Tuesday, February 18, 2014

Not all bitcoin are equal or: One dollar, two prices

A page from  Van Court's Bank Note Reporter and Counterfeit Detector (1843), showing multiple prices for various dollars. Notation: do=ditto, same as above | par=no discount | 20 = 20% discount | 1 = 1% discount | no sale = 100% discount | fail'd=failed bank, 100% discount | clos'd=bank closed, 100% discount

For the past year or so, US dollars deposited at the MtGox bitcoin exchange haven't been considered to be particularly good dollars. The problem is that they are illiquid. Due to a number of reasons (see Konrad Graf), MtGox has limited the ability of users to convert MtGox dollars into conventional dollars issued by the likes of Bank of America, Wells Fargo, and the US Federal Reserve. Withdrawals are slow, uncertain, and red tape abounds.

Current holders of "bad" MtGox dollars would very much like to make their dollar-denominated wealth more liquid. Unfortunately the only reliable route available to them is to exchange their bad MtGox dollars for someone else's bitcoin via MtGox's order book, then move these bitcoin off-exchange in order to purchase better and more liquid US dollars elsewhere. But why would a potential counterparty with spare bitcoin want to engage in this trade? After all, if they do then they'll only end up incurring the very same illiquidity risk that the original owner of MtGox dollars seems so desperate to offload. Better for the potential counterparty to sell their bitcoin for "good" (i.e. liquid) dollars at a competing exchange (like Bitstamp) that doesn't impose dollar redemption hassles.

In order for them to accept the burden of MtGox liquidity risk, potential purchasers of MtGox dollars must be cajoled into the trade by the promise of a large discount. Owners of MtGox dollars eager for bitcoin need to mark down the price at which they are offering their MtGox dollars, or, conversely, they need to mark up the price at which they will purchase bitcoin relative to the price at which it trades on other exchanges.

This, in short, is the most likely reason for the historical premium of MtGox bitcoin over bitcoin on other exchanges like Bitstamp and BTC-e , or, conversely, the discount of MtGox dollars relative to dollars on other exchanges. See the chart below, which shows how the MtGox USD/BTC rate has historically traded at a discount to the Bitstamp USD/BTC trade. (Note that I've flipped the traditional bitcoin price chart upside down to emphasize the dollar-side of the equation). In short, because they are relatively illiquid, MtGox dollars can't purchase as many bitcoin as Bitstamp dollars can.

Now as our chart shows, things have dramatically switched around, with MtGox USD/BTC recently flipping to a massive premium over Bitstamp USD/BTC. I'll get to that in a bit, but before I do so let's imagine the opposite situation to the one outlined above, a limitation on bitcoin withdrawals from MtGox. In this case things would tilt the opposite way. The only way to liberate wealth in the form of bitcoin from MtGox would be to buy MtGox dollars and withdraw them, then buy "good" bitcoin elsewhere. Counterparties would only agree to sell dollars for "bad" MtGox bitcoin if they were offered a steep discount on those coins. The price of MtGox bitcoin would have to be marked down relative to elsewhere, or, conversely, the price of MtGox dollars marked up relative to dollars elsewhere.

Let's imagine another scenario. What if redemptions of *both* MtGox bitcoin and MtGox dollars were halted or at least slowed? Put differently, what if both are equally bad, or illiquid?

In this case, an owner of MtGox bitcoin would see no benefit in offering to buy MtGox dollars at a premium since those dollars would be just as illiquid as bitcoin. Nor would an owner of MtGox dollars have any incentive to buy MtGox bitcoin at a premium, since those bitcoin would be just as unmarketable as dollars. The upshot is that the MtGox USD/BTC exchange rate would show neither a premium nor a deficit relative to the prevailing exchange rate on other markets.

However, the fact that we would no longer see any discrepancy in the MtGox USD/BTC exchange rate relative to the Bitstamp USD/BTC would not indicate that all is normal in the bitcoin universe. The freezing up of MtGox would reveal itself in a market we haven't considered yet: the MtGox USD/Bitstamp USD market. Presumably there is some sort of over-the-counter market on which trusted individuals directly swap the dollar deposits of one exchange for another. With MtGox freezing all withdrawals, we'd expect this exchange rate to be trading at a large deficit, investors willing to pay a much higher price to own liquid Bitstamp dollars. The same goes for the MtGox BTC/Bitstamp BTC market. Normally, this market would trade at par, since a bitcoin held on one exchange should be no different from a bitcoin at another. However, with MtGox frozen up investors would probably prefer to own liquid Bitstamp bitcoin and would pay a much higher price to do so.

Now let's get back to the inversion we see on our chart. There's been some interesting news in the bitcoin universe. It appears that MtGox bitcoin withdrawals have been halted, adding to the already-existing liquidity problems facing MtGox dollar owners. Peter Šurda gives more details here. As I pointed out above, if both bitcoin and dollars at MtGox have now been rendered equally illiquid then the MtGox USD/BTC ratio should not vary from prevailing USD/BTC ratio at exchanges like Bitstamp and BTC-e.

But what we've actually seen is a huge rise in the MtGox USD/BTC ratio relative to that of Bitstamp. The discount has become a premium. This would seem to indicate that though MtGox dollars are still relatively illiquid, they are more liquid than now-frozen MtGox bitcoin. Those with bitcoin-denominated wealth stuck in  MtGox desperately want to trade "terrible" MtGox bitcoin for "less terrible" MtGox dollars in order to flee. They are offering an incredibly high USD/BTC rate to tempt counterparties into taking the opposite—and more illiquid—side of this trade.

Now before you start shaking your head about the strangeness of modern cryptocurrency markets—"how weird is it to have multiple prices for the same money!" —note that we've seen this all before. In the first half of the 19th century, there was a bewildering quantity of different US dollar banknotes, with thousands of banks issuing their own brand. Prior to accepting a certain bill from a customer, a shopkeeper would consult his weekly Bank Note Reporter to determine what sort of discount or premium he should attach to the note.* (See the picture at top). In those days, notes were universally redeemable in a certain quantity of gold. What made things tricky was the fact that redemption could be easy or difficult, depending on how far away the issuing bank was. A note issued by a bank based in rural Pennsylvania and spent in South Carolina would have to take a circuitous and potentially expensive route back to its issuer prior to being cashed in for the metal. This resulted in those notes earning a higher discount. One dollar, multiple prices, was the norm back then.

Now compare MtGox to our rural Pennsylvanian bank. Given the long and circuitous route that both the banknote and the MtGox dollar must take prior to being "cashed out", they both trade at a large discount. Once again we have one dollar but multiple prices. Everything old is new again, it would seem.

*If you're interested, Gary Gorton provides an introduction to Van Court's Bank Note Reporter and Counterfeit Detector (pdf)


  1. Very interesting! I like the issuer discount as a way to think about the price spread between exchanges. I wonder if this is borne out with ripple issuer discounts as well.

    To play devil's advocate, I've noticed that Mt Gox has always quoted a higher USD/BTC price than other exchanges. This was true even in the days of seamless Dwolla integration (the last time I ever used Gox, by the way). Back in those halcyon days, the premium was explained by some as a liquidity premium due to the much higher trading volume available at Gox. I found that explanation plausible at the time. As the liquidity offered on other exchanges caught up with and surpassed Gox, I would have thought the premium would disappear. Maybe it just took some time for the network effects of Gox's early liquidity premium to be eroded. :)

    When Mt Gox started encountering hard times in the press, and yet Gox USD/BTC was still favored, the coindesk muppetry decided the Gox premium was due to the eagerness of the marginal GoxUSD holder to acquire GoxBTC to speed withdrawal of funds from that exchange. I suppose that both explanations could be true (first a liquidity premium, then a dollar illiquidity discount), but I kind of wonder how the market could transition seamlessly between such motivating environments without a more fundamental shift (perhaps such as the shift currently underway at Gox).

    I don't think the Mt Gox USD/BTC premium can really be explained very well during any time period as a market discount. In the pre-malleability crisis days (say, a month ago), if Gox $ account holders were in such a rush to get out of Gox$, I would expect that Gox $ deposits would eventually go to zero, and would not be replaced by new deposits. There would be an initial rush to the GoxUSD exits, which would explode without an upward bound until no depositors were left. Instead, GoxUSD has collapsed vs. GoxBTC, on quite healthy trading volume.

    1. Hi Ezekiel, thanks for popping by. I wasn't aware that MtGox USD/BTC was once quoted at a higher rate than elsewhere. My chart shows a lower rate going back to late 2012. I suppose I should have extended it back further in time in order to get more of a flavor for the full life of the premium/discount.

    2. What I meant is the USD price of Gox BTC was higher, as it was just a few weeks ago. Not the reciprocal.

    3. Ok, I got you now. Some thoughts:

      "Back in those halcyon days, the premium was explained by some as a liquidity premium due to the much higher trading volume available at Gox."

      Even though Gox had higher trading volumes back then, my thinking on this is that this would have simultaneously improved the liquidity of both MtGox dollars and MtGox bitcoin. Which would have added an equal premia to both, this having a canceling effect such that the ratio of MtGoxBTC:MtGoxDollars should not have deviated from its previous level or the level seen in the wider market. (However, we would have seen a premium develop in MtGoxDollars:BitstampDollars and in MtGoxBTC:BitstampBTC.)

      So I'm partial to the "speed of withdrawals" argument, since that effected only one half of the MtGoxDollar:MtGoxBTC equation.

      "I would expect that Gox $ deposits would eventually go to zero, and would not be replaced by new deposits"

      I'd argue that deposits were being replaced, specifically by MtGox dollar owners discounting their deposits to a low enough level that they could motivate outsiders to buy those deposits with bitcoin.

    4. But there was no delay in dollar withdrawals during the time when Gox had good Dwolla integration. Yet there was still a premium (BTC/USD, if I'm not having a dyslexic moment again). How do you explain that?

      Also find it hard to believe that USD withdrawal delays would result in a consistent premium vs. other exchanges. I would expect for the value of GoxUSD to spiral downward much more rapidly. Yes, GoxUSD holders were offering a discount, but there must have been new money being deposited in Gox at the same time. New depositors were seeking GoxUSD, for reasons that I don't understand. Is it possible that Gox is a more convenient exchange for certain customers? Either due to geography, scale, or other reasons?

    5. "But there was no delay in dollar withdrawals during the time when Gox had good Dwolla integration. Yet there was still a premium (BTC/USD, if I'm not having a dyslexic moment again). How do you explain that?"

      Good point. I'm puzzled. But I don't think the higher trading volume theory can explain it. Gotta be something else.

      "New depositors were seeking GoxUSD, for reasons that I don't understand."

      Certainly strange. New depositors were basically selling higher value Fed dollars or banking dollars for bad Gox dollars when they could have purchased better Bitstamp dollars. The FedDollar:MtGoxDollar exchange rate should have been above 1 in order to motivate people to hold bad MtGoxDollars. But it traded at par, indicating that people were willing to make the trade. Perhaps it was just ignorance that drove everyone to MtGox? Good advertising?

    6. Maybe there is some feature of Gox that explains the seemingly irrational appetite for Gox$ -- the trading API, the way the Gox trading algorithm interacts with HFT strategies, banking convenience for Japanese customers...

    7. Based on events of the past week, I will add two more hypotheses: fraudulent manipulation of the exchange rate by insiders; or, accurate reporting of a premium that was created by the need for gox btc deposits to fuel a fraudulent scheme (either perpetrated by insiders, outsiders, or both)...

  2. JP:
    "A note issued by a bank based in rural Pennsylvania and spent in South Carolina would have to take a circuitous and potentially expensive route back to its issuer prior to being cashed in for the metal. "

    I've always wondered about this. Textbooks usually talk about how there were 5000 banks in the US issuing 8000 different kinds of notes, and it sounds chaotic. But of course there were also maybe 5000 towns in the US. In any given town there would have been only a few banks, so only a few kinds of circulating notes. Not chaotic at all. Of course a Pennsylvania note would be hard to spend in South Carolina, but how often would a Pennsylvania note make it that far from home?

    1. Mike, the bank note reporter I linked to quotes note prices at Philadelphia. (Prices would have been different in New York) Given the fact that it provides quotes on notes from all over North America, I'd assume that notes often circulated far from the town of the issuing bank.

    2. Mike,

      branch banking was forbidden in the 19th century US, that might explain why there was such a wide variety and spread of bank notes.

    3. Certainly branch banking prohibitions made the situation worse than in Canada, where all notes exchanged at par. But according to his own research, Selgin claims that by the eve of the Civil War, the total discount on all the notes of Union banks would have been less than 1% in the NY or Chi markets.

      http://www.youtube.com/watch?v=JeIljifA8Ls (from 5:00)

    4. JP, the # of circulating notes in Phi was prob somewhere in between. Gorton writes on p. 13 that it was unlikely that all the listed notes were actively traded; instead, Van Court strove for complete coverage of any note that might possibly be encountered.

      Interestingly, even by 1858 the mean discount for even very distant notes (e.g. Carolinas, Montana, GA, KY) was often quite low (around 1-3%). For states with really strong banking systems (e.g. MA), discounts were consistently very low (less than half a percent).