In this issue:
- In Investing, and in Life, The More Timeless the Discovery the More It's a One-Off—You can boil many successful investors' results down to a set of strategies that, at least today, are trivial to implement. Part of their returns come from human pattern-matching at a time when the automated kind wasn't feasible, and part of their returns come from other investors catching up to their insights. This happens at all levels of finance; the biggest hedge fund fortunes right now come, not from picking stocks, but from picking stock pickers. And the alpha from that, too, will dissipate in time.
- Airlines—Alaska Airlines has acquired the parent company of Hawaiian Airlines for $1.9bn, with the intention of consolidating their businesses and reducing costs.
- Deflation—Prices for durable goods are now declining year-over-year, but services inflation is more persistent.
- VC—Venture funds are utilizing a private equity tool to extend the life of their fund, taking advantage of a limited pool of high-quality companies and the ability to charge fees on more assets for a longer period of time.
- Automation—A Brazilian city passed an ordinance written by an LLM. This demonstrates the utility of LLMs as an interface.
- Vertical Integration—Apple's chip business is part of their general efforts to control the entire stack, and it creates a chip business with more predictable demand and thus better economics.
In Investing, and in Life, The More Timeless the Discovery the More It's a One-Off Gain
A fun hobby among systematic investors is taking the historical track record of a famous discretionary investor, running some regressions, and discovering that they were actually just quants-without-calculators—Bill Gross's returns can be partly attributed to credit risk, avoiding extremely long-term bonds, and preferring mortgages to treasury bonds, Soros's to trend-following within and between asset classes, offset by a slight negative return from the specific assets chosen, or Warren Buffett buying low P/E stocks earlier in his career and switching to higher-quality ones later. That worked, not just because both factors have produced high returns over time, but especially because the "quality" factor, i.e. the excess returns from investing in companies with high margins and steady growth, had a phenomenal multi-decade run ($, WSJ).
It's worth thinking about why that happened the way it did. Quality is more persistent than other factors like value, momentum, and growth. Status as a value stock is transient, because some companies snap back to being fairly valued and others remain cheap-for-a-reason (Jim Chanos noted that "Kodak was cheap all the way down too, trading at five times cash flow, from 70 to 24"). Momentum doesn't last forever, because if it did every stock would go straight to either zero or infinity. That is, in effect, what happens over sufficiently long time periods, but it's not a straight line. And growth also tends to fizzle out: buy a portfolio of companies that grew 30% last year, and you'll find that some are cyclicals that will have seen revenue shrink 25% the next year, some are flukes that mean-revert for other reasons, and even the quality growth companies are growing a bit more slowly than before.
Meanwhile, high margins persist for a long time. If the median investor doesn't understand that, and if the median manager underestimates it, several things will be persistently true:
- Investors buying these stocks will systematically outperform.
- That outperformance will be more extreme in cases where the companies mostly reinvest in their own business, or, if their stock is cheap enough, in buybacks.
That described the status quo when Buffett made the shift towards quality investing, a process that started in the 60s and was mostly complete by the 80s. At the time, the list of quality companies that preferred buybacks to dividends, and that continued to reinvest at favorable returns on equity, was a short one with frequent turnover.
But today, that's not true. Investors understand quality; as the WSJ article on Munger notes, high-quality stocks outperformed low-quality ones by 5.2% annualized over a 34-year period, in part because those companies started getting recognized by investors, and partly because their managers learned how to manage them. If a company's an unsteady grower with poor returns on capital, a buyback will be heavily countercyclical; they'll only have the money to do it when they're over-earning, and both cash flows and valuation will mean-revert. Meanwhile, a dividend is a strong signal from management about earnings' future persistence. A company that's been able to grow pricing faster than inflation, keep unit volume steady or rising, and grow margins over time doesn't need to continuously provide such a signal, so its investor base will tolerate the variance in capital returns inherent in buybacks. In fact, those investors will tend to appreciate buybacks' ability to stabilize the share price; one element of the 2010s bull market was companies shifting their capitalization from equity to debt (i.e. borrowing money to have a buyback that returned more than their free cash flow) because debt was so cheap.
The net result of all of this is that quality stocks are priced as if they're high-quality, and can, in part with generous equity compensation, recruit better managers than other industries can. The signal is still there, but much of the edge has been priced out. It's good to find a signal that works, and it's even better, for a while, to find a signal that other investors are discovering, too: they'll bid up the stocks you've already bought, so your performance gets a boost (for a while, until things overshoot).
It is, of course, unfair to say that great investors didn't add value during their careers, just because we can reverse-engineer their process today. Sure, you could imagine buying a license for Capital IQ—Time Travelers Edition and traveling back in time; a modern laptop probably has enough power to identify every major source of systematic alpha. It's a lot harder to execute those reverse-engineered systematic strategies the way they actually would have been done decades ago: reading SEC filings by going to the SEC reading room, trading an OTC stock by calling half a dozen market-makers and hoping that the price hasn't moved by the time you've assessed the market, dealing with the fact that quarterly reports are nonexistent, and running your discounted cash flow scenario analysis on a calculator powered by a hand crank.
In short: the bargains were there because it was a colossal pain to identify and exploit them.
This kind of transition has happened in the faster-moving parts of markets, too: before Black-Scholes, options were fantastically illiquid and mispriced. One of the big latency decreases in the 1960s was when Carl Icahn started publishing a weekly newsletter with options quotes. Better calculators completely changed things; for one thing, a hand-held calculator powerful enough to run the Black-Scholes formula, or at least to update the most important Greeks, meant that a centralized in-person trading venue made more sense than telephone-based trading by people in physically separate offices. You could get pretty far writing a history of investing and trading defined entirely by calculators, especially if you generalize it to the most ubiquitous and important calculator of them all, Excel. When hardware improves and data improves—and they have the same feedback loop that renewables and batteries do today, or that gasoline distribution and cars had a century ago—then what used to be a proprietary advantage turns into table stakes.
This framework even applies to meta discoveries. There was a time when the big money in investing was made in a literal way, by systematically identifying undervalued companies. But if you look at the categories that are currently minting the biggest fortunes (private equity and multi-manager hedge funds), they've actually moved out one level of abstraction. They're both good layers for connecting returns-seeking capital to investment talent, with a disproportionate share of the excess returns accruing to that intermediary party. They've gone from stockpickers to stockpicker-pickers, trying to snap up an undervalued index arb or load up on cheap vol traders while also hoping to flip their TMT team before it mean-reverts (again).
The more important the discovery is, the more likely it is that the news gets out and the knowledge gets priced in. Anyone can go back and reverse-engineer the classic quality trades; soon they’ll be asking why it was possible to buy a growing network effect beneficiary like American Express at 18x earnings in the early 60s, or to invest in a globally-recognized brand with high and high-margin international growth like Coca-Cola at a similar valuation. Now, it's easy to list quality companies, but within that list you have two choices: either pick a company trading at roughly the market multiple because that quality is offset by some existential risk, or get ready to pay a gigantic premium and pray that the growth model stays intact forever. Genies are sadly reluctant to reenter their bottles, and the big ideas work once before they become part of everyone's vocabulary, and are just one more generic risk premium that produces small excess returns but periodically blows up.
That is, of course, the fun of the whole business; the money is made by using mental models to identify mispriced assets, but with mental models that have a hard to estimate half-life that's usually far shorter than a career. So there's a metagame of getting good at revisiting and revising models, and inventing new ones. The cycle is probably speeding up, too; whoever the best investor of the 2020s is, by the 2030s there will be an ETF replicating whatever they're doing right now. Which means that, if they're a truly generational talent, by then they'll be doing something else.
One reason big tech companies have done so well is that while their growth rates weren't absurdly high for most of their time as public companies; those growth rates have been surprisingly durable ($, Diff). Usually, investors model a deceleration in growth, but sometimes a company's economies of scale and continued efficiency improvements are enough to almost fully offset the difficulty of achieving the same percentage growth rate on an ever-higher base. Over very long periods, cumulative returns approach cumulative growth in revenue per share, so a slower-than-expected deceleration has an outsized impact. ↩︎
"Mostly," but not entirely: one reason to do quality over value is that it's really hard for a big company to be statistically cheap unless there's something seriously wrong with it. But at smaller scales, Buffett was involved in companies that were nothing special but were especially cheap; even when he was one of the world's richest people, some of his personal money was in small REITs and a small-cap electronics company that was slowly liquidating. ↩︎
Comprehensive data, but only up through the date you're visiting, lest you a) create a paradox by reacting to history you already know, and b) even worse, fall prey to look-ahead bias. ↩︎
The first hand-held calculators started appearing in the early 1970s, and the CBOE launched in 1973. This is not a coincidence. ↩︎
Companies in the Diff network are actively looking for talent. A sampling of current open roles:
- A well-funded startup is building a platform to identify compliance risks associated with both human- and AI-generated outputs. They are looking for a frontend engineer with React/Typescript experience to join their team of world-class researchers. (NYC)
- A company building the new pension of the 21st century and building universal basic capital is looking for a GTM / growth lead. (NYC)
- A crypto proprietary trading firm is actively seeking systematic-oriented traders with crypto experience—ideally someone with experience across a variety of exchanges and tokens. (Remote)
- A concentrated crossover fund is looking for an intellectually curious data scientist with demonstrated mastery in analytics. Experience with alt data, web scraping, and financial modeling preferred. (SF)
- An early-stage startup aiming to reduce labor costs by over 80% in a $100bn+ industry is looking for a part-time technical advisor with robotics experience; this has the potential to evolve into a full-time role. (NYC)
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.
Alaska Airlines has agreed to acquire the parent company of Hawaiian Airlines for a total of $1.9bn including the assumption of debt. The equity component of the deal is $18/share in cash, or a 270% premium to Hawaiian's Friday close. Alaska and Hawaiian are part of the dying breed of more-regional-than-national carriers; they're both big in California (Alaska's business in its namesake state is not huge, but is a nice near-monopoly for them). Consolidating these businesses is strategic for two reasons:
- There are some fairly-fixed airport-level costs that are cheaper with more planes. The number of spare parts and maintenance workers needed on hand for surprises is a sublinear function of the number of planes in a given location.
- All airlines have idiosyncratic demand drivers. In the long run, these don't matter much, as long as demand keeps growing. But in the short term, they can give a specific company a bumpy ride; Hawaiian's shares are down 87% over the last five years for some destination- and fleet-specific reasons. A hedge makes an airline less likely to go under, and that makes it cheaper to borrow.
The deal may not go through; Spirit and JetBlue have been trying to merge since July of last year, and are just now getting to the point where a deal might be approved. So this announcement is closer to the start of something than the end.
Inflation is always and everywhere a multi-causal phenomenon. Sometimes it really is about the quantity of money flying off the printing press, and sometimes it's something as prosaic as traffic jams at major ports. Right now services inflation is still running higher than the Fed's target and the pre-Covid trend, but durable goods prices have been deflating year-over-year since June ($, WSJ). Durables will have more swings than services, because the money spent on services often gets spent on—more services! When Walmart raises wages, its employees can afford another visit to McDonald's or a night at the movies, adding to demand for services. Durables will have more cycles, since prices for a given good respond to demand minimally when production is below capacity and then nonlinearly when it moves beyond that point.
Venture funds are borrowing a private equity tool and raising "continuation funds," in which they roll some of their holdings forward into a new vehicle to reset the life of their fund ($, FT). As with similar recent moves in private equity ($, The Diff), there are two big drivers: there's a limited pool of high-quality companies, and if you own one you'll be reluctant to sell. And, of course, being able to charge fees on more assets for longer is always attractive.
Porto Alegre, a city in Brazil with a population of 1.3m, has passed an ordinance written by a large language model. This is partly a reminder that LLMs are useful as an interface, not just a tool for working. There's usually a part of the legislative process when bullet points get converted into legal language and the edge cases get fleshed out, a process analogous to compiling human-readable code in a high-level language into machine code—both in the sense that the output is hard to read for normal people and that the process might reveal bugs in the original spec. Since many laws produce natural-language commentary before or after the fact, there's ample training data for building a product specifically for this purpose.
CNBC has a nice overview of Apple's decade-plus effort to build in-house chips instead of relying on outside vendors. It's a nice demonstration of the compounding returns of vertical integration: Apple has a long history of tying hardware and software closely, which has good strategic utility (i.e. the more of the final product Apple controls, the better they can make the user experience) and some nice practical payoffs (high margins on Apple-branded peripherals like chargers). Adding chips is partly a matter of knowing and controlling the roadmap: if Apple has a good sense of what its hardware will be capable of over a longer timeframe than other companies, it can build around those constraints. And it also means that Apple can solve for what's always been a big problem in the chip industry: predicting demand. Apple's chips business has one big captive customer, and that customer happens to be very good at predicting and influencing demand.