In this issue:
Hedge Funds as Liquidity Providers and the Great Degrossing
Return of the Tontine
Amazon and the Minimum Wage
Granularity and Price Discrimination
The Vaccine Distribution Supply Chain
Rallies and crashes are easy to understand, and so is sector rotation—when energy is up and large-cap tech is down, for example, that's easy to read as a consensus shift towards more immediate economic growth. Hedge fund degrossing is another market phenomenon, sometimes invisible at the surface, but often with more dramatic consequences. In a classic de-grossing, heavily-shorted stocks go up, recent outperformers go down, and, until it's over, being skilled but conventional has a strong inverse correlation with making money.
This is what's happening right now.
The basic mechanics of de-grossing work like this: a typical long/short fund will run with high leverage relative to its equity base. Funds match long and short positions so their net exposure is minimal. So for every $1 of equity, a fund might have $3 of long positions and $3 of short positions. It's hedged—on average, that means zero net exposure to the ups and downs of the market—but it does mean more exposure to individual stock picks. If the hedge fund's investment team is doing their job, that's good; their returns isolate the results of skill, and ignore the general performance of the market.
A good tenet of risk-management is that when volatility increases, it's a good time to reduce risk. This is an article of faith among experienced risk managers. (The ones who don't believe it tend to be either inexperienced risk managers or former risk managers.) It's a practical interpretation of Knightian uncertainty: if your portfolio is designed so that 95% of the time, it loses no more than 0.5% per day, then a succession of single-digit losses implies that a) your model was wrong, and b) while the magnitude of the error is unknown, the direction is towards higher risk.
So, our hypothetical fund might adjust its portfolio to $2.50 of long positions per $1 of equity, offset by $2.50 of short positions. (This involves turning over slightly more than 100% of the firms' original capital, since in this scenario the equity base is also a tad smaller.)
What do they buy, and what do they sell?
If you're a short-term, highly-levered stock picker turning over a portfolio fairly rapidly, your job is not to be wildly unconventional: it's to replicate next month's conventional wisdom today. That doesn't mean ignoring the future, since the job is to predict the consensus view of what the future holds. But it does mean making decisions in a narrow domain: what are slight misconceptions that will be resolved in the next few weeks? In general, hedge funds want to guess where real money will be in the future; real money investors tend to move more slowly (in part because they're running more money, in part because the time scale on which they operate is long enough that they're less sensitive to short-term fluctuations. Capital Group could easily own Netflix for the next two decades. They aren't dumping their position ahead of a bad quarter, although they may adjust it from time to time.) So short-term stock pickers are often consensus-only-more-so. It’s improv: instead of saying "No," they say "Yes, and."
Since I view every investor as a market maker, it's easy to describe this as a liquidity-providing activity. In fact, it's quite close to how market makers behave: a long/short hedge fund's job is to provide liquidity to comparatively time-insensitive investors, and to charge them a small premium for being late to change their minds.
One consequence of this division of labor is that real money investors tend to know a lot more about the business, while long/short funds know a lot more about the market. If you want to know why a stock did something strange today, ask someone who covers it at a hedge fund. If you want to know why one stock is up 10x in the last decade while their competitor's shares are flat over the same period, ask a long-only investor. (This is a generalization that describes two mostly-overlapping bell curves. The best tend to know a lot about both. But since hedge funds pay higher fees per dollar of assets, they tend to reward trading skill relatively more.)
In this model, hedge funds are behaving like a sensible market maker when they lower their gross exposure. Market makers want to take risk when they're getting paid a stable price for liquidity. They don't want to get stuck with inventory when the market moves against them. So they provide less liquidity, and charge much more for it, when the market is more volatile. They're happy to live in a frequentist world with a known distribution of outcomes; when the distribution shifts, they want to back off until they see what the new normal looks like.
We are currently in an abnormal world, at least for equities. The most-shorted stocks had their best ten-day period ever, for example. Some of this is the catalyst for de-grossing, and some of it is the result. Usually, a de-grossing involves selling off some stocks and buying others; the net impact isn't visible, unless you look at heavily-shorted stocks. But in this case, one of the most crowded trades is large-cap tech, which is also a large proportion of the major indices. Meanwhile, many of the most shorted stocks are fairly small; hedge funds have reached a mostly accurate consensus that the current economy is relatively favorable to the biggest companies' prospects.
There are other symptoms of de-grossing: As Andrew Walker of Yet Another Value Blog points out, parent/stub trades (buying a company that owns a chunk of another company, shorting the other company) are also losing money. An ETF that holds pre-deal SPACs, another popular hedge fund trade, dropped 3% today (it's still up 40% since launching in May).
Meanwhile, media reports and the Twitter rumor mill report losses for most of the well-known long/short funds, albeit on a smaller scale than Melvin.
This is not the first de-grossing. There was a big one in September 2019, when momentum stocks sold off and value stocks rose. There was one in the spring of 2014, when mid-sized growth companies sold off. There was a famous one in August of 2007, the "quant quake" that caused the usually bulletproof Medallion Fund to briefly show losses.
But they're usually not a systemic risk. Since the most levered funds are also market-neutral, they have to do as much buying as selling. This can substantially erode their own equity, and certainly cause wild swings in the market. But the long-only funds and less levered players are still there, and have probably been making opportunistic purchases. Meanwhile, funds that de-grossed early are no doubt salivating over the opportunity to short GameStop and AMC on the way down. GameStop stock briefly dropped 30%, losing $6.5bn in value, when a WallStreetBets chat got banned for hate speech.
While this de-grossing is more severe than most, there isn't historically a strong connection between hedge fund de-grossing and market crises. Of the early examples I mentioned, most happened during the long post-crisis bull market, and the 2007 one preceded the crash by a little over a year. They're interesting to watch because they're largely invisible on long-term charts of broad market indicators, but they make and break careers. And they're a bracing reminder that the most extreme risks are the hardest to predict in advance.
 This is an oversimplified view. For example, funds won't just enforce rules about low net exposure to the overall market. They also want low net exposure to individual sectors, or to factors like value or momentum, or even to specific industries.
 Two points: first, that's suspicious timing. Second, it's symptomatic. A rapid-fire chat with poor moderation is a primordial soup for contagious, competitive memes. Some of these memes have been adaptive—if you thought the jokes on WallStreetBets were compelling enough that you went ahead and bought GME calls, you did very well for yourself. (Memes often spread fast when they're both contagious and have some kind of utility. It's not a coincidence that many ancient holy books spend a lot of time on ritual hygiene—the ancient religions that didn't focus on that kind of thing had a higher mortality rate.) But the same format that allows rapid adoption of bull cases for GameStop allows rapid adaptation among trolls, so there was enough content for Discord to justify banning the server.
Subscriber Q&A - GameStop Edition
This week's subscriber Q&A will be at 2pm ET / 11am PT on Friday, January 29th. We'll be talking about GameStop, gamma squeezes, market manipulation, memetic selection, and more.
Submit a Question in Advance
Return of the Tontine
Pensions & Investments writes about how tontines are making a comeback ($). If everyone's retirement savings are optimal on average, then most individuals either saved too much or too little, depending on how long they end up living off their savings. Tontines are an elegant solution to that problem, by paying a sum only to the survivors. It's an ancient idea, with some regulatory and reputational cruft (in the US, they were mostly banned because they were used by swindlers, not for more dramatic reasons. Implemented correctly, tontines are an elegant solution to the individual-level variance of savings behavior that’s high enough in the aggregate.
(I previously wrote about how the economics of tontines line up with the economics of learning COBOL here.)
Amazon and the Minimum Wage
A useful meta-strategy for big companies is to a) adopt something as a standard voluntarily, and then b) try to make it a universal standard. This shows up in literal technical standards, as when Huawei has lobbies heavily for favorable 5G rules ($, FT), but it also shows up in more abstract ways. In 2018, Amazon raised its minimum wage to $15/hour, and now it's advocating the same for everyone else.
The exact effects of minimum wage are still a contentious topic—there is a meta debate about which meta-analyses to trust—but a new variant on the debate is the question of whether minimum wage is a de facto subsidy for big companies. Larger employers are more efficient, for a variety of reasons, so they're better able to pay higher wages, but accepting the argument that $15 is the minimum that someone should earn for an hour's work doesn't mean accepting the belief that it's the minimum anyone should pay for that work. Negative income taxes that create the same after-tax income for low wage earners are an implicit subsidy for small businesses, and supporting small business is politically popular. So the ultimate debate may be over whether the US government will indirectly support small companies through tax policy or indirectly support large ones through wage floors.
Granularity and Price Discrimination
During a pandemic, airlines have two related problems: a general drop in demand for travel, and the deeper problem that every incremental passenger they accept imposes a potential negative externality on everyone else, by making them sick or possibly killing them. One Russian airline has a solution to both problems: using a ride-sharing service to let people charter 96-seat passenger planes. Chartering the plane places the burden of pandemic management on whoever is organizing the trip, which allows the airline to accommodate varying levels of risk tolerance.
Historically, you could divide the energy market into two categories: companies that represented a bet that global warming wasn't a problem, and companies that only made sense if it was an existential threat. Two forces are pushing for convergence between these views: first, climate change is a much more broadly accepted idea than it was a decade or two ago, while the deadline for action is creeping closer (regardless of exactly when that deadline is, if there is one, it isn't getting any further away). Second, lower real rates mean that companies in a terminally declining business, like oil extraction or internal combustion-based vehicles, get a valuation penalty. If they can redirect some of their cash flow towards a business that isn't likely to get regulated out of existence, the all-important terminal value component of their net present value goes way up. A a result:
One interesting element here is the coordination. Companies with an especially long lead-time, like auto manufacturers or LNG exporters, can signal in advance where they think the supply and demand curves will be a decade hence. If their signals are credible, everyone operating on a shorter time scale will have to catch up.
The Vaccine Distribution Supply Chain
Last week I linked to this piece on how vaccines get manufactured. Here's a similar overview of how they get distributed. A key point is that there are some layers that are standardized, like Palantir's Tiberius system, and some layers that are updated on the fly. This may lead to low estimates of vaccine distribution, especially in places that have recently stepped up their efforts:
Clinic workers in multiple states have found the new requirements so onerous they’ve started writing all the vaccination records on paper and entering them into the computer by hand when they have free time.
US vaccination numbers have been improving from the slow initial rollout, and that improvement may be even better than it looks.