Different Approach to Rate Setting

I think we should be favouring automation of processes over having technical implementations restrict the governance constructs. The Base Rate may be one of the first 1:N constructs, but it definitely won’t be the last.

4 Likes

I think we can mitigate some USDC risks by equally setting rates for GUSD, PAX, and TUSD. I hope DeFi doesn’t become too complacent on one centralized stablecoin. Onboarding baskets will help mitigate this as well (yUSD, dUSD, mUSD,USD++). I hope once the other stablecoins are added they’ll have a WBTC effect where usage shoots up.

2 Likes

We can automate the generation, but at some point and with enough collateral types we’re going to run out of block space to perform the edits. We’re already considering removing the drip function calls to save some gas in the executive, as last week’s cast() cost 0.77 ETH and took nearly 3mm gas to complete.

https://etherscan.io/tx/0x9125c2c8795a0a872386e1ceda091c603db94f737a4557e784a96352086bd985

As we continue to add collateral types we’re going to approach a situation where we start regularly pushing against block size limits. We can probably find ways to engineer around that, but I do urge governance to not discount the technical implications of their decisions.

7 Likes

Why not trying negative interest rates first? I think need to try new things and step ahead not backwards

1 Like

On the business side of things I’d love to see low volatility collateral uncorrelated to crypto (ie PAXG) and fixed term (or less variable) rates. Ie I would like to borrow against gold for a longer term business use case.

The variability of rate setting policy combined with the risk to borrowing against crypto given volatility renders the credit unworkable.

2 Likes

Right now an increase to -5.5% base rate is winning. The highest we can go without changing the governance process this week is a -2% base rate, which would be a mild (read: negligible) increase in stability fees for a few collateral types

Not really, that’s roughly 2m

If we are to continue to Signal Request and there is certain agreement with this strategy we need to make a few things clear. This strategy is basically stating to 1) Determine rates more on market/competitive landscape basis 2) Make less frequent changes to rates by keeping Base rate stable and Risk Premium used only to assess risk compensation 3) Defend the peg through lower USDC liquidation ratio.

Few things to consider to make sure what this means in practice.

  • Governance no longer makes weekly changes to Base Rate or Risk Premiums but instead manages each Vault SF with a lower frequency

Currently if governance wants to change rates only for particular collateral it changes its risk premium. However risk premium doesn’t necessarily move 1:1 with the SF that we to use to chargeto Vaults. For instance if we see certain Vaults mostly utilized by yield farmers and we can apply higher rates, this doesn’t mean that Vault is more risky and we need to increase risk premium. In this case it’s even less riskier, because those vault users hold DAI in some farming venue and can pull it out immediately to avoid liquidations.

  • Risk premiums are still used, but only to assess how well we are compensating risk

Risk premiums instead becomes a tool that is used solely as a gut check for SF that governance sets for particular Vault. There isn’t a direct 1:1 relationship between risk premium and SF anymore, although it is still correlated in practice. This gives us some feedback about the SF that is charged and what is the actual surplus for MakerDAO after the expected losses (risk premiums) are covered.

  • Lower LR for USDC-A is used to defend the peg

This one is tricky for various reasons. First of all, if we are to set SF too high, we know the peg will deviate and by keeping low LR for USDC and managing DAI price cap, we are inevitably increasing exposure to USDC. Here governance needs to make sure that a) high SFs don’t cause users to leave and instead of having higher exposure to crypto assets we end up in situation with USDC being the leading exposure and b) we define what USDC LR (equal to DAI cap price) is tolerable because the difference between 101% or 102% might be huge. If one wants to put a DAI price ceiling to $1.01 instead of $1.02 this might mean tens of millions of more DAI selling into Curve currently. So it all comes down to what exposure of USDC governance is willing to accept versus having lower DAI price cap or higher SF.

As for the Signal Request itself, there is one already pending that addresses a lower USDC liquidation ratio. But we need to make sure that governance understands the whole concept. The way I see next steps:

  • Signal request to support the strategy outlined
  • Prepare a plan or working group to determine competitive rates for Vaults that are compensating risk premiums
  • Make a poll on new rates (SF)
  • Find a solution regarding implementation of Base rate and cadence of rate changes going forward
5 Likes

Complete opacity of Maker Foundation financials may have justifications, but it also makes me incredibly uncomfortable.

IMO even if MF is not able to contribute any more financial resources to the project, the protocol could easily raise millions with a 1% dilution. It feels sort of selfish to raise rates when MKR holders could muster the resources themselves.

2 Likes

@BrianMcMakerDAO I must admit this one was new to me. How many parameters do you think we can change at any one time before the block size becomes a problem?

2 Likes

Agree, however it feels a little harsh to call selfish the idea that mkr hodlers don’t want (we don’t know) to dilute their capital after subsidizing 0% rates for quite some time now.

1 Like

Looking at the historical volume weighted adjusted prices, the additional ~50 million Dai generated over the past 7 days has contributed to Dai drawing closer to the PEG by 0.6%, but we are still 0.5% away from 1. The additional Dai supply is working - let’s stay the course with 0% stability fees!
Volume Weighted Average Price (90 days): 1.012
30 day: 1.012
14 day: 1.011
7 day: 1.011
3 day: 1.011
1 day: 1.005

if I understand correctly what you’re saying: The peg is the peg. rate is the rate. since rates have a weak effect on the peg, we might as well decouple the 2.

That would make sense to me! more fundamentally, I believe rate should reflect credit risk and MKR need to embrace itself as something as more of a commercial bank than a central bank (arguably it already is b/c DAI is more analogous to “money as we know it” as in bank deposit than “base money” as in reserves at the fed)

2 Likes

Hi @mspacey4415 and welcome to the community!

The rate does affect the peg, the community knows this as we needed to raise rates to almost 20% at one time in order to save Dai from becoming undervalued (yes - that was a long time ago).

Now we have the opposite situation where Dai, due to high demand, is overvalued compared to the peg. As a result we have reduced effective rates to zero and hoped this would stabilize the value. The net effect has however been very weak.

Presently it appears consensus is building for raising the rates to a small positive and then doubling down on adding collateral types and raising debt ceilings instead.

4 Likes

In general I am okay with proceeding down this route, but I think we should take care not to eliminate the usefulness of the Base Rate as a monetary lever when the supply starts outstripping demand again. I can appreciate that the Base Rate is nearly useless with these huge farming APRs, but I worry that when we re-enter positive interest rate territory we will need to rapidly adjust each of the Stability Fees again which I want to avoid. It may make sense to keep the spirit of the Base Rate, but adjust the formula so it doesn’t have the same size effect on all collateral types. I’d be interested in joining a working group to determine a more generalized formula taking into account Monetary Policy, Risk and Market Competitiveness.

3 Likes

Not understanding spell size as the limitation for the base rate.

If it becomes problematic, can’t there be an on-chain module that actually codifies the base rate on-chain. So you won’t need to be adjusting all the stability fees in the for each collateral type in the spell.

(We’re currently paying $5.5k a week for oracle stipends, last weeks cast cost $270 @266 gwei)


Base rate seems useless now because we’re at the lower bound, but it is a tool to be able to adjust all stability fees. Not really understanding why it should be removed from the weekly cycle.

2 Likes

Looking forward to have you in the working group for a new rate setting approach. As I said, Base Rate does make a lot of sense, the issue is just that it won’t be usable with RWA. It will also create problems with stablecoins (already does to some extent) or any other assets where market rates don’t necessarily move in same direction with same magnitude. And then we might end up increasing base rate and decreasing risk premium simultaneously for certain collateral.

But I do agree that using base rate is easier from governance overhead point of view, especially for volatile assets. Maybe a good compromise is to have different Base Rates for different buckets of collateral types? For instance crypto assets, stablecoins, RWA, etc. Or simply stop using Base Rate for non-volatile assets.

Additionally I also feel governance shouldn’t be changing risk premium, or stopped calling this parameter risk premium (maybe spread or rate markup is a better term). Because if we want to charge fees that are tolerated by market we wouldn’t want to be limited by risk premium. In other words if risk premium for ETH-A is 4% but we can afford 6% SF because rates on market are at 8%, why limit ourselves? But the issue is that once you do this market rate targeting approach for each Vault type, you end up changing both markup and base rate simultaneously and it is a mess. This exactly was the original reason I proposed getting rid of Base Rate. And by knowing how much more governance overhead this will make, rather decrease cadence of rate changes.

10 Likes

This was one of the prime reasons I was against doing a global base rate over all assets. Prime example is a 1% BR rise on something that will just lift off of 0 to 1% (big change) to something that will go from 5-6% (not so big a change).

It wasn’t just the above I was thinking about with all of these rate changes but also the governance overhead that would be required. My thoughts on this was that a Domain manager should just be given the responsibility to set rates and then governance just agrees/disagrees but then we don’t deal with the technical issues around this growing management situation.

All of the above was why some time ago I proposed Maker move to using a facility useage based rate setting mechanism since once governance sets to ‘target rates’ on facilities - markets will then manage the rates themselves simply with usage.

something like the above was proposed with MIP17

Look at compound - how often do you see them change their ‘target rate’ or whatever they call the primary rate parameter used to manage the debt facility rate curve based on usage. Not very often compared to how often Maker governance is changing rates.

I also really want to decouple the rate models from the PEG since evidence now presents that when the PEG is high (high DAI demand) there is a certain point lowering rates won’t do enough to provide the markets liquidity. We literally need another mechanism beyond market incentives to provide liquidity when markets need it. The problem here is that there appears to be no concensus on what that model should be. discussion seems to be divided between doing USDC-A vault with low LR and some variant of the PSM but no matter what we do there both of these were designed around using another stablecoin to make DAI PEG stable. Governance needs to come to consensus whether or how much of the portfolio of assets should be dominated by USDC or whether this can be somewhat unlimited due to market liquidity demands. While I would prefer a vault that has fees vs. PSM that has one time fee I have some real issues with going to LR of 102 or the like. I see a lot of people saying this will put a cap on PEG, but lix is correct in that it will require huge liqidity demands and it is possible Maker protocol ends up with abandoned USDC vaults and having to liquidate collateral. Literally at 1.02 there is no room for a liquidation fee, the protocol risk is unclear.

I remember when comp farming came on-line and I was the only one in chat saying “perhaps this is a time Maker should do something different - instead of lowering the SF - raise it and see if the market even reacts”. My data suggested that even a .5% raise wouldn’t have done anything substantial or long term to the PEG and would have had Maker beginning to capture some profit on the farming craze. Here what a few months later we are contemplaing the same thing. Which brings me back to when things seem stable - is there any idea to perturb rates and look for signals on PEG measurement indicators. (i.e. a probing the system for measurement response so we understand via data what market situations are). Literally no bandwidth on this topic of probing markets with periodic rate changes to measure market response. I will say this is a complex topic with myriads of issues and we are already problem saturated for the most part so I will leave this for when things settle down.

This whole rate thing - we definitely are going to have different classes of assets that are treated differently with rate moves. RWA is just one example, stablecoins are another. It is precisely where the facility based rate model can work the best (ones where we expect rates to be pretty stable unless there are unusually high or low market demans for liquidity).

I do think we need a new rate setting approach. Given what I see with the dynamically changing rates based on facility usage in other contracts I really think dynamic rate curves based on facility usage to be a better solution to reduce governance cadence, and to provide a better product for users. Strong up/down end curves for assets with high volatility, weaker up/down end curves for other ones like RWA. Then to manage rates governance can either expand or reduce DC’s (has the effect of lowering/raising rates if utilization is high without actually lowering a facility target rate) or by generally lowering or rasing the target rate on the facility and leaving the DC alone. It is my expectation that if a more dynamic rate mechanism is put into place that governance won’t have to be doing weekly adjustments to the vast majority of rates helping ease technical considerations and reducing governance bandwidth to this management issue. There is also the added benefit that there will always be ‘some’ liquidity in a facility on the high/low usage ends vs. just hitting hard caps and watching markets distort because there is no new liquidity in a facility.

I am also really big on just allowing a combined business and risk units/EPCs to send their recommendations on these issues and governance to just yes/no these things (after propsal made and working group discussions happen as normal if we want).

I like the idea of different buckets. What about the following segmentation?

Securities-like

For instance ETH, WBTC, … but also PAXG and whatever has a high liquidity and a real time price feed (ETF and stocks for instance). Here we just aim to be the cheapest way to borrow a usd stablecoin to gain market share but without taking too much risk (keeping SF > RP).

SF( collateral) = max(risk_premium( collateral), min_lending_rate_competitors(USD_STABLECOIN_for_collateral_pair) )

For instance, on Aave you can borrow USDC from ETH for around 1% (30d average). That what we should aim. Now obviously there are some corner cases (like TUSD which is inexpensive) so it can’t be fully automated. We might want to move slowly (using 30 days average) instead of being insanely volatile. This might help building a yield curve.

For securities-like token we are making a kind of repurchase agreement. We own the collateral and agree to give it back to the customer at a fixed price (minted DAI + SF) whenever he wants. Fully trustless.

A question might be, are we driving the market or are we following it? Food for thoughts.

Monetary tools

Those are mainly used for peg stabilization and should be used only when DAI demand is higher than supply (which is a good issue to have). Those collaterals are not here to generate fees but to manage markets (exchange rates and interest rates). Use of those shows that there is an underlying issue. It’s more band aid than anything else. I put quantitative easing here as well.

USDC, TUSD, … : We will have to find the best LR and SF so that they can be used mainly when the DAI supply from other sources is not enough.

cDAI, aDAI, …: Those will provide a way to expand the DAI supply without taking a custody risk. They can also control the interest rate on money markets.

Illiquid stuff

In this bucket I put everything else. Those will be more on a case per case basis as they will have specific terms (indexed on LIBOR for instance). Liquidation and oracles will often be custom as well. While we might have a collateral as some form of guarantee, it will most likely not be that easy to liquidate it (for instance a 12 month notice before starting the liquidation) nor would be have direct control on it.

This is the world of real banking and real impact on the real world.

PS: I started this post be separating crypto and RWA collaterals, but found that it’s not correct. Gold is a RWA but works just like WBTC (almost the same custody risk). On the other hand, maybe we can have private equity of crypto protocols, NFT like crypto art which are crypto but not liquid at all?

1 Like

Appreciate the feedback @SebVentures.

In general I agree, but if we are still using Base Rate to simplify rate changes for certain buckets, those buckets members should be defined based on their volatility. Because volatility to some extent defines how high rates are or how frequently they change. For instance in your case PAXG wouldn’t fit well in this category, because when USD borrow rates on crypto assets go up (because crypto is bullish), something like PAXG wouldn’t necessarily exhibit bullishness and therefore borrowing USD on top of it shouldn’t get expensier.

Yes, this is exactly some indicative formula we might want to be aiming at. In extreme cases we could be offering SF under risk premium (as we do currently), undershoot at one asset with high minting potential and then overshoot at others, but on average still compensate portfolio risk.

Here is where it gets interesting. It is actually hard to make direct comparison with other venues. It is because Maker is different in: 1) DAI mints are related to DC instead of platform deposits 2) rates at Maker are somehow more stable or don’t get affected when large whale borrows and increases utilization 3) other platform don’t offer protection for losses as MKR does (this is less relevant for borrowers though) 4) LR is different (usually higher for Maker which is less appealing). And then finally, a lot of users simply prefer Maker versus Compound or Aave, presumably due to reputation related to being operational longest.

So this tells me that someone would need to weigh in all these advantages or disadvantages when analyzing a competitive rate.

Correct, those assets should have rates set up primarily based on monetary needs, less on market rates. We should however assume that something like cDAI is more risky than a normal stablecoin and it will be harder to accept low LR/low SF parameters for it.

2 Likes