Homesteading Mental Models

Plus! Robinhood, Momentum-Squared, Dollar-Weighted, Blocking Trump, Meta-Bundles, Commercial Ties

Welcome back to The Diff. Here are the subscribers-only posts you missed this week:

This year, I'm changing the publication schedule a bit: I'll take one day a week to work full-time on a new book, which will be available for free to paying subscribers.

This is the once-a-week free edition of The Diff, the newsletter about inflections in finance and technology. The free edition goes out to 18,759 subscribers, up 864 since the last edition.

In this issue:

Homesteading Mental Models

A surprisingly useful piece of advice about writing online, which generalizes beyond that, is that you should never get tired of repeating your best ideas. Many of the most useful ideas in the world are abstract inputs to many fields: most people don't watch a rocket launch or take a photo with their brand new smartphone and think "Wow, look at all that calculus!" So if you're a single-source supplier for an analytical input that repeatedly produces novel outputs, you'll only get full credit if you draw attention to it.

This is easiest to see in online media: "Aggregation theory" is synonymous with Stratechery, "Everything is securities fraud" is straight out of Money Stuff, "Charge More!" means Patrick McKenzie, "Software is Eating the World" means Marc Andreessen, "Signaling," is Robin Hanson, "X, Explained" is Vox, etc. In no case was the concept completely invented by the person who now owns it—but in every case they've refined it, found many new applications for it, and consequently owned it. It's the IP version of homesteading, or Lockean property rights: if you put enough effort into an un-owned idea, it's yours. This dynamic produces something that's normally rare in knowledge work fields: a form of intellectual property rights that are valuable to the creator but hard to abuse.

These topics are not quite a beat, but might be closer to a metabeat; there are some topics these writers and publications cover that other writers know to stay away from, both because of expertise and because of branding. There are some stories that basically belong to these writers, and anyone else who writes about them is going to be swimming upstream.

It's a bit like having a strong brand name. As a thought experiment, suppose you design a search engine that, in blind tests, gets rated as 1% better than Google. Google's standalone value is hard to measure, but it's safely in the hundreds of billions of dollars. So what's your search engine worth? 1% more than that?

No. Your search engine is worth a little bit more than $0. It's a very impressive project to put on your GitHub, and will probably get you a nice job building something else. It might be worth something to Microsoft, but would be hard to sell to them, and few other entities have the distribution necessary to justify paying much for a search engine, even if it's high quality.

"1% better" doesn't sell, except in markets that are so commoditized that the 1% is a cost rather than quality advantage. A small advantage can't compensate for the fact that Google is a default on billions of devices, that customers automatically choose it by default, and that "Google" is a verb.

This is not marketing nihilism, though: Google is a verb because Google was so much better than competing search engines. For fragmented, non-commodity businesses, it's possible to build a better product, turn it into a leading brand, and then retain market share leadership because of the brand rather than the difference in quality—there's a form of accounting alchemy where early R&D spending turns out, in retrospect, to have been marketing spending all along. A quality advantage matters, but distribution does, too; "Coke" is synonymous with "soda" in markets the company won first, and in some Slavic languages, the generic term for marker comes from a specific brand.

And this applies in fields beyond marketing, too: in medicine, law, academia, investment banking, and other fields, specializing in one subset is a process of becoming synonymous with it, both at an individual and institutional level.

There are a few themes I hit over and over again:

I didn't invent any of these; the first time I thought about applying financial analogies outside of finance, for example, was in Michael Lewis' Liar's Poker, in which he complained that it was a try-hard habit of investment banking trainees.

To the extent that I have competitive advantage in writing about them, it's from a different kind of marketing: the internal kind. A slogan is not just a way to talk to the outside world; it's a way to establish an internal narrative. This is one reason the Chinese Communist Party has so many policies that get codified in a slogan even if they've been happening for a long time. It's hard to coordinate the behavior of over a billion people without some kind of catchy guidance. So a tagline, attached to a useful concept, is a good way to run every decision or observation through a quick filter: if it matches, it's important; if it doesn't, discard.

Part of the point of slogans and brands is that they involve compressing a concept to the point that it loses information, which makes them a way to apply a general principle in many contexts. "Don't be evil," "a computer on every desktop," and "socialism with Chinese characteristics" are all vague enough to provide guidance in many different situations.

Slogans also provide plausible deniability. "Make America Great Again" means something different if you lost your factory job than if you're nostalgic for a lower top tax bracket, or perhaps have some other concerns you're expressing at the ballot box. A reddit cofounder's book, Without Their Permission, read a bit differently after what some of the site's users did without permission. Which is not to say that that's what slogans are for—Reddit was founded to share interesting links, and having an unrestrictive view of "interesting" created many flourishing communities, only a small fraction of which turned out to be full of terrible people.

Some idea-homesteading happens by accident: Boltzmann apparently didn't call it the Boltzmann Constant, Overton didn't call it the Overton Window, Turing talked about a-machines rather than Turing Machines and "The Imitation Game" instead of the Turing Test. And apparently every model from academic finance fits this rule. It makes sense in academia. Black and Scholes put their name on the paper, not on the model, but "the options model from that paper by Black and Scholes" quickly gets compressed to "Black/Scholes model."

The rise of writers-as-idea-homesteaders is a reflection of the general trend towards fragmentation at one level and concentration at a more meta level. Computers used to be owned only by universities and big companies, and now most people carry one everywhere and many people wear one. But some of the components of those computers have gotten more concentrated, not less. And the supply chain for homesteading ideas is owned by very few companies indeed; they don't spread from campus to campus and firm to firm so much as they spread on Twitter/Facebook/TikTok/Google or not at all. To small-scale idea-homesteaders, this is mostly good news: if any of the output of your work could involve public writing, it's worth it to apply a short and ideally catchy name to a concept that you find yourself using more often than anyone else. It's not an asset in accounting terms, but it's very much like digital real estate: a claim on future attention. And, as in real estate, if you can identify an undervalued concept before it's about to get a lot of intellectual foot traffic, you'll own a very valuable asset indeed.

Further reading: while researching this I came across a delightful essay by Eugene Wei on compressing ideas. And thanks to Diff Reader Nathan Taylor for the comment that prompted this post.

A Word From Our Sponsors

Here's a dirty secret: part of equity research consists of being one of the world's best-paid data-entry professionals. It's a pain—and a rite of passage—to build a financial model by painstakingly transcribing information from 10-Qs, 10-Ks, presentations, and transcripts. Or, at least, it was: Daloopa uses machine learning and human validation to automatically parse financial statements and other disclosures, creating a continuously-updated, detailed, and accurate model.

If you've ever fired up Excel at 8pm and realized you'll be doing ctrl-c alt-tab alt-e-es-v until well past midnight, you owe it to yourself to check this out.


I wrote a piece for Marker on shared office space as a bet on hybrid work-from-home models. If the office arrangement of the future is to work from home most of the time and intermittently meet up with colleagues, then companies that sell flexible office space on the spot market will be well-positioned. As it turns out, WeWork was the worst business model imaginable during a pandemic, but a surprisingly good one after.

Robinhood's Plot to Make IPOs Less Exciting

The IPO process is designed to a) determine the demand for stock at various price points, and then b) reward the initial buyers for taking the risk that the IPO process won't work well. Historically, this has rewarded the investors who participate. And more recently, it's been especially beneficial for them because retail demand is harder to measure in advance, and much less sensitive to valuation. A few weeks ago, I suggested that one solution to this problem is for Robinhood to aggregate more of the demand itself; companies would raise a bit more money, at a slightly higher price, and retail investors would get a better deal. If they're a big enough factor to cause a 113% first-day pop in Airbnb, they're worth raising money from. Institutional IPO investors would lose, but if a sufficiently hot company goes public, they won't be able to object.

A hot company like, say, Robinhood, which is considering giving customers access to the IPO. Robinhood has already gotten very good at using small, mixed-and-matched financial incentives to acquire users and keep them active; Robinhood stock itself fits into this model, with the added benefit that it makes users feel like they're making back some of the money Robinhood gets from selling their order flow. Casinos are a better business than poker nights among friends, but poker nights have a lower cost of customer acquisition.

Momentum-Squared and Value-Squared

The value and momentum factors are the observation that statistically cheap stocks exhibit better returns over time, and that stocks with high recent returns tend to have higher future returns, also on average. Both of these edges are small relative to stock returns, and volatile, but they do seem to exist. But the factors have periods of poor performance; value did terribly during the dot-com bubble, and very well after that bubble popped; it's also had a poor run in the last few years, albeit with better results in recent months. So one question is: do the anomalies themselves have the same characteristics? Does value's outperformance display momentum? Can the momentum factor be bought low and sold high?

As it turns out, the answer is yes, factors themselves do display some momentum. Oddly enough, this second-order claim is easier to explain than the first-order claim that factors exist at all. Many discretionary investors end up preferring one flavor or another; factor analysis is a way to build a Buffett-bot, or more cynically to explain that Buffett is just bot. And since investors have a well-known tendency to back managers based on their track record, any outperformance of momentum investors will lead to more money chasing momentum, which makes momentum perform even better. Traders tend to believe in a vague folk wisdom around "what's working," and this is statistically true: it takes a while for money to move from a cold hand to a hot hand, and when it happens, prices move, too.

Dollar-Weighted Returns

There's a probable urban legend that a major brokerage once analyzed all of its customers to figure out what the best performers had in common, and they found two factors that maximize performance: forgetting the brokerage account exists, and dying. It's a good story, and probably directionally true: for the vast majority of investments, dollar-weighted returns are lower than time-weighted returns because winning funds gather more assets, and winning is partly a matter of luck. (Consider a model with two time periods, where every fund has one lucky year and one unlucky year. All the funds with a lucky year their first year raise more money; the funds with an unlucky first year either don't raise much or shut down entirely. So the next year, most assets are invested in funds that will have an unlucky year.)

Brevan Howard's assets under management have declined by two thirds since 2013, as the fund's mostly macro strategies struggled. 2020 was a very good year for investors who reacted quickly to headlines, and who were especially attuned to policy decisions and their consequences, so Brevan Howard had its best year ever.

Blocking Trump

In yesterday's issue, I noted that one of the challenges online platforms face is that the media are implicitly part of the legitimizing process in an election. It's when the push notifications of called elections get sent that people start celebrating, not before then. Social media have less reputation but much more reach than traditional media, and they haven't been faced with a case like the 2020 election yet. In some cases, it's an easy cost-benefit analysis; Shopify took down the Trump campaign store after several other sites banned him, and Facebook has banned Trump from posting at least through inauguration day. Twitter, unluckily, has the hardest challenge: they took down a video of Trump telling protestors to "go home," because the video included claims about stolen votes—which fit the standards they set back when the issue was claims about election fairness rather than the physical safety of Congress. As it turns out, one element of America's unwritten constitution is that the President's favorite social media app may have to figure out exactly what mix of statements the President can make that calm things down more than they inflame tensions.

The way this kind of problem gets resolved is not through having the best possible rules, but having explicit rules that don't change much. 2020 was a stress test—in many ways—but by widening the limits of plausible constitutional issues an app can face, at least it means the next crisis will have dissatisfying answers instead of much more dissatisfying uncertainty.


A group of media executives is launching Struum, a service that aggregates video content and lets users sample it and then sign up for full accounts if they like it. It's an interesting bet: a mixture of different services has to be compelling, or Struum's users will leave—but if it's too compelling, then the services themselves won't want to be on a platform where they lose access to users. So it has to aim for a narrow window band of quality. Like a dating site, it can suffer from churn by being bad at matching users on both sides of the platform, but also if it's too good.

Commercial Ties

The right way to look at globalization is usually that it's a general term for a series of specific, bilateral relationships. The story of globalization would be very different without Japan, Germany, Taiwan, the US, etc. On the other hand, the general model has one thing going for it: when a country like the US pulls back, other countries, like Germany, step in ($, Economist). Business relations between the US and China have long been driven by the immediate opportunity to access China's manufacturing capacity, and the long-term hope to reach its consumer market. (The former has been much easier than the latter.) Now, large German companies are pursuing the same goals:

“We have to play ball with the Chinese,” says Joerg Wuttke, the German head of the EU chamber of commerce in China. “If you are not at the table, you are on the menu,” he warns.