Off and on, friends ask me why I’m not working at or running a crypto hedge fund. I’m interested, worked in the industry for a while, like to think about esoteric financial topics, etc. But I’m not sure what I’d actually do as a full-time crypto investor. Currencies are a Schelling Point, so unless you think Bitcoin will collapse, the correct trade for a crypto fund is to be long Bitcoin, at which point there are only two questions for the manager to answer:
- How long?
- Long how?
I’m long Bitcoin in my personal account, and consider it an extremely speculative position. However, it’s also a unique asset. For a typical asset allocator — an institution trying to own a basket of assets with favorable risk-adjusted returns — the right cryptocurrency allocation is close to zero, but nonzero.
Asset allocators have to fit every position into their portfolio — to think about risk/reward and correlation to other asset classes. I’ll address both concerns in this piece.
The Bull Case for Bitcoin
Bitcoin is a store of value that got invented backwards.
There are many theories of money, but the most comprehensive is that a money is a Schelling Point: money is that which you treat as money because other people treat it as money. Sort of the same way any word is a Schelling Point; if you decided not to know what “you” meant, you’d have trouble communicating and trouble being communicated with. If you come up with a better “you,” it had better be much, much better, or nobody will adopt it.
“You” is a good example because its current meaning is a somewhat recent innovation. Prior to the 16th century, people said “thou” informally and “you” only when speaking to a social superior. But — back to game theory! — nobody is offended if you’re slightly more respectful than they anticipated, and everyone is offended by blatant disrespect. When social mobility increases, it’s hard to guess how important someone is: maybe the person you’re talking to has a regional accent, but they might still be rich; maybe they’re dressed shabbily because their social status precludes work but they’re economically downwardly mobile. The ubiquitous “you” started out in London, ground zero for New Money and newly-impoverished Old Money, and spread everywhere from there. We hear plenty of “thou” in Shakespeare, but William S. himself was exactly the sort of person who was born a “thou” and died a “you.”
Schelling Points exist because they’re stable given outside circumstances, but those circumstances can change.
For most of history, precious metals were the ideal Schelling Point for currency. Gold and silver are both hard to find in large quantities, and after a few “eureka!” moments we got good at measuring their quality. The benefit of precious metals is that everybody knows they’re worth something and nobody’s equipped to make more. The problem is that if wealth grows faster than the money supply, you wind up with deflation — you can’t mint money fast enough to reflect the abundance of goods and services available, so prices decline and borrowers get pinched.
As I wrote the other day, banks solve this by minting certainty into money. When a borrower can demonstrate that their future income is sufficient to service debts, the bank lends money, which creates money. So the total amount of money available reflects the banking system’s view of the present value of all likely future production.
This is fine, unless the banks are systematically wrong. A big problem with credit-driven growth is that systematic errors tend to hide other systematic errors: if people are paying off credit card debt with money they’ve extracted from home equity, and that allows them to keep spending consumption steady which lets companies service their own debts, then you have an unstable equilibrium. By the same token, a collapse in one financial market can quickly infect other markets, which makes investors more cautious and makes valuations stupidly cheap. Just look at the ValueInvestorsClub.com archives from early 2009. IAC was trading at $13.50/share, with $12 in cash! Oracle was trading at a low-teens P/E! Oshkosh was trading at a P/E under 4! There were wide spreads even in zero-risk trades, like buying super-voting stock.
Central banks and short-sellers provide some ballast, although to be accurate about the relative impact of these groups, you’d want “Central banks” in 72-point font and “short-sellers” in letters approximately one planck length tall.
If you’re worried about the market overshooting, the traditional hedges are treasuries and gold. Treasuries outperform as a deflationary safe-haven, and correlate inversely with equities when inflation is low. Gold outperforms during times of high inflation (because it’s hard to make more of it) and during times of deflation (because other assets’ returns get crushed, and the opportunity cost of owning it is low).
In the near term, Bitcoin fits into the gold framework. Its supply is limited, by design; it doesn’t produce a return, so its opportunity cost is low when rates are low, but in the event that inflation accelerates, the lack of additional supply should cause it to outperform.
If you buy the historical argument for safe-haven assets, you’ll treat gold as the safest haven. If you buy the theoretical argument for why gold is a safe-haven, Bitcoin is slightly more reliable, albeit riskier. There’s regulatory risk to owning Bitcoin, since it can get confiscated, but Bitcoin was created with that risk in mind.
If you think there’s a nonzero chance that Bitcoin will become a safe-haven, you have to buy the argument that it will become a major one. The drawback Bitcoin has relative to gold is that gold is more valuable and more stable; if Bitcoin rises in value, that solves one problem; if that rise in value is because asset allocators and central banks are allocating some of their rainy-day money to Bitcoin, it should be more stable.
Paradoxically, the safe-haven theory makes a ban more likely, because it means every government and every investor is short Bitcoin, so they have an incentive to squash it. We’ll get to the likely event path later.
Bitcoin’s Valuation and Comps
Since the only Bitcoin story that matters is the successful version we can dispense with any comps to other crypto-currencies and unprofitable tech plays. The right comps for Bitcoin are gold and the US dollar. This might sound like an extreme claim, but the way to tie it to reality is to note that Bitcoin’s value is something like (Gold + USD) * (% chance that Bitcoin becomes the default global savings vehicle). The second term in that equation is a small and volatile number, but the end goal is what we’re calibrating valuation against.
There are precedents for this. Back in 2009, Paul Graham thought Facebook’s closest comp was the TV industry. FB’s revenue run-rate at the time was under half a billion dollars a year, compared to around $70bn in US TV ad revenue at the time. While the businesses didn’t look anything alike, both Facebook and TV are fundamentally in the business of converting time-spent-staring-at-screens into advertising revenue. Facebook was smaller, but it was the option to bet on.
The world’s current gold stock is around 190,000 metric tons, give or take. At $1,470/troy ounce, that’s roughly $9 trillion worth of gold. You can throw in the Swiss Franc or the US Dollar as additional comps — CHF is more comparable because more people hold it purely so they own an asset that goes up when everything else goes down, and USD is more comparable because it’s the default thing you end up owning when you sell anything else.
Central banks hold around $6.8tr in USD reserves, and a comparatively trivial sum in CHF, so you could add that to the value of gold and conveniently almost-double your comp.
However, when you move further afield from gold, the comp gets harder to justify. It’s rare to borrow money in gold terms (although it used to be common; when J.P. Morgan bought Carnegie Steel, he paid in gold bonds; France issued the Giscard bond, to their immense regret; and Turkey issues some as well). It’s useful for central banks to hold dollars, in case either the government or corporate borrowers run short. Dollars, then, are a hedge against specific problems; gold is a hedge against unknown problems.
Bitcoin would be comparable to dollars if companies and countries routinely borrowed Bitcoin, but that’s only something speculators do today, so it’s not worth considering. By the time it’s a possibility — a small one, given that gold-based borrowing is rare — circumstances will be so different that it’s not worth speculating about today.
So, gold makes sense as an asset that is both analogous to Bitcoin (few intrinsic uses, scarce, comparatively easy to transfer) and used as Bitcoin could be used (hedge against extreme changes in asset prices).
In this framework, valuing Bitcoin is simple. Not easy, but simple: if you have a required rate of return, a date on which you expect Bitcoin to replace gold, and an estimate of the odds of this happening, you can easily back into a present value. For example, if you think Bitcoin has a 1% chance of replacing gold in ten years, its expected future value in 2029 is $90bn. If you expect Bitcoin to have the same historical Sharpe ratio as equities, and expect equities to deliver a 3% return above the risk-free rate (lower than the historical average, but we should be cautious here), then Bitcoin’s ~50% annual volatility compared to the S&P’s ~10% implies that Bitcoin needs to return around 18% per year.
So, discount $90bn back ten years at 18%, and you get a fair value of… $17bn. But that ignores inflation. Bitcoin’s inflation is fixed, but it’s not zero. Over the next ten years, Bitcoin supply outstanding will increase roughly 16%, so we need to further haircut our number to about $15bn.
This $15bn price target compares relatively unfavorably to Bitcoin’s current market value of $157bn.
But don’t despair!
There are three major levers we can pull:
- Perhaps the odds of Bitcoin replacing gold are greater than 1%.
- Perhaps our expected return is too high. An expected return of 18% assumes that Bitcoin remains just as volatile over the next ten years as it is now, but high volatility is incompatible with reserve-asset classification. If Bitcoin becomes more of a reserve asset, there will be more natural steady buyers, and it will be harder to move the price.
- We’re setting a hurdle rate based on equities, but in the long term Bitcoin shouldn’t behave like a stock. It’s more of an out-of-the-money option on something that will turn into gold.
That allows us to make more generous assumptions. Suppose we assume Bitcoin’s odds of replacing gold are 5%, rather than 1%, and we split the difference on required rate of return — 8% above the risk-free rate, rather than 16%. That gets us to a $170bn price target, or a bit above today’s valuation.
The odds-of-replacement estimate does all the work, here, and long-term Bitcoin investing requires an investor to continuously reassess those odds.
How Bitcoin Fits Into a Portfolio
There’s an old cliche that markets climb a “wall of worry” — when everyone is scared that something bad will happen, equity holders get paid every day that it doesn’t.
Bitcoin is different: it climbs, and descends, a cliff of confusion. Nobody truly knows what’s going on, and everyone who claims to is selling something. People have tried to value it, but the answers round to either zero or infinity. Either Bitcoin is a Ponzi scheme, a Dunning-Kruggerand, a pure vehicle for speculation without even the shininess of gold — or it is gold, only easier to hide, easier to trade, and thus worth more.
There are even contradictions embedded in the bull case. Some people talk about the pace of product development, the elegant solutions to long-standing CS problems, the sheer braininess of Satoshi and the core devs. Other people talk about it as a sort of high-volatility store of value, the kind of asset you buy when you think every chart is going down except the money supply.
So, is it a share of Superhuman stock, or is it the Swiss Franc? There’s a difference.
Specifically, in a portfolio context, there’s a difference between Bitcoin as a hyper-optimistic tech play and as a pessimistic emergency asset. If you’re overweight risky assets, which Bitcoin position balances that out: long, or short?
Why It Isn’t Obvious Whether Bitcoin is Risk-On or Risk-Off
First, let’s take a step back: what do we mean by risk-off and risk-on? I first started hearing the term regularly after the financial crisis. If you look at a ten-year chart of the S&P, you see a steady progression with some wobbles, but for the first few years after the crisis, every single wobble felt like enough to give you a heart attack.
We had the ongoing aftereffects of the crisis — chronically high unemployment, with a historically low recovery rate:
Source: Business Insider, The Scariest Jobs Chart Ever
(Stay tuned for a future post on this: recovery rates were faster when GDP volatility was driven by manufacturers’ inventory levels, so a services economy tends to have slower drops and slower recoveries.)
We also had the ongoing balance sheet effects, most notably in Europe. “Black Swan” insurance was painfully expensive. As Mike Green noted in an interview a while back:
>if you look at 2007 when Nassim Taleb published The Black Swan, if you’d gone into the market and you tried to purchase a put option that paid out in the event that the S&P fell by 50% over the next two years, the market would have priced that somewhere around 5%, a 5% probability, which relative to the empirical, the historical distribution of about 4%, it was like 25% markup, right? And that’s pure profit for an investment bank that is underwriting those puts… By the point that that market peaked in June of 2012, the probability that the market was placing on that 50% decline over the next two years had risen to 47%. And this is three years after the market had been rallying, 3 and 1/4 years after the March, 2009 lows that you were seeing this sort of pricing that exists in the market. And it was simply a function of what I was articulated before, that there were noneconomic buyers of insurance. And the ability to sell those derivatives had collapsed.
So, it was a world hyper-attuned to risk. And in a globalized world where financial capital and trade link economies together, it’s hard to tell a story in which the S&P 500 drops by half — again — and it doesn’t wreck the economies of every country that exports to the US, every country that exports to them, everyone who lends here, everyone who borrows here, and so on.
So, in the post-crisis period, the only question that mattered was: is this the bottom yet? If we’d reached the bottom, it made sense to buy anything that would do well in a global economic recovery: cheap emerging-market equities, high-yield debt, the Australian dollar, copper, oil. And it made sense to sell any kind of crisis-hedge asset, like treasuries, Yen, the dollar, the Swiss Franc, or gold.
This is somewhat ironic, because the “risk-on” trade assumes a recoupled world, while crises tend to be long-term decoupling events. In a boom, when valuations are stretched in rich countries, investors look to riskier places and riskier asset classes. When those don’t turn out well, they go back to stuff they know; German banks bought lots of subprime CDOs in the mid-2000s, then learned their lesson and went back to buying bunds and lending to the Mittelstands.
In a benign, risk-on world, disputes get mediated through markets. China fights Europe and the US for access to raw materials by bidding more. Risk-off moves tend to involve higher price correlations, because levered asset owners have to sell everything at once. But in a truly risk-off world, the underlying correlations have changed: the Yen and Swiss Franc move together when they’re both up 1% on a bad US unemployment number, but in a global depression it matters that Japan is a US ally right next door to a US rival, while Switzerland is basically a giant mountain fortress with a high gun ownership rate.
Risk-on assets are all alike, but risk-off assets are each different in their own way, because their idiosyncratic performance depends on… their idiosyncracies. There are two broad categories:
- Assets that move in response to broad risk preference because levered investors either borrow them or short them: USD, Yen, treasuries (which you’d be short, directly or indirectly, if you owned risky debt and wanted to hedge your interest rate risk).
- Assets that would actually outperform in specific disaster scenarios, like a financial crisis, a major war, a pandemic, a major trade disruption, or an oil shock. Crucially, for these assets the backtest omits the part of the distribution that matters. If fiat currencies collapse, something like gold or Bitcoin would be worth a lot (though Bitcoin would be worth a lot less if the collapse seriously disrupted the Internet, too).Ten years ago, a recession driven by an oil shortage would have made USD a risk-on asset at the extremes, since ten years ago the US was a major importer; today, it would be a risk-off asset, since the US is oil-independent and has a relatively diversified economy with plenty of internal sources of demand. China may have executed a similar flip: ten years ago, a global crisis would be an opportunity for them to expand their sphere of influence and extract concessions from their neighbors; today, China is highly levered, demographically challenged, and extremely dependent on trade.
If you run a simple optimization algorithm based on historical returns and your portfolio’s volatility requirements, the algorithm will tend to spit out giant allocations to one risk-on and one risk-off asset class, with minimal money allocated to others. In practice, portfolio managers split the high- and low-risk buckets, but end up following essentially the same design. In an economy with globalized flows of goods and money, the only two big bets are on good times and bad.
In this paradigm, Bitcoin has to fit into one of those slots. It has the volatility and technological sheen of a risk-on bet, but the long-term story is pure risk-off. Empirically, where does it fall?
Bitcoin vs Risk-On Assets
To analyze Bitcoin as a risk-on or risk-off asset, I’ve compared it to a handful of other assets. In the risk-on bucket, I’m including major indices (S&P 500, Nasdaq 100, Shanghai Composite, Nikkei) as well as other assets that track macroeconomic performance (AUDUSD, oil).
When we track Bitcoin’s rolling 30-day correlation with risk-on assets, we see…
Let’s try the risk-off bucket. Same approach, rolling 30-day correlations:
Zero for two.
There are a few ways to look at this:
- Bitcoin’s price is totally random, and has nothing to with with flights-to-safety; it’s pure gambling.
- Bitcoin’s returns are driven by some factor that doesn’t directly affect other instruments, such as demand for money-laundering. A while ago, I tested this out by looking at other money-laundering options, but didn’t find anything meaningful — Bitcoin prices don’t correlate to either Macau gambling revenues or Vancouver home prices.
- Bitcoin doesn’t correlate with standard risk-on or risk-off trades because it embodies both. In this model, it should trade like an asteroid mining company: a highly variable bet whose payoff is an unusually stable asset.
The first option isn’t compatible with Bitcoin being included in a portfolio. It’s a good null hypothesis; if Bitcoin is a gambling asset, it will eventually go to zero. The other two options, though, make Bitcoin a very interesting thing to own for anyone running a diversified portfolio: as long as you’re optimistic about its ultimate payoff, the low correlation to other assets means it’s a disproportionately strong diversifier.
The Event Path
The basic event path for Bitcoin works like this: when asset allocators decide it has a shot at being a safe asset, they buy a little. This has three effects: it raises the price, obviously; it lowers volatility, because asset allocators rebalance their holdings, which means they trade against the market; and it makes it more likely that Bitcoin will be a safe asset.
It’s Schelling Points all over again. If every major asset allocator decides that 0.1% is the ideal weighting for Bitcoin, the ideal weighting is necessarily higher than 0.1%, so they ratchet up, and continue to until they reach equilibrium. As Bitcoin’s odds of acceptance as a reserve asset grow, its volatility declines.
You might map out Bitcoin’s hypothetical valuations like so:
I’ve shown lots of possibilities but I highlighted the only reasonable path it can take: as Bitcoin’s odds of replacing gold rise, its required rate of return declines.
Of course, as Bitcoin’s value rises, regulatory risk increases. Most governments like the amount of monetary sovereignty they have. If they want it to change, it’s because they want more. So the higher Bitcoin goes, the more worried governments may get.
When governments ban it, it’s because they’ve realized it’s a problem, but it’s a deregulated network that moves at the speed of software, not government, so they’re always too slow. If they ban it because it’s used for drugs and child porn (still!) that can kill it, but its predominate use right now is speculation, so the only reason for governments to incrementally favor a ban is for Bitcoin to appear likely to succeed.
But a ban is untenable if investors genuinely believe that Bitcoin is a good asset, and any government that bans it is implicitly claiming that their own currency is a worse asset in a world in which Bitcoin is commonly used. Even the US government, the most powerful in history, was only able to ban gold for a little over a generation — and gold is relatively easy to confiscate.
We can think of a series of overlapping curves: the odds the smartest people ascribe to Bitcoin replacing gold, the odds asset allocators ascribe to this event, and the odds governments do.
In every case, the expected odds are based on the price of Bitcoin, but as long as your model accepts un-savvy governments, you’re safe buying in the face of a ban. (At least, if it’s not a ban that personally affects you.)
This is a nerve-wracking approach, but it’s a useful framework: when Bitcoin’s price is trending up and its volatility is trending down, its intrinsic value is rising.