Banking When You Can’t Bank on Anything: Part 2

Plus! Taiwan and Covid, Defense Tech, The Right Amount of Fraud, Prizes, Airshows, Reliance, Moderation

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 14,654 subscribers, up 727 week-over-week.

In this issue:

Banking When You Can’t Bank on Anything: Part 2

Last week, we looked at the long and tragicomic history of emerging-marketing banking, full of ill-considered booms and painful crashes. Today, we’ll look at how these banks work today, and what has changed.

First, there are two definitional issues. One: is “emerging market” a meaningful category? And two: Just what are banks for?

“Emerging market” is a marketing term. It was coined in 1981, by a mutual fund founder who was having trouble pitching his Third World Fund to investors. It covers a wide variety of countries—some of which, like China, are most definitely emerging into superpower status, others of which show no signs of emerging any time soon. From a financial standpoint, “EM” means “China”—the MSCI Emerging Markets Index is 42.5% weighted to China, and every one of the other large index constituents (Taiwan, South Korea, India, Brazil) has China as its largest trading partner. While China is the biggest variable reflecting the group, each country has a different structure and different internal constraints. For some countries, the core constraint is insufficient capital to fund growth. For others, especially energy exporters, the problem is what to do with a trade surplus that can’t be profitably invested in the domestic market.

Banks, too, require some definitions. Their function is obvious in a market economy, but banks have uses everywhere. Soviet Russia had banks. China Construction Bank, currently the second largest bank in the world with $3.6tr in assets, was founded in 1954—an explicitly communist country, ruled by Mao Zedong, and just wrapping up a campaign of mass-murdering landlords was also busy starting a bank. Even in a command economy, banks are useful for two things: first, they perform a sort of bookkeeping function, keeping track of how much of the state’s resources are being allocated to which projects. And second, they’re a good spot for foreign currency reserves, which allows a country to import and export goods without all international trade being managed by the central bank.

And that’s a good summary of what banks do today. They manage a literal exchange rate, between different currencies, and an abstract one, between current and future spending.[1] In the developed world, the latter function is by far the most important one. Americans who want to vacation in Bali have no trouble acquiring rupiah if that’s what they want to spend there. But an Indonesian factory owner who needs a new piece of equipment from an American, Japanese, or German firm will need to find the dollars, yen, or euros somewhere.

In last week’s historical examples, this often made banks the weak link in a growing country’s economy. They’d import foreign currency, pass it on to corporate borrowers, who would then use it to buy equipment. Some of that currency would leak out into consumer purchases of imported goods. And when banks lost access to currency, they’d have to call in their loans and the economy would slow (in some cases) or completely collapse (in others).

It makes logical sense that countries that get more integrated with the global economy could switch from being insufficiently financialized (small business owners can’t get loans, so small businesses stay small) to over-financialized (property developers can borrow against one incomplete luxury hotel to build another one), and in short order. These economies are small, and markets are vast, so a slight increase in global lenders' interest in a given market leads to a vast increase in available capital.

That’s part of a general problem in emerging markets. A classic development story is that Western advisors or consultants will parachute into a poor, unstable country, and suggest that they adopt very up-to-date Western institutions wholesale. The usual result is that Western advisors figure out that something that’s worked when codified in law often reflects much older cultural norms. But they only find out when they try to have the law without the norms. When countries adopt Western labor laws and pollution rules, for example, the usual result is that those rules are intermittently enforced and lead to uncertainty and bribery.

Finance, though, is a partial exception, for two reasons:

  1. Over-developed financial institutions let a country absorb more financial flows from rich countries. While that’s not a low-risk strategy, it’s a high-reward one: China’s industrialization was partially funded by outside risk capital, and a century earlier, American railroads and early factories were funded by British investors.
  2. It’s a way to partially reverse the brain drain poor countries face. One mechanism for developing countries over-regulating is that some of their best and brightest attend schools in rich countries, then return home and work for the government, where they implement ideas that they’ve seen work well elsewhere. If they come home and work for financial institutions instead, they at least have some P&L discipline to make them more practical.

This means that banks can productively copy some of the institutional frameworks that richer countries adopt, as long as they keep risk low. And that, as it turns out, is the story of the last ten years. The Basel 3 framework requires banks to maintain significantly more capital and liquidity than they’d previously needed, and to accumulate excess capital during good times so they can safely take losses during bad times. Other financial games, like letting consumers take out loans denominated in foreign currencies, mostly didn’t survive the Swiss franc’s revaluation in 2015, and have since turned out to be far more trouble than they’re worth.

A highly levered, non-transparent bank whose short-term liabilities are not denominated in its home currency is a very exciting speculative asset. A bank that mostly takes deposits in local currency, makes loans in the same currency, and takes on a modest amount of foreign exchange risk, backstopped by a central bank with ample reserves, is a much more sedate business, less of a hedge fund and more of a utility. And that’s the direction most emerging market banks have gone. Meanwhile, currency markets themselves have gotten more sedate, and much more transparent. The BIS produces lengthy reports detailing the world’s dollar funding situation, which a) makes the scope of overall foreign currency funding risk easier to track, and b) shows that there just isn’t as much of it as there used to be. At the peak in 2007, dollar-denominated cross-border liabilities were about 28% of GDP, and now they’re around 26%. Euro-denominated cross-border liabilities have dropped from 32% of global GDP to 18%.

Within banking systems, policy choices also have an effect. A procyclical banking system is bad for the country but generally good for the banks. They lend when it’s profitable, and they pull back when it’s risky. Many great fortunes have been made by financial contrarians, but they need either liquidity or lack of leverage, because a levered contrarian who buys the dip when prices are down 10% can get wiped out if they drop another 20%. In many countries, policy banks act in a procyclical way, making more loans when the economy shrinks, even though credit conditions are worse. This can mitigate recessions, but also makes extreme ones much worse. China narrowly escaped this problem in 1998, when their banking system helped keep the economy afloat despite a contraction in the region. The result was that GDP growth decelerated only slightly, but their largest banks were all insolvent and had to be recapitalized.[2] So developing countries generally have a mix of policy banks, which engage in bad banking but good macroprudential policy, and private-sector banks, which back more intelligent projects but don’t always show up when they’re most needed.

There are still countries that go through classic emerging market crises. But they’re either very small (Laos is in trouble, but its total public debt is around $10bn, or 0.05% of the outstanding value of US public debt) or they’re due to extreme, persistent mismanagement (Turkey’s financial system is in trouble, but this is due to years of pressure on the banking system to source foreign currency, and also probably their president’s novel theory that higher interest rates cause inflation).

Hyper-financialization has created another unusual reason we just don’t have crises like we used to: financial systems are more distributed, and have more diverse strategies, so it’s more common for a financial system to partially break down, and for the government to bail out the part that can be saved. If the financial ecosystem consists of a central bank and some semi-privatized banks that take orders from the government, any failure in one part of the system quickly hits the rest of it. But when there are privately-held banks, policy banks, and shadow banks, governments can let the shadow banks die, top up the balance sheets of the more official ones, and not worry that the system faces threats to its overall legitimacy.

Overall, emerging market banks have finally evolved into a success story for globalization, rather than being a metonym for everything wrong with it. They switched from connecting with the rich world on a short-term transactional basis to copying more of the rich world’s institutional frameworks, especially limits to leverage. And those frameworks got updated quickly when American and European banks learned some harsh lessons about credit risk and liquidity in 2008. The rich world has even gotten better about restructuring loans en masse instead of waiting for each one to default. So banking is less of a swashbuckling business than it used to be, and more of a socially useful way to encapsulate global economic complexity into institutions that are fairly easy to understand and much less likely to blow up.

[1] It’s easy to get carried away with analogies, but this one is surprisingly durable. What is inflation if not a balance of payments crisis caused by the present importing too many consumer goods from the future? What is deflation if not an overvalued currency that artificially holds back consumption?

[2] A great read on this is Hank Paulson’s Dealing With China; he worked on Goldman’s investments in this area, and then negotiated with China as Treasury secretary. I find the book unforgettable for another reason: one of the little details in it was his remark about visiting China during the SARS epidemic and getting his temperature checked before every meeting with government officials. This detail added a bit of techno-authoritarian flavor to the narrative when I read it in December of last year, but I’ve thought a lot about that since. Total SARS infections in China were 5,327, and the peak daily case count was around 400.

I’d like to thank several helpful readers for their discussions and feedback, including Sean Pawley, Lado Gurgenidze, and Marc Rubinstein. Any errors or logical reaches are mine. Recommended further reading:


I wrote a piece on Marker about the rise of Venmo and Cash App. I also joined the Palladium magazine podcast to talk about the economic response to Covid, inequality, and whether the US economy is fake.

Inside Taiwan’s Covid Response

Logic Magazine profiles Taiwan’s hyper-competent response to Covid-19, an atoms-and-bits approach that sourced PPE, quarantined the high-risk, and involved frequent app launches by private and government actors. A sample:

Howard Wu, a programmer and member of g0v, noticed that many of his family and friends were sharing information in LINE groups about which convenience stores still had masks in stock, back when convenience stores were the primary places to buy masks. He built a real-time “Mask Map” which relied on crowdsourced data to display mask stock levels in different stores. Users’ geolocation data would help them find nearby stores. Since there weren’t any existing comprehensive GIS datasets of convenience stores in Taiwan, Wu used Google Maps to obtain this data. Wu’s site had roughly 550,000 visits within the first six hours.

But relying on crowdsourced data wasn’t accurate enough. Digital Minister Audrey Tang showed Wu’s work to Taiwan’s Prime Minister, who immediately understood its usefulness. The government recognized that it could improve the accuracy of such civic digital tools by providing more up-to-date data. On February 4th, two days after Wu released his digital map, the government announced the switch to selling masks from pharmacies. In a coordinated effort with Tang, the Ministry of Health and Welfare released mask inventory data at pharmacies nationwide that was updated every thirty seconds.

One way to understand the mid-twentieth century is that countries get hyper-competent when they face existential risks. The US had an unusually long spate of that (the depression, the Axis, the Soviets), and a correspondingly long period of high productivity growth and highly functional institutions. Taiwan has existed in the shadow of communist China for about as long, and while the country started out much poorer than the US, it’s now a prosperous place, leads the world in some areas, and hasn’t been conquered yet. Low-grade constant panic may be the secret ingredient to national prosperity.

Defense Tech

Anduril has launched a new drone that can automatically detect weapons and soldiers. Defense is one of those fields, like education and healthcare, that faces the Baumol Problem: it’s labor-intensive, and doesn’t seem to get more efficient over time, so countries that want to have a meaningful military end up spending vast sums—and periodically risking lives—to do it. (A friend who served in the army calculates that, inclusive of the post-9/11 GI bill’s benefits, an 18-year-old enlistee’s compensation is equivalent to a pretax salary of $100k.) But, also like education and healthcare, there are hidden efficiencies that are starting to be exploited. Health is a lot cheaper than healthcare for the unhealthy; drones are a way to make sure the other side is the one that has to perform the moral calculus of putting people in harm’s way.

And on the topic of defense, this essay argues that semiconductors are an increasingly strategic resource, and that the US should offer more subsidies for domestic production. This kind of defense spending is hard to justify politically, since the business is more capital- than labor-intensive. There’s a wonkish argument for it as a matter of comparative advantage: the world suffers from a shortage of treasurys, the US from a shortage of strategic technology manufacturing, and increasing the deficit to subsidize a local fab industry solves both problems at once.

The Right Amount of Fraud

Following up on yesterday’s post: California says fraudulent unemployment claims are rising. The best illustration of the tradeoffs:

The state agency, which has been under intense scrutiny for its lengthy backlog, busy phone lines and spotty customer service, is now drawing fire from lawmakers for its growing fraud problem — and measures to stop it. Some say the agency’s new precautions, such as closing claims that “match suspicious patterns,” could make it even harder for their constituents to receive benefits, especially if they’ve been targeted by fraudsters.

That about sums it up: you can reduce fraud, but only by worsening other problems, and the question is whether unemployment fraud is a more pressing issue—and a harder one to solve after the fact—than unemployment itself.


NASA has a novel way to price contracts: instead of paying companies to build rockets and lunar landers, they’re just buying moon rocks. That’s an excellent way to devolve responsibility down to people who are more cost-conscious. It could lead to some duplication of effort (although contractors could cooperate on some parts of the project while going solo on others), but it aligns incentives well. It’s possible to question the prize model, but it does produce results, such as yesterday’s $3m Breakthrough Prize for designing novel proteins, which could be applied to fighting Covid.

The Airshow

There are, globally, two organizations that can design large commercial planes that a disinterested buyer will consider purchasing. Boeing can do it, and Airbus can. Other projects tend to flop. But there are three organizations that try to do it, and the third, China’s COMAC, has the advantage of a captive domestic market to sell to. China has decided to hold its biennial airshow in November, which is good evidence that the country remains serious about making progress in aviation.

The Reliance Model

A few months ago, I wrote about Reliance’s business of offering investors in India safety from hostile regulators. That model is still in effect, at growing scale: Reliance is rumored to be selling Amazon a $20bn stake in its retail arm. When a country’s economy grows, it allows economic actors to specialize, and Reliance seems to specialize in rolling up domestic companies and then using them to sell outside businesses market access.

Moderation and Politics

Twitter and Google have both launched election-related product updates: Twitter will be flagging tweets that claim victory before the official election results and Google will suppress political autocompletes. These are both prudent decisions that will still look political. Twitter’s calculus is pretty simple: while there’s an official procedure for determining the winner of an election, in practice the winner is whoever gets widely reported as the winner in the news. Arguably, “Dewey Defeats Truman” is a bigger threat to the system’s legitimacy than a questionable vote count, since it affects public perception and, unlike a problem with votes, can’t be reversed through litigation. Google’s decision is harder. It’s difficult to be neutral and factual when the selection of facts determines so much. Is “Trump bankruptcy” or “Biden plagiarism” a political result, or a factual one about a public figure? “Biden crime bill” and “Trump coronavirus” are both queries that satisfy public interest, but it’s basically impossible to have search results for these that a) don’t have some partisan leaning, or b) aren’t perceived to because a truly neutral result will look biased to both sides.

Overall, now is a good time to be thankful that the US has a long lag between when voting happens and when the President’s term begins.