Rates Changes Proposal 2 Feb 2021

[ultraschuppi]

Authors: @Primoz @LongForWisdom, @Monet-supply, @SebVentures, @Akiva, @hexonaut, @ultraschuppi

Source: Risk Premiums & Competitive Rates Feb 2021

Table Notes:

• Competitive rates are based on lowest stablecoin borrow rates on each lending venue for particular collateral, except when stablecoins are used as collateral (in such case only DAI borrow rate is compared)
• Negative competitive rates are mostly due to liquidity mining rewards at Compound & Cream and due to rates and rewards accrued on deposited collateral
• Risk premiums were recently updated for every Vault type except stablecoins where we used values from last governance votes. Links to calculations can be found in the table above.
• Lending products between secondary lenders and MakerDAO are not standardized and therefore rates can not be strictly compared.

Lending Market Overview

The lending market keep having high rates as we are still in a bull run. Nevertheless, it’s lowering a bit

Competitive Landscape

Rates at competing defi lenders increased markedly in January versus the previous month, with Aave in particular experiencing several rate spikes. Both Compound and Aave have seen rates fall somewhat towards the very end of January, although they remain above levels from early December.

Source: loanscan.io

Looking forward, Cream Finance will begin supporting Uniswap and Sushiswap LPs as collateral beginning this month, and Aave is adding support for Balancer BAL token. Alpha Homora has also launched v2 which offers support for a variety of LP tokens. Once they are live, the new product offerings may increase competition Maker faces for these collateral assets. Compound is also considering a proposal to increase WBTC collateral factor from 60% to 75% (equivalent to lowering liquidation ratio from 166% to 133%), which will increase competition for BTC backed loans.

Proposed Rate Changes

ETH-A: SF increase from 3.5% to 4.5%

Since the last Rates Change Proposal more than 400M DAI-from-ETH-A have been minted, raising our debt exposure significantly. The increase of the SF is not fully covering the Risk Premium and we will most probably need to make another raise within this month if the pace of minting is not getting slower in the next days. From a pure risk perspective we should raise the SF much higher to prevent further debt exposure, since the only other weapon we have (Surplus Buffer) is only a long term strategy (see ongoing Signal Request).

We are still better than the average competition, anticipating that the rates at competitors will rise as well.

ETH-B: SF increase from 6.5% to 7.5%

ETH-B already hit the Debt Ceiling for almost a month and Maker Community does not want to raise it due to the high risk exposure. The SF-increase is following ETH-A here, hopefully moving some vaults over to ETH-A.

WBTC-A: SF increase from 4.0% to 4.5%

Almost 100M DAI-from-WBTC-A has been created in the last month, raising our debt exposure. Adjusting the SF therefore, bringing it to parity with ETH-A. Raising it even higher might make our product less competitive.

BAT-A: SF decrease from 8.0% to 6.0%

DC and DAI-from-BAT decreased a lot. Also, there was one big vault on BAT-A with a fairly low CR that closed as well. As debt exposure is lowered and the high risk vault is gone, we can safely reduce the SF.

LINK-A: SF increase from 2.0% to 3.5%

DAI-from-LINK has tripled in the last month, making it necessary to adjust the SF. We are still better than the competition, but might need to adjust depending on the future debt exposure.

COMP-A: SF increase from 2.0% to 3.0%

Economically it makes no sense to supply COMP as collateral right now, as Compound is giving far better rates (-6.73%). Nevertheless, there is a single vault responsible for 6M of DAI. The Risk Premium itself is higher due to low liquidity of COMP, but if we match that we would probably force the vault owner out of the system.

BAL-A: SF increase from 2.0% to 3.5%

BAL-A usage - though still pretty low - increased a bit in the last 2 months. Competitive rates on BAL increased as well making it reasonable to adjust to the market rates as well.

YFI-A: SF decrease from 6.0% to 5.5%

Based on the current debt exposure (~7M ) the Stability Fee could be a bit lower, however there is more YFI coming in expected so it would not be wise to go much lower on the rates. We already overshoot once on the SF for YFI (in the other direction though).

UNIV2DAIETH-A: SF increase from 1.0% to 2.0%

The fees for this collateral type were set based on the SF of ETH-A when UNIV2DAIETH-A was initially proposed. It should be half of ETH-A SF, we decided to keep it a bit lower since we want to encourage DAI-from-LPs in general.

DC was maxed within 3 days and the debt exposure will surge as soon as we increase the DC.

Uniswap based LP-tokens just got added to Cream and will get added to Aave soon - we are fine with subsidizing so we keep the first mover advantage.

UNIV2USDCETH-A: SF increase from 1.0% to 2.5%

Same as UNIV2DAIETH-A - a bit more since USDC is having a little more centralization risk and we do favor DAI-based LP-tokens. Depending on the future growth of LP-token-based vaults, we might need to go a lot higher here.

UNIV2WBTCETH-A: SF increase from 1.0% to 3.5%

Same reasoning as UNIV2DAIETH-A - was proposed during times of very low SF for both ETH and WBTC-based vault-types.

No Change

KNC-A: No changes (2.0% SF)

No changes are proposed as debt exposure is still insignificant.

ZRX-A: No changes (2.0% SF)

No changes are proposed as debt exposure is still insignificant.

MANA-A: No changes (5.0% SF)

No changes are proposed.

LRC-A: No changes (3.0% SF)

No changes are proposed.

UNI-A: No changes (3.0% SF)

No changes are proposed.

AAVA-A: No changes (4.0% SF)

No changes are proposed.

RENBTC-A: No changes (6.0% SF)

No changes are proposed.

USDC-B (50,0% SF)

No changes are proposed.

USDC-A, TUSD-A, PAX-A, GUSD-A, USDT-A: No changes (0.0% SF)

No changes are proposed.

Final Note

With more than 4.7B TVL and 1.6B DAI outstanding MakerDAO remains committed to offer competitive and stable rates to our customers. Nevertheless, the growth we are facing might force us to increase SF before the next monthly meeting, mainly on ETH-A, if demand continues to push risk higher. Right now, the SF is the only arrow in our quiver until we have higher Surplus Buffer and Liquidations 2.0.

Proposed SFs will get included into next week’s on-chain poll on Monday 8th February, and if passed will be included in an executive vote on Friday 12th February.

[mrabino1]

taking eth-a as an example, as outlined above, the proposed Stability Fee is 4.5%. This stability fee should include both the “default risk insurance” and then some alpha (risk free spread) that Maker earns (if I understand it correctly).

in general, what is the target “alpha” that the community is targeting per vault? are they all the same?

in my eyes, i read the above, and i struggle with the “why X vs Y?” rate… initially i “get” that we have to start somewhere with rates and move them around. however if one views that most of the “competitive” rates may be a harmonic feedback loop from Maker’s rates (I fall in this camp), if we are pricing the stability fee is default risk + alpha, then the “competitors” rate become irrelevant to the computation.

further, based on the above, the risk simulation that @SebVentures recently showed is a great way to start the analytical research per collateral type in a scenario driven way. In my eyes, we need 10x this… However, even then we have to have the foundation of “what is our objective?” … e.g. alpha of 100bps (as purely an illustrative example) per vault + default risk insurance rate…

do we have an active model on how we are computing the “default risk insurance rate” per collateral?

[ultraschuppi]

@mrabino1

in general, what is the target “alpha” that the community is targeting per vault?

unsure if this even relates with your question:
just my 2ct: right now we should rather grow like hell (by providing low SFs while having a good feeling with the risk). we can start milking the market soon enough

analogue to the DSR-spread (haven’t used this term for a long term… sad!): should be near zero

[mrabino1]

No dispute that we should grow! Moreover, even if we assess our “alpha” to be zero (or even negative) and have MKR assume the risk of the collateral (e.g. what we were doing when the SF % was zero), I get it… the DAO is skipping over the the “what is our objective?” discussion RE what alpha we collectively want per vault… and some of that is impacted with business decisions to grow…

[Jiecut]

@mrabino1

if we are pricing the stability fee is default risk + alpha, then the “competitors” rate become irrelevant to the computation.

But, there’s also a feedback loop between the stability fee and the default risk. Risk Premiums calculated by the Risk team aren’t static, they’re subject to demand forces. And the demand forces are partially affected by the stability fee, and also the competitiveness of the stability fee.

The competitiveness of the SF is still relevant. Default risk isn’t static but also partially a function of demand (vault size).

[SebVentures]

@mrabino1

do we have an active model on how we are computing the “default risk insurance rate” per collateral?

Yes, it is in the link provided each month (Risk Parameters sheet). For instance, for ETH-A the Risk Premium is 6% currently and up to 10% at a DC of 1.7B. Obviously, those are just indications. As my work on the crash in slow motion showed, we are not talking about credit risk here. It’s mainly operational and market risks which are super difficult to assess.

Speaking for myself, I hope increasing the SF will limit the exposure before we need to stop being competitive. The surplus buffer is also increasing meaning we can increase risk (but not at the same pace sadly). Still waiting on RWA and derivates of DAI to be less crypto beta exposed.

[cappaberra]

In addition to the SF changes, is there another proposal in the works on raising the debt ceiling for UNIV2USDCETH-A, UNIV2DAIETH-A, and UNIV2WBTCETH-A since they are pretty well maxed out? If there’s another thread for that, I’d love a link, since I’m guessing this thread isn’t the right place… thanks!

[ultraschuppi]

@cappaberra

In addition to the SF changes, is there another proposal in the works on raising the debt ceiling for UNIV2USDCETH-A, UNIV2DAIETH-A, and UNIV2WBTCETH-A since they are pretty well maxed out?

welcome to the forum @cappaberra

afaik there is no running thread around that, but in the latest signal about increasing the DC it was mentioned, that we will be able to increase the ceilings as soon as the running audit is done. weeks not months™

[Jeffshaw]

This is great work once again rate working group. Thanks for your efforts. One concern I have is that it appears we are not optimizing for best market rates, and this potentially puts us at a disadvantage to grow our market share of the stable coin market. At this stage of Makerdao, much like any early startup, we should be optimizing for growth (and corresponding network effects), rather than trying to squeeze out any profits more than what we need to run the DAO and protect against systemic risks from a black swan event. When I look at Dai’s position in the market, we are firmly in 3rd place in many metrics including token operations, total supply, active addresses, and transaction volume. I understand that we need to raise rates to protect against risk, but are we thinking as creatively as possible to keep rates very low? For example, we are now building up our surplus buffer and over the coming months it will hopefully grow to 30 million, but what can we do more immediately to protect from a black swan event that will allow us to keep rates low? I still like offloading risk with a partnership with Nexus Mutual and appreciated @Hugh_Karp’s presentation in the governance meeting. Since we don’t give out farming rewards, we need to subsidize low rates to reach escape velocity and catch up to USDC.

[SebVentures]

I would separate growth from vaults SF. You are right looking at the DAI position in the market. But the link with vaults growth is limited. Those are now decorrelated thanks to the PSM.

Having a huge USDC position is also a way to ensure that DAI keeps the peg should the DAI demand diminish.

Black Thursday and the inability to keep the peg have hurt the DAI position. Increasing the DSR is an option to add new DAI users. There is most likely a lot of other stuff we should be doing. But dealing with the rate is, in my opinion, a weak tool to increase DAI adoption.

[mrabino1]

@SebVentures

Yes, it is in the link provided each month (Risk Parameters sheet). For instance, for ETH-A the Risk Premium is 6% currently and up to 10% at a DC of 1.7B. Obviously, those are just indications. As my work on the crash in slow motion showed, we are not talking about credit risk here. It’s mainly operational and market risks which are super difficult to assess.

I suppose that is really my point. The market risks of how much ETH could move (downward) and its cascading impact on liquidations should drive the risk premium for the loss.

I easily could be missing the logic that was applied to the model, but I don’t see where that simulation-style risk analysis determines that Risk Premium (default risk where MKR has a loss). In my eyes, that is paramount to determine.

[Primoz]

If you click on the link under the table and see second tab where risk premiums are listed, you can get links for each vault’s RP estimation. This numbers are based on simplified model for estimating expected losses, but @Andy_McCall makes additional calc using General Model, where VaR also gets calculated (for major vault types and on portfolio level using correlations).

[mrabino1]

I see that. Curious. Regulations aside, have we looked at the DAO acquiring PUTs to reduce the catastrophic risk with a known time in the future? I glanced through the model (and am by no means an authority on how it is structured), but I am curious why emphasis is placed on the $ amount rather than the market liquidity to acquire at a known price. Granted i fully acknowledge that simulations are complex in nature (and liquidity at a known price can disappear at a moments notice); however i would think the key characteristic is the ability for the dao / protocol to divest at a known price (hence the put comment at the top). Maybe some of the algos that determine options pricing are things the DAO could consider…

PUTs in this context (w/ ETH as collateral) would work better than other collaterals as the derivatives market is emerging w/ ETH… for sure derivatives for other collateral types are way behind.

[Andy_McCall]

@mrabino1

I glanced through the model (and am by no means an authority on how it is structured), but I am curious why emphasis is placed on the $ amount rather than the market liquidity to acquire at a known price.

FWIW the model that @Vishesh wrote incorporates liquidity a good deal more. It is the main thing that is looked at to regress the slippage curve. The models used in the sheet linked above sort of do a “what if” analysis that supposes 40% slippage on liquidation. You can basically think of these numbers as hedging against that kind of loss.

There is work being done to potentially use the general risk model for the monthly rates adjustment, but there is more work that is needed before that is ready.

[mrabino1]

at the end of the day I can imagine the model does need to take a snapshot of what the weighted average blended liquidation strike is… and then basically “recompute” based on historic liquidation data w/ associated slippage and then the trend of the market… no doubt there is crazy statistical analysis and simulation data that can help refine the premium needed to adjust for the then market risk.

[Primoz]

@mrabino1

Regulations aside, have we looked at the DAO acquiring PUTs to reduce the catastrophic risk with a known time in the future?

So ETH puts would be welcomed if we knew we are in bear market and we would basically want to boost our treasury/surplus buffer assets, but we would still somewhat “gamble” with reserves. Otherwise ETH puts could be pretty ineffective. Imagine DAO bought ETH puts in December last year, strike price around market price. Even if we had a 50% crash tomorrow, they wouldn’t help us much. You really need structured insurance that hedges from these tail event random crashes.

[Jiecut]

BAL has quite the high RP.

One thing to note about BAL, maybe it has low volume some days but there’s a ton of on-chain liquidity.

2% slippage for a $1.3M order.

[Primoz]

@Jiecut

One thing to note about BAL, maybe it has low volume some days but there’s a ton of on-chain liquidity.

Agree. We are relying on volumes because a lot of trading is still out of AMMs (LINK being one good example). We need to adjust it somehow so that AMM pool size is used as a metric. Usually large pool size also means higher volume, but at BAL volume is unusually low compared to pool size. And pool size is actually most informative for slippage estimation.

[ultraschuppi]

Rates Changes Proposal vote is onchain now. Please participate.

If the poll passes, the proposed Stability Fees will go into the exec vote on Friday.

[ElProgreso]

Plus 1 on this one – Please vote–you can take advantage ( not sure if there is such a thing right now) of LOW Gas prices between 4AM UTC to 10AM UTC (11pm EST-6am EST).