[Signal Request] Adjust the Surplus Buffer (March/April 2021)

Overview

Last time we raised the Surplus Buffer we had around 1.6B total DAI and by raising the Surplus Buffer to 30 MM we managed to almost match a ratio of 2% of Surplus Buffer / total DAI.

Since then, a good Billion of new DAI was generated, bringing our ratio back to 1.15% again (and we haven’t even filled the Surplus Buffer yet).

We still have a couple of weeks left until the Surplus Buffer is reached (60 days according to makerburn.com) - but as we need to be at least at 2%, we should prepare to make another move here again.

We have to decide on two things regarding this topic:

  • How high do we want to set the Surplus Buffer
  • Do we want to put all the fees into the Surplus Buffer or do we want to slowly lerp it so some of the fees get into the MKR-burning

To make it more tangible, I added some tables. Please bear in mind, that the fees might go up (and down) based on vault-utilization, Stability Fee changes - so I based the tables on the current state of the System (2.6B DAI, 3.75% average fees)

Assuming we stay at the current 2.6B DAI (and the chances it keeps growing are pretty good) we would end up with the following ratios:

Surplus Buffer in MM Surplus Buffer/DAI
30 (current setting) 1.15%
50 1.92%
60 2.31%
70 2.69%
80 3.08%

Now looking at the effects on increasing the Surplus Buffer. The two dimensions are: how much do we want to increase it, how much of the fees accrued shall be used for MKR burning. This result is the amount of weeks it would need to get the Surplus Buffer getting filled. Please bear in mind that Weeks until SB filled up counts from the point in time where we reached the current (30 MM) Surplus Buffer. The guestimate about the weeks is probably pretty conservative and might be a bit lower (due to more DAI in circulation or higher Stability Fees).

Surplus Buffer in MM Increase Surplus Buffer by MM MKR Burn Fraction Weeks until SB filled up
50 20 0% 10.7
50 20 25% 14.2
50 20 50% 21.3
50 20 75% 42.7
60 30 0% 16.0
60 30 25% 21.3
60 30 50% 32.0
60 30 75% 64.0
70 40 0% 21.3
70 40 25% 28.4
70 40 50% 42.7
70 40 75% 85.3
80 50 0% 26.7
80 50 25% 35.6
80 50 50% 53.3
80 50 75% 106.7

(spreadsheet to clone)

Increasing the Surplus Buffer

Update: please have a look at the details provided by @SebVentures and @Primoz

Pros:

  • likelihood of flopping new MKR in a “too much bad-debt”-event gets decreased
  • I hope we feel a bit better about keeping up with the DAI-demand without going crazy on SFs.

Cons:

  • DAI in SB is DAI taken out of circulation (probably not a big issue anymore, the supply side is elastic due to the PSM)
  • no / less MKR burning

Please vote on ALL options you would support in an onchain-poll.

  • 30 MM (no change)
  • 50 MM (+20 MM)
  • 60 MM (+30 MM)
  • 70 MM (+40 MM)
  • 80 MM (+50 MM)
  • Abstain

0 voters

How much of the fees should go into the MKR burning?

0%: no lerping, the Surplus Buffer will be directly set to the target value. No MKR burning.
>0%: we slowly increase the Surplus Buffer over time to try to burn some of the Surplus.

Pros:

  • We burn some MKR again

Cons:

  • It takes longer for the Surplus Buffer to get filled (see tables above on the effect, up to 4x)

Please note that the actual % cannot get guaranteed, as we would basically just calculate the end-date before adding it to the exec - if we accrue a lot of more fees than expected the burn-ratio will go up, if we accrue less, the ratio will go down (maybe even to 0).

Please vote on ALL options you would support in an onchain-poll.

  • 0%
  • 25%
  • 50%
  • 75%
  • Abstain

0 voters

Next Steps

The Poll will run until 2021-04-01T13:00:00Z; its outcome will result in a on-chain-poll assuming the outcome of the poll deems it necessary.

7 Likes

Shouldn’t we tie the size of the surplus buffer with the risk-weighted assets rather than total DAI in a formulaic way rather than having to have a signal request every month?

Would love to have risk weigh-in on what they feel is an appropriate level based on our current collateral profile but 3% of risk-weighted assets feels like the minimum. Given how fast total DAI is growing, I would rather have governance continue growing our Surplus Buffer and protect ourselves from diluting at unfavorable prices in the event of a bear market and continue investing in growth. We are seeing a lot of Core Unit proposals being submitted and I wouldn’t advocate for burning any MKR at this time until we have more visibility into budgets and expenses.

3 Likes

These polls should probably be separate as I don’t want to increase the surplus buffer unless it’s 50% burn. If I vote for 60M buffer and 50% this could be paired in a way I don’t agree with.

30M/0% or 60M/50%.

I don’t like all this money being unproductive.

10 Likes

I support lerping to slowly allocate fees towards MKR burning now. Instead of having some threshold where we start burning a bigger portion of fees.

4 Likes

We will provide an answer with @Primoz. Probably next week.

3 Likes

Polled for 80MM SB–with 2.6B outstanding DAI and an upcoming move to progressive Decentralization via Core Units + RWA + Liquidations 2.0 + Optimism bridge–we should have at least 3% of reserves in the SB, IMO.

“The best product doesn’t always win, but the winning product becomes the best product.” -Ben H.

1 Like

We do not necessarily have to have those reserves in the SB, but as long as there is nothing else - that’s the place to have it.

It is easy to decrease the SB later on

2 Likes

If we agree that the current surplus buffer is too low, wouldn’t we rather have it fill up quickly and get us to safer buffer levels than try burning and adding to the buffer at the same time?

Aside from MKR holders not seeing their token go up in value, is there any objective reason to care about a delay to reignite the MKR burn?

3 Likes

I feel that at some point, such a big reserve should be doing something for us. Even if you see it as an insurance, even insurance companies make profit on the time between reserve accumulation and claims. A $30 million buffer, if invested at 1% could offset the cost of a full-time dev.

7 Likes

Dai are already invested unless they are in usdc.
Actually even those one are invested in usdc at 0%

So when I bring up using lerp, I’m also thinking about long term MKR value, not just short term. The benefit of mixing a bit of burning is that the protocol gets to burn over a longer time period. Possibility of being able to burn more MKR in the long term instead of waiting to max out.

Assumptions:

  • A 30MM DAI buffer doesn’t present an existential risk to Maker, but we’d still be committed to growing surplus reserves.
  • Once we get an 80MM buffer, MKR will be a lot safer. Price might be a lot higher. May result in less MKR burned.

We’ll see what the Risk team has to say about how dangerous the current reserves are. The SB isn’t the only thing regarding protocol safety. Bigger impact is the risk exposure from vaults and the robustness of the liquidations system. So great steps forward with Liquidations 2.0 underway. And also had some recent improvements to the current liquidation system with a higher ‘beg’ on some vaults, meaning less risk of loss. And a 2x in throughput of the liquidation system. 30MM DAI → 60MM DAI in 12 hours.

Anyways this isn’t a binding poll, we could always resignal to start lerping when we get more System Surplus. And with the future growth of the DAI supply maybe it’s fine to not start burning until next year.

7 Likes

Regarding increasing the Surplus Buffer, this post provides some risk and business-related elements elaborated by @Primoz and myself. Feel free to ask more questions.

In short, we are advocating to increase the Surplus Buffer to 2.5% and 5% of Maker’s RWA (risk weighted assets), e.g. something between 70M and 80M xxM. Nevertheless, each community participant should take an informed decision. At the end of the day, it is a business decision on how much risk MKR holders want to take.

Here are some data points you could consider:

Risk-weighted assets

One way to objectivize the risk of the protocol is to use the leverage ratio. The leverage ratio is surplus buffer/assets (all loans). This is around 0.5% right now. Nevertheless, this doesn’t take into account the underlying risk of each asset type. Indeed, holding some USDC isn’t risky except on the tail events (regulatory ban of USDC on Maker Protocol). ETH-B is strictly riskier than ETH-A or ETH-C. Additionally, we currently face a lack of Keepers for Uniswap LP vaults making them risky (as we can have 0 bids).

A better way to think about risk is to use a risk-weighted approach when each asset is weighted by a risk factor. This is called Core Equity Tier (CET1) 1 ratio in the real-world.

The bank regulation set the minimum level of the leverage ratio to 3% and the CET1 ratio to 4.5% but those are not necessarily adequate to evaluate the risk of MakerDAO.

Real-world assets (MIP21 and MIP22)

Currently, both MIP21 and MIP22 are quite aggressive regarding loan impairment. Indeed, they all start with a soft liquidation that is transferred to a hard liquidation after a delay (MIP21) or after a governance decision (MIP22).

When the hard liquidation is triggered all the remaining tab (debt) is declared as sin which decreases the Surplus Buffer and might lead to MKR minting. This is not necessarily a good representation of the reality and it might well be that most of the loan will be recovered in the future (days, weeks, months, …). Nevertheless, that’s the best solution at this stage.

For this reason, it could be safer to have a Surplus Buffer at least the size of the biggest RWA vault loan.

Regulatory risks

It might well be that regulations about stablecoin will be enacted in the future.

One is MiCa (European regulation on stablecoin that apply also on USD-pegged stablecoins) which requires a 2% leverage ratio. Should this regulation apply in 2024, it might cut us from EU-based exchanges/centralized products.

If the bank regulation is taken (Basel iii), the minimum leverage ratio should be 3%. This might be the case with the STABLE Act (US-based) that requires stablecoin issuers to apply for a bank license. Again non-compliance might cut us from most US-business.

MakerDAO is currently far from being regulated under those legislations. Moreover, there are plenty of other rules that MakerDAO doesn’t match (meaning having a proper Surplus Buffer doesn’t solve everything). Nevertheless, in case such capital levels are needed, it would take significant time to reach those objectives (assuming we don’t want to dilute MKR holders).

On MKR burn lerp

If the community decides on a surplus buffer size as adequate for the community risk appetite, it doesn’t make objective sense to delay aching the goal by burning MKR. Nevertheless, the need to burn some MKR along the way might be needed for PR.

At this stage, we are generating $100M of revenues annually from lending and we could forecast around $30M of operating expenses annually (Core Units mainly). This lead to a Surplus Buffer increase of around $6M/month.

Impact of the Surplus Buffer on the MKR earnings

The Surplus Buffer is on the liability side. It is therefore not resources we can use for anything per se.

Nevertheless, assuming the demand of DAI to be constant no matter the size of the Surplus Buffer, the only impact would be the size of PSM-USDC-A. This is not an earning asset yet, but some MIPs are proposing to use those USDC in a more lucrative way.

At this stage, and even without earnings, the level of the Surplus Buffer doesn’t impact significantly the MKR return on investment. Indeed, if the Surplus Buffer is 1% of MKR market cap, the impact on earning is 1%.

Competitive landscape

Without going into much detail, the existence and the level of the Surplus Buffer is a key element of the value proposition of DAI.

USDC, PAX (incl. BUSD and HUSD) don’t have Surplus Buffer equivalents in their design. Nevertheless USDC has a surplus of assets of 0.5% of the issued USDC currently. Moreover, their investments are risk-free assets (T-Bills and FDIC-guaranteed banks deposit mainly).

DAI has a Surplus Buffer but also has more risky assets (crypto-backed loans mainly, RWA soon). This provides a much better return for MKR holders but DAI might want a share of the cake (as they take some of the risks). DAI holders also don’t get $ convertibility. This might lead to a needed increase of the DSR to compete with USDC-like stablecoin if we can’t be better on some other aspects.

Other competitors in the future can be lending platforms (CeFi like BlockFi or DeFi like Compound). They might want to issue a stablecoin to decrease their funding cost and increase profits.

Nevertheless, it should be noted that the DeFi space currently doesn’t care as shown in the USDT case and the existence of algorithmic stablecoins like FRAX.

Risk of having a big Surplus Buffer

As designed currently the surplus buffer belongs to DAI holders at the emergency shutdown of the DAI Protocol (as discussed here). The limit of the Surplus Buffer (before game theory suggests triggering an Emergency Shutdown) is hard to define and depends on the quality and liquidity of the collateral that can be retrieved. Intuitively, you can say that the Surplus Buffer shouldn’t be above a few percents of the DAI outstanding.

Solutions to this problem are either:

  • Using the mandated actors multi-sig (which require significant trust in the mandated actors) to store excess funds (above 1% Surplus Buffer)
  • Implementing Strategic Reserves (which demands SC work)
  • Finding a solution to have residual collaterals (after all DAI holders are given $1 of collateral) given to MKR holders (or mandated multisig or Strategic Reserves) after an Emergency Shutdown. Probably hard to solve.

The first two solutions are not perfect as nothing guarantees MKR holders will refund DAI holders if they get less than $1 at Emergency Shutdown.

Risk premiums and VaR

According to the latest risk premiums table, the weighted risk premium for MakerDAO’s portfolio is about 6%. This means expected losses per year would amount to about 115m on a portfolio of 1.9bn (excluding stablecoins). This metric is however based on the Liquidations 1.2. Further, the 99% VaR value from Andy’s presentation in January measured about 10% of debt exposure (again excluding stablecoins). Applying this to today’s debt levels, 99% VaR would likely amount to around 200m. Again this number doesn’t assume implementation of Liquidations 2.0 and its positive effect on expected and tail event losses.

Assuming huge improvement with Liq 2.0 we could say that the reasonable amount of Surplus Buffer would be between 50m and 100m (also depends on the community’s confidence interval to be hedged against worst case losses, in other words what % of VaR is chosen). This potential range of Surplus value would represent between 2.5% and 5% of Maker’s RWA (risk weighted assets) and would be also somehow in line with Basel recommendations for banks.

Conclusion

Assuming a middleground figure of 75m targeted Surplus Buffer and net income of 6m per month (after DAOs costs), we would need to wait until the end of 2021 to get Surplus Buffer to target levels. This also means no MKR burning in 2021 at this pace of net income.

From MKR holders perspective the issue really lies in current inability to achieve some minimum safe yield on Surplus Buffer assets (currently Surplus Buffer counterparty is non-interest yielding USDC in the PSM). Imagine a 5-10% yield that would be achieved on a 75m of Surplus Buffer.

We therefore think that from a risk perspective a Surplus Buffer of 70m to 80m would be preferred going forward with no burning or a minimum amount of burning (25%) if we are able to make yield on Surplus Buffer assets by implementing strategic reserves or similar type of treasury management concept.

11 Likes

literally could pair the DAI with USDC in a LP and earn some return on this while still being accessible from a liquidity standpoint. Not sure if this would still satisfy regulatory requirements though.

I would suggest to get working on one of the first two in the list as I have been advocating for SR pretty much since I came to Maker. I have also been advocating for a minimum of 1% SB forever which we still have not reached. To put this in perspective BT cost the protocol about 4% of total outstanding which at the time was something like 150M if I remember right. 4% of 2.7B is now 100M which if you believe the MKR price will only drop 50% would mean a dilution of about 100K (or 10%).

BTW: Regarding negative MKR event like BT I realized Maker could have done a 10%/yr bond offer and saved itself something like 5M and not diluted one MKR. This would have been paid off in about 6 months. So again I would urge research into different DSR facilities that could be used during liquidation events to backstop the system and or to offer up as longer term deposit facilities (which could pay out against and manage interest rate fluctuations against longer term loans capturing interest rate deposit loan spreads - think RWA here).

4 Likes

100% agree with this conclusion. One way to view it–Fred Ehrsam wrote this letter recently:

Ask yourself: “If crypto goes through a protracted down cycle, do I have enough cash to survive?” Cash that is easy to come by today may not be tomorrow. At Coinbase, we found ourselves running low on cash shortly after 2013-14. Luckily, we decided to fundraise before
crypto was in serious winter; had we not, we may not have survived until the 2017 spring.

This :point_up_2:t4: is an excellent read on how to Survive a Crypto Cycle. We should make all preparations by establishing parameters ASAP

2 Likes

Thanks a lot @SebVentures and @Primoz for providing all those details! I believe this gives some guidance for everybody participating in this signal request.

1 Like

hi , i am new here.

Like to give some feedback . Is it not better to do both have a buffer and burn.
Lets say the worst happens and we used the buffer and also need to use mkr to cover the loans.
And the shock of just not having enough buffer would drive mkr price down even more and we be auctioning mkr for less then what it is to buy back the dai. If we had a buffer and also burning a % of the fees we have a strong mkr also. So in a event of failure we can apply bought measures if needed.
say an 75% buffer and 25% mkr model. Having a buffer and doing nothing is putting money and losing value. I like to share this idea.

2 Likes

Welcome to the Maker Community @myh

Yes, I believe your point about having a buffer and burning a % is what @SebVentures and teh Risk Team are suggesting. I mostly speak about an 80M DAI because I tend to have a long-term view/focus, as oppose to pushing the burn MKR narrative, as soon as possible. But I am totally okay with what you suggested of 75% SB and 25% burn.

I’m more focused on getting a nice SB so we can sleep a little more at night.

6 Likes

HI, Thanks

I think an buffer with out a top limited, would be better , just keep feeding it the 75% and burn 25%.
Then adding specific buffer tops. the buffer should act as a permanented insurance fund.
I am for a buffer. Just think we need a better buffer model.

4 Likes

Ooooh–you want the hump parameter to be increased gradually–say, 30M, 40M, 50M, 60M, etc. – and once each hump parameter is met – you want to accumulate 75% in the Surplus Buffer and divert 25% to the Flapper and auction it for MKR burn. Cool. Never thought about that. Assuming that’s what your suggesting…

3 Likes

Yeah that could work. This way we have both the shield aka the buffer and the sword aka mkr burn.
And limited less buffer will give us protection and a strong mkr can be uses in the future to protect mkr value.

1 Like