfill margin requirements, its aggregate size would be different than a synthetic store-of-value (which is what UST attempted to be). Thus $100,000 may be sufficient to help fulfill specific sets of on-chain uses (such as those around derivatives or prediction markets). []
  13. As a friend recently pointed out: “an ‘algo stablecoin’ like Luna / UST is a form of collateralized stablecoin just different from external collateral. In this case, TFL and others were making their own collateral and hoping it retains value. They seemed to believe the amount of Luna backing UST was relatively high enough it could absorb redemptions without going into a spiral because, say, people valued those Luna tokens independently from redemption sufficiently high enough due to its governance rights over the entire blockchain that had other important commercial applications. A small coin that had limited systemic impact could be used as some sort of collateral basis and potentially survive indefinitely.” []
  14. Ultimately all public chains need base layer transactional demand to survive post-block reward. “Hypothetically, an algorithmic stablecoin could survive in the long-run, if it were to have ongoing transaction-related demand (similar to a fiat currency)” from Global Markets Daily: The Economics of Algorithmic Stablecoins, by Rosenberg & Pandl at Goldman Sachs on May 16, 2022. []

Leave a Reply

Your email address will not be published. Required fields are marked *