System Reserves

While governance is currently focused on restoring the peg and improving liquidity, I’d like to open up a conversation about system reserves in preparation of a return to normal conditions (item 12 of the initial problem space from MIP1c2).

Bank capitalization metrics can provide some useful insight. The global banking system generally requires regulated financial institutions to maintain core capital ratios of 4%, meaning that their balance sheet common equity needs to be 4% of their total assets. However, in this context the common equity is not the total market capitalization of common shares, but the book value of equity (assets minus liabilities).

In the context of the MCD system, the 500k DAI surplus buffer compared with the current DAI supply of ~85MM DAI is equivalent to ~0.6% core capital ratio. If the MCD system had 4% core capital on Black Thursday (4% of ~100MM DAI, ~4MM DAI), this would have been sufficient to absorb the majority of the losses.

As a first step, I’d be interested in gauging sentiment on changing the DAI surplus threshold from the current 500k DAI. I’d also love to hear any other perspectives on reserve management :slight_smile:

Summary of Thread and its comments

  • less than 500,000 DAI
  • 500,000 DAI (current value)
  • 750,000 DAI
  • 1,000,000 DAI
  • 1,500,000 DAI
  • 2,000,000 DAI
  • more than 2,000,000 DAI

0 voters


not directly related to the above number but I’ve been curious what the thoughts are in the community about holding non-dai reserves.

This was touched on a bit by @cyrus on the 4/14 governance call. IMO it would be a good idea but I’m not sure what exactly would be the best asset. Maybe PAXG (or similar) given gold’s long history as a reserve asset? Maybe just holding eth itself?

Just curious if people have any thoughts on non-dai reserves since we are talking about raising the reserve value up a bit. Do we like them? What would be the best asset to hold? What is the right split between DAI reserves and non-DAI reserves?


I think it’s important to address the stupidity that the protocol bought MKR at high prices and sold MKR at lower prices. That’s backward. If we collect lots of DAI into the surplus buffer during times of positive sentiment then we’ll have plenty of DAI available to initiate flap auctions on the next downturn.

Taking this idea to the next level, we could add capabilities to the protocol to allow the protocol to place limit orders in the MKR/DAI order book (e.g., on Oasis). Using the surplus buffer, we could place limit buy orders at the bottom of the range and limit sell orders at some crazy high P/E ratio (100-200).

I’m willing to hear advocates, but I’m skeptical; this seems crazy to me. Wouldn’t this drastically increase the complexity of the risk simulation?

1 Like

Well, I won’t advocate too much for crypto assets as a reserve because it is unclear to me how they would fare in a bear market, and recent events seem to indicate that they don’t fare too well.

So what kind of “real world” assets make sense? There is the obvious choice of USD, or in this case synthetic USD in the form of DAI. This is probably a great choice as a reserve but it comes with the downside of actually taking DAI out of the markets decreasing the token’s overall utility. I kind of agree with Cyrus that we could probably go overboard there. Taking out say 4% of the DAI supply as mentioned in this thread is probably fine but I would start to become skeptical once you had say 10% of the DAI supply earmarked for the surplus buffer.

I highly recommend watching the risk team’s presentation on VaR at this point because it is kind of a prereq for the rest of my discussion. Depending on what you found the VaR to be it may not be necessary to hold anything other than DAI for reserve purposes, but if the VaR is much larger than ~8MM (10% of DAI supply) you might need to consider holding something else.

So what is our second choice? Normally I would think “risk-free” assets like 10-year T-Notes would be a good option for something like this, but IDK that there has been any effort to bring those on-chain at this point. As such, I don’t think that would even be an option in the medium term. There probably is some way that you could set up a trust or similar legal entity that would be funded by the protocol and buy the T-Notes for you, but that starts to introduce counterparty risks. Additionally, it’s probably worth noting that uncle sam would likely not be fond of US debt being used to back a stablecoin like DAI.

So what, IMO, is a decent 3rd choice? Gold. Gold historically fairs well during market downturns. It tends to be cheap to buy during bullish market cycles and usually sees price appreciation during times of market stress. There probably are fewer regulatory concerns with gold-backed synthetic assets as things like Gold certificates and Gold ETFs have existed for some time. There is precedent for holding gold as a reserve asset. The US for instance only officially abandoned the gold standard in 1976. Finally and most importantly in my opinion is that there are already on-chain mechanisms that can be used to gain gold exposure. I already mentioned PAXG as an option there but if we thought the counterparty risk was too high in that instance there are other options such as sXAU.

I don’t expect this to be a popular opinion though, so criticism is welcome. Obviously the question of the system buffer size is more of a pressing issue as it is the only option that could be feasibly implemented in the short term. Like I said, I’m just interested in hearing how people feel about holding non-dai assets for reserve purposes generally, and wanted to share my thoughts on what might be reasonable options.


I agree, it’s crazy to buy at high prices and sell at low prices. I cringed a bit as it was happening. However, the stability fees in SCD were paid in MKR, so there was really no other option. In MCD, we didn’t really experience much buy and burn prior to Black Thursday, so wasn’t much of an issue. This should be rectified soon, but I don’t see an easy path forward other than hammering away at some discussions for a few weeks and then putting a vote out.

Regarding other assets, it’s not that crazy. The idea of off-chain reserves has been floated around for a while now. But certainly it’s a long-term plan.


Given the current low SF, it will take a long time for a substantial surplus to build up just from operating cashflow. I know that the idea of “pre-printing” MKR was brought up on the call last week, and while I would be supportive of issuing capital to build up the surplus buffer, we could potentially issue other types of capital instead of just equity.

Traditional banks often issue financial instruments called contingent convertible bonds as a supplemental source of reserves in addition to common equity. During normal times, the instrument works like a bond, paying a fixed rate of return from the date of issuance until the full face value of the bond is repaid at maturity. However, the bonds have built in triggers which allow for them to be converted into common equity at a predetermined price under certain circumstances.

An example of how this could work in the Maker system:
Each convertible bond is represented via an ERC20 token (BOND). Each BOND has a face value of 200 DAI, and will be redeemable on or after the maturity date of the BOND. BONDs are sold via auction at a discount to the face value price of 200 DAI, with the discounted price and term length implying the buyer’s yield to maturity. Proceeds of the auction would be held in a segregated account, and would be topped up to 200 DAI per BOND from the system’s DAI surplus to make up for the original issue discount.

Some numbers to make this more concrete:
Par value per bond/value at maturity: 200 DAI
Term: 91 Days (13 weeks)
Cost at issuance: 195 DAI

Based on the cost at issuance, face value, and term, the buyer of the above BOND would earn an annualized return of roughly 10% over the 13 week term. Note that Maker would need to top up the 195 DAI supplied by the buyer with 5 DAI from the surplus in order to fully back the BOND.

Whenever a flop auction occurs during the term of the BOND, proceeds from the segregated account would be used to make the initial bid of 50,000 DAI (200 DAI per MKR). Any uncontested auctions won by the BOND account would convert a portion of each BOND’s maturity value from DAI to MKR.

BOND holders would be able to earn a fixed return on their DAI in excess of money market rates, in exchange for taking on the risk of potentially acquiring MKR at an above market price.

MKR holders would be able to limit the insolvency risk of the system overall, and in the case of MKR price performing well, the overall cost of capital would be lower than printing MKR tokens directly. In the medium term once DAI supply improves, issuing “debt” capital instruments like this could be a more cost effective way to control the liquid DAI supply, and on a macro level it could help the system sustain a wider DSR spread.

This also avoids some potential issues @cyrus eluded to about the risks of holding substantial assets on the balance sheet. Because Maker doesn’t hold the DAI free and clear, it would not be at risk of governance attack or other asset stripping maneuvers.


I’m pretty sure that I agree with you that there is a probably a problem with the protocols capital reserves, and with the SF heading to 0% who knows when that situation will improve.

However, why is having the protocol take on debt a preferable option in terms of risk mitigation as opposed to something like raising the liquidation ratios.

1 Like

I think of contingent convertible bonds more like purchasing insurance than taking on debt. But to your point I don’t think vaults would be competitive with other lending/leverage options if we required too high of a liquidation ratio.

I pitched the debt idea internally several months ago, but it didn’t gain any traction =(

Why is the convertible useful here? Why is it “purchasing insurance than taking on debt?”

1 Like

One thing to note regarding monetary policy. Raising DAI in this current situation, through bonds or equity, puts more pressure on the peg.

Though convertible bonds are cool. Sort of the opposite of a normal convertible bond, it’s more like MKR holders are buying a put option. An interesting way of purchasing insurance.

I do think it’s a good idea when the stability fee is high to build up DAI reserves. There are many benefits:

  • a buffer in case of liquidation losses
  • won’t need to cyclically sell and buy MKR.
  • Also, the big one is that holding DAI has a positive return for MKR holders. DAI held in reserve, effectively earn the SF.
  • It may also suppress the SF a bit which is good for growth of DAI, less sf volatility
  • DAI reserves could possibly be used to purchase non DAI reserves. (when low sf AND dai persistently above peg)

(Also MakerDAO will also be effectively able to earn the sf on the convertible bonds too, it also becomes more like a fixed rate product for bond holders if they value the cost of selling a put option as low)

(These put options could also be less costly for MakerDAO to sell if they have a floating interest rate (earn the DSR as a coupon), depends whether there’s more demand for a fixed interest rate product or variable)


Agree with basically all of this.

I feel like having a much larger Dai reserve makes sense for the mid-term, and we look into holding reserves in other assets at a later date.

I think the sensible next iteration of reserve policy is to:

  1. Accumulate large Dai reserves during periods of high SF (reserves equalling 5-10% of total Dai.)
  2. Sell accumulated reserves to burn MKR when the Dai peg is high and we’ve hit 0% SF / DSR.

This would appear to have the following advantages:

  • Dai liquidity at the time we most need it.
  • Reserves to balance potential bad debt.
  • MKR burn at a time when MKR is probably trading at a lower price.

At the cost of:

  • No MKR burned during reserve accumulation.
  • Potentially difficult to release the reserves quickly without overwhelming keepers. Assuming we had 1mil DAI of reserves, even dropping 1mil DAI per week would require 100 parallel FLIP auctions per week (with current auction parameters)

On balance it seems like a win.


The big problem with using the buffer as a system reserve (which is not its intended use) is that it means the game theory of emergency shutdown and its effect on the peg gets very screwy, since the contents of the buffer go to dai holders in an emergency shutdown.

In times of uncertainty when emergency shutdown become a possibility, the price of Dai will trend towards the expected payout in an emergency shutdown. If that payout is significantly more than 1 USD per dai, then it creates a self-reinforcing effect where uncertainty causes the peg to break, increasing the likelihood of an emergency shutdown. All in all, while the intention of increasing the size of the buffer is to increase robustness of the system, it may actually have the opposite effect and make it more fragile and susceptible to cascading failure.

The solution is the theoretical guardian system of legally bound entities that hold reserves and use them according to pre-defined logic to prevent MKR dilution and reimburse dai holders in the event of a dai haircut. They would need to be distributed across many jurisdictions to minimize the likelihood of failure, and it will take a lot of effort to get it set up in a legally compliant and safe manner, but it is so far the best option available that is future-proofed, and doesn’t negatively interact with the game theory of the core system mechanics.

IMO it is less problematic to mess around with the buffer if it is just for shorter durations, but a very bad direction to go when looking for long term solutions.


why do the contents of the buffer go to DAI holder in ES?
if there is no bad debt (no uncollaterized DAI) then it should go to MKR holders


@monet-supply Personally I feel the bigger risk is us loosing faith with investors. Growth is great and all, but probably going to be difficult in this climate despite our best efforts.

We just had a round of “angel funding” that we used to write-off the system debt. Here (with the BOND option) we would presumably be asking those same people to come to the table with a few million more DAI so we could sure up our balances, so what happens if it is not enough and MKR issuance happens again? Do those investors keep coming to the table for this money pit? IDK.

My vote would actually be to start slowly raising the LR a few basis points at a time to de-risk, and we can worry about making DAI loans more attractive once things stabilize in the financial markets globally. Just my opinion though.

Hey that kind of sounds like my option two :smiley:! I personally think this is a great idea, but I worry it would take years to properly setup. In the meantime though, we are unable to keep enough cash on our books, so we might want to consider a trade-off. I mentioned building the reserves onchain with synthetic gold assets as my 3rd choice. Curious what your thoughts are on that @rune. To me it seems like the preferable option in the short/medium term because it seems straight forward to implement and could always be dissolved/redistributed once these “trusts” were setup.

If we ended up in this situation of uncertainty and the peg broke upward for this reason would we not just be able to release the reserves at that point? It seems like this would have a significant effect due to the combination of reducing the ES payout, and also flooding the market with Dai.

Any Dai reserves of that nature should reduce uncertainty anyway, and they would almost certainly be deployed before any sort of planned shutdown in the event of the peg rising significantly above $1.

Would it not be possible to operate an on-chain guardian system in the form of smart contracts? I’m guessing there is a reason this isn’t desireable, but if you could spell it out for me that would be great.


This is a simplified explanation: The system cannot know in real time if there effectively is bad debt or not, so if you give the contents of the buffer to MKR holders it is possible to have a situation where Dai holders get an effective haircut (even if the system doesn’t think there is bad debt, because the price feeds haven’t updated yet), and yet MKR holders would get value from the buffer if it is still positive. Leaving MKR holders with value while letting Dai holders take a haircut would be a violation of the social contract since MKR holders are supposed to be the backstop.

you should remember that the contents of this buffer was supposed to be used to burn MKR and we are deciding to not burn MKR and use it instead to build reserves to add another layer of collateral… so even if in ES the buffer wasn’t used to help with the bad debt it will just go back to MKR holders

also, you are talking about a very specific situation

1 Like

I’m having trouble imagining a situation in which we’d have this problem. The ES would have to be both expected and unexpected at the same time.

If the ES was expected by the market in the form of an increased Dai price, we’d release the reserves (or let them be burnt up by bad debt) before contemplating shutdown. Any non-malicious MKR Holder should make this choice, everyone wants the system to survive. If we have a malicious shutdown then I’m not sure how we’d be more screwed than in the general case? It’s also hard to see how a malicious shutdown would be expected by the market.

If the ES was unexpected by the market then we wouldn’t have the issue of the market pricing in the value of the reserves in the case of ES.

Like bottom line, MKR Holders have no incentive to shutdown while reserves are unbalanced by bad debt, we don’t gain anything, it just gives Dai Holders more collateral. In that situation we can always (and should always?) wait for the reserves to be erased. The average Dai user will probably know nothing about this, but it’s fair to say the ‘advanced’ Dai user would both know that Dai Holders get a share of the reserves and that because of this there is no reason for MKR Holders to ES the system while reserves exceed bad debt.


I think we can ignore the system reserves for now - it will not be too late to discuss them when we fix the peg. In general, I think it’s better to develop new DAI design (which can include system reserves) than to try to fix the current design with system reserves.

If we’re doing synthetics, why gold? Why not just synthetic short term treasuries that can be easily converted into Dai when needed?