How Terra’s Burn Used to Work
As the Terra Classic Community embarks on its signature collective initiative, the 1.2% burn tax, it’s also beginning to consider alternatives to the burn tax which may better optimize growth potential vs. near-term supply deflation.
As we compare the current approach with Terra Classic’s pre-crash approach, as well as various longer-term burn proposals over the next several months, it’s worth reviewing how Terra Classic’s burn mechanics worked before its May 2022 downfall.
How did LUNA’s famous burn work?
The founders of Terra gave deep thought to taxes, and how taxes could be used to incentivize or discourage different behaviors on the protocol. They also experimented aggressively with different options, from a moderate tax rate to a “Treasury” module which programmatically reallocated tax revenues to reward dApps based on their success in drawing new users to Terra Classic. However, Terra ultimately discarded almost all redistributive tax policies in favor of extremely low internal taxes and reallocation of all growth surpluses (transaction fees & growth in UST supply) to LUNA token burns, after the Columbus-5 network upgrade.
However, that didn’t mean that Terra had no taxes at all. During its heyday of relatively sustainable growth (3Q21–1Q22), Terra Classic’s original “burn” and “tax” was structured very differently from the 1.2% burn mechanism recently enacted by the Terra Classic community. Its drivers were:
UST adoption: LUNA was burned directly proportional to growth in UST adoption
Every additional $1 of UST adoption, in addition to expanding LUNA’s market cap by $1 through the swap module, “burned” $1 worth of LUNA token supply, thus concentrating LUNA’s increased market cap among a smaller number of tokenholders
As UST’s market capitalization grew from $100m in December 2020 to $17bn by the end of March 2022 (170x), the token price of LUNA went up 275x, from $.40 to a peak of ~$110
This “burn” was heavily offset by dilutive issuances of LUNA to fund the drivers of rising UST adoption, especially the notorious Anchor Reserve and the Luna Foundation Guard wallet, set up to defend the UST peg during times of systemic stress
Mercantilist tax policy
Very low taxes for point-to-point transactions within the Terra ecosystem
Low taxes (.5% minimum) for exiting UST into LUNA
Rapidly increasing exit taxes (up to 100%) for funds leaving the system during high systemic stress
High convertibility: you could withdraw at least some of your money (sort of) under any conditions.
The LUNA burn: Theory vs. Practice
Drawing on some historical data, we can gather the following results.
Zooming in on the nominal changes between UST market cap (in US$) and rough estimates of LUNA supply shrink (in US$):
Estimating the supply shrinkage using the average of t(0) and t(-1) token prices in the first table (average of $.40 and $18.75, average of $6.62 and $18.75, …), every $1 in UST growth correlated to roughly $2 of LUNA burn (supply shrinkage). The total supply shrink is also net of perhaps $6 billion of LUNA issuance to build the Luna Foundation Guard (as the LFG’s initial $3.5B had to be replenished after a not-mark-to-market-related drawdown in February with an additional $2.5B in April), $550m to fund the Anchor Yield Reserve, and $1.5bn bailout in May 2021, which we think was ultimately paid in the form of 2H21 LUNA issuance to a third party (so $8bn in additional dilution). If someone in the community calculated the VWAP of LUNA burned during all these intervals (wink wink!) we could arrive at a much more precise estimate, but in the meantime, these back-of-envelope estimates strongly indicate that UST did its part in heavily reducing LUNA token supply over time; roughly 50% over an 18-month period.
Pros and Cons of the Old Model
Terra’s old burn model was thus highly effective at retiring LUNA supply and translating growing UST demand into additional accretion for LUNA tokenholders. However, its foundation had several severe flaws.
It worked as advertised on a day-to-day basis.
It incorporated a significant BTC reserve firebreak, which (20/20 hindsight notwithstanding) could’ve mitigated far more damage to the protocol than it ultimately did.
At the ecosystem level, it combined a “Hotel California effect” with high convertibility: You could always check in, but (in times of systemic stress) you couldn’t really check out (unless you paid a steep exit tax if you ran for the exits at the same time as everyone else.) However, there was no limit on withdrawals. This tax could theoretically reach 100% but in practice topped out at around 40% during Terra’s implosion.
Artificial UST demand from Anchor: An entire white paper can be written about Anchor’s flaws, its function as Do Kwon’s central bank, and its central role in the failure of the Terra Protocol. Initially, Anchor was a conventional DeFi money market which also offered a way for staked LUNA holders to access some liquidity at market-standard LTVs, i.e., it was a good product with organic demand for borrowing as well as lending. Additionally, because of the rate of USD money supply expansion in 2021, Anchor’s (20% | 26%) APYs were not only “normal” by DeFi standards — they were actually sustainable, so long as they tracked changes in USD money supply growth as other DeFi protocols did.
However, as consumer adoption of Anchor’s 20% (26%) deposit rate swamped the original core use case of borrowing against staked LUNA, Anchor created an extremely addictive feedback mechanism for the expansion of money supply that built up explosive systemic risk
Unlike other popular DeFi protocols such as Aave and Compound, whose borrowing and lending rates were set by market supply and demand, Anchor’s rate was centrally controlled by Do Kwon himself
Anchor’s “20% rate” (which was actually ~26% because devs didn’t understand the difference between APR and APY) didn’t programmatically drop as the crypto bear market of 2022 deepened, unlike its peers’. Thus, Anchor’s advertised deposit rates became increasingly attractive relative to those of peers (whose deposit rates rates plunged as borrow demand plunged), making Anchor a much more attractive place to deposit money, as long as Do Kwon advertised about his willingness to keep topping up the Anchor Reserve, which he loudly did
Thus, outside money poured into Anchor throughout January-April 2022 at the expense of Terra’s stablecoin competitors, built on the foundation of unsustainable Anchor yields
However, these Anchor deposits were zero-maturity deposits that weren’t being locked up doing anything productive (like funding a 30- or 90-day loan to a real-world business or real-world loan asset — the subject of another white paper). These deposits were mercenary capital that could leave anytime.
Other artificial, “ponzi” UST demand sources. Before Anchor became a widespread phenomenon, Terra worked with MIM/SPELL to create the infamous “Degenbox,” where Anchor’s 26% interest rate could be “looped” 4–5 times by degens for 110%+ APYs. For Kwon, it was a massive demand engine for UST (and thus LUNA pumping). The Degenbox’s absurdly unsustainable yield caused enough outrage within the Anchor community (despite Kwon throwing money at Anchor’s Yield Reserve to foot the bill for it) that it was discontinued after ~3 months.
Insufficient reserve: Terra’s BTC reserve was originally set at zero. After Terra’s crash in May 2021, Do Kwon and Jump Crypto embarked on a strategy to gradually scale up LUNA’s decentralized reserve, to ~20% of UST in circulation at the time of the crash. It wasn’t enough.
Centralized reserve: Terra’s use of reserves, nominally under MultiSig administration, was in practice controlled by Kwon and Jump. The use of the LFG Reserve during UST’s crash was extremely centralized, non-transparent (off-chain), and in hindsight, way too hasty.
The reserve’s second buildup in April 2022 was essentially funded by zero-maturity Anchor deposits. Anchor deposits less liabilities (which peaked between $11B and $15B depending on what % of Anchor borrowings you consider genuine) grew at a far faster rate than the LFG Reserve, and acted as an contra-reserve during the onset of Terra’s May 2022 liquidity crash.
If we were to launch a new dollar-indexed Algorithmic Fungible Token (we should consider dropping the term “stablecoin” because that’s not what we are doing, let alone what we should be advertising), what key variables of Terra’s old economic model (taxes, burns, deposit rates, reserve requirements, and asset/liability matching) need to change?
Domestic (Terra to Terra) tax rate: In the Quant Team’s view, Terra’s very low domestic tax rate (Terra-to-Terra txns) was a huge selling point of the network which facilitated an excellent overall UX, attracted world-class dapp developers, and laid the foundations of a vibrant and sustainable ecosystem.
Conclusion: Terra’s prior regime of very low internal taxes, and very low exit taxes during times of stability, should be retained.
Exit tax rate: Terra’s exit tax rate, which operated on a sliding scale of .5% to an effective max of 40–50% and theoretical max of 100% depending on the rate of net capital outflows from UST, was a brilliant financial innovation. However, it a) didn’t work as designed, and b) cannot be counted upon to work perfectly during the next crisis, regardless of how many improvements we make upon it.
Conclusion: Make as many improvements on Terra’s exit tax regime as possible, but assume that it will suffer significant system failures from time to time, and have a backup plan ready to avert catastrophe when it does.
Deposit rate: The Anchor Protocol began as a great onboarding hook for new users combined with a good starter use case (lending against staked LUNA collateral). However, when money supply growth inevitably reversed and Anchor’s centralized deposit rate didn’t, Anchor turned into unwitting ponzi capital.
As CeFi deposits from Celsius and others poured into Anchor in February-April 2022, Anchor’s governance process was completely unable to keep up with the tsunami of outside capital. Two activist investors (Arca and Polychain) launched a contentious activist proposal to fix Anchor’s severe market distortions by slashing its rates in March of 2022, but their proposal was fiercely opposed by Do Kwon and TFL employees for reasons partly operational (affecting the composability of aUST, the Anchor deposit certificate) and partly fatally shortsighted (Kwon by this point was single-mindedly fixed on UST’s market share relative to DAI and other competitors). The original Arca proposal, authored by Matt Hepler and Jeff Dorman, was heavily diluted after a governance tug-of-war, and became “too little too late”
Conclusion: The only bulletproof solution is letting the market balance deposit vs. lending rates, as Compound and Aave successfully have.
Reserve: Terra’s BTC reserve level was too low. We envision a “Reserve 2.0” where the decentralized reserve level is much bigger (ideally ~60% of UST circulating supply at time of BTC purchase).
The reserve’s impact on LUNC & UST safety: A much larger reserve, following prespecified parameters, would reduce the drawdowns LUNC holders would face during , thus reducing dilution from May 2021-equivalent events and drastically reducing if not eliminating the possibility of a catastrophic (95%+) LUNC loss.
The reserve’s impact on the burn: the size of the reserve is inversely proportional to UST’s impact on LUNC supply. A 60% reserve maintenance level would effectively cause the protocol to burn (1–60%) of LUNC supply for every $1 of USTC minted. This would burn at roughly half of LUNA-UST’s effective rate.
Conclusion: A 60% BTC reserve rate, combined with better capital controls, would drastically improve UST’s long-term stability and reduce volatility-induced dilution in LUNC. It would slow down the rate of burning somewhat, but also significantly reduce the dilutive effects of non-catastrophic drawdowns, and be an enormous long-term positive for systemic stability.
Conclusion: The reserve must be decentralized and defend the peg according to predictable, programmatic rules. [this will be the subject of an imminent white paper]