Reserve Factor Consensus for the Compound Protocol

[pauljlei]

Simple Summary

To facilitate community discussion and establish consensus regarding the next steps on Reserve Factors for Compound II, Gauntlet provides the following 3 options. Below, we outline the tradeoffs for each option in more detail. Gauntlet recommends that options 1 and 3 are the most prudent options. Since adjusting Reserve Factors is as much a matter of business strategy as it is of market risk, it is important for the community to find consensus.

1. Decrease Reserve Factors by 0.025 for USDT and DAI
2. Increase Reserve Factors by 0.025 for USDT and DAI
3. Wait until Compound III is live

Motivation

The community made excellent points on the purpose of reserve factors and their utility beyond mitigating insolvencies. To continue the discussion with the community, we want to provide insights around important questions raised. As a point of clarification: we are aware that mechanics around reserves will change with Compound III. Nonetheless, we find the discussion around user elasticity relevant to optimizing protocol risk and returns. This analysis and recommendations focus specifically on Compound V2.

Before we dive into the discussion, let’s recenter on why the reserve factor is a valuable parameter in the first place. The reserve factor is an attractive parameter for navigating volatile markets since it can be set at the asset level (as opposed to other parameters like liquidation bonus which is global and applies at the protocol level). The reserve factor represents the portion of protocol revenues (paid by borrowers as interest) retained by the protocol as a liquidity backstop or insurance fund and potentially to supplement the treasury in the future based on community feedback. Reserve factors are generally viewed as supply side incentives because they are expressed as a portion of revenues not paid to suppliers. Conceptually though, the reserve factor is the delta between what the protocol takes in from borrowers and what it pays out to suppliers. A change in reserve factor could be passed along to borrowers as an incentive if an equivalent change were made to borrower interest curves. The reserve factor could therefore be used to incentivize either borrow or supply behaviors. Having robust models for this behavior is important, especially in times of high volatility, to set this parameter at optimal levels that maximize revenue for the protocol.

Specification

What drives reserve growth on Compound?

Reserves are a function of interest rates and total amount borrowed. As you can see in the graphs below, borrow usage on Compound has declined significantly in recent months. Since the reserve growth rate is driven by borrow balance, the annualized reserve growth has also declined.

Looking at the borrow balance by token, we observe that the USDC/USDT/DAI stablecoins represent nearly all of the borrowed value and the lion’s share of reserve growth. Since USDC will be part of the initial Compound III launch, any changes to reserve factors on Compound 2 for USDC would result in noisy data. We, therefore, chose to focus our analysis and recommendations on just USDT and DAI.

How might reserve factor changes impact borrower interest rates?

Reserve factors have not changed on Compound in a long time (since December 2020), and markets have changed significantly since then. Consequently, the protocol does not have accurate estimates of borrower and supplier elasticity with respect to reserve factors. In lieu of that, we have modeled the scenario of perfect elasticity to uncover what would happen if users reacted proportionately to the changes. We acknowledge this likely overestimates the true elasticity of users, who should be stickier for various factors like switching cost, relative risk level vs. alternative protocols, etc. This does illustrate the risk and potential upside of adjusting RF up or down. Borrower interest rates are a function of supplied liquidity, so we must first estimate the impact of the reserve factor on supply.

As explained in our previous post, to estimate the effect on supply, we use the following formula:

If the protocol increases reserve factors, we expect the amount supplied to decrease, driving up borrow usage and therefore borrow interest rates.

The borrower interest rate for the USDT/DAI stablecoins is calculated as the utilization ratio (total borrows / total supply) times a constant factor (a much larger constant factor is used once utilization surpasses 80%, but for simplicity, we will ignore this tail case for now) plus a constant base rate which the rate cannot fall below (currently set equivalent to 0).

For the purposes of this analysis, we assumed borrowers were unit elastic to the borrower interest rate. This enables us to compute the expected borrower interest rate that incorporates this change in supply, shown in the figure and table below:

The current reserve factors are indicated with black dots above: USDT is currently set to 0.075 while DAI reserve factor is set to 0.15. Some point estimates for borrower interest are outlined in the table below.

When benchmarked against rates in the market, borrower interest rates are already higher on Compound:

The current reserve factor USDT is lower than the market (.075 vs. .1). It is already higher for DAI (at .15 vs. .1). Raising reserve factors would increase the spread in the market between Compound and the market rate for both assets. Supposing any level of elasticity on the part of the users would project a decrease in borrows as a result (due to rising interest rates) at a time where our models show risk is well contained (Value at Risk, currently equal to essentially $0 at the 95% tail scenario). Given the decreased volume of total borrows as compared to previous quarters (reserves are projected to grow $2.5 million over the next year at its current growth rate), the opportunity cost of decreasing reserve factors is currently low, with the potential upside of driving borrows (calculation with unit elasticity below).

With risk well managed, this becomes predominantly a question of business strategy. We presented numbers in both directions (increasing and decreasing RF for the top 3 assets) to help inform the discussion, as we defer to the broader community regarding business strategy.

Several available options for Compound are:

1. Decrease Reserve Factors by 0.025 for USDT and DAI - since reserves are currently high enough to cover the 95% tail scenarios showing $0 VaR for all collateral assets, and borrows have sharply declined on the protocol, reducing Reserve Factors is an opportunity to bring interest rates close to the market rate with the goal of driving borrow activity while also gathering data around borrower and supplier elasticity. These assets are also not part of the initial Compound III roll-out and, therefore, relatively isolated from that market.

   Limitations and Tradeoffs: Since we do not have data on borrower elasticity, we cannot model borrower behavior. If users are inelastic to interest rate changes, the reduced reserve factor might result in lost revenues for the protocol. This risk may be acceptable given the low borrow volume at the moment.

2. Increase Reserve Factors by 0.025 for USDT and DAI - proponents of this option may assume user behavior is inelastic to interest rates, therefore allowing the protocol to capture a larger share of the revenue. Any change in reserve factor would enable the community to gather data on user elasticity to build better models in the future.

   While we support either option to gather data and improve models, Gauntlet does not recommend this option since any user elasticity with respect to this change would push interest rates farther away from the market rate as liquidity leaves the protocol.

   Limitations and Tradeoffs: same as above - since we do not have data on elasticity, we cannot model behavior. If users are elastic to this change, an increase in reserve factor could result in decreased revenues for the protocol, the extent to which is unknown given the limited data.

3. Wait until Compound III is live - with the changing dynamics to be brought on by Compound III (reserve factor will no longer be set as a global parameter), the community may decide to wait to change reserve factors.

   The tradeoff here is that we will collect no information on user elasticity.

We see options 1 and 3 as the most viable options. Our concerns are outlined under Option 2. We will initiate a Snapshot poll and look forward to further input and feedback from the community.

Next Steps

• Targeting a Snapshot vote on 8/15/2022 with the above 3 options.

^{By approving this proposal, you agree that any services provided by Gauntlet shall be governed by the terms of service available at gauntlet.network/tos.}

[pennblockchain]

Hey Paul! Thanks for this proposal!

Kirill here from Penn Blockchain. We support the decrease in reserve factors by 0.025 for both USDT and DAI because, as much as cash flow generation is an important aspect of the protocol, we believe that moving towards the market rate is a better idea than away from it in an attempt to analyze the elasticity of users with regards to the interest rate.

I’d argue, however, that reducing the reserve factors by 0.025, especially for DAI, would be insufficient to bring Compound significantly closer to the Market Rate or to significantly impact its market share. Have you considered lowering them further, making Compound more competitive in those markets?

[pauljlei]

Thank you, Kirill and @pennblockchain. We appreciate your feedback. Reducing reserve factors by 0.025 would be an initial step, and we can then reevaluate the next steps after gathering data. The reason why we are limiting the change of reserve factor per symbol is to avoid the polarized and drastic solutions which may shock the Compound ecosystem. By making relatively smaller changes, we can gather more fine-tuned data which can better inform the models and future parameter change decisions.

[Sirokko]

An so it is right back at pretty much same point with Gauntlet again. I don’t know what made Gauntlet so convinced that lowering RF would somehow benefit protocol, as not much indicate that.

But let’s start with common ground we have here. Yes, RF is important parameter and maybe it is time to discuss it’s values.

While most arguments below will address points made by Gauntlet, i suggest for reader not to treat it as a standalone post, and carefully read Gauntlet post prior to it to be somewhat in context, especially if not being an actual user of protocol.

First of all, we should clearly understand that RF by itself does not have direct impact on borrow rates, and the impact on supply rate is rather minor. That rates are determined by relation of amount borrowed to amount supplied, as indicated by formula, kindly presented by Gauntlet. Simply speaking, RF is a sort of a tax, imposed on revenue, collected from borrowers before it is distributed to suppliers. So, if borrower is paying 2% APR, then with 0 RF all of that revenue will go to Suppliers, and with RF 0.1 about 10% from that 2% APR collected will be substracted before distributing remaining to suppliers. However in both cases borrower will still pay same 2% APR regardless of RF being 0.1 or 0. Changing RF does not mean reducing Borrowing APR by itself, but it does have immediate impact on amount of revenue of protocol coming to reserves. (It have a bit more complex relations, as amount of total reserves accumulated in market somewhat decrease rate for borrowers, but as reserves are relatively small comparing to total supply, impact is negligeble, so we not going to discuss that further)

Gauntlet logic implies, that by decreasing RF, amount, distributed to Suppliers will increase, thus will produce iniative to Supply more, which in turn will decrease borrow rate, as being a function from Supply/Demand. That, in turn would incentivise to Borrow more, which will then increase revenue, as lower RF from higher total Borrow can potentially produce more revenue, given Total borrow will increase big enough to make up and even beat losses from RF decrease. Simply speaking, it is suggesting directly decrease revenue in assumtion that IF Supply will increase and IF Borrow will follow increase of supply AND IF that increase will be big enough to cover the losses from decreasing revenue at the first place, than WE MIGHT see some benefit.

While such relations do somewhat exist, there is just so many IF, that it REALLY raises a question.

To begin with, logic is completely backwards. Supply is pretty much irrelevant. WE don’t need to incentivise Supply.

The core problem we have in macro now is actually financial Supply. It’s just too much of it and it is not productive by itself. No matter how much money you print and disperse, it ain’t going to produce more products or services. It’s just inflation. You will get more money chasing same amount of goods and services and that is how inflation is created. Compound is good showcase to demonstrate it.

What we DO need, is actually more Borrowing, Not more of Supply. Supply is overbloated, there is just no place to actually put it to productive use. But before i’m going to go a bit deeper into topic, let’s look at the Compound as a service.

As we all know, it’s a borrow protocol. It allows you to borrow assets given you are able to provide enough Collateral. You can also Lend, obviously, but that as long as there is demand for your assets. Rates are generated by a function, based primarily on total amount supplied vs borrowed for a given market.

For simplicity, we can observe 3 big groups of assets, supported by protocol: Collateral assets, Stable coins, S…t coins. S…t coins group doesn’t imply those assets have no value at all, but rather reflects the fact that their monetary value is very speculative and could easily become zero or close to such. They do have a value assigned by market, but can easily loose it, sometimes overnight.

For collateral assets available on Compound we can see established assets, which unfortunately gives just 2 assets: WBTC and ETH. While WBTC by itself is just a derivative, it inherits it’s value from being backed 1 to 1 by BTC, which is established
asset. However it’s still not pure asset. It’s just a reciept for actual btc which you might or might be not able to redeem at a time of need, so it actually bears quite a bit of more risks than btc itself.

Then we have ETH, the native token of Ethereum, chain, on which pretty much everything what matter resides.

Now let’s look at our prospective userbase. We can assume that vast majority of Compound users are basically traders and speculators. Primarily because using Compound as a funding base for some real world enterprises is kind of difficult because
of lack of Quality Collateral assets. Yes, borrow rates might be amazing to fund any sort of business, but collateral is just crap. Literally best collateral you might use are ETH and WBTC and both of those can easily plummet up to 80%. Not type of collateral you want to have in borrowing money for year or longer.

That’s not Compound fault, but it’s just how it is. So we can’t really expect a lot of capital using Compound as a source of funding for some actual productive business in real economy. Sure, maybe someone could be able to manage it, but most likely we can’t expect big numbers. Thus Compound users are going be primarily for some financial operations inside crypto itself.

And now let me explain why i went to such long description of userbase. (though i tried to make it short)

Let’s look at that Total borrow graph, presented by Gauntlet, where we can observe Total Borrow to plummet from 7b to 1b.

What we can observe in that time perioud?

Well there were 2 big events, which matter for us:

First, of course, the crypto downtrend. ETH went from 3k to below 1k, and BTC went from 45k to below 20k. That important for us, because those 2 are collateral assets. Those people, who borrowed before just can’t afford it anymore. Their collateral at
least halved, they had to repay. And now even with eth and btc going up from their lows it is still impossible to borrow same amount of as before because collateral just not allow it anymore (even if we assume everybody preserved exact amount of tokens through going down)

Then you could argue that stable coins also decreased and rightfully so. Well, the group that supply stable coins to borrow eth, btc, or any small token always was very small. If you look at total borrowing of those assets retroactively, not much demand was there. And for a good reason. It’s just too risky to short crypto. That group of people who actually consistently do it, and for prolonged time and end with a profit is so small.

And so we come to another considerable event. Cutting down COMP distributions in half. Kudos for Gauntlet not actually voting for that.

That was another big initiative for bigger Supply and Borrow of stables at that point. It was profitable to supply stables and borrow stables in recursive way, farming rewards. While it was not designed activity, it also kept Supply and Borrow of stables at higher rate as it would be otherwise. As profitabilty of such activity vaporised, TVL followed,
however not instantly as people were prematurely celebrating at that time about stickiness of liquidity.

So we arrived here, where we are now, at a much different macro reality, much of uncertanty, higher inflation and higher fed rates as well. Do we really expect 0.6% APR at Compound to be attractive when you can buy treasury at above 2% APR? I don’t think so.

Now let’s roll back a bit to a Borrow rate. Should i remind everybody that besides of flat Borrow rate there is also Effective Borrow rate? As well as Effective Supply rate? And Effective Borrow Rate = Borrow Rate - Initiatives?

No, Compound Borrow rates are not above Competitors, if you mean that by a market rate.

Effective DAI borrow rate is actually below 0.6% APR
Effective USDT Borrow rate is below 3% APR also.

While COMP distribution isn’t really primarily intended as subsidy, it however do bear market value regardless, and thus can not be just excluded from equation like nothing is there.

Actually if a person is up for borrowing USDC or DAI it’s hard to find lower rates to borrow than at Compound. WE don’t really need to make it lower to begin with. BUt if we would like to, it’s easy to make it even cheaper, even negative and without actual expense for protocol reserves. We can just ramp up COMP distribution rate for Borrow side of those markets. And that, unlike RF has direct impact on borrow rates, without many “IF”.

But of course, if we look at broader perspective, what do you really expect? Compound isn’t here in vacuum, it have competition. After total valuation of crypto markets shrinked, remaining money concentrated at most profitable positions.

Compound never expressed interest to deploy V2 on perspective L2 (Optimism, Arbitrum) or sidechains, like Polygon. Anybody seriously expect to win competition for TVL, when there is aave on optimism distributing optimism tokens for next 3 month to
come?

Do you really believe that change from 0.6% to let’s say 0.75% APR for stable coins going to attract liquidity, when you can get from AAVE above 3% APR for Dai AND another about 0.8% in OPT tokens on top of that? Or 2.5% APR for USDC with 0.8% in OPT on top Or 2.7% APR for USDT with another 2% on top? Just NOT GOING TO HAPPEN. Of course, as the rates are variable that might and probably will change overtime, but the point here is if you going to compete with other protocols for shrinked liquidity, you can’t expect to win when your rates are several times lower.

Compound can compete for farmers, question is do we need to? We have more Supply side than we need actually and there is just no organic demand on borrow side. Sure, we could stimulate it with some distributions, but truth is there is just not much demand for stables for now.

For example, sure i could, for example borrow more stables from Compound. Great rates, yea. And what should i do with them? Farm on Optimism? Well, that could be done to extent, risk profile doesn’t allow to pour everything into one place. Buying
ETH doesn’t look great idea given merge upcoming with unpredictable results price wise. Buying wbtc? Macro not exactly rainbows and butterflies again.

I’m sorry for wall of text. It actually starts to feel like a job compiling arguments here. To keep it clean i’ll move actual suggestions to separate post.

[Sirokko]

Ok, now when i hope we all are on same page and within context, here is alternative.

I suggest instead of trying to change things in area when we not going to win anyway, we concentrate our tuning efforts in area where Compound probably could beat anyone else and take a look mostly on collateral assets for now. I suggest we change RF:

ETH from 0.2 to 0.75
WBTC from 0.2 to 0.75

To balance such a change i suggest at the same time increase COMP distribution speed for Supply side only:

for ETH x3 (eth is a bigger market and with upcoming merge i expect even more interest in it, thus i want to promote Effective Supply rate to be higher)
for WBTC x2

We can clearly observe increased market interest in borrowing WBTC and ETH from Compound, as effective borrow rates are close to zero. Why such a big change in RF? I believe suppliers of those assets couldn’t care less about Supply APR at all, as it is less than quarter of percent APR (less than 0.1% for ETH). On the other hand, if most of these go to reserves, we could have meaningful reserve growth in those assets overtime. At the same time, to make it effective improvement for users as well, by increasing distribution we will provide effective supply rate bigger than it is now for a relatively small price of COMP tokens. I expect it to stimulate usage of those markets even more, which in longterm would benefit protocol and can potentially lead to more usage of stable coin markets as well. Those markets deserve much more love in governance distribution in general, as while being pretty much as important as stable coin ones, they always were getting scraps from distributions.

For stablecoins on the other side

DAI leave intact
USDC leave intact for now

For USDT market i’m very tempted to support Gauntlet idea of decreasing reserves just for a sake of observing results. I do personally believe nothing good would come of it, and if anything, RF for other stables should be increased instead to about 0.1, which is very reasonable tax rate, neither too stimulative, nor too prohibitive. But well, if we need to make an experiment USDT might be a good enough candidate for it.

We can come back in 2-3 month to evaluate and make adjustments for RF if needed. BY that time we will be after merge and possibly with v3 deployed, and can take a look at RF of stablecoin and collateral markets again.

[blockchaincolumbia]

Thank you Gauntlet for this proposal and @Sirokko and @pennblockchain for your input. This is @remusofmars (twitter) from LionDAO (ColumbiaU).

Rather than a one-off vote, this proposal seems like the beginning of a longer term discussion around how the Compound protocol ought to treat its reserve factors. Put simply – how do we balance constant cash flow generation vs market rate optimization, and what are the longer term effects of focusing especially on either one? I think that maximizing protocol revenue ought to be the goal, but that that is also a long term consideration, and we fortunately have an opportunity to do some testing before the release of Compound III.

Before I digress, I think it is crucial that we get data on elasticity and would support an increase in the DAI RF and a decrease in the USDT RF. The reason for increasing DAI over USDT is simply that it has a higher TVL and is thus likely to produce more revenue during this experiment, but I do not think it matters so much which increases/decreases – just want to collect data on each possibility. (Ideally I would actually increase USDC and decrease USDT, as those are the closest comparables and least likely to have supply/demand fluctuations against each other due to exogenous events.)

If only given the options in the poll, we vote for option 2 . Option 3 is the least preferred since we collect no data, and option 2 is likely to increase protocol revenue in the short term (direct effect of raising RF, while we observe the second/third order effects).

Collecting this data before the launch of Compound III would be valuable to help us make an informed decision/policy then.

Ok, now let me dive into some of the logic of rate adjustments and their effects. Here is a simple step-by-step visualization of Gauntlet’s reasoning for the two active scenarios (* denotes an effect that is not a real change):

Please correct me if I am wrong on how Supply Interest acts, but all of the others are correct (and more important to our debate).

@Sirokko makes the point that suppliers in DeFi are probably quite inelastic, as other protocols offer better APRs on the same assets. Additionally, he says that Compound hasn’t made efforts to build/incentivize via L2s like AAVE has with OP (and is distributing OP rewards). I don’t entirely agree here – I think that Compound benefits from its trustworthiness and that these (OP) types of incentives are mostly short term beneficial and long term inconsequential (except insofar as attracting new users). Either way, I do not really know and am sure Sirroko has more direct knowledge here than myself, but would be interested to see how this change would play out if we agree to treat this mostly as a trial period.

Now, assuming that suppliers are at least a bit elastic, here are the possibilities around revenue given different borrower actions. If our goal is maximizing protocol revenue, which happens by maximizing R = Total Borrows * (RF * Borrow Interest), the actual effect on R depends on borrower elasticity. If we increase the Reserve Factor (as is option 2 in this vote):

1. If borrowers are very reactive (elastic) to the increase in borrow interest, then the decrease in borrowers will overcompensate for the increase in borrow interest rate, and thus will decrease protocol revenue.

2. If borrowers are not so reactive (inelastic) to the increase in borrow interest, then the decrease in borrowers will not be enough to offset the increase in borrow interest rate, and thus will increase protocol revenue.

I am not sure how elastic borrowers are, but would guess that they are probably more elastic in the past few months than during periods with higher growth - this is a prime opportunity to collect near worst case data on preferences.

Please do reach out to me if there are any questions/comments/concerns on this post. Its important that there is mutual understanding.

[Ross]

Hi everyone,

Ross here from a16z. This is a thoughtful proposal from Paul and the Gauntlet team that considers the nuances of both risk mitigation and protocol revenue generation. Given current borrow volumes and low protocol risk (defined by Gauntlet’s modeling of value at risk), this is an appropriate time to assess the impact of reserve factor modifications on protocol usage. A reserve factor reduction of 0.025 for both USDT and DAI will (1) slightly nudge borrow rates toward the market rate and (2) allow the collection of user elasticity data. Both of these are worthwhile experiments to run in our view. Some quick follow-on questions from this:

1. What drove the selection of a 0.025 magnitude change? i.e. is there data supporting 0.025 as the optimal increment size for safe yet meaningful changes to the reserve factor or was 0.025 arbitrarily selected as a safe magnitude?
2. What specific criteria would validate further decreases in these reserve factors? What do you think the timeline looks like for both observing that validation and the subsequent proposal of further decreases?
3. As @blockchaincolumbia has pointed out, an increase in one reserve factor and a decrease in another may provide more data on user elasticity than simply decreasing both. If this hybrid approach is employed, what specific metrics should be tracked and at what point should subsequent modifications to the reserve factors be made?

Thanks so much!

[pauljlei]

Thanks, @Ross @blockchaincolumbia @Sirokko for your thoughts. Given the open questions and community engagement here, we will wait to publish the Snapshot vote and will give a heads up before we do so. We will come back to the forums shortly to respond to the open questions.

[Sirokko]

1. Changing RF will not let us collect data on borrower elasticity and would not move Borrowing rate towards “market rate”. RF factor have no direct impact on Borrower rates. Raising or lowering that parameter will still result in same Borrower rate.

2. Compound Effective Borrowing rate after taking in account COMP distributions are already among the best if not THE best in industry. There is simply no need to make it lower, because if we don’t have more borrowing demand while asking half rate than competition, we ain’t going to increase borrow demand if we lower it a bit more. Thus, no encreased earning to reserves.

                   USDC Borrow          DAI Borrow            USDT Borrow             
  
  AAVE (eth)          1.58%              2.29%                  2.3%
  
 Compound             0.65%              0.67%                  2.6%
(effective rate)

*USDT stands aside here, as distributions mostly were concentrated in USDC and DAI historically, and USDT is not usable as collateral and thus bearing less cumulative effect on protocol.

If you in market for borrowing stables you just not likely to find better deal. Already. Compound rates are NOT higher, but are WAY lower than “market” for 2 of 3 stables coins.

It is true, though, that it would allow to collect data on Supplier elasticity, as that is where the change would be observable, while still relatively minor. Since it would be small, i fully expect it wouldn’t bring Supplier rate to high enough levels to be interesting for Suppliers. So speaking in elasticity terms, while i do expect Suppliers to display elasticity in response to change of RF in general, we unlikely be able to observe it, as Supply rate will still remain quite low as we already have too much Supplied for current market Borrow demand.

I want to give special note on increasing DAI rate. RF for that market is already high, at 15% taxation rate we should be very careful to push it higher. The reason i don’t suggest lowering it, because it works fine as is. Effective Borrow rate is exceptional, effective Supply rate is 1.3%, which while not something amazing, it can grow fast if at some point Borrow demand will catch up. And it might, because when it will, it will come to Compound first, because our Borrow rates are just better than elswere.

Increasing RF for DAI short term will increase earnings to reserves, but in long run might put unnesessary down pressure on Supply rate. If anything i’d suggest to experiment on relatively isolated USDT market FIRST if it’s just for data, before potentially breaking something which works well with adjustments not called for.

Please, listen me up, collegues. Here is a napkin math for you. Let’s take USDT market and what we are discussing here. With current Total Borrow of 233M and RF 0.07 we are getting about 1295$ in reserves daily. With RF 0.1 we would be getting 1850$
in reserves daily. 500$ increase. Nothing groundbreaking actually.

Now lets look at WBTC. With current Borrow of 24M and RF 0.2 we getting 493$ in reserves daily, with RF 0.75 we going to be getting 1850$ in reserves daily. AND we don’t really take anything from suppliers, we take out their flat dust interest they
recieving, and replacing it with increased COMP distributions. Effectively, they are going to recieve more yield on their position than before. And that is for mere doubling COMP distributions to Supply side. Currently we disperse about 70 COMP totally for both sides of that market, doubling supply side will make it 100.

Effectively we trading our “treasury” COMP for increased reserves in WBTC. And doing so without selling single COMP into market as protocol. And users benefit from that change. WE DO want more reserves in ETH and BTC than more COMP in stash, because that COMP is very volatile
.
AND if we want actually to increase our earnings, rather than slightly optimise what we currently having, we can double not only Supply, but also Borrow side. That action i expect to be VERY elastic at Borrow side and will actually bring more borrowers. That can also bump stable coin Supply markets, because, yes, those guys, who short wbtc and/or eth are likely using stables as collateral, not anything else.

Suggested increase in RF for ETH and WBTC actually benefits both users and protocol, allows us to capitalise on our market strength (amazing deal on borrowing eth and btc, for those users who want to bet on that) and in the long run reserves accumulated from such adjustments are likely way overshoot anything we going collect from stable coin markets in usd value in current market conditions. Though, let’s not call that profits, as currently protocol spends more than it earns, so it’s kind of non-profit system really.

Yes, i am aware it’s probably hard to push something like this through governance, and for those people, that going to complain about how we could take away mighty 0.15% APR away from suppliers. Well, if you supply 1 million of usd in WBTC, that going earn about 1600$ per year, or about 133$ per month from 1M supplied. That is dust. It only makes accounting more complex, rather than bear any financial meaning. On the other hand, that supplier is going to recieve more than 1600$, but distributed in COMP instead. He can sell and convert it to wbtc if he choose so and come ahead. Or keep it in COMP.

Interest for Supply side of ETH and WBTC markets is dust for pretty much any individual user. But all interest combined do have meaningful value for a protocol as reserves. By replacing that interest with COMP for users, we do good service for both users and protocol. It’s a beautiful and meaningful improvement, while moving RF by 25 basis point for stable coin market is likely not. Especially when motivation is like: we are not sure what going to happen, so let’s try that.

Yes, that RF increase for WBTC and ETH isn’t going to be forever, of course. COMP are not unlimited, and while we can continue distributions for years to come at the current rates eventually they will run out. It’s a temporary measure indeed, maybe for several quarters, maybe for a year. That is going to be adjusted down when we gather more data on effects. But in the mean time we going to get bigger reserves and likely higher total borrow for those markets.

To Summarize it i suggest:

INCREASE RF for WBTC and ETH markets to 0.75 with simultaniously doubling suply side distribution of COMP for those markets to compensate Suppliers.
LEAVE DAI RF intact for now until we have results from USDT adjustments.
INCREASE USDT RF to 0.1

Increase of USDT RF will increase reserve growth for protocol with possible decrease of Supply side of USDT market, which short term might increase reserve growth even more. Long term if borrow side turn to be elastic to Supply side decrease it might decrease reserve growth. I’d expect it will still net higher reserve growth than before increase of RF anyway.

Alternatively we can decrease USDT RF to 0.05 instead. That action will immediately decrease reserve growth for that market and slightly increase Supply rate for that market. It might increse total Supply for USDT, (especially as long as borrow dai/usdc, converting it to usdt and resupplying it as USDT is profitable activity in Compound), which in turn will decrease borrow rate, which might increase borrow side. I do not expect though, that it will be enough to compensate for loss imposed by decrease of RF at first place, so we highly likely are going to end with net loss of reserve growth for that market.

I strongly believe that RF adjustments for ETH and WBTC is an improvement for both protocol and users, while adjusting RF for stables is kind of less impactful at the current conditions. I still would expect that bringing RF for USDC and USDT to 0.1 in the long run likely going to work well for protocol without asking to much from a users.

Going above 0.1 RF for stables will likely work (as it works for dai), but that is trying to extract more value for protocol at expense of users. Question here is protocol about profit, OR protocol is about common good? Not for profit, but only for covering own expenses. (which means protocol isn’t tuned to generate profits for itself, but rather as balanced system, on top of which users can build for profit enterprises)

I think we need more options for snapshot, as increasing RF for DAI isn’t something i am willing to advocate for. Let’s concentrate only on USDT for now.

[pauljlei]

We received some great questions from the community (some of which we paraphrased for brevity). Our feedback is below. Our upcoming Snapshot will present the same options outlined initially above.

We plan on publishing a Snapshot vote on Wednesday, August 31.

1. What specific criteria would validate further decreases in these reserve factors? What do you think the timeline looks like for both observing that validation and the subsequent proposal of further decreases?

TVL (supply) of DAI and USDT is where we’d expect to observe the primary impact. Two factors that create noise in the system are market volatility and the behavior of large whales, who can have an outsized impact on TVL. For this reason, we will analyze not just total numbers but also individual accounts to observe behavior after the change.

Next, it is valuable to consider secondary effects on total borrows against the collateral. If the reserve factor decrease does result in increased supply, we expect borrowers to adjust their positions to maintain similar borrow usage as before. We previously observed this behavior in response to collateral factor changes.

Finally, to estimate the impact to the protocol, we will compare the net interest generated for the protocol after the change (reserve factor * borrower interest paid) and compare that to the interest generated the day before the reserve factor change went into effect. If total supplies, borrows, and reserves increase due to the change, we would consider impact as a positive outcome validating that users are elastic to reserve factors. As mentioned before, this will need to be considered in the context of external factors such as market volatility and whale behavior.

Regarding timing, we’ve observed with collateral factor changes that some users react very quickly while others can take weeks to adjust their positions. We recommend waiting for ~one month between subsequent reserve factor changes to observe the effect.

2. Why did you choose .025 as the change increment?

The increment was selected based on the scale in statistical analysis. This increment is expected to be large enough to show impact if users are elastic to the change since the relevant stables here have reserve factors that are already low. This represents a 33% decrease for USDT and 17% for DAI. It is prudent to avoid making drastic changes since yield farming is largely an all-or-nothing practice. Once yield farming becomes unprofitable, we’ve observed that users pull their liquidity from a protocol and that liquidity might not return.

3. How about increasing one reserve factor and decreasing another?

Changing reserve factors in two different directions, as suggested by @Blockchaincolumbia, creates noise in the data. While the suppliers of each collateral are generally separate profiles, making opposing changes introduces an additional variable where suppliers might flow from one asset to the other. It would be hard to distinguish if the change was due to a change in strategy/risk appetite or to elasticity to the yield impacted by the reserve factors. For this reason, we do not recommend this approach.

4. Effective rates are already higher on Compound for LPs than the market due to COMP rewards. Should we raise RFs and incentivize supply with additional COMP instead?

The COMP rewards are indeed what make supplying on Compound more profitable due to their impact on the effective interest rates, but these returns fluctuate with the price of COMP. More importantly, a strategy that relies on incentives in this manner presupposes that those incentives will last forever, a strategy which @sirokko already pointed out is not sustainable, particularly in a bear market. If the goal is to uncover elasticity of users to the real interest rates and reserve factors, then we should not add additional inflation to the interest rates by increasing COMP reward spend.

The behavior currently observed related to USDT borrowing is only profitable due to COMP rewards (and is therefore not sustainable long term). See the following example.

Rates without COMP incentives

Token	Supply Rate	Borrow Rate
USDC	0.73	2.14
DAI	0.72	2.22
USDT	1.39	2.96

Rates with COMP incentives

Token	Supply Rate	Borrow Rate
USDC	1.15	1.01
DAI	1.23	0.90
USDT	1.52	2.71

If one were to borrow USDC, convert it to USDT, and resupply it as USDT, they would make around -0.75% interest without COMP incentives and 0.51% interest with COMP incentives, plus however much interest they are making from the original collateral they put down. For instance, if someone were to supply DAI, borrow 83.5% of it (the maximum allowed) as USDC, convert it to USDT, then resupply the USDT, they would make around 0.09% interest without incentives and 1.66% interest with incentives. If we were to decrease reserve factors for USDT, the first order effect would be an increase in supplier interest rate for USDT, which may make this more profitable without additional COMP rewards.

5. Should we change the ETH reserve factor?

The challenge with this suggestion is that the ETH merge is upcoming. It would be difficult to attribute observed supply/borrow behavior changes to a reserve factor parameter change with that backdrop.

[pauljlei]

Snapshot is live for voting below. Voting ends September 9th, at 2 PM PT.

https://snapshot.comp.vote/#/proposal/bafkreibb7hbrrgz5a2xdg55uzhynfotw7xadn45wefzr3uecazhcnu2oeq

[rleshner]

Note: it would be great if the Snapshot could include cCOMP somehow.

Compound III is still in a preliminary testing phase with limited supply caps; I would like to contribute the following thought:

If Compound III is demonstrated to be more efficient, it may make sense to migrate reserves (from a version of the protocol with excess reserves) to a version with too few reserves–or deployments on multiple blockchains. This would change the calculus for the v2 Reserve Factors–potentially to one in which the community chooses that the better direction is to leave them unchanged, or even raise them.

My individual preference is to “wait” productively–by designing a framework for Target Reserves for any given deployment based on total borrowing volume, or some other metric, that the community could then use to determine which deployments are most over or under reserved given the level of risk to users.