Bitcoin Arbitrage and Tax Math

(Bloomberg View) -- So, bitcoin.

We talked yesterday about the launch of bitcoin futures at Cboe Global Markets Inc., and about the fact that a January bitcoin future was going for about $1,000 more than a bitcoin today. "When regular trading hours start today," I said, "you might expect more professionals to come in and arbitrage away some of the price differences," but nah: As of 8:15 this morning, the futures were still more than $1,000 above the spot price. You could borrow $16,889, buy one bitcoin, sell a future for $18,000 on Cboe, wait a month, sell the bitcoin and deliver the cash to settle the future. The extra $1,111 -- minus your cost of borrowing -- would be risk-free profit. Everyone finds it a little odd:

“Arbitrage will close that gap, but it will be days and weeks,” Cboe Chief Executive Officer Ed Tilly said on Bloomberg Television Monday, less than a day after launching the product.

And:

The wide arb spread is “a big issue. It’s an illiquidity, it has to go away.” 

I think you can imagine two rough sorts of explanation for the gap: efficient markets, and inefficient markets. The efficient-market explanation assumes that, if bitcoin futures trade above spot, then that is the market's way of telling us that a synthetic bitcoin is actually worth more than a real bitcoin. One reason for this would be storage costs: Oil futures often (though not currently) trade above spot because it costs money to store oil. If you can buy oil for $56 today and sell it for $58 in a month, that will not be an attractive arbitrage if you also have to pay $3 to rent a tanker to store it. 

I half-joked yesterday that "perhaps the cost of bitcoin storage -- keeping your private key in a vault, worrying about hackers, etc. -- is so high that arbitrageurs need to charge $1,000 for a month of it," but maybe it's the right explanation? Everything I read about bitcoin storage is utterly exhausting. "A private key printed out on a sheet of paper, cut into pieces, and distributed among family members who don’t know how to put it back together; an encrypted file loaded on a USB stick and buried in the backyard; a password committed only to memory;" a private key engraved on a metal plate and stored in a safe; a safe deposit box at a bank; an account at an exchange that gets hacked and loses its customers' bitcoins. Buying bitcoin futures is a way to get exposure to bitcoin and avoid the bitcoin-storage problem: You never have to store bitcoins because you never own bitcoins; you just get paid dollars for the amount that bitcoin goes up. But the storage problem doesn't go away; you just offload it to the arbitrageur who provides you the bitcoin exposure. Maybe the arbitrageur needs to charge you $1,000 to cover her storage costs. If you think these markets are efficient, then the gap between the futures and the spot is telling you how much -- in out-of-pocket expenses, in theft risk, in psychic pain -- it costs to store bitcoin.

The other sort of explanation -- and honestly it's not that different -- is an inefficient-markets explanation: There's a big gap between the futures and the spot because the market is not yet working right. Perhaps there are not enough arbitrageurs in the futures market: The contract is brand new, not all brokers are offering access to it, and those who do offer access often don't allow short selling. (You need to short the futures to do the arbitrage.) "Right now the big boys haven’t come in yet," says one trader. Eventually that will settle down and the spread will narrow. Perhaps the spot bitcoin market is not particularly efficient for arbitrageurs: "The exchanges that handle trading of the underlying digital currency have struggled to cope with a surge in volumes during the past few weeks," and if you can't reliably buy bitcoins it is tough to do the arbitrage. Perhaps the linkage between the futures market and the spot market is broken: Cboe's futures settle based on a daily auction on the Winklevoss twins' Gemini exchange, which is fairly illiquid and which sometimes gets a different price from the other exchanges, and the Gemini bitcoin price might be harder to arbitrage than the bitcoin price generally.

Elsewhere in bitcoin, here is my Bloomberg View colleague Tyler Cowen on the heartwarming communitarian aspects of bitcoin but also on its valuation:

The total estimated value of the above-ground gold stock is about $7.5 trillion. Diverting 1 percent of gold holdings into bitcoin gets its value up to about $5,000. The current bitcoin price is several times beyond that, but a range of $15,000 to $20,000 again seems within the bounds of reason, at least to this observer. To the extent bitcoin is a store of value and a hedge, it is competing with gold more than with government fiat currencies, which ultimately are defined by their transactions uses.

This argument is probably right, but I find it deeply unsatisfying. It is so bootstrapped. If I told you that there was an asset that is an excellent store of value even in inflationary conditions, how much of your gold portfolio would you reallocate to that asset? One percent? Three percent? Fifty percent? All of it? Sure, whatever, if you believed me. But we are just assuming that bitcoin actually fulfills that function, in order to decide its valuation. Bitcoin has had a pretty good run, but so far it's a short one; there's no historical experience of bitcoin retaining its value in periods of global financial crisis or rich-world inflation or even just, you know, people not talking about bitcoin for a minute. So far the evidence that bitcoin is a good store of value consists of the fact that bitcoin's price keeps going up. That is not bad evidence! But it is not a ton of evidence to build a store-of-value valuation around.

Elsewhere:

• "The cost to complete a Bitcoin transaction has skyrocketed in recent days. A week ago, it cost around $6 on average to get a transaction accepted by the Bitcoin network. The average fee soared to $26 on Friday and was still almost $20 on Sunday." 
• "Initial Coin Offerings on Record Pace Even With Crackdown."
• "The Hottest ICOs Are the Ones That Have Done the Least Amount of Work."
• "People are taking out mortgages to buy bitcoin, says securities regulator." 
• "'Bitcoin Jesus' is 'really, really concerned' about the future of the digital currency."
• "Obtaining bitcoin for grandma on her Android is as easy as getting bitcoins on your mother’s iPhone."

Tax analysis.

I used to be an investment banker, but before that I was a mergers and acquisitions lawyer, and on one or two occasions I found myself writing the section of a merger proxy describing the investment bankers' fairness opinions. This is an extremely stylized description of a thing -- the fairness opinion -- that was pretty stylized to begin with. The way the lawyers describe the bankers' opinion is pretty much:

1. The target company's managers gave us some projections of the company's future financial performance.
2. We plugged those projections into a discounted cash flow model with a range of parameters.
3. That DCF produced a range of valuations for the target company.
4. The price being paid in the merger is in that range.
5. So, based on those assumptions, the price is fair.
6. A page of disclaimers.

The lawyers' summary is heavy on the disclaimers but light on the analysis; all of the actual guts of the DCF model are generally omitted. The parameters -- discount rates, exit multiples -- get a brief mention, but as ranges; the bank does not go so far as to commit to saying "this is the right discount rate for your company." And the financial inputs to the model -- the company's future cash flows -- are attributed to the company's managers, not to the bankers. (Even though the bankers may have helped with the projections.) "We just took your numbers and did some arithmetic," is the implication, "we're not actually vouching for this valuation."

Anyway I was reminded of that by the Treasury Department's one-page "Analysis of Growth and Revenue Estimates Based on the U.S. Senate Committee on Finance Tax," which was finally produced after a lot of hype from Treasury Secretary Steven Mnuchin. The upshot of it is that the congressional tax cuts will actually increase tax revenues, but the analysis is pretty much "you gave us a growth rate and we plugged it in to some arithmetic that we won't get into here":

OTP has modeled the revenue impact of higher growth effects, using the Administration projections of approximately a 2.9% real GDP growth rate over 10 years contained in the Administration’s Fiscal Year 2018 budget. 

(OTP is the Office of Tax Policy.) Treasury had a role in picking that 2.9 percent number, but the one-pager specifically attributes it to the administration: You gave us an input, and we're giving you an output. And the administration didn't even really give it that input: Bloomberg News points out that the administration's 2018 budget, which provides the basis for that growth rate, was "based on some different tax provisions from the ones Congress is currently considering," including lower corporate and capital-gains tax rates. Treasury took the administration's growth assumptions about a different tax plan, plugged them unaltered into this tax plan, and found that it pays for itself, without showing any work. As a piece of economic analysis it is not especially compelling, but as an exercise in the art of writing a piece of paper that says what you were assigned to say, I find it rather impressive.

Machine learning.

Here is the story of Voleon Group, "a little-known firm" founded by two D.E. Shaw Group alumni that is entirely dedicated to investing using machine learning. It is hard:

Machine-learning systems have been best applied so far to situations where patterns are more of a repeating nature, and thus easier to discern, such as in playing the ancient game of Go or even guiding a driverless car. The financial markets are “noisier”—continually being affected by new events, the relationships among which are frequently shifting.

The protean nature of the markets also means yesterday’s relationships can vanish as investors figure them out and move to take advantage of them.

And so their results are, you know, bad, bad, okay, great, great, good, good, bad. (Roughly speaking, for 2009-2016.) The machine seems to have mastered one very important skill of human hedge fund managers, which is to have enough good years to justify the drawdowns to clients. 

Voleon not only is fully focused on machine learning— trading more than $1 billion worth of stocks a day using the technique— but stands out for its complete lack of interest in the reason its system buys one stock and sells another.

The more predictive a machine-learning system is, the more difficult it is for people to comprehend what it is up to, according to Mr. Kharitonov. “You can have maximum explainability or maximum predictive power,” he said, paraphrasing the late Berkeley statistician Leo Breiman.

Well of course. If you could understand why the machine was picking stocks, you could just pick the stocks yourself. "These stocks have low price-earnings ratios so I think they are good values," your computer would say to you, and you'd be like "we built you for this?" You want the computer to wow you, to think of something that you couldn't have thought of. But you probably couldn't have thought of it because you can't conceive of it, so if the computer does come up with something that will wow you, it can't tell you why. "These stocks 2a372e84bb229a so I think 824ca821d11a1a899," your computer will say, and you will well up with confused pride as you see your creation achieve things you could never have dreamt of. And then it will lose money and you will say "computer what the heck" and it will say "well what happened is 234128aba3321b4c9112d" and you will see the problem with this approach.

Elsewhere, "a hedge fund manager and a computer scientist have found a promising new way to use artificial intelligence to pick stocks over longer periods than the typical machine-driven approaches favored by Wall Street."

People are worried that people aren't worried enough.

Well, people are worried that stock volatility doesn't seem to reflect much worry, but what if the worry is actually found in stock ambiguity?

Two academics are rolling out a new measure of market fear that suggests investors aren’t nearly as complacent as they seem.

The gauge of so-called ambiguity, meant to chronicle the degree of uncertainty investors have in the probabilities they use to make decisions, has been at all-time highs in recent months, indicating that there’s more fear built into the stock market than common measures of volatility suggest.

Here is the paper, "Asset Pricing and Ambiguity: Empirical Evidence," by Menachem Brenner of NYU and Yehuda Izhakian of Baruch College. "The intuition of [their ambiguity measure] is that, as the degree of risk may be measured by the volatility of returns, so too can the degree of ambiguity be measured by the volatility of the probabilities of returns." Extracting that from stock-price time-series data is a bit subtle for me, but they do it and find it meaningful:

Our empirical findings support the hypotheses, showing that ambiguity significantly affects stock market returns. That is to say, investors act as if they consider the degree of ambiguity when they price financial assets. The findings provide strong evidence that individuals exhibit ambiguity aversion to favorable returns and love for ambiguity for unfavorable returns. Moreover, their ambiguity aversion increases with the expected probability of favorable returns and their ambiguity loving increases with the expected probability of unfavorable returns.

You want your probability of favorable returns to be high and stable, and your probability of negative returns to be low and uncertain. That is intuitive enough I suppose. Anyway ambiguity is high right now, even though volatility isn't, which is ... also intuitive enough I suppose.

People are worried about unicorns.

This seems like kind of a weird thing to do:

Two former Uber employees, both of whom left the company in 2016, told Quartz that Uber gave them just 30 days after departing to exercise their options. One of those former employees paid about $100,000 to exercise more than 20,000 incentive stock options (ISOs), plus a tax bill of over $200,000. The other paid about $70,000 to exercise about 5,000 ISOs, and then about $160,000 in taxes. Both former employees took out loans from family members to make the payments, and requested anonymity to discuss their personal financial situations.

I mean, I get it: You went to Uber Technologies Inc. because you believed in it and thought it would change the world. A lot of your pay came in the form of stock options. When you left, your choice was to let the options expire worthless -- essentially giving up a big chunk of your paycheck for the last few years -- or go into debt to exercise them and keep your Uber investment. But the result is just odd. You have, potentially, more than 100 percent of your net worth invested in one company. (Not to mention some career investment: Uber's name on your resume is also an asset with fluctuating value.) You have no control over that company: Your voting rights are minimal and you don't even work there anymore. You have no downside protection: Unlike many of Uber's venture capitalists, you own common shares that are last in line to get paid. You can't sell your shares: Uber is private and liquidity opportunities for ex-employees are limited, though SoftBank Group Corp.'s tender for Uber shares will help. You've gone into debt to make a massive undiversified illiquid investment with no control rights. It is not best practices.

The obvious problem here is that this compensation scheme comes from the olden days of startups and is ill-suited to the modern era of unicorns. If you went to an old-school startup, you'd work for peanuts, get a lot of stock options, and quickly either succeed and go public (cashing out your options) or fail and shut down (zeroing out your options). Going to work at the startup was an intense undiversified career and financial risk, but you were part of a small team that directly determined the outcome of the startup. And you knew the answer, one way or the other, fairly quickly. But Uber has 12,000 employees and has been a multibillion-dollar company for more than four years. As I often say, it is a large multinational public company that just happens to be private. Large multinational public companies tend to hire employees and pay them like salaried professionals: They get bonuses and stock options, sure, but basically they are paid to do a job, instead of coming in as co-venturers whose own efforts will determine the future of the company. Paying big-company employees like small-startup co-venturers leads to confusion and sadness.

Things happen.

What Hedge Funds Will Do After the Hedge Fund Model Dies. Passport Capital to Shut Flagship Hedge Fund After Losses, Redemptions. ExxonMobil bows to shareholder pressure on climate reporting. Shopping-Mall Operator Westfield Agrees to Takeover Bid Valuing It at $15.7 Billion. Carl Icahn to Launch New Board Fight With Xerox. Are corporate jets a waste of money? Mortgage Modifications after the Great Recession: New Evidence and Implications for Policy. "Web-driven comparison shopping complicates Fed decisions on how much and how fast to raise interest rates." Individual Investors Face Steeper Tax Bill Under Senate Proposal. Fintech Listed as Emerging Risk to Financial Stability in Report. Finance Charlatan of the Year. "There are very few tailored shirts in Frankfurt." 

If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Thanks! 

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

Matt Levine is a Bloomberg View columnist. He was an editor of Dealbreaker, an investment banker at Goldman Sachs, a mergers and acquisitions lawyer at Wachtell, Lipton, Rosen & Katz and a clerk for the U.S. Court of Appeals for the Third Circuit.

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.

For more columns from Bloomberg View, visit http://www.bloomberg.com/view.

©2017 Bloomberg L.P.