Crypto as an Overheated Emerging Market

Plus! Rolling Blackouts, Floating Power Plants; Scalable Review Fraud; The Risk of Writing Options; Socially Responsible Hackers

Crypto as an Overheated Emerging Market

There's a common pattern in emerging market countries where growth starts with something fundamental—a new government that launches reforms and enables catch-up growth, a recovery from civil unrest or war, or just an increase in the price of the country's main export. Some of this growth compounds; higher exports beget higher domestic investment (the profits have to go somewhere) which creates more avenues for growth.

But, over time, there's an unfortunate shift: more growth comes from borrowing, rather than reinvesting, and more of that growth is channeled into property development and the loans that back it up. As the country's financial system grows, a Minksy dynamic emerges: bad loans can be serviced with more loans, and as long as the supply of credit keeps growing, unprofitable investments can get flipped before the debt comes due. This starts to crowd out other kinds of investment: money flooding in tends to cause currency appreciation, which makes exports less competitive, and real estate deals are higher-status than operating farms, mines, and factories. And then, the growth of interest obligations exceeds the growth of credit, and the system falls apart.

This cycle happens in all sorts of economies. It describes Japan at its 1990 peak, for example, when the four most valuable companies in the world were all Japanese banks. It also describes the US in the run-up to 2008, when American banks were powerful and bank-like entities like AIG and GE were growing fast by levering up to buy or guarantee structured products. But it's most extreme in poorer countries, because of denominator effects: an impoverished country that gets a bit richer can, for a while at least, sustain a modern-looking financial center and a growing balance sheet. And even a small shift in global asset allocation means a torrent of money into a country with a low GDP and low ratio of market cap to GDP. So this effect has been most extreme in places like Thailand, Brazil, Russia, or Argentina: big enough to get attention, but small enough that even a little attention from New York, London, and Tokyo distorts their economy.

The crypto market is going through a similar cycle right now. As with emerging markets, there is real value creation happening, but there's also real crowding-out from levered speculation. Just as real estate is a sort of index fund on an underlying economy, the cryptocurrencies themselves represent a broad-based bet that blockchain-based protocols will be an important way to transfer value in the future.

But the crypto-lending market is different. Consider the roster of investing options on These represent opportunities to buy cryptocurrency, and lend it through some protocol that will allocate loans to the borrowers who a) meet credit limits, and b) are willing to pay the most.

The actual concept of DeFi is pretty compelling: build a financial system out of smart contracts that talk to each other, rather than humans that talk to each other. Finance is already drifting in that direction. A fundamental investor at a hedge fund might lean on a data science team to extract a fundamental signal from reams of data, and then execute trades through an algorithm that reduces market impact. But full automation is a long way away; systematic strategies still use human input, both to create and update the strategies and to manually decide when something is going wrong. An algorithm is perfectly happy to convert 212 orders into 4 million orders due to a bug, but it takes a human being to notice that there's a broken loop in the code somewhere and to pull the plug.

Within known bounds, computers are miles better than humans at repeatedly executing algorithms. Computers are also much better at backtesting strategies to see if they work based on historical data. But full automation doesn't play nicely with dynamic markets, both because of potential bugs and because any exploit of a fully automated system can be repeated at arbitrary scale. If a person notices that they've tossed a coin over a hundred times in a row and gotten heads each time, they'll suspect that something unfair happened. An algorithm may well do the sophisticated thing and store the improbably low odds as a double rather than a float.

DeFiRate has many rates posted for many types of currencies, but there are only two important questions you need to answer to see if this is sustainable:

  1. Is there anyone who, in the course of their business, naturally wants to have liabilities denominated in Bitcoin, Dash, the Basic Attention Token, etc.?
  2. Is there any counterparty that can generate real-world returns that make 10%+ interest on demand deposits a good deal? And, if so, why don't they get access to capital through some conventional, and much cheaper, means?

The answer to question #1 is not quite "no." Crypto miners, for example, have revenue denominated in whatever currency they're mining, while their costs are generally denominated in terms of whatever the local currency is where they buy electricity. But very few parties need liabilities denominated in USDC, a stablecoin pegged to the dollar. Certainly, those who do are generally better off borrowing actual dollars, which are much easier to earn than USDC.

The answer to question #2 is pretty much "no": there are not borrowers who are investing in real-world projects such that a) they can earn enough that borrowing at 10% is worth it, and b) they can't get funding elsewhere. Crypto projects themselves are fundable, both by selling equity and by issuing tokens. And the non-crypto economy has plenty of ways to access capital.

Short-term debt has an interesting adverse selection dynamic, where the market in riskier debt doesn't clear. If someone is borrowing for a five-year term, it can make sense that they'd borrow at a high spread to other borrowers. They might be financially distressed now, but a lot can happen in five years, and the lender is being paid to take risk. But shorter-term funding doesn't function that way; if someone is a sufficiently unreliable creditor that they can only borrow at 800 basis points above a good creditor, then lending money to them on a short-term basis implies the intent to pull that money out (or, equivalently, stop rolling over the loan) the instant it looks like they won't be able to pay their loan. And since that short-term liability is presumably the one they have to pay first, the outcome of a shaky borrower taking out a series of short-term loans at high rates is, in effect, that they give up all of their remaining liquidity to the short-term lender and promptly become insolvent once the loan is called in.

At least, that's the equation if there's a real operating business. DeFi loans are, of course, intended to fund leveraged speculation in crypto assets, not loans. That is a market where borrowers are willing to pay high interest rates (since cryptocurrencies are some of the most volatile assets on the planet, crypto traders are already willing to take risk; levered crypto traders are just taking it a little further). Take the first five assets listed on DeFiRate:

For each of these products, the underlying expectation driving interest rates is that speculators will get a positive return on their speculations, even after paying interest, and that if they lose money they'll get margin-called before their collateral runs out. At best, BlockFi might represent a way to borrow against appreciated assets to defer taxes, but that's a pretty limited use case.

US equity markets used to function the same way. Right now, margin requirements are strict enough for retail investors that margin rates are extremely low, and reflect price discrimination rather than any real credit risk. But it wasn't always so: margin rates used to fluctuate depending on both the demand for speculative credit and lenders' nervousness about asset prices. Some of that call money subsidized real growth: a speculator buying shares in an electrical utility that frequently issued stock was turning that cash into generators, transformers, and power lines. But eventually the high rates for call money drew in capital that would otherwise go to more productive uses; by the peak, companies were selling stock to levered buyers and parking the money in the call money market where it funded more of the same speculation.

The concept of decentralized finance is an interesting one, and it will no doubt produce some very useful financial institutions in the future. But rules-based lending is naturally attracted to the kinds of loans that require as little human discretion as possible; an algorithm can issue margin loans by looking at collateral, volatility, and market depth, but lending against working capital or hard assets still needs some human input. So DeFi is naturally scaling in the direction that it's best suited to right now, which is in providing liquidity for crypto traders. Liquidity produces underlying value when it's a complement to real-world investment, but it adds friction when it's a substitute for it, and unfortunately crypto lending is stuck in substitute mode.

This could unwind gracelessly in a crash (although there are enough institutions with real-money crypto positions that a crash wouldn't send prices to zero). It could also shift in a more gentle way, if regulators decide that DeFi is close enough to banking, and USDC is close enough to dollars, that DeFi needs to be regulated. There isn't a direct way to regulate a smart contract, since the source code doesn't have any notion of what the rules are. But regulation can affect which pools of capital can touch smart contracts. One possibility is that the onshore/offshore separation between crypto ecosystems will deepen, and since there's ultimately more institutional capital that wants to be onshore, the offshore piece will slowly wither away. That would be bad for prices but good for crypto1—it would mean that there's a technological ecosystem for decentralized finance, and plenty of brand awareness, but that actual DeFi projects will only start collecting assets and making loans again when there's something in the real world for them to invest in.

Disclosure: I own some Bitcoin. I have generally sold to diversify, but I don't actively trade it because I'm optimistic about its use as a reserve asset and pessimistic about my ability to time market inflections. Given the reflexive nature of reserve assets—you want to use the same ones everyone else does—an overheated market is a weak timing indicator since it increases adoption even though it raises risk. I also own some Urbit, a far weirder and more interesting asset that is crypto-adjacent but very different.


Rolling Blackouts, Floating Power Plants

South Africa plans to lease ships with power plants ($, FT) to provide emergency power after the state-run utility imposed rolling blackouts. Normally, there's a kink in the supply elasticity chart for electricity: it's either available instantaneously or with a years of lead time. But as it turns out there are flexible short-term options which, while expensive, risky, and polluting, do at least keep the lights on longer. One thing this story illustrates is that, when state capacity is limited, it's important to have reservoirs of competence in the right places. Raising money, putting together a tender, and choosing a vendor are not easy tasks, but it's a lot better for a country to deploy its resources towards providing baseline services rather than providing expensive patches when they fail.

Scalable Review Fraud

A service that sells fake Amazon reviews left its database exposed, revealing 75,000 accounts that posted reviews. Since the breach covers messages, it's a good look at how these reviews get past Amazon's filters: minimum word counts, media types, a realistic delay between receiving the product and writing the review, etc. Any scalable growth strategy lends itself to scalable parasitic strategies; it's the corporate equivalent of a monoculture that's vulnerable to a single parasite. Amazon has its own internal immune system for catching this, and presumably tests out buying and selling illicit reviews to see what patterns it can detect. As Amazon matures and growth slows, fake reviews will likely drop through attrition: a merchant who gets caught using them will get kicked off the system, and the bigger the site is the higher the barrier it can impose for any new merchant.

(Disclosure: I own some AMZN.)

The Risk of Writing Options

The CFTC whistleblower program pays whistleblowers 10% to 30% of the value of fines that result from information they offered. To fund these payouts, the CFTC has a whistleblower fund with $100m in it, which gets topped up by fines. Unfortunately, they are now potentially going to pay out over $100m to a single whistleblower ($, WSJ), and can't afford it.

In one sense, this is a boring story about intra-agency accounting. The US government can certainly afford to spend $100m on something, and if the government is collecting a much larger fine, the payout doesn't even need to be financed. But it illustrates the dangers of writing options. I've argued before that this program actually encourages people to delay reporting fraud, at least as long as they know they can be the first to report it. Someone who has a solid paper trail indicating bad behavior owns an option on the fine for that behavior with an unknown expiration date, and their optimal choice is to hold the option as long as possible. Any financial actor that systematically writes options is going to deal with some surprisingly large losses, in part because the availability of cheap options affects everyone else's behavior.

Socially Responsible Hackers

The hacking group that shut down the Colonial Pipeline has promised to exercise more discretion in choosing targets in the future. They say that the pipeline itself wasn't their target, just the corporate records of its owner. This actually makes sense: the pipeline was shut down as a precaution, not as a direct consequence of the hack. In a way, the good news that hackers aren't directly trying to shut down crucial infrastructure illustrates the bad news that they can. They already operate in a strange gray area where they a) break the law, but b) follow certain rules, like "don't go after targets in Russia." By building the capability to go after more crucial targets, but choosing not to, their behavior is the cybersecurity equivalent of a nuclear weapons test: a way to demonstrate offensive capabilities before it's time to use them.

  1. It's a cliche to respond to every piece of bad news with "actually, this is good for Bitcoin," but pre-registering this view is different.