D3M - DC Increase & Target Borrow Rate Decrease Proposal

Hey all,

D3M launched this week and has seen the debt ceiling fully utilised up to 10M DAI. @PE has received draft audits from ChainSecurity finding no issues. Finalized audits will be provided in the coming two weeks. Protocol Engineering is therefore comfortable increasing the debt ceiling to 50M (hedged by the Surplus Buffer).

This post is therefore proposing a DC increase to 50M next week and will go on to explain some of the D3M risks so that decisions about further increases can be easier in coming weeks.

On our side we’d like to make a few comments on D3M risks, although the majority was already covered in our D3M Risk Assessment. Before I get into details I would like to share a dashboard we made that lets us estimate the amount of D3M exposure at particular DC and Target Borrow Rates which are both governance parameters: MakerDAO Risk Dashboard | Block Analitica. We are also working on a separate D3M monitoring dashboard that lets us monitor credit risks on Aave v2 DAI market and coverage of their Safety Module.

As already mentioned in the initial Risk Assessment, we recommend limiting the D3M exposure at 20% of the Aave v2 DAI supply market. This puts the D3M DC at around 400m right now. This is because in equilibrium, the DAI market is always about 80% utilized (because of the rate curve of which the slope increases at 80% utilization) and we want to assure we don’t face liquidity risks if Maker wanted to shut down the module for some particular reason. In reality we would want a bit more buffer and we’d probably recommend up to 300m DC.

Few words on credit risks: Aave v2 DAI market is about 80% utilized and therefore about 1.6bn DAI is borrowed out of 2bn DAI supplied. DAI being borrowed is collateralized by other assets, and this is the credit risk we are facing. However, a large part of DAI borrowed is collateralized by DAI (recursive leverage DAI liquidity mining) and is delta hedged. We estimated that only about 250m DAI (17%) out of 1.5bn borrowed is collateralized by volatile collateral. And majority of this collateral is either ETH or BTC. There is one larger loan collateralized by MKR, but doesn’t seem large compared to the exposure. This means the DAI borrow market carries relatively low credit risk, currently.

However things can change fast and Aave still doesn’t have a concept of debt ceiling implemented. For instance, if the last exploit at Cream happened at Aave, the DAI market could be drained. The biggest risk of D3M is probably the inability of the module knowing that something is wrong at Aave and instead of pulling money out, it would supply additional DAI because the rates would spike in a bank run/exploit scenario. This is why we recommend caution and extra safety that can be somehow controlled by DC-IAM parameters and instant access of shutting down the module if needed.

Few more words on Target Borrow rate which is currently set at 4%. Because of the DAI interest rate curve at Aave v2, the rate tends to somehow be in equilibrium around 4%. This is because rate significantly increases after 4% and 80% market utilization and forces the market to either repay DAI or supply more DAI and push rate back down to 4% range. We have observed many occasions where the rate is either slightly above 4% or slightly below. This is however annoying for Maker, because the D3M exposure would either be 0 or a positive number, making it less efficient and earning less. This is why we would recommend lowering the Target Borrow rate just slightly to 3.9%. This wouldn’t change the competitive angle by too much, but significantly increase D3M ability to have constant exposure and earn fees.

Risk Core Unit therefore proposes:

  • Increase D3M DC line to 50m
  • Increase gap to 25m
  • Lower Target Borrow Rate bar to 3.9%

Unless there are notable objections from the community or other mandated actors, we would like to include these changes in the on-chain poll next week.

26 Likes

great to see we are moving quick on D3M :slight_smile:

any feelings on the next steps? when are we going to go for the last-stage (300 MM)?

1 Like

I would like to object, actually. Mostly on grounds of size of exposure and some worries I have (outlined below).

Aave Backstop
There are some worrisome whispers (from folks forking Aave on other chains) circulating that Aave’s backstop will not perform as expected in the event of a large amount of bad debt. I would request that @Protocol-Engineering take a peek if they have not already done so.

Additionally, Aave is extremely thinly capitalized. Discounting the boatload of AAVE, they have only $4m in reserves according to this as of today which is down substantially from the ~$25m seen in @Hasu and @monet-supply’s recent article using data drawn from the same source.

ETA: Looks like Aave has $22m in non-AAVE reserves, but the $4m is probably more accurate because it discounts a-tokens. aUSDC, for example, won’t be much use to them if the bad debt is USDC

ETA2: Actually looks like only around $400k in unwrapped tokens.

For reasons I wholly agree with, Monet and Hasu argue that buffers in a native token (whether staked or printed) should be viewed somewhere between “worth something” and “worth nothing” that is hard to quantify and should not be relied upon.

More generally, their financials are fairly opaque and do not help shed light on what is happening here. It’s not clear how much (if any) money they actually make.

Aave Risk Management

Aside from the lack of buffer, I do not feel Aave has as rigorous risk management practices as Maker does. The fact that their AMPL market has been a basket case for several weeks and remains unresolved is a good example. That this problem is one of economic design and not technical oversight – the AMPL token rebasing messes with their rate curve, which should be no surprise – is not reassuring.

The quick removal of liquidity earlier in the week also shows the ability for deposits to be locked in place at Aave, and in a market downturn, I suspect Aave is “safe until it’s not.” The cross-collateralization of different tokens should make things safer, but likely increases the odds of a cascading event if a major collateral has problems. This is unlike Maker where if a single collateral type falls out of bed, the other vaults should be unaffected unless DAI breaks peg.

Competition with PSM for capital flows

The D3M module mints DAI that is either unbacked or backed by Aave’s debt, depending upon your preferred semantics. This is fine. But it does mean that it competes with the PSMs for managing the size of DAI’s monetary base. Ceteris paribus one would expect that 1 DAI minted via D3M to result in 1 DAI unwound from a PSM. Real life is, of course, messier than that, but the direction of the pressure is correct.

When it comes to the PSMs, they are excellent tools at managing the peg and absorbing/providing liquidity for capital flows into/out of DAI when demand for DAI != demand for vaults. More specifically, they are automatic. This is actually much more efficient than what actual central banks do.

The D3M, on the other hand, is meant to target a rate and do so imperfectly (because the market at Aave may be frozen or experience random gyrations unrelated to Maker’s actions), which is much more like what the Federal Reserve and other central banks do. Unless we are very clear that the D3M is used mainly for revenue and keep it sized accordingly, we risk the D3M taking the lead on peg management. Because it presents liquidity risk and (in my opinion) greater credit risk, I think we should steadfastly establish a policy that the D3M is strictly for revenue and not monetary/peg policies. From a monetary/peg standpoint, the D3M should be seen as a substitute for bonds/Treasuries/other parking spaces, and we need an institutionalized rule to remind us of that. Keeping all D3Ms at some percentage of all PSMs is probably the correct solution here, and I would ask @Primoz to think about what that might look like. If that is hard coded into our process here, we can simply perform regular sweeps on a weekly or bi-weekly basis to keep these two tools in balance.

Raise D3M at all?
For now? I suspect we can increase it a bit, but I think 50m is too much. In an unlikely, worst-case scenario, that would leave Maker with less than a year’s operating budget on hand. Given that the only D3M at the moment is with Aave, a catastrophic Aave event would likely impact our other business lines as well, and we would want to have that cushion. Our estimated expenses on an annualized basis are currently around $40million – I think we should make sure we always have that + some actual buffer for bad debt on hand.

TLDR
-Aave’s risk management is questionable in my opinion, and they are the ultimate borrower
-Aave looks broke and their financials don’t suggest that will change
-I advocate for a rule that keeps all aggregated D3Ms to a fixed % of all aggregated PSMs so as not to present liquidity problems in the face of unexpected failure of D3M counterparties

7 Likes

The DAI market size on Aave is around 2B, so if we don’t raise the ceiling eventually to be hundreds of millions we might as well not run the D3M because it will not be able to enforce the borrow rate. Leaving it at 10M or even 50M we may as well of not even created the module.

Is there particular criteria you are looking for before we go to 50M+?

2 Likes

I would suggest we just use it for income.

Aave has $400k in actual cash on chain, notwithstanding their Safety Module. I think we should be careful accepting aDAI at par value in amounts we can’t afford to lose

1 Like

What do you mean by “use it for income”? Are you wanting a high target borrow rate with minimal exposure? Something like 8% or something instead of our current 4% or 3.9%?

2 Likes

In that 4% is above average yield on out own vaults. But not let it interfere with monetary/peg policies. Just treat it like a loan to Aave (which it ultimately is)

At 4% target borrow we get 3% interest or 300k / year at 10M. This is not even worth the dev time on the module imo and doesn’t help users on Aave.

4 Likes

I am a bit confused here. Wasn’t the whole point of D3M that we can basically enforce that DAI borrowers at Aave can be sure that the rate will never go higher than what we set through Maker Governance? (reference)

The MIP passed, if we are not thinking about rolling back from this strategy we should focus on how quickly (and safely) we can get the DC to a proper level enabling this feature and not think too much about the income we generate for us. That’s really just a side effect imho. It is more about the UX people will have when borrowing DAI

5 Likes

We can target a rate, but if we get it wrong it’ll mess with the peg. The PSMs are flush, so that’s not much danger today, but I also don’t trust us to keep our eye on the ball.

I really don’t think we should use D3M to target rates. We’d have to manually adjust it any time the PSMs were drawn down, and we can’t do that quickly with governance, and potentially couldn’t withdraw liquidity from Aave if we needed to.

We are price takers on rates. Actually moving the rates will work against our peg maintenance. We can’t really serve both masters simultaneously.

(Paging @MakerMan to come say “impossible trinity” here)

1 Like

This dashboard might help with the health of AAVE provided courtesy of @inkymaze & the Gauntlet squad: Gauntlet Autogov

1 Like

Doesn’t seem quite up to date, though. It’s missing several collaterals, including the troublesome AMPL market. Also don’t see GUSD. And the MKR utilization rate definitely isn’t the 30% shown.

Thanks, though. Will poke around some more. Aave is suddenly very interesting and want to know more about how it functions.

1 Like

I think this $4m reserves data you got is wrong, but I won’t dispute that their capital is not as large and liquid as we would want it to be. Their other most liquid part of capital is in Balancers stkABPT pool, which consists of $95m ETH. Even though this looks to be more than Maker’s current Surplus Buffer, it isn’t stable and it is questionable how much of that ETH disappears through AAVE token dump before potential coverage of loss.

I don’t see a clear connection between D3M and PSM. Sure someone might borrow this DAI on Aave and sell it for something else and drain PSM, but same holds for every DAI borrowed from whatever vault type at Maker? Why would this be a special case at 50m DC? And also people borrowing DAI might just farm with it somewhere (which is the case mostly) and this wouldn’t impact the PSM.

EDIT: I see where you are going - keeping DAI rates on Aave too low would stimulate DAI borrowing and unwinding of PSM. True, we’d probably not want to overshoot with a DC and I think we won’t with these numbers. If PSM starts draining to dangerous levels we’ll simply lower D3M exposure in time? I think PSM drainage has more implications on fixed rate vaults than D3M though.

All in all I don’t see how these concerns would make a good argument not to increase DC. I personally think this DC is within the risk tolerance compared to all the other vault offerings at Maker, but I let the community decide. It is true that any black swan event (such as the one at Cream) can lead to bad debt and questionable coverage from their reserves. But if we start assessing every collateral with a black swan probability, we should probably put on brakes at many collaterals.

Yet I do think we need a discussion when we start increasing DC above that figure. In the post I made I haven’t even counted all the other risk factors that I would want community to weigh in, for instance:

  • how do we feel about increasing exposure to Chainlink Oracles indirectly?
  • do we want to hold off with this cadence of DC increase until Aave implements debt ceilings that were rumoured to be implemented at some point - the biggest missing risk mitigation parameter at Aave IMO
  • do we want a bigger protection against liquidity risks and ask Aave to implement a rate curve that lowers the “kink” from 80% utilization to something lower?
  • do we want a lower gap parameter that protects us to some degree in the bank run scenario (assuming DC isn’t fully utilized)? This makes D3M less efficient though and comes with other risks.
3 Likes

This type of thinking will kill any large scale initiative. If you start from the assumption that governance will never be able to change its parameters in time to save the peg, maybe we should lower the cumulative DC of all ilks so the USDC PSM is guaranteed to save the peg? I am a little amazed that governance could consider giving up the plan that was discussed at length when the D3M came up before it’s even truly enacted.

4 Likes

So i guess this is no longer a rumour, Aave v3 posted just minutes ago: Introducing Aave V3 - Governance - Aave

7 Likes

I am a little amazed that governance could consider giving up the plan that was discussed at length when the D3M came up before it’s even truly enacted.

Let’s not get ahead of ourselves here. You have one community member voicing their opinion. This will go to a formal vote and MKR holders will have the ultimate say.

Personally I’m in favor of the 50M DC increase. Audits have come back clean and D3M is a critical piece of our distribution strategy for Dai. D3M has the potential to generate an enormous amount of revenue for the protocol. Short term spikes in rates were something Maker could never actually monetize because of our governance processes being extremely slow. D3M changes that. I’m looking forward to those days where a new farm gets launched and Aave borrow rates spike up to 70%. A DC based on the utilization curve makes a lot more sense than trying to constrain it based off of Aave’s balance sheet. You’re not going to find any DeFi protocols that are as well capitalized as Maker (tbf we’re not even well capitalized). Should our appetite for risk evolve over time? Absolutely. But right now we have to play in the sandbox we have. Let’s not shoot ourselves in the foot here.

EDIT: @Primoz It will take a while for Aave V3 to get released, plus afaik we’d have to make changes to D3M and get new audits, etc. I don’t think the introduction of V3 at some point in the future should stop us from executing on our current path.

7 Likes

Seems like they might just upgrade the V2 contracts, which is mildly scary.

We talked with them in Lisbon, they said it would be a new deployment, so agreed, that’s concerning for sure. Will follow up. Good that it’s public now so we can talk about it. If they do the upgrade, we should probably cut the D3M DC to 0 in the week leading up to the migration to V3 to give the new version some time to bake in. Should still be a few months out from our discussions with them.

3 Likes

To be clear, I’m not saying not to use it. Just that we shouldn’t use it to target rates because it will eventually be at odds with our peg. It should be extremely lucrative even then.

Worth noting that this only impacts AMPL depositors, not the integrity or solvency of the rest of the Aave markets. So I think this partly explains why Aave hasn’t taken more forceful action to adjust the aAMPL market (although I hope to see this resolved in the near future).

IMO it’s not problematic for Aave to hold their reserves in aTokens, if the protocol faced losses it could either (1) forfeit their aTokens to reduce pool liabilities, or (2) deposit their reserves to increase pool assets. It should have equivalent impact either way.

I also think the competition for PSM flows is possibly a good side effect of using the D3M - Maker earns a very low return on PSM assets (potentially 0% return if proposal to remove fees passes), D3M may offer a better risk adjusted return and also gives the ability to maintain profitability even if we need to activate the DSR to maintain DAI demand.

3 Likes