Tuesday, May 28, 2019

Revisiting stablecoins

Source: Gravity Glue (2014)

Cryptocurrencies were supposed to destroy the traditional monetary system. Ten years on, where are we?

Bitcoin has been wildly successful, but as a financial game--not as a medium of exchange. It's a fun (and potentially profitable) way to gamble on what Keynes once described as what "average opinion expects the average opinion to be." But no one really uses it to pay for stuff. It's nature as a gambling token makes it too awkward to serve as a true substitute for banknotes and credit cards.

A number of stablecoins have emerged over the last five or six years. (I first wrote about stablecoins four years ago). Like bitcoin, stablecoins exist on a blockchain. But unlike bitcoin, these tokens have a mechanism for ensuring their stability. Stablecoin owners can convert tokens at par into underlying dollar balances maintained in the issuer's account at a regular bank. So stablecoin entrepreneur have basically built a new blockchain layer on top of the existing financial stack. This is interesting, but not very subversive. It's not that different from what PayPal does, or a mobile money operator like M-Pesa.    

Which gets us to MakerDAO. MakerDAO is the name of the decentralized organization that manages the Dai stablecoin. Dai is unique because like bitcoin (and unlike other types of stablecoins), it has no connection whatsoever to the traditional financial system. So Dai has all the rebelliousness of bitcoin. But unlike bitcoin it isn't a gyrating Keynesian beauty contest. Which means that it has a much better chance of becoming a generally-accepted medium of exchange than bitcoin.

This post is for monetary economists and others who would like to know how MakerDAO works, without necessarily getting into the specifics. Since cryptocurrency jargon, like all jargon, is complicated, I'm going to explain it by comparing it to something we can all recognize, a bank.


The Dai system is in many ways like a regular bank, say Citibank. Citibank create 'stablecoins', specifically deposits, out of unstable assets like personal promises, property claims, flows of future business profits, etc.

The process of creating Citibank deposits begins with a loan. Jim pledges his house that is appraised to be worth $1 million to the bank, and the bank creates $500,000 digital Citibank dollars for Jim. He spends the $500,000 into the economy, which ends up being held by Terry, who is most comfortable investing in safe assets like Citibank deposits. Thus Jim's unstable house has been transformed into Terry's stable deposit.

The creation of Dai tokens works the same way. Jim pledges $1 million in assets to the Dai system, and in return he gets $500,000 Dai. After Jim spends those stablecoins into circulation, they end up with Terry, who wants to hold a stable cryptocurrency.

One difference between Dai and Citibank emerges pretty quick. To get his hands on Citibank dollars, Jim pledges his house as collateral (or some real world instrument, like business inventory or a boat or equity shares). But with Dai, Jim can only pledge assets that exist in blockchain space. Because Dai exists on a particular blockchain--Ethereum--the key pledgeable asset for a Dai loan is Ethereum's native token, ether, a volatile cryptocurrency.

What ensures that Terry's Dai tokens will be worth the same as a Federal Reserve dollar? First, lets revisit why a Citibank dollar is always worth a Federal Reserve dollar.

Citibank maintains a network of ATM machines and tellers that will redeem Terry's deposits at par with paper currency. Since he knows that he can always cash them in 1:1 at a Citibank outlet, Terry needn't ever sell his Citibank dollars at a discount on the open market.

Unlike Citibank, Dai doesn't maintain a network of dollar-filled ATMs. There is simply no way to redeem or cash out of Dai, as there is with other stablecoins. To provide a cash-out mechanism would contradict the whole point of a fully decentralized stablecoin. Dai is trying to recreate a virtual version of the dollar, but entirely within the world of blockchains. It can't rely on out-of-blockchain dollars to secure the system.

So how is the price of Dai kept at $1? 

Let's go back to our Citibank illustration. Imagine that Citibank were to announce that henceforth all its deposits are inconvertible. Its network of ATM machines is to be shut down and cash can no longer be withdrawn at the teller. If Terry can no longer return his Citibank deposits to the bank at par, will their value collapse? Or will the 1:1 exchange rate somehow hold?

The short answer is that the exchange rate will hold... to a degree. Remember that Jim is still obligated to repay $500,000 to Citibank. Even if Terry can no longer directly bring his $500,000 worth of Citibank dollars to Citibank for redemption into Federal Reserve dollars, he can do so indirectly, by offering to sell them to Jim for Federal Reserve dollars, who in turn is obligated to bring deposits to the bank to clear up his loan.

Say that Jim's debt is due and he has decided to take up Terry on his offer. The price that Jim decides to pay Terry for his deposits depends on how many other buyers he must compete with. On any given day, a number of Citibank borrowers will have to purchase Citibank deposits in order to settle their existing debt to Citibank. If they are all anxious to settle their debts, Jim may have to offer Terry as much as $1.05 or $1.06 for his deposits. Again, with ATMs and tellers no longer providing 1:1 convertibility, it is possible for these odd exchange rates between Citibank dollars and Fed dollars to emerge.

Terry isn't the only Citibank depositor. There may be many other depositors who are anxious to sell Citibank deposits that day. If Jim is one of the only buyers, he may be able to convince Terry to accept 93 or 92 cents for each Citibank dollar.

So under inconvertibility, the price of Citibank deposits relative to Federal Reserve dollars depends on the short term demand for deposits and desire to settle debts to Citibank. If there is a large demand to settle debts on Wednesday, and few sellers of Citibank deposits, then the price can spike well above $1. But if everyone wants to sell on Thursday, and no debtors want to settle, it could collapse to well below $1.

The soft Citibank peg I'm describing is exactly how Dai functions. You can actually see below how relaxed Dai's peg is below. Sometimes Dai trades far below $1, sometimes it trades above:

Source: dai.stablecoin.science

This flexibility is not so much a bug, but a feature. It's the same sort of behaviour that Citibank's inconvertible deposits would exhibit.


Returning to our Citibank analogy, there are limits to how far the price of Citibank deposits can stray from $1. When the price of Citibank deposits falls too low, say to 90 cents, then existing Citibank borrowers will smell a deal. They can buy cheap Citibank deposits, cancel their loans (and thus unencumbering their housing collateral), and then proceed to another bank (say Wells Fargo) in order to re-open the same loan (using the same collateral). The whole process of closing and re-opening the loan will result in a 10% profit. The pace of Citibank debt cancellation will increase as a result, thus shrinking the supply of Citibank deposits and bringing its price back up towards $1.

Conversely, when the price of Citibank deposits gets too high, say $1.10, then borrowers will be eager to mortgage their homes with Citibank (and not another bank). After all, they can mortgage a $1 million home with either Citibank or Wells Fargo and get $500,000 in deposits. But Citibank deposits are worth $1.10 which means that a Citibank borrower gets 10% more bang for buck. A splurge in new Citibank loans will increase the supply of Citibank deposits and drive the premium back down to $1.


In addition to these automatic forces that push Citibank deposits towards $1, Citibank can use monetary policy, specifically interest rate changes, to keep the exchange rate between their dollars and Federal Reserve dollars close to $1. 

Say that there are is a glut of Citibank depositors who want to get rid of their deposits, and their desire to sell has temporarily pushed Citibank dollars down to 95 cents. By increasing the interest rate on existing loans, Citibank makes it more onerous for those who have Citibank debt to meet their interest payments. These borrowers will start to buy up Citibank deposits in order to cancel their burden. This wave of buying will counterbalance the glut of depositors who want to sell, pushing the price of Citibank dollars back up to $1.

Citibank can also set monetary policy using the rate it pays to depositors. Say that a horde of debtors have lined up to repurchase and cancel their debts to Citibank, pushing the price of Citibank deposits up to $1.05. By lowering the interest rate it pays depositors, Citibank reduces the incentive that people have to hold Citibank deposits. Depositors will flock to sell, thus pushing the purchasing power of Citibank deposits back down to $1.

MakerDAO manipulates a rate called the stability fee to a level that is consistent with $1 Dai. The stability fee is the rate that Dai borrowers must pay. MakerDAO is in the midst of implementing the Dai Savings Rate. This savings rate provides Dai holders with a reward, much like how Citibank depositors are paid interest.

Does Dai monetary policy work? The price of Dai recently fell to a large 3-4% discount to the dollar. In response, MakerDAO jacked up the stability fee. I documented what happened in this series of tweets:

This effort seems to have successfully brought Dai back to $1.


Under times of stress, how can these systems continue to ensure that the value of their deposits/tokens stays close to $1?

Each inconvertible Citibank deposit is twinned with a lender who will eventually have to repurchase it. Say that Jim and a few other debtors go bust and can no longer pay back their loan. Now there are a bunch of orphaned deposits. This spells disaster for the peg. There won't be enough debtors to repurchase Citibank deposits from Terry and the remaining depositors. And as a result, the price of Citibank deposits will slide far below $1.

But Citibank has a tool to prevent this. Remember that Jim provided collateral in order to get his loan. When Jim can no longer pay his loan, the bank can seize Jim's collateral--his house, inventory, boat, or whatnot--and sell it to repurchase Citibank deposits. All the orphaned deposits can be withdrawn, driving the exchange rate back up towards $1.    

The same goes for Dai. But rather than seizing debtor's houses, MakerDAO takes control of the cryptocurrency collateral that Dai debtors have provided.

Another feature that helps keep Citibank inconvertible deposits near par is the fact that Citibank can always wind down its operations and go out of business. If so, all debtors must settle their debts, which means buying up Citibank deposits and thus cancelling out what is due to Citibank depositors. Debtors who can't pay their dues will have their collateral seized and sold, the proceeds used to pay remaining depositors US$1 for each Citibank deposit.

As long as Citibank has properly appraised the value of the collateral that has been deposited with it, then it will be able to make everyone whole. The proximity of a wind-down, the mere chance that this event can always occur, should be enough to help push the price of Citibank deposits towards $1.

MakerDao also has an equivalent feature called global settlement. Global settlement occurs when the Dai system is shut down and all Dai holders are paid out an equivalent of US$1, with debtors to the system getting all that remains. The odds of global settlement being invoked should help keep the price of Dai close to $1.


There is a lot of skepticism surrounding stablecoins. Folks like Preston Byrne, for instance, are convinced like stablecoins are Dai inherently doomed. And some of them have collapsed. (Just read my old post on the demise of Nubits.)

I'm more sanguine. As I've illustrated, Dai isn't that strange of a beast. Apart from the fact that it is inconvertible, a Dai token is very much like a Citibank deposit. Both Citibank and MakerDAO take unstable assets and turn them into stable-priced ones.

These sorts of water-into-wine institutions have been a regular feature of the financial landscape for centuries. Yes, banks have often failed. But they can also be incredibly durable. Here in Canada, the Bank of Montreal has been operating since 1819, some 200 years, without going under. And for those who attribute the Bank of Montreal's longevity to government sponsorship ad support--nope. Canada only got a central bank in 1935 and a deposit insurance scheme in the 1960s.


Will the new digital upstarts like Dai be able to unseat the incumbents, as many of its fans believe?

Relative to convertible Citibank deposits, inconvertible Citibank deposits really aren't that great of a product. Thanks to Citibank's convertibility mechanism, regular Citibank deposits are fungible not only with Federal Reserve dollars but all other brands of bank deposits including Wells Fargo dollars, Bank of America dollars, JP Morgan Chase deposits, and more. To be fungible means to be perfectly interchangeable.

Harmonization, or interoperability, is pretty useful. People can walk into a store and purchase goods with whatever brand of dollar they want. Neither the buyer nor seller need think twice about which one is being used. But not so with inconvertible Citibank dollars or Dai. Lacking direct 1:1 convertibility into underlying Federal Reserve dollars, the price of these "soft-pegged" versions of the dollar will never be quite the same as other dollars. Any purchase that is made with these exotic dollars would be a bit like walking into a Taco Bell in New York with euro banknotes or Canadian dollars.

I mean, the purchase can still go forward, but there is an extra layer of awkwardness that must be endured. The exchange rate between inconvertible Citibank dollars and Federal Reserve dollars at that instant must be determined, conversion fees must be incurred, and foreign exchange risk absorbed. Likewise with a purchase made with Dai. Sure, Dai tokens are relatively stable. But they aren't fungible with the underlying instrument they are trying to represent--U.S. dollars--and that hobbles their payments functionality.

The same awkwardness occurs when taking out a loan in inconvertible Citibank dollars, say to invest in a business or renovate a house. Businesses and individuals earn income and salary in regular Federal Reserve dollars (and all the other dollars that are interoperable with Fed dollars), but if their loans and interest must be repaid in Citibank dollars, they effectively owe what is a foreign currency.

This undoes one of the most useful features of dollars or yen or pounds, which is that they can be used as general medium for short selling, or put differently, a standard for deferred payment. Standard of deferred payment is "that other function" of money, the one no one thinks about because it is overshadowed by the triumvirate of medium-of-exchange, unit-of-account, and store-of-value.

Briefly, since income is earned in local currency, and income is fairly predictable--especially salaries--a borrower (i.e. a short seller) has a pretty good idea ahead of time how much of their future budget they will be required to pay to cover the bank loan. But when the units borrowed are different from the units that make up most of one's income, all of that pleasurable certainty is lost.


But what about decentralization? Doesn't this feature give Dai an advantage over other types of dollars?

Unlike inconvertible Citibank deposits, which are issued by a centralized financial institution, Dai tokens are decentralized. What does this mean? The organization that maintains the system--MakerDAO--doesn't exist in a fixed physical location. It resides on the Ethereum blockchain, which is maintained by a crowd of validators that is distributed all across the world. Whereas the authorities can easily exert pressure on Citibank--they know its address--MakerDAO's crowd of decentralized nodes cannot be so easily controlled.

Citibank relies on people in offices to do much of the work of running the bank. MakerDAO uses smart contracts: automated bits of code that cannot be tampered with. Governance of the system occurs over the internet, with MakerDAO shareholders voting on resolutions such as interest rate changes. MakerDAO shareholders needn't reveal their identities, which means the authorities can't exert pressure on them as easily they might on Citibank executives.

Decentralization allows for subversiveness. A regular bank is obligated to meet a long list of regulatory requirements including those on how much capital they must hold, customer identification practices, and more. But since the authorities can't easily get a bead on MakerDAO stakeholders in order to punish it for infractions, the Dai system may be able to avoid all sorts of costly regulations. And these cost savings means that Dai borrowers might be rewarded with lower interest rates than Citibank borrowers, and Dai stablecoin holders with higher interest rates than Citibank depositors.

There are a set of actors who are have been censored from the banking system. For instance, thanks to embargo threats emanating from the U.S. Treasury, Iran has been mostly cut off from accessing U.S. banks. American marijuana companies can't get bank accounts because banks consider them to be too risky to serve. MakerDAO is (in theory) much more resistant to censorship than Citibank. Dai tokens can filter into all sorts of unserved and risky markets because those who run the Dai system needn't worry about being punished by regulators.


So there is certainly a natural clientele for decentralized dollars. But whether the benefits arising from decentralization--lack of regulation and censorship resistance--are enough to overcome the awkwardness of non-fungibility remains to be seen.

I also wonder how genuine the decentralization of MakerDAO is. I mean, say that Dai became popular for skirting Iranian sanctions. Wouldn't the U.S. Treasury have a number of levers it could pull in order to reverse this? Many of the MakerDAO developers are public figures, as are MakerDAO shareholders. If the U.S. threatened to arrest them for breaking sanctions rules, would they fall into line and write Iran out of the system? If so, Dai is about as subversive as PayPal or Citibank. A centralized and non-fungible stablecoin doesn't seem to offer many benefits.


  1. I've lately been thinking about what dai could be pegged to in the future. Some have mentioned peggin dai to CPI. I was thinking with crypto you could create a decentralized NGDP futures market (augur kinda did this already) and make dai convertible into NGDP futures contracts (a la scott sumner) and have a completely automated monetary policy?

    Obviously this is far off in the future, if ever, and pegging to $1 probably makes the most sense for the foreseeable future, but interesting to think about.

    1. Yep, they could peg Dai to all sorts of fancy targets.

  2. Great post! I've been pondering on the MakerDAO/Dai system for some time (I think more in the terms of derivatives or synthetic dollars as opposed to bank deposits but it seems your framing works just as well).

    Ultimately, the basic trade that is going on here makes sense. Some people want to speculate and gamble on the asset, and some people don't, but want the asset for some other reason (censorship resistance?). Those who don't want to speculate can construct a derivative that shifts the upside and downside of holding the asset to the counterparty, who gets to go leveraged long in exchange.

    The cost of this trade depends on the supply and demand of people willingness to hold the long position. What a bank (or MakerDAO) does in this situation is act as a third party intermediary to combine all of those who want to go long and short together to create a fungible, stable asset (Dai is fungible with itself I mean, regardless of the CDP that created it).

    Overall it's a very interesting system, but I don't think MakerDAO has been as upfront with the risks as they should have been. There is a limit to these systems - the are highly dependent on the liquidity of the underlying cryptocurrency ecosystem, if there is a big liquidity crunch in the ETH/USD market, I expect Dai to have a lot of trouble.

    1. Ben, good points.

      Yes, there's a couple of ways to think about these systems. I find it more comfortable to think of them as banks, but going the derivatives route works to.

  3. MakerDAO team seems to be eager to move forward with "everything tokenised collateral", including tokenised real world assets. The system is dubbed as Multi Collateral Dai (MCD). This would make the "bank" more robust, but introducing real world assets makes the system inherently incompatible with censorship resistance.

    So I would not be surprised if MakerDAO chooses to become "bank 3.0" and emphasis cryptoanarchy less in the future. As there is more money to be made this way.

    1. I think you make a really good point. Once an Ethereum-based token has a real world connection, then regulators can presumably exercise control over it. If the Dai system wants to accept these tokens as collateral, then it too will have to comply with the wishes of regulators. If not, then the regulators can force the token issuer to prevent the asset from being used as Dai collateral.

      But why would anyone other than hobbyists want to use a regulated Dai system? At that point, it's just a regular bank.

  4. fair comment Ben but I'd point out that the underlying colleteral asset (eth) has seen a 90% price drop and the peg was still maintained...impressive imo.

    As dai moves to multi-colleteral backing, this liquidity risk should be reduced

    1. Oh, I definitely find Dai's holding of the peg from 2018 to today impressive, but the crypto-markets maintained liquidity during that time (the price collapsed, but there was never any
      great uncertainty in the price itself) which is the event that would cause the most problems.

      I actually think multi-colleteral Dai is a mistake. The only native asset on Ethereum is ETH itself, any others suffer from correlated external risks, since they represent assets outside of the ETH system, whatever they may be. Uncertainty in the price of these assets could cause trouble, especially if an outside party (governments) move to constrain them.

    2. The Dai system certainly faces liquidity risks, although this applies to all banks in general. I mean, during a downturn Citibank may have to foreclose on a bunch of properties and sell them. There's a risk that the market for that property type is not liquid.

      I suppose that's where it helps to have a central bank. Citibank keeps a portfolio of liquid securities (MBS, Treasury bonds, commercial paper etc) that it knows it can always sell to the Fed if it needs temporary liquidity. That's not the case with Dai.

  5. As usual a great post!

    I have a question on: 'This [exchange rate] flexibility is not so much a bug, but a feature'.

    I'm not really seeing why. If I want to use a cryptocurrency to trade and am not interested in speculation then dia is clearly better that bitcoin because its less volatile. but wouldn't zero volatility be even better ? What (other than perhaps making it easier to implement) makes this small volatility a feature and not a bug ?

    1. Hi Rob,

      Let me clarify what I meant by that. Commentators often poke fun at Dai because its price varies from $1. I'm only saying that this isn't a flaw or an oversight. That's simply how the Dai system works.

      A stablecoin like Paxos is backed by actual dollars held at a bank. Because Paxos tokens can be converted on demand at par, they're far more stable than Dai. But there's a compromise. By virtue of its connection to a bank account, Paxos is a regulated institution. So Paxos tokens can only be used in certain ways by approved people. Anyone can use Dai. But the corresponding sacrifice is a loss of perfect stability.

  6. You might already have seen this, but I found this recently released paper (https://arxiv.org/pdf/1906.02152.pdf) to be an excellent discussion of decentralized stablecoin systems similar to Dai. You should take a look at it.

  7. JP, what's your opinion to deflationary token?

  8. Great,

    We love your writing skills.
    keep sharing things like this.