A full retrospective of the MakerDAO migration from single-collateral to multi-collateral DAI.
This article was co-authored by Block Analitica, which develops on-chain analytics for the Ethereum network, more specifically MakerDAO and DeFi. Block Analitica was founded by Primož Kordež, who regularly contributes DAI-related metrics to the MakerDAO governance and wider DeFi community.
Intro: The Migration
On November 18th, 2019, MakerDAO’s Multi-Collateral DAI (MCD) update went live on the Ethereum mainnet. It marked the most anticipated milestone on the largest defi project to date.
Before November 18th, the MakerDAO platform allowed users to draw loans in the SAI stablecoin using just ETH as collateral (SAI stands for the token on the former single-collateral DAI [SCD] platform). The November update was designed to allow users to draw loans in the new DAI stablecoin using many forms of collateral, ETH included. During this update, SAI holders were able to redeem their stablecoins 1:1 for the new DAI. SAI still exists, but at some point the community will vote to shut down the single-collateral SAI legacy platform, ensuring the defi community continues with the multi-collateral system.
Note: To clarify, SAI refers to the token in the single-collateral DAI platform, abbreviated as SCD. DAI refers to the token in the multi-collateral DAI platform, abbreviated as MCD.
The collateral options in the new platform are not set in stone. They are “extendable” — meaning collateral types can be added (or removed) by virtue of the Maker governance process, which relies on community votes. As of January 2020, the collateral types offered in the MCD platform are ETH and BAT (Brave’s digital advertising token). Granted voter approval, however, other tokenized assets can be made available as collateral.
With the support of different types of collateral, two additional updates occurred. First, the new system now stipulates that stability fees are to be paid in DAI instead of the MKR token, as they were in the former SCD system. Additionally, the stability fee is now added to your debt on a continuous basis. Whereas in SCD, if the price of ETH remains the same you will never be liquidated, in MCD, since your debt is always increasing, you will eventually be liquidated.
The second major improvement includes the DAI Savings Rate, or DSR, which is the capability that DAI owners now have to earn interest with their DAI. This mechanism doubles as a way for MakerDAO to also control the demand side, whereas beforehand changing the stability fee was the only monetary policy lever. The stability fee will act as the lever for managing supply, and the DSR as the lever for demand. Using the DSR will be as simple as transferring DAI to a smart contract, with the added benefit of no counterparty risk.
A Lesson in Successful DeFi Governance and Execution
So far, DAI creation has been a success. On December 17th, more than 50% of the SAI in circulation was successfully migrated to DAI, less than a month after the announcement on November 18th.
BAT was the first collateral to be voted part of the new MCD system. Though the adoption rate of BAT as a new collateral has been considerable, WETH still makes up the majority of the collateral (in USD) behind DAI (BAT = ~$4.5m vs. WETH = ~$355m [01/30/20]). However, when we look at the relative market caps of both tokens, we see a similarity: 1.5% of all the BAT in circulation and 1.8% of all the ETH supply in circulation are now locked in Maker.
Note: even though the BAT vote was overwhelmingly in favor of adding the token (99.82%), it was not the first choice of the voters for the first new collateral on Maker. Augur’s REP token initially won the vote (figure 1.5), but Maker’s risk team assessment concluded that Augur’s impending v2 launch and the possibility of the REP token’s worth going to 0 was too risky to add it as collateral.
MKR has two primary functions: governance and utility. Holding MKR comes with voting rights (including MCD collateral) that are proportional to the amount of MKR held. As a utility token, MKR is used to pay fees by SAI CDP users. This fee is then burned, making MKR a deflationary token. The original supply was fixed at 1 million MKR. Around 11k have been burned, and the current supply is around 989k. The November update and migration required the SAI CDP positions to be closed, triggering the payment of stability fees. The expectation was that MKR burns would increase. In the weeks preceding and immediately following the update, we can see the burn rate did increase (figure 3):
There are currently over 17k MKR token holders.* The top 3 largest holders collectively own 38% of the total supply (with just the top 6 holding over 50%). These top 3 addresses are: 1) MakerDAO’s governance contract (202k MKR), 2) MakerDAO’s Multi Sig wallet (117k), and 3) Andreessen Horowitz’s wallet 60k (purchased ~6% of the MKR supply in September last year at a 25% discount).
* Holders were obtained by taking the distinct addresses that participated on either side of one of the ~611k MKR token transfers, and getting their balances. Those with balances >0 made up the current holders datast.
On the SCD platform, there have currently been more than 154k CDPs opened. Most of these CDPs are still open; just 19,609 positions have ever been closed, and only 2,587 since the migration began.
The largest open CDP (by amount of collateral locked), CDP #3088, was opened on August 28, 2018 and currently has 171,008.72 PETH locked as collateral. The owner of this CDP, address 0x…50a9, has drawn over 10 million SAI, paid back 23% of these (having a current outstanding debt of 8,280,868. 93 SAI), and has accrued over $800k in fees.
This CDP has a collateralization ratio of 331%, way over the required 150% limit. This address locked $15 million dollars more than it would have been required to by the minimum collateralization ratio. When aggregating across all the existing SAI CDPs, the collateralization ratio is currently at over 300%. This means that half of the ETH locked as collateral for Maker SAI positions is just being used as a safety net in case of sudden ETH price drops. It would take ETH prices to drop under $80 for the overall legacy platform’s collateralization ratio to drop under 150% (caeteris paribus).
However, it is interesting to see how the legacy platform positions reacted to actual fluctuations in ETH price — recently, November & December 2019. In the first month following the launch of MCD, ETH prices dropped by >30% from $183 to $122 ($183 to $140 between Nov. 18–25 and from $145 to $122 between Dec. 11–18). But it’s not slow, steady declines that affect users — it’s the quick, almost-overnight drops that do the most damage.
There was a total of ~38k ETH liquidated between these dates, with the largest amount on November 22nd for 19,404 PETH (the largest single amount a CDP was liquidated for was 9,839.89 PETH, on CDP #16,843).
On average, a drop of 7.2% in ETH prices is enough to trigger liquidations (this figure was calculated by getting the average price drop in the days where liquidations for more than 1,000 PETH occured).
With some very rough calculations, we could say that market keepers made somewhere around $170k. Note: under the assumption that liquidated collateral gets the full 3% discount offered, but that is not always the case due to gas prices, liquidity, accuracy of the price feeds etc. We’re also using the average ETH price registered in this timeframe, which is $149.
The smooth migration flow that we witnessed in the last 2 months was possible due to the implementation of a migration contract, which holds migrated SAI and makes it available for SCD CDPs to utilize it and perform fast migration of their SCD debt position to multi-collateral DAI.
The contract that stores SAI is called saiJoin and the total inflow of SAI over the migration period amounts to ~99m SAI (97% of the SAI supply before the migration started). However, a lot of SAI liquidity was provided by fresh SCD CDPs during the migration period. This was expected, as SCD CDPs needed SAI liquidity to unwind their positions in addition to borrowers from secondary platforms such as Compound. Therefore, more than 38m SAI was minted during the migration period to enable smooth migration and sufficient SAI liquidity for borrowers to deleverage and migrate.
Figure 6demonstrates how SAI inventory in the migration contract has been continuously drained within a day or two until the last 2 weeks, where the SAI inventory began to pile up. This demonstrates recently reduced interest among CDP owners to migrate.
Remaining SAI Holders and Borrowers
MakerDAO’s governance is considering when to perform an emergency shutdown for the SCD platform to finalize the migration over to the MCD platform. Analyzing remaining SAI holders and SAI debt held by CDPs and their activity becomes an important variable in decision making.
Out of the 100 largest CDPs (which together represent 86% of total SAI debt), 30 CDPs were totally inactive during the migration period. They represent 11.5m (41%) of the top 100 CDP remaining SAI debt and they were last active between 6 and 12 months ago. This group of debt holders also presents a considerable amount of accrued fees. In total, this group owes 1.2m SAI out of the total CDP unpaid fee amount of 2.2m SAI. The largest share of unpaid fees is carried by CDP #3088, which owes over $800k (10% of its remaining SAI debt).
CDPs that were performing debt repayment activity amount to only 5m or 20% of SAI debt. These may be the only CDPs seriously thinking to migrate soon. The other part of the debt is represented by CDPs who were either minting additional SAI or adding ETH to collateral (about 7m SAI, or 28%).
Similar to SAI CDPs, SAI holdings are also quite concentrated. The top 100 SAI addresses hold 75% of the remaining supply, currently at 28.5m SAI. However, many of these top addresses are pools such as Compound cSAI pool (which holds 2.4m SAI and represents hundreds of users). Known addresses revealed by Etherscan (mostly secondary lending platforms and trading venues) hold more than 6m SAI. This supply should soon be migrated to DAI, as SAI is becoming less liquid and is yielding less compared to DAI lending markets.
There are 1,140 wallet addresses that hold more than 10 SAI and have not been active for more than a year. Together, they account for 5.2m SAI — nearly 20% of the current supply. The average age of inactivity weighted by SAI value is almost 18 months and the top 10 inactive addresses hold more than 2m SAI. It could also be that some amount of SAI held in those addresses is lost forever.
There has been a huge drop in SAI trading activity across decentralized exchanges (DEX). Currently, only Uniswap and Kyber offer decent liquidity for SAI, with daily trading volumes on average reaching $250k in the last week. Uniswap SAI liquidity is at $1.2m compared to DAI liquidity of $2.9m. This has become a serious problem for remaining SCD CDPs wanting to deleverage, particularly for those trying to increase their collateralization ratios before getting liquidated when ETH price is crashing.
Secondary Lending Markets
Introduction of DSR in the MCD platform created a more competitive landscape for secondary lending platforms. It seems MakerDAO overtook Compound as the #1 spot for DAI savers, also because Compound implemented DSR for its un-utilized supply. Market share also significantly decreased for dYdX as its value of DAI deposits dropped by about 80%. The total value of DAI deposits represent about half of current DAI and SAI total supply, which is interestingly less than ATH of 66% reached during September last year. This may be due to less-active remaining SAI lending markets and higher interest rates seen last year (savers could earn 13% yield during summer last year versus 6% currently).
The market share of DAI borrowing has also dropped significantly for secondary lending platforms. The total value of DAI and SAI outstanding loans on secondary platforms currently reaches 22m — only 15% of MakerDAO’s borrowing or total supply. This ratio was much higher in the past, when in September of last year 36m SAI (46% of total supply) was borrowed at secondary lending platforms. The recent increased market share of MakerDAO borrowing is probably related to lower borrowing rates at MakerDAO that were voted for migration purposes.
The number of daily active DAI addresses stands at ~2k, which is 10% below the same SAI metric estimated for the 6 months before the migration. Note: the large count in active addresses for SAI during summer 2019 is associated with Coinbase Earn DAI [now, SAI] campaign and therefore is not directly comparable with current DAI address activity.
DAI Locked in DSR
DAI locked in DSR has been steadily increasing since the beginning of the migration period. The most notable increase was seen in the third week of December when Compound implemented DSR for its un-utilized DAI and during the second week in December when the DSR rate increased from 2% to 4%. Interestingly, the increase of DSR from 4% to 6% did not increase the pace of DAI being locked in DSR. In fact DAI in DSR decreased in the third week of January and pushed DSR utilization down even more severely as DAI supply increased by 45% since the beginning of the year and outpaced the amount of additional DAI locked in DSR.
Altogether, the MakerDAO migration from the single-collateral to multi-collateral platform has proven immensely successful. Overwhelmingly, the ecosystem responded to the news by taking the expected actions to provide momentum and network legitimacy to the adoption of the new platform. The success of the migration can be seen as a strong signal of the continued and growing attention given to defi platforms.
MakerDAO’s migration isn’t complete yet. An important governance decision of when to terminate the single-collateral platform is impending. The closure of the former platform is necessary for the defi ecosystem to decidedly shift to a more effective and versatile platform. However, a tremendous amount of people’s personal funds is still locked in the SCD platform. MakerDAO has proven a success in defi evolution, adoption, and community communication. Now, it will serve as a test of on-chain governance as it looks to its next steps.