Wall Street

Bitcoin Exchange Had Too Many Bitcoins

It's enough to make you wish for a blockchain. Oh, wait.

This is what the hard fork looks like, inside the blockchain.

Photographer: Manuel Romano/NurPhoto/Getty Images

A few months ago we talked about a weird legal dispute over the Dole Food Co. buyout. Dole's chief executive officer, David Murdock, had taken it private for $13.50 a share in 2013, but shareholders thought it was worth more. So they sued, and won, and Murdock was ordered to pay shareholders an extra $2.74 a share plus interest, and shareholders were told to submit claims for their money.

What Bitcoin's Split Means for Traders

But there was a problem: People submitted more claims than there were shares. This turned out not to be fraud, or carelessness: People really owned more shares than there were shares! It's just that other people owned negative shares. In rough numbers, there were 37 million shares outstanding, and people owned 49 million shares, but other people were short 12 million shares. The way short selling works is that X borrows a share from Y and sells it to Z. So Y owns one share, and Z owns one share, and X owes one share, and everything balances out and there's only one share outstanding. So the millions of extra shares made complete sense.

But that doesn't answer the question of what to do about it. When the buyout closed, back in 2013, it was straightforward enough: If you owned a share, you got paid $13.50. If you were short a share, you had to pay $13.50. So shareholders got paid 49 million shares times $13.50 a share, and short sellers paid 12 million shares times $13.50 a share. (And David Murdock paid 37 million times $13.50 for the shares he was buying. 1 ) But in 2017, it was more complicated. If you owned one of the 49 million shares back in 2013, you were due the extra money in 2017. If you were David Murdock and you bought 37 million shares in 2013, you had to pay the extra money in 2017 -- but only on the 37 million shares you bought. But what if you were short the stock in 2013? Was your obligation discharged by paying the $13.50 in 2013, or are you still on the hook to come up with more, four years later?

The answer is a mess. The Delaware judge who heard the case sort of punted this issue to the Depository Trust Co., which keeps track of all the shares of all the companies, and told DTC to follow its procedures to figure it out. The practical answer seems to be that the short sellers' brokers were on the hook to come up with the extra money, and that the brokers' ability to go after the short sellers depended on the margin and stock-lending agreements they had with those sellers. 2 Anecdotally, it seems that the brokers have mostly tried to seek payment from their customers who were short -- and that some of those customers feel pretty aggrieved about it. You can see why: They closed their trades four years ago, and whatever irregularities occurred in Murdock's buyout of Dole weren't their fault. Why should they have to pay for them? Even beyond the arguable unfairness, it is just administratively messy: Someone has to find all those short sellers. If you were short Dole shares in 2013, and subsequently closed your account, wound up your fund, or died, how would a broker get you to pay up?

It's all such a mess, in fact, that I wrote: "It does seem like a half-competent blockchain would be faster and cheaper and more transparent" than the messy current system of share ownership. Just, you know, blockchain it up, keep track of who owns what and who borrowed what from whom, and have a permanent legible record to keep track of these weird webs of contingent obligation.

Ha ha ha, what a fool I was. Here is an announcement from Bitfinex, a bitcoin exchange, that is mind-blowing and wonderful and far weirder than anything a Delaware court could come up with. It has to do with Tuesday's hard fork of bitcoin, in which each holder of a bitcoin ended up with both (1) the original bitcoin, on the original bitcoin blockchain, and (2) a new bitcoin, on a new "Bitcoin Cash" blockchain, which is trying to become a viable alternative flavor of bitcoin. 3 (The convention seems to be to call original bitcoins BTC, and the new Bitcoin Cash bitcoins BCH or BCC. I'll use "BTC" and "BCH" here.) Jian Li writes:

To use an imperfect analogy from corporate finance, you could think of the fork as a spinoff. For most of PayPal’s life, it was owned by eBay. Holders of the EBAY ticker owned the parent company eBay, which encompassed eBay proper as well as PayPal. On the day of the spinoff, eBay stockholders received, for each EBAY share they owned, one PYPL share. At the same time, they got to keep their existing EBAY shares.

Something a little like that is going on with the bitcoin fork, although it is a bit stranger metaphysically. 4  It is also a bit stranger economically. In a spinoff, you'd expect the original company's value to drop by roughly the value of the spun-off company, which after all it doesn't own any more. 5  BCH spun off from BTC on Tuesday afternoon, and briefly traded over $700 on Wednesday (though it later fell significantly). But BTC hasn't really lost any value since the spinoff, still trading at about $2,700. So just before the spinoff, if you had a bitcoin, you had a bitcoin worth about $2,700. Now, you have a BTC worth about $2,700, and also a BCH worth as much as $700. It's weird free money, if you owned bitcoins yesterday.

But what if you owned negative bitcoins yesterday? What if, that is, you had borrowed bitcoins in order to sell them short? Well, in stock lending situations, the normal way that this works is that the short sellers (stock borrowers) have to come up with whatever is distributed on a stock. If you are short a stock and it pays a $1 dividend, you have to come up with $1. If it spins off a subsidiary, you have to go out and buy a share of the subsidiary to deliver back to your stock lender. If it is acquired in a leveraged buyout for $13.50, you have to come up with $13.50. If it distributes a pony to each shareholder, you have to come up with a pony. 6

You could imagine bitcoin lenders taking the same approach: If you were short a bitcoin going into the fork, now you have to deliver one BTC and one BCH to your lender. Or not! In fact, when bitcoin distributed a pony of indeterminate value to its holders, Bitfinex decided -- not unreasonably -- that it would be unfair to make bitcoin borrowers come up with it. The value of Bitcoin Cash is uncertain and volatile, and forcing bitcoin shorts to go out and buy Bitcoin Cash to cover their shorts might create artificial demand for it and push up the price. So Bitfinex announced, last week, that short sellers would not have to come up with any BCH. 7

This creates a problem: If people are long 125 bitcoins, and other people are short 25 bitcoins, then there are a total of 100 bitcoins at the exchange. If there are 100 bitcoins, then 100 BCH will be distributed on them. But if people own 125 of those 100 bitcoins, and if you get only 100 BCH, and if the shorts don't have to come up with the shortfall, then you can't give one BCH to each bitcoin holder. One option here would be to just not give them any BCH, and ignore the whole thing, which seems to be the approach that several bitcoin exchanges took. 8  But Bitfinex took the more customer-friendly -- though pretty ad hoc -- approach of just divvying up the BCH evenly among all the long holders of bitcoins. With my stylized numbers, if people were long 125 bitcoins and short 25, then each long holder would get 0.8 of the 100 BCH distributed to Bitfinex, and the short holders wouldn't have to come up with anything. 9  

But this creates another, funnier problem: That's so easy to game! Here's what you do:

  1. Set up an account, borrow one bitcoin, sell it short, collect $2,700.
  2. Set up another account, buy a bitcoin, spend $2,700.
  3. When the fork happens, your long account ends up with +1 BTC and +0.8 BCH.
  4. Your short account ends up with -1 BTC and -0 BCH (because Bitfinex doesn't require you to come up with the BCH).
  5. Net, you have $0, 0 BTC and 0.8 BCH.
  6. The 0.8 BCH were worth as much as $560.
  7. That money was totally free. 10

This is such a dumb obvious arbitrage that lots of people tried it. Bitfinex was not happy. "After the methodology announcement on July 27th, several accounts began large-scale manipulation tactics in an attempt to obtain BCH tokens at the expense of exchange longs and lenders on the platform, causing the distribution coefficient to artificially plummet," said the exchange's Wednesday announcement, which claimed that "this kind of manipulation - including wash trading and self-funding shorts - is in violation of Bitfinex’s terms of service." So:

Upon careful review and analysis, we have decided to disallow any hedged BTC balances in excess of any such hedged balances that may have existed at the time of the July 27th distribution announcement. While this may be disappointing to some, it is welcome news to the many users with bona fide BTC exposure through settled wallet balances. This adjustment increases the distribution coefficient from 0.7757 to 0.8539.

Umm? That strikes me as fair, I guess? 11 It is also a mess, though: Bitfinex announced a policy to deal with the fork, 12  people took advantage of the policy, and Bitfinex changed its mind after the fact. Each of its decisions was rational, and quite plausibly the fairest option available to it. None of those decisions were required by, like, the nature of bitcoin, or of short selling: There is no single obviously correct solution to these issues. Instead, each decision was sort of weird and contingent and reversible: not the immutable code of the blockchain, but just humans sitting around and trying to figure out which approach would cause the fewest complaints. In that, it's a bit like the Dole settlement process -- only instead of a neutral judge making decisions based on written contracts and established precedent, 13 it's the people running each exchange making their own judgment calls. 14

Part of me still stands by my dumb statement about Dole that a blockchain would help. It would help here too! A ledger of who was long and who was short before the fork would perhaps make sorting out these issues easier after the fact. But the bitcoin blockchain doesn't work that way: There's no primitive way to short bitcoins into the blockchain, so exchanges exist in part to provide lending services for people who want to be short. 15 The blockchain has a certain stark logical completeness, 16  but it doesn't address all of the actual human uses required of it. And so it has become encrusted with other human institutions. 17 And those institutions turn out to be unsurprisingly human.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
  1. I am oversimplifying this in several ways; there were both a fiduciary-duty lawsuit and an appraisal lawsuit, and the class at issue in the case we are talking about excludes the appraisal claims. So actually Dole had more than 37 million shares. But none of this matters much for the story.

  2. The Master Securities Loan Agreement governs some of these transactions, though presumably many of them are instead governed by brokerage firms' margin agreements. Section 8 of the MSLA gives a frankly rather unsatisfying answer:

    Lender shall be entitled to receive all Distributions made on or in respect of the Loaned Securities which are not otherwise received by Lender, to the full extent it would be so entitled if the Loaned Securities had not been lent to Borrower.

  3. The reason for it is about block size and transaction speed and so forth, and about who gets to control bitcoin's future, but whatever. Here is some background from Nathaniel Popper. Here at Bloomberg View, Leonid Bershidsky endorsed the hard fork.

  4. Did your BTC last month also encompass a BCH? In some obvious sense, no: The fork is not separating two things that were previously together but rather creating a new competitor. And yet in some other sense, sure, why not: One old BTC represented the total present and future economic power of the bitcoin network, and now that that network is splitting in two, that economic power is also splitting into BTC and BCH.

  5. Li writes:

    After the spinoff, the price of EBAY dropped from around $66 per share to $27 per share, an amount close to the $41 value of the new PYPL shares. So for each $66 worth of stock, you ended up with approximately $68 worth of stock.

    You'd expect the sum of the parts to be worth a little more than the original stock, since the whole point of the spinoff was to create value! But not much more: Just rearranging things that already exist shouldn't vastly increase their value.

  6. Again, Section 8 of the Master Securities Loan Agreement covers distributions, in perhaps less detail than you might like.

  7. Its stated reasons included:

    It would be an undue burden to create negative BCH balances for BTC/USD shorts and XXX/BTC longs. Furthermore, lenders may be locked in for up to 30 days and would be adversely impacted. In the end, we believe that this solution creates the least amount of disruption and facilitates a timely distribution of BCH.

    It also called BCH "a spurious forking of the Bitcoin project with little mining support," which perhaps influenced its decision not to force short sellers to buy it. I cannot vouch for the factual accuracy of that statement, but if it's true, it's a good  reason not to support it.

    Imagine if I announced tomorrow that I had created a new blockchain, called Bitcoin Matt, and that everyone who owned a BTC today will tomorrow own both a BTC and a BCM. Fine, great, you all own BCMs, congratulations. But also anyone short a BTC today will be short a BCM tomorrow, and will be forced to go buy in those BCM shorts. Even with no economic support for BCM -- with nobody mining it, or using it, or treating it as a store of value -- I have magically created demand for it, just because BTC short-sellers will be forced to buy it in to cover their shorts. And if no one else is using it, then it will trade very thinly, and it will be very expensive to cover. (And anyone who does sell it will make a lot of free money.) Nothing stops me from just announcing that I've cloned a copy of the bitcoin blockchain for BCM; the only way to avoid this abuse is for people to ignore it -- and that means not forcing short sellers to cover it.

  8. From CNBC: "Several exchanges, such as BitMEX, Bitstamp and Coinbase, have said they will not support or allow trading of Bitcoin Cash on their exchanges, which means investors holding bitcoins on these sites will not receive any new tokens." Presumably those exchanges hold BTC on behalf of their customers, and will now get BCH that they won't give to those customers. What will they do with them?

  9. Bitfinex's method of dealing with margin accounts is slightly more complicated than that but that's the general idea.

  10. Actually you paid some interest to borrow the bitcoin in step 1, but still.

  11. Honestly the outcome here seems a little fairer than what happened in the Dole case? Really charging the shorts there was rough, and just divvying up whatever you got among the longs would have been fine.

  12. Which was different from its previous policy for dealing with forks, which would have required short sellers to come up with the BCH!

  13. I mean, to be fair, the judge mostly punted the question to the brokers anyway. It's a lot like the Dole process, actually. Except much faster!

  14. Oh, by the way, because each exchange was making its own decisions, even dumber arbitrages were available. Reader Tom Lefevre emailed me to point out that, beyond the hypothetical Bitfinex arbitrage, "there was a real executable arb opportunity that was created because the exchanges did not coordinate on their Bitcoin Cash fork allocation treatment." Some exchanges ignored the fork, while others announced that they would distribute BCH to their holders. So the trick was to borrow and short bitcoins on one of the former exchanges, while simultaneously going long bitcoins on one of the latter. Then, after the fork, you'll be long BTC and BCH at the second exchange (which implemented the fork), but short only BTC at the first exchange (which ignored it), giving you free BCH.

  15. Also: To convert bitcoins into usable currencies.

  16. Though not that much! After all, it is forking because people can't agree on how it should work!

  17. There is an obvious parallel to the fact that the blockchain is meant to be secure and decentralized and trustless, but lots of people hold their bitcoins in exchanges and wallets that require them to trust a centralized intermediary -- and those intermediaries always get hacked.

To contact the author of this story:
Matt Levine at mlevine51@bloomberg.net

To contact the editor responsible for this story:
James Greiff at jgreiff@bloomberg.net

Before it's here, it's on the Bloomberg Terminal.