Welcome back to The Diff. Here are the subscribers-only posts you missed this week:
China's Recovery and its Consequences looks at the manufacturing- and export-driven outperformance of China relative to other large economies. In some ways, 2020 reversed major trends, but it reinforced China's centrality to the global economy.
Bubbles and Easy Money: Quality, Not Quantity explores the theory that financial bubbles can always be explained by growth in money and credit. It's not entirely true, and often the tail end of a bubble is associated with tighter money. But changes in the form of money can, and do, propel bubbles.
Hedge Funds as Liquidity Providers looks at the second-order consequences of the GameStop situation. Hedge funds are covering shorts and selling long positions. This kind of event happens every few years, but since it involves buying and selling, it's usually invisible in broader market indices.
Sign up today to read them, and to join the subscriber call where we’ll be discussing GameStop and what comes next.
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 19,312 subscribers, up 224 since the last edition.
In this issue:
Messaging is an interesting business because the network effects are so granular. If a messaging app makes it easy to send messages, then it's easy to receive them, and that leads to more sending. The network effect looks durable because it's reinforced every few minutes during the entire waking day—but any time a new platform can demonstrate higher usage per user, it's instantly fundable because better user-level network effects win. When Snapchat was private, it consistently looked overvalued from the outside, but all its investors had to know was that when someone in its target demographic picks up their phone and sees one notification from Facebook and one from Snapchat, they tapped Snapchat's first. As it turns out, Facebook and Snapchat fought their way to a rough stalemate, where Snapchat's iteration speed wasn't enough to overcome Facebook's denser network effect among older users (and its willingness to match Snapchat feature-for-feature).
But that's an older messaging war. The new one is over privacy and moderation. It's created some strange bedfellows: extremists in the US and Europe, anti-government protestors in Belarus and Iran, drug dealers, cryptography geeks, trillion-dollar corporations, and Elon Musk fans.
Many companies have had challenges monetizing messaging directly. If the product lives and dies on how easy it is to use, any ad that's visible enough to be worth paying for also disrupts the core viral loop. And a privacy-focused app makes it harder to target ads. The reason Facebook can earn $25.50 per monthly active user each year is that their ads are laser-targeted based on numerous signals. The targeting is good enough that "Facebook targets ads by listening to your conversations" makes it into the news cycle about once a year. LINE has been able to make money selling in-app sticker collections, but most messaging apps justify their existence more indirectly. WeChat, for example, does messaging and basically everything else. Messaging, because of its high engagement, is a very strategic product, so each of the messaging platforms uses it to advance completely different strategic goals:
iMessage, which may be the world's second most-used messaging app, has a different model. The core business case is twofold: first, keep Apple devices tightly integrated, making it inconvenient for anyone with an Apple laptop to switch to Windows. And second, to have a privacy-sensitive communications tool that reduces Facebook's influence. Apple has successfully positioned itself as a more privacy-friendly player than Facebook, which is perhaps the best light in which their comparatively restrictive system can be viewed. They've been willing to go to great lengths to preserve this privacy-first approach; they refused to cooperate with the FBI in a terrorism investigation, for example. This gives Apple great credibility when they argue that their platform's privacy-preserving features are essential to preserving human rights. If they can take on the FBI, the thinking goes, they can probably handle FB.
Signal was founded by a well-regarded security researcher and funded through a private foundation backed by WhatsApp cofounder Brian Acton. It's a more ideological enterprise, meant to defend privacy as a value rather than as a business proposition.
Telegram, too, is a values-driven company. Its founder, Pavel Durov, founded a successful Russian social network, and was forced out in 2014 after refusing to cooperate with Russian authorities. He's since created Telegram as a sort of Putin-proof service. (The app's relationship with the Russian government remains complicated.)
H. L. Mencken pointed out that "The trouble with fighting for human freedom is that one spends most of one's time defending scoundrels." (He wasn't complaining. Mencken obviously got a big kick out of scandalizing non-scoundrels with his views.) End-to-end encryption comes amazingly close to being a technological solution to this human problem: if every message is private, the scoundrels are hard to identify. A messaging app can have a bad reputation, but it's technically infeasible to find out if that reputation is true.
Some messaging apps, either in a concession to the growth imperative or as a useful feature, offer public chats. Having a privacy-first approach with optional public chats means giving pseudonymous or anonymous people a chance to say whatever they want, and it gives the media a chance to look at the worst kinds of communication the app enables. This gives the media ample material for pieces about Telegram's unmoderated extremist group chats.
Encrypted chat providers are in a strange situation, where they're worried about two kinds of regulation: from democratic countries worried that authoritarian extremists are coordinating on their platforms, and from authoritarian countries worried that pro-democracy activists are doing the same thing. It's very hard to design a technology that protects dissidents who want to bring down the government of Belarus and not have it do an equally good job of protecting dissidents who want to bring down the government of the United States. It's a bit like asking the physicists at Los Alamos to please tweak physics so nuclear weapons don't work on US soil. Telegram does take down some public calls for violence, but, by design, can't affect the private ones.
Telegram is run by an ideologically libertarian billionaire with technical chops and a personal vendetta against government restrictions on speech. It's very unlikely that they'll change their ways. And if the race is to appeal to free-speech absolutists, any compromises Telegram makes will make Signal a better choice. (And since the encryption protocol Signal uses is open-source—is, in fact, used by WhatsApp and Facebook Messenger—it's possible for an even more absolutist app to show up if both start compromising.)
The Mencken Problem isn't fully solved through technology, but it's often solved through a combination of aggregate transparency coupled with individual privacy. There are many good reasons to personally value privacy; Tim Cook isn't just talking about that because of his fiduciary duty to crush Facebook and redistribute its margins to Apple shareholders. And there are many good reasons to have a strong preference for free speech, or at least to promote a system where speech is extremely free at low granularity (person-to-person, within small groups) and only lightly restricted beyond that. That's a value judgment, but it's a value judgment shared by people with the economic and technical resources to make it a universal. And as it gets more universal, using the apps gets less suspicious. Using Signal used to send a signal of its own, but as the app gets more common, it's just background noise. Telegram users might have wanted privacy for a dubious reason, but now it's a place to get updates from world leaders.
It's hard to stop a trend when there are multiple motivations for it, and unrelated pools of capital backing it. One feature of privacy is that it goes away easily and is hard to get back. If you say something unspeakable, you can't easily unspeak it. Since social media expands the hostile-reader surface area, lax control over privacy turns into a liability that compounds over time. As it turns out, the smartphone addicts who first adopted Snapchat were onto something: people like privacy, and if aggregated user information is an asset to some companies, it may be a corresponding liability to their users. For better or for worse, more secure messaging will continue to grow.
 The company has denied it, repeatedly, and made credible claims that it's not feasible based on app permissions and the amount of battery life and bandwidth required to either maintain a continuous audio feed 24/7 or accurately transcribe text on the device itself. As Antonio Garcia Martinez described a related theory in his excellent memoir of working at Facebook "This was like being accused of fathering Scarlett Johansson's love child. I wish I could even reasonably be suspected of pulling that off."
 In general, people are more willing to say interesting and controversial things in private but under their real name; the model for promoting the most incendiary possible commentary is public but anonymous. And an app that supports privacy in small group chats will tend to support it in larger, open venues.
 The NYT says the app "faces scrutiny" over its content. This is sort of like when a bank robber hands the teller a note saying "This is a bank robbery." Technically, accurate and timely information! But a little short of full disclosure.
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.
A simple model of a stock brokerage is that it's an almost fully-reserved bank. The brokerage has clients, the clients have positions, and in one sense each position is an asset on the broker's balance sheet, while the customer's ownership of that position is a liability. So your broker is a sort of bank, that takes deposits in dollars, but also in shares of IBM, treasury bonds, far-out-of-the-money call options on AMC ("It's like a Netflix—but you can catch Covid"). Normally, this bank tanks very little risk, but there are a few things that can go wrong: trades take time to settle, which creates a brief liability mismatch. Brokers put up collateral to ensure that, in the event that they go out of business before the trade completes, their customers won't lose their assets. When trade volume rises, and the volatility of the assets rises, this creates a larger demand for collateral.
This is, day-to-day, a problem brokers are able to manage. Their capitalization needs don't change that much. But when users are all piling into the same volatile trades, the need for capital rises suddenly. As a result, Robinhood stopped processing trades in GameStop, except trades to unwind existing positions. (As did several other brokerages: Interactive Brokers, WeBull, and Public, for example.)
This story certainly sounds sinister; it matches the appearance of strings getting pulled in order to bail out hedge funds. But the brokers' actions also look like the actions of any financial intermediary faced with a sudden increase in uncertain obligations. (And there were not stories about brokers like Vanguard and Fidelity, which have fewer day-traders, blocking GameStop trades.) When volatility is high, brokers often act in their own interest. This has happened before: one major short-selling firm was basically shut down mid-crisis because their prime broker raised margin requirements. (The prime broker in question denies much of this.) While it's rare in the US, brokers can go under because of client losses. FXCM, for example, had many clients who were borrowing the Swiss franc to fund other currency bets. It was a stable currency with low rates, until the Swiss gave up on keeping it stable and allowed it to float; it instantly rose 45%, wiping out many clients many times over and forcing FXCM to get rescued by a larger financial institution, and to rescue many of those clients in turn.
Which doesn't excuse Robinhood, WeBull and the rest. They communicated inaccurately, but not poorly, because an accurate description of the problem was "We're more of a bank than we realized, and we're in danger of insolvency." And that would lead to a run on the bank—and definite insolvency. WeBull's CEO did clarify this later on, and with enough time to digest it, and new funding from their venture backers, the bank-run risk is minimal.
In a way, WallStreetBets' GameStop experience is the culmination of efforts to give retail investors an institution-quality experience. As Josh Brown points out, WallStreetBets is a scaled-up version of an idea dinner. It might seem more raucous than how financial professionals behave, but competitive hyper-bullishness and hyper-boorishness are not restricted to reddit and Discord. Most individual investors don't lever up enough, or get into crowded enough trades, that their broker raises collateral requirements at the most inconvenient possible moment, but this does happen to institutions. And professional investors often develop somewhat conspiratorial instincts—the more research you've done before a trade, the more losing money on it feels like the result of sinister forces trying to thwart you. After many layers of indirection, WallStreetBets and Robinhood have given retail investors a version of the professional experience.
(In related news, joining a class-action lawsuit against brokers has been productized already.)
Too Much of a Good Thing
In December, I wrote about how Roblox and Affirm were delaying their IPOs because the market was too good ($), and they didn't want to miss the opportunity to raise more. Affirm is now public (first-day pop: 98%). Roblox's IPO has been delayed again, due to SEC scrutiny of their accounting. Specifically, the SEC worries that their revenue recognition is too conservative. I wrote a bit about this in my Roblox writeup:
Some of their digital goods can be consumed immediately (say, a virtual pizza). Some can be used over long periods (a virtual hat). But since every game on the platform has its own arcane rules, it’s actually hard to tell which is which. What if the virtual pizza gives your avatar virtual calories, and that particular pizza prevents your in-game character from starving to death. Now the pizza is a long-lived asset!
After what I can only assume were some very memorable discussions between accountants and game designers, Roblox concluded that the only appropriately conservative policy was to treat all of the in-game goods as durables, and to recognize revenue over the expected time a player would participate in Roblox, which currently averages 23 months.
This is insanely conservative. The real-world equivalent would be Chipotle selling you a burrito, and recognizing 32.6 cents per month of revenue from now until October 2022, starting only after you take the first bite. It’s necessary, because digital goods are messy, but it means Roblox’s GAAP revenue numbers are basically irrelevant.
In general, companies that sell digital goods have difficult accounting, although measures of their bookings and cash flow provide a good look at the health of the business. For Roblox, it's especially complicated, because so many of the digital goods are designed by users, not the company, and because their currency circulates in-game. Buying Roblox means making a bet on the Robux Economy, and measuring the P/E ratio of a country is a difficult exercise.
The Economist has a good writeup on China's property market ($). People have been calling it a bubble, and expecting it to pop, for decades. But in a country where public officials are promoted or demoted based on GDP growth, and where the property sector is a quarter of GDP, that's a problem the government is keenly aware of as well. Essentially, the policy has been a complex system of quotas, subsidies, and restrictions, which generally keeps prices from crashing but tries to tamp down bubbles. It has many unpredictable second- and third-order consequences—you know financial regulations are getting pretty fine-grained when they start creating fake divorces—but the combination of regulating supply and consumer demand at a local level, while adjusting the money supply and property development firms' leverage at the high level, has kept the market surprisingly stable.
The day after Nissan announced plans to sell its last non-electrified cars in major markets by the 2030s, GM set a 2035 deadline for going fully electric. Automakers are doing interesting things to the long-term demand picture for oil, and giving their suppliers some lead time to scale up production of batteries and other components. There's a PR race to have the earliest deadline for zero net emissions, constrained by the difference between current EV economics and what the cost curve will look like a decade from now.
Taiwan and China
China has used its success at curbing Covid-19—positive GDP growth, pool parties in Wuhan—as an argument in favor of its system. And it is true that the CCP was able to marshall more resources, more quickly, than most other states. But one of the many trend deviations caused by 2020 was that Taiwan reported faster economic growth than China for the first time since 1990 ($, FT). Technocratic competence can exist in many systems, though it seems to spike under specific circumstances and then gradually decay over time.
Local newspapers have been declining, in number and in content per paper, for decades. Nextdoor didn't cause this—blame cable news, then online news, and also the shift in work schedules that led to one-paper towns. But it's exacerbating the trend. Nextdoor has very different editorial standards than a conventional paper, but its success can be attributed to carrying the same kind of it-bleeds-it-leads content that is a staple of local news. Nextdoor doesn't put individual crime/accident stories in a broader social context, but the broad social context is usually not why people read a news story about a murder that happened in their neighborhood. In a way, it's a division of labor: national news organizations talk about national trends, and local ones, including Nextdoor, highlight instances of the broad class that are salient only to their readers.