Web3 Lending Protocol Classification Detailed Explanation Interest Rates

Comprehensive Explanation of Web3 Lending Protocol Classification and Interest Rates

Author: David Ma, translated by Shanooba and LianGuai

In a previous article, I started a project to classify lending protocols in Web3.

Quick recap, a lending protocol is a set of rules that govern how a borrower temporarily takes possession of an asset from a lender and commits to repay those assets. Such protocols will define how interest is collected from the borrower to the lender, and how collateral is built to protect the lender. Part 1 of this series explored topics within the interest term: zero term and positive term, loan rolling and perpetualization, selectivity, and classified a few protocols within this framework.

In this article, I continue to build the classification to discuss how various protocols determine interest rates.

• Avara: A Web3 Marvel with a Dash of Humor
• OP+Plasma’ a new combination, speed-reading Lattice chain dedicated L2 Redstone.
• Web3 Fueling Innovation, not Just Another ‘Money Grab’ for Brands – Insights from a Tech CEO

Interest Rates

An interest rate is the rate at which a borrower accumulates additional payments to the lender. For comparison purposes, this rate is usually quoted in an annualized format. It is expressed as an Annual Percentage Rate (APR) or Annual Percentage Yield (APY). The difference between the two is that APR assumes no compounding, while APY assumes compounding. For example, an annual interest rate of 10% compounded semi-annually would result in an APY of 10.25%.

The two are related by the following formula: APY = (1 + APR / k)^k — 1 where k is the number of times compounding occurs in a year.

In Web3, most loans are continuously compounded (k = large) as most loans are zero-term loans. Therefore, they quote APY to tell users how much money they would make in a year, assuming the interest rate remains the same. For fixed-term loans, it is more common to quote an annual interest rate.

By the way, if the 2020-style rudimentary agricultural yield chasers return, be wary of predatory quotes in APY because none of those opportunities will last long enough for compounding to matter. Mental math is much easier with APR. Doubling TVL for fixed reward pools means halving the reward rate. Simple.

Now that the definitions are settled, we can discuss pricing.

Pricing

Pricing is the mechanism for calculating how much interest lenders and borrowers pay to each other. While not comprehensive, this article will introduce some of the mechanisms:

• Order Book Pricing: Most flexible and market-driven, but requires trade-offs in user experience.

• Utilization-based Pricing: This model has found product-market fit in defi but is not 100% efficient and performs poorly in extreme cases.

• Auction: Good pricing, efficient lending, but requires advanced planning by users, secondary market fragmentation, low friction, etc.

• Ajna’s Usage: An adaptation of the classic utilization-based approach that works in oracle-less protocols.

• Tazz’s Perpetual Financing Loan: A new p2pool lending primitive that allows the market to price rates, making collateral fully modular.

• Manual Pricing: or governance-led pricing.

Order Book Pricing

The most common way to price assets is through a market-driven mechanism: the order book. Borrowers and lenders place limit orders, specifying the amount and interest rate they are willing to lend or borrow. When orders match, the trade is booked.

However, the order book also has its drawbacks:

• Inexperienced users don’t know how to price their orders. These users simply want to transact without paying huge costs.

• Placing a limit order is like writing a free option. The worse the market liquidity and the slower the block time, the more valuable the option becomes. In other words, the more the theoretical true price fluctuates without the order being executed, the greater the option value embedded in these limit orders.

• Running an order book requires active management. You need to cancel outdated limit orders. You need to play the bidding game with other participants.

• It requires a lot of trading.

This is why the order book remains unpopular on the blockchain. Instead, automated market makers, auctions, and inquiries are more suitable for blockchain products.

In the lending space, the order book faces greater challenges:

• Order book trading creates peer-to-peer loan matching. Default risk is irreplaceable.

• Continuously issuing term loans generates positions that cannot be fully swapped with each other. Instead, protocols like Pendle and Notional choose to issue loans with fixed terms on specific dates. The trade-off is that the term of available loans is always odd, such as 37 days, 159 days, and so on…

• Short-term loans create more transactions. You can automate the rolling, but how do you price the next loan?

All of this leads to market fragmentation, or at least a complex trading experience. That being said, NFT lending platforms like Blur and Arcade.xyz still rely on a user experience similar to the order book. Each of them has come up with features to mitigate the poor user experience.

• Blur incorporates a form of perpetual lending to eliminate the term dimension.

• Both Blur and Arcade have “portfolio collateralization,” where lenders view the entire NFT collection as fungible, and any NFT within the collection can be used as collateral.

AMM is a subclass of the order book. If a protocol can gather enough people into a fungible lending market, AMM is a good choice. Interest rates are easier to regress to the mean compared to tokens, so using AMM to LP the rates is safer than LPing tokens. That’s how Pendle, Notional, and Tazz work.

Let’s review, while the order book can handle simple assets well, the lending order book has too many dimensions to consider and requires clever methods to reduce user experience complexity.

Utilization-based pricing

The utilization rate of a lending asset pool is the ratio of the borrowed asset amount to the total lendable asset amount.

Utilization-based pricing defines interest rates as an increasing function of utilization rate.

The first and largest on-chain lending platform, Aave, adopted this approach. It remains the most popular interest rate pricing method for liquidity and alternative assets.

Note that Aave didn’t start with this design. EthLend (before the Aave rebrand) outlined peer-to-peer fixed-term order books in their 2018 whitepaper. The utilization pattern was introduced in their 2020 whitepaper. This kind of radical rethinking of financial primitives (Uniswap AMM is another example) is part of the fun of DeFi.

Since the whitepaper does not explain this, my speculation about the motivation behind this approach is simple. Consider the Econ 101 model–when interest rates are high, fewer people are willing to borrow money while more are willing to lend. The “optimal” interest rate refers to the perfect match between the number of borrowers and lenders, with a utilization rate of 100%.

When interest rates are too low, there are more borrowers than lenders. The utilization rate will reach 100%, but it doesn’t tell us where the model stands. Additionally, lenders cannot exit the loan.

When interest rates are too high, a large supply of loans will be idle. The APY spread (1-utilization rate) increases with rising interest rates. This is before the platform charges fees. The chart on the right is simplified. The actual rate received by lenders is lower than r.

Due to the unobservable lending curve changing with market conditions, the challenge is to maintain interest rates close to the optimal rate while retaining a certain buffer for lenders to exit.

The term “interest rate model” used by Aave above is a bit misleading. Math enthusiasts like to call it a PID controller, but it’s only partially automated. First, Aave chooses a target utilization rate (e.g., 90%) and a twist curve. If the utilization rate frequently remains above 90%, Aave’s governance (thus partially automated) will steepen the interest rate curve, attempting to reduce utilization. If the utilization rate is too low, the opposite occurs.

Sometimes, the market encounters special situations where the PID controller reacts too slowly. For example, during the Ethereum merge in September 2022, the pre-merge Ether will fork into Proof-of-Stake Ether (the Ether we hold now) and Proof-of-Work Ether (sponsored by Justin Tron, kinda like). The market values PoW Ether at about 2% of PoS Ether. Market participants saw this and wanted to hold as much pre-merge ETH as possible in their wallets. One way is to use stablecoins as collateral and borrow Ether. As long as the cumulative interest within the borrowing period is less than 2%, it’s a reasonable trade. Earning 2% in less than a week means an annual interest rate of over 100%. Aave’s interest rate ceiling is 100%. Needless to say, in the days leading up to the merge, Aave, Compound, Euler, Inverse, and every PID controller lending protocol for Ether reached their limits. If I remember correctly, Inverse didn’t set an interest rate ceiling, and the annual interest rate reached 1000%.

When it comes to usage-based pricing, it naturally fits well with a peer-to-peer structure, making it a zero-duration loan. Therefore, we often see these attributes go hand in hand.

In summary, the benefits lie in the user experience under normal market conditions. Borrow, lend, repeat. But when the utilization reaches 100% (such as during a merge period), the lender is left in a state of non-recourse. Other downsides include 10% loan buffer assets leading to inefficient capital efficiency, and the inability to provide fixed-term loans.

Auctions

Auctions are a tried and tested method for issuing new bonds (primary market issuance) over time. US Treasury bonds are the most liquid government securities globally, and they use auctions to price new debt. At a high level, borrowers and lenders submit sealed bids to regularly held auctions. The market-clearing interest rate is found, and new debt is issued to participants at that interest rate.

Term Finance is a relatively new protocol that is inspired by this mechanism. The details of their auction implementation are worth a read. Their documentation is clear and concise.

Auctions can effectively match lenders and borrowers. Unlike order books that require capital to be locked up for orders waiting to be executed or utilization-based loan pools that require lenders to have withdrawal buffers, there is no idle capital. The only non-productive period is when the assets are locked during the auction period.

Auctions also generate high-quality pricing because market participants converge to a Schelling point to aggregate their private information.

The disadvantages are that auctions require some pre-planning and are less user-friendly. While this is a good option for US Treasury bonds, participation in the fixed-term loan market is not yet high enough in the cryptocurrency space. Another challenge is market fragmentation. Cryptocurrencies have many similar but non-interchangeable assets. This would be a more difficult product to launch, but I hope that one day Term Finance can issue Ethereum Treasury Bonds with full support from the Ethereum printer.

Arbitrary Utilization

Ajna is one of the few lending protocols that does not rely on oracles. A full discussion of how Ajna achieves this goal is beyond the scope of this article. Instead, its high level of setting interest rates is worth discussing.

The lender first selects the collateral (e.g., ETH) valuation for which they are willing to lend the quoted token (e.g., USDC). The borrower accumulates matches from the highest valuation downwards. The borrower with the highest loan risk (highest loan-to-collateral ratio) sets the “Highest Threshold Price” (HTP), and borrowers with lower valuations do not earn interest.

The lender does not want to set the valuation too high because they will be the first to suffer losses in the event of default. The borrower does not want to set the valuation too low because they will not receive any interest.

The interest rate is also determined by the utilization function, but the calculation here only considers setting the collateral valuation higher than the HTP of the lender. The interest rate starts at 10% and is multiplied or divided by 1.1 every 12 hours, depending on the comparison between the utilization rate and the “target utilization rate”.

The main advantage is that, despite using a pair-to-pool design, this mechanism does not require any oracle to operate. On the other hand, the lender needs to constantly monitor their valuation. Like other utilization-based interest rate pricing mechanisms, the loan’s annual interest rate is also influenced by the unused borrowings.

Although borrowers and lenders can leave at any time (zero duration), considering that Ajna’s initial fee is 1 week and the interest rate half-life is 3 days, Ajna’s design period is at least one week. Since Ajna has not been in the market for long enough, it is still too early to fully understand the advantages and disadvantages of this mechanism.

True permissionless and immutable protocols are rare because they are difficult to execute correctly. But when they do happen, they become the cornerstone of composability. I sincerely hope that Ajna will become the Uniswap of the lending world.

Tazz’s Perpetual Loan Financing

Tazz is an upcoming loan protocol that introduces a new interest rate pricing primitive. Again, the full description of this mechanism is beyond the scope of this article. Stay tuned for future in-depth articles.

The debt begins with interest-free perpetual loans, similar to Aave. The debt tokens (ATokens in Aave’s terminology) can be traded on any dex. Apart from protocol insolvency, Aave’s Atoken trades almost always at a 1:1 ratio, while Tazz’s debt tokens (ZTokens) are priced by the market. The price of ZToken determines the interest rate at which nominal debts accrue (i.e. funding). If the nominal debt continues to increase, the collateral parameters will trigger liquidation.

The continuous funding payment is proportional to k * (1 – ZToken price relative to the token’s TWAP). The lower the constant k, the longer the debt duration and the more susceptible it is to interest rate risk.

Note that in this mechanism, the collateral is completely modularized from the rest of the protocol. You can set up collateral-free, NFT collateral, LP tokens, illiquid tokens, locked tokens, oracle-based collateral pricing, or pooled pricing with one-time pricing. It doesn’t matter because the market can determine the interest rate pricing required to bear the risk.

It is worth noting that this allows for:

• P2P lending
• 100% loan utilization, hence lower spreads
• Merged liquidity divided only by collateral terms and duration
• Any collateral type

One potential drawback is that it requires monitoring pool prices (but less than Ajna). If unrealistic prices persist for too long, it can lead to unrealistic interest rates. The liquidity market for ZToken will allow borrowers and lenders to not have to closely monitor it.

Manual/governance pricing

This is worth mentioning given the GHO depeg. There are stablecoins with debt positions (CDPs). Maker’s DAI is the largest, followed closely by Liquity’s LUSD, Lybra’s eUSD, Prisma’s mkUSD, and others.

Although CDPs may not look like loans, they are actually quite similar. Borrowers collateralize with ETH (Maker v1), LST (Prisma, Lybra), or other assets. Borrowers create a CDP, and the protocol’s oracle calculates the value in USD at a 1:1 price. The CDP can be sold on the open market, allowing the borrower to “borrow” another asset and the lender receives the CDP. The loan is permanent, with an unknown term, and may not be fixed at $1. Borrowers pay interest to the protocol, and lenders can receive another interest rate from the protocol (e.g., Dai savings rate). Sometimes, there is a backstop fund called a “stability module” to prevent self-reinforcing CDPs from becoming uncollateralized.

The downside of manual pricing is that it is influenced by governance processes, lengthy debates, voter apathy, and quorum requirements. As a result, it is slow to react. The benefit is that human processes are harder to manipulate than code that may have edge cases.

Aave’s GHO is a CDP with manual interest rates. Currently, GHO’s borrowing rate is 3% (lower than treasuries and Dai’s 5%), and their lending (savings) rate on Aave is 0% (unless you know where to find gamified incentives). As a result, there are too many borrowers and too few lenders, leading to a drop in GHO price.

Debates on the Aave governance forum have been ongoing for months. The essence of the debate boils down to whether to fix the exchange rate or maintain stable interest rates (thus variable rates but positive term). GHO cannot have both until it gains more market dominance.

Conclusion

In this article, we have introduced various methods of interest rate pricing in lending protocols. Of course, there are many other approaches, but the goal of this series is to establish a taxonomy. So far, we have considered interest term and interest pricing as the two main perspectives for analyzing and categorizing protocols. In the next article, I will discuss collateral.

We will continue to update Blocking; if you have any questions or suggestions, please contact us!