Currently, there is a lot of debate over whether the stability fee as a policy rate for borrowing DAI and backing it with ETH properly represents market dynamics within the MKR ecosystem. MKR works in such a way that the stability fee is voted on, not defined by markets — something like what central banks do, but in a decentralized fashion.
I believe the best way to assign an interest rate to any asset is to look at what the market is signalling. For MKR, this means we should take a look at similar products where users are borrowing assets and backing them with ETH, ideally with the interest rate being defined purely by the market.
When assessing the stability fee for the MKR system, we monitor two important parameters:
1. Market risk of collateral
2. Supply and demand of DAI: Supply comes from margin traders and market makers, whereas demand comes from DAI being utilized and locked in Dapps and from people holding it for hedging purposes (as with tether)
Keep in mind that people who open a collateralized debt position (CDP), withdraw DAI, and spend it (leave the system) to buy goods has a net zero effect as the created supply has found demand. As noted above, if an increase in DAI’s supply isn’t offset by passive holders or utilization of DAI within the ecosystem, market makers will reduce the supply by closing their CDPs in order to make profit and bring DAI back to its peg. The same goes for margin traders, who withdraw withdraw DAI to buy ETH which follows the narrative of going ‘long’ — someone needs to buy that DAI and hold it, otherwise there will be a supply glut, causing DAI to fall below its $1 peg and cause inventories of market makers to increase when they defend that peg.
The market risk of collateral can be measured qualitatively by evaluating certain coins and quantitatively by deriving the risk premium of the collateral adjusted for liquidity and other parameters such as correlation risk and exposure risk, which are well defined in the MKR governance framework. The risk premium of a particular asset would normally be extracted from past price volatility, though I believe this method is inappropriate in crypto, where volatility patterns change constantly. Implied volatility would be the preferred indicator here, but we don’t really have well developed derivative markets for ETH, let alone for other ERC20 assets.
The other alternative option for measuring the risk premium is looking at other money markets. The most obvious venue for such data is the Compound Dapp, which works similarly to CDPs (same 150% overcollateralization limit) but has borrowing interest rates defined by market participants, though not purely. Compound uses this algorithm to define the interest rate formula, calculating rates by using supply and demand figures from the market. You could say that markets define the interest rate at Compound, but only to some extent. Note that Compound doesn’t carry a lot of liquidity yet and hence shouldn’t be treated as an ideal benchmark for MKR. The total ETH locked in Compound amounts to about 30k, only 1.5% of the amount of ether deposited at MKR.
Let’s assume Compound’s interest rates are purely market defined and try to use them to get an indication of DAI lending market interest rates. Note that DAI at Compound that is lent out is collateralized mostly by REP, DAI and ETH, and we expect the interest rate for borrowing DAI to be a bit higher. Below is the data extracted from Compound by Vishesh Choudhry. You can see that the borrowing rate for DAI in the past month was between 8% and 10%, almost 10x higher than it would be if you would directly issued DAI through CDP. This means that market participants are prepared to pay more for borrowing DAI and having to back it with the same 150% overcollateralization amount. In other words, they are much more bullish about ETH in collateral, assigning it a higher risk premium and expected return.
When trying to benchmark lending rates for ETH collateralized borrowing products, we can also take a look at other large ETH money markets to find risk premiums associated with ETH. The only obvious candidates are those money markets offered by centralized exchanges such as Bitfinex and Poloniex, where people lend to margin traders. There is currently about 300k ETH (about 15% of the amount of ETH locked in MKR) being loaned on Bitfinex and Poloniex, with interest rates of about 1.5% and 8% respectively. In the last couple of months, the range of the ETH interest rate varied between 1% and 10% for Bitfinex and 3% and 50% for Poloniex. The market thus tells us that the risk premium and expected return for ETH is way higher than MKR estimates. Also, if you wanted to create leveraged long ETH exposure (similar to what CDP owners at MKR are doing), borrowing USDT would currently cost you 10% per year.
However, comparing interest rates between MKR and money markets at centralized exchanges isn’t exactly the best benchmarking method. On exchanges such as Bitfinex, there is the benefit of margin collateralization (only 15% margin required); people would normally be willing to pay a higher interest rate for this benefit. However, they are still exposed to a central counterparty (a discount) and would need to hold tether (another discount) to have the same benefit compared to opening a CDP and withdrawing DAI. On the other hand, MKR has the benefit of high decentralization with no counterparty risk but is exposed to some black swan events associated with Ethereum as a layer and with smart contract bugs.
When trying to assess the risk premium of borrowed assets collateralized by ETH, we are limited in the number of venues we can look at. Centralized exchanges are currently the most liquid venues where such analysis can be done but for obvious reasons cannot present the most appropriate benchmark when assessing risk premiums that would help determine a stability fee for MKR. DeFi Dapps such as Compound are more appropriate, but we need to wait for them to grow and have high enough liquidity pools and a larger number of market participants to get a proper representation of what the market is telling us. Also note that the stability fee is a function of a wider range of factors, and collateral risk and DAI adoption are just two of them.