The Impact of the Merge on Ethereum Lending Protocols

Summary

1. Because of the presumed value of the hard fork of Ethereum, there was excess borrowing demand for ETH in DeFi lending markets. On top of this, stablecoins were expected to be worthless on the forked chain, so they fueled a race for assets that would have valued on the forked chain.
2. Aave and Compound halted/limited ETH borrowing respectively, because of market structure and liquidation concerns. Euler Finance decided to leave lending markets unchanged.
3. This caused Euler Finance to hit 100% utilization of ETH in their markets, creating the potential for a run on their protocol.
4. However, no liquidations occurred, due to low ETH volatility. Because of this, market structure was not tested.
5. Compound and Aave’s decision to halt lending markets was supported by the utilization spike on Euler. This decision should be made again in the future if another large exogenous value event occurs (airdrops or hard-forks).

Introduction

On September 15th, 2022, Ethereum transitioned to Proof of Stake (PoS) from Proof of Work (PoW). There were no interruptions in the normal operations of Ethereum and everything continued as planned on the protocol. However, on some decentralized finance (DeFi) protocols, a deleveraging event occurred. Before the merge, large lending protocols such AAVE turned off ETH borrowing entirely, as the protocol feared insolvency if position liquidations became impossible¹. Others, like Compound, limited WETH borrowing to a fixed cap².

However, Euler Finance decided that limiting ETH borrowing or increasing the max interest rate from 100% was not necessary³. The team felt that the borrowing market would stay efficient by pricing in floating interest rate risk, and that the liquidation concerns on Euler would be controlled by alternative liquidation flows. Because of this, we can analyze issues and mechanics behind lending markets before, during, and after the Merge.

How do Ethereum lending markets function?

Lending markets on Ethereum function similarly to money market mutual funds in traditional finance. Lenders provide assets and borrowers then lend them at some rate. Ethereum lending protocols such as Aave, Compound, and Euler expanded on these ideas by creating permissionless lending pools that allow anyone to lend and gain yield and for any borrower to borrow these assets. In traditional finance, only a specific subset of groups could lend and borrow from these funds, creating gated yield opportunities. In exchange for opening these opportunities up to the public, Ethereum lending markets require overcollateralization of the borrowers positions to ensure the lending protocol can be made whole if the loan becomes undercollateralized (the assets borrowed are worth less than the collateral). To automate this, lending protocols are able to easily facilitate market transactions compared to traditional finance by a few novel innovations.

First, price oracles allow the lending protocol to know the approximate price of the collateral used against a loan. Many traditional finance assets used as collateral are impossible to calculate, and if they are mistakenly priced they can lead to bad debt (debt worth less than the assets loaned out) for the protocol. Mispriced collateral caused many banks in the 2008 financial crisis to have too much bad debt and go bankrupt. Second, lending protocols benefit from the composability of the Ethereum ecosystem. Lending markets can easily tap into the liquidity of decentralized exchanges to facilitate the liquidation of undercollateralized positions. If a position gets close to becoming bad debt for the lending protocol, the protocol (or operators working on behalf of the protocol) can liquidate the debt using decentralized exchanges. Both of these innovations allow for more users to access better yield opportunities and smaller borrowers to access more efficient borrowing opportunities.

What is the difference between Euler Finance and Compound/Aave?

Euler Finance innovated on Compound/Aave by allowing lending and borrowing of long-tailed crypto assets in addition to pairs supported by Aave/Compound. They did this through a few additional protocol innovations, such as improved liquidation auction design and more price oracle choices from Uniswap v3. This met an untapped demand from asset managers and traders to short long-tail tokens by borrowing, and allowed token holders to earn yield on their long-tail tokens.

Figure 1 — Growth of TVL among major lending protocols

Source: Delphi Digital citing DefiLlama

It is important to note the double y-axis on this chart. While Compound and Aave have lost TVL over the past year, Euler continues to gain. One hypothesis is that Compound and Aave have been hurt by better yield opportunities elsewhere (both in crypto and traditional finance), while Euler has found a product-market fit — lending on long-tailed tokens — that is different from Compound and Aave, which have a manual process to add new assets.

How were lending markets impacted by the Merge?

Unlike Euler, Compound and Aave’s ETH markets were relatively unimpacted by the Merge. This was a calculated decision to subvert market dynamics by the governance of these protocols by limiting/stopping ETH borrowing for Compound and Aave respectively.

Figure 2 — ETH Borrow Rates on Compound

Source: Website of Compound Finance

Compound maxed out at around 6.86% APY with 113.72 million dollars of ETH borrowed. This is far below efficient market prices because of the limit in ETH borrowing. Market implied rates are expected to center around the future expectation of the price of ETHW, so borrowers would be willing to pay enough in interest rates to offset this receiving an asset at this price. Aave turned off ETH borrowing entirely, so their ETH markets were unchanged.

Figure 3 — ETH Borrowing Rates on Euler

Source: DuneAnalytics Query by author

Unlike Compound and Aave, Euler reached the end of their possible interest rates approximately one day before the merge. There was an earlier spike in interest rates on September 10th, which returned to around 2% on September 11th. Lending protocols such as Euler generally have kink interest rate structure where their interest rates increase slowly until a certain kink point, then move rapidly. Small fluctuations (~3–5%) in borrowing and lending lead to these massive swings.

Figure 4 — ETH Utilization and Availability of borrow

Source: DuneAnalytics Query by author

On September 10th, almost 20 thousand WETH was removed from Euler in one hour. This immediately caused the first spike in interest rates as utilization of WETH in Euler hit 100%. This WETH was removed by three large wallets both paying off WETH debt and calling back WETH lent to the protocol. (These wallets returned their WETH to Euler within a few days after the Merge.) This suction of liquidity and spike in interest rates created a brief scramble for yield as new lenders entered the market, which then returned interest rates back to normal.

After this, more and more WETH was borrowed until a day before the Merge, at which point utilization of WETH in the protocol hit 100%.

What is the problem with 100% utilization?

Lending protocols face a unique problem when borrowing of one asset reaches 100%. This is analogous to a bank run in traditional finance. A bank run is when everyone attempts to remove their money from a bank at the same time, forcing the bank to liquidate assets at a lower value to pay back their lenders. The opposite is true in a lending protocol run. Lenders borrow so much from a protocol that loans collateralized with that token cannot be liquidated, as there are no assets in the protocol to liquidate.

This has happened with many lending protocols on the Terra ecosystem, which had the UST (an algorithmic stablecoin that was intended to stay at $1) hardcoded to $1. When UST broke one dollar, users lent assets to the protocol knowing that their collateral was worth less than the protocol expected, and essentially borrowed so much from the protocols that they were no longer solvent.

Issues such as these were the reason that Compound and Aave turned off/limited the amount of ETH borrowing. The governance of these protocols wanted to prevent a lending pool run.

Euler governance felt that these concerns were unfounded for their protocol. They had reason to believe that the Euler Protocol was uniquely positioned to weather these problems because of its unique liquidation mechanism.

What is different about Euler?

Unlike Compound and Aave, Euler has a unique liquidation mechanism through which a liquidator can take the loan on themselves using their own collateral. Liquidators are rewarded for unwinding this bad debt by acquiring it at a discount. Because of this lending mechanism, the Euler team felt comfortable with no changes to their Protocol’s behavior⁴.

However, it is unclear if this was viable. Taking on the loan for a premium requires the liquidator to essentially run a book of loans themselves. To do this, they need to source collateral and hedges to remain profitable and safe from market fluctuations. Liquidators in DeFi generally run lean books focusing mainly on atomic arbitrages where they strictly hold zero capital. It is unclear if there are actors in the space equipped, willing, or able to support this alternative liquidation strategy.

What about the type of collateral on Euler?

Figure 5 — Borrowing behavior by asset on Euler and Compound

Figure 5a — Borrowing on Euler

Source: Dune Query by author

Figure 5b — Borrowing on Compound

Source: Dune Query by author adapted from Dune Query by Pierre Yves

Another big difference between the Euler and Compound is the type of borrowing activity. The majority of Compound’s borrowing activity involves using risky collateral to borrow safe assets. Comparatively, Euler has significant borrowing demand for risky assets like WETH and Lido-staked ETH (stETH).

Figure 6 — Leveraged stETH and ETH borrowing

Source: Dune Dashboard by index_coop

This brings up another potential issue with allowing floating ETH interest rates. A common tactic for yield in lending markets is leveraged lending and borrowing of WETH and stETH. When ETH interest rates are low and the stETH peg is in-line with ETH, these positions are net profitable. Some borrowers will 3x leverage their ETH positions to gain extra yield. With high interest rates, these positions have to unwind into brutal market conditions, this causes interest rates to fluctuate more and the stETH peg to deviate from ETH. This further exacerbates the difficulty of unwinding these leveraged positions. This was noted as a reason for halting Aave lending markets (another protocol with high stETH lending)⁵.

Leveraged stETH positions on Euler were underwater for an entire day before the day before the merge (Figure 6), but the market was able to absorb the unwinding of these stETH positions. The unwinding of leveraged stETH positions was a major shock to ETH liquidity during the crash in May, and forcing these positions to de-leverage before the Merge was an unnecessary risk.

Perhaps Euler felt that their operators were prepared for the market fluctuations. Liquidations / TVL on both protocols are similar in the past 90 days⁶. Because Euler is onboarding much riskier collateral against risky assets, expectations would be that Euler would have more liquidations. Because their liquidations are lower than Compound per unit, this could lead Euler to feel their marginal borrowers are more sophisticated than marginal borrowers on Compound.

Who was right?

Ultimately, no borrower defaulted on Euler in the time period before the merge, so no liquidation cascade occurred⁷. This was due partly to historically low ETH volatility and partly due to infrequent liquidations on most platforms. stETH positions held firm through the negative yield, as it was believed at the time (and would come to be true) that ETH yields would stabilize.

Figure 7 — ETH Volatility

Source: Dune Analytics by author

However, another situation will eventually cause a mass borrow event to occur on lending markets. It could be another fork or an airdrop. With this in mind, the ecosystem should halt lending markets to ensure that mass runs on lending markets don’t occur.

Conclusion

The Merge from PoW to PoS caused market participants to flood Ethereum lending markets with excess borrowing demand for ETH. While lending markets capped/halted ETH lending to ensure that markets functioned soundly, Euler Finance decided this was not needed. Euler proceeded to reach 100% utilization of their borrowing markets for almost a day before the merge. While no mass liquidation event occurred due to several market factors, it serves as a lesson for future lending protocols to halt lending when exogenous events mass impact lending demand.

[1]: https://governance.aave.com/t/arc-aave-ethpow-fork-risk-mitigation-plan/9438

[2]: https://www.comp.xyz/t/proposal-adjust-eth-interest-rate-model/3493

[3]: https://forum.euler.finance/t/rfc-don-t-alter-the-eth-interest-rate-model-ahead-of-the-merge/370

[4]: https://forum.euler.finance/t/rfc-don-t-alter-the-eth-interest-rate-model-ahead-of-the-merge/370

[5]: https://governance.aave.com/t/arc-aave-ethpow-fork-risk-mitigation-plan/9438

[6]: See https://dune.com/queries/1443860 and https://dune.com/queries/1443861 by author
Euler 90 day liquidations / TVL = 0.002268 and Compound 90 day liquidations / TVL = .00149

[7]: See https://dune.com/queries/1443059 by author