Decentralizing Credit 104
Pushing the boundaries of ETH collateral
Last time, we discussed the scaling question.
It’s now time to get into the meat of things regarding how GUILD holders can set lending terms to outcompete legacy markets like Aave, Compound, and MakerDAO, all of which use trusted oracles, maintain large amounts of idle capital, and lack scalable governance mechanisms. We’ll focus on ETH, the preeminent onchain collateral.
More and better lending options for ETH holders
On any of Aave, Compound, and MakerDAO, you can borrow about $0.80+-$0.04 against $1 of ETH, which is comparable to the newly launched Coinbase Perpetuals offer of 5x leverage against ETH and BTC, but much less than the 20x+ leverage offered now and historically by Binance and other exchanges. For reference, TradFi may allow the borrowing of as much as $0.98 against $1 of quality short duration bonds, and options markets offer another venue for highly leveraged exposure to (or cost effective hedging of) the price of a wide variety of underlying asset. Besides the relatively limited leverage, there is a massive spread (2%+) between the rates paid by lenders and earned by borrowers. We can do better.
A non-callable, variable rate onchain debt position is like a one-block long option that is perpetually rolled over. The premium is the interest charged from the current block. The option holder (the borrower) can exercise at any time, until the value of their position reaches zero due to either options expiring worthless (liquidation due to price movement), or interest payments depleting their balance. By depositing additional capital to or withdrawing idle capital from the market, lenders can adjust the current premium on a block by block basis. While it is not obvious to lenders that their exposure is similar to options writers, this is only because these markets focus on conservative options that are far enough out of the money that lenders are unlikely to take a loss.
A non-callable, fixed rate onchain debt position is more like a traditional option, where the interest rate is the premium, and lenders are more cognizant of the risks they take on entering these markets.
To review, in a fixed rate loan, the lender sells an option that says:
“swap X units of borrowed asset to reclaim Y units of collateral at time Z”
In a regular CDP, the lender sells a perpetual rolling option that says:
“swap X units of borrowed asset to reclaim Y units of collateral next block”
A callable loan is the same, but the lender has the right to repurchase the option by paying the call fee. This right of repurchase constrains the maximum loss the lender can take, and also the maximum profit the borrower can make, on any given option/loan. This is essential for building deep liquidity in longer dated lending markets onchain, as it allows for a rational pricing of the loans.
This leads us to the important realization that lenders are like options writers, and so the basis of deciding on lending terms is looking at historic volatility data as well as the future volatility implied by the market. Based on these, GUILD holders must take into account the duration of the loan (call period + time for the liquidation auction). If a loan has a collateralization and interest rate that are too low relative to its call fee, call period, and expected volatility, it will have a negative value at issuance even if it is ostensibly overcollateralized (that scenario is also what’s going on when DAI or LUSD trade below peg).
Some readers may be familiar with the tremendous body of work and ongoing industry around this type of analysis that happens on Wall Street; others like me may feel their heads swim when Greek letters appear on the page. We as a core team will work to make open source analysis tools and educational content available to all GUILD holders, but also hope that the system will attract users who have lending and risk analysis expertise. Please join the conversation if interested.
Let’s work through an example in an alternate reality, all numbers made up. In Fantasyland, over the last two years the most volatile hour saw the ETH price move 5%, and the most volatile month saw it move 50%. Therefore, Fantasyland GUILD holders judge that for loans with a one hour call period, 10% overcollateralization is ample, while loans with a one month call period will need at least 60% overcollateralization. For the former, they offer an interest rate slightly lower than that on Aave or Compound, but higher than borrowers there are earning. For the latter, they charge an origination fee on top of the interest rate, based on the prices currently available on Deribit.
In either case, there is no absolute mathematically correct interest rate. The interest rate must be determined by both external market conditions, and the nature of the specific lending term in question. A key tenet of the Ethereum Credit Guild is that no fully automated or immutable oracle system can properly observe the changing conditions of the global market, and so this must be left in the hands of the GUILD holders themselves. Hopefully, if you’ve read this far you are thinking like an options writer and not treating overcollateralized lending like a risk free operation.
Next time, we’ll focus on the set of genesis lending terms and the special considerations around bootstrapping liquidity. Tldr:
The ECG can and should outcompete legacy lending markets by offering more aggressive leverage against ETH and staked ETH, comparable to LUSD’s non-recovery mode limit of 110%.