Compound D3M Risk Assessment

Legal Disclaimer: This communication is provided for information purposes only. This communication has been prepared based upon information, including market prices, data and other information, from sources believed to be reliable, but Maker has not independently verified such information and makes no representations about the enduring accuracy of the information or its appropriateness for a given situation. This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any digital assets and the use of finance-related terminology are for illustrative purposes only, and do not constitute any recommendation for any action or an offer to provide investment advisory services. This content is not directed at nor intended for use by the MakerDAO community (“MakerDAO”), and may not under any circumstances be relied upon when making a decision to purchase any other digital asset referenced herein. The digital assets referenced herein currently face an uncertain regulatory landscape in not only the United States but also in many foreign jurisdictions, including but not limited to the UK, European Union, Singapore, Korea, Japan and China. The legal and regulatory risks inherent in referenced digital assets are not the subject of this content. For guidance regarding the possibility of said risks, one should consult with his or her own appropriate legal and/or regulatory counsel. Charts and graphs provided within are for informational purposes solely and should not be relied upon when making any decision. The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others.

  1. Summary Proposed Risk Parameters
  2. Overview
  3. Metrics and Analysis
  4. Risk Parameters

Summary Proposed Risk Parameters

Line: starting with 10m, later recommending up to 20% of real DAI supply (126.6m)

DC-IAM gap: 10m

DC-IAM ttl: 12h

Bar: 4.2%

A link to our Compound D3M Model specification with inputs and outputs can be found here.

Overview

The notion of a Direct Deposit DAI module (D3M) was first introduced by Stani Kulechov from Aave in a Maker forum post in April 2021. That particular iteration of the D3M was primarily focused on the Aave V2 protocol. At the same time, hexonaut from the Protocol Engineering Core Unit published MIP50: Direct Deposit Module. The MIP50 introduced a smart contract implementation of a Direct Deposit Module. The implementation allows the Maker Protocol to connect with any third-party lender that MakerDAO governance approves. The Aave D3M was successfully implemented on November 2, 2021.

Recently there has been discussion to expand the D3M product offering to other protocols. An informal poll was published by PaperImperium on November 6, 2021, asking if MakerDAO should prioritize a Compound D3M implementation. Three key benefits of implementing a Compound D3M were raised in the post:

  1. Diversified revenue
  2. Diversified credit risk from PSMs and Aave
  3. Timely inclusion of Compound does not make it appear Maker favours Aave

On December 4, 2021, GFXLabs posted a Signal Request asking the same question. The Signal was approved by the community on December 17, 2021. An on-chain poll is currently live, polling the Maker community on whether onboarding a Compound D3M should be a priority. The poll ends on January 13, 2022.

This report will now assess the risks related to collaborating with Compound and implementing the Compound D3M.

The Compound D3M will operate in a similar fashion to the Aave D3M. In order to stabilize the Compound DAI market borrow rate, the D3M will mint and directly deposit DAI into the Compound protocol in periods of high demand, and remove liquidity in periods of low demand. This report will address related risks and use a report structure similar to the Aave D3M Risk Assessment. Some risks in the Compound D3M are almost identical to the Aave D3M. These risks have been reiterated in the report.

Metrics and Analysis

Solvency Risk

Compound Risk Mitigation Mechanisms & Risk Parameters

When users deposit cryptoassets into the Compound Protocol, they receive cTokens in return. cTokens can be understood as interest-bearing redeemable certificates of an underlying asset deposited to the Compound Protocol. By holding cTokens, users earn interest through the cToken exchange rate. The value of the minted cTokens increases over time relative to the underlying asset. cTokens can be used as collateral for borrowing within the protocol.

The Compound Protocol has a number of mechanisms and parameters in place to cover protocol deficits and help mitigate insolvent debt. The Compound Protocol utilizes a risk management layer called the “Comptroller”. With respect to solvency risk, the Comptroller determines how much collateral a user is required to maintain when borrowing tokens. This is referred to as the borrow limit, which is a function of the Collateral Factor parameter. Collateral Factors are assigned to each individual cToken type. The Collateral Factor can vary between cToken types, normally within the range of 0-90%. The parameter can be changed by Compound Governance. Larger and more liquid assets, such as ETH, generally have a higher Collateral Factor compared to smaller and more illiquid assets. The total value of a user’s supplied assets is multiplied by the market’s Collateral Factor. This is then subtracted by the borrowing balance to derive the borrowing limit (account liquidity).

If an account’s borrowing balance exceeds the borrowing limit, liquidations can occur. In such a scenario, a liquidator’s assets, denominated in the liquidated debt, are exchanged for the borrower’s underlying collateral of the liquidated loan. In other words, liquidators pay back the liquidated debt to the protocol by buying the borrowers underlying collateral. Liquidators are incentivised to do this since the borrower’s underlying collateral can be purchased at a discount rate. The discount rate is determined by the Liquidation Incentive parameter, which is covered in more detail in the “Liquidity Risk” section.

Compound Markets

At the time of writing, the Compound Protocol offers 16 assets for borrowing and lending. The total value supplied to the protocol is circa $18.52 billion. Approximately $7.57 billion is currently borrowed. Five stablecoins are available for borrowing and lending: (i) DAI, (ii) USDC, (iii) USDT, (iv) USDP and (v) TUSD. The DAI market is currently the largest stablecoin market on Compound with approximately $4.64 billion supplied and $3.72 billion borrowed. The second-largest stablecoin market is USDC with $3.83 billion supplied and $2.79 billion borrowed. Compound’s largest non-stablecoin markets are (i) ETH ($6.78 billion supplied and $2.22 billion borrowed), and (ii) WBTC ($1.68 billion supplied and $111.94 million borrowed). These values include recursive leverage positions. The protocol still accumulated reserves for recursive leverage positions, but in the case where COMP rewards would be discontinued, the recursive positions would end up unprofitable. Recursive leverage is covered in more detail in the “Compound Loan Book Positions & Collateralization Metrics” section outlined below.

Cross-asset collateralization is possible in Compound. This means that loan positions in Compound, compared to the Maker Protocol, can become relatively complex. For example, Compound users can build unique types of exposures that are not available in the Maker Protocol. These include (i) synthetic short positions (e.g. LINK collateral and ETH debt), and (ii) normal short positions (e.g. DAI collateral and LINK debt). The increased complexity of cross-asset collateralization means that Compound users can be exposed to both general cryptoasset market crashes as well as high volatility between listed assets.

In some instances, Compound offers lower collateral requirements compared to Maker. This is however not always the case. For example, the collateral factor for ETH is 80% on Compound. This corresponds to a 125% liquidation ratio in Maker terminology. At Maker, the liquidation ratio is 145% for ETH-A and 130% for ETH-B. The liquidation ratio for WBTC on Compound is approximately 153.8% (65% collateral factor). At Maker, the liquidation ratio is 145% for WBTC-A and 130% for WBTC-B. This means that the health of the Compound Protocol, compared to Maker, is not necessarily more susceptible to sudden price drops. However, the Maker Protocol makes use of some additional mechanisms which are not currently available in the Compound DAI market:

(i) Debt ceiling: In order to limit the risk exposure of certain collateral, MakerDAO uses a debt ceiling concept to control the amount of DAI minted from a particular vault type. The Compound Protocol, in some instances, applies borrow caps to cToken markets. An asset’s borrow cap determines the amount of that asset that can be borrowed in a specific cToken market. It should be noted that the borrow cap is an aggregate parameter. As such, it does not necessarily limit the amount individual positions can borrow. Currently, UNI, COMP, MKR, AAVE, and YFI have borrow caps. All other markets, including DAI, are currently set to an “unlimited” borrow cap. Unlike a debt ceiling, a borrow cap does not limit the exposure of the underlying collateral within a borrowing position. Hence, risks are only partially mitigated. For example, users can theoretically borrow unlimited amounts (all available supply) of DAI using any available asset as collateral. A supply cap would help to manage and limit the risk exposure of individual collateral types. Such a supply cap would work in a similar fashion to MakerDAO’s debt ceiling.

(ii) Dust parameter: MakerDAO utilizes a debt floor parameter for collateral types, called “dust”. The dust parameter controls the minimum amount of DAI that can be minted for a specific vault type. This parameter mitigates the risk of a niche attack vector where large positions can be divided into many small ones during low gas fee periods. Small borrowing positions can result in the accrual of bad debt if high gas costs make it prohibitive and unprofitable to liquidate such borrowing positions. Compound does not have a debt floor parameter in place. According to ChainSecurity, the Compound Protocol has 27,600 USD worth of bad debt which is due to a lack of incentives to liquidate very small positions. This suggests that a debt floor parameter would be beneficial. However, the protocol wide insolvent debt in Compound is currently at a very low level.

Protocol Reserves

A portion of the accrued interest from a specific cToken market is kept as reserves in a Reserve Pool. The primary use case of a Reserve Pool is to mitigate borrower defaults and liquidation malfunctions. The amount of interest that is allocated to a Reserve Pool is determined by the Reserve Factor parameter. For example, the cDAI market’s Reserve Factor is currently 15%. This means that 15% of the interest paid by borrowers is allocated to the DAI Reserve Pool. The Reserve Factor differs between markets and can be changed by Compound Governance.

Accrued Reserve Pool funds are denominated in the same asset as the particular market. For example, DAI is accrued in the Reserve Pool for the DAI market, ETH is accrued in the Reserve Pool for the ETH market, and so on. At the time of writing, approximately $15.54 million is allocated to the DAI Reserve Pool. Total reserves for the Compound Protocol amount to $41.82 million. The current dollar-denominated Reserve pool distribution is shown in the chart below. The dollar-denominated value can vary due to the volatility of the underlying assets.

Source: Compound Markets

Reserve Pools in the Compound Protocol are similar to the Ecosystem Collector in the Aave Protocol. Aave, however, has other funds and sources which can be used to cover potential protocol deficits, such as the Aave Safety Module and Ecosystem Reserves, which are mostly denominated in AAVE tokens. In addition, Aave has the ability to mint new AAVE tokens in a scenario where the aforementioned sources are not sufficient. Between the two protocols, Aave protection mechanisms seem to be more developed. However, Aave generally has a more aggressive approach when it comes to available asset markets.

Compound Loan Book Positions & Collateralization Metrics

We took a snapshot of the largest 3131 accounts in Compound on December 18, 2021. These accounts represent the vast majority of assets in the Compound Protocol. The main objective is to examine positions that include DAI. We sorted the positions into (i) recursive leverage positions, (ii) supply only, (iii) long positions, and (iv) short positions.

Recursive leverage positions are positions that are created by supplying DAI, borrowing the same asset, and then supplying the proceeds of the loan back to the protocol. Such positions can maximise liquidity mining yield. Without liquidity mining rewards, this strategy would be unprofitable. The process can be done several times, as each cycle adds new collateral which can be used to borrow again. These positions can be built manually or in one transaction using flash loans. Supply only positions consist of users that only supply assets and do not borrow. Long positions consist of volatile collateral and borrowed fiat stablecoins. Short positions consist of borrowed volatile assets and either stable or volatile collateral types.

The DAI related wallet strategies for Compound are illustrated in the pie chart below.

Source: BlockAnalitica

The chart above shows that 72% of DAI borrow positions on Compound are recursive leverage. This trend is similar to Aave’s loan book. Probably because both protocols offer liquidity mining incentives. Such positions are generally not exposed to market risk. However, if poorly managed, they can get liquidated due to accrued interest. The credit risk from the Compound DAI market is primarily determined by organic DAI borrowing positions. At the time of writing, approximately 28% of organic DAI borrowing positions are long positions.

The pie chart below shows the distribution of underlying collateral for DAI borrowing positions, excluding DAI as collateral. ETH and WBTC make up the vast majority of collateral (92.1%). The value of all collateral is $1.72 billion. The real DAI borrowed (actual DAI debt at risk) is approximately $632 million. This means that the real collateralization ratio is at a healthy level of approximately 272% that is primarily composed of high-quality assets such as ETH and WBTC.

Source: Block Analitica

It is important to note that the credit risk of the Compound DAI market can vary over time as DeFi market conditions change. For example, if yield farming incentives diminish on Compound, users may change from recursive leverage positions to leveraged long positions. This would increase the credit risk of the DAI market. For this reason, it is important to continuously monitor the state of the Compound DAI market and change strategy according to new and evolving market dynamics.

Liquidity Risk

In any given cToken market, specific interest rate models are applied to taper borrowing activity in periods of “too high” borrowing utilization. Too high utilization happens when a market gets dangerously close to, or reaches 100% utilization. Full utilization means that all supplied assets are being borrowed. Such a scenario is suboptimal and could lead to illiquidity and increase the “run on bank” risk. Suppliers will not be able to withdraw their supplied capital, and borrowers will not be able to take loans. Hence, an interest rate model needs to be able to find and sustain an adequate utilization level to maintain the health of the protocol.

For the Compound DAI market, adequate utilization levels are achieved by implementing a linear kinked interest rate model. The Compound DAI market currently uses the JumpRateModelV2 to determine the rate curve. The “target utilization rate” is the desired utilization level of a market which an interest rate model is designed to sustain. The “kink” parameter determines where the interest rate curve changes shape. The “jump multiplier” influences the slope of the curve after the “kink”. A higher interest rate after the “kink” should drive borrowers to repay their loans while attracting additional suppliers. This should naturally push back the utilization rate to the equilibrium target rate.

The chart below illustrates approximately (according to the interest rate model specifications) what the borrow and supply rates would be at different utilization rates. Note that Compound’s interest rate model is a function of how many Ethereum blocks are mined. Hence, the model shown in this report may vary from other sources, since different block time assumptions may be used. At the current target utilization for cDAI (80%), the borrow rate should be 4.58% and the supply rate 3.11%. When the utilization rate exceeds 80% the slope of the interest rate curve becomes much steeper. At 100% utilization the borrow rate increases to approximately 29.54% and the supply rate to circa 25.11%.

Source: Compound: The Money Market Protocol & JumpRateModelV2

The chart below illustrates the historical supply rate, borrow rate, and utilization rate of the DAI Compound market for 2021. The desired target utilization rate has been set to 80% for the whole year.

The Aave Protocol uses a similar interest rate mechanism to control utilization rates. As discussed in the Aave D3M Risk Assessment, Aave reached its highest DAI utilization rate of 2021 on February 14, at 94.24%. At this level, the supply rate was 21.41% and the borrow rate 57.58%. It took 6 days to reach a utilization rate around the 80% level, at 81.25%. Conversely, the highest utilization rate recorded in the Compound DAI market was 90.37% on February 10, 2021. This corresponded to a 14.36% supply rate and 19.08% borrow rate. It took 25 days for the utilization rate to reach around 80%, reaching 80.32% on March 7, 2021.

After the kink, the interest rate change in the Aave DAI market is more aggressive when compared to the Compound DAI market. This means that in the Aave DAI market, there are larger incentives for new suppliers to enter the market and for current borrowers to repay their loans. In the example given above, Compound needed 19 more days (4x days more than Aave) for the utilization level to reach the target rate. This suggests that Compound’s DAI interest rate model may take longer to stabilise the utilization rate in times of excessive utilization. All else being equal, this implies that the number of suppliers who can exit the Compound DAI market will probably be smaller compared to the Aave DAI market. Hence, the “bank run” risk may be higher in the Compound DAI market when compared to Aave.

The Compound Monitor page in the Maker Risk Dashboard tracks “real DAI supply & borrow amounts”. Since the majority of DAI supplied in Compound is reborrowed DAI through recursive leverage positions, the real DAI supply is only circa $633 million, approximately 20% of the total $3.17 billion supplied. For the Aave D3M, we recommended setting the D3M line (debt ceiling) to 10-20% of the real Aave v2 DAI supply. For the Compound D3M, we believe a similar approach would be prudent. In our opinion, the best course of action initially, would be to first set the line (debt ceiling) to 10m and eventually increase it to a maximum of 20% of real DAI supply at Compound (126.6 million DAI).

Finally, as discussed in the D3M - DC Increase & Target Borrow Rate Decrease Proposal forum post, in a run on bank/exploit scenario, the Compound D3M would be disproportionally affected. The D3M cannot identify that something is wrong at Compound. Instead of pulling money out, it would supply additional DAI because of the rate spikes. This is one of the biggest risks associated with the D3M. Hence, adequate DC-IAM parameters need to be put in place. We recommend a gap of 10m and a ttl of 12h. Similar to the Aave D3M, an instant shutdown of the module (i.e. no requirement of the GSM delay) with a 7 day cooldown period is also important to ensure protocol safety.

D3M Exposure Simulation

With the Aave D3M Risk Assessment we created a D3M exposure simulation to examine how the D3M would have performed historically. For this assessment, we created a similar simulation for the Compound D3M for the time period January 1, 2021 - December 14, 2021. The simulation illustrates what the D3M exposure to Compound would have been at different Bar (target rate) parameters. Maker Governance will determine the actual Bar parameter.

The chart on the left illustrates the exposure, in DAI amounts, that the Compound D3M would have had with different target rates. The chart on the right shows the share of DAI deposits from the Compound D3M, out of all available DAI on Compound, for different target rates. The highest share of DAI deposits would have happened on February 10, 2021, when Compound reached its highest utilization rate for the whole year, at 90.37%. On February 10, 2021, at a 3% target rate, the Compound D3M would have supplied 42.2% of all DAI in the Compound DAI market. This is very high considering liquidity and bank run risks. We recommend a bar of 4.2% initially. In the historical simulation, this would have resulted in a maximum exposure of 18.90% on February 10, 2021. The maximum amount of DAI supplied would have been approximately 547 million, on April 14, 2021. This would have corresponded to 11.50% of all DAI supply at the time. However, the line (debt ceiling) will play an important role in the level of Compound D3M exposure, which is not taken into account in the historical simulation.

We believe that an initial 4.2% target rate will limit the exposure of the D3M to healthy levels while still being able to successfully sustain borrowing rates and stay productively engaged in the Compound DAI market. According to the Block Analitica Compound D3M Model, on January 12, 2022, a 4.2% target rate would result in 113.45 million DAI deposited or 3.6% of the DAI supply in the Compound DAI market, but actual exposure will be 10 million due to the debt ceiling. This is much lower than one month ago since markets have fallen substantially in recent weeks. A 4.2% target borrow rate implies an effective 3.6% APY (including liquidity incentives) earnings, and 3.78% APY when a 10m debt ceiling is considered for the D3M. A 3.6% APY should be sufficient to cover the credit and solvency risks addressed in the report.

Oracle Risk

The Open Price Feed

The oracle of the Compound Protocol is called the “Open Price Feed”. The current Open Price Feed architecture consists of one main reporter and one so-called “Anchor” price feed.

The main reporter changed from Coinbase to Chainlink Price Feeds in June 2021. The reason for changing the main reporter was to mitigate certain risks and flaws present in the previous architecture that was subject to the Coinbase Oracle. For example, in November 2020, around $85 million was liquidated from the Compound protocol due to an oracle exploit. During a short time period, the price of DAI on Coinbase Pro traded at unusually high prices across DAI/USDC, DAI/USD, and DAI/ETH pairs, reaching 1.3 USD/DAI. This, in turn, led the Compound protocol to believe certain loans were undercollateralized, and ultimately caused 124 addresses to get liquidated. 10 months later, on September 17, 2021, after a Compound governance vote, all borrowers who got liquidated were reimbursed the 8% liquidation penalty fee. In total, 6.8 million DAI was paid out from the DAI market reserve.

The Compound protocol also uses a so-called “Anchor” price-feed. The Anchor is the time-weighted average on-chain price of token/ETH pairs on Uniswap V2. According to Compound Documentation, a View contract (“Price Feed”) is used to verify that the reported price from the main reporter falls within a certain bound of the Anchor price feed. This is used in order to increase the validity of the main reporter price feed. The “Community Multisig”, a 4-of-6 multisig, has the ability to engage a failover which will switch the oracle from the main reporter to the Anchor price feed.

It is also worth noting that stablecoins such as USDC, USDT, and TUSD are fixed at $1. The price of DAI, however, is still fetched through the Open Price Feed oracle.

Risks associated with the Open Price Feed

Compound’s Oracle setup gives rise to a number of potential risks for Maker. In the previous Aave D3M risk assessment, we addressed certain risks associated with Chainlink price feeds. Since Compound also uses Chainlink, these points have been reiterated in points 3 and 4 outlined below.

  1. Dependency: By adopting the D3M module, Maker will become directly dependent on Compound’s price oracle setup. MakerDAO will become reliant on, and place trust in Chainlink oracle price feeds as well as the Anchor Uniswap V2 price feed.

  2. Trust in the community multisig: The Anchor price feed can be activated through a 4-of-6 multisig. The multisig has the ability to change the main oracle from Chainlink to the Anchor price feed. This increases the amount of trust Maker puts in the hands of a few Compound community members.

In the case of a Chainlink oracle malfunction, Maker will have to trust that the Compound community effectively switches to the Anchor Uniswap V2 price feed and that the Uniswap V2 price feed functions correctly.

If malicious actors are able to change the main oracle price feed from Chainlink to the Anchor, while also influencing the price of specific trading pairs on Uniswap, they will be able to manipulate the price information feeding into the Open Price Feed.

  1. Reduced security: Maker Oracles are entirely controlled by Maker Governance. There is no multisig control, hence Sybil guarantees are assured, meaning that no single individual can exert malicious behaviour. In addition, the risk of a third party managing to compromise the keys of multisig signatories is not the case with Maker Oracles. However, these risks are potentially present when there is multisig ownership of oracle price feeds, as is the case with Chainlink.

  2. Transparency and data guarantees: Maker Oracles query data directly from exchanges and use a Data Model to process the data in a transparent way. With Chainlink, on the other hand, node operators fetch data from data aggregator endpoints. Both methods have advantages and disadvantages. Some of the advantages of Maker’s system were highlighted in an analysis by Nik Kunkel, which can be found here. The advantages include:

  • MakerDAO oracles do not have to trust external data services. There is no risk of API service failure or intentional censorship. The Maker Protocol does not have to trust external data services that could potentially provide malicious data.
  • The Maker Protocol can process data using the exact data model determined by Maker Governance. Iterations and adjustments can be made quickly depending on market circumstances. This gives Maker Governance more flexibility and control compared to having to request external service providers to change their API’s to suit Maker’s needs.

By adopting an oracle that does not include the advantages outlined above, there is a risk of compromising full transparency and data guarantees on the oracle price. As well as diminished flexibility and agility to respond to unexpected market circumstances.

Liquidation Risk

Liquidations - Fundamental Concepts

Apart from the Collateral Factor, which was explained in the “Solvency Risk” section, there are two other parameters that are important to grasp in order to understand liquidations in the Compound Protocol. These are (i) the Liquidation Incentive and (ii) the Close Factor.

The Liquidation Incentive is the additional amount of cToken collateral that a liquidator receives when successfully liquidating a defaulted loan. Offering liquidators additional collateral is possible due to the overcollateralized nature of Compound loans. This parameter is used to incentivise liquidators and arbitrageurs to continue liquidations so that the protocol functions correctly. Currently, the liquidation incentive (or liquidation penalty for the borrower) is 8%. Out of precaution, the liquidation incentive was raised from 5% to 8% during the volatile market conditions in March 2020. Note that the whole liquidation incentive is allocated directly to the liquidator. Nothing goes to the Compound Protocol. This differs from the Maker Protocol, where a liquidation penalty accrues to the protocol’s Surplus Buffer that can help mitigate the accrual of bad debt.

The Close Factor determines the portion of the borrowed assets that can be repaid by a liquidator. This parameter can range between 0-100% and can be changed by Compound Governance. The Close Factor is currently 50%. This means that a liquidator can repay 50% of a liquidated loan. The liquidation process can be continuously called until a loan is less than the borrowing capacity and no longer in default. Gauntlet stated that the Compound protocol faces a trade-off between user adoption and protocol security when setting the Close Factor. A high Close Factor, or no Close Factor at all, reduces liquidation risks but results in more severe liquidations for the borrower. A low Close Factor makes borrowing more attractive to the end-user as the potential loss of being liquidated decreases. However, liquidations get increasingly delayed the lower the Close Factor is since liquidators will have to liquidate a loan in several transactions.

Liquidations - Historical Analysis

As previously mentioned in the “Solvency Risk” section, the DAI borrows collateralized by volatile collateral measures about $632 million. This means that Compound’s DAI credit risk is relatively small in nominal terms, despite the high total value locked (TVL).

The chart below illustrates all of Compound’s liquidations from January 1, 2021, to December 6, 2021. For the whole period, approximately $353.73 million worth of loans were repaid by liquidators. The day with the highest liquidation amounts happened on May 19, 2021, when ETH dropped 46% intraday. Approximately $86.59 million worth of assets were liquidated. This represents circa 24.47% of all liquidations during the observed time period. The day with the second-highest liquidation amounts happened on February 23, 2021. On this day, $80.14 million worth of debt was liquidated.

The two pie charts below illustrate (i) the share of liquidated debt assets, and (ii) the underlying collateral of liquidated USDC and DAI debt. Note, this is the same observed time period: January 1, 2021 - December 6, 2021. The pie chart to the right, depicting the share of underlying collateral, only represents the matching value of USDC and DAI liquidated debt. It does not account for the liquidation incentive (i.e. the total seized dollar amount). The majority of debt liquidated during the observed period was denominated in DAI and USDC (76.7%). The majority of the underlying collateral of liquidated USDC and DAI debt was denominated in ETH (75.3%). This relationship suggests that DAI and USDC loans with ETH collateral, represented the majority of liquidated positions in 2021.

The two charts below break down Compound liquidations from May 19 by repaid amount. The total repaid amount was approximately $86.6 million. The vast majority of debt liquidated was denominated in DAI ($33.79 million) and USDC ($51.56 million), backed by ETH. As illustrated in the chart to the right, approximately $1.24 million worth of other debt assets were liquidated (cUSDT, cETH, cBAT, cUNI, cWBTC2, and cCOMP), also mostly backed by ETH. By way of comparison, on May 19, stablecoin loan liquidations amounted to $152 million at Aave, out of which $23.6 million were DAI debt loans. The amount of liquidated debt at Maker amounted to $41.3 million the same day. This means that Compound, compared to Aave, experienced less liquidated debt in general, but higher amounts of DAI debt liquidations. Maker experienced the least amount of liquidations out of the three protocols.

It is worth reiterating that Compound, just like Aave, does not use a 1h oracle delay “grace period”, as Maker does. This is probably the reason for the more intensive liquidations on days of high volatility. Liquidation risks at Compound seem to be reasonably low. According to ChainSecurity, only 27,600 USD worth of bad debt has accumulated on Compound. However, Compound would benefit from incorporating some sort of protection against oracle dependency and auction throughput limitations in tail event scenarios.

Competitive Risks

It is important that MakerDAO avoids cannibalizing its own product offerings and core market while expanding the D3M to other protocols. It is therefore relevant to assess the potential competitive risks associated with implementing the Compound D3M, and how MakerDAO can work with Compound without losing any significant market share in its core business.

Competitive areas

We have identified two distinct areas where actors within the Compound ecosystem compete with MakerDAO. These are outlined below.

Stablecoin issuing

In December 2020, the Compound Chain Whitepaper was published. The Compound Chain is a stand-alone blockchain that will allow Compound users to lend and borrow governance-approved assets from any blockchain platform that is connected to the Compound Chain, through so-called “Starports”. Compound CASH, a stablecoin, will be used as the native unit of account within the Compound Chain. All transaction costs will be denominated in CASH. Users will be able to mint new CASH through borrowing, similar to the way that DAI is created within the Maker Protocol. CASH will automatically accrue interest within the Compound Chain. By utilizing Starports, users will also be able to port CASH to other blockchains, including Ethereum. While this new development may lead to increased competition between Compound and MakerDAO in the future, it is currently too early to assess the impact that a Compound D3M integration would have on this particular competitive area. The mainnet version of Compound Chain has not yet been released. We currently do not see how implementing the Compound D3M would benefit the Compound Chain and CASH stablecoin. The subsequent effects in regards to the Compound Chain can only be accurately quantified once it is live and relevant data is available.

Permissionless overcollateralized debt

The Maker Protocol and the Compound Protocol compete to capture both small and large debt positions. In this context, Maker may want to maintain lower, or relatively equal, interest rates to Compound to stay competitive in the DeFi lending market.

If the Compound D3M is implemented, the bar parameter will play an important role in determining how competitive Compound DAI rates will be compared to Maker. As discussed in the [Signal Request] - D3M Parameters & Guidance for MOMC forum post, a higher bar parameter leads to higher DAI rates on supported third-party lending protocols, which helps keep the Maker Protocol more competitive in terms of DAI lending. However, it also leads to less DAI exposure for the D3M. A lower bar leads to higher DAI exposure for the D3M but lowers Maker’s competitiveness due to the lower DAI rates on the supported third-party lending protocols. It also leads to higher credit and liquidity risks because of the higher D3M exposure.

It is important to keep in mind the implied effective borrowing cost when setting an appropriate bar rate. The chart below compares the historical rate movement for depositing ETH and borrowing DAI on Compound and Aave since November 5, 2021. The chart includes liquidity mining rewards to illustrate the effective borrowing cost when using ETH as collateral. Note that the effective borrowing cost can vary depending on the price of AAVE and COMP since liquidity mining rewards for the two protocols are denominated in these assets.

Source: Blockanalitica

In the observed time period, Compound DAI borrow rates (with rewards) are on average 0.54% higher than Aave DAI borrow rates (with rewards). This makes the competitive risk of the Compound D3M smaller compared to the Aave D3M.

At the time of writing the effective DAI borrow rate collateralized by ETH at 200% collateral ratio is 2.71%. If Maker sets a 4% target rate, the implied borrowing cost of DAI (including rewards on borrowed DAI and deposited ETH) on Compound would be 2.28%, 47bps lower than the current 2.75% ETH-A stability fee. The Risk Core Unit recommends a 4.2% bar initially. This would result in a 2.48% effective borrowing cost of DAI on Compound.

Aside from borrowing rates, Compound and Maker provide other trade-offs that may affect borrowing demand for different users. We recommend that the community takes all of these factors into consideration when deciding an appropriate bar:

  1. Liquidation Ratio: For both Compound and Maker, ETH is the most used collateral. The Compound ETH market requires a minimum collateral ratio of 125%. Maker offers a LR of 145% for ETH-A, and 130% for ETH-B. As Compound only allows for partial liquidations, liquidations are generally less capital intensive compared to the Maker Protocol. Maker, however, offers the OSM delay, which gives borrowers a 1h “grace period” which to some degree makes up for the higher liquidation ratio.
  2. Rates Behaviour: Borrowing rates for Compound are more affected by large instant borrowing volumes or withdrawals. For Maker, rates are much more fixed. For this reason, larger borrowers generally favour the Maker Protocol.
  3. Liquidation Penalty: The liquidation penalty of 8% at Compound is lower than Maker based loans of 13%, plus keepers profit share.
  4. Collateralization: Compound offers cross-collateralization, which Maker does not. This seems to be an important feature for increased retail demand.
  5. Dust: Maker incorporates dust parameters for all collateral types. The dust parameter has recently increased for certain collateral types and may increase further for other collaterals in the future. Maker has increased dust levels to protect against unprofitable liquidations and the accrual of bad debt. However, it has also led to less accessibility for smaller borrowers. Compound does not incorporate a dust parameter, which favours smaller borrowers.

Due to the increase in dust, MakerDAO is only accessible to larger borrowers and hence borrowing concentration is increasing, while retail borrowers prefer, or can only afford, to borrow on platforms such as Compound or Aave.

On the one hand, a case can be made that by setting lower rates through the D3M on lending protocols such as Aave and Compound, Maker is able to provide DAI to smaller borrowers who could otherwise not afford to borrow directly from Maker. Note also that Maker will not have complete control over the Compound DAI rate, since the D3M will only be able to control the rate up until the maximum debt ceiling (line) is reached.

On the other hand, by setting lower rates on third party lending protocols, the case can be made that Maker is cannibalizing its longer-term business objectives. The Multichain Strategy and DAI Wormholing is materializing. This will make borrowing on Maker more affordable. In addition, Compound liquidity mining will most likely not continue forever. When liquidity mining eventually stops, effective borrowing costs will go up on Compound. This will allow Maker to set a higher bar value on Compound and ultimately maintain Maker’s competitiveness. From this point of view, if the D3M begins to replace core collateral types, Maker should consider increasing the spread between the D3M target rate, and vault stability fees.

Governance Risks

Compound uses an on-chain governance system to manage the Compound protocol. Elements of the Compound protocol that can be changed through the governance process include (i) changing system parameters, (ii) adding new markets, and (iii) adding new functionalities to the protocol. These changes can subsequently impact the risk profile and market dynamics of the Compound D3M.

Governance decisions normally begin with discussions on the Compound forum before a proposal is submitted on-chain. Governance participants use the governance token, COMP, for either voting or delegating. A COMP owner can delegate voting rights to any Ethereum address. An address with a minimum of 65,000 COMP tokens is able to create governance proposals. Once a proposal has been created, it enters a two day review period. Thereafter, voting weights are recorded and the voting begins. Active proposals are listed in the governance dashboard on the Compound website. The voting lasts for 3 days. If a majority, and at least 400,000 votes, is cast for a proposal, the proposal is queued in the so-called “Timelock”. The Timelock contract has a minimum delay of two days. Once the delay period is over, the proposal can be executed. In total, any change to the Compound Protocol will take at least seven days.

MakerDAO will inevitably be more dependent on Compound governance if the Compound D3M is implemented. The MakerDAO community, and the Risk Core Unit, in particular, will need to monitor Compound governance discussions and on-chain proposals in more detail going forward. The Compound governance timeline is approximately 7 days. MakerDAO’s proposal timelock period is 2 days. This means that we will have about 5 days to organise and pass a proposal to react to governance changes that may negatively impact the Compound D3M. We can minimize governance risk by engaging with the Compound community and expressing any concerns before proposals are put up for vote.

Summary of Notable Risks and Red Flags

  • Solvency Risk: The analysis suggests that the majority of DAI debt on Compound, much like Aave, is recursive leverage (72%, or $1.9 billion). Recursive leverage carries a much lower solvency risk. Real DAI debt at risk is approximately $632 million, with $1.72 billion in collateral. The majority of collateral is denominated in ETH and WBTC. This means that the Compound DAI debt currently has a healthy collateralization ratio with high-quality collateral. However, credit risk may increase for the following reasons: (i) Compound does not use a debt ceiling or dust parameter concept, (ii) the majority of positions may change from recursive leverage to leveraged long positions with volatile collateral if liquidity incentives diminish and (iii) Compound may onboard a risky asset or have misconfigured LTV ratios that could lead to a loss.
  • Liquidity Risk: Liquidity and bank run risks are partially mitigated by the fact that the majority of the DAI loan book consists of DAI recursive leverage. If Compound liquidity incentives are disabled, the total DAI market risks being reduced considerably. In such a scenario, utilization rates may increase quickly. Compound’s DAI interest rate model is less aggressive compared to Aave’s. So while Compound may successfully mitigate close to 100% utilization rates, it may take longer to reach target utilization equilibrium after a period of excessive utilization. In a run on bank scenario, the Compound D3M would be disproportionately affected. The D3M cannot by itself recognise that something is wrong at Compound. Instead of pulling money out, it would supply additional DAI during rate spikes. This is one of the biggest risks with the Compound D3M. Adequate DC-IAM parameters, as well as instant shutdown capability, is needed to ensure protocol safety.
  • Oracle Risk: By implementing the D3M in the Compound Protocol, Maker would become directly dependent on Compound’s oracle setup. This includes a dependency on Chainlink oracle price feeds, as well as the Anchor price feed, in the case that Chainlink, for whatever reason, would malfunction. Maker is, however, already vulnerable to Chainlink malfunctions for at least two reasons: (i) collateral assets available in Maker that rely on Chainlink are tied to the performance of Chainlink Oracle price feeds, and (ii) Aave V2 also uses Chainlink. Implementing D3M in Compound would further increase the Chainlink Oracle dependency for MakerDAO. This dependency increases in relation to the debt ceiling amount for both the Aave D3M and Compound D3M.
  • Liquidation Risk: Liquidation risks at Compound seem to be reasonably low. Most organic leverage on the DAI market is built through liquid tokens such as ETH and WBTC. According to ChainSecurity, only 27,600 USD worth of bad debt has accumulated on Compound. However, Compound would benefit from incorporating some sort of protection against oracle dependency and auction throughput limitations in tail event scenarios.
  • Competitive Risk: Maker and Compound offer different trade-offs in regards to permissionless overcollateralized debt. Benefits of the Maker Protocol include (i) the OSM price delay, (ii) unconstrained liquidity amounts, and (iii) relatively fixed price of debt. Compound, on the other hand, offers (i) liquidity incentives, (ii) the Close Factor parameter - which benefits borrowers, (iii) the ability for smaller borrowers to participate in the protocol (due to the lack of a dust parameter) and (iv) cross collateralization. MakerDAO needs to weigh these trade-offs against each other to ultimately decide an appropriate Bar (target rate) for the Compound D3M. If the D3M begins to replace core collateral types, Maker should consider increasing the spread between the D3M target rate and vault stability fees.
  • Governance Risk: The adoption of the Compound D3M will inevitably lead to more dependency on Compound governance. The governance risks are mostly related to Compound DAI rate curve changes and other risk parameter changes. If Compound governance decides to change the rate curve, Maker will most likely need to react before the changes are implemented in the Compound Protocol. Changes to the Compound protocol take at least 7 days. If the Compound D3M is implemented, MakerDAO, and the Risk Core Unit, in particular, will have to monitor governance risk exposure at Compound on a regular basis.

Risk Parameters

Line: starting with 10m, later recommending up to 20% of real DAI supply (126.6m)

DC-IAM gap: 10m

DC-IAM ttl: 12h

Bar: 4.2%

Depending on when the Compound D3M is launched to mainnet, the 4.2% target rate (bar) may need to be changed to reflect the market dynamics at that time. With a 126.6 million debt ceiling (line), a target rate (bar) of 4.2% would lead to approximately 113.45 million DAI exposure and an effective borrow rate of 2.48% incl. COMP rewards on deposits and borrowed assets. At a 10 million debt ceiling (line) and a target rate (bar) of 4.2%, the exposure would be limited to 10 million DAI, and the effective borrow rate would be 2.63% at the time of writing.

A link to our Compound D3M Model specification with inputs and outputs can be found here. We are still working on improving the underlying data in these calculations, which may cause some minor changes to the output in the coming days and weeks.

After the implementation of the Compound D3M, we will make a similar poll to the [Signal Request] - D3M Parameters & Guidance for MOMC to ask for community feedback on competitive risks and gauge the appetite for max debt ceiling amount.

Authors: @Sean @Primoz @rema

Sources:

16 Likes

Excellent & super informative risk assessment. Thank you for the diligent research, and for providing a thorough and comprehensive risk analysis. Super cool.

I was wondering what are the initial thoughts of the Risk CU on implementing DSS Gate for scenarios where the D3M is exposed to a protocol that has an unlimited borrow cap? Have you had a chance to review and analyze how effective of a tool DSS Gate can be for risk mitigation?

Also, since there is no supply cap within Compound, are there any scenarios for griefing attacks, oracle manipulation, or D3M manipulation that you have model/simulated around?

And last but not least, to GovAlpha— @Patrick_J @LongForWisdom @prose11 would it be possible in the near future to refrain from posting a Poll until a Risk Assessment like this one has been presented to the community? I ask because this assessment is super informative, and helpful. But by now we are in Day 3 of a 4-day Poll, where both Delegates & MKR token holders have either already voted, or will do so without being aware of this post. Wondering what your thoughts are around that, or if there’s a need to improve communication between Core Unit teams? Juts thinking out loud and wondering what your thoughts are around managing such. TY in advanced!

7 Likes

Thanks for the shout out.

From my perspective, a positive outcome in the current on-chain poll is a signal from MKR holders to Core Units that they would like assessments, like this one, to be prioritized so that we can move to a formal onboarding poll and then an executive vote. I would view a positive outcome in the current poll to be an instruction from Governance to the CUs which essentially says “we want a Compound D3M, please prioritize the necessary work to get this done”.

We certainly wouldn’t post a formal onboarding poll without these assessments being done as an onboarding poll would include the proposed risk parameters, listed in this post. Example.

It would seem counter-productive to wait to post a prioritiziation poll until after the relevant CU work has been done as that would just be delaying an onboarding poll at that stage.

@LongForWisdom has previously posted on the subject of prioritization.

Perhaps the wording of the on-chain poll could have made this distinction more clearly. As the author of this particular poll I’d be open to suggestions to how I could have made it clearer. Happy to discuss further.

4 Likes

As @Patrick_J said, the poll is more of an instruction to the core units to prioritize making these assessments. The reason we got this before the poll concluded is mainly just because @Primoz and the Risk Core Unit have been working for the DAO for a while, and at this point they’re good at anticipating what people want.

Basically the poll is saying ‘please look into this’, but given that the core units self-direct to a large extent, the risk core unit was already working on it.

The process for doing things currently looks like this:

  1. (Optional) poll or signal from governance asking core units to work on something.
  2. Relevant Core Units produce domain assessments for the work item.
  3. Once we have assessments from all the relevant core units, we create an onboarding poll including details of parameters etc for the work item.
  4. If the onboarding poll passes, then we move to an executive containing the changes.

This may have been more confusing than necessary in this case. Previously we’ve tried to phrase signals + polls at the #1 stage as: “Core Units should investigate doing x.” Which better communicates that governance aren’t directly voting for onboarding.

4 Likes

Thank you for the reply LFW & MKR/Breaker–understood. This has been the trend since the pre-delegate days. All good.

But with that in mind–recently, we have seen a lot of grit & hustle from an RWA MIP6 applicant, @Scarlet_Chen --where she has ask delegates who have already voted and abstain (because of lack of clarity) to ask questions, as to what is missing–in order for said delegates to vote, Yes, or No. And we All know that these delegates are abstaining because we are voting on an opaque collateral type, as we all need to lean heavily on the experts, in this case a transitioning RWF model.

With the oncoming Arranger model and the propose RWF CU teams (underwriting/compliance):

  • Does GovAlpha believe that putting on the brakes on RWA MIP6 applications is the right move until the Arranger structure & the proposed RWF Collateral Risk level guidelines are set in stone?
  • Or, should we continue to get an overwhelming response of Delegates abstaining?

I ask because that seems to be the trend and I don’t believe its fair for MIP6 applicants (you might agree as well) to go through the process and get an "abstain"vote. Just wondering what GovAlpha’s thinking is around this subject. And I know its been brought up many times by others (@PaperImperium ) but need to retouch the subject as we kickoff the new year.

Thanks in advance for your services and input. Much appreciated!

3 Likes

A post was merged into an existing topic: [RBLD] MIP6 Collateral Onboarding Application - Real Estate - NEW