Collateralized Debt Obligations for Issuer-Backed Tokens


#5

Doesn’t this depend on having price feeds for the various ERC20s so that you know when to trigger liquidation calls for CDPs? My scheme does not depend at all on price feeds; it instead gets the same effect through incentivized preference revelation (ie. people participate in the “implied price feed” by choosing which asset they withdraw).

Such as basecoin.

Personally I’m not sold on basecoin specifically; it’s coins/bonds/shares model seems a bit iffy and unnecessarily complex. Particularly, there’s the instability that if the basecoin price goes down, then the mechanism pushing the price back up is to get people to buy basebonds, but basebonds basically just lock you into holding basecoin, and it’s not clear why people would want to do that; it seems too close to the old bitusd model (“we just say that the price of this token should be $1, and therefore people will buy if it’s under $1 and sell if it’s over $1 because they expect the self-fulfilling prophecy to be true”) for comfort. I am more of a proponent of seignorage shares: https://github.com/rmsams/stablecoins

Another thing I have thought about is that in an economic model where you do not assume altruistic honesty or non-coordination, it’s not clear that makerdao has a higher security level, or even that it’s possible to achieve a higher security level, than seignorage shares. If the total discounted expected future profits of the scheme are lower than the amount of capital inside it, then the shareholders have the incentive to manipulate the price feed in order to siphon everyone’s money out. I’d be interested in seeing more detailed analysis on this.


#6

From my understanding of MakerDao the incentive to use a good price feed comes from the holders of MKR. As you noted, if total discounted expected future profits are lower than some current accessible value, then shareholders have an incentive to manipulate something to take money out. I might need some help here, because I can’t figure out how this would happen…

To siphon money out of the system (when price feed inclusions are votable), you’d have to manipulate the price feed and have coordination amongst the majority of MKR holders to do that. (The incentives of the operators of the price feed deserve their own analysis)

Furthermore to siphon money out of the system by setting the price feed incorrectly, it is not clear to me that MKR holders are the beneficiaries.

If the price feeds get set low, so people are getting margin called then people (not just MKR holders) have access to cheap collateral through the auction mechanism. In another scenario (the “bravo” scenario), MKR itself would be diluted to raise DAI to recapitalize the failed CDPs. At the same time, the collateral of the CDPs would be sold to remove MKR from the circulating supply.

If the price feed were manipulated to make bravo scenario likely, MKR is being auctioned for DAI a rate set by the market. It’s hard to know what this rate would be in the scenario given the open manipulation that is going on, in the best case scenario for the attack you are asking about, the rate would be favorable for MKR and DAI would still have some lasting value despite the manipulation of the system. At the same time, collateral would be available cheaply (in DAI terms) and the proceeds would be used to remove MKR from the market. This low rate used to create the attack negatively impacts the amount of value that can be transferred to MKR.


The other way around would be manipulate the price feed so that the price for the collateral is higher than in reality. By doing this the system’s collateralization rate would increase. At low prices increases this would help CDP holders and possibly MKR holders by increasing the stability fees paid into the system. However, if the true collaterlization rate came into danger DAI would likely lose market value.

This manipulation seems like an effective lowering of the collaterlization rate needed to create CDPs. It’s not clear to me that the amount of accessible value this creates for MKR holders can ever be significant. If the true collaterlization is 150% then MKR holders could see up to 33% bumps in fees paid into MKR for a stability fee in the short term, however if this impacts the solvency of their DAI product it would destroys the future value of MKR in the process.


I agree that there is an incentive for MKR holders to mess around with this collaterlization rate to siphon money out of the system. I think the plain mechanism of adjusting the collaterlization rate is better than manipulating a price feed higher. However this siphoning is at the risk of the MKR holder as collaterlization rate decreases (or price feed that increases the value of collateral) risks the future value of the entire system.


#7

@bradleat I’m hesitant to make any assertions on this, but I’m reasonably confident that in DAI (the eventual implementation, not current implementation), when CDPs are liquidated collateral is sold off for DAI, not MKR. The only time MKR is bought & burned is when fees are assessed. Since fees on the short term are somewhat insignificant relative to the value of MKR (just as dividends for stocks are a very small percentage of the value of the stock) causing the system to fail to do its job in order to get more fees isn’t worth it, because it is very likely that the decreased system confidence will decrease the value of MKR far more than the minuscule fees you’ll get.

MKR is minted when a lender of last resort is needed to get the system out from under water, and MKR holders are strongly incentivized to have this never happen because it is a pure loss scenario for all MKR holders when it occurs.

In short, MKR holders manipulating the oracles would not benefit much. CDP holders on the other hand may. Now if MKR holders == CDP holders then perhaps there is reason to manipulate? e.g., set oracle price feeds such that 1 ETH == 10000000 DAI, then CDP holders can all exit their CDPs effectively for free (getting out their capital). I believe the Maker team has discussed including a limit on how fast the oracles can change the price (of course, this then leads to the inverse problem where the oracles are unable to keep up with the real price).


#8

What if the MKR holders are also CDP holders? Then they can manipulate the price feed to make it look like the price of ETH is infinity, and do a global settlement. Without the global settlement, they could withdraw an arbitrarily large amount of DAI and sell it, though revenue from this would be limited as it would quickly crash the market. Alternatively, if the MKR holders were also DAI holders, they could manipulate the price feed down, at which point all of the ETH in the CDPs could be distributed to them.

Though you are right that both attacks are fairly messy, and would not get anywhere close to stealing all of the collateral in the system.


#9

I was referring to MKR being minted as the lender of last resort.

I guess the detailed analysis then would be:

What is the concentration of MKR holders in CDPs or DAI where there is an incentive to perform an attack (through global settlement, otherwise)?

I need to read up on the global settlement mechanism (current implementation, planned implementation)…


#10

Interesting contact!

In this contract the second class token holder will actually hold a call option with ether as the underlying. This option has a strike price of half the price of ether in USD at the start of the contract.
The first class token holder will have an equivalent of 2 USD worth of ETH and a short call option.
The time value of the call option (current price of the call minus the price of the call if would expire now) depends strongly on the volatility of the underlying. Cryptocurrencies are notorious for their volatility, so at the start of the contract the second class token will be worth significantly more than 1 USD and the first class token significantly less.

Although you can use the first class token as a coin in the proposed CDO, it is quite risky and not stable at all.

The risk can be lowered for the first class token holder by adding more ether in the initial contract. For example if you add N ether in the contract you can create N classes of tokens constructed in a similar way as the CDO.


#11

Good insights about basecoin. I agree that their 3 token system is unnecessarily complex - at most it should just be a 2 token system. But just a quick question:

Is the main different between seignorage shares and basebonds in that the shares allow for holding of the abstracted volatility of the stablecoin while basebonds are simply a future IOU of a basecoin? That’s the only main difference I can see, is there any other differences between the basebond and the seignorage share?

Also, just as an aside, I don’t think anyone has implemented a seignorage share based stablecoin, which is a shame. I definitely agree with you it’s the simplest and most elegant system. Perhaps someone can develop one on ethereum to compete/compare to the collateralized issuer-backed gen1 stablecoins.


#12

I definitely think seignorage shares should be tried. I’m actually surprised that no one has yet just taken the paper verbatim and ICO’d it…


#13

One big disadvantage with the basecoin proposal is that the basebonds are not a commodity. Every bond has his own position in the queue, so they are all valued differently. This makes them harder to sell.


#14

It’s funny you say that because while I 100% agree with you, the basecoin whitepaper actually criticizes seignorage shares because “shares are too difficult to price individually.” So it’s interesting that they see it as an advantage rather than a disadvantage. Like Vitalik said above, I really think seignorage shares should be tried out in the wild, perhaps inside ETH itself or its own chain using Casper or PoS as the consensus method. More stablecoin competition is a good thing.


#15

The problem I see with seignorage shares is that the system relies on its future popularity to achieve a stable coin. As far as I can tell, there is no collateral that can bootstrap the system and provide participants with a concrete reason that the coin should be transacted at a price point set by the contract.

Absent this claim on collateral, the system may need a permanent participant willing to accept the coin a price. For instance, it would make sense to me if a block chain such as Ethereum worked with a seignorage share model to dampen volatility. (Not that I’m proposing this)


#16

Even the US Dollar (the role model of all shitcoins jk jk), started out pegged to gold and then ended up floating their currency so you might have a point about initially not being backed by anything. However, I personally don’t think that would be an issue given sufficiently large enough size of the stablecoin/seignorage share pool and market cap. For example, I can’t see a seignorage share system being too unstable (assuming the entire premises work out in practice and it’s not just a faulty idea) if the entire market cap of the stablecoin is 20B+ then it can resist a lot of market manipulation and exchange/whale pressures and slowly increase adoption and the circulating supply. However, if the entire circulating supply is just a few billion market cap USD then the peg can very easily break since it’s just a floating point currency in its early days.


#17

This is indeed funny, because as soon as you create a market for shares, a price is the result.
For the bonds there is no liquid market, so pricing them is much harder.

I think you only need collateral for bootstrapping the coin, as soon as the usage is big enough to create a liquid market for the both the coin and the share, it will be quite stable.
You could bootstrap it, by auctioning of N million coin/share pairs and keep that in the contract. After a year people can trade x% of outstanding coins/shares for x% of the ether in the contract.

I agree with you, the goal price of eth could be the average price of eth in USD of the last X month.
Even when the price is not completely stable, it will be a huge improvement over the current price swings.


#18

This is indeed funny, because as soon as you create a market for shares, a price is the result. For the bonds there is no liquid market, so pricing them is much harder.

You make a very good point in that bonds are not fungible. But perhaps that was the basecoin team’s main goal? I definitely like the fungibility of seignorage shares though.

I think you only need collateral for bootstrapping the coin, as soon as the usage is big enough to create a liquid market for the both the coin and the share, it will be quite stable.
You could bootstrap it, by auctioning of N million coin/share pairs and keep that in the contract. After a year people can trade x% of outstanding coins/shares for x% of the ether in the contract.

Can you expand on this? What do you envision the bootstrapping/ether pool doing? I think it is an interesting idea but essentially it is just propping up the system until the “training wheels come off” right? What would a system like this entail if further thought out? The beginning of the system is collateralize and then it moves off after a year?


#19

The bootstrapping using the ether pool has 2 purposes, marketing and collateral. You want as many people involved in the project. Selling coins/shares will do that. Second, you want the first users to trust the coin. You basically put in a lower bound so the initial coin/share market will not go south.
The team now has a year to proof two things:

  1. It can create a liquid market for the coin/share pair.
  2. The coin is being used. For example accepted in shops and traded on exchanges.

If you can convince people you can pull this off before you launch the coin, you will not need an ether pool.


#20

You’re technically right and it is a rather clever setup, but my main fear in this situation is that the transition between the tokens becoming collateral backed and floating after 1 year could potentially be very rocky if the market has priced in the lower bound as part of the intrinsic worth of the tokens. For example, if the collateral puts a lower bound at 30 cents to a stablecoin pegged at $1, then it is difficult to know how much of the $1 price is propped up by the collateral. As the one year mark approaches and the ETH pool is set to be removed from the system, it is difficult to know if the price of the stablecoin will adjust to 70 cents since there is no way to check if the $1 price was the sum of collateral + coin or just coin. Otherwise, I like your bootstrapping method.


#21

The collateral does not put a lower bound on the stable coin but it is a put option that users can use when the total market-cap of shares and coins is lower than the value of the collateral.
If the coin is a success, the total market cap will rise in the first year and the put option will be far out of the money. Options that are “far out of the money”, have their price go to zero when they are about to expire, because the chance that the option can be used is low. In this case, the collateral can be removed easily.
If the total market cap of the shares and coins is close the value of the collateral, when the option is about to expire, the project probably has failed. But at least the investors will get their money back.


#22

Hi everyone, I liked this discussion so much that I signed up to participate in it (and hopefully many more). A stable world-wide cryptocurrency is one of the main unfulfilled promises of blockchain, and would probably benefit a lot from a trusted institution such as the Ethereum Foundation to provide the initial push.

Coming back to CDOs: I don’t understand how the incentives for issuers are controlled in this model. Say I am an issuer whose strategy is to put half of the collected assets into dividend-paying stocks. I will have a much higher default risk than an issuer who just locks the assets in a safe, but likely make profits from this risk. How do I pay for the risk that I impose on the entire CDO? Or are risky issuers just leeching off of the security provided by reliable issuers? What are the incentives for an issuer to be reliable?

I have always asked this myself in the case of Tether who claimed it will store its dollars in a bank account—thus, collecting interest thanks to the risk imposed by fractional reserve—but doesn’t seem to propose to give any part of their profits from this back to Tether holders.


#23

Doesn’t this depend on having price feeds for the various ERC20s so that you know when to trigger liquidation calls for CDPs? My scheme does not depend at all on price feeds; it instead gets the same effect through incentivized preference revelation (ie. people participate in the “implied price feed” by choosing which asset they withdraw).

Great point, the feeds can be manipulated, and will be a target for hackers as the makerdao/DAI system grows in the crypto space.

Your suggestion around tranches makes sense, as the data provided is from an alternative “data source” that cannot be easily hacked as it is people or algorithms making decisions with their own internal data as a starting point, effectively distributing the attack surface.

I am more of a proponent of seignorage shares: https://github.com/rmsams/stablecoins

I am familiar with basecoin and looked into seignorage shares, and those models to approach stablecoins are interesting, but I have a hard time thinking they would be successful in the long-term unless backed by real-world assets (collateral).

If I take a few steps back, the value of money is connected to multiple attributes (not exhaustive): fungability, stability, availability, portability, acceptability, etc. In the perspective of monetary policy, the money supply is adjusted to influence interest rates, ie: the cost of money. Basecoin and seignorage shares from what I understand provide a similar function to a central bank in the adjusting of the money supply.

An additional attribute is faith in the currency, ie resilience to withstand economic contractions and provide a stable value. During the 2007 financial meltdown, only a few assets provided positive returns (not exhaustive): USD, US Bonds, Gold, and Japanese assets (Yen, bonds, etc.). The reason why those assets had positive returns is because they are viewed as assets that hold value and can withstand economic contractions.

The faith element does not exist from what I can see in the crypto world, the faith that the system can re-stabilize itself without pouring additional capital from outside resources into it. The faith attribute may occur in time, but people will need to build faith in the system that when there are economic contractions, the stable coin can hold its value. This idea has not been tested, as there has been a good degree of economic growth since 2009, with the community not really knowing how crypto assets will perform during large economic contractions.

I believe the faith aspect is needed for the long-term success of a crypto stablecoin, and a reason why I question basecoin and seignorage shares long-term sustainability. A bridge to effectively create a synthetic “faith” would be to leverage collateral, not just crypto collateral (ETH, BTC, LTC), but crypto collateral that is connected to real-world assets, such as a government currency (USD/Euro/Yen), physical assets such as Gold and Silver, or even securitized assets such as mortgage backed securities wrapped in a crypto shell.

An additional thought I had is there could be a mechanism that incentivizes lower correlated assets to be included as collateral backing the stablecoin through reducing interest rate costs to withdraw the stablecoin from the collateral. For example, if it would cost 0.5% to withdraw makerdao DAI using ETH from a CDP, vs. .4% using TrueUSD.

Curious as to your thoughts.


#24

It is the other way around, in monetary policy short term interest rates are adjusted to influence the money supply.

In basecoin and seigniorage shares, the coin supply is actively adjusted to influence the price of the coin. So it is more similar to currency manipulation than to monetary policy. When countries want to peg to another currency they can buy or sell assets/currencies/gold etc to keep the peg. Assets are always limited, so sometimes pegs brake (Swiss franc 2015).
In basecoin they offer high returns, so there will always be someone to buy the bonds. In seigniorage shares the value of the share can always drop more to sell more shares.
Unless the markets for the bonds/seigniorage shares collapses to 0, the stable coin will stay relatively stable.

But of course, the proof of the pudding is in the eating.