December 6, 2018 . 12 min read
Crypto Chat #64
The rapid emergence of the Ethereum-enabled Decentralized Finance movement has brought with it some curious and heroic tales.
Maker CDP 3228 was the first such tale to attract attention, with industry analysts watching in real time as 50,000 Ether (ETH), valued at close to $7m, was liquidated in orderly fashion after a major move down in the markets on the 19th November.
It has taken just over two weeks for another fascinating saga to arise, this time intersecting three decentralized applications — Compound, Maker, and Paradex.
Compound is an Ethereum-based permissionless money market. The sleek user interface allows long-term crypto asset investors to seamlessly put their otherwise idling assets to use, lending to speculators on a variable yield basis.
Compound forgoes the need for borrowers and lenders to negotiate terms by leveraging an Interest Rate Model, which calculates interest rates on an aggregate supply and demand basis: when demand is low, interest rates are low, and, perhaps unsurprisingly, when demand is high, interest rates are high.
When Compound launched in late September, Annual Percentage Rate (APR) across all markets — BAT, DAI, ETH, REP, and ZRX – and on both sides of Supply and Borrow was low, as, understandably, the platform had yet to attract significant usership.
In recent weeks, however, usership has surged, with over $11m of assets available for borrowing at the time of writing. As a result, both Supply APR and Borrow APR have adjusted: users can now supply ETH at 0.42% APR, ZRX at 0.03%, BAT at 0.01%, and borrow at 7.73%, 5.30%, and 5.10% respectively.
The outlier is the DAI market.
DAI is the over-collateralized stable coin product of the MakerDAO network. Users can obtain DAI by either opening a Collateralized Debt Product (CDP) via the Maker Portal or, more simply, by buying DAI on a variety of exchanges. One might choose to open a CDP in order to tap into the liquidity of one’s ETH holdings without having to give up long exposure, and/or, if the DAI is then used to purchase more ETH, to essentially leverage long using existing ETH as margin collateral. For more on intricate dynamics that keep DAI stable, I recommend reviewing this detailed explainer from Scalar Capital’s Cyrus Younessi.
As the chart below indicates, CDP creation has continued to grow over recent weeks, with speculators presumably seeking to take advantage of the relatively low ETH prices to build leveraged positions. CDP owners pay 2.5% APR and, as CDP 3228 owner knows well, must carefully maintain a minimum 150% collateral ratio at all times.
So here’s where things get interesting.
If anyone can open a CDP and draw DAI at 2.5% APR, we would expect borrowing rates to never extend beyond 2.5% APR on the Compound market. In practice, perhaps we should add some small percentage premium to take into account the ease of borrowing via Compound over opening a CDP: the exact percentage premium this superior user experience should command is beyond me, but if I had to guess I would estimate a cap of maximum 1%.
Likewise, we should expect Supply APR for the Compound DAI markets to never extend beyond that same 2.5-3.5% range: if APR is higher than that, and the market remains efficient, then arbitrageurs can easily profit by opening a CDP at 2.5% APR and lending on Compound. Over time, any subsequent addition to the Supply pool would lead to APR returning to the equilibrium rate.
And yet, at the time of writing, Supply APR for DAI on Compound sits at 8.87% (down from ~20% earlier this week) and Borrow APR sits at 15.62%.
Fortunately — or unfortunately depending on the extent to which you think that this activity should be public — the open source nature of these protocols allows enterprising analysts, like EthHub founder Eric Conner, to trace user activity and thus determine why rates might be as high as they currently are.
Here’s what has happened.
1. A user (who I will subsequently refer to as U) decided to borrow DAI via Compound.
2. U deposited 724,000 REP, which equates to 7% of the circulating REP supply, as collateral, allowing her to borrow close to 1.6m DAI.
3. U then sent this borrowed DAI to non-custodial exchange, Paradex, where she exchanged this DAI for ETH.
4. Finally, in somewhat of a twist, U used this ETH to add to her own CDP, which has 88,000 ETH as collateral, presumably to protect it from liquidation in the scenario where the ETH/USD price continues to decline.
Sounds simple...ish! And what a marvelous, high-stakes illustration of the way in which anyone can access these various financial products in a permissionless and intersectional capacity.
And yet, after a moment’s pause, this fantastic detective work only serves to raise more questions. I have crowdsourced some answers to each, ranging from ‘probable’ to ‘conspiratorial’, with ‘plausible’ in between.
1. Why did U bother converting her borrowed DAI to ETH instead of paying back some of the DAI drawn from their CDP? After all, paying back DAI lowers the liquidation rate more effectively than adding ETH as further collateral.
Plausible: The CDP in question is already heavily overcollateralized — the Maker system requires you to keep a 150% collateralization ratio and the CDP currently stands at 244%.
As such, we might infer that risk reduction was only a second-order priority for U: she could primarily be looking to build long ETH exposure. Any significant appreciation in the ETH/USD price in the near-medium term would outweigh losses from 15% Borrow APR and the likely slippage from acquiring ETH through Paradex’s thin order books.
2. Why not open a CDP at 2.5% APR instead?
Probable: Maker is yet to transition to a multi-collateral system, so borrowing via Compound is the only way for U to access liquidity from her REP holdings.
3. Why not directly borrow ETH on Compound, which has a lower APR than DAI?
Plausible: U is hoping that DAI APR eventually returns closer to the 2.5%ish equilibrium, as discussed earlier in this piece. By contrast, there is no 'natural' equilibrium Borrow APR for ETH.
4. Why not sell their REP directly for DAI or ETH instead of going through this convoluted borrowing process?
Plausible: Selling REP would trigger a taxable event and likely crash the REP price at that volume (7% of circulating supply). By buying ETH, U is presumably increasing their ETH cost basis, which should lead to lower capital gains on liquidation.
Conspiratorial 1: If U thinks that liquidation of her REP holdings will see more than 33% slippage, a more profitable, although still loss-inducing, strategy is to borrow up to 66% of current REP price and let the Compound loan default.
As an aside, this strategy raises another question as to who the ‘buyer of last resort’ would be in a decentralized system.
Like MakerDAO, Compound uses a ‘liquidation discount’ in order to incentivize investors to step in and reduce the borrowers exposure.
The Compound liquidation percentage currently sits at 5%.
However, my sense is that this discount may not be high enough for certain assets. In this scenario a 5% discount may not be attractive enough to take on a large supply of a relatively illiquid and risky asset. Moreover, as a major REP holder, U’s antics may be possibly interpreted as a reputable warning from an industry insider that REP still has some way to fall.
Conspiratorial 2: This operation creates more incentive for arbitrageurs to ramp up DAI issuance, which should lead to appreciation in the MKR token as CDP fees have to be paid in MKR. Thus, if U also has a large MKR position, the various costs involved through borrowing and purchasing may be offset by MKR appreciation.
Many thanks to Eric Conner, FollowTheChain, Hex Capital, and all others who contributed to this fascinating discussion. I look forward to following further updates from The Tale of Three dApps and monitoring the continued development and maturation of these various decentralized financial products.
And lastly, I encourage everyone to take 10 minutes to play around with these applications, which provide insight into the future of open, permissionless, compositional finance.
Until next time,