What's a Ponzi Scheme? What Isn't?

Welcome to the Friday edition of The Diff! This newsletter goes out to 24,830 subscribers, up 134 from last week. (Welcome, Browser readers!) In this issue:

In this issue:

  • What's a Ponzi Scheme? What Isn't?
  • Meta
  • ROE and CAPM
  • Expensive Cheap Energy

What's a Ponzi Scheme? What Isn't?

Ponzi schemes have been happening for a very long time, even before Charles Ponzi first plied his trade. Charles Dickens writes about a fictional one in Martin Chuzzlewit. One of the schemers explains that the model of the Anglo-Bengalee Disinterested Loan and Life Assurance Company is to sell very cheap life insurance policies, to be paid for by selling more of the same. Another asks: "It’s all very well, while the office is young, but when the policies begin to die—that’s what I am thinking of," and the response is:

"[W}e had a couple of unlucky deaths that brought us down to a grand piano... I give you my sacred word of honour," said Tigg Montague, "that I raised money on every other individual piece of property, and was left alone in the world with a grand piano. And it was an upright-grand too, so that I couldn’t even sit upon it. But, my dear fellow, we got over it. We granted a great many new policies that week (liberal allowance to solicitors, by the bye), and got over it in no time. Whenever they should chance to fall in heavily, as you very justly observe they may, one of these days; then—" he finished the sentence in so low a whisper, that only one disconnected word was audible, and that imperfectly. But it sounded like "Bolt."

And that is roughly the model: get money in, promise a high return, pay it for as long as the money keeps flowing, and when it stops—then bolt.

A ponzi scheme is not just a money-in/money-out situation. It's a money-in/money-and-story-out one. For the Anglo-Bengalee Disinterested Loan and Life Assurance Company, that story plays up the exoticism of investing in far-flung parts of the British Empire ("It a devilish fine property... to be amenable to any claims. The preserve of tigers alone is worth a mint of money..."). Other ponzi schemes have also had some kind of topical narrative to tie them together:

  • Charles Ponzi claimed to be engaged in a postal service arbitrage: there was an international system for buying stamps in one country that could be redeemed in another, and the prices of the stamps were fixed based on prevailing exchange rates before the First World War. There was actually an arbitrage based on how much rates had drifted, but the problem was the denominations. So Ponzi had a credible way to promise high percentage returns, but no way to achieve the stated dollar returns—the hard part was not trading cheap postage coupons for expensive stamps, but of somehow realizing millions of dollars in revenue one postage stamp at a time.1 Ponzi's scheme was thematically appropriate: he was taking advantage of US economic hegemony after the First World War. The numbers were fake, but the thesis was real.
  • Ivar Kreuger in the 1920s built a match-selling empire—which did in fact sell matches, and, after being nationalized and then privatized, is still publicly-traded today. His method was to borrow money from US and European investors, and use it to bail out failing governments in exchange for local monopolies on matches. This was actually a fairly old model, with the new twist of globalization. Kreuger, too, had a thematically savvy pitch: smoking was getting more common (per-capita cigarette consumption in the US rose 372% in the 1910s and another 123% in the 20s), and companies had an easier time accessing capital markets than deeply indebted governments.
  • MMM, a Russian ponzi scheme in the early 90s, was always vague about its actual operations (its TV ads mostly focused on how much money investors made), although there were certainly people earning the returns it promised in Russia at that time. It was an interesting echo of both Soviet propaganda about how much capitalists benefited from controlling capital and Soviet propaganda about how financiers were swindlers; both its growth and collapse tapped into popular memes.
  • Bernie Madoff's massive ponzi scheme was widely rumored to be a fraudulent operation, but there was disagreement about the nature of the fraud. Since Madoff owned a major market maker, some of his backers simply assumed that his fund was front-running customer orders to generate steady, low-risk profits. Others thought that Madoff ran profitable trading strategies with low fees, but directed orders to his market-making business to profit that way. Madoff's own secretiveness made it easy to infer that he was hiding something, and there was circumstantial evidence about what it was that he was hiding. Which turned out to be wrong (another benefit of secretiveness was that investors didn't know how much Madoff had; if everyone assumed he was running a small hedge fund attached to his market-making business—a business that at one time was responsible for 5% of NYSE volume—the story was more plausible).

These enterprises all started out as legitimate companies, with the possible exception of Madoff's, which may have been a scammy spinoff of a legitimate company instead. And they all evolved into pure ponzi schemes over time. But the line between something that’s a ponzi scheme and something that’s not and not is a fuzzy one: many companies require continuous cash investments to keep growing, and in many cases the justification for those investments is nonlinear future returns instead of immediate ones. Airlines, for example, have network effects (for the hub-based carriers) and economies of scale (for all carriers): each incremental route should deliver higher margins, all else being equal. And other businesses require a continuous investment of time and labor rather than cash; from Goldman Sachs to McDonald's, most businesses collapse pretty fast if everyone stops showing up to work.2 For companies in manufacturing, there’s a need to continuously source more raw materials. And raw materials providers themselves are constantly depleting whatever it is that they’re selling, meaning they’re always on the hunt for more—sometimes an expensive proposition.

Some ponzi-esque environments can be completely non-monetary: a community, a platform matching two sides of a transaction, or even a party, will only work if people show up, and they'll only show up if other people show up. A continuous investment of people is required, and the payoff is in kind. If only a handful of people showed up to your party, and that's also why they all left early, congrats: you are a social Ivar Kreuger. (Like Kreuger, you presumably started out with the best of intentions here.)

And the concept of ponzi schemes sometimes gets generalized to other domains. An investment product that promises returns based entirely on the funding from future investors describes a ponzi scheme, but it also describes Social Security, as critics of the program have pointed out since the beginning. (Granted, the government can and will step in to top things up when benefits exceed payroll taxes, so this is partly a question of the terminology used to describe Social Security, not its operations; it ceases to be a ponzi scheme if you describe it as taxes and benefits that are not connected except in a fairly tenuous, historically-contingent way. On the other hand, it was a much easier policy to advocate when US demographics and life expectancy ensured that it could easily run a surplus; one of the easiest ways to sell those payroll taxes was to treat them as an investment in a pension rather than just a tax).

But we can extend that further: if Social Security is a ponzi scheme, so is having kids: your kids will have a fairly unpleasant time in their 70s, 80s, and 90s if there aren't plenty of people in their 20s, 30s, 40s, 50s, and 60s to run the economy, not to mention providing healthcare. (You think parents pressuring non-parents into having kids is all about the endowment effect, or perhaps that sleep deprivation interferes with memory formation—no! We're just talking our book.) There is always the hypothetical possibility that having kids is a sort of utilitarian ponzi scheme, where it's not fun but it's a whole lot worse if nobody else does it. But it's been self-sustaining for a pretty long time, and at some point the Bayesian argument is that the cycle will continue indefinitely.

Ponzi-looking characteristics will show up in any business where the long-term economics look very different from the short-term ones. It's a common knock on cryptocurrencies, for example: people buy Bitcoin and then instead of using it for global remittances, they just add laser eyes to their Twitter avatars and cyberbully the fiat fans. But that speculation and hostility is also a form of activation energy that ultimately recycles funding back into the crypto ecosystem, where some of it finds its way to useful projects. Some aggressive crypto advocates bully the salty nocoiners, but anyone who is evangelical about buying crypto but not about building on it can just as easily get bullied for being a salty nocoder instead.

The fact that a legitimate business can evolve into a ponzi scheme is evidence that there's a continuum, and that different enterprises slide along it all the time. And that means it's not especially useful to pattern-match, especially early on. An early-stage company often appears to be in the business of turning funding into hype and turning hype into funding, with no endgame in sight. But today's level of overhyped is best measured by what free cash flow looks like ten years from now; smoking guns will still matter, but circumstantial evidence is mostly noise.

As starting an early-stage company gets more capital-intensive—the capital-light ideas are mostly taken, and there's a lot of capital out there—more and more of them will pattern-match this way. Which is a good thing to notice, both because it's a clue about what kind of due diligence will pay off and because it's good to be aware of a source of false positives so it doesn't swamp the real signal.

Post prompted by this tweet from the excellent @Misha_Saul.

(Disclosure: I own Bitcoin.)

A Word From Our Sponsors


There's so much going on in the world of crypto and web3, it would take all day for one person to stay on top of the latest news.

Fortunately, we made Crypto Dispatch so you don't have to.

We scour the web every week for the most important crypto stories and summarize them for you in one fast, entertaining email.

Crypto Dispatch keeps you up to date on the latest trends, tech, and opportunities in the crypto and web3 space in one weekly email that only takes 5 minutes to read.

Oh, and best of all? It's 100% free.

Subscribe for free to Crypto Dispatch today.



Facebook is now Meta, and is now a metaverse company. It was, in operating terms an increasingly metaverse-focused company this year: $10 billion in metaverse spending out of a total of $70-71bn is a serious investment, and if the company expects people to move some of their online socializing straight from Instagram and Facebook to the metaverse, then some spending on the core product is really metaverse spending in disguise. Here's the official announcement (which notes that, like Alphabet, the company is coupling this with new disclosures) and a letter from Mark Zuckerberg.

This Ben Thompson interview with Mark Zuckerberg is worth reading for more context around the decision. Some interesting bits:

We don’t think about this as if different companies are going to build different metaverses. We think about it in terminology like the Mobile Internet. You wouldn’t say that Facebook or Google are building their own Internet and I don’t think in the future it will make sense to say that we are building our own metaverse either.

I do say that they're privatizing parts of the Internet, though; they do this in direct ways, like AMP and Instant Articles, and more subtly by shaping consumer behavior so that the easiest way to use some open protocols is through a closed, company-owned-and-operated platform. That's hard to avoid; given the opportunity, rational economic actors try to privatize a commons if they find it, and this gives them an incentive to reinvest in it (Chrome, Project Zero, Messenger, and Instagram are all great things). An interoperable metaverse may be a technical reality, but it won't necessarily be an economic one.

An aside from Zuckerberg:

Think about what we’ve had over the last couple of years during the pandemic where everyone’s been on Zoom, and one of the things that I’ve found very productive is you can have side channel conversations or chat threads going while you’re having the main meeting. I actually think that would be a pretty useful thing to be able to have in real life too where basically you’re having a physical conversation or you’re coming together, but you can also receive incoming messages without having to take out your phone or look at your watch and even respond quickly in a way that’s discreet and private.

This may be a symptom of an IM-first communications model (personally, I find side channel discussions in the middle of a conversation quite distracting). It's also an interesting view of engagement: Meta monetizes time, but in an auction-based ad model it's really monetizing attention, and already one view of the metaverse is that it offers novel ways to fragment attention during social interactions.

On the financial model:

I think the first job that we need — well, I guess the first job is getting the foundational technology to work. Then after that, our next goal from a business perspective is increasing the GDP of the metaverse as much as possible

This is a thought process that some companies will no doubt find flattering. Payments and ads are two areas where "increase the GDP of X" is actually a great near-term guideline. Payments will tend to take a smaller cut of a given transaction, but can cover more of them; a lot of purchasing decisions are not directly influenced by ads, or at least not in a way that's easily attributable. The main upshot of this line, for investors, is that the metaverse will take a long time to fully monetize; if they can keep the platform's tax rate ($) low to encourage growth, that's what they'll do.

In one sense, Meta is unprecedented: outside of government, there just haven't been many individuals who can devote that much financial and engineering firepower to solving a new problem long before there's an economic model that supports it. The nearest comparison might be the Gates Foundation, but they spend $2-3bn each year, already well below Meta's metaverse spending. And while this is a trivial point, it is worth noting that the "Meta" name will probably stick3: Meta has a controlling shareholder who has decided to focus on a new market, having survived many near-death experiences ($), many of which came from relying on someone else's platform. So they're applying their considerable resources to owning their own.


Ryan Petersen of Flexport argues that port congestion is caused by companies' obsession with high returns on equity. Parts of this are correct: a company that reduces how much inventory it needs to reach a given level of sales will naturally need less equity to produce a given level of sales, and a company that outsources capital-intensive manufacturing to a third party will also shift some big assets off its books.4 But ROE is not the sole model companies use, and many of them care about risk as well as short-term return; a CEO with a multi-year options package that only vests based on long-term performance is probably not going to trade a good quarter this year for a disaster later on. Meanwhile, part of raising return on equity is leaving a smaller buffer against disaster, but a bigger part over time is figuring out where it's possible to substitute good planning and good data collection for mere capital. Walmart excelled at this early in its existence; it needed less inventory because it tracked inventory better than competitors. That didn't mean Walmart was taking massive supply chain risks and hoping they'd pay off; CEOs with a substantial stake in their company and a name that doesn't rhyme with "dusk" generally don't bet the company that way. Since investors tend to punish companies for missed projections more vigorously than they reward them for beating expectations, there's a bias towards low volatility; better to have a modest but predictable return than a highly variable but slightly higher one, and to let investors lever up if their risk tolerance is higher.

Thinking in return on equity terms means asking: how much investment does it take to produce a given outcome? And generally trying to raise ROE means trying to find clever ways to make a business more efficient, which is generally a good thing. Obsessing over near-term ROE, to the exclusion of everything else, is not ideal. For most companies, the lesson of 2021's logistics issues is that they should have more assets on the books, not because they necessarily need spare capacity but because they want to have more control when shipping is at a premium. Many investors look at companies' returns on equity over a full industry cycle; there are some companies that can put up good numbers at the peak but destroy value over time. (This kind of analysis is especially important to private equity companies; companies are illiquid enough that it's hard to tell what broader economic conditions will look like when the PE owner wants to sell. They're usually not selling at the bottom of the cycle unless they absolutely have to.) The difference now is that more investors have a better sense of what the worst point in the cycle will look like for a company reliant on imports.

Expensive Cheap Energy

The Economist looks at the trend of countries shifting to more emissions-heavy sources of power as natural gas prices spike ($). One interesting piece of this is the argument that explicit subsidies for fossil fuels are under 1% of GDP, but implicit ones from externalities are closer to 7%. Those implicit subsidies are heavily weighted to coal and oil, with coal growing relatively faster.

Diff Jobs

A new round of roles for Diff Jobs. If you're interested in one of these roles, or generally interested in seeing what's out there, please get in touch.

  • A company with a novel systematic strategy is looking for an execution trader with equities experience. Knowledge of Python is a plus.
  • A fintech company is looking for front-end engineers: Vue.js, Nuxt.js.
  • We're talking to multiple companies in fintech looking for product managers.
  • Engineers and engineering managers at an edtech company: they are looking for people with experience in functional languages, especially Haskell (does not have to be professional experience).

If you're hiring (or about to be) and you suspect that the people you're looking for are readers of The Diff, please reach out.

This kind of thing happens from time to time. In an interview, Sam Bankman-Fried of FTX described one leg of a cross-country Bitcoin arbitrage that required the trader to simultaneously be Japanese and not Japanese in order to withdraw profits and convert them to US dollars.

A good stress test for a company that prides itself on automation is to ask: how long would it take to collapse if everyone stopped working on it. The answer for any company should be a finite number, of course; they won't run themselves forever. But some companies have tested this: the original Google doodle was meant to indicate that the company's staff had gone to Burning Man and wouldn't be available if the site went down, and there are stories about companies holding all-employee offsites (or throwing all-employee parties) in their early days, and not noticing that their site went down.

Helpfully for Facebook, they chose a name that they already have an indirect claim to. The Chan-Zuckerberg Initiative had a project called Meta, which until recently owned the Twitter handle @meta. (Archive.org shows that handle going private overnight—but with a new logo and link to the Facebook.com page about the new name—and the handle now appears to belong to the former @facebook account. Companies can launch products for which they don't own the brand name—at launch, iPhone was a Cisco trademark—but this kind of transition is a lot easier when the founder has been reaching for the same abstractions for a while.

As one Fintwit writer points out, Japanese companies have been notoriously indifferent to return on equity for decades, and this didn't save their asset values during Covid. (Although, to be fair, Toyota had a chip stockpile—bad for ROE to keep rapidly-depreciating assets on the books!—which helped them for a while but not forever.)