The DSR/SF spread should be as tight as possible to maximize growth

In my opinion we should aim to run DSR at a rate that takes nearly all the revenue from CDP holders to facilitate the quickest healthy growth of the system overall. I don’t think it makes sense to be funnelling 50%+ of the revenue into MKR profits at this stage when it could be used for growth. In my view the DSR / SF spread should be 0 (or possibly negative if utilization is low).

We are first to market with a brand new product. We need to capture as much of the market as possible. I believe we can start pulling new types of users into the Ethereum/Maker ecosystem that aren’t interested in highly speculative gains, but instead safe returns on money that would otherwise be sitting in their checking account.

Running a loss leader is a standard business practise that makes sense in a new market like this. I believe we should press on the growth throttle as much as possible.

Please discuss.

10 Likes

This is a super interesting idea. Let me paraphrase some of my comments from chat on how this would affect risk.

Standard operating procedure for risk would be to apply a DSR / Stability fee spread to account for the collateral risk. So, let’s say a risk team evaluates a collateral and pumps out a number something like 4%. Then, if the DSR were at 2% (assuming full utilization), then the stability fee would be 6%. 2% to Dai holders and 4% to MKR holders as compensation.

At this level, the Maker Protocol is running (theoretically) at breakeven. In order to change one rate (say, the DSR), you’d need an accompanying change in the other in order to keep Dai stable. And they would move opposite each other. So, for example, if the DSR were to go up by 1 point to 3%, then the Stability Fee would need to (probably) move downwards by 1% to increase supply as well. Increasing/decreasing the spread essentially increases or decreases the risk premium, which is the profit source for MKR.

Converging the spread is essentially forgoing profits for increased growth. So, if right now we are at 2% DSR and 4% Stability Fee, we could move it to 3%/3% and the peg would still be stable, but MKR holders would likely start to lose money (its not a linear function in rates but whatever). This would be more attractive for Dai holders and Vault owners. One could even go 10% DSR and -10% Stability Fee if you really wanted to fuel growth.

So, the downside of this is that with extra losses, the MKR market is lower (ignoring projected growth increases due to increased users), and the tolerable debt ceiling would be lower. So once we start discussing what tolerable debt ceilings are (soon™), we can try to predict what type of monetary policy would be needed to achieve our “growth” goal.

DISCLAIMER: I am not in favor or against this idea, as its a business-oriented decision that should be decided by the MKR community. I can only help quantify the tradeoffs and maybe share an opinion or two down the line.

5 Likes

great discussion between @cyrus and @hexonaut . here is a simple table that highlights the relation between spread and risk premium https://docs.google.com/spreadsheets/d/1-NKGNzz7gQMEmyUvdNuhitOBn93nOFdpT9gSSR-RLAY/edit#gid=0

2 Likes

I like the idea of fast healthy growth of DAI supply by temporary lowering MKR holders’ revenue.

but we need to know how to define healthy and secure growth

at what DAI supply we should stop subsidizing DSR?

what percentage of total ethereum market cap being held in maker vaults can be considered the upper safe limit (5% ,10%?), how about other less secure collaterals?

also I remember that with MCD we should be able to give different SF to different collateral ratios to encourage safer loans!

1 Like

You want quick DAI growth?
We do that through quickly onboarding tons and tons of collateral types and setting debt levels really nice and high. That will push DAI growth much harder compared to just sitting on our hands and fiddle with the Stability Fee and the DAI Savings Rate. No point in subsidizing anything. Just onboard collateral types and let greed do the work.

4 Likes

I like this idea in general, but I have two concerns:

First, Cyrus’s point regarding risk is important, if we ignore the risk premiums then we are essentially saying that we are accepting a loss on average (because we are not adequately compensated for the risk we are taking on.) As Cyrus said, this is a business decision, but it’s important to remember that while it might look like we are break-even, or turning a small profit, that comes at the cost of failing to compensate for risk.

Second, we’re in an industry where everything is very open, and expectations are very high. There is a belief that rent seeking should be minimised as much as possible and this belief is enforced by the threat of protocol forks*. If we start up MCD with a loss leader approach, at some point we will need to end that approach. At that point, we are at risk of a backlash or a aggressive fork in response. Despite the dislike of rent seeking, I think the risks of this are less if we set the expectation early that we are always going to be taking a small profit.

*As a side note, the protocol is rent-seeking, but it needs to be otherwise the incentives don’t work. In my view this negative is balanced by the fact that MKR Holders are the lender of last resort and are on the hook for their decisions in a way that traditional bankers aren’t.

4 Likes

I sympathize with the temptation and I would only be weakly opposed to some very short term, limited growth boost. Still here’s why I think it’s a bad idea, generally speaking: it will be hard enough to maintain the DAI price while managing MKR risk. Beyond the DSR we may have to directly tune the SF for supply reasons (rather than risk reasons). That’s already one source of tension (in the “monetary trilemma” sense). I would like to refrain from adding a growth/profit function to the SF knob, let growth come the way @Planet_X described it, and let profits come from good risk management and whatever margin the landscape will allow. On top of that @LongForWisdom’s concerns about what happens when we reverse course are very much justified.

2 Likes

Folks I’m still trying to work on a proposal in relation to what I’m going to post here so to keep this short.

Where does one put growth relative to risk and how is this managed?!

Literally having a fixed SF even when utilization is nearing or at 100% of the debt ceiling is NOT really the best way to address these market issues.

I am toying with two ideas ignoring the very important implementation aspects (just looking at this from a how to manage risk, liquidity, peg, against growth).

  1. Instead of the DSR why not just take a % (that would be managed by governance votes of the incoming revenue stream and designate it to the DSR fund - call this DSR-SF-PERCENTAGE). This would give a floating DSR that is based purely on the MCD-XXX-SF fee collection as well as the DSR utilization and could always be set to under say 90%) to always give MakerDAO at LEAST 10% profit on all SF fees collected. This would then have the added bonus of adjusting the DSR based both on DSR utilization and SF fees?

  2. Make the SF’s ‘SF targets’ based on utilization curves. Simple example say the MCD-XXX-SF target is 4%. Make a MCD-XXX-SF a function based on the utilization to produce DAI (loan generation). MCD-XXX-SF(U, MCD-XXX-SF-TARGET, MCD-XXX-U-TARGET), where U=utilization, MCD-XXX-SF-TARGET = MCD-XXX-SF as currently designed, and MCD-XXX-U-TARGET=the desired MCD-XXX utilization target percentage. What the form of this function is the community can decide but it should have the effect of lowering the MCD-XXX-SF below the MCD-XXX-SF-TARGET by up to 2x if utilization is low, and increasing it above the MCD-XXX-SF-TARGET by at least 2x if utilization is high.

There are additional changes I want to propose in conjuction with the above two but I want to point out a number of important factors that come about if the above model were to be implemented.

  1. the DSR or (DSR-SF-PERCENTAGE) by definition/construction is always a rate that never causes MakerDAO to lose money as contrasted with current situation that if DSR deposit interest exceed loan income MakerDAO loses money.
  2. the DSR rate will change up or down based on both DSR and all Maker/loan SF utilization allowing MakerDAO DSR to always stay competitive with minimum of governance votes. The DSR-SF-PERCENTAGE is still a lever governance can use to modify behavior to sop up or loosen DAI liquidity based on savings behaviors.
  3. If designed properly debt ceilings should rarely if ever be hit and utilization will always be some amount above 0. This will mean there will always be available DAI liquidity to be minted if the DAI loses it’s PEG (both in bad and good markets) and will manage loan risk as an added benefit.
  4. Maximizes SF return to MakerDAO owners.
  5. Allows all fees to float to be market competitive with the least amount of governance voting. If we can design a system that allows the markets and market participants themselves to act for us to manage the system - it will.
  6. Allows for ‘good growth’
  7. Still have the MCD-XXX-SF-TARGET levers as well as an additional MCD-XXX-U-TARGET lever

My proposal only leaves a few parameters to be managed.

  1. MCD-XXX-SF-TARGET (XXX stability fee target)
  2. MCD-XXX-U-TARGET (XXX-utilization target percentage)
  3. DSR-SF-PERCENTAGE (percentage of fees that goes into the DSR payout pot)
  4. MCD-XXX-DC (debt ceiling)

May need to adjust this but

  1. MCD-XXX-SF as a function of 1,2 which should only need to be changed maybe once or twice in the lifetime of MakerDAO to optimize the following parameters
a)   least need for governance votes on 1-3
b)   maximizes utilization and growth
c)   minimizes risk
d)   optimizes DAI liquidity and PEG during bull/bear market.
e)   allows MakerDAO to be more competitive against the secondary markets.
f)   allows markets and market participants themselves to manage the DSR rate (if utilization high and DSR rate low - people will move to secondary markets for better rate and vice versa)

As to the technical design. One thing I think will need to be changed is that the DSR probably should be tokenized ala compound or DAI-HRD type functionality so the DSR-SF-PERCENTAGE payments just float into the collection account increasing the mDAI value on withdrawal and accurately accounting for interest accumulation using multiple floating SF rates.

Some great counter points here.

@LongForWisdom I hear and agree with your concerns about the perception of tempting people in then turning up the SFs. I do not support rent seeking beyond a modest amount as protocol forks are always a danger of being too greedy. What I am arguing for now is to forego the modest profits we should aim to have in a mature system to keep things growing as fast as possible in this new market.

I sympathize with the temptation and I would only be weakly opposed to some very short term, limited growth boost.

@swakya I do not believe this should be a short term plan. We should forego profits until we reach market maturity or at least somewhere decently far along the adoption S-curve.

Still here’s why I think it’s a bad idea, generally speaking: it will be hard enough to maintain the DAI price while managing MKR risk. Beyond the DSR we may have to directly tune the SF for supply reasons (rather than risk reasons). That’s already one source of tension (in the “monetary trilemma” sense). I would like to refrain from adding a growth/profit function to the SF knob, let growth come the way @Planet_X described it, and let profits come from good risk management and whatever margin the landscape will allow. On top of that @LongForWisdom’s concerns about what happens when we reverse course are very much justified.

This should be no more difficult to maintain then any other plan as we are just setting the DSR to be some constant factor away from the ETH-SF (could be 0). As @cyrus said above something like raising the DSR to 3% and lowering the (BAT/ETH)-SF to 3% makes sense to me since the fees work in opposite directions for peg stability. After that you would raise and lower those numbers together for monetary policy.

I like the idea of fast healthy growth of DAI supply by temporary lowering MKR holders’ revenue.

but we need to know how to define healthy and secure growth

at what DAI supply we should stop subsidizing DSR?

what percentage of total ethereum market cap being held in maker vaults can be considered the upper safe limit (5% ,10%?), how about other less secure collaterals?

also I remember that with MCD we should be able to give different SF to different collateral ratios to encourage safer loans!

@mmoossttaaffaa I am calling this the fastest possible healthy growth because we are maximizing both the supply and demand to the largest amount possible. In theory minimizing the SF will maximize supply growth and maximizing the DSR will maximize DAI demand growth.

So, the downside of this is that with extra losses, the MKR market is lower (ignoring projected growth increases due to increased users), and the tolerable debt ceiling would be lower. So once we start discussing what tolerable debt ceilings are (soon™), we can try to predict what type of monetary policy would be needed to achieve our “growth” goal.

@cyrus As we discussed in the chat, I don’t think taking losses upfront will necessarily lower the MKR market cap as the market cap is not a measure of the current MKR burn, but both current and projected MKR burn. If we are making smart decisions to grow the user base then the MKR market cap should increase even with short/medium term losses.

The debt ceiling is probably the main piece that I am uncertain about because there become non-trivial risks if we grow the supply too fast. I’m looking forward to hearing more about where the limits of that are.

2 Likes

To clarify, there are 3 values that in principle may need to be set independently, but there are only 2 degrees of freedom. Given a DSR (which modulates demand) and a risk spread (SF-DSR), the SF is determined. Given a SF (which modulates supply) and a risk spread, the DSR is determined. In practice, as you & cyrus mentioned, it will often be possible to move DSR/SF in opposite directions to maintain the price, but generally speaking the forces may be asymmetrical and we’d then have to favor either the peg or risk management. Hence my point about adding yet another possible function (profit/growth) to those knobs.

1 Like

You are correct. In your example I am suggesting modulating the DSR and setting the risk spread to 0 or near 0 which would make the DSR and SF the same number.

For week-to-week monetary policy we only need to adjust the DSR and the SF falls out of that as you said.

I agree that having a minimal average spread is critical when DSR utilization is low (Not 0, as that might create weird incentives to max out debt ceiling).

Obviously it is critical that in the long term, the average spread needs to follow (and be higher than) the average loss from insolvent liquidations.

But in the short run there are more important factors to consider. One being the DSR utilization being very low, which by itself should always trigger a minimal spread until the utilization hopefully maxes out close to 100%. As long as DSR utilization is low attempts at risk modeling the spread are futile, since the actual spread will be mainly determined by the utilization.

But beyond the DSR utilization issue, theres also the positive effect a low spread will have on the MCD migration, which will bring in much greater fees than a high spread would in the short term.

More importantly theres the fundamental dynamic of MKR liquidity and price following growth of MakerDAOs ecosystem rather than MKR burned directly, so focusing on growing the ecosystem will actually provide greater security to Dai holders than trying to extract value in the short run.

And finally of course the fact that a lower spread, but on a greater supply, ultimately results in high fees. Either in the short run, or in the long run as the spread is slightly increased again to properly be set above whats necessary to cover insolvent liquidations.

3 Likes

I agree with the rationale to support policies that encourage dai growth over immediate mkr earnings (burns). As a mkr holder, I would support operating the protocol at near breakeven (very little mkr burn) for the next year to increase dai growth rate. In the end, the mkr peg ratio (mkr price to earnings / mkr earnings growth rate) will benefit and mkr holders will eventually see the mkr value appreciate due to the network effect of mass adoption of Dai. I do agree with @LongForWisdom on the risk of backlash when rates are “normalized”. For this reason, I believe the community should signal that mkr holders are “subsidizing” these attractive rates to bootstrap adoption of this new technology.

1 Like

I get the idea, however, given the lower borrowing costs compared to the analog world (and the certainty around being able to borrow), I question how needed this “really” is… It really should be at a level above MKR… (the equivalent of a bank given an introductory loan)…

That said, fundamentally MKR as a quasi repo line should want to insure against a value at risk premium (the risk premium)… the questions is when is enough DAI enough? another 100mm or 500mm? etc…

the ability to grow the DAI space is purely limited now to the community’s imagination (and comfort around the risk parameters)… for my sense of things I would prefer to have a correctly priced RP for collateral… and move the DSR as needed to digest the excess DAI…

we will need to iterate on this DAI “mop-up” process… the concern is that when we mint too much and the DSR rises, it will price some assets “out of the market”…

by conveying all of the value to the DSR, it could make sense if it was a well-communicated short-term event that ALSO INCLUDED the iterative decrease of the SAI debt ceiling to zero.

that said without the debt ceiling component, it would not make sense to bring on risky assets… really only those that are not-risky… (e.g. treasuries)… however, those are exceptionally rate sensitive…

the answer here is that there is no good answer… the specific challenge here is one where when we have 50 collateral types this will no longer be an issue… or at least it will be one where the excess DAI minting will be light wave compared to the aggregate… for now, there really is no good way to do this part without getting bruises along the way… if we introduce risky collateral (e.g. one where the risk premium can be high) we need that RP to cover for MKR loses… (not theoretical loses… but dilution of MKR). that said, when we need to increase the DSR they will be less rate sensitive to increases… as it is already risky… vs… bringing on less risky assets that wont have near the same quantity of losses. but will more rate sensitive…

the real recommendation is to launch a mini-basket of them with a minimal debt ceiling for all of them (with correctly non-subsidized RPs for each)… and then raise the DCs over time… and let the DSR do what it is going to do …

Wouldn’t an immediate goal be to incentivise Compound savings rate users to migrate over to MCD’s DSR? If that’s the case, we need to probably get the DSR up to 6% and then adjust the SF to keep the Dollar peg.

Interesting conversation. It seems that the protocol should run as close to breakeven as possible to maximize revenue growth. This is similar to how start ups and young companies operate. Once growth is achieved, then rent-seeking and profitability will be possible. The threat of fork, once growth is achieved, is minimal assuming Dai has reached escape velocity and built network effects.

4 Likes

Here we are with DAI leaking below $1. Isn’t it time to start raising the DSR? Only 15.24% of DAI receives the DSR, so we should not hesitate to temporarily raise the DSR above the weighted average SF.

2 Likes

The Maker foundation is bringing on hundreds of new integration partners for DAI. They will need DAI market cap to be in the billions soon. The only way to attract that much collateral is to offer the smallest risk premium we (MKRH) can while remaining profitable. This should be accompanied with the highest possible DSR to attract as many DAI money market users as possible. This is necessary to bring DAI liquidity way up and support the massive influx of DAI integration partners which is bound to happen.

2 Likes

If we decide to forgo short term profits for increased growth, we should consider not only subsidizing the lower spread, but also “labor”. It seems to me that one of the bottlenecks is currently risk team - and this is not critique. Any additions to the risk team would be very beneficial to the project.

Ofc, for me, as a community member, it’s also hard to evaluate Foundation strategy and their spending of resources. I almost certainly don’t have enough information/skills to form opinion on their actions.

Some more thoughts/clarifications on this:

  1. When I say the DSR/SF spread should be as tight as possible I am talking about the global SF minus the DSR. This does not include the additional per-collateral SF for risk premium. In this sense I really do think that the global SF minus the DSR should be 0, and that we should use the per-collateral SFs to capture the risk premiums.
  2. Here is how I am thinking of the system parameters overall.

Global Stability Fee - Used for week to week monetary policy. This should be viewed as the MCD equivalent of the SCD SF adjustment.
Global SF / DSR Spread (Global SF minus DSR rate) - Should be relatively constant and reflect how much of a profit margin we want to run. Setting this to 0 will tune the system to maximum growth whereas a higher number will increase profit margin.
Per-vault type SF - This should be mapped directly to the risk premium and we should aim to run breakeven numbers here as we can capture profit in the global SF.

In this mental model we can separate out the monetary lever, profit/growth lever and the risk premium levers into the respective system values above.

As an example)

Let’s say the risk premium for ETH-A is 2% and after making the choice for maximum growth (DSR spread of 0) the supply/demand equilibrium of the market is calling for a 4% SF. Then the cost of opening an ETH-A vault is (4% global SF + 2% risk premium) = 6% in fees. The DSR is set to 4%.