Reserve Factor Standardization

Background
The Reserve Factors of supported markets are inconsistent, and have evolved through a variety of governance proposals and new market launches.

Reserves are taken out of interest paid by borrowers, and reduce the corresponding supply interest rate in the market. Reserves accumulate in each cToken contract, and can be deployed by the Governance process for any variety of use-cases.

Current Reserve Factors
USDC: 5%
USDT: 20%
DAI: 5%
ETH: 10%
WBTC: 10%
BAT: 50%
ZRX: 50%
UNI: 20%
COMP: 20%

Philosophy
Reserve Factors have been a long-standing discussion in Discord, and there are a variety of lenses through which to view them. Historically, discussion has centered around:

  • Reserve factors are the protocol’s fee to offset risk (which is incurred when there is borrowing outstanding)
  • Reserve factors incentivize behavior, by acting a component of interest rate models
  • When reserves are large relative to borrowing demand, reserves functionally add liquidity to a market

Changing Reserve Factors does not immediately change the risk of the protocol (compared to changing a Collateral Factor), which should make them flexible levers to consider adjusting.

Potential Approach for Discussion
Reserve Factors should be standardized in a consistent format–just as interest rate models have begun becoming standardized across markets. By using the risk associated with borrowing the asset as a parameter (e.g. volatility of the asset), we have a starting point to frame new reserve factors.

I would suggest (and love everyone’s feedback on), the following Reserve Factors:

USDC: 5% -> 7.5%
USDT: 20% -> 7.5%
DAI: 5% -> 10%
ETH: 10% -> 20%
WBTC: 10% -> 20%
BAT: 50% -> 25%
ZRX: 50% -> 25%
UNI: 20% -> 25%
COMP: 20% -> 25%

These create sequentially higher Reserves from borrowing stablecoins, to semi-volatile stablecoins, to large liquid collateral, to more volatile collateral assets. Going forward, as new assets are added to the protocol, they could follow this format as well.

7 Likes

Numbers above look pretty good to me!

One potential modification, I think it would be interesting to add a surcharge onto the reserve factor when total reserve quantity for a market is below x% of total assets borrowed. If this is feasible on a technical level, it may be an effective way to discourage excessive leverage and help bootstrap reserves for new markets.

4 Likes

That’s a cool idea.

The cToken contracts accept Reserve Factor as a Governance function; DAI, USDT, UNI, and COMP (the “gen-2” upgradable cTokens) could evolve to use more complex logic, but the other markets (ETH, USDC, WBTC, BAT, ZRX) need to be set manually. For simplicity, all of these could apply a “layer” in the manual Governance process.

The other perspective could be that Reserves are somewhat fungible; there’s nothing to stop Governance from using the Reserves from an old market, to support a new market somehow (this was actually done a long time ago, to transfer the SAI reserves from v1 to DAI in v2).

I also suggest having 2-step approach to Seting Reserve factor. I believe we should determine target reserve pool for every market. For example, it could be something like 5% of supplied assets. So, when reserves for market are lower than target (5%) reserve factor is set into accumulation mode and set to higher rate, to stimulate reserves grow.

When reserves for market are at target level or above it, reserve factor is set to normal mode, which is lower.

USDC: 5% -> 10%
DAI: 5% -> 10%

Reserves for both this stable coins are very low in comparison with size of market, i personally would like to see even higher reserve factor than 10% in accumulation phase.

USDT: 20% -> 20%

I suggest to stay alert towards that particular stable coin. Despite of it being widely used, there is no evidence whatsoever that it’s backed 1to1 by USD, as there are no audits provided by issuer. It might be, it might be not, but risk is definitely elevated with that coin, and reserves for market are miserable (126 000)

ETH: 10% -> 20%
WBTC: 10% -> 20%

I’m not against accumulating reserves for that markets, as they are very small, but it’s worth mentioning that even such increase might be futile. There’s very low demand on borrowing side, and interest is very low also. Actually it’s so low accumulation, that might be it even worth consideration to subsidize these markets from reserves accrued from other markets (aka primary stablecoins)

BAT: 50% -> 25%

This one is special, as being the one heavily used in initial farming of COMP, which in combination with initial high reserve factor let this market to accrue considerable reserves. This pool i beleive doesn’t really need to grow reserves and can benefit from lower reserve factor indeed.

UNI: 20% -> 25%
COMP: 20% -> 25%

We might consider having higher reserve factor for these tokens, as these markets are new and reserves are still low.

1 Like

To address USDT specifically; I (and the majority of the community) completely agree that there is considerable risk of USDT declining in value. It’s not fully backed, it’s not transparent, it’s printed whimsically, etc etc.

This risk is reflect in USDT’s 0% Collateral Factor, e.g. if it collapses in value, the protocol doesn’t suffer. But, it doesn’t have any more upside volatility than USDC – borrowing USDT doesn’t jeopardize USDT suppliers more than borrowing USDC jeopardizes USDC suppliers. IMO, they warrant the same Reserve Factor.

On the other hand, DAI is a somewhat more volatile stablecoin–we saw considerable up-side price volatility in March 2020, which poses slightly more risk that DAI borrows aren’t properly liquidated–leading to losses for DAI suppliers.

This is why, IMO, the framework for Reserve Factors should look at volatility as a starting point.

Second, your perspective for “seed the reserves” is a fair argument for having slightly larger Reserve Factors across the board. :+1:

2 Likes

Okay! My thoughts:

  1. Regardless of what our philosophy on reserves are, I believe we all would agree having larger reserves in 5 years is better than having smaller ones.

  2. Tokens with the lowest reserve rates pay the greatest amount of interest into the reserves. Tokens with high reserve rates pay the smallest amount into reserves. Yes, I know correlation does not equal causation but this is a factual observation

  3. We are very early as protocol, in absolute numbers these reserve rates will not significantly de-risk the protocol. (i.e moving wBTC from 10% - 20% will change reserve growth from $20 a day to $40 day).

  4. A primary value proposition of decentralized finance is dis-intermediating rent seeking third parties. The protocol taking 25% of all interest paid to lenders is goes against this ethos.

  5. The primary value proposition of the protocol is earning yield, optimizing for maximum yield rates will lead to faster growth.

Given this, I would propose the following rates:

USDC: 5%
USDT: 5%
DAI: 5%
ETH: 10%
WBTC: 10%
BAT: 10%
ZRX: 10%
UNI: 10%
COMP: 10%

Stablecoins 5%, non-stable coins 10%.

Questions

  • My main assumption here is that higher reserve rates right now will not meaningfully de-risk the protocol when we look at the actual numbers. We should look at this question more closely

  • The counter argument I see to this is that right now borrowing and lending is being incentived with COMP and so therefore it’s a good time for the protocol to be more extractive. This argument makes sense to me but as total borrowing and supplying grows the relative subsidy of the COMP will shrink

Your observation is correct, but i do completely disagree with your colnclusions. Actually there’s much more sense to do exactly opposite.

Let’s take a look at WBTC market:

Current Reserve factor: 10%
Current Reserves: 1.2383 WBTC
Interest Paid/ day: $192.15 (10% of that will go to grow reserves)
Current WBTC Supply: $489 Million ( 90% of interst Paid/ day are up for distribution between suppliers, which is $172.8 )

It’s pretty much safe to assume that interest accrued on Supplied WBTC is pretty much negligeble and close to zero. It doesn’t really impact Suppliers if they get their 0.01$ per month or not at all.

Now let’s take closer look at Eth market in similar way:

Current Reserve factor: 10%
Current Reserves: 100.02 Eth
Interest Paid/ day: $1315.15 (10% of that will go to grow reserves)
Current Eth Supply: $747 Million ( 90% of interst Paid/ day are up for distribution between suppliers, which is $1183.5 per day )

In similar way, we can observe that interest gained by Suppliers is negligeble.

Conclusion: Interest is not a primary if not at all motive in Supplying Eth or WBTC to Compound, but rather serving as a Collateral.

Proposal: Set Reserve factor for ETH and WBTC markets at 100% for next 3 month and evaluate after that. Revisit Reserve factor for these markets in 3 month and decide on if it should continue as is, or be adjusted to lower Reserve Factor.

Argument: For every single WBTC or ETH supplier interest accrued is a dust amount, however for Reserves while small, but still overtime noticeble amount, which would support seeding of Reserves for that markets. I don’t see a noticeble downside for suppliers as they were not actually gaining any substaitional amount to begin with. By forfeiting their pennies suppliers will help to create some reserves for protocol, which ensures more stability and thus financial security for Suppliers in future.

2 Likes

One thought I was considering (along with an earlier proposal) is to essentially enable COMP holders to use their COMP in creating/adding to reserve pools - potentially across the board in terms of serving all markets.

Generally in agreement with @rleshner’s proposal to have a uniform notion of reserve factors - and also driving a better understanding of the risk framework associated with borrow demand. However - an increase in market driven rates, as pointed out by @lay2000lbs, wouldn’t really shift the needle in terms of enabling building up larger reserves in the short/mid-term + it has the effect of dampening supply rates and thus the enthusiasm of backers to participate in a market that reduces their capability to earn interest for taking the risk of participation.

To make this fair to people who are just using the platform to borrow/lend assets and may not be fully incentivised to be long-term participants/contributers to the growth of Compound, we should consider one of two alternatives:
i) Increase the COMP distribution (in conjunction with vesting) in order to offset some of the losses they’d be facing + also potentially incentivise them to be more engaged in Compound
ii) I don’t understand the technical feasibility of this, but I’d really like to see the emergence of a common COMP reserve pool, with the capability for COMP holders to add liquidity to the same. The idea would be very similar to the notion of CET in the Basel III spec, with the notion that this reserve could serve as the minimum requirements needed to serve the risk parameters of a given market, in addition to serving as a function to signal the appetite to seed current/new markets.
Additionally, this would mean that COMP holders who are incentivised towards the growth of the protocol would have a very meaningful way of doing the same by chipping in liquidity to this reserve pool - with the understanding that this pool would be subject to liquidation risks as well as a potential to earn some part of the interest in the form of further COMP incentives/a few basis points off the supply rates. The additional reserve factor (as it is structured today) could then take the additional form of a market-driven assessment of risk in the underlying asset - and be used to contribute rewards to the COMP reserve pool holders.

If this were truly possible to implement, this would be a much neater way of classifying risk ratios in terms of risk-weighted assets, outflows and net exposure. Basel III specifies seeding markets with 4.5% of RWA’s - if the community is interested, could pull up some numbers on loans on the book for the past month/3 and use that as a basis to set out to find a number that makes sense, as opposed to a finger-in-the-air estimation of 1% to start out with.

Any thoughts?

1 Like

Hmm. This is interesting. It’s a great distinction to make between assets people are depositing to earn interest vs. assets people are depositing for collateral.

However, I still would push that reserve rates to be minimized.

I like these ideas! Overall I like the idea of COMP being able to serve as a reserve of last resort / insurance for the protocol in some fashion.

Based on COMP distribution of 2,312 per day and current price of ~$125, the protocol is distributing ~$290,000 per day in incentives. With current daily interest and reserve factors, the protocol is taking in roughly ~$17,000 per day to the reserves, or ~17 times less than COMP outflows.

Hey, monet appreciate your work in trying to get Uniswap governance moving anywhere. :slight_smile:

But i think you have a wrong way of thinking about COMP. Protocol isn’t distributing 290000$. COMP isn’t money, it have zero backing and cost zero to protocol to create and distribute. Same as with UNI. Basically that $290000 per day isn’t distributed by protocol, it’s actually taken from market, protocol have zero cost of making that.

It is a combination of distribution of tokens AND buying demand, which makes it’s value. But for protocol itself it’s not money. It could potentially be money if that COMP would be converted to Eth, or btc, or stable coin. But that is IF. Maybe market would be able to absorb it, if it was to be sold, but highly unlikely without price impact. It’s wrong to percieve COMP reserves as big money bag sitting there. Because today it might look as 125$ per comp and in 1 month it may very well be 10$ per COMP. It’s extremelly dangerous to treat something with highly variable value as a reserves.

Reserves on the other hand, especially in stable coins and maybe in eth or btc to certain extend are money already.

However, that brings us to another interesting option for protocol. For example, portion of COMP could potentially be auctioned (solld) to create some reserves buffer for Compound pools rather than distributed to liquidity providers. And that might be another viable option for making reserves aside of Reserve Factor.

Overall, I agree with @rleshner assessing the level of reserves for each market for the time being.

This has been an ongoing tangential discussion, but never really been addressed due to lack of research. I think extensive research should be done in the future to choose optimal rates beyond the finger in the air, but for now, synchronizing them as Robert proposed makes the most sense to me.

I like @Sirokko idea on the 100% reserves for ETH and BTC, but I do think that the competitive disadvantage would outweigh the potential reserve gains. I think the idea of 100% reserves will not sit well with the end user.

2 Likes

Most of the liquidity provided in compound, comes with the idea of farming COMP.

This is not going to be the case in maybe 4-5 years when all the COMP became available in the market, but if we start to look at COMP as a valueless asset it will affect the overall liquidity in the protocol.

COMP has value and we should consider it a liquid asset. Maybe we can work in fixing the inflation of the token, so whales stop dumping it as soon as they collect it.

Increasing the reserve is a good way to do this.

I agree with you, the borrowing side (DAI) needs to be strengthened a bit and ETH and BTC as basic protocols are oriented towards long-term holding and the most logical solution as collateral and I am of the opinion that no need for rising that reserves.
For UNI and and especially COMP we need large reserves because their tokenomics are still in development and it is still quite uncertain where the value capture will be.

1 Like

I generally like the numbers that Robert suggested. I might also suggest using a formula for the reserve factor that’s a function of price volatility (standard deviation calculated from the relevant asset price feed). The reserve factor could be the max of the Robert value vs the price volatility based model.

I think the best way to create value for Comp token holders in the future (only once the platform is stable with enough operational history) is to distribute reserves that accumulate above a certain threshold to holders.

2 Likes

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

This seems like a reasonable first pass. Ideally, these parameters should vary function of the default probability, but that requires more analysis. Given the issues with Dai on 11/26 and the need to have higher reserves for assets with weaker liquidity, we will support this proposal.
-----BEGIN PGP SIGNATURE-----
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=bcN0
-----END PGP SIGNATURE-----

2 Likes