In this issue:
- Momentum, Intellectual Humility, and Missing Pieces of the Model—One way to look at momentum is that it's an indicator of the real factors a model is missing. Momentum tells you that some assets have started to correlate with one another, and it doesn't tell you why, but often by looking at what's doing well in momentum terms, you can figure the why out. This general approach applies to real-world examples of momentum, from flow states to habits to elite formation.
- War Finance Meets Green Finance—Guyana has an increasingly oil-dependent economy, but is also selling emissions credits. A diversified portfolio of foreign investors is particularly important for countries worried about getting invaded.
- AI-Assisted Creativity—A novelist in Japan reveals that she used ChatGPT to help write a novel.
- Property Rights and Mortal Institutions—One advantage companies have in negotiating with unions is that the corporation is immortal but union leaders don't personally benefit from choices that make the union stronger in the long run. But when a union becomes a family enterprise, and operates in a jurisdiction with weak property rights, the relationship flips around.
- Complements—Shopify has a strategic interest in keeping the non-Amazon e-commerce ecosystem healthy and well-funded.
- Hedge Funds—Big hedge funds reported record profits, because changes in the structure of the industry more than offset a shrinking opportunity set.
Momentum, Intellectual Humility, and Missing Pieces of the Model
The Diff frequently discusses systematic investing, and, in that context, often brings up factors. Depending on the point you want to make, there are two good examples of factors:
- To the extent that you want to treat factor investing as codified common sense—take a pattern that sounds like it should make money, and study market history to show that it does make money—you'll probably use a factor like value or quality. Yes, buying statistically cheap stocks does tend to make more money than a broadly-representative strategy would. Investing in companies with high and stable margins has also worked out very well in the long run (though quality had such an amazing run in the last three decades ($, WSJ) that investors are now paying dearly for that specific kind of quality).
- If you want factors to sound mysterious and counterintuitive instead, talk about momentum. "Buy high, sell higher" actually works over long periods, in different geographies and asset classes. (Most of them; it's mysteriously weak in Japan.) The two simplest explanations for momentum are opposite phenomena. One is that investors are excessively prone to irrationally over-extrapolating price movements, whether that's in the form of losing one's head and diving into a trend (I hear AI was a big topic at Davos!) or institutional investors wanting to show a portfolio with winners rather than losers when they report their top holdings. The other standard explanation is that investors are irrationally under-extrapolating from fundamental changes. A momentum portfolio will own lots of companies going through transformative turnarounds, and will end up short lots of frauds as they accelerate towards zero.
But there's another explanation for momentum, arguing both for how it works and how systematizers can avoid the trap of treating everything their model doesn't capture as some kind of unpredictable error term. In this interview, @Macrocephalopod frames momentum differently:
[T]he fundamental insight is that assets always move together for a reason, even if that reason is something which is invisible to you. And that reason is likely to persist into the future, as well... so when you form baskets, based on just a very simple idea of stuff that has moved together in the past momentum basket is an example of this. It’s a basket which is long as stock which is up a lot and short stuff, which is download your to some extent grouping based on these kind of latent hidden explanations of why prices are moving.
Consider early March 2020: someone using a sophisticated risk model would have a way to see that there were some factor moves going on, like China weakness and worse performance for companies tied to economic growth. But they'd also see a weird grab-bag in the momentum set: the low momentum stocks were disproportionately travel-related, but also included in-person entertainment like movie theaters. Meanwhile, the early winners were companies like Zoom. So what the momentum factor's constituents illustrated around that time was that the Platonic ideal of a factor model would include, in addition to sensitivity to interest rates and oil prices, a factor for sensitivity to global respiratory pandemics.
The 2020 election and vaccine announcement naturally had a market impact, with value rallying. But within that, there was also a shift towards reopening trades. That's something momentum clarifies, too.
One of the funny things about momentum as a factor is that it's hard to explain well in financial terms, but also seems to exist in the real world. That's not necessarily true for other factors; there's a lot more real-world adverse selection in being a "value" consumer than in being a value investor, for example. But momentum exists at many levels:
- It's much easier to keep working on something than to start it; divide your day into five-minute slices and you'll see productivity feed on itself, and also see that even minor interruptions that break the flow of concentration can put a stop to it.
- Habits accumulate, especially when they get to the point that you're building them into your schedule. The transition from "it takes mental effort to do this" to "it takes mental effort to skip this" is surprisingly short.
- The concept of the "miracle year" might be overfitting, but it does imply something about people's production functions, especially narrowing things down to science. It's very hard to predict extreme outlier accomplishments, so it's at least a bit funny that one of the best predictors of making an earth-shattering discovery is having already made one in the last year.
- Selective institutions that limit applicants to people who have passed a different selective institution's filter are implicitly selecting for a high-momentum cohort. The emergent result is that among people in their late twenties, the highest earners in New York and San Francisco or the ones with the most influential jobs in DC and slightly different parts of New York and San Francisco have never had to make a big and difficult decision in their lives (they've always taken the highest-prestige track) and have never had to deal with failure (which, given how hard it is for standardized criteria to tease out meaningful differences within the 99th percentile, means they've been lucky).
These forms of momentum also make sense as latent factors. Procrastination, for example, is partly a form of uncertainty: if you're not sure what to work on, it's hard to start working. But intellectual work tends to produce ideas for the next project; very few people start out with a fifty-year study plan and work through it right on schedule, and it's common to average one new idea to pursue per idea thoroughly explored.
Habits are basically a phenomenon coupled with its own explanation: persist in them long enough and there's a habit-shaped gap when you don't. But they usually don't involve figuring something out and then applying it (or, if they do, it's retrospective).
What about miracle years? There are a few reasonable theories here. Some of these come about because someone unlocks a broadly applicable technique that unlocks other possibilities—it's not a coincidence that calculus and universal gravitation coincided. But the other factor is free time: a surprising number of big discoveries come from people getting their first unbroken stretch of free time to tackle a big question (in addition to the examples linked before, I'd add Stripe, Microsoft, and Facebook being founded during Harvard's reading period, and Jack Kilby developing integrated circuits because, during his first year working at Texas Instruments, he hadn't accumulated enough vacation days to take time off in the summer and thus had the office to himself). And one way these become serially correlated is that someone who, when left to their own devices, accomplishes something incredible will probably be left to their own devices more often.
For people on elite tracks, momentum has grimmer implications. The organizations that hire them tend to be long momentum in their human resources decisions in the same way that investors with strict stop-losses will tend to be long momentum in their portfolios. Unless there's a deliberate effort to find people who were unlucky, it creates a brittle system where participants aren't used to handling adversity. This might be another reason to support failing upward ($, Diff): they're sampling from the same distribution as the long-lucky-streak cohort, but they've seen more of the left tail.
Disclosure: Long MSFT, META
Imagine telling someone twenty years before that the first big beneficiaries of a Republican losing reelection would be oil and banking executives! ↩︎
Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:
- A company reinventing the way Americans build wealth for the long-run by enabling them to access "Universal Basic Capital" is looking for a product designer with fintech experience. (NYC)
- A fintech company using AI to craft new investment strategies seeks a portfolio management associate with 2+ years of experience in trading or operations for equities or crypto. This is a technical role—FIX proficiency required, as well as Python, C#, and SQL. (NYC)
- A well funded seed stage startup founded by former SpaceX engineers is building software tools for hardware engineering. They're looking for a full-stack engineer who enjoys working with customers to design and build software. (Los Angeles)
- A company that helps clients use alternative data to make better decisions is looking for a data scientist/analyst with experience in the finance sector and with alternative data. (Remote)
- A vertically integrated PE-backed cannabis company is looking for an Excel wizard with a background in supply chain. (Remote)
Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.
If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.
War Finance Meets Green Finance
Guyana's economy has a diversified portfolio of carbon-related activities: beautiful rainforests, but also the world's 17th-biggest oil reserves. The reserves have gotten the attention of Venezuela, which recently pushed through a referendum that, they claim, showed 95% support for Venezuela's territorial claims there. (It goes without saying that this is probably not a true reflection of Venezuelan popular opinion, and that the people living in the disputed areas were not consulted.) Meanwhile, Guyana is trying to sell carbon credits backed by its plans to continue not cutting down trees ($, FT). The boring way to read this is that developing countries are perennially short of capital, and use every funding tool they can. A more fun possibility is that the more financial ties they have to different kinds of investors, the safer they are militarily—Exxon can certainly lobby for the US to intervene if Venezuela invades, but that's hit-or-miss depending on who is in office. Having ESG-focused investors who also have a financial interest in the region's stability is another way to ensure it.
Rie Kudan, a Japanese novelist, revealed that "about 5%" of the text of her prize-winning Sympathy Tower Tokyo, was copied directly from ChatGPT outputs. Given that the book talks about AI, this was not as impolitic as it would have been in other contexts; it's closer to Neal Stephenson writing the 4,000+-page Baroque Cycle longhand. It's useful to view the novel-writing process as somewhat analogous to other AI-enhanceable tasks, where there's variation in how much creative thought needs to go into every single piece. Sometimes, characters just need some excuse to go to a particular location, some reason to learn a particular fact, etc., and even if an AI tool doesn't have the same quality as human-generated content, it can lead to a better overall output if it preserves momentum by minimizing interruptions.
Property Rights and Mortal Institutions
Long ago, this publication argued that one reason ports eventually got automated was that shipping companies are immortal corporations while unions care about current members but don't have an economic stake in the outcomes of future ones. That isn't strictly true, especially when union membership is desirable but headcount is limited; family members tend to find their way in. In Argentina, the truckers' union has had the same leader for over three decades, three of his kids work for the union in senior roles, and another son runs a separate union ($, Economist). Depending on how stable property rights are and how strong labor is, this time preference question can flip: arguably the unions have a longer time horizon, and are thus in a better position to make sacrifices today to ensure that they're still well-placed in the future.
Flexport is raising $260m from Shopify ($, The Information). There's a cluster of of companies that perform a variety of different services but have the strategic commonality that they're not Amazon. Competing with Amazon is daunting, and using their infrastructure sometimes seems to start a countdown to competing with them; it's unclear if any given piece of the Amazon puzzle is more of a standalone business or a strategically useful one that feeds valuable data back to the mothership. So it makes sense for these companies to collaborate a bit. It also makes sense for them to specialize: part of what makes them safer to work with is that they aren't a vertically integrated and can't do the same information-amortization as Amazon.
Disclosure: Long AMZN.
The twenty most profitable hedge funds collectively produced $67bn for investors last year, an all-time record ($, FT). Tracking performance of a fixed slice at the top of some industry is a messy statistic, since it's tracking both absolute performance and the share that accrues to a fixed number of companies. (This is one problem with looking at long-term trends in CEO pay versus worker pay by restricting the analysis to the S&P 500. In the last forty years, the valuation percentile cutoff for S&P 500 membership has moved up as the economy has grown.)
In general, the story with hedge funds since about the year 2000 is that absolute performance has declined, risk has declined faster, and they're increasingly selling pure alpha. That pure alpha has a high fixed cost, but the fixed cost means that they can capture more revenue per dollar of alpha generated by an individual portfolio manager or analyst. So what metrics like this really track is the maturity of the hedge fund ecosystem: the people who would have been running small independent funds are increasingly still working at large fund companies; it's harder to raise money for small funds, and easier to get paid at big ones.