Summary:
Lending platforms tend to disproportionately lend stable coins and collateralise with volatile cryptos like BTC or ETH.
This makes depositors on DeFi lending platforms long BTC/ETH, not market neutral.
By extension, the reputation of lending platforms is heavily long crypto.
Collateralisation levels on lending platforms are very low if you compare crypto prices now to prices at March 2020 levels.
This means lending platforms are highly dependent on their liquidation protocols and markets to run very smoothly to avoid depositors going underwater in a crypto crash.
I keep an updated copy of this post here.
Are Lending platforms market neutral?
First, some terminology I’ll use for this piece:
When I say “market neutral” I mean indifferent – from a collateralisation standpoint – to changes in the value of the crypto market.
By “crypto market” I mean the US dollar denominated value of volatile cryptos like Bitcoin (not stablecoins).
One example of an approximately market neutral lending platform would be one that primarily lends bitcoin and takes ethereum as collateral. Provided Bitcoin and Ethereum move roughly together, collateralisation levels would remain roughly constant.
Now, let’s look at what lending platforms actually hold within their smart contracts, starting with Compound:
And then Aave – in two parts (because I can’t fit the full table in one screenshot):
Ok, what do you notice about assets on these platforms?
A lot of people want to lend out WBTC and ETH, and few want to borrow – so interest rates are low on WBTC and ETH. (Read: Borrowing BTC and ETH is risky because that means you are going short while the price has been skyrocketing. BTC and ETH are also the biggest cryptos by market cap so there is a lot of BTC and ETH out there to lend). [WBTC is a ethereum token that is backed by one BTC.]
Lots of people want to borrow stable coins (USDC, Tether, DAI) relative to lenders – so the interest rates of high. Take at USDC borrowings versus Ethereum.
Interestingly, the dollar value of stables (USDC, Tether, DAI) being borrowed is much higher than the dollar value of BTC/ETH type assets being loaned out.
So, why is there a lot of stablecoin borrowing? Here’s one answer:
People who own BTC/ETC deposit that as collateral on Aave/Compound
Using that collateral, they borrow stables (e.g. USDC)
They use that USDC to buy more BTC/ETC
This gets them a leveraged long position on BTC/ETH! (Note: I don’t recommend it nor do it myself).
So are lending platforms market neutral?
To answer this, we have to ask what happens when the price of BTC/ETH goes down.
Those who have deposited on the platform are – on average – lending USDC (or other stables) that is being collateralised by BTC/ETH. If BTC/ETH goes down in value, the collateral backing the loans of USDC goes down. Therefore, depositors on lending platforms are – on average – long crypto. They are not market neutral.
How much of a drop in collateral can lending platforms tolerate?
Let’s start with the benchmark that BTC/ETH prices have risen by 15X+ since March 2020. My view is that platforms should be prepared for such a drop or worse in future, even if only temporary.
Now, let’s look at average collateralisation that lending platforms require. If we take Compound as an example, and focus on WBTC (which BTC backed ethereum version of Bitcoin). A typical collateralisation factor for WBTC is 40% (or 2.5x), meaning that for every $1 of crypto borrowed, collateral of $2.5 must be placed in escrow on the platform.
Let me restate the last two paragraphs: In a market that has risen by 15X since march, typical collateralisation factors are around 2.5X.
To be concrete: Let’s say someone borrows USDC using BTC as collateral. For every $1000 of USDC, they provide $2,500 worth of BTC. Now, let’s say the price of BTC falls by 10X leaving $250 worth of collateral backing $1,000 of USDC. Those who supplied/loaned USDC out to compound are now underwater.
Technically, the lending platform (say Aave) should be able to sell off the BTC as the price goes down before the collateral gets below $1,000. This is done by the platform (e.g. Aave) allowing third party arbitrageurs to sell the collateral off at a discount to the immediate market price. This requires a well functioning and liquid market. This also requires that the platforms smart contracts have oracles that are very fast at updating the prevailing market rates for assets in the case of a crash. In a deep market crash, both requirements may be big assumptions.
Let’s walk fully through the example above where $1,000 of USDC is being borrowed with $2,500 of BTC as collateral:
Let’s start with a rapidly falling market. The collateral is now down to a value of $1,500 and the lending platforms starts allowing 3rd parties to liquidate the collateral.
Arbitrageurs start selling off BTC at a discount.
But the market price for BTC is still falling falling, and there is increased selling of BTC to cover borrowings in platforms like Aave/Compound. This further drives down BTC prices.
The markets are going crazy, crypto prices are now different on different exchanges and the oracle (the mechanism in the lending platform that determines at what price assets can get liquidated is giving off rapidly fluctuating values).
Depositors on USDC now want to withdraw their lendings, so that also starts to put downward pressure on collateral – although this would be somewhat mitigated with time-locks on deposits.
If the collateral falls to $1,000 in value before it can be liquidated, then the depositors of USDC are underwater.
Will the arbitrageurs be able to execute liquidations (and be willing to take on the associated risk) before the collateral falls to $1,000?
That’s the key question – and – it’s very difficult to model. In deep crashes there are many second order effects driving price downward that a Brownian motion model with some price jumps would not capture. Evaluating risks requires a back to basics approach and a close look at collateralisation.
Impact on lending platforms of a market crash
Notice that I still haven’t fully answered the question of whether lending platforms are market neutral. Here’s my answer (and please do comment below if you think this is wrong – even if it’s not fully wrong I’m sure I’m missing some of the picture):
In terms of first order effects, it is the depositors on lending platforms that are not market neutral.
However, if lending platforms like Compound/Aave have depositors that are burned in a crypto market crash, I think there is a second order effect that is bad for the platforms’ reputation and tokens.
In the case of Aave, there is a safety module where Aave token holders can stake to get rewards (about 5.5% annually) in return for having 30% of their stake slashed in the case of a collateral shortfall (yes, I hold staked Aave, eek). This means that since depositors on Aave are heavily long crypto, the Aave token is also long crypto and is not market neutral.
Similar challenges are present in severe downmarkets for Maker (behind the Dai stablecoin) as Maker collateralises its stablecoin with cryptos including BTC/ETH. The stability of Dai in these environments warrants examination. I think aspects of the Celo stablecoin protocol may also face this challenge.
Lastly, lending platforms are long crypto at the macro level because if crypto does badly, less people will use crypto lending platforms. This is really a separate point but I’m including it for completeness.
Some initial recommendations for lending platforms:
Safety modules* – as Aave have in place – are a good idea. As an improvement, I would recommend setting the interest payable on the safety module at a level that achieves a target staking percentage, e.g. increase interest rates until staking percentage is 75%. This would allow a market based assessment of the platform’s risk. If the interest rate proves to be very high, that’s a sign that collateralisation levels should be increased (The current interest rate of ~5.5% seems low for the risk involved.)
Required levels of collateralisation should have some simple link to previous market lows – for example collateralisation to meet the previous market low (March 2020). Yes, platforms will hate this because it means much more collateralisation. That’s life though, crypto is making inroads on adoption but is still very volatile.
I’m on the fence about this one, but introducing fiat backed stables (like USDC) into the safety module would add security. Opponents will argue that there is unquantifiable counterparty risk with fiat backed stables and I can’t disagree with that. It kind of defeats the point of DeFi, but maybe it’s a bridge and the end justifies the means.
Lending platforms should post a clear summary of their levels of collateralisation on their website (btw, I’d appreciate if someone could point me towards any live feeds showing collateralisation of platforms). They should break out stable coin collateral from volatile crypto collateral. They should avoid fancy metrics like VAR that led the financial industry astray. They should avoid the false confidence of Brownian motion models in preparing for extreme events.
*As an aside, harvest.finance effectively did a retrospective safety module by providing partial compensation to those who lost funds in the October 2020 hack. Moving this to an a priori approach would be an improvement. Kudos to harvest for this though, and for Aave for starting a safety module (I’m staked in FARM and Aave).
In Summary
Depositors on lending platforms are heavily long crypto markets, as is the reputation of the lending platform themselves.
Platforms would be wise to monitor carefully their collateralisation and be prepared for strong crypto value crashes that seem inevitable in the peak and trough nature of crypto adoption.
Disclosure: I’m long BTC, FARM, AAVE, CRV, UNI, CELO. If you enjoyed this article, you can subscribe on Pinotio.com and also find my DeFi Mini-course there.