[ARFC] CRV Interest Rate Curve Upgrade


title: [ARFC] CRV Interest Rate Curve Upgrade
Author: @Llamaxyz - DeFi_Consulting, @dydymoon and @scottincrypto
Dated: 2023-01-11
Revised Date: 2023-01-31


Simple Summary

@Llamaxyz presents a proposal to update the CRV interest rate parameters on the Aave Ethereum v2, Ethereum v3 (when deployed), Polygon v3, and Polygon v2 Liquidity Pools.

Abstract

The CRV SupplyCap on the Polygon v3 deployment is currently at 100% with borrowing costs of 24.66% with a Utilization of 47.76% relative to a 45% Uoptimal value, [1]. This proposal intends to increase the CRV SupplyCap on Polygon v3 deployment, improve the capital efficiency of each liquidity pool, and increase the amount of revenue generated across the various CRV reserves.

Capital efficiency is improved by increasing the Uoptimal value. Revenue is increased by introducing a Base interest rate, increasing the Slope1 and aligning the Reserve Factor to 20% across all pools. This is expected to generate more revenue for the DAO and significantly reduce the amount of free CRV in aggregate across the reserves.

Motivation

With the CRV reserve on Polygon reaching its SupplyCap, this is an ideal time to implement an upgrade that addresses the SupplyCap and Interest Rate parameters. The current CRV Interest Rate reflects the original parameters implemented when the liquidity pools were deployed. The Interest Rate Parameters do not take into consideration the alternative use cases for CRV and consequently, Aave has missed out on revenue from the sustained borrowing demand.

Llama expects the sustained borrowing demand to continue due to the veTokenomics construct presenting several opportunities for CRV borrowers to earn yield. In this proposal, the Uoptimal parameter is to be increased from 45% to 70% improving the capital efficiency of the active reserves.

image

image

Specific to the Polygon v3 deployment, increasing the Uoptimal parameter from 45% to 70% will lead to an increase in the BorrowCap, which is a function of the SupplyCap and Uoptimal parameter.

BorrowCap = SupplyCap * (Uoptimal + 0.1)

The new SupplyCap parameter has been prepared by Chaos Labs, 1,125.24K. This generates a new Borrow Cap of 900.19K, [2].

With borrowing disabled on some Aave deployments, lending rates will drop and it will become rational for profit-driven actors to transition liquidity to the Reserve that offers the greatest yield. This has been playing out with demand on Polygon v3 increasing relative to other deployments.

The interest rate at the current Optimal point is 7%, which is low relative to current demand of 24.66% on Polygon v3.

image

This proposal recommends introducing a Base and increasing the Slope1 parameter to 3% and 14% respectively. This will lead to a substantial change to the current curve. However, the current borrow rate still exceeds the borrow rate at the Uoptimal utilization point. Over time, we will be able to monitor the reserve and amend the Slope1 parameter t to optimize for utilization and revenue generation for Aave.

The graphic below shows the changes in the interest rate.

image

The Reserve Factor on Polygon v3 is to be increased from 10% to 20%, bringing it in line with all other Aave deployments.

Specification

The below table shows the current and proposed changes to the CRV Reserve on Polygon v2, Polygon v3, Ethereum v2 and the soon to be deployed Ethereum v3 liquidity pool.

Polygon v3

Parameter Current (%) Proposed (%)
SupplyCap 937.70K 1,125.24K
BorrowCap 640.44K 900.19K
Uoptimal 45 70
Base 0 3
Slope1 7 14
Slope2 300 300
Reserve Factor 10 20

Polygon v2 - Frozen

Parameter Current (%) Proposed (%)
Uoptimal 45 70
Base 0 3
Slope1 7 14
Slope2 300 300
Reserve Factor 20 20

Ethereum v3 - Not yet deployed

Parameter Current (%) Proposed (%)
Uoptimal NA 70
Base NA 3
Slope1 NA 14
Slope2 NA 300
Reserve Factor NA 20

Ethereum v2 - Frozen

Parameter Current (%) Proposed (%)
Uoptimal 45 70
Base 0 3
Slope1 7 14
Slope2 300 300
Reserve Factor 20 20

References

[1] Aave - Open Source Liquidity Protocol
[2] AAVE V3 Borrow Caps Methodology
[3] Creative Commons — CC0 1.0 Universal

Copyright

Copyright and related rights waived via CC0, [3].

6 Likes

Thanks, @Llamaxyz - Gauntlet is conducting analysis on the market risk side for this proposal. There may be nuances with the risks here associated with CRV, and we target providing analysis in ~2-3 weeks.

1 Like

Hi @Pauljlei,

@ChaosLabs has already provided the revised SupplyCap

We plan on posting a Snapshot later this week.

We defer to the community preferences on the expediency required here. Gauntlet will continue our modeling and publish it as soon as possible regardless.

In the interim, our IR methodology demo and other research might be helpful for the community to make an informed decision on next steps.

Hi Everyone :wave:

We have created a Snapshot starting tomorrow and extending to Sunday.

https://snapshot.org/#/aave.eth/proposal/0x2cb10cfb57a79bb97c3aed1cc3e9847227fb0f6a843916921ae315b9d8ad11d3

Thank you in advance to everyone who participates in the vote.

1 Like

While we would have preferred waiting for @inkyamze Gauntlet recommended parameters for this proposal

The current ACI position on this proposal is YAE. we will wait until the last day of the snapshot vote or the publication of the recommendation from gauntlet to post our vote or change our position.

We understand that risk teams have backlog and not infinite resources, and we have to balance this with the strategic aspect of making the protocol evolve and stay competitive will sometime lead to moving forward without both reports.
.
That being said, when it’s an option, having both Chaos & Gauntlet reports are true net positive for AIPs.

4 Likes

Thanks for identifying this opportunity to optimise CRV revenue. While it’s ingenious to make some changes, I’m surprised to see a single snapshot bundling so many changes for CRV, each with their own risk implications.

What are your considerations (besides revenue) in the interest rate model calibration?

  • Doubling slope 1, means a volatile borrow rate at low levels of utilisation, increasing risks for users
  • Increasing the Optimal Utilisation to 80% and the borrowing cap to 90%, would encourage high utilisation leaving little liquidity for liquidations and withdrawals

Today, users are paying 58% to borrow on Polygon V3, this would take us to 84% utilisation based on this Llama model, with only 16% of liquidity available, not enough to liquidate PolygonV3 top supplier of CRV who is actively borrowing against.

I’m in favour of optimising the interest rate model for revenue but not increasing risks so significantly. I believe its important to maintain low rate volatility at low utilisation for usability, while preserving available liquidity for solvency, for example:

  • U 70% → Max borrow cap 80%
  • Base 3%
  • Slope 1 7%
  • Slope 2 250%

→ Users could borrow up to 75.5% of the asset pool for cheaper than they are paying today

Still, I see no reason to skip the reasonable delay for Gauntlet to make their recommendation. They just expanded their mandate to cover interest rate model, with a robust qualitative model. Risk contributors represent a significant expense for the DAO, it doesn’t seem reasonable to make risk decisions without leveraging this ressource.

3 Likes

Hi @Alex_BertoG,

With a slope1 parameter of 7%, all communities with veTokenomics tend to have a Reserve utilization consistently around the Uoptimal value. The gradient of when the utilizations meets the following criteria, Uoptimal < Utilization < 100%, is fairly steep on the existing interest rate curves, 5.33 current versus 6.8 proposed.

When,
Uoptimal < Utilization < 100%

Then gradient,
(Slope2-Slope1)/(100-Uoptimal)

Current gradient on Polygon v3, is 5.33 and proposed gradient is 6.8. Meaning, as utilization increases by 1%, the borrow rate increases at 6.8%. If the Slope2 parameter was changed to 120.5%, this would reduce the gradient equal to the existing interest rates when utilization exceeds the Uoptimal value. Having a greater rate of change (6.8 > 5.33) when utilization > Uoptimal is desirable for Aave. This encourages utilization to stay < Uoptimal value.

Why the Base and Slope1 parameter increase ?

When,
0 < Utilization < Uoptimal

Then gradient,
(Slope1-Slope2)/(Uoptimal)

The current gradient is 0.1556, the proposed is 0.1375. In fact, the newly proposed interest rate curve have a lower gradient than the existing curves which means the “volatility” (more accurately defined as the rate of change) is lesser than the current proposal. The proposed curves actually lead to a better User experience as utilization can consistently fluctuate across a wider range of utilization with a lower rate of change.

The above statement is factually incorrect. The gradient of the proposed interest rate curve is lesser than the existing gradient. This reduced rate of change, gradient, reduces volatility when the following condition is met, 0 < Utilization < Uoptimal.

Lets touch on the Uoptimal value, current 45 and proposed 80, the intent here is to improve a very inefficient underperforming CRV Reserve. 80 is high that is true. With a Slope1 of 14%, the desired affect is for actual utilization to be < Uoptimal. However, we do not intend on jumping Slope1 aggressively to achieve this. In our opinion, doubling the Slope1 parameter is probably the maximum single increase we should perform. This approach is derived from more holistic thinking that takes into consideration the user experience and knowing that we can simply delivery another upgrade at a later date. Future proposals may increase the Base and Slope1 value to achieve this over time.

We expect utilization to increase and for the market to find equilibrium beneath Uoptimal utilization. Our view is firm on the 14% Slope1 value, however pivoting Uoptimal to 70 is something we can do and in time we can walk this Uoptimal value higher pending what the markets feedback is.

If Aave was to have a policy that assumes all liquidation liquidity is derived from Aave itself, then yes the point around BorrowCap being to high is true. However, this neglects all other available liquidity pools and Market Makers who hold the asset, it is a restrictive conceptual limitation. Given the linear relationship with liquidation fees, the fee offsets any price impact from purchasing CRV. We believe liquidation liquidity should be sourced more widely than just Aave and thus prefer a higher BorrowCap. However, similar to the Uoptimal value, we can start lower and walk this up over time.

Llama will likely be expanding its services to offer an alternative quantitative model to Chaos and Gauntlet. For this proposal the SupplyCap was provided by Chaos Labs and the CRV is at 100% capacity for some weeks before Llama put forward a plan.

1 Like

Gauntlet Recommendations

After examining Llama’s IR curve recommendation, Gauntlet would recommend making the following changes to the proposal:

  • Setting the Uoptimal to 70%
  • Maintaining the Slope 2 at 300%.

Rationale:

CRV’s interest rate on Polygon v3 has averaged 32% (31% median) since Jan 1st. Llama’s proposed IR curve parameters would result in a utilization rate of ~83% in the Polygon v3 market if the borrow interest rate stays at 32%. As noted by Llama, borrowing demand has increased on Polygon v3 since the beginning of 2023. The demand for borrowing is expected to continue based on yields being offered to CRV holders and historical borrowing demand.

We recommend setting the Uoptimal at 70% and maintaining Slope2 at 300% for the following reasons:

  • The market borrow rate on Polygon v3 is greater than the recommended Slope1 rate of 14%. Utilization above 80% would place the liquidity pool at high risk of preventing suppliers from withdrawing collateral and increase the risk of failed atomic liquidations. The proposed interest rate curve is not calibrated to manage liquidity risk and maintain utilization below the Uoptimal if market demand continues at 32% (or higher). Thus we recommend to calibrate the Uoptimal point lower and maintain Slope2 at 300% in order to increase the likelihood of ample liquidity within the pool.
  • Whale user 0x7a16ff8270133f063aab6c9977183d9e72835428 has a balance of 83M CRV, representing 75% of supplied assets in the v2 ETH pool. Even though borrowing is disabled, there are still risks associated with increasing the Uoptimal to 80% and decreasing the Slope2 rate. The CRV borrowing market on v2 ETH is still active with borrows at $41.8M (6th largest on ETH v2 market) and utilization rate at 41%. Changes to Uoptimal and Slope2 curves would minimize the IR risk levers by discouraging new supply to come into the pool if suppliers were to start withdrawing. Lower incentives for suppliers when the CRV ETH market reaches high utilization might create liquidity issues for this highly concentrated market. We recommend more conservative IR parameters in order to maintain ample liquidity within the pool. Similar to the ETH v2 market, Polygon v3’s market is compromised of large positions as well. The top 5 suppliers on Polygon v3 represent approximately 54% of supply in the liquidity pools.

CRV Account Supply Concentration per Market

We recommend recalibrating the Uoptimal point to 70% and maintaining the Slope2 rate at 300 and to monitor how the market performs after implementing the proposed IR parameters prior to increasing Uoptimal or adjusting the Slope2 lower.

Parameter Current Proposed Gauntlet
Uoptimal 45 80 70
Base 0 3 3
Slope1 7 14 14
Slope2 300 150 300
Polygon v3 Reserve Factor 10 20 20
3 Likes

The recent discussions indicate the need for consensus on this proposal.

The initial proposal was complex and considered 4 differents changes on an asset, we invite the community to reach consensus on this and re-run this proposal in the form of 4 different votes that will allow the governance to express their opinion with more granularity.

For now, the ACI cast a NAY vote for the snapshot proposal and would like to thank llama, Gauntlet & Alex from aaveCO for their work on this proposal.

We will monitor the evolution and are ready to support a re-run with more maturity & consensus reached.

3 Likes

To clarify, Gauntlet’s Interest Rate Curve recommendations here apply to all the CRV markets listed in Llama’s proposal. We generally recommend IR curves to be the same across markets in order to reduce the likelihood of dynamics that would incentivize outsized utilization on one market (although there are edge cases).

As such, we would not be opposed to having one Snapshot that bundles all these IR Curve changes, instead of having multiple Snapshots.

2 Likes

Hi Everyone :wave:

We have updated the proposal to reflect @Pauljlei (Gauntlet) feedback and unless additional feedback is shared, we will resubmit this proposal for Snapshot in the coming days. It will most likely be submitted Sunday, so that voting can start on Monday 6th February.

2 Likes

Hi Everyone :wave:

A Snapshot has been created.

https://snapshot.org/#/aave.eth/proposal/0x56aaf192f5cad8277b0e7c82abad030c62bb8fcfe4f2640ce5b896ab04397c20

Thank you in advance to everyone who participates in the vote.

Even if the proposed parameters are technically correct in this and other cases, we would like to insist and raise awareness in the community on how critical the increase of the Optimal utilization point of the rate strategy (Uoptimal) is.

Lack of liquidity for liquidation of volatile assets, especially those with no “special” dynamics (e.g. not wstETH) can lead to simply a scenario of apathy from liquidators and an increased probability of bad debt, as other community members and providers already commented. This should not be taken lightly for the sake of any revenue optimization, the security of the protocol is first.

In addition, we think at least 1 of the risk providers (Gauntlet, Chaos Labs) must always fully agree with any proposal from the community. If some analysis time is required (reasonable), assurance goes before speed.

4 Likes