The Menu Method - a brainsnack

Planet_X · January 6, 2021, 9:40am

The current workflow of adjusting Vault parameters is linear and time consuming. It is linear in the sense that if Maker wants to double the pace of Vault parameter adjustments it we will have to double the number of heads working on this. It is time consuming in the sense that this task is performed manually, something that is an ill fit with the stated objective of becoming the Linux of money.

Can we do this differently? Sure we can! Below is an outline method for Vault parameter adjustment that can be fully automated after the initial setup. Do note that this is not a MIP or pre-MIP and is not in any way a critique of any existing or ongoing work. This article is mental entertainment only and will not be pursued further unless picked up by more capable hands.

Why bother with this? The reason is to allow Maker to function in a more automated manner, moving towards being a protocol. With the current method of work we are adjusting Vault debt ceilings and risk premiums, with the Menu Method we only adjust debt ceilings. As the debt ceilings are getting automated by MIP27 the Menu Method allows for a large step towards the goal of Maker being a protocol for money.

What is the difference?
Presently Maker has one, or in some cases two, Vault types per collateral. Think ETH-A and ETH-B as examples. The collateral type is subject to intensive analysis before a Vault with various parameters is presented to the market. At regular intervals the parameters of the Vault is adjusted.

The Menu Method does this differently - it operates with a range of Vault types for each collateral type from the very beginning. None of these Vaults have been subject to any heavy intellectual effort, they intended to be as standardized as cookies. All have different parameters and all start with a very low debt ceiling. This meny of Vaults, hence the name Menu Method, is then presented to the market and after a while we simply see what works. What works is a combination of what is attractive for Vault holders and what is profitable for Maker. Vault types that works for both Vault holders and Maker have their debt ceiling adjusted upwards by MIP27 or similar. Vault types that do not fulfill the above criteria have their debt ceiling reduced. The procedure is simple enough to be fully automated.

But there are just too many vault parameters for this to work? Are there now? Of the nine parameters a Vault type has, six of them are scalars with nearly identical values across all of Maker’s Vault types. What remains is the Debt Ceiling - which can be offloaded to MIP27, the Liquidation ratio - which tends to fall into brackets of 175%, 150% etc, and finally the Risk Premium. So when it comes down to it it is really only the Risk Premium that is the critical variable.

Example - you may skip this if you understood the idea
As an example we use the onboarding of the generic token XYZ

Generic token XYZ
XYZ-A: Target Debt Ceiling DAI 0.1 million, Stability Fees 2.00%
XYZ-B: Target Debt Ceiling DAI 0.1 million, Stability Fees 2.50%
XYZ-C: Target Debt Ceiling DAI 0.1 million, Stability Fees 3.00%
XYZ-D: Target Debt Ceiling DAI 0.1 million, Stability Fees 3.50%
XYZ-E: Target Debt Ceiling DAI 0.1 million, Stability Fees 4.00%
XYZ-F: Target Debt Ceiling DAI 0.1 million, Stability Fees 4.50%
XYZ-G: Target Debt Ceiling DAI 0.1 million, Stability Fees 5.00%
XYZ-H: Target Debt Ceiling DAI 0.1 million, Stability Fees 5.50%
XYZ-I: Target Debt Ceiling DAI 0.1 million, Stability Fees 6.00%
XYZ-J: Target Debt Ceiling DAI 0.1 million, Stability Fees 6.50%

The reader will immediately notice that the number of Vaults have expanded from single item to 10 items, each with a different Stability Fee. The market is then presented with the Vault Menu and a certain period is allowed before a review. Based on experience this period should be between one to three months.

Upon review it is found that:
XYZ-A: Target Debt Ceiling DAI 0.1 million, Stability Fees 2.00%, utilization 100%
XYZ-B: Target Debt Ceiling DAI 0.1 million, Stability Fees 2.50%, utilization 100%
XYZ-C: Target Debt Ceiling DAI 0.1 million, Stability Fees 3.00%, utilization 100%
XYZ-D: Target Debt Ceiling DAI 0.1 million, Stability Fees 3.50%, utilization 100%
XYZ-E: Target Debt Ceiling DAI 0.1 million, Stability Fees 4.00%, utilization 100%
XYZ-F: Target Debt Ceiling DAI 0.1 million, Stability Fees 4.50%, utilization 38%
XYZ-G: Target Debt Ceiling DAI 0.1 million, Stability Fees 5.00%, utilization 0%
XYZ-H: Target Debt Ceiling DAI 0.1 million, Stability Fees 5.50%, utilization 0%
XYZ-I: Target Debt Ceiling DAI 0.1 million, Stability Fees 6.00%, utilization 0%
XYZ-J: Target Debt Ceiling DAI 0.1 million, Stability Fees 6.50%, utilization 0%

Based on observed behavior of the market we can now see that the market presently will fully support the 2% to 4% Stability Fee. It will partially support the 4.5% Stability Fee, but will presently not support the 5% Stability Fee. Maker on the other hand of course is less interested in these lower Stability Fees as these do not generate as much income.

The Vault Menu for token XYZ is now automatically adjusted at regular intervals:
Vault Menu items XYZ-A to XYZ-D will have their debt ceilings reduced, possibly to 0.
Vault Menu items XYZ-E will see a increased debt ceiling as this is the highest stability fee the market will presently fully support and will be able to provide a floor in case of fall in demand.
Vault Menu items XYZ-F will see a increased debt ceiling as this is the stability fee that has edge case support in the market.
Vault Menu items XYZ-G will see a increased debt ceiling as this is the stability fee that has possible support in the market in the case there is increased demand for XYZ as collateral.
Vault Menu items XYZ-H to XYZ-J will have their debt ceilings reduced, possibly to 0.

A slight consideration of the Menu Method is the fact that a nimble first mover can secure a small Vault with an artificially low Stability Fee, which is why the Target Debt Ceiling is set deliberately low during the initiation period. This is the economic cost of moving the ongoing risk analysis work from Maker and onto the market.

Take aways

The Menu Method eliminates the need to adjust Stability Fees, instead only the Debt Ceilings are adjusted across a range of Vaults with already differing Stability Fees.
The Menu Method changes the effective sum of fees paid to Maker without changing the Stability Fees for established Vaults. This has major positive implications as established Vaults effectively become fixed rate.
The Menu Method allows Maker to match Defi competitors offering fixed rate lending.
The Menu Method allows Maker to offer guaranteed defi competitive rates if we so choose.
The Menu Method allows onboarding of big ticket real world assets such as large US Treasury Bills or Euro debt which have too low yield to be risked in a variable rate Vault. Unless these go to a PSM of course.
The Menu Method lowers risks for capital firms such as 6s which now does not need to guard against rate changes for established Vaults.
The Menu Method is step towards Maker being a protocol, not a startup company or a central bank.

Did you like your brainsnack? Feedback appreciated.

SebVentures · January 6, 2021, 10:25am

Interesting idea.

Two points I can think of:

Vaults don’t have maturity so we are taking an interest risk if we keep SF fixed (opportunity loss). What if DAI drop?
Acceptable interest rate for the market is a moving target.

swakya · January 6, 2021, 11:16am

Very interesting and tempting. Some drawbacks :

The operational side for vault holders becomes more complex. You have to switch vaults as the menu changes weights, split your collateral to minimise your fees…
Rates for us are not just set by market supply and demand but by risk analysis, which does not appear in the Menu Method. Althouhh the min fee on the menu could be chosen by risk.
As @SebVentures mentioned the fixed rate is a source of risk (lost revenue and systemic risk). It also unduly favors historic vault holders. Imagine a massive vault at 1% and prevailing rates reach say 8%. Our only defense would be to reduce the LR and force liquidation but that’s very agressive compared to just raising the rates.

As a possible alternative, what about the Pressure Cooker Method? There is a single vault type. DC is insanely high. At 0% utilization, the fee is 0%. As the debt increases, the fee increases. We’re now back in variable rate territory but that means governance / risk just has to choose the max fee value and a function from utilization to fee. The market determines the exact fee and there is no advantage given to historic vault holders (as there is in the Menu Method).

Planet_X · January 6, 2021, 12:34pm

Maker is already taking a non-visible opportunity cost. Example: ETH-A currently has 2.5% Stability Fee. Possibly this could be increased to 2.75% without declining outstanding debt. But with the current system this loss is invisible to us, although it is a opportunity cost non the less.

If DAI drop the Debt Ceiling increases could be skewed towards Vault types that have a higher Stability Fee, i.e. Increase a Vault type with 4.0% SF instead of 3.5%. That will effect new Vaults only off course, so it is slightly weaker compared to what we currently have. But - as we both know - the peg was not fixed by rates, it was (fingers crossed) fixed by the PSM.

Yes. Which is why the Menu Method will see frequent Vault types having rather frequent revisions. If we can fully automate them then once per week should be no problem.

iammeeoh · January 6, 2021, 1:07pm

I agree with the general idea (and also with the points raised by the others) of your proposal, but overall I feel it could be made simpler.

You main idea, that of using 10 vaults XYZ-A, XYZ-B, …XYZ-J, effectively does one simple thing: market price discovery on the SF value on the XYZ vault type.

I believe this could be done in a simpler way with an automatic SF-adjustment algorithm (IAM like) which:

gets a minimum SF (e.g., 2.5%) which we consider the minimal to protect MakerDAO from Risks).

Gets a maximal DC, which we consider maximal to protect the MakerDAO from risks.
Get a desired utilisation (e.g., 70%) which is both healthy (good returns) and but guarantees that there is some DAI available (30%) to mint if circumstances changes.

And it just adjusts constantly (using moving averages, and other basic ideas) the SF (always above the minumum) to keep the utilisation at the desired utilisation.

This in the long(er) term, allows Governance to just adjust the minimal SF and the maximal DC. Both parameters should be based on Risk analysis and I don’t think should change every month or 2, they should be kind of fundamental and stay valid for several months.

Edit: I think the following comment of @NikKunkel in the chat is relevant:

Planet_X · January 6, 2021, 1:32pm

mmm…no. You do not have to switch Vaults, but maybe you will not be able to increase outstanding debt in your existing Vault - that is correct.

It is correct that the market to a larger degree will set rates. This is however not a bug - it is a feature.

This is why the Menu Method starts with a low debt ceiling at first. See the above worked out example. We will need to dial in the correct rates first and then increase the debt ceiling.

Love it!! It is really good to have a discussion around various ideas regarding Vaults. There has also been mentions of individual fees for individual Vaults. The higher the outstanding debt compared to liquidation ration, the higher the fee.

Planet_X · January 6, 2021, 1:51pm

Also good! “Keep it at 70%” type algorithm. Sounds very interesting - and as you say - probably not too hard to implement. You might want too keep those to Vaults Types with quite a number of users as the opportunity cost would otherwise be a bit high. I really love to hear ideas like this!

swakya · January 6, 2021, 2:03pm

Technically you don’t have to but if a 1% vault type gets much bigger and you’re at 3%, you’ll have to move to enjoy the new fee – previously you could just do nothing and the new fee would apply to you. I guess that’s what fixed rate means though

Since I expect the market rate to be below the optimal-risk rate a good chunk of the time, this feature is both very exciting and very terrifying!

The scenario I have in mind is: yesterday the largest vault type was the 1% vault type (it started very low, price discovery led us to increase it ceiling, and it became very utilized). Today we see that the market can bear 8%. So now there is this huge amount of debt enjoying a 1% rate for possibly a very long time. This costs us money and/or increases our risk!

Let’s keep giving dining-related names to our ideas