An update on DSR and initial values

Hey all. A blog was just posted on DSR and a little bit about its mechanics. https://blog.makerdao.com/an-update-on-the-dai-savings-rate-in-multi-collateral-dai/

There’s also a portion relevant to risk/governance that I want to highlight:

"
Governance and Risk

The Dai Savings Rate is expected to play a central role in MakerDAO governance, primarily as a tool to help maintain Dai’s peg to the US Dollar. In Single-Collateral Dai thus far, the lever used to maintain the Dai peg has been the Stability Fee. Changes to the Stability Fee, which are determined by Maker governance, incentivize CDP owners to borrow or repay Dai, consequently pushing the Dai price up or down until it equilibrates. By targeting Dai generation, this parameter impacts the supply side of Dai.

The DSR, however, offers an alternative tool for Maker governance. Because the DSR will influence Dai holders’ behavior (as opposed to the behavior of CDP owners), Maker governance can affect the demand side of Dai. End users will be incentivized to buy or sell Dai in the open marketplace based on how much additional Dai they may earn through the DSR. If the Dai price is too weak, the DSR may be raised, motivating users to acquire more Dai, and thereby pushing the Dai price up , closer to the peg. If the Dai price is too strong, it is expected that the DSR will be lowered to alleviate excess Dai demand, moving Dai prices down closer to the peg. Maker governance should be cognizant of these dynamics as they adjust the Dai Savings Rate.

Who Pays for the Dai Savings Rate?

Functionally, a design goal of the Dai Savings Rate is to ensure that excess Dai is not printed “out of thin air” (i.e., make certain that Dai minted for the DSR remains adequately backed by collateral). The implication here is that the DSR is a cost burden on some ecosystem actor. So, Maker governance should take into consideration who ultimately carries that burden and absorbs the cost.

At the accounting level, Dai awarded through the DSR is recorded in the same line item as the one used to record Stability Fees collected. In other words, the Dai created for the DSR is recorded as an offsetting adjustment of Stability Fees collected for the Surplus Auction. If the total amount of Stability Fees collected in Dai does not cover the total amount of Dai minted for the DSR, the difference is recorded as bad debt, and MKR is printed to cover the cost.

While MKR holders bear the ultimate cost of the DSR, the expectation has always been that CDP owners would effectively pay for it through a commensurate increase in the Stability Fee. Conceptually, the Stability Fee should be comprised of two components: 1) a collateral-specific risk premium that is a value transfer from CDP owners to MKR holders, and 2) a DSR adjustment that is a value transfer from CDP owners to Dai holders. Essentially, CDP owners compensate the two distinct ecosystem actors: MKR holders for the risk of collateral, and Dai holders for the risk of Dai instability.

There are additional complexities to consider with respect to collateral types. Specifically, global increases in Stability Fees due to the DSR may render certain collateral types as infeasible for inclusion. For example, if the DSR were set at 5% (thus causing a base Stability Fee of 5% on all collateral types), it would no longer make financial sense to include certain collateral assets that require a low cost of capital. A mortgage, for example, might have difficulty remaining competitive at >5%.

Taking these determinations to their natural conclusion, Maker governance will need to consider the following questions:

  • If a high DSR is required due to overexposure of a volatile asset, such as Ether, how should the other collateral types be managed? Does it make sense to penalize collateral X due to collateral Y’s popularity?
  • Should Maker governance assess collateral-specific DSR adjustments proportional to the asset’s effect on the Dai price? If so, can governance adequately calculate what that specific adjustment should be?
  • Assuming there is a large pool of assets, can governance accurately assess which collateral type is contributing the most toward a weakening/strengthening Dai price?
  • Is Maker governance prepared to efficiently make what could be a myriad of Stability Fee changes?

These are difficult questions, even in the presence of a long history of data analytics. As a result, the Interim Risk Team will propose to Maker governance that the full amount of the DSR adjustment not be passed globally onto all collateral types until a robust solution is devised and put into place. Instead, the portion of the DSR passed on to the Stability Fee should allow for inclusion of all collateral partners into the Dai Credit System. The shortfall will end up being a cost borne by MKR holders at the benefit of a flourishing collateral ecosystem.

To illustrate, consider a collateral pool of three assets with a cost of capital of 3%, 5%, and 10%. If the DSR adjustment required by the market is 5%, and that cost were to be passed on to the Stability Fee, then the first collateral would be priced out of the protocol. Therefore, Maker governance should consider passing only a 3% cost into the Stability Fee, and absorbing the other 2%.

Comparison to Other DeFi Platforms

As the Dai Savings Rate presents another savings vehicle for Dai holders in the broader DeFi ecosystem, it is important to understand its risk profile relative to other DeFi platforms. The key insight for understanding DSR is that it presents exactly the same risk as holding regular Dai. Unlike centralized savings products whereby a counterparty lends deposits in order to generate interest, DSR doesn’t have counterparty risks, as the Dai generated by it is programmatically minted, and it is guaranteed by the same backstops as regular Dai, namely MKR dilution. This makes DSR definitionally the least risky savings vehicle for Dai. Other savings vehicles in the DeFi ecosystem offering higher rates do so only at the cost of higher risk.

DSR Starting Values

Prior to the launch of MCD, the Maker community will have an opportunity to select (through a governance poll) a starting value for the Dai Savings Rate. Options will be chosen by the Interim Governance Facilitator and derived from a signal-gathering discussion in the MakerDAO Forum. While we expect plenty of debate on what the rate should be or which benchmark (if any) it should track, we suggest a simpler trial-and-error approach. After the initial parameter has been selected via a governance poll, the community should prepare for a rapid, iterative process until the DSR reaches its natural equilibrium, as evidenced by a stable Dai price.

Eventually, the DSR can be determined based on empirical data. As the models become more mature, Maker governance can start to manage the balancing effect necessary to attribute global or collateral-specific DSR values.
"

I’m hoping that we can spark up some good discussion on where we might think DSR should be initialized.

9 Likes

A few very quick takes.

First: congratulations. This is well-written and strikes the right balance between clarity and conciseness. You also do a good job separating what is set in stone, what is to be decided and what your own suggestions are.

At the very start, the DSR should be 0. It will probably increase very quickly to some nonzero value, but it is simpler to split the technical transition to MCD from the monetary transition to a with-DSR system. For now I have no opinion on what the DSR should be after that and I’m very interested in hearing everyone’s thoughts on the topic.

The suggestion to fund any DSR-SF leftover with MKR dilution is interesting. It’s about relaxing a hard DSR ceiling that I believe was implicit in previous discussions, and deserves some thought. It is tempting, but also unsustainable (ignoring growth) and so just kicking the can down the road. I don’t have a strong opinion right now.

2 Likes

Excellent that we have come so far!
With regards to DSR my feelings are that we should try to avoid doing what we did with the stability fee (SF) . SF started at 0% and then the can was just kicked down the road until we had to do a lot to bring the peg back to 1:1. How about this time we start it at a semidecent level and then use smaller iterations?

My gut feeling: US bank account interest rate are 1.5-1.8%. EU (German) rates are 0.1-0.3%. So if we say 2% to match US bank and then 2% to cover for the crypto inconvenience we get 4%. This could be a start and then we can iterate from there. Alternatively we could match Compound, but I feel that will not scale.

1 Like

I agree with @swakya that on the begining DSR should be 0 making it somehow similar to SCD and by that reducing chaos of signals caused by transition from SCD to MCD.

Important question is, how we will change it from there. Are there any guidelines how to proceed and decide if DSR should be increased or rather SF ?

If i remember correctly

SF=DSR+RP

where RP is Risk premium, do we have any way of assesing what is RP for certain collaterals type ?

I think numbers based on US/Germany interest rates are no use for us by any means

I think much more informative is Compound DAI Deposit rate.

What are arguments for it being lower than DSR and what are arguments for it being higher?

My arguments for DSR being lower is :

  1. There is no bank run risk, while Compound operates on fractional reserve
  2. Maker is more decentralized and because of that there is smaller risk of defraudation

My arguments for DSR being higher:

  1. Compound code is battle tested while MCD will not be (that is temporary state)
  2. Compound deposit has additional use as a collateral for loans while MCD as far as I’m aware has no such use.

On the Blog Post

First of all, I just want to second @swakya and agree that this is a fantastic write up, it’s great to start to see more details on how these systems will work in practice.

This was understanding, it’s great to see it confirmed.

Likewise, it’s great to see this recognised, for those that are unaware, this has previously been discussed at some length in this thread: Is anything wrong with the DSR and Stability Rate calculations as planned for MCD?


On open questions

Yes and no. This is complicated. On the one hand if we let the market forces work uninterrupted then we end up making more money overall at the cost of a less varied pool of collateral (in the worst case, the collateral pool is entirely made up of the most volatile asset as everything else has been priced out.) On the other hand, if we allow higher-risk collateral to subsidize lower risk collateral, then we cut into the systems short term income, with the hope of more positive long-term outcomes.

It seems like it will be very difficult to figure out which collateral package is having the largest effect on the dai price under normal circumstances (obviously if we see a spike then the answer is more clear cut.) Personally I am still in the camp that we should not make DSR specific adjustments to individual collateral packages, and we should instead apply the DSR modifier proportionally to each collateral package based on its risk premium (this would require a smart contract change to do automatically, but we could do this manually to test its effectiveness.)

This was kind of included in the question above. In non-extreme circumstances. No.

No, we really shouldn’t try to do this on a package by package basis. I would say that in the short-medium term it might be feasible, but the more collateral packages we add, the worse this is going to get. Long term this is completely unsustainable.


On Starting Values

Sounds good, how soon should this start? Is @rich.brown going to be managing the signalling process/discussion?

Agree, there is no way we’re getting this right first time, iteration will definitely be needed.

I agree with @swakya that we should start the DSR at 0. There is no reason to complicate the technical transition to MCD by having the DSR set to a non-zero value on launch. To be honest, I feel the same way about the new collateral types. I’m a strong proponent of the following process:

  1. Move SCD to MCD with only Eth and SCDai as collateral, ensure that the system is working and stable and that the migration doesn’t cause unforeseen problems. (We don’t need to wait until SCD globally settles, but we should wait at least a month, if not two.)
  2. Turn on DSR in MCD, and perform the iterative process to ensure a stable Dai.
  3. Bring in the new collateral assets one at a time in order of priority.

Just to clarify, this part isn’t a suggestion. If the system is net-negative that means that Dai is not fully supported by collateral, the only option here is to print MKR to recapitalise the system. The issue is more a question than a suggestion: ‘What do we do if we need to increase the DSR above a collateral asset’s Stability Rate?’ My view is that we can increase DSR over the Stability Rate in theory but in practice we need to be extremely careful how, when and how long we do this.

I think that we’re probably looking at between 0-3%. We should not attempt to match the Compound Rate. @Adam_Skrodzki makes some good points on the pros and cons here. Specifically this:

We may well see a reluctance to lock up Dai for the first six to twelve months of the system’s operation. We will have to factor this in to our thoughts on the DSR.

In general I still believe that the DSR should be lower than Compound, the key here is the lack of counter party risk. Though it is yet to happen, at some point Compound will reach 100% utilisation at a time where multiple actors want to withdraw collateral. I believe once this happens the benefits of the DSR will be better appreciated.

3 Likes

I agree. But I’m all ears for arguments for a more aggressive timeline.

Thanks, you’re right to clarify! I meant that the suggestion was about relaxing a norm – unlike with dilution due to sudden price falls, it is possible to tune the parameters so that it never happens (right?). I did take that as an implicit design constraint in earlier discussions, but I’m open to the idea in theory – basically what you said:

Also:

A valid concern! But
a) We have more experience and would react faster. I’d even argue the biggest mistake was: not reacting aggressively enough when it became clear the SF had to go up. Keeping the SF at 0% earlier wasn’t really bad in comparison.
b) We could have a target DSR in mind, let’s say somewhere between 0% and 4%, then have a clear plan to reach it incrementally. Such a process would assuage your concern about complacency, assuage my and others’ concern about starting with a nonzero DSR, and open the door to trial-and-error iteration if something comes up.

2 Likes

@LongForWisdom
I have question to that.

Do You agree with my other argument that DSR collateral cannot be used (in opposition to Compound) to take a load

Therefore Compound deposit has some additional utility in comparition to the DSR, or You believe that utility is so minor that its effect can be ignored.

Yes. I think I understand you correctly, and I agree.

I’m not sure I would classify this as utility for savers. It’s an essential part of Compound’s system that the DCS doesn’t share. Having a counterparty lets you potentially get higher saving rates on your Dai, but the trade-off is more risk. The DCS is a lower-risk lower-reward system.

I think there will be people willing to use both systems. I also think that the Compound rates will reduce a bunch when the Stability Fee (presumably) comes down with MCD. I expect them to end up only 1-2% above the DSR.

DSR is just a tool we should use it to achieve certain goals. so we should identify the goals that we are trying to achieve then we can determine how to use the DSR as a tool to help us achieve these goals.

in SCD we have one goal (keeping the peg) and we have just one tool (the SF).
so governance is relatively very easy with SCD.

but as we have experienced the organic demand for DAI was relatively low and that’s why the governance was forced to raise the SF to very high levels to keep the supply in balance with the low demand.

DSR is the tool that will help us to increase the DAI demand.
and since the bottleneck for increasing the DAI supply right now is the low demand this means that DSR is designed exactly to fix this bottleneck.

so with MCD we will have 2 main goals: 1- increasing DAI adoption (supply) and of course 2- keeping the peg… also we have 2 tools SF + DSR.

This means that the governance should identify the targeted Dai supply… after that we use the tools (SF + DSR) to achieve our goal without losing the peg.

The Dai Savings Rate is expected to play a central role in MakerDAO governance, primarily as a tool to help maintain Dai’s peg to the US Dollar.

I don’t agree with this… I think DSR is mainly the tool to increase Dai adoption. we are already successful in keeping the peg with the SF.

although that doesn’t mean that DSR won’t be helpful with the peg. but we should mainly use it as a tool to achieve the targeted supply.

3 Likes

I’m going to try to zoom out and then back in to give my take on the initial savings rate we should set. My understanding is that the Makerdao project primary goal is to become the most efficient and stable financial system. This can be accomplished by providing better rates (both saving and borrowing) compared to other financial alternatives while maintaining stability in the system.

Other important goals are:

  1. decentralization
  2. permissionless
  3. trustless (or lowest counter party risk possible)
  4. censorship resistant
  5. transparency
  6. Did I miss any?

To achieve the project’s goals, the objectives are:

  1. Increase Dai supply

While:

  1. Maintaining US dollar peg
  2. Not creating any undesirable risks (for example, systematic risk to any collateral type)

This can be accomplished if we can provide the highest value for both Dai suppliers and borrowers. What this means to me is we should aim to initially start the DSR at 0%(prove stability in system), and then raise in 0.25% increments until we provide a higher savings rate than traditional large us banks. Currently, major us banks offer interest rate on deposits up to 2%, so I’d like to see if we can offer something more. Say 3%. I’d recommend the stability rate for each collateral type be adjusted during this process to keep the peg. Determining the best stability fee for each collateral type will be a huge challenge, but with the goal of maximizing Dai supply while maintaining the peg at an acceptable risk level, this will be a very worthwhile endeavor. Thoughts?

I think it is critical to start DSR at a very attractive rate. This is because of the somewhat risky nature of the SCD to MCD upgrade. If the upgrade isn’t as smooth and as quick as possible, it could create a lot of uncertainty about the future of either SCD or MCD as users will be unsure about which asset to pick if they are only thinking about short term utility.

If there’s no positive DSR, then a SCD holder isn’t compelled to upgrade in the short run and would only do so when a majority of users have already upgraded because the only direct benefit of upgrading would be greater liquidity of MCD (but that only happens once a critical mass has already upgraded, so it’s a chicken and egg problem). This problem is further exacerbated by the fact that many users will have SCD deposited into Compound where they’ll be earning a steady and comfortable return, and it’ll be inconvenient for them to have to upgrade so they’ll need a strong and compelling reason.

The problem also isn’t just that any individual holder doesn’t feel compelled to upgrade, but also that they know others in the ecosystem aren’t either. So it could create uncertainty around the liquidity of MCD, as every user would want others to go first in order to make sure they don’t jump into MCD only to be stuck with an illiquid asset, and everyone would be aware of the inherent chicken and egg situation this creates. By setting an attractive market and risk optimized DSR we deal with this issue by creating an immediate and very compelling reason for every SCD holder to upgrade on day 1 of MCD, and we should be able to leapfrog this issue entirely.

One thing that I think everybody here got right is that whatever the starting value of the DSR is, it is totally arbitrary and will immediately start changing in response to the peg, until it finds a proper equilibrium range after the SCD -> MCD migration is largely done, where it will then move around in smaller increments.

I think a better way to put it is that the spread between DSR and average stability fees help with balanced adoption (on both demand and supply side) at the cost of risk to the system. Trying to just artificially increase DSR without also decreasing SF (taking the consequences on the spread and thus systemic risk into account) just breaks the peg upwards by creating a surge in Dai demand without a corresponding increase in supply.

DSR and stability fee adjustments themselves happen reactively in response to the peg, once the spread has been decided by governance on the basis of risk tolerance and collateral portfolio quality.

If the starting value of DSR will be the main tool that will convince SCDAI holders to switch to MCDAI then would it be possible to set the DSR to a very high level for a relatively short period in order to grease the transition? We could even set the DSR substantially higher than the SF for a limited time.

Also - how do we handle Compound during this period? On one hand we want a clean quick switch to MCD, but on the other hand we do not want to cause Compound to run into liquidity problems which could easily happen as DAI makes up roughly 25% of their total collateral.

Would it be possible to have a “switch-week” where the DSR is set to a really juicy level in order to ensure a quick switch? Or even a “switch-day” with even higher DSR?

IMO that wouldn’t go too well since it could have been seen as a bait and switch similar to very low SF in the early days of SCD which then skyrocketed to 20%. I think the best is just try to give MKR holders a range of options and then let them pick whatever they think is a reasonable level to start with given the spread that the risk models require.

Since the starting value also won’t matter for too long as it will quickly change, it’s more of an important signal to give in advance of MCD launch so everybody knows its worth it to switch over since they can expect something similar to that. Personally I think 5% is a good value, but it’s arbitrary so really I just want to see what MKR holders support and then have them start modifying the value post launch through the same process as what we now use to adjust only the SF.

2 Likes

I admit I hadn’t considered this aspect of the problem. As a gut feeling I wasn’t worrying about this because of the whole ‘new is always better’ paradigm. People are pretty incentivized by society in general to ‘upgrade’ without actually considering whether it is the rational action. That said, it was certainly be good to make sure there aren’t lingering feelings of regret over the decision.

I’m afraid I’m going to be ‘that guy’ and ask more specifically what you feel would be a ‘very attractive’ rate. I guess to clarify the question further, very attractive in comparison to what? Compound? Central Bank Rates? Retail Investor Rates? Your answer implies Compound Rates, which would mean like 10%?

Can this not be resolved on the other side of the equation? If the SF for Eth is lower in MCD than in SCD then it will incentivse CDP holders and Compound Borrowers to switch to MCD. This will decrease the utilization rate in Compound and further incentivise SCDai holders to switch to MCDai to earn better rates. I would expect to see them reach an equilibrium level which is then resolved by the DSR increasing over the medium term.

On a another note, do we know how Compound plan to deal with the migration? Will they be disabling borrowing on SCD?

I’m yet not convinced this will be an issue, but I can see how it could be. Are there other solutions to this? Again, it seems like this can be better resolved on the other side of the equation. What happens if we reduce the debt ceiling to zero on SCD during the migration period? That would force borrowers into MCD and ensure a gradual migration.

Yep, we really need to avoid that happening again. I’m not sure that we can both avoid that and signal 5% (or whatever) in advance. The responsible thing to do would be to announce 0% at launch.

Any situation where we incentivise users to switch with a high/medium rate when we are fairly sure the rate is probably going to end up lower is a bait and switch, isn’t it? Only the degree changes.

2 Likes

I am very curious to hear people’s thoughts on this specific point. Is it unfair to users if we impose a 0 debt ceiling beginning 1-2 weeks after the launch of MCD?

My sense is that it shouldn’t be an issue, since users will be able to use the migration tool to move their CDPs from SCD to MCD directly, and same for converting Sai(current Dai) to Dai (MCDai).

I’m also not convinced that this would be enough of an issue to need to set a competitive rate from the start. With the migration tools and the use of the 0 debt ceiling it should be straightforward enough imo.

I agree with LFW that:

Any situation where we incentivise users to switch with a high/medium rate when we are fairly sure the rate is probably going to end up lower is a bait and switch, isn’t it? Only the degree changes.

1 Like

Thinking more about this, I’d like for the DSR to start at 0%, and step up in 1% increments every two weeks until we get to an industry leading rate of 12%. We can monitor the amount of Dai being locked up utilizing the DSR at each interest rate and learn through the process.

Hello everyone. A rather new (in past months) MKR holder and hoping to contribute here.

I have a few comments regarding MCD, SCD migration.

Maker markets MakerDAO have control over.

MCD SF, SCD SF and DSR rate.

What we have are secondary markets in USDC and DAI the two biggest players Compound and DYDX for seconary markets and would be the only ones I would be concerned about right now.

I have DAI predominantly in DYDX earning interest, and other assets in Compound earning interest.

I think there are a number of issues.

  1. Is depositing ETH and borrowing SCDAI going to be turned off when the MCD opens?

  2. I think generally the difference in the SF on the MCD and SCD could be used to drive movement of CDPs but then one has to consider both the MCDAI/SCDAI pegs. Keeping all the pegs while driving the transition is going to be difficult. I have some issues with how the PEG is being maintained with SF on the SCD currently but will discuss these elsewhere.

  3. The DSR rate. We can start this at zero, but I think the DSR should at least be close to what we are seeing on the secondary markets otherwise if the DSR rate is lower than the secondary markets there will not be incentive for depositors like me to make the switch. There are other considerations regarding the DSR which is the composition of the backing capital is (how much BAT vs. ETH). I think it will be mostly ETH in terms of fiat value so in terms of risk profile will match current SCD in terms of risk profiles. My point here is that people don’t seem to care much about risk and are really looking for return but from the data I have the secondary markets are tracking the Maker SF pretty well. If you have noticed the USDC borrow and supply rates pretty much have come in line with SCDAI now. Personally I would advocate no more than 100bps spread between the MCD SF and the DSR rate.

My question is whether MakerDAO has the capacity to have the DSR above the SF on the MCD or not as this would be a nice lever to encourage SCDAI holders like me to switch.

Ideally we’d like to drive the migration from both the SF sides along with the DSR. But one has to keep in mind the PEGs on both of these which is lately driven by liquidity (notice we hit $100M SCDAI) yet the PEG is still > 1 in many places. Literally a ~20% change in supply over a change in the SF from 16.5 to what 4.5% now? This is speaking volumes that the peg is related to liquidity vs. rate. I think the USDC markets also drive the other side of this. So I think when the MCD is opened the SCD should be kept capped at $100M and start decreasing as SCDAI is freed up. If there is no way to deposit or borrow this cap is irrelevant as the lack of ability to open new CDPs on SCD effectively acts as a cap and one way - out direction. (Like a financial diode)

Keep in mind we could also drive this transition against the USDC rates (both borrow and supply) since if the DSR rate is low this might be a reason for people like me to free up SCDAI and move to USDC. But if DAI deposit rates are higher elsewhere and higher than USDC there will be no reason for people to switch.

The idea of a ‘one’ time reward if SCDAI holders switch I think should be considered, or a floating rate based on market conditions and PEGs). Meaning put up a certain amount of MCDAI (say 1.01 MCDAI exchange for 1.0 SCDAI) might be an interesting way to encourage the switch. BTW doing this offers another ‘lever’ to manage the PEGs during this transition independant of MCD-SF and SCD-SF and DSR.