Crypto lenders forego profits

TL;DR: DeFi often faces misunderstandings about capital utilization and borrowing rates due to its focus on supply rather than demand, benefiting traders but not long-term stability. While high utilization in lending pools leads to skyrocketing interest rates in DeFi, such swings occur only in niche markets of traditional finance because stable rates are attractive for risk-averse individuals. In this post, we will delve into how we can parameterize lending pools using statistical methods and look at the preferences of both borrowers and lenders. Should lending protocols pay more attention to the needs of borrowers?

How shall we parameterize lending pools?

In DeFi, we’re used to very volatile interest rates — if utilizations of lending pools reach certain levels, these interest rates skyrocket. In contrast, we don’t see that phenomenon in TradFi, and sometimes we wish the rates were more stable in DeFi. But so far, we haven’t acted on it. I argue that is because we’ve cared too much about the market’s supply side (the liquidity providers) and too little about the demand side (the borrowers): We’ve promised lenders to rent out their capital while still always being able to use it; however, the borrower paid the price for this flexibility. Nevertheless, this was good for experienced traders, short-term liquidity, and POCs (proof of concept).

Back to the roots: The importance of capital utilization for profitability.

Let’s look at a fundamental form of lending first: Renting. Imagine you are Warren Buffet. That is already a great feeling, isn’t it? Now, imagine you buy a pinball machine, rent it out in a local barber shop, and split the revenue with the barber. Did you know that this is exactly what Warren did? I’ll leave the source below so you can fact-check.

With the machine, you have transformed your cash into working capital sitting in a barbershop. But it doesn’t earn money if it is not utilized, and nobody pays to use it. So, if we want to earn interest rates on our assets, we must put them to work. A recent analysis (NYC Uber & Lyft Utilization Rates) of the New York rideshare market shows how Uber and Lyft try to maximize the utilization rates of drivers and cars. This logic is omnipresent: The US capacity utilization (Capacity Utilization: Total Index (TCU) | St. Louis Fed) only once dropped from its long-term trend to 65%. This was during the global financial crisis. What do we learn from this? Low utilization is undesirable, at least if we look at physical examples of capital.

Renting things out means you can not use them — unless you are in DeFi.

With the above in mind, should we aim for 100% capital utilization in financial markets? Probably yes and most probably, we’ll have to incentivize utilization with more or less predictable rates until full utilization. Nevertheless, lending protocols are hesitant to allow 100% capital utilization, especially in lending. Aave is advertising that its capital is only fully utilized 1% of the time.

But you might ask: “This is great risk management by Aave, though, right?” Higher utilization ratios indicate that a significant portion of the asset pool is in use, which could increase the risk that lenders cannot withdraw their funds when desired. To compensate for this risk, Aave charges borrowers more, ensuring that lenders are adequately compensated for the increased risk they bear when liquidity is tight.

This illiquidity is not credit risk!

Most DeFi lending implicitly mixes maturity transformation with lending (liquidity transformation). In DeFi lending pools, all lending is always due, but only borrowers doing arbitrage or speculation have a planning horizon of zero. Borrowers with non-trading motives typically need the capital for longer and cannot ensure they can always return it. DeFi Lending protocols have to do liquidity risk management because they promise the lender that they can almost always get their money back — this guarantee is what the borrower pays for. The risk here is not credit/default risk; it’s the risk for lenders not to be able to exit the pool at any time.

How variable prices and rates work outside of Web3

But this risk has nothing to do with the “classical lending risk” — i.e., the default of the lender because their undertaking (business, construction project, family) fails. If you lend out something, a pinball machine or money — you might only be able to get it back after the agreed-upon end of the renting period. The DeFi promise that you can always get back what you just lent out is expensive — for instance, because the guy at the pinball machine doesn’t want to give up the controls while breaking the high score… well, you get it. Optimizing the flexibility of the liquidity providers is mostly achieved by making interest rates very reactive to utilization, such that borrowing gets way more expensive if the utilization goes up. This leads to a highly variable interest rate for borrowers, which is undesirable for long-term borrowers.

You might say — but what about Uber? They change prices all the time!
Yes, they do, but they never change the price while you are on the ride, even when prices triple.

If you have a variable interest rate on your mortgage, the rate may also change overnight. It is common in international commerce to make contracts conditional on FX rates or overnight interest rates.

But here is the catch: Contracts use rates like the Fed Funds Rate, ESTER/€STR (successor of the Euribor and LIBOR), or the dollar as a base currency. These rates do typically not move a lot, and your risk can be expressed in basis points, rather than percentage points. And even still, ample liquid tools are available to help hedge against the tiny fluctuations of these rates.

Still, these tiny fluctuations are costly for the industry and macroeconomy. The recent Fed interest rate hike (from around zero to 5.3%) has caused significant turmoil for the US real estate industry, followed by the most pronounced decline in commercial real estate prices compared to previous periods (US Commercial Real Estate Remains a Risk Despite Investor Hopes for Soft Landing).

What is a good range for DeFi interest rates and the parameters of lending protocols?

The answer is easy but might not suit you — you have to work it out according to your users’ preferences. Ask your target users and validate what they tell you by observing their usage of your product. Measure demand and supply sensitivities regarding the interest rate change. Here is one recent example from Harvard Business School on Estimating Models of Supply and Demand.

Let’s look at two simple cases abstracting from liquidations and defaults:

A) You lend out 10k USDC on a protocol; the interest rate is 11% but shoots up to 60% because the pool hits full utilization but then returns to 10% because of a large capital injection. On average, you earn more than 10% but can withdraw anytime.

B) You lend out 10k USDC on a protocol; the interest rate is 11% and increases to 18% as the pool hits full utilization but stays there until incoming interest payments enable you to withdraw your funds partially. On average, you earn 18%, but your funds have been on loan for a while and were inaccessible.

The preference between these two scenarios is mainly driven by two questions:

1) Does the lender return if their capital was locked but compensated → What’s the right compensation for the lender? (Capital Supply Curve Estimation)

2) What is the typical endeavor of the borrower, and what is an adequate rate for borrowers to not sit on the capital but achieve their goals and return the capital on time? (Demand Curve estimation)

In any case, we can provide some guardrails for long-term interest rates — if we abstract away FX rates and inflation/dilution.

First of all: There is an effective lower bound that is more or less equal to the storage costs of money. If the interest rate falls below this threshold, people will buy other assets that are easier to store and provide similar liquidity.

Global real rates of return on housing, equities, bonds, and bills have historically ranged between 1 and 8%. If your rate of return is way outside this range, your product may not be sustainable.

What are “too high” interest rates?

VCs typically aim to 10x their investment over ten years. For this, you would need a 26% return (10 ≈ 1.26¹⁰), but the air gets very thin above 20% annual rates. At such annual rates, the rate itself significantly impacts the borrower’s default probabilities, and you are most probably in the realm of very risky lending. Also, very likely there will be strong liquidity constraints in place. However, there are counterexamples: Amazon stock has generated an annualized total return of more than 32% for its entire history as a publicly traded company. So, Jeff could have afforded to take on a loan with a 30% interest rate instead of selling Amazon equity to investors.

Then there is day trading: If you are implementing a Volatility Range Breakout Strategy and your weekly goal is 1,5%, you will make 110% a year. So, lending to the winning percentile of day traders at 50% interest per year could be a win-win for both the lender and the borrower.

Things get even more extreme if we look at overnight interest rates and repo markets. Sweden has seen overnight interest rates of 500%, and the central bank even prevented a banking crisis by setting such high rates. As we can see, good reasons exist to penalize illiquidity, and even 500% interest rates might (sustainably) occur between specialized parties.

Conclusions

Getting back into DeFi, we can say that protocols such as Aave have excellent applications for short-term borrowing and lending operations because they implement adaptively fluctuating rates. However, such variable rates are not attractive for long-term borrowers who do not constantly monitor their positions and are hence disconnected from the “real interest rate” that economists and everyday users are interested in. That is also the reason why Aave lenders might lose potential profits in the long run. Their capital is not utilized efficiently. Let’s exaggerate a bit for illustrative purposes: Either a large fraction of the capital is idle, not earning anything, or the rates are so high that the lenders are incentivized to return it.

The current interest rates model in DeFi seems at odds with the real world. They could converge in the long term in two ways: We’ll see Uber adopt the Aave model, and you might be kicked off your ride because prices have just risen. Or we might soon see more stable rates on DeFi as legacy capital enters the arena. Which one do you think is more likely?

If you want to discuss how you can estimate your supply and demand curves, contact me at jan@oaksecurity.io. Let’s tackle your economic needs together!

PS: As promised, here is the Warren Buffett link: Warren Buffett bought a pinball machine for $25 in 1946 and started ‘the best business I was ever in’