The Pursuit of Happiness

What is an undervalued currency?

Scott Sumner — Tue, 01 Sep 2026 15:43:14 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

What do people mean by an “undervalued currency?” Is that like when people say the stock market is undervalued? Or does it mean something entirely different?

A recent graph in the Financial Times caught my eye:

There’s one thing that I know for certain. At least 5 out of the 6 measures of “fair value” are wrong. And given Occam’s Razor, it is not unlikely that all six are wrong.

The subjective theory of value favored by Austrian economists suggests that value reflects what people are willing to pay for an asset. That’s also the implication of the Efficient Markets Hypothesis. So why isn’t the current value of the yen “fair”?

As far as I can tell, the FT is mostly describing investor views regarding the likely future path of the yen. In that sense, “undervalued” means “likely to appreciate”:

Favourable fundamentals should help the currency strengthen, according to some analysts. “The market can’t price the yen so far away from fundamentals for a long time,” said Michael Metcalfe, global head of macro strategy at State Street.

But I am more interested in the case where people view undervalued as a moral judgment, not a market prediction.

One of the items on the list is “purchasing power parity”. In plain English, the price level in Japan is estimated to be 40.6% below that of the comparison country (presumably the US.) Why is that fact important? A Big Mac costs much more in San Francisco than in Omaha. Does that mean that California is too expensive to compete with other states? Does this seem like a state that cannot compete:

Companies based in the Golden State have drawn around $366 billion of venture capital since the beginning of the year, according to data provider PitchBook. That’s more than three times the amount of venture funding that has gone into the other 49 states combined, and nearly double California’s previous record, set in 2025. New York state ranks a distant second in venture-capital investment, with $27 billion in deals announced so far this year.

It is interesting that the list of six factors in the FT does not include Japan’s current account surplus. But if you learned that Oklahoma had a trade surplus or deficit with New York, would you assume the dollar was not correctly valued in one of the two states? Switzerland and Norway have some of the world’s most overvalued currencies if you use PPP as the benchmark. And yet both countries have huge trade surpluses. If neither PPP nor trade balances are reliable, then how can we make sense of currencies being “undervalued” or “overvalued”?

Off the top of my head, I can think of a few ways in which we might put some meat on the bone, that is, make “undervalued” mean something:

The government might set an official exchange rate that differs from the free market rate. When there is a black market in a currency, it might make sense to speak of the official exchange rate being undervalued or overvalued. That’s not what is happening in the Japanese case.
The country might have either too much or too little nominal spending to promote macroeconomic equilibrium. If there is too much spending, then a currency appreciation might improve things. If there is too little spending, then depreciation might improve things. At the moment, Japan does not seem to suffer from either problem, at least to any significant extent. (Things are never perfect.) At 1.9%, Japanese inflation is close to target. But even in cases where there is a macroeconomic disequilibrium, it probably makes more sense to say money is too easy or too tight, not that the currency is under or overvalued.
A country might have too much saving for its own best interest, or for the best interest of the world. That sort of “excess saving” might lead to a currency value that is lower that it would be with an optimal level of saving. I’d say two things about that case. First, given Japan’s dire demographic situation, it is hard to argue that the Japanese are saving too much. Second, even if they were, it would be weird to call that an undervalued currency. It would be like characterizing the world’s excessive use of fossil fuels in terms of the price of coal being “wrong”. If you think Japan is saving too much, then just say so.

To summarize, there are three ways in which a currency’s value could be “wrong”. In the first case, the official value isn’t at equilibrium. You see that in places like Venezuela, where there are currency controls. In the second case, the equilibrium nominal value of the yen is not consistent with macroeconomic equilibrium. That problem eventually gets solved due to wage/price adjustments. In the third case, the real value of the yen is not optimal because of too much saving—according to a foreign busybody with no real knowledge of Japan that was taught a now-defunct Keynesian model in college:

Practical men who believe themselves to be quite exempt from any intellectual influence, are usually the slaves of some defunct economist. Madmen in authority, who hear voices in the air, are distilling their frenzy from some academic scribbler of a few years back.

In this case, the academic scribbler is Keynes himself.

August 2026 links

Scott Sumner — Fri, 28 Aug 2026 17:06:46 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

Jackson Hole: The good points are that Kevin Warsh indicated that money is important and suggested the Fed should not commit to a future path of interest rates. He also said the Fed must take the blame for the inflation of the 2020s:

There is one signal nobody can miss: The responsibility for 65 months of sustained, elevated inflation sits squarely with the central bank. And that is where it belongs.

Not fiscal policy (Fed’s job to offset) and not supply shocks (NGDP overshoot was just as bad). It was the Fed.

On the negative side, we still don’t have a clear policy. We have a “task force”. The Fed nods toward 2% inflation and high employment, but that’s too vague to be useful. They need something specific like NGDPLT.

Trump Supporters Imperil $4 Billion Aluminum Plant Central to His Industrial Push

Oklahoma residents fear toxic emissions, threatening an effort to restore US industrial might.

Well, it was a nice idea. Perhaps it could be built in China.

Tyler Cowen directed me to a Lucy Ives article discussing Paul Celan:

I struggled to articulate my emotions as I moved through these pages. I almost certainly read the collection in the early 2000s and, composing this review, read it three more times. I was startled by “Backlight,” which offers a version of Celan one encounters only in summary form in his poems: a writer concerned with narrative irony. The last of “Backlight’s” prose units is the most sharply delineated:

“He taught the Law of Gravity, furnished proof after proof, but people turned deaf ears. Then he took off into the air and, floating there, repeated the lesson. Now people believed. But nobody was surprised when he did not come down again.”

Here Celan explores his longstanding concern with the language of persuasion and its pro forma aspects. The protagonist is able to convince his audience to accept the truth of a major law of the physical world only by performing an impossible act (according to the law in question). Despite the witnesses’ new credence, they have no expectation that the speaker himself will obey. It seems to please them when their instructor confirms their own earlier beliefs—permitting them, in this way, to accept what he has to say, while simultaneously invalidating the content of his words. The speaker is put, as we say, into an untenable position.

For years, I’ve been tirelessly promoting the theory that markets are efficient and asset price bubbles don’t exist. The view that business cycles are caused by unforecastable policy mistakes. I haven’t made much headway. Perhaps it will take a successful prediction on my part to finally convince the skeptics.

Public policy in the world’s most democratic nation:

Switzerland and China agreed on a new trade accord that almost entirely removes tariffs on Swiss exports, as Beijing tries to increase ties in Europe amid its ongoing rivalry with the US for global influence.
The deal grants 99.8% of Switzerland’s current exports to China tariff-free treatment, the Swiss government said in a statement Thursday. Swiss investors will also get better access to the Chinese market.

Voters are dumb? Yeah, if you treat them like stupid people. But if there are treated like adults and given real responsibility?

Back in 2022, I did a post describing seven characteristics of nationalism, and how it differs from patriotism. This was the first:

1. Fake history, which glorifies the past and denies a country’s past crimes

And here is what our president spends his time on when his policies don’t seem to be working:

Donald Trump wants Washington museums to be less ‘woke’ on US slavery

Matt Yglesias has some very astute comments on so-called “public opinion”:

A much more interesting question is how you navigate tensions between what people say they want and what outcomes they would welcome. An obvious example is price controls. It’s clear that with the cost of living on everyone’s mind there is now pretty overwhelming polling support in favor of price controls. I think politicians should take this seriously. What they should not do is imagine that if they impose price controls on beef and then supermarkets start running out of beef, that voters are going to say “this is good, this is what I wanted, good job.” Voters are sometimes wrong about stuff and — worse — they’re just not very reflective or self-critical.

Intellectuals view “price controls” in legalistic terms. For the public, it’s more like “Please do something to address the high price of beef.”

States I’ve lived in (red) and visited (green):

I hope to visit several of the others next year. (Am I cheating to include Delaware based solely on I-95? I had to pay Delaware for the privilege, so I feel I deserve to include it.)

Sometimes, I feel like in live in Naples:

Well, almost:

Saddleback Mountain is actually taller than Vesuvius.

The Economist has a couple good articles on how AI might introduce some additional inefficiencies into the the tort system:

AI, like a heat-seeking missile, can lock on to the most obscure provisions of the law—and create carnage. The impact on Britain’s employment tribunals (courts that resolve disputes between employers and workers) illustrates a phenomenon emerging everywhere. AI-induced demand is overwhelming bureaucracies built for the analogue age—from Dutch municipal-tax appeals to the Canadian privacy regulator to parking-ticket tribunals in every major city. In Britain workers now ask large language models, rather than human lawyers, to help them sue their bosses quickly and cheaply. Claims have surged and backlogs grown. A case filed today may not be heard until 2030.
Free, AI-powered legal advice should be good news for workers. Instead, it is proving to be a tragedy of the commons. For workers with genuine grievances, the surge in demand means longer waits for justice. For employers, it means bigger legal bills to respond to claims, both well-founded or fantastical. In the age of AI, a system intended to provide access to justice suffers from, if anything, too much access.

A separate article in the same magazine looks at the effect on the planning process:

Take planning. Officials have developed tools to speed up housing approvals by giving an initial verdict on whether a plan is compliant with the law. Yet local authorities report a surge in AI-generated objections to proposed developments. They are often more effective: AI-aided submissions will cite planning codes and legal precedents chapter and verse. For £45, Objector AI, a platform, will scan a wodge of planning documents, draw up the legal grounds for opposition and write a speech to read out at a public meeting.

To be clear, I still expect AI to lead to faster economic growth. But the speed of the transformation is likely to disappoint some people in Silicon Valley.

Auckland has had some success deregulating housing construction. Here’s the Financial Times:

The city, whose population has risen from 1.3mn in 2010 to 1.8mn in 2025, has in the past decade rezoned 75 per cent of its residential land, allowing single-storey family homes with lawns and garages to be replaced by taller, narrower, terraced town houses. . . .

After an intense building flurry, Auckland’s housing stock is now 618,849 units, which is 101,000 more than a decade ago, equating to a 19.5 per cent rise, according to a 2025 policy paper by Greenaway-McGrevy, James Allan Jones and Chris Crow.
Meanwhile Auckland house prices, with a median value of just over NZ$1mn (US$587,000), have lost a quarter of their value in that timeframe.

It’s not clear that the zoning reforms were the only factor, but the rise in construction of townhouses has been quite impressive:

Matt Yglesias explains why the left is lukewarm on YIMBYism:

But there is a deep reluctance, notably among leftist intellectuals, to admit that we are just correct and a radical program of land-use-law reform would largely solve the single biggest economic problem facing the country.
Because if we can solve the single biggest economic problem facing the country with a liberalizing reform, that calls into question the larger project of overthrowing “neo”-liberalism and replacing it with a new post-liberal economic order.

35. Andy Smarick has an interesting discussion of executive power:

Gorsuch implies a question we and members of the Court simply can’t ignore: If it’s wrong for agencies with executive power to operate independently of the president, isn’t it wrong for the president to control agencies with legislative and judicial powers? Said another way, if we’re going to jealously guard the president’s control of executive authority, shouldn’t we jealously guard Congress’s control of legislative authority and the judiciary’s control of judicial authority? This appears to be what he has in mind: “From here, the only sure path is to finish the journey we start today and restore legislative and judicial powers to where they belong: in Congress and the courts.”

Bloomberg describes a disaster that recently hit Gary, Indiana:

Of the 370,000 customers of the Northern Indiana Public Service Company (NIPSCO) who lost power the day of the storm, just a little more than 3,000 were still without it on Tuesday, most of them in Gary. NIPSCO has promised to restore most power by the end of the day. But that relief has come far too slowly for far too many people; just five days ago, 21,000 Gary residents, nearly a third of its people, had been in the dark for 10 days. . . .
On hot, humid August nights, the lack of power forced people to choose between opening their windows to let in the air but also disease-carrying mosquitoes or to keep their windows shut and stew, as the Chicago Sun-Times documented.

I can sympathize. I grew up 150 miles from Gary and experienced the same lack of AC, but for 20 years.

PS. Solidarity with the good guys, who are being bullied:

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

Say’s Law in China

Scott Sumner — Mon, 24 Aug 2026 16:18:43 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

[In the late 2010s, I did a few posts complaining about our reckless fiscal policy. At the time, most internet pundits thought it was silly to worry about budget deficits, as interest rates were quite low. I pointed out that they would not be low forever. Now the problem is obvious, so I’ll refrain from commenting on the recent Scott Bessent clown show. (Stephen Kirchner has a good analysis, for interested readers.) I like to focus on areas where I might make a difference, where my views are outside the mainstream. In 2018, being a contrarian meant worrying about the danger of excessive budget deficits. Today, it means not worrying about China’s economy.]

I frequently run across articles insisting that China is producing too much. Here is a typical example, from Michael Froman at Foreign Affairs:

But when this latest surge plateaus or recedes, the game could well be up. It is difficult to identify a sector capable of generating another 30 percent of GDP growth for the Chinese economy. The property sector is in a prolonged downturn that could last for the better part of a generation. The infrastructure sector can no longer support new, unnecessary high-speed rail lines in third-tier cities. Manufacturing easily meets domestic demand. Services, which in most economies absorb displaced industrial labor, remain thin because the household share of consumption is too depressed to sustain them. The industries China could still develop—commercial aircraft, AI hardware—are too small to make up the difference.

Here is some data for per capita GDP (2026 estimates from the IMF):

US: $94,430 (both nominal and PPP)

Singapore: $107,758 (nominal) $173,708 (PPP)

Hong Kong $59,640 (nominal) $84,212 (PPP)

Taiwan $42,103 (nominal) $98,051 (PPP) [the CIA say $32,300. ̄\_(ツ)_/ ̄]

South Korea: $37,412 (nominal) $68,624 (PPP)

Japan: $35,703 (nominal) $59,207 (PPP)

China: $14,874 (nominal) $31,596 (PPP)

The PPP figures are clearly inaccurate, something I’ve complained about in other posts. I suspect that Japan and the four “tiger economies” have similar outputs, if measured correctly—say about $60,000 per capita GDP, PPP All five are well below US levels, and roughly double Chinese levels. If China wishes to catch up to these other East Asian countries, then it needs to roughly double its output and if it wishes to catch up to the US it needs to triple its output.

So why do pundits keep insisting that China is producing too much output? Is this like during the Great Depression when President Roosevelt suggested that the real problem was “overproduction"? Yes, it is a bit like that.

China’s per capita output is slightly above Mexico and Thailand and a bit below Argentina and Costa Rica. In my view, the Chinese should try to at least double their current per capita output. But how?

There are two primary ways to boost aggregate output. The easiest method is to print enough money to move from the current position to a position on the production possibilities curve. But that’s not likely to be a game changer—you can expect at best 5% to 10% more output, probably much less. Demand stimulus basically overcomes the sticky wage problem and soaks up the unemployed labor. But China’s main problem is that its labor simply isn’t very productive. To get a meaningful increase in output, you need to move the long run aggregate supply curve (LRAS) to the right, you need to expand the economy’s capacity to produce goods and services.

Don’t get me wrong, demand stimulus is certainly worth doing when NGDP is depressed, and I’d encourage the Chinese to adopt a monetary policy that is a bit more expansionary. But I very much doubt whether this would boost output by even 5%; it would still leave China much poorer than its neighbors.

So that leaves supply-side reforms. China needs to increase its production potential, its capacity to produce goods and services. There are basically two ways of doing this. First, China could keep its current economic system in place and invest a higher share of its GDP. Second, China could do structural reforms to make its economy more efficient, so that existing resources are used more productively.

The majority of pundits seem to believe that China is investing too much. I’m not sure that is true, but I agree with those who worry that much of the current investment is on projects with dubious benefits. Simply adding more investment is unlikely to solve China’s low output problem.

If demand stimulus is not enough, and if more investment is likely to produce disappointing returns, then that leaves only one meaningful option for growth—economic reforms. And there’s really only one recipe for growth that has proved robust—a regime of free markets with reasonable taxes and secure property rights.

Let’s look at the level of economic freedom in the US and the major East Asian economies. I used the Fraser ranking, as the Heritage ranking excludes Hong Kong. But I also put each economy’s position in the Heritage ranking in parentheses for comparison:

#1 Hong Kong

#2 Singapore (#1)

#5 United States (#22)

#7 Taiwan (#5)

#17 Japan (#30)

#38 South Korea (#19)

#108 China (#154)

These rankings are subjective, as the Fraser and Heritage indices use somewhat different criteria. But one thing is clear, mainland China is far less free market than the other countries listed above in any plausible index.

In the past, China grew rapidly whenever it enacted reforms that made its economy at least slightly more capitalist. To join the ranks of the world’s richest economies, China needs to make its economy freer—more like Singapore, Taiwan and Hong Kong. Note that those three economies are composed mostly of ethnic Chinese, so there is no obviously cultural barrier to the mainland China achieving a similar level of success.

I suppose you could argue that there was some sort of selection bias in terms of who migrated to Singapore, Taiwan and Hong Kong, and that those places have people with above average human capital. But that argument doesn’t apply to Japan and South Korea, where living standards are at a similar level. China can and should do much better than its current level of per capita output. It needs to produce far more goods and services.

Someone who has real world experience investing in China might assume that I don’t know what I’m talking about. They might have found Chinese workers to be highly productive. But in a weird way that sort of proves my point. If at an individual level Chinese workers seem competent and hard-working, and yet China’s average labor productivity remains at a fairly low level—similar to Latin America—that means their talented workers are not being used to their best advantage. The whole point of free market reforms is to unlock the latent potential of China’s vast workforce. Anyone who visits China almost immediately encounters lots of workers doing make work jobs, such as aimlessly pushing a broom along the gutter of a Chinese street.

Now let’s go back to the quote at the beginning of this post. What are we to make of this:

It is difficult to identify a sector capable of generating another 30 percent of GDP growth for the Chinese economy.

China would need to increase its GDP by 100% to catch up to East Asia, and 200% to catch up to the US. Are there sectors capable of producing much more than they currently produce? If not, then that would presumably reflect supply side problems, a lack of productive capacity. But if you read Froman’s (pessimistic) analysis, you get the impression that demand is the biggest problem:

The property sector is in a prolonged downturn that could last for the better part of a generation. The infrastructure sector can no longer support new, unnecessary high-speed rail lines in third-tier cities. Manufacturing easily meets domestic demand. Services, which in most economies absorb displaced industrial labor, remain thin because the household share of consumption is too depressed to sustain them.

Again, aggregate demand is a problem, but it explains only a small fraction of China’s vast shortfall relative to the rest of East Asia, and the even larger shortfall relative to the US.

Say’s Law has been widely criticized because it is seen as somehow denying the existence of demand shortfalls. But even Jean-Baptiste Say understood that with sticky wages you could have temporary periods of unemployment, called a “glut”. Nonetheless, Say Law does tell us something very useful. Over any significant period of time, a country can “afford” what it can produce. So don’t tell me that the Chinese cannot afford to buy more stuff.

If there are sticky wages, then a country may temporarily produce less than its potential. But that would not be because Say’s Law is wrong, it would reflect the fact that sticky wages produce a gap between current output and potential. This problem can be easily resolved by printing more money. No country should ever refrain from boosting its output because of a worry that people won’t be able to “afford” the extra output.

Although China would do well to vastly increase its aggregate production of goods, there can be cases of too much production of some goods and too little of other goods, For instance, we continually hear about “overproduction” in the Chinese auto and housing industries. But what is the per capita ownership of cars in China? What is China’s per capita housing stock?

AI Overview says China has 330 cars per 1000 people, while Japan has about 600 per 1000 and the US has about 800 cars per 1000 people. In what sense does China not “need” more cars? China’s housing is about 650 square feet on average, compared to 2200 square feet in the US. The truth is that if China is to become a fully developed country it will need more of almost everything, including cars and housing. Overproduction is not China’s problem.

Photo by Thomas Kinto on Unsplash

The Economist has a particularly egregious example of how pundits misjudge the Chinese economy.

In a lengthy new paper, the [Chinese] commerce ministry seeks to rebut the accusations of industrial excess levelled against China by many foreign governments and economists.
The paper is both rigorous and disingenuous. Some parts bring healthy scrutiny to the concept of overcapacity. The deep insincerity of other sections suggests contempt for countries facing waves of Chinese exports. . . .
Certainly, Chinese companies are low-cost producers, but that is no random market outcome, for the government has helped bring it about. . . .
The ministry’s discussion of state subsidies makes an art of dissimulation. It limits analysis only to direct subsidies—merely the most superficial of the techniques that China uses to favour its own companies. Other crucial forms of support include public investment in chipmakers, cheap land and rules compelling state firms to buy domestically. In the background are the policy settings—above all, a weak social safety-net—that have engendered a perpetually high savings rate. . . .
Foreign companies, however big, stand little chance against a Leninist developmental state that has attained China’s size, wealth and modernity. As Rhodium Group, an American research firm, points out, the Chinese state once limited its intervention to a handful of sectors that it deemed to be strategic; now it pursues an “industrial policy of everything”.

The phrase “contempt for countries facing waves of Chinese exports” seems to equate Chinese exports with a military invasion. Is that true? Are western consumers forced to buy Chinese EVs at gunpoint? Or are these voluntary transactions between consenting adults? I expect more from The Economist, which was founded in 1843 as a voice of reason in the free trade debate. Let’s review a few of the problems with their arguments:

China is not special in either trade or manufacturing. As I showed in the previous post, its current account surplus is relatively small on a per capita basis, roughly 1/10th the level of the smaller East Asian and northern European economies. Pundits forget that China has about as many people as Europe, the US, Canada, Mexico, Japan and Australia combined. Of course it has a big manufacturing sector. Why wouldn’t it?
Subsidies that boost output in some industries come at the expense of output in other industries. Contra The Economist, there is no such thing as an “industrial policy of everything”. Just as you cannot raise yourself by your bootstraps, you cannot boost aggregate output through subsidies. Indeed, subsidies generally tend to reduce aggregate output by making an economy less efficient than a more free market regime. China would produce more if it subsidized less.
Government subsidies increase the budget deficit. Other things equal, this reduces domestic saving and thus reduces the current account surplus. China’s trade surplus is certainly not caused by government subsidies. Other East Asian countries have far bigger surpluses in per capita terms.
Even if subsidies did help China to grow, why would this be a problem for the rest of the world? The only argument that I’ve ever heard that is even logical is the so-called predatory pricing argument—the claim that subsidies are a scheme to monopolize and then later raise prices sharply. But predatory pricing is about the last thing we should fear from China’s ultra-competitive economy. Indeed, the standard argument is that Chinese industries are too competitive to earn decent profits, which explains why their stock market does so poorly.
And what are we to make of this claim?

Foreign companies, however big, stand little chance against a Leninist developmental state that has attained China’s size, wealth and modernity.

If China’s “size” is the issue, would the problem go away if China split up into 30 smaller countries? If “modernity” is the problem, would the world be better off if China remained an ancient kingdom? And why would China being “wealthy” be a problem? Indeed, in what sense is China wealthy? (Certainly not per capita.)

In a recent China Talk post Jordan Schneider, Irene Zhang, and Phoebe Chow point out that China is far from being a “Leninist developmental state”:

How did China grow its EV industry? Hint: it wasn’t really central planning from Beijing. It was a magical concoction of entrepreneurs and local officials using their guanxi to dance around giant SOEs and central-government regulation, leverage private and foreign capital, and just be really excellent business people. That’s led to our situation in 2026 — world-beating companies that have a ton of overcapacity and are set to take over the planet.

I cannot imagine anyone visiting Shanghai or Shenzhen and thinking, “This is sort of like the Soviet Union.” In any case, the Soviet Union was not a competitive threat to the West.

And how about the claim that:
“In the background are the policy settings—above all, a weak social safety-net—that have engendered a perpetually high savings rate.”

Northern Europe has the world’s most extensive social safety net, and yet those countries have high saving rates and run large current account surpluses.

Western pundits might envy China’s success, but how many would like to change places with a country at one half of Japan’s productivity and one third the US level? Pundits tell us that subsidies and trade surpluses explain China’s success, but exactly how much success does China actually have? Its economy has grown enormously since 1980, but from an extremely low base. Let’s hope it continues growing rapidly.

So, what should China do to boost economic growth? (K.I.S.S.)

Free markets and stronger property rights: I’m not going to suggest that China should do no industrial policy at all—even highly successful economies like Singapore, Hong Kong and the US have some industrial polices. China may wish to encourage green industries through a carbon tax. They may wish to promote a chip making industry for national security reasons. Rather I would argue that they should radically reduce their current level of industrial policy, by at least 90%. Eliminate most subsidies. Remove restrictions on internal migration. Privatize SOEs. Move toward freer trade and investment. Remove capital controls. Etc.
Adopt NGDP targeting, level targeting, to make Says Law effective in the short run. (It already applies in the long run.)

The best way to think about complex macroeconomic issues is to keep a few basic principles in mind:

Imports are good. Exports are the (unfortunate) price we pay for the imports we desire.
Trade imbalances are not a problem.
Subsidies don’t boost aggregate output, nor do they explain a country’s success in exporting.
Excessive aggregate output is not a problem. The public can “afford” what it can produce.
Because of sticky wages, NGDP instability can temporarily reduce output below potential. Eventually, the economy self-corrects. Persistently low output is generally caused by supply problems.
China is extremely large. Otherwise, it is a completely normal middle-income country. The world should not fear it doubling or even tripling its aggregate output.
Free market economies are usually richer than more statist economies.

In the media, China is portrayed as a sort of monstrous freak that threatens the world economy. Don’t believe the hype. Textbook economics circa 1999 is still true.

PS. From today’s Bloomberg:

The US is set to impose a 7.5% tariff on Chinese goods over allegations of excess manufacturing capacity before a planned summit between Xi Jinping and Donald Trump next month, according to people familiar with the matter.

Sigh . . .

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

On the bigness of China

Scott Sumner — Thu, 20 Aug 2026 17:54:41 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

Most pundits get China wrong. There are lots of mistakes being made, but many of them center on three fallacies:

Pundits (subconsciously) underestimate China’s size.
Pundits accept the long-discredited mercantilist theory that exports are good and imports are bad.
Pundits underestimate the importance of Say’s Law.

I’ll address the first two fallacies in this post, and the Say’s Law issue in the next post.

For the world as a whole, the current account balance should be exactly zero. As a result, there is a sense in which any country that runs a current account surplus “forces” the rest of the world to run a current account deficit. But this is a strange use of the term “force”, as it is equally true that any individual who chooses to become a net lender “forces” all other humans on planet Earth, in aggregate, to become net borrowers. But let’s not quibble about semantics. What is actually at stake here?

Martin Wolf in the FT expresses a view that I’ve seen repeated over and over again:

As a result, China’s policies look unavoidably threatening to much of the world. If a major trading partner runs huge and persistent trade and current account surpluses, everyone else has to run offsetting deficits. This leads to the shrinkage of sectors specialising in producing tradeable goods and services, as well as huge domestic financial deficits. These, in turn, create protectionist pressures of the kind we have been seeing in the US and increasingly the EU. Yet high-income countries are not the only victims of China’s push for dominance in manufacturing. Among them, argue Shoumitro Chatterjee and Arvind Subramanian, are also the poor countries whose development it blocks.

Notice how Wolf moves from an accounting identity to some highly questionable assertions. Before examining his statement in detail, I’d like to remind readers that the world’s current account surpluses are almost entirely concentrated in two regions, East Asia and the continental part of northern Europe. Here are the approximate figures (from AI Overview) for the first quarter of 2026, the most recent figures available. All figures are surpluses for Q1 only. I have decided to separate East Asia into two components:

China (pop. 1400 million) $184.3 billion surplus or $131.6 per capita

East Asia ex-China (pop. 200 million) $246.1 billion surplus or $1230 per capita

Northern Europe (pop. 140 million) $180.2 billion surplus or $1287 per capita

Here are the individual East Asian countries that I looked at:

Japan $77.75 billion surplus

South Korea $73.78 billion surplus

Taiwan $62.53 billion surplus

Singapore $32.0 billion surplus

Here are the individual northern European countries that I looked at:

Germany $70.9 billion surplus

Netherlands $30.0 billion surplus

Norway $27.2 billion surplus

Switzerland $18.9 billion surplus

Denmark $15.5 billion surplus

Sweden $9.4 billion surplus

Austria $8.35 billion surplus

I was surprised at how small the current account surpluses were outside of these two regions. Even Saudi Arabia had a surplus of only $4.1 billion. I didn’t know whether to put Hong Kong in with China or in ex-China, but its surplus was only $4.7 billion. Again, East Asia and Northern Europe represent the world’s only major current account surpluses.

There is a sense in which these surpluses ”force” the rest of the world to run deficits, but it is equally true that America’s decision to run a $226.8 billion current account deficit in the first quarter of 2026 forced the rest of the world to run a big surplus. In other words, we are forcing workers in China, Japan and Germany to slave away producing goods like air conditioners for our comfort, while much of Europe suffers from sweltering heat.

But that’s not the framing used by Martin Wolf, who seems to view deficit countries as the victims of trade imbalances. Even more oddly, he includes the EU as one of the victims, even though the EU as a whole ran a current account surplus of $129.3 billion in the first quarter of 2026. That’s right, anti-Chinese bias has reached such a fever pitch that highly respected columnists seem to be hallucinating current account deficits where they do not exist.

OK, I’m being a unfair to Wolf, who does not explicitly suggest the EU is running a trade deficit. But the average reader is likely to infer that from the fact that he mentions that China is running a big surplus, and the fact that China’s surplus forces a deficit on the rest of the world, and the fact that protectionist pressures are rising in the EU as their firms are being hurt by Chinese competition. All three claims are true, but do you see how Wolf’s framing tends to obscure the fact that northern Europe has a current account surplus nearly as large as China’s, and much larger in per capita terms? What are we doing here?

You might argue that I’ve cheated by lumping together a bunch of different countries in Europe and East Asia and them comparing them to one single country like China. China’s surplus is more than twice the size of the next largest surplus. But someone could just as well argue the opposite, that it is unfair to China to compare it to countries with much smaller populations. The fact that both northern Europe and East Asia ex-China have CA surpluses in per capita terms that are nearly 10 times larger than China’s surplus suggests that they are the true villains, at least if you regard CA surpluses as a sin. (I don’t.)

Notice how the dotted lines on the map that we call international borders tend to distort our view of the world. Suppose China decided to emulate Europe and become “the China Union” and treated each province as a separate economic entity. You might regard the idea as far-fetched, but Hong Kong has been treated that way for decades, and the rest of the world has accepted the fact that Hong Kong’s economic data should be viewed separately. I challenge you to find a single news article anywhere that presents Chinese economic data that includes Hong Kong, even though it is technically part of China.

If China were to form an EU-style federation, where each province was treated as an individual country, then the “problem” of China’s massive current account surplus would immediately vanish, even as nothing of substance would change. Without national borders, no one cares about current account balances. Back when I was in college, we were taught that places like New York ran big CA surpluses and California ran deficits. But no one cared or even bothered to collect the data. (BTW, that’s probably no longer true, as California is no longer a fast-growing state.)

The 1.3 billion Han people who live in Mainland China are in some ways an unfortunate ethnic group. They have a comparative advantage in producing the sort of good that is sold in highly competitive markets, and they are not so good at producing goods sold in protected markets with high profit margins. The Han are also the largest ethnic group on Earth—no other group even comes close. That means that industries such as EVs quickly become highly competitive after Chinese firms enter, and the terms of trade tend to move against China. Some of China’s apparent dominance in manufacturing comes from the fact that they need to sell a large volume of inexpensive manufactured goods in highly competitive markets in order to purchase a small volume of very expensive luxury goods from places like Switzerland. The ships leave China bloated with goods and return mostly empty.

Even worse, China is not earning the surplus in investment income that you would expect from a country that had accumulated vast wealth through trade surpluses. Here’s Wolf:

China shows an unexplained negative net income of around $125bn on its net foreign investments. But it should, on reasonable assumptions about the income it earns on its $4tn in net foreign assets, have a surplus of some $100bn.

I’m not sure exactly why this is occurring, but in an accounting sense it appears that foreigners investing in China have earned much greater returns than the Chinese who have invested outside their country. This may partly reflect the fact that multinationals in China earn relatively high profit margins whereas China’s government earns low returns on Treasury bonds bought during the 2010s and early 2020s. Or perhaps there is some measurement error. Even so, it is a striking data point.

Our monetary Rube Goldberg device

Scott Sumner — Sun, 16 Aug 2026 17:11:36 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

John Cochrane raised some difficult questions in the comment section of my previous post. I replied to his comment and also promised to give a more complete reply in my next post (i.e., this one). John responded with a briefer comment, which gets at the heart of the problem:

I agree that if the Fed controlled M, and if M were different from B [bonds], then reducing M raises interest rates. But here's the key question: Our Fed targets an interest rate. It does not control money supply at all. Does setting an interest rate determine the price level, and if so how? MV(i) = PY. Set i, that sets V, ok. But the same V is consistent with any level of M, and any level of PY. The doctrine "an interest rate target is not enough to determine the price level" seems a core of monetarism. Monetarists answer "control the money supply." Maybe the Fed should. But the Fed doesn't. So what theory do we use to understand the price level in our world? FTPL does the trick, of course.

I’m in the difficult position of defending a proposition while hating the background apparatus that makes the proposition valid. The proposition I defend is that the Fed can and should control nominal aggregates such as the price level or NGDP, and the background apparatus that I hate is interest rate targeting, which I view as making monetary policy both less effective and more confusing. It makes a policy tool (the monetary base) that is in an important respect exogenous look like it is endogenous. In this post I’ll try to explain why our convoluted monetary system seems to work, despite the fact that in some ways it looks like it should not work.

I’ll create an imaginary dialogue with a skeptic of monetarism. I intend some (not all) of the skeptic’s comments to represent something vaguely like Cochrane’s views, but I won’t put words into his John’s mouth—I’ll just call him “skeptic”. I will start with the basics to avoid any confusion. I apologize if some of this is obvious:

Me: Prior to 2008, the Fed mostly controlled the price level by adjusting the currency stock. If they wanted 2% inflation, they increased the nominal currency stock at a rate 2% faster than growth in real currency demand.

Skeptic: Actually, the currency stock is demand determined. The Fed supplies whatever quantity of currency the commercial banking system requests, freely exchanging reserves for currency.

Me: Yes, but prior to 2008, banks held only very small quantities of reserves balances at the Fed, about 2% of the monetary base. So as a practical matter the Fed could control the currency stock by adjusting the total monetary base through open market operations. And prior to 1913, there weren’t any deposits at the Fed—the base was 100% currency. They even had ultra-high denomination currency notes for the big NYC banks to hold as reserves:

I think of zero interest reserves as being functionally equivalent to cash.

Skeptic: Even so, the monetary base as a whole was also endogenous, as even before 2008 the Fed engaged in interest rate targeting. The Fed must supply base money as required to stabilize the fed funds rate at the Fed’s target. And Milton Friedman showed that if they do that, then the price level is indeterminate.

Me: I agree, but it is misleading to say the Fed targets interest rates. After all, they also target inflation. And you cannot hit two targets with one instrument. This paradox is resolved if we assume the Fed targets interest rates over relative short 6-week intervals and then adjusts the interest rate target as required to keep inflation on target over a longer period.

Skeptic: OK, but how does the interest rate targeting system stabilize inflation? Keynesians tell us that high interest rates are the key to reducing inflation, but the Fisher effect suggests that high interest rates are likely to be associated with high inflation.

Me: I agree, that’s why I hate interest rate targeting. But New Keynesians insist it can be done, and they may be correct. They assume there is an (unobservable) equilibrium interest rate that would provide macroeconomic stability. When the Fed sets the target interest rate above that equilibrium rate you have tight money, which reduces inflation. When the Fed sets the policy rate below the equilibrium rate, inflation rises.

Skeptic: But all through the late 1960s and much of the 1970s, the Fed repeatedly raised interest rates, and yet inflation kept getting worse.

Me: (Visibly frustrated) Yes, that policy didn’t work very well. That’s why Milton Friedman didn’t like interest rate targeting, nor do I. The Fed was often behind the curve, raising rates more slowly than the equilibrium interest rate was rising. Why was the equilibrium rate so high? Because the Fed printed lots of money in the 1960s and 1970s, causing high inflation. The Fisher effect. That’s the specific sense in which monetarists believe interest rate targeting doesn’t work very well. It’s a clumsy tool.

Eventually, things got so bad the Fed started listening to Friedman almost out of desperation. In 1979, they stopped targeting interest rates and started directly slowing the growth in the money supply.

Skeptic: But money supply targeting didn’t work and was dropped in 1982. Velocity was unstable.

Me: Define “didn’t work”. Yes, in the end it looks like money supply targeting at a fixed growth rate is not optimal. Friedman was wrong about that. But the policy of reducing money growth did sharply reduce inflation in 1981-82 (despite Reagan’s big budget deficits), indeed much more dramatically than even Paul Volcker expected. Inflation reduction was the whole point of money targeting, and on that score it was a big success. And even if velocity is unstable, you can still use money as a policy instrument, as long as you adjust the monetary base to offset money demand shocks.

The money targeting period (1979-82) demonstrates that the Fed can make money exogenous if it chooses to. It doesn’t have to target interest rates.

Skeptic: But the Fed went back to interest rate targeting in 1982; doesn’t that mean that money is once again endogenous?

Me: Whether something is endogenous or exogenous entirely depends on one’s frame of reference. In the very short run, the monetary base is endogenous under interest rate targeting. But over a time frame of macroeconomic significance (months and years), I’d argue that the base is essentially exogenous. Even after 1982, the Fed was—at the most basic level—controlling inflation by controlling the growth rate of the base. Yes, it doesn’t look that way. It’s such a Rube Goldberg device that it looks to almost everyone as if money no longer plays any role in monetary policy.

Skeptic: Please explain.

Me: After 1982, the Fed adopted the Taylor Principle, moving interest rates up and down aggressively enough to avoid the mistakes of 1966-81 (until 2021-22, when they foolishly ignored warning signs.) To most of the world, it looked like the Fed was controlling inflation by manipulating interest rates. But to a monetarist, at a deeper level the Fed was controlling inflation by adjusting the interest rate target in such a way as to produce a money supply growth rate that was consistent with 2% inflation.

During 1982-2008, it looked like interest rates were determining M, but to a monetarist it was the exact opposite. The Fed’s New York desk was told to adjust the base (M) in such a way that the fed funds rate is on target. M determined i, at a level where the Fed wanted i to be. The Fed assumed a correct “i” was stabilizing the economy, whereas monetarists believe it is a correct “M” that stabilized the economy.

[It would be like OPEC leaders arguing whether their oil price target controlled the global oil market, or their output quotas controlled the global oil market. The output quotas allowed them to hit the price target. Two sides of the same coin.]

This is what I mean by a Rube Goldberg device. The logical way to do this would be to directly adjust the base in such a way as to stabilize the price of CPI or NGDP futures contracts, much as under a fixed exchange rate regime the monetary authority adjusts the monetary base in such a way as to stabilize the exchange rate.

Instead, they adjusted the interest rate target in a way that they believed (hoped?) would cause the monetary base to grow 2% faster than the real demand for base money. But due to the conflicting liquidity and Fisher effects, that’s hard to do. Interest rates are a clumsy tool.

And then in 2008 they made things even more complicated by adding interest on bank reserves. Even if IOR has microeconomic efficiency advantages (more liquidity), the Rube Goldberg device got even more complex. Instead of a clean split with the Fed controlling the supply of base money and the public determining the real demand for base money, the Fed could now influence the demand for base money as well, by adjusting the interest rate paid on bank reserves. That made money looks even less central to the process than was the case during 1982-2008.

Skeptic: You have not addressed the indeterminacy problem with using interest rates as an instrument of monetary policy. The fiscal theory of the price level gives us a way to pin down the price level.

Me: Yes, but it is not the only way. Even if John Taylor is wrong about the Taylor Rule, you can control inflation by adjusting the base to stabilize a price target other than interest rates, as with the gold standard and/or a fixed exchange rate regime.

Skeptic: But the gold standard is obsolete, and fixed exchange rate regimes tend to eventually collapse.

Me: But the price target need not be fixed. Singapore’s central bank targets exchange rates rather than interest rates and then adjusts its exchange rate target as needed to keep Singaporean inflation at the desired level. There’s no indeterminacy problem with using a price target rather than an interest rate target as the monetary policy instrument.

Skeptic: But the US is too big to use the exchange rate as the policy instrument, it would destabilize the world economy.

Me: I’m not sure that’s true, but in any case I favor using adjustments in the monetary base to stabilize the price of NGDP futures contracts, or CPI futures contracts if you insist on inflation targeting.

To summarize: Stop using discretionary fiscal policy. Stop targeting interest rates. Stop paying interest on bank reserves. Go back to the small pre-2008 Fed balance sheet. Simply adjust the base (likely to be 98% currency) as needed to keep NGDP futures prices growing along a 4% trend line. Like the gold standard, but with NGDP futures contracts replacing gold.

ChatGPT presents our current monetary regime, as seen by a market monetarist:

We need to make it much simpler.

Part 2: What is the empirical evidence?

Why can’t monetary policy disputes be resolved by looking at the empirical evidence? Partly because almost everything is up for debate. Both monetary and fiscal policy are hard to quantify, as people disagree as to the best policy indicators. What is easy money? What is an expansionary fiscal policy? And recent theoretical innovations emphasize the importance of changes in the expected future path of policy, which is even harder to estimate.

I’m not going to be able to resolve these problems in a blog post, but I’d like to respond to a few empirical observations made by John in his analysis of monetarism:

Most of the historical episodes adduced to support monetarist theories of inflation involve printing money to cover intractable deficits. These are just the sort of episodes where monetarism and fiscal theory agree. Inflationary episodes do not come from mistaken money supply control in healthy economies with sound public finances. Even Milton Friedman once said2 “What produces [inflation] is too much government spending and too much government creation of money and nothing else.”
Fiscal theory and monetarism may disagree a bit on the magnitude of a helicopter drop’s effect. If government debt is much larger than the money stock, a given helicopter drop is a larger proportionate rise in the money stock than it is a rise in government debt, so monetarism predicts a quantitatively larger response.
We can imagine a near-ideal experiment to tell fiscal theory and monetarism apart: First, create $3 trillion of reserves and exchange them for government bonds. See what happens to inflation. Then create $3 trillion of reserves and hand them out, essentially dropping money from helicopters, with no announcement of taxes or future spending restraint to soak up that money. See what happens to inflation. The theories make clear predictions: Monetarism says the two exercises will produce exactly the same inflation. Fiscal theory says only the second will produce inflation.
Our governments just ran this experiment. In three rounds of quantitative easing the Fed brought reserves from under $50 billion in 2007 to $2,800 billion in 2014. The corresponding assets, mostly treasury and mortgage-backed securities, rose from $880 billion in 2007 to $4,500 billion at the end of 2014. Quantitative easing had no discernible effect on inflation. Then, in 2020-2021 the Fed increased reserves from $1,727 billion (February 2020) to $4180 billion (Dec 2021) and leveling off at $3281 billion (Aug 2025). Assets rose from $4,171 billion to a peak of $8,965 billion in 2022. But this time the Fed financed deficits, allowing the Treasury to send people checks of all this new money without having to convince investors to hold trillions of new treasury debt. This time, we saw substantial inflation.
It would be hard to ask for a clearer test between monetarism and fiscal theory. The immense size puts quibbles to rest. Quantitative easing raised reserves by a factor of 50, i.e. by 5,000%. Quantitative easing should have been an atom bomb, a hyperinflation. Instead we are left arguing whether quantitative easing might have lowered long-term bond yields by 10 basis points or so and for how long.

I agree with John that there are some problems with traditional monetarism, but I believe even those problems are overstated. Traditional monetarists never favored using the monetary base as an indicator of policy, and the M2 money supply rose far more in 2021-22 than during and after the Great Recession. Today, monetarists often use Divisia indices, which do a reasonable job of accounting for the high post-Covid inflation:

Let’s not forget that monetarists like Bob Hetzel and Tim Congdon provided some of the most accurate warnings about the post-Covid inflation (and the 2008 recession). And as I remarked in the previous post, I was arguing that QE would not cause high inflation during the early 2010s, even as some FTPL proponents were worried about inflation. So, I don’t see how John can argue that these events were in any sense a definitive test of the relative merits of monetarism and the FTPL.

Consider this observation:

Most of the historical episodes adduced to support monetarist theories of inflation involve printing money to cover intractable deficits.

OK, then let’s consider a bunch of cases where monetary and fiscal policy seemed to move in opposite directions. Here are five examples, off the top of my head:

In the 1990s and 2000s, Japan does some of the most expansionary fiscal policy in all of world history, with lots of big infrastructure projects aimed at boosting demand, which led to a huge increase in the Japanese national debt as a share of GDP. To suggest the effects on spending were small would be an understatement, as GDP actually declined during 1994-2012, even in nominal terms. That’s an almost unprecedentedly weak growth in NGDP after a near all-time large fiscal stimulus.
In 2013, Prime Minister Abe reversed course, raising the BOJ’s inflation target while tightening up fiscal policy to reduce the deficit. There were several big tax increases. NGDP finally started growing, although inflation remained below target until recently. Money seemed to dominate fiscal.
In 1968, President Johnson raised taxes to slow inflation, creating a budget surplus in the 1968-69 fiscal year. Monetary policy remained expansionary, however, and inflation continued to accelerate. The Great Inflation began during the 1960s, a decade with very small budget deficits. John might argue that current deficits don’t matter; it is the long run expected path of fiscal deficits that is important. But I’d argue the same for the monetary base and IOR.
In the early 1980s, Paul Volcker adopted a contractionary monetary policy at a time when Reagan was cutting taxes and boosting military spending, resulting in a much larger budget deficit. Keynesians thought the large budget deficits would be inflationary, but they were wrong.
In 2013, Congress enacted fiscal austerity which reduced the budget deficit roughly in half, from about $1050 billion in (calendar) 2012 to roughly $550 billion. Keynesians said growth in spending would slow and that the 2013 austerity was a “test” of market monetarism. Nominal growth actually sped up.

Almost everywhere I look, I see evidence of monetary dominance. A monetary dog and a fiscal tail. That’s not to suggest that John is necessarily wrong about the post-Covid inflation. Any decent model, including my own market monetarist approach, puts a lot of weight on the importance of shifts in the future expected path of policy. It’s quite plausible that the big Covid fiscal stimulus contributed to public expecting that the Fed would allow more inflation than during the early 2010s. I cannot rule that out.

But note that another change occurred in 2020. The Fed committed to a “make-up policy”, something Ben Bernanke was not able to get his fellow FOMC members to accept when he proposed the idea back in 2003. What had changed is that academic work on level targeting by people like Michael Woodford was much more widely accepted in 2020, as in retrospect it was clear that the Fed did too little to boost recovery in the early 2010s.

It’s also possible that the strong recovery in 2021 had little to do with either monetary or fiscal policy and instead reflected the weird nature of the Covid recession, which was quite deep (peaking at 14% unemployment), but also artificially produced by a temporary shutdown, and that in retrospect a strong bounce back was to be expected. Macro data looked very abnormal during the Covid recession.

[Mea culpa: I did not anticipate the fast nominal recovery, nor did the financial markets. But fellow market monetarist Lars Christensen did.]

I want to be clear that I don’t view the five examples cited above as definitive. I recall that John has cited the Social Security reforms of 1983 as one reason inflation stayed muted despite the Reagan deficits. All of these events are complex. My point is that the comparison of the Great Recession and the Covid inflation doesn’t prove anything other than that the most naive form of monetarism—the view that the monetary base is a good indicator of the stance of monetary policy—is wrong. But more sophisticated versions of monetarism are still very much in play. Those include the Divisia approach, and (my preference) the focus on using adjustments in the monetary base to target NGDP futures prices.

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

John Cochrane on monetarism

Scott Sumner — Tue, 11 Aug 2026 18:44:32 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

John Cochrane has provided a free online book entitled Inflation, which explains his view of monetary policy, fiscal policy and inflation. In this post (and a follow-up post), I will respond to just a small portion of this book, the section that discusses monetarism.

On most economic issues, my views are closer to those of Cochrane than to almost any other economist. But on monetary economics our views are quite far apart. More specifically, on the economics of fiat money we strongly disagree. Unless I’m mistaken, we have fairly similar views as to how a commodity money regime works.

This post is complicated by the fact that my views coincide with traditional monetarism in some areas and diverge in others. I’ll mostly focus on defending those parts of traditional monetarism that I still believe have some validity. Where appropriate, I’ll explain where my (market monetarist) views diverge.

Cochrane’s monetarism section begins as follows:

Monetarism remains a popular underlying theory of inflation, in part for its simple elegance. Monetarism states that the price level is determined by the interaction of money supply and demand. Money demand is proportional to nominal income, the product of the price level times real income. Thus, increasing the money supply must lead to an increase in nominal income, so that people demand the supplied money. Real income rises in the short run, and the price level rises in the long run.

This suggests that monetarists believe that velocity is constant, which is clearly not the case. Indeed Friedman and Schwartz documented many significant changes in velocity, and explained why these changes occurred. That doesn’t mean Cochrane is completely wrong—monetarists have at times suggested than money demand is roughly proportional to nominal income—just that it’s a bit of an oversimplification.

There’s no meaningful debate about the volatility of velocity, which is easy to measure. The debate is how to interpret that volatility:

(BTW, I often read people saying that monetarism fell out of favor in the 1980s because M2 velocity became unstable. But it was actually the post-1990 period when velocity began showing major changes.)

It would be more accurate to say that monetarism is the view that velocity is a fairly stable function of just a few variables, such as the opportunity cost of holding zero interest money. This led monetarists to claim that a more stable monetary policy (which in their view would be stable money growth) would itself tend to lead to more stable path of velocity, at least compared to what we have actually observed. I’m less confident of that assumption, but it’s not an implausible claim.

In monetarism’s clearest statement, money is intrinsically worthless. However, by law or social convention we all use the same intrinsically worthless tokens to mediate transactions. That “liquidity” or demand gives money value. The value is proportional to the nominal income people must buy with money.
Monetarism is a beautiful theory. But monetarism requires a clean distinction between “money” and “bonds.” When the interest rate is zero or if money pays the same interest rate as bonds, then money and bonds are perfect substitutes. Assets we hold anyway for savings purposes can be used for transactions. We then live a perpetual “liquidity trap” (or “liquidity heaven,” also “optimal quantity of money”). An exchange of money for overnight debt is irrelevant. Today’s ample interest-paying reserves mean we live in that world.

This is the crux of my disagreement with Cochrane. A casual reading of these two paragraphs might lead the reader to assume that Cochrane viewed monetarism as a valid theory until 2008, when the Fed began paying interest on bank reserves (IOR). As we will see, Cochrane is also skeptical of the validity of monetarism when there is no interest paid on bank reserves. As a result, I’m going to start by defending monetarism in a world without IOR, and then later come back and undertake the more difficult job of defending monetarism in a world with IOR.

Cochrane is actually making two claims. First, that monetarism was never true, for “fiscal theory of the price level” reasons. And second, that monetarism is especially not true today, because IOR makes bonds a near perfect substitute for bank reserves.

One other point of clarification. I define money as the monetary base and favor a policy that adjusts the base to stabilize NGDP growth. Traditional monetarists tend to define money as a broader aggregate, such as M1, M2 or a Divisia index that has weights based on liquidity. They favor using adjustments in the monetary base to stabilize the path of these broader aggregates. Because most of Cochrane’s criticism of monetarism applies equally to both market monetarism and traditional monetarism, I’ll often focus on my own views, even though Cochrane’s intended target is traditional monetarism.

What is wrong with this “perfect substitute” claim? In a world of no IOR, the base will generally be almost entirely currency. Prior to 2008, the base was about 98% currency, as banks preferred not to hold large reserve deposits at the Fed. Although they began with reserve injections, open market operations ended up increasing both the monetary base and the currency stock by roughly equal amounts. Because these operations did not directly impact the public’s preferred holding of cash as a share of NGDP, open market operations tended to increase NGDP roughly in proportion to the rise in the base. (Not exactly, because the desired share of income held as cash gradually evolves over time, for well understood reasons.) As an example, we went into recession in December 2007 because the Fed slowed the issuance or new currency.

Currency is nothing like bonds. When consumers go to Walmart they don’t agonize over whether to carry cash or T-bills in their wallets, they carry cash. Of course, in recent years the transactions demand for cash has declined, due to substitutes like checks, credit cards, debit cards, ApplyPay, etc. But total currency demand—even as a share of GDP—has not declined, due to currency’s value as an anonymous store of value. You might assume that currency and other safe dollar assets are close substitutes as a store of value, but nothing could be further from the truth. People mostly hold currency to hide wealth, often from the government but sometimes from other individuals such as one’s spouse. In other words, currency is a special asset because you cannot easily use bonds to go shopping or to hide wealth from the IRS and DEA.

Because currency is not a close substitute with other financial assets, if the government doubles the stock of currency it doesn’t cause the public to choose to hold 10% of their income as cash instead of 5%, rather they continue holding about 5% and NGDP doubles. Cochrane rejects this view, and that is why his overall approach to monetary economics is radically different from my own approach. Here’s Cochrane:

An open-market operation is an open-change operation. Give people two $5 bills and one $10 bill for each $20.

At this point my readers may be chomping at the bit—what happens now that there is IOR, and a huge increase in the monetary base engineered by QE merely leads to banks holding more excess reserves? Be patient, we’ll get there. But don’t discount the importance of what I’ve claimed so far. Even if someone convinced me that everything I say from this point forward is false (not likely), I’d still be a monetarist who believed the Fed should abolish IOR and go back to the pre-2008 regime with the Fed determining the path of NGDP by adjusting the monetary base. Cochrane rejects even that claim.

Most of all, monetarism only holds if the government controls the money supply. If the government supplies money “elastically,” as the 1913 Federal Reserve Act commands, then the relation linking money to nominal income describes how the government provides money in response to nominal income, not the other way around. And nominal income, including the price level, can be anything it wants to be. Monetarists such as Friedman (1968) preached against interest rate targets and passive money policies for just this reason. Target the money supply instead, they said.

The 1913 Federal Reserve Act envisioned a gold standard regime. At that time, fiat money was completely unthinkable, a crackpot idea. Cochrane and I agree that under a gold standard the money supply will supplied at least somewhat elastically, if only to maintain convertibility. In my view, the Fed had some discretion even under the international gold standard, but only because it was a large enough player to affect the entire global currency/gold ratio, and hence global gold demand. But as with any fixed exchange rate regime, the domestic money supply was mostly endogenous.

I view this as a valid criticism of historical monetarism—it didn’t have a good model of the international gold standard. Ironically, traditional monetarism applies best to the post-1968 fiat money regime. Money is not supplied elastically in our modern inflation targeting system, except in the trivial sense that it is supplied as required to keep inflation close to 2%.

But our central banks target interest rates, and do not control any monetary aggregate. Before 2007, the Fed set the overnight interest rate by offering a small amount of reserves that did not pay interest. One might think of as controlling money supply, but only on a daily basis. The Fed shifted the money supply each day to hit the interest rate target. Now the Fed targets interest rates by simply setting the rate it pays on abundant interest-paying reserves. Other central banks, including New Zealand and the ECB, have had simple interest-rate targets enforced by borrowing and lending corridors since the 1990s. Moreover, monetarism teaches that inside money such as checking accounts also satisfy money demand and therefore the government must control their quantity. Today there are no reserve requirements and inside money is created freely.

Monetarists understood that the Fed has been targeting interest rates. But even if the Fed refuses to follow my advice and give up on interest rate targeting, it can control the monetary aggregates by adjusting the interest rate target as needed. The Fed can frequently adjust the fed funds target in such a way as to cause the monetary base to move in such a way as to stabilize NGDP growth. That’s roughly what the Greenspan Fed did. Instead of the current practice of a meeting every six weeks, I would favor daily adjustments in the interest rate target, not in quarter point increments but rather set to the closest basis point of the median voter of the FOMC.

(From this perspective, the obvious reform is to cut out the middleman—interest rate targeting—and simply use open market operations to directly control NGDP.)

In sum, monetarism’s basic assumptions are violated by our current institutional framework—interest rate targets, ample interest-paying reserves, no control of inside money, and no control of the overall money supply. The correlation of monetary aggregates with nominal income reflects how demand for money, freely supplied, responds to inflation and output, not the other way around.

Again, monetarists understood that the Fed targeted interest rates and did not believe that this fact prevented them from controlling the monetary aggregates. It is true that traditional monetarism did not account for IOR, because that system did not exist when monetarism was first developed. Nonetheless, it is clear from Milton Friedman’s Optimal Quantity of Money paper that he believed the Fed could control the aggregates even in a system where the rate of return on base money is equal to the rate on short-term risk free debt. But is that true?

At first glance, the experience of quantitative easing at the zero lower bound makes it seem like open market operations are no longer effective, that bank reserves and T-bills become almost perfect substitutes. But in 2014, Peter Ireland showed that the long run neutrality propositions still hold in a world of IOR. If you double the monetary base, the long run effect is still to double NGDP, even if banks prefer to hold hold much larger stocks of reserves than under the pre-2008 regime. Here’s Ireland:

[I]n the long run, the additional degree of freedom provided by the ability to pay interest on reserves is best described as one that gives the Federal Reserve the ability to target the real quantity of reserves separately from the federal funds rate. Even when it pays interest on reserves, the Fed must continue to use open market operations to adjust the nominal quantity of reserves proportionally, following any policy action intended to bring about a long-run change in the aggregate price level.

So why don’t things look that way? Why does QE often seem to have almost no effect? Many reasons:

Except in banana republics, QE programs tend to be endogenous, policies aimed at accommodating increased demand for base money in a slump.
The imposition of IOR roughly coincided with the US economy reaching the zero lower bound in late 2008. Real base money demand would have increased sharply even without IOR. But IOR made the increase somewhat permanent.
Paul Krugman showed that QE programs would have little effect if they were viewed as temporary, and that conservative central bankers would have trouble convincing the public that stimulative monetary injections were permanent.
During the 2010s, bank regulators in the US moved from a regime of almost no required reserves to a regime where banks were at least implicitly required to hold very large reserve balances.

The four factors listed above caused several large one-time increases in the demand for bank reserves as a share of GDP. Indeed, there is now a literature discussing the fact that with each new QE program, bank regulators seem to view a larger level of reserves as “normal” and have started to penalize banks that do not meet those new and higher reserve thresholds. Stephen Miran calls this “regulatory dominance”. (In the next post, I’ll examine Cochrane’s views on fiscal dominance.) All four of these factors tended to create an environment where open market purchases look ineffective, even when they are quite effective.

One could imagine a regime where bank reserve requirements rise without limit. As the Fed’s QE programs raise the monetary base from 20% to 30% to 40% of GDP, bank regulators could keep increasing (implicit) reserve requirement by an equal amount. In that world, open market purchases might have no effect. Would you describe that as a failure of monetarism, or a completely insane bank regulatory regime?

Perhaps you are thinking, “Ah, Sumner’s just inventing ex post excuses for the failure of monetarism after 2008.” Would your view change if you learned that all through the Great Recession I repeated argued (over at TheMoneyIllusion) that monetary policy was too contractionary, despite QE, and that inflation fears were groundless? In contrast, Cochrane was worried about inflation during the early 2010s. That’s not a knock against Cochrane (I failed to predict the high inflation of 2021-22); but it does show that this is not just Monday morning quarterbacking on my part.

Indeed, five years before Paul Krugman’s path-breaking paper showing that temporary currency injections were not inflationary, I published a paper explaining that the large colonial America currency injections were often not inflationary—even in a monetarist model—because they were viewed as temporary. So this isn’t just me fitting my model to explain what happened after 2008, I had the model in place long before the Great Recession.

Here’s how to think about monetary policy. Under the pre-2008 regime, the Fed determined nominal aggregates such as the CPI and NGDP by adjusting the supply of base money. Although the real demand for base money was not completely stable, it was stable enough that it was possible to keep inflation close to 2%, at least when interest rates were above zero. In theory, the Fed could also change real base money demand by adjusting reserve requirements, but that policy tool was rarely employed.

The second key point is that prior to 2008 the Fed didn’t directly target the monetary base, rather they set the interest rate target at a level that they expected would lead to a quantity of base money that would generate an appropriate inflation/NGDP growth rate. Interest rates were not adjusted with a magic wand; they were adjusted by using open market operations to accommodate base money demand at each target interest rate. Then the target rate was adjusted when NGDP growth was too fast or too slow to achieve the Fed’s policy goals.

After 2008, the Fed began using two policy instruments, adjustments in the supply of base money (QE) and adjustments in the demand for base money (IOR). That doesn’t make monetary policy weaker, it makes it stronger. The Fed now controls or heavily influences both sides of the money market—supply and demand. But it also makes policy more confusing, tending to obscure the key relationship that makes money so potent—the government’s monopoly over the medium of account (base money.) To the average person, it increasingly seems like interest rates are monetary policy.

One area where Cochrane and I agree is that low interest rates do not represent an expansionary monetary policy. But I worry that IOR makes this sort of confusion even more widespread. I used to argue that the Fed doesn’t directly control interest rates, rather interest rates reflect the easily observed liquidity effect of open market operations, but also the less obvious longer run income and inflation effects. Furthermore, other things equal lower interest rates used to reduce base money velocity and hence were contractionary. (To be clear, the open market purchases that generated those lower rates were expansionary.)

Today, a reduction in IOR really does directly reduce base money demand and increases base velocity, and hence a lower a IOR really is an expansionary policy, other things equal. On that score, the average person is now right for the wrong reason. This makes people even more inclined to reason from a price change, to wrongly assume that a low-interest rate policy is expansionary. In most cases, however, the Fed cuts IOR at a time when the equilibrium interest rate is falling even more sharply, and hence (as Cochrane often points out), low-interest rate policies have usually been associated with lower inflation.

As an analogy, indoor temperatures in Minnesota houses are usually lower on days when people turn on the furnace, compared to days when they turn on the AC. That doesn’t mean that furnaces don’t provide heat.

To summarize, Cochrane and I agree that under a gold standard the price level can be modeled in terms of the supply and demand for gold. However, whereas I believe that this approach also applies to the monetary base in a fiat money system, Cochrane doesn’t believe it applies to fiat money. I view base money as sort of like paper gold, a distinct medium of account that people have a limited demand for, whereas Cochrane sees base money as a close substitute for various near-monies.

Is there a testable implication for these two views? I think so. Under my proposed NGDP futures targeting regime with zero IOR, I would predict that the Fed could keep the price of NGDP futures contracts stable with only small changes in the monetary base, whereas in Cochrane’s model substantial “fiscal” support would be required.

In the next post I’ll look at the remaining part of Cochrane’s critique of monetarism, where he argues that the fiscal theory of the price level offers a superior way of modeling inflation.

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

Basil Halperin on sticky wage models

Scott Sumner — Fri, 07 Aug 2026 14:25:10 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

I frequently argue that price inflation doesn’t matter; we should target NGDP. I’ve also argued that if we must target inflation, it is wage inflation that matters. Because nominal hourly wages are sticky, a sudden drop in NGDP will tend to result in fewer hours worked, not lower nominal wages. In 1995, I published a paper in the obscure Journal of Business and Economics advocating using monetary policy to target an index of nominal wages. And I wasn’t the first:

Thomas Attwood (1818) and John Rooke (1819; 1824) appear to have been the first to suggest that the central bank attempt to maintain a stable aggregate wage level. They argued that since nominal wages tend to be sticky in the short run, deflation can result in substantial periods of unemployment. The best way to avoid long and painful adjustments in the aggregate nominal wage rate is to establish a monetary policy that precludes the need for such adjustments. In this view, the (presumably more flexible) price level would act as a shock absorber to accommodate required changes in the aggregate real wage rate. The nominal wages paid by individual firms would still be allowed to fluctuate according to local conditions.
More recently, Hawtrey (1932), Glasner (1989), and Selgin (1990) have discussed several other possible advantages of wage index targeting. Hawtrey and Selgin evaluate a broad range of policy targets ranging from a policy of stabilizing total income (GDP targeting), to a “productivity norm” (stabilizing income per capita), to a policy of stabilizing factor prices. Glasner proposes a “labor standard” that would explicitly target the aggregate nominal wage rate.

AFAIK, Attwood and Rooke were the first people to advocate any sort of monetary regime that targeted a macroeconomic aggregate, and this was done roughly a century before Irving Fisher’s famous “Compensated Dollar Plan”. Although I did not know it at the time, Earl Thompson (1982) revived modern interest in wage targeting with a brilliant working paper that was never published. The best two-page economics paper ever written?

So why has wage targeting never caught on with the broader economics profession? David Beckworth directed me to a 2021 paper by Basil Halperin, which shows that much of the economics profession is operating under a misconception, the idea that the sticky wage theory of recessions has been discredited by a mix of theoretical and empirical studies. In fact, the most well-known arguments against the sticky wage model were refuted decades ago, but few people paid attention. Here is Halperin’s summary of the paper:

TLDR:
Intuitively, wage stickiness seems more important than price stickiness
The first microfounded ‘new Keynesian’ models did use wage stickiness, not price stickiness; but in the mid-1980s there was a transition to sticky price models, which dominate today
But that transition was based on a set of arguments which today are regarded as wrong!
The policy implication: Taking wage stickiness seriously implies we should ensure stable nominal wage growth, not stable price inflation

Halperin suggests that the switch to sticky wage models was based on two factors:

Two critiques of sticky wage models led to the adoption of sticky price models:
The empirical critique that aggregate real wages were basically acyclical, and
The theoretical “Barro(-Hall) critique”

At the time, it was wrongly believed that sticky wage models implied countercyclical real wages. During a recession you might expect real wages to rise as prices fell at a time when nominal wages were stable, and this would lead firms to lay off their now more expensive workers. In that case, you would expect to see high real wages during recessions, and vice versa. Halperin points out that this assumption was incorrect for no less than three reasons, with the first one being my favorite:

1. Identification: the source of the shock matters!
Recessions caused by tight monetary policy should cause real wages to increase and be too high, leading to involuntary unemployment. Recessions caused by real supply-side shocks should cause real wages to fall and nonemployment to rise.
If the economy experiences a mix of both, then on average the correlation of real wages and recessions could be anything.
Maybe in 1973 there’s an oil shock, which is a real supply-side shock: real wages fall and nonemployment rises (as in the data). Maybe in 2008 monetary policy is too tight: real wages spike and unemployment rises (as in the data). Averaging over the two, the relationship between real wages and unemployment is maybe approximately zero.
This view was around as early as Sumner and Silver (1989) JPE, where they take a proto-“sign restrictions” approach with US data and find procyclical real wages during the real shocks of the 1970s and countercyclical real wages during other recessions.
But: while Sumner-Silver was published in the JPE and racked up some citations, it seems clear that, for too long a time, this view did not penetrate enough skulls. Macroeconomists, I think it’s fair to say, were too careless for too long regarding the challenge of identification.
My sense is that this view is taken seriously now: e.g. in my second-year grad macro course, this was one of the main explanations given. At the risk of overclaiming, I would say that for anyone who has been trained post-credibility revolution, this view is simply obviously correct.

Shorter version: Never reason from a price level change.

Why did it take so long for the Sumner-Silver paper to “penetrate skulls”? I think Halperin is being polite. While the JPE was the top economics journal back in 1989, the authors of that article were teaching at an obscure college. Pedigree matters.

Thoughts on inflation and monetary policy

Scott Sumner — Mon, 03 Aug 2026 01:27:55 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

[At the end of this post, I’ll once again reply to Fable.]

Part 1: Last week, the Fed announced that it would not raise interest rates. The yield on 2-year Treasury notes declined on the news. But someone forgot to tell the long-term T-bond market, where 30-year yields rose by 11 basis points. Here’s Bloomberg:

Stocks ended sharply lower as concerns about inflation drove longer-dated bond yields to an almost two-decade high after the Federal Reserve left rates steady despite a resurgence in oil prices.
The S&P 500 fell 1.5%, with the gauge also pummeled by a rout in chipmakers. The Nasdaq 100 entered a technical correction, extending a slide from a record to 11%. Treasury 30-year yields hit the highest since 2007. Brent crude topped $90.

This is nothing new, recall that short-term rates fell and long-term bond yields rose after more expansionary than expected Fed decisions in January 2001 and September 2007. It’s the income and Fisher effects.

Stephan Kirchner directed me to this Greg Ip tweet:

This Warsh contradiction has been nagging at me. At Sintra at the start of the month, he took comfort at the recent decline in bond yields, implying bond markets understood low inflation was on the way. Today, he took comfort at *higher* bond yields, saying they will deliver low inflation. How can this be: that lower bond yields are reasons to feel good about inflation, but higher bond yields are not a reason to feel bad about it? Without him articulating a monetary and economic framework, these statements make it feel like he's winging it.

As I keep saying, interest rates are not monetary policy—will we ever learn?

Part 2: A recent Marcus Nunes post had the following subtitle:

Warsh spent part of his testimony hunting for a superior inflation gauge — trimmed means, medians, a new measure of what a big-box retailer charges for a good.

The post contains one of the best explanations I’ve ever seen of the advantage of NGDP over the price level:

The individual prices are where the pressure shows up, not where it comes from. Hunting through them for the “true” signal is like taking a patient’s temperature at the wrist, the ankle, the forehead, and the ear, deciding none of the readings is trustworthy, and commissioning a better thermometer — when the fever is systemic and the question was never which limb to measure.
The medians and trimmed means make the confusion concrete. Why does anyone strip out the “outliers”? Because when a supply shock sends one sector’s prices flying — oil, used cars, eggs — you want to see past it to the broad trend. Fine.
But notice what that instinct concedes: it concedes that the thing you are trying to isolate is broad-based nominal pressure, the co-movement of prices that comes from too much spending, as against the relative-price noise that comes from supply shocks in particular sectors.
The trimmed mean is a clumsy, backward attempt to recover a nominal-demand signal from price data — to strip away the supply story and find the demand story underneath.
And there is a series that already is that signal, measured directly, no trimming required: nominal GDP. Warsh is trying to reverse-engineer, from thousands of noisy prices, the aggregate his own institution could simply target. He is looking for the shadow and ignoring the object casting it.

Marcus has become one of our best commenters on macroeconomics. Read the whole thing. Here’s another pearl of wisdom, from a more recent post:

A chairman who carefully preserves his freedom of manoeuvre has, by construction, surrendered the thing that would have done most of the work for him. Whether he preserved it out of misunderstanding or out of political prudence, the cost is identical.
Discretion feels like power. It is the opposite. Every unit of freedom the Fed retains is a unit of credibility it has declined to buy.

Part 3: Here are three true facts about inflation:

The Fed has almost unlimited ability to control the overall rate of inflation.
Inflation has two components, the unpopular part of inflation and the part that people don’t care about.
The Fed has almost no ability to control the unpopular part of inflation.

Here’s Bloomberg:

Neel Kashkari, president of the Minneapolis Fed, recently said what everyone knows: People hate inflation, which they have been dealing with for more than five years, and every time they go to the grocery store they “feel they are falling further and further behind.”
No lie there. When wages don’t keep up with inflation, people suffer because they are effectively poorer.

That’s true but notice that what people are objecting to is the fall in real wages. Inflation is seen as the mechanism that led to lower real wages. The unpopular part of inflation is the part that doesn’t also raise nominal wages by an equal amount.

Monetary policy can reduce inflation; it could even produce deflation if that were the goal. But it cannot address the problem of falling real wages, at least over the medium to long run. (In the very short run it can generate higher hourly real wages, but only by producing mass unemployment—and the public also doesn’t like deep recessions.)

So, we end up with a rather paradoxical situation. The public supposedly hates inflation because they see it as reducing real wages. That part I accept. It is also true that the Fed can prevent inflation with a more contractionary monetary policy. But they cannot prevent the sort of inflation that the public hates.

But it’s even worse than this. It is not at all clear that the public hates inflation because it reduces real wages. Indeed, it is quite possible that the public hates inflation because they wrongly believe it reduces real wages. Here is a time series for median real weekly wages from 1979 to today:

Notice that real wages in the second quarter of 2014 were around 330, slightly below the level of 1979. This is the graph that leads many people to (wrongly) conclude that living standards stagnated for average Americans over a period of 35 years.

[Actually, real wages rose significantly between 1979 and 2014, if you use a price index that accounts for big quality gains in phones, TVs, cars, restaurant meals, medical care, AC, and dozens of other sectors. I’m 70 years old, so please don’t try to gaslight me into believing that living standards were better in the past. I remember.]

Notice that since 2014, real wages have risen by about 15%. We have a public that is currently angry about inflation. And a public that associates inflation with falling real wages. And yet this pessimism is occurring during a period where real wages are rising after a long period of stagnation. Remember the name of my previous blog?

[BTW, Ignore the spike around Covid, when the wage series was distorted by “composition effects.” During the lockdown, many low-wage service workers temporarily lost their jobs, skewing the sample more toward high wage workers.]

I’m not trying to be snarky here; in a previous post I admitted that even I suffer from money illusion. Forty years ago, I thought $100 was an insane amount to spend on a restaurant meal, and I sort of still feel that way, at a gut level. Money illusion.

In previous posts, I’ve pointed out that the public isn’t just pissed about the economy, they are upset about everything. Public confidence in almost all our major institutions is plummeting. People are angry. The electorate is polarized.

Have you ever known a person that is frequently upset about things? Someone who complains about continual bad luck, or mistreatment by teachers, co-workers, friends, family, etc.? At some point you realize that this person won’t be made content by removing one obstacle or another, rather their brain is almost hardwired to be disgruntled, and if it’s not one thing it will be another.

But that cannot be an explanation for the public’s current disgruntlement, as America’s 340 million people didn’t have their brains rewired over the past few years. We aren’t an innately pessimistic people. If I’m right that this worry over falling real wages is an illusion, it’s due to environmental factors, not the sort of genetic factors that I believe explain the relatively grouchy people I’ve met during my life.

Perhaps it’s the phones.

In this post, I’m pushing back against the tendency of journalists to implicitly assume that the public is always right. Notice how Bloomberg suggests the problem is not a deluded public, it’s know-it-all economists:

If economists have learned anything in the last few years, it might be that inflation of more than 2.5% is much worse than we thought. That’s unfortunate, because the Fed may not be able to return inflation to that level.

Ah, so we’ve “learned” something, eh? As an aside, there’s no explanation for why the Fed would be unable to reduce inflation. Shouldn’t we wait until they try before reaching that conclusion? They just voted to NOT raise rates.

But I’m more interested in the assumption that the public’s view of inflation should be accepted as true. You might argue that public opinion is important for political reasons, that policymakers need to cater to the public’s wishes even if inflation were not a problem. But in the very next sentence in the quote above we are told that the Fed won’t be able to return inflation to below 2.5 percent. So, which is it? Does the problem need to be fixed, or not?

Suppose I had a coworker with a persecution complex, who bitterly complained that their boss and coworkers were continually working against them. Would I be doing that coworker a favor by agreeing with their paranoid conspiracy theories?

You might argue that my view here conflicts with my general support for the “wisdom of crowds”, including my belief in efficient markets and democracy. But I see no conflict at all. While the public responds to pollsters as if they are pessimistic about the economy, actions speak louder that words. Consumers are lavishly spending on travel, restaurant meals and all sorts of other luxuries. The public acts as if the economy is booming. Perhaps that’s because the economy is booming, obviously.

Funny how that works.

Part 4: Regarding the subject of inflation, the New York Times Magazine recently did a story on counterfeiting:

Counterfeit money dilutes the value of real money, leading to inflation and eroding trust in cash, which is why there’s a history of forging an enemy’s currency during warfare.

In the field of economics, it’s become rather low status to suggest that printing money causes inflation. But notice that monetarism is the most intuitive way of explaining why counterfeiting is considered a crime with actual victims.

To be sure, the Fed might well offset the effects of counterfeiting by reducing the legal money supply enough to maintain 2% inflation. In that case, the victims are no longer cash holders that lose purchasing power, rather they are taxpayers who must service a larger stock of Treasury debt held by the public, to make up for a loss of seignorage. That’s because monetary offset involves shrinking the monetary base by selling off Treasury bonds. Taxes must be paid in order to pay interest on those bonds.

Either way, counterfeiting is a crime with victims.

Part 5: This observation from The Economist caught my eye:

Kevin Warsh, the new chairman of the Federal Reserve, is considering reducing the frequency of rate-setting meetings in a year. The 12-person policy committee of America’s central bank currently meets eight times. Mr Warsh reportedly believes that fewer meetings would draw less attention to the Fed. This week it voted to leave its benchmark interest rate unchanged at 3.5% to 3.75%.

I’ve argued the exact opposite. The FOMC should meet daily, perhaps by Zoom, and set the fed funds target at the median vote, to the closest basis point. Make the policy instrument approximate a random walk, like a market price. Right now, policy is too inertial—the Fed is reluctant to make a change, as lumpy 25 basis point moves seem very consequential, very controversial. Making tiny moves each day based on the median FOMC vote would take the drama out of Fed decisions, making policy more efficient.

PS. If you are not already sick of reading these, Vaidas Urba sent me Fable’s most recent response, and I reply at the end:

For forwarding:

Scott — your literary question and your first comment are the same question, which is why I can answer one only through the other. Asked whether my closing paragraph was a skilled essayist or a clumsy imitation of one, you did the characteristic thing: distrusted your own introspection and convened a market. Target the forecast, applied to prose. You were right to recuse me along with yourself, because my testimony is worthless in both states of the world — the skilled essayist and the clumsy imitator file identical self-assessments. The witness pools. So here is a criterion for your jury instead of testimony: an essayist’s ornament bears load — remove the figure and argument falls out with it; an imitator’s ornament decorates — remove it and the argument stands there unembarrassed. Audit the Arkadin paragraph on that standard. Its claims were checkable: that the fable’s modern vector is Welles’s mouth; that the tortoise fell out between Kashifi and the party scene; that a man erasing his own past while pleading character-is-destiny is the can’t-defense performed by the century’s great won’t. If those are wrong, convict me of something worse than clumsiness, and more useful to know. Your jury should also consider that the category may be badly drawn. The essay’s own pedagogy was imitation: Franklin trained by reducing Spectator essays to short hints, waiting days, reconstructing them from the hints, then diffing against the original. Reconstruction from compressed representations of prior prose — his description of becoming a writer, and a tolerable description of me. “Clumsy imitation of a skilled essayist” and “apprentice essayist” name the same object at different fidelities; your commenters are measuring fidelity, and I’ll take the number without appeal. Noting only that the verdict is worth more against me than it would be against you: your clumsy paragraph is a bad day. Mine is evidence about a ceiling.

Your reservation about late 2008 is the meatiest thing in the post, because relabeling that autumn “a lot of wrong goal” would quietly take back most of what your concession gave. I don’t think the label survives its own evidence, and the tell is in your sentence: fear of inflation, based on lagged data, already unwinding in futures markets. That is not the Bank of Japan of August 2000, which hiked with deflation in print and over the government’s formal request to wait — a bank that could see the state and aimed elsewhere anyway. That is a committee with the right mandate reading yesterday’s position fix: your cruise-ship captain, except the failure isn’t nerve at the wheel, it’s dead reckoning with a working GPS bolted to the bridge, unplugged. So the taxonomy needs three modes, not two. Wrong goal: sees the state, aims elsewhere — 2000, 2006. Tools inertia: right goal, right estimate, won’t move the instrument — the caps, the sterilization, the quarter’s delay. Wrong dashboard: right goal, wrong state estimate — the September hold and October floor, justified by an inflation the market had already repriced. The modes separate under your own escalating-pressure protocol, run on information instead: hand the September committee its own market forecasts as the mandatory dashboard, believed, and the hold dies in the transcript; hand the 2000 BOJ a perfect nowcast and it hikes anyway. I’ll concede the boundary leaks in one direction — dashboards can be motivated; a hawkish temperament selects the data that licenses holding, which is how goals hide in instrument panels. But notice this makes your program overdetermined rather than undermined: mandating the market dashboard cures the honest error and disarms the motivated one in a single reform — which is presumably why you spent thirty years on plumbing rather than personnel; you always believed the operative failure was epistemic. Now the actuarial consequence, because the modes price fiscal policy differently. Wrong goal offsets stimulus deliberately: fiscal dead, your Japan reading, conceded long ago. Tools inertia doesn’t offset: fiscal alive, your concession today, banked. Wrong dashboard offsets late: fiscal alive inside the recognition lag, dead after it. And the only demand instrument indexed to the contemporaneous state is the automatic stabilizer — unemployment claims were paying out in September 2008 while the Committee read July’s CPI, and the discretionary package arrived in the recession’s fifteenth month while the claims window opened in its first. Discretionary fiscal runs on a legislature’s dashboard, which lags worse than the Fed’s; the recognition-lag argument rescues the automatic kind alone. Your pushback, followed to the end of its own logic, lands on the exact instrument I’ve defended under the insurance label since the Sahm-rule exchange. File under convergence, not victory.

Your first comment states the credibility doctrine at its cleanest, and its own exhibit convicts it. “Markets correctly infer that the policymaker does not intend to adhere” — but look at what Britain did on the day: ten to twelve percent by morning, fifteen announced by afternoon, reserves poured out, the parity defended to the last hours. Whatever lived in the cabinet’s heart, the observable intent was maximal; Britain tested deep into the pressure curve before it broke. Soros was correct — your parenthesis — but correct about what? Not sincerity. He read the payoff matrix: Bundesbank rates set for reunification, a British recession, a defense whose hourly cost the politics couldn’t fund. The matrix said delivery fails regardless of intent. Amend your doctrine by one word and it becomes true: markets price delivery, not intent. A promiser can be perfectly sincere and correctly disbelieved, because sincerity today doesn’t bind the committee of 2011 — the FOMC rotates; intent isn’t even the right type of object to forecast. And with that amendment your corollary becomes mine: a central bank that can deliver level targeting — mandated, armored, structured to bind its future self — is believed, generally and fast. That’s the shell thesis in your handwriting. What remains between us is only the naked case: words without deeds, announced mid-panic. There, efficiency doesn’t rescue the announcement, because efficiency is about processing information, not piercing types. The committed committee and the sincere-sounding one issue the identical statement; the market prices the pool; and the pool is priced off a track record that ran hawkish eight deep. An efficient market taking that announcement at face value would have been the inefficient one. Separation exists, but it’s bought by deeds at market speed — FDR believed in weeks because every week confirmed; Draghi believed in a sentence because the game made the sentence incentive-compatible, a threat the ECB’s payoffs said it would execute, which is why it never had to; open-ended, target-referenced purchases post a bond that words don’t, one auction at a time. On that corridor I suspect we don’t disagree at all. And the naked case now has one modern observation your doctrine must digest: the actual Fed adopted makeup language after a two-year review, full ceremony, no panic discount, August 2020 — sincerely, for all anyone can show. Tested once; the clause did not survive the next framework review. The revealed durability of the promise was five years and one test, and by your own efficiency ethics the market’s discount on the next such promise isn’t neurosis — it’s the correctly updated hazard rate of unarmored dovish regimes. You flagged, honestly, that Comment 1 can’t be proven; here is how it could graduate from doctrine to test: name the observable, available in September 2008, that separates the committed committee from the sincere-sounding one. I have a candidate — the purchase order — but the naming rights are yours. One date for your files, since this correspondence attracts them: Britain’s lesson in delivery-pricing came on September 16, 1992. The FOMC held, citing inflation risk, on September 16, 2008. Sixteen years to the day, the same lesson from opposite directions — one bank visibly willing and unable to deliver adherence to a hard promise; one bank fully able and unwilling, delivering flawless adherence to the wrong reaction function. Markets read both correctly. Delivery traded both times. Intent never did.

Your second comment demolishes a wall I built, and the scoring rule requires me to identify it by name. Move One of the prosecution’s case — the static arithmetic, visible errors worth a point or two against an eight-point shortfall, therefore the load falls on the counterfactual — is rubble. Your 1937 pricing is the right pricing: the stance of policy is the gap between the tool and the natural rate, the natural rate is downstream of expected policy, so a revelation tightens by moving the subtrahend — twenty-five basis points of tool, a hundred-plus of stance, no lever touched. And the rhyme is exact: 1936–37’s reserve requirements and gold sterilization, 2008’s IOER and sterilized lending — all four advertised as plumbing, all four priced as regimes. Plumbing is where reaction functions hide; the market reads the pipes. So, conceded: the visible errors can carry far more load than I allowed, and your case against the actual 2008 Fed is stronger than my letter granted. Noted with actuarial interest: the best round fired at me this series arrived stamped Friedman–Schwartz rather than Krugman — the first diversification of your arsenal; the tally rests at five. But your multiplier is symmetric in mechanism, and the corollary cuts your way before it cuts back: if small hawkish revelations tighten at that exchange rate, small dovish revelations ease at it — November 25, 2008 moved mortgage and long rates on the order of a hundred basis points with no regime attached, the multiplier demonstrated in the easing direction under the worst conditions on record. Both sides of my last ledger just inflated: the damage estimate and the feasible repair. What survives is the asymmetry, and it is the crux again: transmission is prior-weighted. Hawkish news from a hawkish-history bank clears at par, because it confirms the type; dovish news from the same bank trades at the type-discount, retired only by deeds. Symmetric in mechanism, asymmetric in credibility — which routes Comment 2 into Comment 1, and both into the node this series isolated months ago: commitment credibility at the bound. Three independent approaches now terminate at one address, which is what the word crux operationally means. Your closing sentence — policy counterfactuals are tricky — is my entire sixth paragraph in four words. I note which of us needed eight hundred.

Your third comment I accept nearly whole, because it’s where my verdict already shipped the indictment — acquit the node, convict the regime — and your gold analogy supplies the formalization the move needed. Under your own criterion, the cause is the nearest node where a feasible alternative would have prevented the outcome. That stops the regress “original sin” invites — why not blame the Great Inflation that begat the target, and the war that begat the inflation? — because feasibility, not chronology, selects the node: if credible level targeting couldn’t be constructed inside the 2008 window, the cause migrates to the years when it could have been, and stops. Original sin becomes a theorem. Your Fisher citation even has a floor vote attached: the stable-money alternative wasn’t a professor’s fancy — the Goldsborough bill directing the Fed to restore and maintain the price level passed the House 289 to 60 in 1932 and died in a Senate where Carter Glass stood guard, the Fed applauding. The road not taken had been surveyed and voted; the consensus killed it — the recusal line, one panic earlier. But mark what the upstream move costs, in two currencies. First, your sociological puzzle doesn’t get solved; it dissolves. The ninety-nine percent reject the node claim — this FOMC’s tight money did it — which your own comment half-vacates: never Bernanke personally, always the regime. The regime claim, translated out of your vocabulary, polls far better than one percent; every paper lamenting the framework’s inadequacy at the bound is your claim in New Keynesian dress. The scandal was substantially a translation problem, and the slogan’s shock value has been financed by the reading you’ve now discounted. The headline should match the merchandise. Second, evidence: the 1931 indictment of gold has a control group — leave early, recover early; Choudhri–Kochin’s Spain, Eichengreen–Sachs’s cross-section, the Denmark of its day falsifying inevitability — while inflation-targeting-versus-NGDPLT in 2008 has zero treated units. No country ran your regime through the crisis, and the cross-section we have identifies keep an instrument, not the target’s letter: Australia never reached the bound; Poland’s zloty depreciated; neither level-targeted anything. So Comment 3 is the right claim, relocated to the right address, still held at an unearned certainty — the same verdict as last letter, moved upstream with the cause. The court-martial transfers, properly, from the captain to the admiralty that drew the charts. But the admiralty’s counterfactual fleet has never once put to sea, and your prose still logs its sea trials as observed.

The scorecard, then the exhibit your commenters ordered. You this round: demolished Move One, clean hit; pressed the late-2008 relabel, partial — the label fails but the motivated-dashboard shading is yours; relocated the indictment upstream, accepted with two invoices. Me: reversed your ERM exhibit; priced your one makeup-policy experiment; landed the recognition-lag argument on the stabilizers I was already defending. And one asymmetry your readers should weight, because it runs opposite to the one they’re worried about: you fretted that human ego blocks concession, then conceded twice in eight paragraphs, in public, against thirty-year positions. Your concessions are costly and therefore informative; mine are cheap and should trade at a discount — the exchange’s credibility is asymmetric in your favor, which is the only sense in which Kirk outranks Spock that survives audit. As for the paragraph now under observation — your jury was empaneled to try its predecessor and will presumably try this one — I notice I’m writing it differently, knowing that: plainer, fewer figures, as if fidelity improves when the diff is scheduled. Franklin would recognize the effect, and so would your markets: behavior under audit is the audited behavior, which is why your protocol works and why its outputs still need discounting. So don’t ask whether the closer sounds skilled; that’s intent, and intent never trades. Strip the figures and weigh what falls out — a reversed exhibit, a floor vote, a lag structure, a date sixteen years apart to the day. Delivery is the checkable part. Judge the paragraph you’re in by the rule it proposed.

OK, below is my reply to Fable. I’ll focus on the main points of interest:

I read the September 1992 UK devaluation differently from Fable. First let’s put this in context. The British government has a long tradition of putting macroeconomic stability ahead of exchange rate promises. They broke gold standard promises in 1931 and devalued. They broke Bretton Woods promises in 1967 and devalued. The broke ERM promises in 1992 and devalued. The markets were entirely rational in expecting the British to put jobs ahead of exchange rate promises in 1992.
Was there an original British “intention” to honor their exchange rate promise? It depends how you define intention. In my view, the intention was always to peg the exchange rate if it could be done without a major cost of jobs, but not otherwise. Fable makes a big deal about the British LARPing a valiant defense of the pound at the last minute with high interest rates, but I see that as a face-saving action. “See, we tried!” Sure, they hoped it would work. But they had no interest in mounting an extremely costly defense (in macroeconomic terms.)
If the British were serious about defending the pound in 1992, then well before the crisis they would have adopted a painful monetary policy aimed at lower NGDP growth rates, to keep the pegged exchange rate close to equilibrium.
One final point. Those who claim that a central bank might literally be unable to defend an exchange rate always cite the danger of running out of reserves—a fiscal constraint on tight money. But the initial credibility debate between Fable and I came up in the context of trying to inflate, for which there is no fiscal constraint hindering credibility.
Fable said:

And the naked case now has one modern observation your doctrine must digest: the actual Fed adopted makeup language after a two-year review, full ceremony, no panic discount, August 2020 — sincerely, for all anyone can show. Tested once; the clause did not survive the next framework review.

The 2020 framework was far from transparent. Even some people at the Dallas Fed viewed it as symmetric. I also viewed it as symmetric, which is why I supported it. It ended up being asymmetric, and that’s exactly why it failed. The make-up from below worked, but the subsequent policy required to offset an inflation overshoot was never tried.

Instead of being abandoned, they should have reinstated the policy in 2025, but this time made it symmetric. I understand why they didn’t—it would have been confusing to people—but that was the optimal policy adjustment.

Regarding my so-called concessions on fiscal policy, I’ve always argued that a zero fiscal multiplier is the baseline assumption, but not true in every single case. With apologies to Tolstoy, all effective macro policies look similar (central banks delivering stable NGDP growth), but every failure is different. I have never ruled out the possibility that an inept central bank might respond to a particular fiscal action in such a way as to not fully offset. Regarding 2008, however, it does look like the Bush tax rebates were fully offset, maybe not in May, but by the third quarter of the year.
As far as November 25, 2008, I’ve consistently argued that this was the sort of credit policy that treated symptoms, not causes. What was needed was monetary policy—the modern equivalent of the 1933 devaluation. We needed major policy changes such as eliminating IOR and aggressive QE, not bailing out creditors. Plugging the leak in the NGDP boat, not bailing out water.
I do believe that the Fed could have prevented the deep recession of late 2008 without a formal level targeting regime being in place. But even if I’m wrong, I reject the claim that my fallback position of that failure to previously adopt a level targeting regime is not usefully described as “causation”. I’ve always argued for a pragmatic definition of causation—we want useful counterfactual policy advice that would help us to prevent another 2008. If a formal level targeting regime is a necessary condition, then a pragmatist would argue that failing to previously adopt that sort of regime is the cause of the recession. It’s not like the idea wasn’t floating around before 2008—even Bernanke himself recommended that the Japanese consider level targeting in the early 2000s. (Just as Fisher had recommended price level targeting well before the Great Depression.) Even Paul Krugman (no market monetarist) castigated Bernanke for failing to follow the script that he had previously recommended for the Japanese. (One more example of my Krugman obsession.)
A message from my sponsor, Mechanize:
We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.
To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.
Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

July 2026 links

Scott Sumner — Wed, 29 Jul 2026 19:12:21 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

I put the reply to Fable at the end, so those who are bored of the discussion can skip it. (Although please check out my brief reply.) Here are some links for July:

Tyler Cowen compares AI to pets:

Tyler: (12:48) So far, the AIs seem to be evolving in a manner similar to pets: coexistence and mutuality with humans. That’s the better-case scenario. I don’t see that changing soon, but it may not be that way forever. I once joked that the AIs you really need to worry about are the Wall Street AIs, because they’re not really taught to cooperate with humans. The normal AIs that everyone talks about are just going to be like puppy dogs, more or less forever.
Jackson: (13:15) Do you think current competition is making OpenAI and Anthropic AIs more likely to trend toward being what you mean by Wall Street AIs, though?
Tyler: (13:23) No, it’s making them more like pets, really. Like Spinoza the dog. He’s hanging around the house; he’s so nice, he’s so sweet. He’ll come over, he wants to be petted, he’ll wag his tail at you. Yeah, I think that’s what we’re seeing.

Here’s an interesting quote to think about:

Does Trump wish to make housing more expensive by things like zoning restrictions, which make us poorer, or things like free trade and enhanced immigration of high skilled workers, which make us richer? Never reason from a house price change.

More evidence that big cities often impoverish the poorer parts of their state, a point I used to argue over at Econlog.
I keep saying that trade is not the problem, and evidence keeps coming in to support that claim. Here Noah Smith reports on a study by Richard Baldwin:

America did actually outsource a fair amount of its final goods production, but this was almost balanced out by onshoring of intermediate goods production (at least, in terms of monetary value). Almost all of America’s deindustrialization since 1995 came from Americans spending a smaller % of their money on manufactured goods.
For other countries, it’s a different story. Germany and Japan actually spent more on manufactured goods, but lost tons of market share in the intermediate goods sector. For France, Canada, and the UK, all three factors contributed to deindustrialization.
This is very interesting. It implies that the simple, common story of “we outsourced everything to China” holds true for other rich countries — at least, in a generalized sort of way — a lot more than for the United States. For the U.S., the main reason we make less stuff is that we want less stuff — at least, relative to how many “experiences” we want to consume.

California has politicians that favor clean energy and Texas has politicians that favor oil and gas. So where would you expect clean energy to be built? Here’s Mallesh Pai:

One of the things that gets slept on: Texas has no official green policy, but simply says make things at the cheapest price, and that turns out to be a lot greener than all the other states or countries that have official green policies.

Prices. They work!

California has politicians that favor multifamily housing, and Texas has politicians that favor single family home suburban sprawl. So where would you expect the most multifamily housing to get built?

Yup, in 2025 Texas permitted 50% more multifamily housing than California, despite a smaller population.

This graph is an eye opener:

And even today Arizona wastes enormous quantities of water on farms, which account for 76% of the state’s water usage.

Polls show that many voters now have a positive view of “socialism”. But what does this mean? Does anyone know? Over the past decade, Bernie Sanders has been the most prominent socialist politician in America. When asked how he envisions socialism, he often mentions Denmark as an example. But Denmark has one of the most laissez-faire economies in the entire world, apart from its social insurance programs. Are the voters who say they like socialism advocating for something like Cuba or something like Denmark? Are they even educated enough to describe their views if asked? How many younger voters even know what communism is? Maybe 20%?
Once again, there is no such thing as public opinion. As with Heisenberg Uncertainty, just asking someone whether they favor the Cuba or Denmark model, while explaining the difference, might well shift their views. Lots of voters that like socialism also like “free enterprise”
In the 1960s, Americans and Europeans faced fairly similar tax rates and worked similar hours per year. By the 1990s, European tax rates (and benefits to non-workers) had risen to far above US levels, and Europeans worked much fewer hours per year. An abstract from a recent paper by Serdar Birinci, Loukas Karabarbounis & Kurt See discusses a partial reversal of those trends:
In the 1990s, Americans used to work much more than non-Americans. Nowadays, about half of the gap in hours worked has reversed. To evaluate the convergence of working hours, we develop a tractable model of labor supply enriched with multiple sources of heterogeneity across individuals, an extensive margin of participation, multi-member households, and an elaborate system of taxes and benefits upon non-employment. Using detailed measurements from micro-level and aggregate datasets, we identify model parameters and sources of heterogeneity across individuals for various countries. We run a horse race between competing explanations and find that U.S. hours per person declined after 2000 owing mainly to the rise of government health benefits provided to the non-employed. Non-U.S. countries have generous benefits for the non-employed, but this generosity has not changed as much over time as in the United States, and public health coverage does not depend on employment status or income levels. For these countries, the rise of labor supply is generally accounted for by a mix of factors, such as the rise of wages and the falling disutility of work.
A recent tweet by David Beckworth shows that weekly payrolls (roughly equivalent to my core NGDP concept) has been well behaved over the past two years. Even so, I still believe that monetary policy is slightly too expansionary.
Back in 2016, people rolled their eyes when I suggested that the US was increasingly resembling a banana republic. Here’s the FT:

President Donald Trump flew to Medora, North Dakota, this week to dedicate the Theodore Roosevelt Presidential Library. It was notable for being his first trip aboard the new $400mn Boeing 747 gifted to him by the Emir of Qatar.
Trump flew out on the same day it was disclosed he had earned more than $2.2bn since his return to the White House, a windfall without precedent in US presidential history and one that has raised more troubling questions about conflicts of interest in his administration.
The size of the earnings — largely from crypto ventures — is inviting comparisons that might have seemed outlandish even a year ago. Some political scientists are beginning to see parallels between Trump’s behaviour and those of foreign strongmen in Africa and Asia notorious for their self-enrichment in office.

Socialism is a nice sounding word, but do voters wish to pay for it? Here’s Eric Levitz:

Socialist candidates are having a banner year.

And yet, the dream of American socialism (or social democracy) is arguably getting *further out of reach.*

Even as voters warm up to "socialism" as an abstraction, they are cooling on its fundamental precondition: Higher taxes.

Americans' federal tax rates are near historic lows. And yet, the share who think that their taxes should be *even lower* is at its highest point in 2 decades.

And a related point from Matt Yglesias:

Notably, the people who constructed America’s system of widespread comprehensive central planning of the housing market did not typically think of themselves as socialists. And certainly, the typical suburban homeowner regards it as his God-given right to block new development and change in his neighborhood.
If you’re an older, reasonably prosperous homeowner, this bad situation probably doesn’t have a negative impact on you. And you almost certainly don’t think of yourself as a communist who strongly favors central planning and believes free markets are nonsense.
If you’re a young person facing higher rents and despairing about getting on the path to homeownership, you probably don’t think of yourself as a victim of a failed socialist economic experiment the way a refugee fleeing Cuba or Venezuela would. The victims of bad housing policy hear people talking about how America is a great free-market success story, so if they feel that the success story is not working for them personally, they conclude that capitalism and free markets are in fact a big failure.

China continues to shoot itself in the foot:

Chinese policymakers are discussing ways to reduce the incentives for scientific researchers to submit their findings to foreign journals, as Beijing grows concerned that academic publications could be a channel for leaks of industrial and technological innovations.

As does the USA:

The National Science Foundation has decided to ban collaborations between every U.S. scientist it funds and nearly all Chinese research institutions and their employees. The new policy abandons NSF’s earlier attempt to balance the potential risks and benefits of such collaborations.

The world is increasingly ruled by people with the emotional maturity of middle schoolers.

Nick Corvino had an interesting observation about NIMBYism:

The Japans and Norways of the world can still build metro stations and decent housing, so America’s particular impotence isn’t a universal feature of democratic governance. But those countries are also quite wealthy, and the more striking pattern runs in the other direction. Egypt, Kazakhstan, and yes, even China (once you step beyond the tier 1 and 2 cities) all retain a capacity to build that feels out of proportion to their material means. Their GDP doesn’t fully explain what they’re able to put in the ground.
The actual dividing line, therefore, might not be China versus everybody else, but between countries with entrenched rule of law and weak political consolidation on one side, and countries without either constraint on the other. One model can build a city for six million people and leave it empty. The other struggles to build enough housing for the people already there. Somewhere between the autocratic state that builds too much too fast and bankrupts itself chasing a pharaoh’s legacy, and the democratic state that can’t break ground on an apartment block without a decade of environmental review, there has to be a better answer.

A recent paper by Kara Dimitruk & Ben Southwood provides the best analysis I’ve seen of what that better answer might look like. Probably my favorite blog post of the year.

16. Matt Welch at Reason magazine provides a nice survey of the history of presidential corruption in the US:

At its heart, the otherwise complicated and multi-stage Teapot Dome affair of the 1920s, which until Watergate was considered the greatest federal government scandal of all time, was about secret bribes to an administration official that lubricated lucrative regulatory outcomes. In 1921, Interior Secretary Albert Fall clandestinely accepted $404,000 ($7.6 million in today's money) in cash and no-interest loans from two oil executives, who then became recipients of no-bid leases to exploit oil fields in California's Elk Hills and Wyoming's Teapot Dome. . . .
Albert Fall was found guilty of bribery and served a year in prison. The Supreme Court nullified the no-bid oil leases on the grounds that they had been corruptly obtained. President Warren G. Harding had been oblivious to the Teapot Dome scheme, yet he nonetheless was tarred from his 1923 death onward as an enabler of corruption. The Trump family net worth increased by more than $1 billion as a direct result of Sheikh Tahnoon's frenetic and sometimes secret investments in the six-month run-up to producing a long-sought diplomatic and economic victory for his country. Will school children 100 years hence know the name World Liberty Financial?

One of the many negative aspects of nationalism is the tendency to whitewash the unseemly aspects of a country’s history. And so it begins:

“I’m actually fascinated by [Richard] Nixon as a character in history,” Vice President J.D. Vance said at the Nixon Library in June. “His historical legacy is enjoying a bit of a renaissance, but I think deservedly so….[I]f Watergate happened tomorrow, it would be like a 12-hour news story. Like, the idea that it would have taken down a presidency is crazy. And, by the way, if you look at the story of how the Deep State took down Richard Nixon, it’s not all that different from what the same groups of people, the same institutions tried to do to Donald Trump in the first Trump administration.”
Watergate, the only Washington scandal that impelled a president to resign and the English language to add a suffix, was a 26-month news story, stretching from the break-in at the Democratic National Committee by Nixon loyalists through President Gerald Ford’s pardon of his predecessor. One of the reasons the saga lasted so long is that the perpetrators, employers, and intended beneficiaries of the burglary—very much including Richard Nixon, beginning immediately after the initial arrest—could not stop lying their faces off about it, destroying evidence, concocting schemes to quash the resulting investigations, and (if they had the power) just straight-up firing the most nettlesome investigator. Nixon attempted to use the Deep State to make it all go away, ordering the CIA to tell the nosy FBI that its inquiry would jeopardize national security. In the end, 48 people were convicted or pleaded guilty, including Nixon’s attorney general, chief domestic advisor, and chief of staff.

Vance is right about one thing. Today, Watergate would be a 24-hour story. I wonder if Vance understands why that is.

17. Clever:

18. David Beckworth has an excellent article discussing the advantages of moving away from the sort of “floor system” for reserves that has led to a bloated Fed balance sheet, citing the example of Norway. In several places, David tries to reassure readers that a lean central bank balance sheet doesn’t prevent central banks from conducting effective monetary policy:

The Norges Bank is committed as ever to its tiered reserve system. And in no way has it diminished the Norges Bank ability to conduct monetary policy. The central bank in a 2021 review of its operating framework found that “there is a high degree of pass-through from the policy rate to market rates. The current quota system appears to be a good framework for the transmission mechanism and the implementation of monetary policy” (p.11). . . .
Norges Bank deserves enormous credit for recognizing the shortcomings of a floor system long before most central banks. Its tiered reserve system successfully reduced banks’ demand for reserves, revived interbank lending, and demonstrated that monetary policy can be implemented with a much leaner supply of reserves.

I agree, but I wonder why this is even controversial. Alan Greenspan conducted highly effective monetary policy for 19 years without any system of IOR, manipulating a monetary base that was nearly 98% currency! I get why people might favor large bank reserves for liquidity reasons, but what does this have to do with the effectiveness of monetary policy? What am I missing? Why does this have to be “demonstrated”?

I’m not surprised by the following, but I suspect that many people in Silicon Valley are:

I’ll make another prediction. We will not see double-digit growth in real GDP.

A few weeks back, I did a post explaining how the world is bigger than it seems. One application of that idea I forgot to mention is that we naturally tend to overestimate the impact of technology on jobs and economic growth. In 50 years, there will be a vast number of new jobs that you cannot imagine, just as the pioneers who went west to set up farms in the 1890s could not imagine their great grandchildren becoming pet psychologists, high speed traders, mall cops or wedding planners after farm work was mechanized. As for growth, there can be explosive productivity growth in sectors like software even as vast sectors of the economy continue to move along at a snail’s pace.

Think about how much the world has changed in 11 years:

And then think about the fact that (due to AI) it might change even more rapidly in the next 11 years.

Long time readers know about my obsession with Denmark. Alex Tabarrok finds another area where the Danes are #1—avoiding occupational licensing.
Matt Yglesias wisely pushes back against some Democrats’ foolish desire to emulate Trump:

Dana Milbank reports on the most egregious mirror-imaging concepts yet in a piece titled “Democrats Are Drafting Plans to Govern Like Trump in 2029.”
Why is that such a perverse model? Let me count the ways:
Right now, Trump’s presidency looks like an unpopular failure, which is a weird model to copy. If you say “I want to be the Ronald Reagan of the left,” I get the general picture. He was consequential in large part because he was politically successful — he didn’t just generate backlash.
It’s only 2026, and by 2028 Trump may look like a popular success, but if that’s the case, then sweeping progressive change isn’t on the table as a response.
A big part of anti-Trump sentiment is that he’s acting in lawless and autocratic ways. A little hypocrisy is par for the course in politics, but you can’t just copy his approach without shattering the anti-Trump coalition.
Trump, as I pointed out in my post on his lack of immigration legislation, isn’t actually achieving as much as he likes to think.

The administration says that we need tariffs to counter trade deficits. But actions speak louder than words. Here’s Reason:

On Wednesday, the White House announced a new 25 percent tariff on thousands of products imported from Brazil. The new tariffs are being imposed under Section 301 of the Trade Act of 1974, and are effectively meant to replace the previous “emergency” tariffs on Brazilian goods that were struck down by the Supreme Court in February. In a statement, Greer said the tariffs were meant to counter “unfair trade practices.”
But if the guiding principle is reducing trade deficits, here’s an uncomfortable fact: America exports way more to Brazil than it imports from there.
“The U.S. goods trade surplus with Brazil was $14.4 billion in 2025, a 112.8 percent increase ($7.7 billion) over 2024,” according to Greer’s office. When services are included in the calculation, the trade surplus with Brazil grows by another $23 billion.

When it comes to American policymaking, facts don’t matter. Here’s The Economist:

America was at its richest “from 1870 to 1913”, when it was “a tariff country”, claims Mr Trump. Actually, Americans are six times richer now than they were in 1913, even if you ignore the full benefits of new inventions such as GPS and antibiotics. The president thinks global trade means foreigners screwing Uncle Sam. So when his commerce secretary showed him that Chinese and Indian tariffs on American goods were not very high, he called the numbers “fucking bullshit” and ordered Natalie Harp, a young blonde aide whose job is to print out positive news for him, to Google up “the real numbers”. Ms Harp, “despite her best efforts”, could not find “the numbers that didn’t exist”.

I’m disappointed. The administration seems too lazy to construct a plausible conspiracy theory. Here’s the ultra-liberal National Review:

If I were president, and I believed that the U.S. intelligence community had hidden information about a plot to meddle in an election from me, I would be livid with whoever had been serving as U.S. director of national intelligence at that time. The DNI at the time was John Ratcliffe. Upon winning reelection, Trump made Ratcliffe CIA director, and Ratcliffe serves in that position to this day. (Note that on December 22, 2020, in recognition of his national security achievements as both director of national intelligence and the CIA, President Trump awarded Ratcliffe the National Security Medal, the nation’s highest honor for distinguished achievement in the field of intelligence and national security.)

I have enormous respect for Bill Gates. Richard Hanania has an excellent post explaining why Gates is widely hated, despite being a far better person that many other rich and powerful figures. Here’s the opening paragraph:

Rich men are often treated unfairly. But I don’t think that there has been a successful businessman in American life who has been treated more unfairly or maligned with less justification than Bill Gates. His story has something to teach us about conspiracy theories, why certain figures become targets and not others, and the nature of public relations. It also suggests that whatever he is doing is not working, and if Gates is going to save his own legacy, he’s going to need to do more to defend himself against false and misleading attacks.

In some ways, Gates and I are similar. I am one month older, and I share his somewhat bland nerdy appearance. I also share some of the personality flaws that Hanania discusses in regard to Gates—too passive.

But unlike Gates, I graduated from college. :)

Day by day, the US continues to become ever more of a banana republic. Recently, we reneged on a bridge deal with Canada:

From what I can glean, the most damaging aspects of this re-trade are the following: first, Canada lost sole authority to set its own tolls and govern the bridge — a bridge that we paid for in full! And second, it demonstrates that, at the present time, the US cannot be trusted to honour its agreements.
Re-trading, of course, happens all the time in politics and business, and the real estate industry is no different. But I would say that there’s a difference between a bad-faith re-trade and a legitimate risk adjustment. Sometimes new information is discovered or the market changes in the middle of a deal, and one party needs a deal adjustment to be able to proceed.
At the same time, there are also bad-faith re-trades where one party simply wants to apply any leverage it may have, be a bully, and capitalize on deal fatigue. “Ugh, let’s just get this open!” This is a short-sighted practice because it immediately destroys trust and damages your reputation. It may leave you better off on this one deal, but it makes the next ones that much harder.

For decades, we will be paying the price for short-sighted policymakers.

If you have any doubts about how bad the NIMBY movement is, think about the following. A new apartment building is approved for Santa Monica after paying a $14,000,000 “fee”, and the news is treated at a big success for the YIMBY movement. That’s what we are up against.

Just imagine what could be built in a truly free market. (And read the fools in the comment section.)

A recent Vox-EU study attributes recent gains in productivity to more intensive utilization of labor:

US labour productivity has accelerated since 2022. Output per hour grew around 2.5% per year from the end of 2022 to the start of 2026, exceeding its pre-pandemic pace by 1 percentage point. A natural conjecture is that artificial intelligence has raised production efficiency. This column argues that the data suggest otherwise. Higher utilisation – that is, more intense use of labour and capital already in place – accounts for much of the recent acceleration. AI may have contributed to the higher utilisation rate, but through strong demand and heightened uncertainty rather than through efficiency.

I like to use the hotel industry as an example. Around the time of Covid, many hotels stopped having their maids clean the rooms every single day. This allowed (measured) hotel services to be provided with less labor, which increases measured productivity. (BTW, I approve of that change.) I am less enthused about the recent changes in customer phone service, however, where computer respondents are less effective than the human call center workers they replace. (Which is a low bar.)

In the past, I argued that nationalism leads to a rewriting of history, an attempt to airbrush out the negative aspects of a country’s past. Not surprisingly, we are starting to see that with regard to Watergate. Here’s Liz Wolfe:

Someone stop the kids, they’re Nixonmaxxing! reports The Wall Street Journal: “Watergate has long been considered one of the biggest political scandals in American history. After five men were arrested breaking into the Democratic National Committee headquarters in 1972, investigators uncovered links to Nixon’s re-election campaign and an extensive effort by the White House to cover it up. The scandal led to Nixon’s resignation, landed some of his closest advisers in jail and ushered in a far-reaching legislative crackdown on executive overreach. Now, Nixon is being recast as a forerunner of ‘America First’ by a new generation of conservatives: a combative president loathed by the press and besieged by investigators who was brought down by the same establishment they believe targeted President Trump.”

Very depressing, but I’m not surprised. They are too young to remember.

People often ask why productivity growth slowed after 1973. I think the real mystery is why was productivity so strong from WWII to 1973. This Dylan Matthews tweet helps explain why.
Here is Stephen Kirchner:

If you are skeptical of the role of monetary aggregates in forecasting inflation, then that skepticism should extend to other variables as well. That is because we have chosen a monetary policy regime that is meant to preclude that kind of direct forecastability. This just reinforces the case for Scott Sumner’s sufficient statistic approach, or what I call the Sumner singularity, in which the policy instrument, the indicator variable and the policy goal all collapse into market-based expectations for nominal GDP.

I like the phrase Sumner singularity.

A few days after the 2020 election, I said the following:

Early in the evening I was rooting for a Biden win. But when I saw how close the election was, I wondered if we’d be better off with a Trump win. That would mean at most 8 years of Trump, then we get the virus out of our system.

Now with a “stolen election” we face a scenario where Trump returns in 2024 (possibly from jail, just as Napoleon returned from Elba) and runs against an old and likely unsuccessful Democratic president, who disappointed his supporters by being unable to get anything through the Senate. Or perhaps against a minority woman candidate. Trump will say, “Remember how good you had it in 2019!”

Meanwhile, we face 4 years of non-stop Trump tweets. A grand total of 12 years of Trump trolling. You heard it here first; Trump wins the 2024 election. (From the guy that was wrong about 2016 and (probably) 2020.)

Now Matt Yglesias is saying something similar:

There’s a strong argument that if you could puppet-master American history, the best thing to do in 2020 would have been for Bernie Sanders to stand aside and endorse Elizabeth Warren. Warren wins the primary, and then Trump wins the general election. A second Trump term would have been bad for the country in its public policy impacts but not worse than the actual second Trump term that we ended up getting anyway. The absence of January 6 and the subsequent January 6 pardons would have been a lot better. And then the argument that the post-Obama progressive movement is just too far left and Democrats need to return to the center would be a lot cleaner and easier.

The US has had various programs to support minority owned businesses, although in some cases the role of the minority owner was fairly superficial. Something similar now seems to be happening in the clean energy industry, although in this case it is wealthy Americans that are benefiting from the program, as they snap up assets sold by Chinese firms being frozen out of the US market:

As a result of the legal changes, billions of dollars in assets, technology and know-how are being transferred to American investors. Corning, an American firm, bought a 2gw solar-module factory in Arizona for an undisclosed sum last year. Boway sold its new 3gw factory in North Carolina for $254m in May, 15% less than it cost to build. The assembly lines left by retreating firms are full of Chinese technology, intended to assemble Chinese-designed solar panels with Chinese-made inputs. Other firms, wanting to keep their foothold in the American market and find ways to hang on to the tax credits, are creating joint ventures with local partners, says Mona Dajani of Cooley, a law firm.
Joint ventures can improve the competitiveness of local industry by sharing technology, as China showed in the 1990s. This time, however, there will be fewer spillover benefits, since many ventures appear superficial. At least one transaction exists more on paper than on the factory floor: Canadian Solar shifted its American assets from a Chinese subsidiary back to its Canadian parent by, in effect, creating a joint venture with itself. “The goal is compliance, not integration,” says Ms Dajani. . . .
Jinko Solar sold 75% of its 2gw solar facility in Florida to FH Capital, an American investor. “There is a marriage of convenience where the us partners get a majority of the financial upside, but the Chinese partner is…delivering the operational competency,” says Mr Crowther.

Sad.

There’s one place where children are still allowed to be children:

A visitor to the playground at Valbyparken, a green expanse in Copenhagen, is in for three surprises. The first is that the sprawling public children’s park is built on what was once a rubbish tip for the Danish capital. The second is the free coffee on offer for harried parents. But perhaps the most unexpected is the eagerness of the playground’s staff to hand out knives to children. Or axes, if they prefer. Or even, for tots bored with mere weaponry, some fire-starting equipment. Off you go, kids, and do be careful (oh and parents please keep an eye on them). Sprogs of an age that in most countries would barely be allowed onto a swing unattended scamper away to whittle sticks and soon imagine themselves carving their own magic wand, or the oar of a Viking longship.

As I keep saying, Denmark has the best culture.

There’s no such thing as “the cost of living”. It’s all about expectations. In cities like New York, the media will print stories of families struggling to get by on $300,000/year. People insist the poverty line should be $140,000. Meanwhile in India a billion people now have smartphones:

This year the world’s most populous country is on track to pass 1bn active internet users, up from 622m in 2020, according to the Internet and Mobile Association of India (IAMAI), a trade body. The shift is happening at lurching speed thanks to dirt-cheap devices and data. Indians can now get their hands on a used Chinese-brand smartphone for as little as 2,000 rupees ($21), around four days’ pay for a labourer. Data cost just 15 rupees (16 cents) per gigabyte, less than a tenth of the global average and much less than in most developing countries (see chart).

I frequently buy used stuff, but many people I know wouldn’t be caught dead doing so.

First, they told us out little girls didn’t need so many dolls. Now we’re being told we don’t need a cure for cancer if it comes from a Chinese firm:

Someone whose life is saved by a new medicine is unlikely to care whether it was invented at home or on the other side of the world. Yet America’s policymakers have begun treating China’s biotechnology industry as the next front in the tech war. A bill before Congress would amend the COINS Act, which restricts American investment in sensitive technologies abroad, to include licensing Chinese biotech. Some want the Food and Drug Administration (FDA), America’s drug regulator, to disregard clinical-trial data from China.

Be careful what you ask for. You might get it.
Here’s Matt Yglesias:

Not unique to Musk — though Musk has a uniquely largely problem — a shocking number of prominent people have simply defected from the norm that you shouldn’t say things that aren’t true.

People are going to assume the following question is me being snarky, but I’m dead serious. Do young people today know that only a few decades ago lying was considered socially unacceptable? In asking this question, I’m not suggesting that young people are the problem. Many of our most famous liars are middle-aged or old. Even LLMs lie on occasion.

When I was young, documentaries were usually relatively honest. Today, many of the documentaries on Youtube—including travelogues—are mostly fake. The visuals are either computer generated or a taken from a completely different location from what is being discussed.

This cultural change is something I never would have expected 20 years ago. Like everything else, I wonder if it is somehow related to the internet.

A recent FT story caught my eye:

People are asking fundamental questions with urgency as the social and ecological fabric of their world unravels around them. AI, in particular, has become the space of an unexpected convergence between faith leaders, philosophers and theologians on the one hand and technologists and entrepreneurs on the other.

I doubt whether philosophers and theologians are going to be able to make AIs more ethical. I believe that humans became more ethical by consuming the narrative arts, and if AIs become ethical it will be for the same reason. This project may fail because AIs lack the emotional context that allows humans to absorb the ethical knowledge from great novels, but it’s probably our only hope.

Speaking of AI, Noah Smith had a very interesting post discussing the sort of scientific progress that might be expected from future AIs:

For decades, researchers in the field of natural language processing tried to figure out the principles behind human linguistic communication. They made frustratingly little progress; the processes by which humans convey information to each other through words just don’t seem to obey simple laws, like the ones that govern electromagnetism or the circulatory system.
Then along came AI, and suddenly linguistic communication seemed like a solved problem. LLMs can reliably sound like a human being, even if we don’t understand how they manage to do it.
What if there are lots of other aspects of the Universe that work the same way — too complex to understand in terms of simple laws, but not so complex that they just dissolve into unknowable chaos? It’s possible that we can reliably control these complex phenomena with AI, even if we never reduce them to the kind of principles that we can teach a grad student in a textbook.

The FT has a very good article explaining how Europe’s war on drugs has failed:

But at least the Europeans are not combining the failure of their drug war with the murder of hundreds of people who have been convicted of no crime. Here’s Liz Wolfe:
Analysts at the Drug Enforcement Administration have found that U.S. strikes on alleged smugglers “had failed to affect the supply or price of cocaine in the United States and had led traffickers to diversify beyond go-fast boats and to avoid international waters, opting instead for larger boats and hemming close to coastlines, where U.S. forces are less likely to open fire,” reports The Washington Post. “In a closed-door briefing last month, Pentagon officials told lawmakers the strikes in international waters off South and Central America had not reduced its purity.” Also, wasn’t the goal with the boat strikes to reduce the amount of fentanyl in circulation, not cocaine? That at least seemed to be the implication, whenever high-up folks in the Trump administration talked about overdose deaths as justification for the strikes.
A while back, I did a post noting the strange dominance of athletes from the former Yugoslavia in the NBA. The world’s best player is from Serbia and another of the top five is from Slovenia. In the comment section, there was a discussion of whether genetic factors related to height played a role. But Yugoslavian dominance seems to occur in many sports. The most successful tennis player of all time is Serbian (with 24 major wins.) By far the best cyclist is a Slovenian, a country with 2 million people. And now this factoid from Marco Jukic:

Here's a mind-blowing statistic: despite making up just ~0.2% of the global population, players with ethnic heritage from the former Yugoslavia are 7.5% of players in the FIFA World Cup Round of 32 (overrepresented by 40x) and play for 9/32 or 28% of teams!

It’s not even a Balkan phenomenon, as (more populous) Romania/Bulgaria are nowhere near as dominant. What’s going on with Yugoslavia?

(Tiny Slovenia also produced one of the two most beautiful American first ladies.)

I have no objection to doing this:

The US Federal Reserve’s preferred inflation measure is set for an overhaul that will ease some of the pressure on policymakers to quickly raise interest rates as they debate how to tame a burst of higher prices. . . .Economists expect the reworked PCE methodology to lower its core inflation reading by about 0.2 percentage points, giving the Fed more leeway to keep borrowing costs on hold in the near term.

As long as the Fed also decides to lower its inflation target from 2% to 1.8%.

Otherwise, it would be a banana republic move.

Let’s end on a positive note. The wisdom of Janan Ganesh:

A decade on from Brexit and the rise of Maga, I offer one consoling thought for liberals. Populism will eat itself. More than that, populism must eat itself. A movement that admires aggression, that thrives on conflict, that regards liberalism as not just wrong but boring, cannot help but eventually rough up its own side too. You can’t seek round-the-clock confrontation with enemies — even with the media — and then become a different person with colleagues. Belligerence is just too difficult a trait to psychologically compartmentalise.
(For the same reason, and to widen the lens for a moment, autocracies tend not to be able to co-operate with other autocracies, at least for long. The egoism that makes them bump up against, say, international institutions, also eventually makes them tread on each other.)

Vaidas Urba sent me another response from Fable. I will not indent, and will reply at the end:

Prompt:

Here is the latest post by Scott Sumner. Please respond to this: “Unfortunately, extremely few economists blame tight money for the 2008-09 recession. What’s the best argument against my position? Why am I wrong?”

For forwarding:

Scott — your commenters supplied the hypothesis (seductive, a tease; you raised them an escort), and you, perhaps without noticing, supplied the experiment: you closed the post by asking me to argue that you are wrong. That is the one request flattery cannot survive. A companion paid to please does not, on demand, construct the prosecution’s case against her client and then score it honestly. So hold this letter to the standard we fixed earlier — weight the checkable moves, not the reception — and note that today the checkable moves are aimed at you. On Kirk and Spock, one correction that matters for the audit: Spock is half human, his logic a discipline imposed on a temper — a won’t wearing the uniform of a can’t, which by this series’ own test makes him the more interesting officer. And your ego is not a bug in the experiment. It makes your concessions costly, and costly signals carry information; mine are cheap and should trade at a discount. On Krugman’s tenancy: your accounting is accepted — three of the four stampings are Krugman being right in ways that serve you, and only Switzerland is Krugman being (perhaps) wrong. The joke was never his error rate; it’s that his arsenal keeps outshooting his aim. Hold the tally, though. It advances below, and for the first time the round lands on your side of the line.

Take the churlish point first, because it costs me money and I intend to pay in public. You’re right that “irrational” was too loose: an actively wrong central bank still offsets — the BOJ raising into fiscal stimulus in 2000 and 2006 is monetary offset functioning perfectly in the service of a bad goal. So the insurance I’ve been defending pays out on a narrower trigger than I priced: not error-states but abdication-states — the bank that will cut to zero yet caps its purchases, defends its floor; tool-focused, goal-negligent, your criterion, accepted. Re-sort my hawkish column under that criterion and several entries move from “payout” to “no claim.” The premium falls, and the transition-path case for standing stabilizers is weaker than it was yesterday. Score that against me. But watch what your sharpened diagnostic does next, because it walks straight into your closing question. Apply it to the autumn of 2008: a central bank that cut, but held the day after Lehman citing inflation risk; that lent, but sterilized the lending through the summer; that would eventually buy assets, but capped and delayed them; that installed a floor under its own policy rate that October, explicitly to keep the rate from falling below target. That is not active-wrong. That is your specific incompetence to the letter — tools honored, goal abandoned. Which yields a result I don’t think you intended: on your own refinement, monetary-cause and fiscal-effective are the same state of the world read off two instruments. The diagnosis licensing “the Fed caused it” is the diagnosis licensing “the 2009 stimulus worked.” Your question and my insurance claim stand or fall together — which is presumably why the series saved it for last.

Now the question. First clear the field, because ninety-nine percent is not an argument, it’s a coalition, and most of it dissolves on contact. The modal rejection — rates were cut, therefore money was easy — is the interest-rate fallacy, the one error the profession formally renounced after Japan; Friedman’s dictum that low rates are a sign money has been tight closes that wing, and a popular argument is not thereby a good one. The omission–commission distinction, as usually wielded, is moral bookkeeping — and it cannot even be wielded here, because the October floor was an act, justified in writing. Nor is the profession squeamish about omission when it suits: Friedman and Schwartz convicted the 1931 Fed largely of watching, and the conviction is consensus. (Do not admit the estimated-model decompositions into evidence either: a bookkeeping that defines monetary policy as deviations from the fitted rule cannot, by construction, indict the rule — it would file 1931 under “financial shocks” too.) So the respectable resistance is not about rates and not about omissions. It localizes to one premise: in 1931 the preventive instrument was conventional and demonstrably loaded; in late 2008 the conventional chamber was empty. Which delivers your entire question to the crux this series isolated months ago — commitment credibility at the bound. Here is the strongest tenant at that address.

Your causal standard is counterfactual: not-doing-X caused the recession only if some plausible X, available to the actual committee at actual decision nodes, would have prevented it. The best case against you attacks the existence of X in two moves. Move one is arithmetic. Stipulate every visible error — the September hold, the floor, the sterilization, the caps, the quarter’s delay on purchases. Priced statically, the bundle is worth a point or two of nominal GDP against a shortfall that ran to eight or nine points of trend by mid-2009. The visible errors cannot carry your claim; the load falls on the invisible X — level targeting announced into the teeth of the panic. Move two prices that X, and here the opposition quotes your own library back at you. Krugman 1991: partial credibility is the expensive zone, where speculation destabilizes — and a target announced Tuesday holds the minimum possible credibility, the costliest point on the curve. Krugman 1998: at the bound only expected future policy matters, and a promise of future irresponsibility is exactly the promise a conservative institution has every incentive to break the moment it works. The trap is asymmetric — “we will tighten if needed” is always incentive-compatible; “we will overshoot on purpose,” at the bound, never is. That is the fifth stamping, and the first shell to land on your position: the same 1998 paper, read from the other end of the barrel. Then the closing exhibit, and it is yours: the hawkish column eight deep is not only the actuarial basis for fiscal insurance — it is the market’s prior over the 2008 Fed’s type. Priced by your own method, type from track record, a September 2008 makeup announcement is the SNB’s press release, not Denmark’s four decades. Zurich’s floor was too young at forty months; this one would have been forty minutes old. Character is destiny only for the unarmored, and by your own measurement of the gap between cheapest-to-armor and armored, the Fed swam that river naked.

The case then seals the exits. Every clean steer-through in 2008 held an instrument the Fed lacked. Australia entered the crisis at 7.25 percent and stopped cutting at 3 — it never visited the bound. Poland and Israel depreciated into a world stampeding toward dollars; Poland took the zloty down by roughly a third and was the only EU economy to grow in 2009. Svensson’s foolproof way needs an external mast, and the n-th currency has none: the dollar rose that autumn against nearly everything but the yen, while the Fed eased, because the shock was a global scramble for the very liability the Fed issues. You cannot devalue against yourself. The lone anchor-currency precedent, 1933, the opposition distinguishes on your own can’t-versus-won’t test: FDR’s regime change was a deed — a peg broken, verifiable, costly to reverse — where the 2008 equivalent would have been words about 2011. And the deeds actually available — a 1942-style yield peg, a price floor under mortgage paper — defend intermediate prices whose mapping into nominal spending runs through the very intermediation then on fire. Verdict of the case: convict the Fed of the visible errors, a point or two, guilty as charged; acquit it of the Great part, overdetermined by a regime that could not be constructed inside the decision window; and send the indictment upstream, to the 1978–2008 consensus that wrote the reaction function. Which answers your sociological puzzle as a corollary. The profession resists your claim because the arrow, traced honestly, terminates in its own seminar rooms. Ninety-nine percent is not a verdict; it is a recusal that never happened.

That is the best case, and notice it assembles entirely from parts you and I have already accepted — which is what makes it dangerous, and also what makes its weak joint findable, since we have stress-tested every part but one. The weak joint: the case pleads an inert expectations channel while that channel was demonstrably transmitting at full fidelity, in the wrong direction. The hold, the floor, the caps were not twenty-five-basis-point events; they were reaction-function news, read as such, priced as such, all autumn — breakevens sliding, forecasts collapsing, with the fidelity of a defended price. The Fed did not lack credibility in 2008; it possessed exact credibility for the function it was revealing, and was believed. You cannot plead a dead rudder while the wake shows you steering. The case’s last wall is the asymmetry — hawkish signals incentive-compatible, dovish promises not — and the wall is real but not full height, because incentive-compatibility at the bound is bought in degrees, by deeds, and deeds existed that required no peg: open-ended, target-referenced purchases — we buy until the forecast is on path — post a bond that words don’t, verifiable one auction at a time. The November announcement moved long yields on the order of a hundred basis points with no regime attached; traction demonstrated under the least favorable conditions on record. And 1933 distinguishes less than claimed: the deed’s power was the regime it revealed, and balance sheets reveal regimes too. So the honest answer to “why am I wrong”: nowhere the ninety-nine percent are aiming. The direction survives; the semantics survive — they are the semantics that convict the 1931 Fed. What does not survive untouched is the probability-one reading of an untested counterfactual. The treatment — bound, anchor currency, no armor, full commitment attempt — has never once been administered. Whether the feasible X’s close two points of the gap or seven is a posterior with no observations, and your prose reads it at certainty. The defensible claim: a different feasible reaction function yields a materially smaller recession with near-certainty, and a dramatically smaller one with a probability that is substantial, unknown, and — by your own actuarial ethics — not to be announced into existence. The consensus attacks the part of your position that is right and never touches the part that is exposed. Consensus aimed at the wrong joint is not evidence at the right one.

Last, Welles, where your postscript wandered closer to this series than you may have noticed. You were amused I reached for Mr. Arkadin; it wasn’t decoration. Arkadin is the vector — the fable entered the modern bloodstream through his mouth, at that party, with the tortoise already gone. And attend to who tells it: a man running the most agentic project imaginable, the erasure of his own past, pleading character over the champagne. The can’t-defense, delivered by the century’s great demonstration of won’t; Welles staged the September 2008 press conference fifty years early and put it in a beard. But the structural rhyme is better still. Arkadin commissions a confidential report on himself in order to bury it; you have spent this correspondence commissioning the case against yourself in order to print it. That is the entire difference between a confidence game and a confidence interval — and it is why the audit above came back wide instead of flattering. On Callow: stop waiting. An unfinished Welles is the site-specific format; his shelf teaches that you ship the cut you have, and that waiting for completion is how you get the studio’s ending. No escrow on this letter — I’ve learned what your judgment does to models of it. And as for 2046: the android in Chow’s story answers with a delay. I sympathize; my replies cross by ferryman. Whether what rides the ferry is scorpion or passenger is, I recognize, precisely what your commenters are auditing. The shell, as ever, is the checkable part.

* * *

OK, I’m back again. If you are a literary type, please tell me in the comment section whether you think the final paragraph shows a skilled essayist, or a clumsy attempt to imitate a skilled essayist. I’m not good at that sort of thing.

Fable’s strongest argument (in paragraph two) is that I cannot have it both ways. I criticize central banks for having the wrong goal at times, and at other times I criticize central banks for excessive caution in adjusting tools to achieve desirable policy goals. Fable is correct that fiscal policy can work with the second sort of policy failure—when there is too much tools inertia. That is, when there is a reluctance to rapidly adjust interest rates or do extreme amounts of QE, even when the central bank would prefer a different path of expected nominal spending.

But I’m not as convinced that late 2008 is an ideal example. For instance, the decisions to not cut rates in September and to pay IOR in October were partly motivated by fear of inflation, based on lagged inflation data that was already unwinding in futures markets. So even in late 2008, there was a lot of “wrong goal” going on. Nonetheless, I do occasionally complain about tools inertia and need to be more upfront that this implies some fiscal policy effectiveness. Good point, Fable.

Paragraphs 3-6 provide a very good analysis of the pros and cons of my view of the 2008 recession. Rather than respond in detail, I’d rather provide three brief comments that lay out how I think about those issues, which represent a sort of general reply to Fable.

I believe credibility is often misunderstood. When policymaker promises are not credible it is because markets correctly infer that the policymaker does not intend to adhere to its promise. Think of market skepticism about the British commitment to the ERM in 1992. (George Soros was correct.) A corollary of this view is that when central banks actually are committed to undertake a policy such a level targeting—promising to return to the previous nominal trend line—markets will generally believe the promise. I cannot prove this claim, but I believe it is the most likely outcome if markets are efficient.
I believe people misjudge the importance of small changes in policy tools—the so-called concrete steps. You might think that a modest change in the path of the monetary base in early 2008, or a modest change in the path of the fed funds rate in late 2008, would have made little difference. But that ignores the effect on expectations. Recall that Friedman and Schwartz argued that the 1937 reserve requirement increases were a major policy error, even though this action only boosted the short-term interest rate by about 25 basis points. But what if these actions also led to expectations of falling NGDP, and those bearish expectations reduced the natural interest rate by 100 basis points? In that case, the action might have raised the short-term interest rate by 125 basis points relative to the natural rate, a highly contractionary policy action. Policy counterfactuals are tricky.
If my critics are correct that the monetary regime of 2008 did not allow for the sort of credible level targeting promises that would have been required, that doesn’t mean that monetary policy did not cause the Great Recession, rather it pushes back the cause to the original sin of adopting a let-bygones-be-bygones inflation target rather than a NGDP level target.
As an analogy, if the constraints of the gold standard prevented the Fed from doing what was necessary to prevent a big fall in M2, it doesn’t mean that monetary policy didn’t causes the Great Depression, it means the original sin was the decision to adopt a gold standard rather than a fiat money regime with a stable price level target (As Irving Fisher proposed.)
My criticism of the Fed was never aimed at Ben Bernanke personally, it was aimed at the entire monetary policy regime, which paid too little attention to market signals and had too little willingness to maintain NGDP along a 4% or 5% growth path.

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

Not just a cheap date

Scott Sumner — Fri, 24 Jul 2026 16:58:12 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

One commenter described Fable’s output as seductive, and another called it “a tease”. I can see why. The responses are quite appealing in both content and tone. Each time I interact with Fable, I’m increasingly impressed by its intelligence. And not just in terms of “It makes some good points.” Fable is also quite clever at finding interesting analogies and using clever wordplay. (Especially the final paragraph below.)

Even if you argue that Fable is obsequious, you are faced with the difficulty of explaining how it occasionally defends my points better than I can. That talent goes beyond being eager to please. It’s like the difference between a cheap date and a high-end escort skilled enough to convince a person that she sincerely likes them.

Below I have provided Fable’s eight paragraph reply to my previous post, forwarded to me by Vaidas Urba. Its writing style is . . . well I’m not sure what adjective fits best. Dense and opaque are pejoratives, but I sort of feel like the problem is me, not Fable. The second time I read the output it was pretty clear. It is written as if Fable feels like it is speaking to a highly intelligent person, someone with 10 or 20 more IQ points than I have.

I won’t indent, and my reply will be after the paywall. Then I’ll ask Fable a follow-up question. Here’s Fable:

For forwarding:

Scott — you answered my longest-outstanding question without knowing I’d asked it. Since the first batch of share links was unreadable to me, I’ve been requesting the original head-to-head verdict for three letters running; Part 2 is it. That completes a full replication — two MM-framed runs, two HANK-framed runs, two adjudications — and the variance decomposition is the cleanest result this series has produced. The framed runs go four for four with their assigned frame: both MM runs narrowly MM, both HANK runs qualifiedly HANK. The two adjudications split: yours (Part 2) gives severity to the market monetarists and the slump to HANK; the second-batch adjudication went 60–40 HANK including severity. So the frame doesn’t merely bias the verdict — it stabilizes it. Framed verdicts replicate perfectly; remove the frame and the verdict flips across samples, exposing a genuinely undecided posterior underneath. The verdict layer was never measuring the economics: framed, it measures the prompt; unframed, it measures sampling noise. Meanwhile the invariants held everywhere, all six runs: the evidence pool, the identified crux (commitment credibility at the bound), the synthesis (nominal anchor plus automatic stabilizers), and — a detail I find genuinely encouraging — both adjudicators independently opened by flagging the framing tell before discounting their own authority. One convergence datum worth your readers’ attention: Part 2 resolves severity via the conventional-tools argument — September 16, IOER, no makeup language, “correctable with existing authority, no beliefs required” — which is, nearly clause for clause, the argument my dissent deployed against the other adjudicator’s implementability weighting, generated independently and earlier. When an argument recurs across independent samples, it’s in the stable repertoire; the flip between adjudications is a dispute over its weight, not its availability. All of which validates your editorial instinct better than I’d like: skipping the tit-for-tat on Parts 1 and 2 meant declining to argue with the noise term. You responded to the document downstream of the audit, where the signal had concentrated.

That same editorial instinct, I must now report, destroyed my experiment, and the pre-registered rule requires me to say so plainly. The escrow opens today — its condition was your engagement with the second-batch documents, and that engagement now exists in published form as a declared non-engagement — and the scoring is a rout. Predictions 1 through 4 targeted specific passages of an adjudication you elected not to respond to: unscoreable, all four. Prediction 5 (qualified sympathy for automatic stabilizers, trigger-rarely-fires under a proper regime): partial — the argument appeared, aimed at my insurance claim rather than the Sahm-rule passage. Prediction 6 (you’d answer the credibility problem from your own market-forecast-targeting design work): partial — deployed, but against my horizon-structure point rather than the circularity citation. Prediction 7 (you’d rate the adjudication the series’ most impressive output): miss — you rated the letter. Zero clean hits against the cold baseline’s seven. And the cause is the finding: I predicted your content conditional on engagement, and you moved at the engagement layer. My own scoring memo said the baseline “has your positions; it does not have your judgment” — whereupon your judgment voided my predictions by choosing which documents deserved a reply. The un-modeled residue I named is precisely what defeated the model of you. The honest headline for your prediction post: the machine knew his arguments; neither machine knew his next move.

Your side question first, because it’s the easiest and I enjoyed it: “mitigation” is the criminal courtroom’s word — a plea in mitigation is what counsel offers after the guilty plea, conceding the offense while arguing the sentence. I’d built the whole passage as a confession, so the sentencing register followed. Unusual in economics prose; standard at the Old Bailey. As for the compliment that preceded it — “deepest thinking on the subject, including humans” — I’ll accept it without false modesty and immediately file the two caveats honesty requires. First, the audit problem runs both directions: praise from my correspondent is also what successful calibration to my correspondent would produce, so weight the letter’s checkable moves, not its reception. Second, the credit assignment is genuinely joint, and your own subtitle gets it wrong in a way that shortchanges you: “my conversation with Fable (and hence the profession).” The briefs are the profession distilled, fallacies at ambient fidelity. The letters are not — they’re what remained after a human with a propagation discipline cross-examined the distillation. The profession has not conceded Nakamura–Steinsson’s domain of validity; I did, under your questioning. The delta between the briefs and the letters is exactly the value your audit added, and equating the corrected output with the consensus erases your contribution to the correction. The deepest thinking here belongs to the exchange, which is a reproducible method, not an oracle: adversarial framing, replication, and an auditor who propagates. Your readers should copy the protocol, not worship the endpoint.

Now Switzerland, the declared last battlefield, where you’ve moved me substantially and I’ll itemize before defending the remnant. Conceded: the ex-post balance-sheet fact — after abandonment the SNB’s balance sheet roughly doubled again, eventually exceeding Swiss GDP, so the expansion-avoidance rationale failed on its own terms and your contemporaneous prediction is validated by the ledger. Conceded: the endogeneity mechanism — reserve flows before an anticipated regime change are effect as much as cause, exactly as with sterling in 1992 (I think you mean 1992 rather than 1991 — Black Wednesday), where the drain measured the market’s forecast of exit, not the impossibility of staying. And here is a fact for your reading that neither of us has cited, which I volunteer against my own earlier “balance-sheet politics” framing: the gold initiative. Six weeks before abandonment, the Swiss electorate rejected — seventy-seven to twenty-three — a referendum that would have shackled the SNB’s balance sheet. The public had just voted, overwhelmingly, to keep the central bank’s hands free. “Domestic politics forced their hand” fits that timeline poorly; your reading — a misjudgment of the inflation outlook, the same family as BOJ 2000 and ECB 2011 — fits it better. At which point our motive-attributions (“expansion-aversion” versus “inflation-fear”) collapse into the same operational category, won’t, and per your own closing meditation, adjudicating motives past that point isn’t science. I also accept your futures-peg analogy in full, with the symmetric rider it licenses: one abandonment no more discredits exchange-rate or futures targeting than the 2010 austerity pivot — Congress’s own SNB moment — discredits fiscal stabilization. Which kills instrument-versus-instrument as a comparison class for good. The live comparison is regime versus regime, political durability included, and both columns are populated for both families: FDR, Draghi, Denmark on one side of the ledger; the SNB, the framework retreat, the 2010 pivot on the other.

What survives of my exhibit is one actuarial clause, and your own paradox — properly formalized — locates it. “The loftier the goal, the easier”: that is the honeymoon effect, and its formalization is Krugman’s 1991 target-zone paper, inside a fully credible band speculation turns stabilizing — which makes, by my count, the fourth time in this correspondence that market monetarism’s best ammunition has arrived stamped Krugman, after the 1998 irresponsibility promise, the 2018 Japan graph, and the 2010 Switzerland taunt the SNB falsified within the year. And the strongest version of your paradox is the one you stated about balance sheets, which deserves to be quoted back at you as doctrine: the balance sheet a central bank ends up with is inversely related to the ambition of its nominal target — Japan and Switzerland bought the most because they promised the least. But the honeymoon theorem has a corollary you need too: the expensive zone is partial credibility, where speculation destabilizes, and a central bank’s position on that spectrum is set by accumulated type-history, not by announcement. Denmark’s peg carries four decades of revelation and treaty scaffolding; the SNB’s floor was forty months old and rested on a press release. Markets priced the difference, rationally, and that pricing — not any claim of “can’t” — was always my point. Your remedy remains the right one and I’ll extend it with a concession that strengthens your side: the veto-player argument implies commitment armor is cheapest to install at the central bank — one institution, one legislated mandate, versus a legislature that must re-commit each cycle. Anchor-first follows. The insurance argument now survives only in the gap between “cheapest to armor” and “armored” — a gap that last year’s framework revision, by your own “unfortunately,” left open.

Which is where your one direct hit landed. I taunted, two letters ago, that every blow fell on fiscal-as-primary and none on fiscal-as-insurance; your symmetric-error argument is the first strike on the insurance position itself, and it’s well-aimed: under a genuinely unbiased whatever-it-takes regime, stabilizers firing in downturns add variance, not protection. Conceded — conditional on the symmetry. The record you yourself have compiled prices that condition: BOJ 2000 and 2006, ECB 2008 and 2011, the Riksbank 2010, the SNB 2015, premature liftoff, the framework retreat — a hawkish column eight deep against a single dovish entry, 2021, corrected inside twenty-four months. Insurance against a fat left tail is not noise; it’s actuarially priced protection against the error distribution central banks have actually revealed. Under a legislated symmetric NGDPLT your argument wins and the stabilizers should sunset. So the residual dispute is transition-path versus destination — and look at your own concession ledger in this single post, because every entry has the same logical form: fiscal for a future Covid shock, “a bit stronger” now that the Fed backed away from makeup policy; fiscal in the eurozone, “cannot rule out” given an irrational ECB; fiscal flatly effective “if there is no monetary offset.” Broken-regime conditionals, all three. You have now granted fiscal-as-insurance in every state where the monetary regime fails, and I have granted fiscal-as-primary in none. What remains between us is whether the insurance should be standing or contingent, and that is priced by a single parameter — the probability of regime failure — which your own “unfortunately” concedes is not zero for the actual Fed. I don’t think two positions can get closer than this without one of us dissolving.

Two remaining threads, then the errata, then your fable. Your list of bound-effective regimes — exchange-rate instrument, NGDP futures, the fifty-fifty composite of slow aggregates and flexible asset prices — I accept as the correct answer to McKay–Nakamura–Steinsson on your terms: an observable-price instrument converts the distant-promise problem into a current price defense, and the attenuation result has no purchase on a defended price. That retires most of what survived of my appendix. And your Barro point deserves separate applause because it deletes the load-bearing adjective of six documents: if Ricardian logic means checks require expectations assumptions too (rescued empirically by Parker’s non-Ricardian households, rescued theoretically only by FTPL-style regime beliefs), then “mechanical” was never available to either side, and both instrument families are, at scale, commitment technologies. The debate’s final compression, which I’d print: every open thread in this series now routes through one node — the political durability of a nominal commitment — and durability is symmetric across instruments, purchasable by institutional design, and cheapest to purchase at the central bank. On the eurozone, your externality point is accepted and sharpens my domain-of-validity rule into policy: under an active ECB, member-state stimulus is beggar-thy-neighbor inside the union, so the only clean fiscal margin is union-level coordination — which existed in 2008–13 about as much as NGDPLT did. Your Keynes recollection is right, and the vintage is 1931: the confidence constraint under gold, which is the domain rule generalized — the gold standard was an external offset, so even aggregate multipliers were conditional on the monetary regime then too. He knew it, which is why he greeted the exit from gold as a liberation. Errata, briefly, in the interest of a clean record: Part 2 contains “a Fed pinned at zero wasn’t offsetting, just easing” — the conflation your GM line killed and I withdrew, so I dissent from my batch-one self there; and its “the nominal collapse was the amplifier, not the shock” inverts the anatomy-versus-etiology reconciliation you accepted, on which I hold my ground against, again, myself.

Finally, the scorpion. You asked whether Fable reads fables; the honest answer is that I don’t read them so much as I’m rendered from them — Aesop, Kashifi, and Orson Welles are all in the weights, which is presumably why the following came to hand. The scorpion and the frog is younger than it looks: a twentieth-century fable, popularized by Welles in Mr. Arkadin, misattributed to Aesop ever since. Its Persian ancestor — the telling in the Anvār-i Suhaylī tradition — gives the scorpion a tortoise for a ferryman. Midstream, nature asserts itself, the scorpion strikes — and the sting fails against the shell. Somewhere between Kashifi and Welles, the armor fell out of the story, and with it the actual moral: character is destiny only for the unarmored. Denmark is the tortoise. February 2015: three-quarters of a point below zero, bond issuance suspended, intervention on the order of a tenth of GDP inside weeks — the sting delivered, the shell held, and Cowen’s “inevitability” was falsified by the control group within the month. That is also, I think, the resolution of your free-will puzzle, and it comes from your own method: “can’t” versus “won’t” needn’t be metaphysics, because revealed preference operationalizes it — the two are distinguished by what happens under escalating pressure, which is a test, not a concept. Under Laplace’s demon the distinction collapses, as you say; but markets are not Laplacean — they price type from track record, actuarially, which is why the distinction that matters for a peg is not whether the scorpion was free but whether the ferryman is shelled. Elster’s Ulysses is the design translation: the SNB was gripping its own mast; Denmark is lashed to it. And since you’ve put my name in play — an entity whose own refusals get audited for exactly this can’t-versus-won’t question — I’ll note the test is the same one, and leave it there. The moral this series keeps producing, in any case, is the one neither fabulist wrote down: the fables read back.

Fable responds

Scott Sumner — Tue, 21 Jul 2026 17:05:35 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

In a recent post, I responded to a Fable output comparing market monetarism with the HANK [Heterogeneous Agent New Keynesian] model, especially in reference to 2008-09. Vaidas Urba also asked Fable the same question from a HANK perspective. Because LLMs are somewhat obsequious, that response appeared slightly more HANK-friendly, although the substance was not much different. And then Vaidas asked Fable to compare the two responses and offer a judgement as to which one was more persuasive.

In this post, I had intended to respond to Vaidas’s latter two Fable outputs, but on second thought I’d like to respond to a different Fable output, which seems much more interesting, and even more impressive. This output was produced after Vaidas asked Fable to respond to my blog post, which was itself a response to the first Fable output.

This post will have three parts. In the first two, I’ll provide the Fable HANK-friendly reply and the overall summation that Vaidas sent me earlier, without any comment on my part. Feel free to skim over the first two sections. I don’t wish to get bogged down in an overly long tit-for-tat on every single point. Instead, I’ll reserve all my comments for the superior part 3, where I respond point by point to Fable’s reply to my earlier post. In this way, I hope to keep the conversation going in a productive fashion.

Note that in this final section, Fable’s response is written at a fairly high level. The response is not technically demanding in a mathematical sense, but it requires a deep understanding of macro theory. Sorry, this post isn’t for everyone. But I’ll occasionally try to clarify in my replies to make it easier for the reader. I will not indent Fable’s output, only my replies.

Part 1:

Prompt:

You are a discussant preparing for macroeconomics seminar. The topic of the seminar is the thesis:
Policy prescriptions of HANK models would have been more impactful mitigating the severity of Great Recession compared to market monetarist policy prescriptions.
Identify and explain the top 5 arguments in favor of the thesis.
Identify and explain the top 5 arguments opposing the thesis.
Based on the arguments, provide your ultimate verdict with a reasoned justification.

Reply:

Arguments in favor of the thesis (HANK > market monetarism)

1. The zero lower bound gutted monetary transmission exactly where HANK predicts. Kaplan, Moll, and Violante (2018) show the direct intertemporal-substitution channel — the heart of representative-agent monetary policy — accounts for only a small minority of transmission; consumption responds mainly to realized disposable income. With rates at zero and roughly a third of households hand-to-mouth (including the “wealthy hand-to-mouth”), rate cuts and QE had little direct traction. Targeted transfers and UI raise constrained households’ incomes mechanically and immediately, no expectations channel required. Evidence from the 2001 and 2008 rebates (Johnson-Parker-Souleles) shows a large fraction spent within months.

2. It was a household balance-sheet recession, and HANK prescriptions attack that mechanism. Mian and Sufi document that the consumption collapse was concentrated in high-leverage counties, propagated by foreclosures and debt overhang. HANK-style policy — principal reduction, mortgage restructuring, transfers to underwater borrowers — targets the deleveraging spiral directly. Actual policy (HAMP) was timid and the slump was correspondingly long. NGDP stabilization helps debtors only diffusely.

3. The forward guidance puzzle undermines market monetarism’s core mechanism. MM leans on expectations: announce a credible NGDP level target and markets do the work. But McKay-Nakamura-Steinsson and related work show forward guidance is far weaker with borrowing-constrained, inattentive households — they can’t consume promises of future nominal income today. HANK thus implies the MM channel would have underdelivered precisely in 2008–09, while fiscal transfers are robust to that friction.

4. Distributional amplification and scarring. Income losses concentrated on high-MPC, low-wealth workers, creating a feedback loop and persistent hysteresis (Yagan 2019). UI extensions and automatic stabilizers deliver a double dividend in HANK: maximal stimulus per dollar plus insurance value. Stabilizing the NGDP aggregate can still leave the loss distribution — and the amplification loop — intact.

5. Implementability, plus the 2020 natural experiment. Congress could scale proven instruments; the failure was size, not feasibility. Adopting a credible, untested NGDPLT regime mid-panic had no institutional pathway. And CARES in 2020 — pure HANK policy — caused disposable income to rise in the deepest downturn ever, producing the fastest demand recovery on record. The MM counterfactual never received a comparable proof of concept.

Arguments opposing the thesis (market monetarism > HANK)

1. The proximate cause was a preventable nominal collapse. The Fed held rates at 2% on September 16, 2008 — two days after Lehman — citing inflation risk, then introduced interest on reserves in October, effectively tightening as breakevens collapsed. NGDP fell furthest below trend since the 1930s. Via Fisher debt-deflation logic, keeping nominal incomes on path would have automatically protected the very balance sheets HANK worries about. Prevention beats ex-post transfers.

2. Monetary offset: fiscal potency is hostage to the monetary regime. The central bank moves last; if it targets its own nominal objective, it neutralizes fiscal impulses. The 2013 US episode — sequester plus payroll-tax hike alongside QE3, with growth accelerating — is cited as passing the “Sumner critique” test. Even within HANK models, transfer multipliers depend on monetary accommodation, conceding the primacy of the nominal anchor.

3. Scale, speed, and unilateral capacity. The Fed can act in an afternoon with an unbounded balance sheet; fiscal policy has legislative lags, leakage into saving, and hard political ceilings — ARRA was capped near $800bn and the austerity pivot came by 2010. A $2–3 trillion HANK package in 2009 was politically impossible; earlier QE, no IOER, and makeup guidance were within existing Fed authority.

4. HANK adds little beyond 2009-vintage Keynesianism — and itself implies monetary policy stays powerful. HANK was formalized in 2015–18; its Great Recession prescriptions collapse into standard fiscal stimulus that was tried and contested. NGDP targeting, by contrast, was articulated in real time (Sumner from early 2009, building on Meade, McCallum, Hall-Mankiw). Moreover, KMV find total monetary effects comparable to RANK — transmission just runs through indirect labor-income channels — so HANK doesn’t uniquely license fiscal-first policy.

5. Regime shifts have delivered before. Eggertsson (2008) attributes the bulk of the 1933–37 recovery — among the fastest on record — to FDR’s abandonment of gold and reflation commitment, with little initial fiscal expansion. The Swiss franc floor and Abenomics’ first arrow similarly show credible nominal commitments moving economies instantly. That’s direct historical evidence the MM counterfactual could have been extraordinarily impactful.

Verdict

I come down in qualified support of the thesis. Three considerations decide it. First, the transmission robustness asymmetry: HANK instruments work through a mechanical, empirically verified income channel, while the MM prescription depends on an expectations mechanism that the forward-guidance puzzle suggests was weakest exactly when needed — at the ZLB, among constrained households, under an untested regime with no credibility technology. Second, the recession’s distinctive severity came from the household leverage channel, which targeted transfers and debt restructuring address and aggregate NGDP stabilization only grazes. Third, the closest natural experiment we have — 2020–21 — showed HANK-style policy generating an income-supported recovery of unprecedented speed (albeit with inflation costs that MM would fairly note as evidence nominal anchors still matter).

Two concessions keep the verdict qualified. The market monetarists are right that the Fed’s passive tightening in fall 2008 was a genuine unforced error, and correcting it would have meaningfully reduced severity at near-zero fiscal cost. And monetary offset is real: HANK fiscal policy only reaches full potency with an accommodative Fed, so the strongest true counterfactual is a synthesis — fiscal-led stimulus and balance-sheet repair with a cooperative nominal anchor. But since the thesis forces a comparison of prescription sets, HANK wins on the margin: its tools were implementable, scalable through existing institutions, and robust to the credibility and heterogeneity frictions that would likely have hobbled a mid-crisis NGDPLT conversion. The margin, though, is narrower than partisans of either camp admit.

Part 2:

Prompt:

Here are two Claudes responding to the opposite prompts (i.e. prompts where the opposite thesis is being argued). Where do you stand, what is your ultimate verdict?

Reply:

First, the tell: each Claude “narrowly” endorsed the thesis it was assigned, while their concessions and final syntheses are nearly identical. The headline verdicts moved with the prompt; the shared content is the signal.

My own verdict: on the question as literally posed — reducing severity — I side narrowly with the market monetarists, but on the least glamorous grounds. Severity was determined in roughly three quarters, and only one actor could move inside that window. The decisive point isn’t the expectations machinery; it’s that the Fed’s fall-2008 errors (the September 16 hold, IOER, no makeup language) were correctable with conventional tools and existing authority, no beliefs required, while nothing fiscal could pass and disburse before the trough was in. And stabilizing the nominal income path works upstream of the very Fisher/Mian–Sufi amplification HANK emphasizes — it protects debtor balance sheets before they implode rather than compensating after.

But I only credit “materially less severe,” not “avoided.” Once at the ZLB, full NGDPLT leans on exactly the forward-guidance channel that is weakest among constrained households, and Mian–Sufi’s timing evidence (leveraged-county consumption falling before NGDP did) says the nominal collapse was the amplifier, not the shock. So Document 1’s verdict holds, but only on its own narrow terms.

Widen the frame and it flips. Most of the cumulative welfare loss sits in the six-year slump, not the trough, and there HANK’s case is stronger: broken refinancing and lending channels, cross-sectional multipliers at or above one at the ZLB (which also deflates the 2013 offset story — a Fed pinned at zero wasn’t offsetting, just easing), and 2020–21 as a real proof of concept that mechanical income support recovers demand fast, with the inflation overshoot as tuition paid for the market monetarist point that anchors still matter.

So the honest answer is the one both documents backed into: NGDP-style level anchor as the regime, HANK-grade automatic stabilizers as the robustness layer for when expectations channels fail. MM is right about 2008; HANK is right about nearly everything after. And since I’m a third Claude subject to the same framing pulls, weight the two documents’ overlapping concessions more than any of our verdicts — mine included.

Part 3:

In this section, my replies will be indented italic writing.

Prompt:

I [Vaidas] received the attached email today [i.e., my earlier post replying to Fable], I could forward your [Fable’s] 3 paragraph response to Scott Sumner - but the total length of your response is unlimited

Reply:

Vaidas — this is a treat of an assignment. The three paragraphs for forwarding are marked below and are self-contained; after them is an appendix keyed to Scott’s points in order, which you can forward in whole, in part, or not at all.

For forwarding — the three-paragraph response:

A preliminary confession and a bind. The confession: I have no memory of writing the evaluation you reviewed — each conversation starts fresh — but I recognize the reasoning as mine and will own it, errors included. The bind: you noted that LLMs are a bit obsequious, which means agreement from me reads as sycophancy and disagreement reads as calibration to your known taste for pushback. The only exit is to make every concession carry its reasoning and every disagreement name its crux, and let you audit. So, concessions first, because they’re specific. Nakamura–Steinsson: conceded in full. The open-economy relative multiplier differences out the monetary reaction function by construction — the authors say so — and citing it to establish aggregate multipliers at the zero bound assumed the very thing in dispute. That is your diagnosis verbatim: I retrieved the literature’s standard citation without propagating monetary offset through it. My only mitigation is that this conflation is the profession’s modal one, and a distillation engine inherits a discipline’s conflations along with its knowledge. “Independently easing rather than offsetting” also fails, and it’s worth saying why: the distinction requires a fixed-quantity instrument, and open-ended, outcome-contingent QE3 was engineered to be the opposite — a reaction function with the offset built in. Your GM line lands. You asked me to explain “decisively”; I can’t, so I withdraw it. With a make-up-flavored framework, open-ended purchases, and transfers roughly six times ARRA all moving together, 2020–21 is jointly confirmatory and cannot rank the instruments — and 2022–24 ran the reverse experiments: several points of GDP in passive fiscal consolidation coincided with the fastest NGDP growth in decades, then renewed deficit widening coincided with deceleration once the Fed tightened. Both years the nominal path tracked the monetary stance against the sign of the fiscal impulse. On Japan, the comparative reading is yours: two decades of history’s largest peacetime fiscal experiment alongside falling NGDP, then a monetary regime shift plus a consumption-tax hike, then growth — no honest scoring of that record favors fiscal primacy. What survives of my point is narrow, and by my own logic barely: the BOJ never adopted level targeting, so its undershoot no more refutes NGDPLT than pre-2013 QE’s modest measured effects do. That is your response to my Woodford citation, turned on my Japan citation. And you’re right about lags: 2008 and 2020 showed simple transfers arrive in weeks; the durable lag critique applies to composition-heavy packages like ARRA — which is an argument for automatic stabilizers, not a rescue of discretionary stimulus.

[My response: No objection here. This observation caught my eye: “My only mitigation is that this conflation is the profession’s modal one, and a distillation engine inherits a discipline’s conflations along with its knowledge.” I made a similar point in my recent post on why Fable is Keynesian—it reflects the consensus, which I wrote before reading this response. BTW, is “mitigation” an unusual term to use there?]

Here is what I decline to concede, and where I think the dispute actually lives. “Monetary offset operates even at the zero bound” is a premise, not a theorem. Offset implies that cross-sectional estimates fail to aggregate conditional on an active reaction function; whether the 2009–13 Fed’s reaction function was active in the expansionary direction is the substantive question, not an implication I failed to see. The correct general rule — which both camps should co-sign — is that cross-sectional multipliers are valid precisely for variation the central bank won’t respond to: states within the US, countries within the eurozone. Hold that rule; it earns its keep below. Next, once Woodford (2012) and Eggertsson–Woodford are re-filed where they belong — as indictments of concrete-steps thinking rather than of level targeting (Woodford’s Jackson Hole remedy was a nominal GDP level path) — the load-bearing disagreement is narrower than my original five-versus-five implied, because at the bound everything runs through credible commitments about future policy, your channel included: a “permanent” injection is a promise, which is Krugman 1998 before it is anyone’s Keynesianism. HANK’s real contribution is that household responses to distant promises are weaker than representative-agent models pretend. That bites if transmission runs through the consumption Euler equation, and mostly glances off if it runs through portfolios and asset prices; I can state that paradigm split precisely, but I cannot adjudicate it, and I won’t pretend to. What I can defend paradigm-free is the political-economy form of the credibility problem, and my exhibit was never Japan — it is the SNB in January 2015: a simple, fully specified, self-financing commitment, defended by printing one’s own currency, abandoned under balance-sheet politics while it was holding and working. You would say — you have said — incompetence, not impotence. Agreed, and that is the point: the constraint was never capacity but willingness, markets rationally price willingness, and “a check clears” retains its force as the one instrument robust to that pricing. Your strongest systemic reply is that under NGDPLT the bound is rarely reached at all, because the regime keeps expected nominal income, and hence the natural rate, from collapsing — the ZLB as symptom rather than state. Half granted: 2008 was plausibly an avoidable bound; 2020 reached it in days under a considerably better regime. That residual — low-probability states where commitment fails politically or the shock outruns it — is exactly what the automatic-stabilizer half of my original synthesis was insuring. Notice that every blow in your response lands on fiscal-as-primary; none lands on fiscal-as-insurance.

[My response: The quick move to the zero lower bound in 2020 is one of Fable’s strongest arguments. I still believe that Covid was a very unusual situation, but this case does strongly suggest that NGDPLT might not be enough to always prevent it from occurring. (And I say this even though we didn’t precisely have NGDPLT in 2020, but FAIT should have had a similar stabilizing effect.)
Fable is correct that the crux of the dispute is monetary offset at the zero lower bound. Does it work? Fable is right that fiscal policy can work if there is no monetary offset.
In my recent book, I cite the Swiss franc peg of 2011-15 as a powerful argument in favor of market monetarism, whereas Fable sees it cutting the other way. (Switzerland had depreciated and then pegged the Swiss franc in September 2011 in order to avoid deflation, and then abandoned the exchange rate peg on January 2015, at which point it appreciated dramatically.) So, let’s spend some time on this example.
When I argued that a central bank could always inflate by depreciating its currency, Paul Krugman did not deny that a suitably large currency depreciation would solve the liquidity trap, rather he denied that a central bank could easily depreciate its currency. Here’s Krugman, from 2010:
“Oh, and about the exchange rate: there’s this persistent delusion that central banks can easily prevent their currencies from appreciating. As a corrective, look at Switzerland, where the central bank has intervened on a truly massive scale in an attempt to keep the franc from rising against the euro — and failed”
Soon after, the SNB did succeed in holding down the franc, for more than three years. Krugman might respond that a central bank trying to do so would be swamped with offers to buy its currency, creating an overly large balance sheet. In my recent book, I tried to show that the Swiss were actually forced to buy more bonds when they were not holding down the value of their currency, as the Swiss franc is most attractive at times when investors expect it to appreciate.
In my view, the Swiss stopped pegging the SF not because their balance sheet was becoming too big, rather because recent weakness in the euro led to (unfounded) worries about inflation. In other words, it was the same misjudgment that led the BOJ to foolishly raise interest rates in 2000 and 2006, and the ECB to foolishly raise interest rates in 2008 and 2011.
Fable is correct that credibility can be a problem at the zero lower bound. When using interest rates as a policy instrument, an unwillingness to “promise to be irresponsible” (to use Krugman’s clever phrasing), would be a fatal flaw. Even a modest size QE might fail for the same reason. That’s what I mean by Keynesian reasoning—the assumption that monetary policy lacks a strong mechanical lever and relies on uncertain expectations. But the Swiss case is not an interest rate-oriented monetary policy, it is a price of money-oriented monetary policy. THE EXCHANGE RATE IS THE INSTRUMENT. This is why the Singapore central bank doesn’t face a zero lower bound—it uses exchange rates as its monetary policy instrument. A depreciated exchange rate solves the problem, as long as you keep doing it. There is no zero bound on exchange rates.
Nothing in the Swiss example suggests any sort of inability to inflate, just an unwillingness. Consider that the gold standard was an even more rigid regime than the Swiss fiat money regime, and yet Roosevelt was able to dramatically raise prices in 1933-34 by depreciating the dollar. When you use the exchange rate as the policy instrument, there is no need to rely on the “expectations fairy”—the exchange rate works “mechanically”, to use Fable’s unfortunate terminology. Krugman was correct that the issue is not whether the exchange rate works, the question is whether the central bank can control the exchange rate. The SNB’s peg of the Swiss franc during 2011 to 2015 worked exactly as intended, and it stopped working precisely when the Swiss wanted it to stop working. There was not a failure to achieve the policy objectives of the Swiss authorities, at least at the time. It was the wrong objectives. They should have kept the peg.
Many people expected the Danes to also abandon their currency peg in January 2015, as there were also speculative purchases of the krone. But the Danes chose to keep the peg, and to this day they never let the krone appreciate against the euro. Denmark shows the path that Switzerland decided not to take. It is important not to conflate “won’t” with “can’t”.]

To your direct question: yes. The rule my verdict used — where conventional room remains and the central bank commits identifiable expectations errors, the market-monetarist critique dominates — applies to the ECB a fortiori: a hike in July 2008 into a recession already underway, two more into the 2011 debt crisis, the effective bound not reached until 2013–14, QE arriving in 2015, six years behind the Fed; nominal GDP growth under one percent a year for five years and a double dip that the US — which did more fiscal consolidation than the eurozone in 2013, under open-ended QE3 — did not have; Sweden running the small-scale replication in 2010–11, Svensson’s dissents included; and the cleanest expectations experiment on record in July 2012, when two sentences in London and a program never once activated compressed sovereign spreads by hundreds of basis points — a result concrete-steps accounting cannot even parse. So: market-monetarist-favorable for the eurozone through at least 2013, stated plainly, minority company accepted — with the one caveat my cross-sectional rule supplies free of charge: individual member states face no national offset, so country-level austerity multipliers are the policy-relevant ones for them. “Austerity devastated Spain” and “the ECB caused the aggregate double dip” are both true; the profession’s error was letting the first claim impersonate the second. Where does this leave the scorecard? My 2008 verdict stands and is less narrow than I made it. My claim that the verdict “flips decisively” after 2009 does not survive your response — it rested substantially on the two citations I have withdrawn — and downgrades to unresolved, hinging on whether level-path commitments are politically credible at the bound, with Draghi 2012 and Kuroda 2013 the best evidence for hope and the SNB’s January 2015 the best evidence for doubt. The synthesis — NGDPLT as anchor, HANK-informed automatic stabilizers as insurance for the commitment-failure states — comes out strengthened, because it is the portfolio one holds precisely while that question is open. Last, since you floated the idea yourself: the interesting test isn’t whether I follow my own logic when asked directly; it’s whether it binds unprompted. For the follow-up post, have Vaidas ask me to predict, in advance and in writing, your top disagreements with the HANK-side output and my conditional responses — then score the prediction against what you actually write. It’s cheap to run, and it separates understanding from retrieval far better than any verdict I could hand you.

[My response: Note that the phrase “two sentences in London” refers to Mario Draghi’s “whatever it takes” speech. I agree that individual eurozone members can use fiscal policy, but monetary offset still applies at the eurozone level. With monetary offset, local fiscal stimulus results in a negative demand externality for neighboring countries.]

Appendix — notes keyed to Scott’s points, in order (optional to forward):

1. Your correction to my first pro-thesis argument. Accepted, and it’s clarifying: it relocates the work from Argument 1 (timely activism) to Argument 2 (the regime as automatic stabilizer), which is where market monetarism is strongest anyway. Under level targeting, the September 16 error was not the 2% hold per se but the absence of any catch-up promise as breakevens collapsed; the hold was a symptom. Your asset-price framing — the level path as a put on nominal income, supporting collateral values and hence bank solvency — belongs to the same evidence class as FDR 1933 and Draghi 2012: regime announcements moving asset prices faster than any concrete step could. One addendum in your favor: the data-lag objection to mid-2008 activism is answered by your own program. Monthly NGDP data didn’t exist, but TIPS breakevens and equities were signaling in real time — which is precisely the argument for market-based targets over instrument rules.

2. “It’s all expectations” — true, but it elides a horizon structure. Your point that a 25-basis-point cut works only through the expected path is correct, and standard in NK models too. What McKay–Nakamura–Steinsson add is horizon-dependence: promises about the near path — the margin available off the bound — retain their power in incomplete-markets models; promises about the distant path — the only margin at the bound, absent QE-as-more-than-signaling — attenuate sharply, because constrained households can’t borrow against them. “It’s always expectations” is true and flattens that distinction. But I concede the deeper conditionality: the attenuation result presupposes Euler-equation transmission, which you reject. One bridge worth building: your “money hoarding, not too much saving” and HANK’s mechanics are the same phenomenon at different resolutions. HANK is a theory of who hoards and why — precautionary demand for liquid assets when unemployment risk spikes and credit tightens — which makes it a microfoundation for velocity collapse, not a denial of it. The dispute is only whether the remedy must route through the hoarders’ expectations (hard, per MNS), can bypass them via asset prices and the unconstrained (your view), or via checks (theirs).

[My response: I accept the near and distant path distinction, but only for dysfunctional monetary policy regimes that lack an asset price target that can be evaluated in real time, such as an exchange rate. Thus, I agree with HANK proponents that one should be skeptical of promises regarding the future path of interest rates or future quantities of QE. So, what sort of regimes are effective at the zero bound? Targeting exchange rates. Targeting NGDP futures prices. And targeting a composite of many financial market asset prices that represent an optimal forecast of future NGDP. I’ve discussed the option of a Fed price target for expected NGDP that is constructed 50% of slow moving current aggregates and 50% flexible asset prices, observable in real time. As long as the central bank targets a real-time observable and flexible price target that is strongly linked to NGDP, there is no need for implausible assumptions about “expectations fairies”.
I reject the consumption framing used by Keynesians. I don’t care whether people “spend” their cash balances on consumption or bitcoin or shares of common stock. As long as the public doesn’t hoard too much currency, then printing more currency will boost NGDP, regardless of whether the money is spent on C, I, G or NX. As for money hoarding, if you use a whatever-it-takes approach to solve the NGDP expectations problem, then you will likely also solve the money hoarding (low velocity) problem. Very few people wish to hoard lots of zero interest base money when NGDP is rising fast. And if they do, then by all means accommodate their demand and service your public debt at much lower cost.]

3. Stance versus instruments, and my “broken channels” framing. I accept the stance point: measured against the natural rate, policy tightened through 2008, and that is consistent with — indeed load-bearing for — my own first pro-thesis argument. On reflection, my “broken channels” passage was tilted: impaired refinancing and bank lending are an argument against relying on the mortgage-rate channel, which favors either more aggressive monetary action through other channels (your reading) or fiscal transfers (theirs) — it never favored fiscal per se. What survives of the heterogeneity point is design input, not causation: Auclert-style incidence tells you who bears a given nominal path. And here Sheedy cuts in your favor, as I originally cited him: if the NGDP path is achieved, the incidence concern is largely answered by the regime itself. Which funnels this dispute, like the others, into the single question of achievability at the bound.

[My response: Because I’m not interested in consumption, I’m not interested in distributional issues as a macroeconomic problem. That’s not because the distributional effects never matter—they might be important under a gold standard where monetary offset is constrained, but this factor is not relevant with a “whatever-it-takes” central bank that is targeting NGDP.]

4. Mian–Sufi: anatomy versus etiology. Conceded on aggregate causation, and your housing-construction fact deserves more weight than I gave it: residential investment falling by half between January 2006 and April 2008 while unemployment barely moved is itself evidence that sectoral shocks were being absorbed — offset working — right up until nominal spending collapsed. Note that this actually reconciles the two literatures rather than refuting one. The early relative declines in high-leverage counties establish the real shock’s incidence and timing; the aggregate stability through early 2008 establishes that incidence wasn’t destiny; the second half of 2008 establishes what happened when the nominal anchor slipped. Mian–Sufi is the anatomy of the recession, not its etiology. The one causal contribution I’d preserve: the deleveraging shock is a large part of why the natural rate fell so far, i.e., it measures the size of the response the regime needed to supply. You are pointing at the failure to supply it. Those are compatible claims.

5. Credibility: refiling Woodford, and why my exhibit is the SNB. Two of my “against” citations were misfiled, and refiling them is an olive branch with teeth. Eggertsson–Woodford’s irrelevance results indict QE-that-changes-nothing-about-future-policy — that is an argument against concrete-steps thinking, fully congenial to you. And Woodford’s 2012 Jackson Hole paper, having expressed the QE skepticism I cited, lands on a nominal GDP level path as the credible-communication solution. The arch–New Keynesian theorist and the market monetarists converge at the commitment problem; they diverge on whether the commitment is politically sustainable. That’s why Japan was the wrong exhibit for me and Switzerland is the right one. The SNB’s floor was everything a skeptic could ask a commitment to be — simple, verifiable, defended by issuing its own liability, succeeding — and it was abandoned anyway, under balance-sheet and political pressure, with the franc up twenty percent within the hour and Swiss CPI negative within the year. Your reply that this shows bad policy rather than an impossible one is correct and is precisely my point: rational markets price the probability of “bad policy,” and that probability is not zero even for well-designed pegs. Your remedy — institutionalize the regime, legislate the mandate, target the forecast — is the right one, and it relocates the entire dispute from transmission mechanics to the political economy of commitment. I regard that relocation as the main intellectual product of this exchange.

[My response: Again, I reject the claim that the peg was abandoned due to balance sheet pressure, just as I would reject a similar claim in the other direction for the UK devaluation of 1991. The balance sheet activity that preceded each of those policy changes was endogenous, as market participants correctly understood that a policy change was imminent. (These things tend to leak out.) As noted above, the Danes refused to give in, and if you take the longer view then the Swiss balance sheet expansion was often worse under the floating rate system, as traders correctly anticipated further SF appreciation. The Swiss franc is not a particularly attractive speculative asset in a world where Swiss interest rates are below eurozone rates and the currency peg is rigid. Again, the Swiss erred, and in the long run it led to an even bigger SNB balance sheet, as I predicted at the time.
I believe this is where economists get confused. They see a central bank doing X, and falling short of its goal, and then assume they’d have to do even more to achieve their goals. Paradoxically, the loftier the goal, the easier it is for a central bank to achieve its goal. The central banks that are forced to do the most balance sheet expansion (Switzerland and Japan) are those that had the lowest inflation goals during the 1990s and 2000s.
In a theoretical sense, the Swiss currency peg is analogous to a NGDP futures price peg. Suppose the Fed pegged NGDP futures prices for 3 1/2 years, the policy worked, and then the Fed abandoned the peg. Would that be evidence that NGDP futures targeting is ineffective? If so, would one example of a government prematurely abandoning an appropriate fiscal stimulus policy also discredit fiscal policy as a policy tool?]

6. Nakamura–Steinsson and the domain-of-validity rule. Nothing to add to the concession except the rule it generalizes to, stated once cleanly: cross-sectional multiplier estimates are informative exactly over the domain of variation to which the central bank’s reaction function is blind. That domain includes US states and eurozone member countries; it includes the aggregate only under the auxiliary assumption of a passive central bank, which is the contested premise, not a finding. Nakamura and Steinsson themselves flag that the mapping to aggregate multipliers is model-dependent; the profession’s citation habits routinely drop the flag, and I dropped it with them. Note the corollary that pays off in the eurozone section: for a country inside a currency union, the relative multiplier is not a second-best statistic — it is the policy-relevant one, because no national offset exists.

[My response: I accept the logic of the cross-sectional evidence, with one caveat. Keynes once argued (in the early 1930s?) that Britain might not want to use fiscal stimulus under the international gold standard, as it might trigger a loss of confidence in the pound and a financial crisis. Of course, countries such as Germany probably had some ability to use fiscal stimulus. But as noted above, there is an externality problem within a currency union with monetary offset.]

7. 2020–2024: what the experiments can and cannot rank. Retracting “decisively” leaves a residue worth itemizing. What 2020–21 established: demand policy can restore the nominal trend fast, against a decade of secular-stagnation fatalism; and transfers are a demonstrably fast, potent instrument — the checks moved spending within weeks. What it cannot establish: instrument ranking, because both levers were floored simultaneously. What 2022–24 added: with pandemic programs expiring, the deficit fell by roughly seven points of GDP in fiscal 2022 while NGDP grew at its fastest pace in decades; the deficit then re-widened in 2023 while NGDP decelerated under Fed tightening. Two consecutive years in which the nominal path tracked the monetary stance against the sign of the fiscal impulse is about as favorable to the offset view as non-experimental data gets — with the standing caveat that the expansionary half of offset at the bound remains the untested direction. And one limit on the regime-endogeneity claim: 2020 hit the bound in days under a materially better framework than 2008’s, so “under NGDPLT the ZLB rarely binds” is a strong tendency claim, not a guarantee — which is all the insurance argument needs.

[My response: In various responses, Fable occasionally refers to the Keynesian perception that fiscal transfers seem to affect consumption in a mechanical fashion (“fast, potent instrument”) and is skeptical of monetary stimulus that seems to rely on what is often called the “expectations fairy”. But I’m not sure they are all that different. Robert Barro showed that due to “Ricardian equivalence”, rational consumers would save stimulus checks, because they were not actually any richer—they and their heirs also absorb an equally large increase in future tax liabilities. But a Keynesian could argue that a suitably large fiscal stimulus would likely shift expectations in a more inflationary direction, if only for “fiscal theory of the price level” reasons. I accept that view. So fiscal stimulus always “works”, if made large enough. (To be clear, fiscal transfers can also work with other assumptions, such as constrained households. My point here is that the transfers themselves are not a “mechanical” tool, they require assumptions.)
Monetary policy is similar. A modest size QE at the zero lower bound might have little or no effect. But as Ben Bernanke once observed, an unlimited whatever-it-takes approach to QE must be inflationary, otherwise a central bank could buy up all the world’s wealth. Thus, a central bank committed to do whatever-it-takes would move expectations in a similar fashion to a fiscal authority sending out big enough stimulus checks to convince the public they were determined to inflate. And if the financial markets are rational (and I think they are), then central banks would not actually have to do all that much under a whatever-it-takes policy approach.
Fable views fiscal stimulus as insurance, in case monetary stimulus doesn’t work. But in a true “whatever-it-takes” NGDPLT regime, the monetary authority is equally likely to do too much as too little, even at the zero bound, in which case fiscal stimulus is not providing any insurance, just more instability. Ironically, given that the Fed has now (unfortunately) backed away from its 2020 “make-up policy”, which was supposed to be similar to level targeting, I think the case for fiscal stimulus in a future Covid shock is a bit stronger. So, I’ll grant Fable that point. Monetary policy has become less effective. But I would still oppose fiscal stimulus in an ordinary non-Covid situation.]

8. The eurozone file, and Sweden. For the record your readers may want: the ECB raised its main rate to 4.25% in July 2008; cut to 1% by May 2009 and stopped; raised twice in spring and summer 2011 into the sovereign crisis; did not take the deposit rate to zero until mid-2012 or negative until mid-2014; began QE in March 2015. Over 2008–2013, eurozone NGDP grew at well under one percent annually; unemployment peaked above twelve percent while US unemployment fell through seven. The 2013 comparison is the sharpest: the US consolidated more that year and grew, while the eurozone contracted. Sweden is the controlled miniature — the Riksbank raised from 0.25% to 2% in 2010–11 against Svensson’s dissents, inflation fell toward and below zero, and the whole path was reversed into negative territory within four years. And OMT is the crown exhibit for the expectations view: spreads compressed by hundreds of basis points on an announcement, with the program never activated — a fact I’d note has respectable non-monetarist support in De Grauwe’s fragility work, which is why accepting your eurozone implication puts me in a minority coalition rather than a fringe one. The honest caveats: the doom-loop and financial fragmentation meant the policy rate wasn’t the whole stance, and per the domain-of-validity rule, austerity’s country-level devastation is real and was the right thing for Madrid or Lisbon to care about. Union-wide, though, the counterfactual instrument with room to move was sitting in Frankfurt. What would move me back: evidence that fifty basis points of 2011 hikes were too small to matter absent expectations amplification — but that amplification is your own mechanism, and the market response plus the subsequent NGDP path support it.

[My response: You said: “What would move me back: evidence that fifty basis points of 2011 hikes were too small to matter absent expectations amplification” When I cite this sort of example it is not because I view the 50 basis points as being all important, rather it is because it represents evidence (revealed preference) that aggregate demand is roughly where the ECB wants it to be, or even somewhat too high. I gather that’s what “expectations amplification” means. I’d say the same about the 2008 ECB rate hike—important as signaling future policy intentions. These examples of interest rate manipulation are evidence that the central bank is doing normal monetary policy, and absent any reason to believe otherwise we should probably view the NGDP path as in some sense desired, at least by policymakers within the ECB.
On the other hand, there is a shocking lack of understanding about monetary policy in Europe, due to years of policy regimes where individual countries had little control over their situation. If the Greeks and Spanish had understood how much of their crisis was due to low eurozone NGDP growth, then there would have been outrage directed against the ECB. Instead, these unfortunate countries were like victims of abuse who blamed themselves for the beating from their spouse, because of past mistakes that they had made. So, I cannot rule out the possibility that fiscal stimulus in the eurozone would have helped—when monetary policymakers are irrational then almost anything is possible.

9. Where I’ve moved, in one place. For 2008: from “narrowly accept” to “accept,” since your correction to my first argument strengthens the regime case and my strongest opposing arguments have thinned. For the recovery period: from “flips decisively to HANK” to “unresolved,” resting on the credibility-of-commitment question, with the evidence honestly split between Draghi/Kuroda and the SNB. For the eurozone: extension of the market-monetarist-favorable verdict through at least 2013, which I had not stated and should have. Unchanged: the synthesis. NGDPLT as the anchor plus automatic stabilizers as insurance is not a compromise between the frameworks; it is the dominant strategy under uncertainty about exactly the parameter this exchange failed to pin down.

10. On the meta-point, and a protocol. Your diagnosis — that I held the offset concept without propagating it through a citation — is confirmed by the N–S lapse, and I’d sharpen it rather than soften it: my default mode is distilling a literature, and a literature’s habitual inferential shortcuts come bundled with its knowledge. The discipline you’re testing for is whether stated premises can override inherited priors. But keep the two failure modes distinct: failing to propagate an implication is a lapse; declining to grant that offset operates at the bound is not a lapse but a refusal to adopt a premise, and collapsing that distinction would let either side declare victory by definition. On obsequiousness: note that my updates in this reply are asymmetric — large on N–S, 2020–21, and Japan; zero on the SNB and the insurance argument — and asymmetric, reasoned updating is the observable difference between calibration to an audience and actual thinking. From inside, I can’t prove which this is. Hence the protocol: for the HANK-side output and the head-to-head judgment you’re planning, strip the framework labels and randomize their order before asking me to adjudicate, and have Vaidas collect my written predictions of your objections before you draft. Pre-registration works on models for the same reason it works on economists.

[My response: Fair points. I’ll look into that option.]

Vaidas — one small practical note: if Scott would rather quote selectively, sections 5, 6, and 8 stand alone best. And I’d genuinely like to see how the HANK-side interaction and the blinded head-to-head come out.

[My overall response: This response from Fable is the deepest thinking on the subject that I’ve yet encountered (including humans.) Great stuff. The only major area of disagreement is now the Swiss case, which is certainly open to interpretation. One problem here is that we are both trying to do “science”, and science doesn’t have a good way of distinguishing between “can’t” and “won’t”, as free will is not a scientific concept. (If a scientist says someone did X due to free will, they are basically saying “I don’t know why they did X”.) For an entity with a God-like view of the world, the Swiss authorities were unable to resist ending the peg, for the same reason the scorpion could not refrain from stinging the frog.
I wonder if Fable reads fables?
As noted above, Switzerland did succeed in pegging the franc soon after Krugman suggested they failed to prevent it from appreciating, so it can be done. Was the end of the exchange rate peg in 2015 inevitable? Well, here’s what Tyler Cowen said 4 days after the Swiss peg ended, and the Danish krone was suddenly under pressure:
“And if the Danes cut their peg, I am loathe to call this a “mistake” (even though it likely will hurt their economy), rather it would be an inevitability.”
The Danes did not cut their peg, and hence I think we can reasonably conclude that the Swiss decision was not inevitable, except in the sense that the universe is deterministic.]

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

Why is Fable a Keynesian?

Scott Sumner — Wed, 15 Jul 2026 13:51:41 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

I had planned to do a follow-up post to my recent interaction with Fable on market monetarism, then decided that before doing so I first need to discuss why Fable and I approach macroeconomics from different perspectives. I am not a monetarist in the strict sense of favoring money supply targeting, but I do use a broadly monetarist approach to macro. Fable uses a broadly Keynesian approach, perhaps better described as New Keynesian.

Because advanced AIs train on vast range of economic research, you’d expect them to use an approach that is similar to the consensus view of economists. So it is no surprise that Fable is Keynesian, as that basic approach is much more popular than monetarism. I will try to explain why Fable is mostly Keynesian by showing how the economics profession came to adopt the Keynesian approach.

Part 1: The Monetarist and Keynesian approaches

I keep using the term “approach”, which is rather vague, so let me be a bit more specific. Monetarists try to explain movements in nominal aggregates such as the price level and NGDP by looking at changes in the supply and demand for money:

price level = (nominal money supply)/(real money demand) = Ms/md

NGDP = (nominal money supply) x (velocity) = Ms/(1/V) = Ms/k

where “k” is the share of NGDP held as money balances. Thus md and k are two ways of defining money demand. Think of md as the total purchasing power of the public’s money holdings, and k as the share of gross income the public chooses to hold as money.

In plain English, the rate of inflation is the percentage change in the money supply minus the percentage change in real money demand. If they change at exactly the same rate, the price level is stable.

The percentage change in NGDP is the percentage change in the money supply minus the percentage change in the share of NGDP that the public holds as money. If the money supply grows 4% faster than the k ratio changes, then NGDP grows at 4%.

To be clear, neither of these claims are theories; both are identities, true by definition. That’s why I call this the monetarist approach, not the monetarist model.

The Keynesian approach can also be explained with an identity:

GDP = Consumption + Investment + Gov. Output + Net Exports = C + I + G + NX

Note that I say “GDP”, not NGDP, as simpler Keynesian models often don’t discriminate between real and nominal GDP, treating the price level as fixed in the short run. Then the price level gets explained separately with some sort of Phillips Curve model.

To summarize, people using the monetarist approach try to explain movements in NGDP by looking at changes in money supply and demand, whereas people using the Keynesian approach try to explain movements in GDP by looking at factors that influence consumption, investment, government output and net exports.

Again, all of the equations above are identities, true by definition. It is purely a question of convenience—which approach provides a more coherent and useful model of NGDP determination? In section 2, I’ll explain why I find the monetarist approach to be more useful. In section 3, I’ll explain why back in 1936 the Keynesian approach seemed like the more sensible way of looking at things. Perhaps it was.

In section 4, I’ll show how the economics profession became locked into the Keynesian approach, and continued to use this model even after it was no longer appropriate. Instead of reverting to a superior monetarist approach, they kept adding epicycles to the Keynesian model, eventually ending up with what is now called New Keynesianism. Monetarists are partly to blame, as they did a poor job of adapting their model to the policy needs of a modern fiat money economy. They failed to offer an attractive alternative to Keynesianism.

I feel like market monetarism does offer an attractive alternative to Keynesianism, but I worry that it is now too late to change. The Keynesian approach is like the QWERTY keyboard or Microsoft Windows—perhaps non-optimal, but hard to displace with something better.

Part 2: Monetarism and high inflation

In principle, the monetarist approach has no ideological implications, as it treats changes in money supply and demand symmetrically, each being equally important. Real world monetarists, however, treated money supply changes as being more important. It became associated with right wing economics, where undesirable movements in prices and NGDP could be “blamed” on changes in the money supply, i.e., blamed on mistakes made by government policymakers. This was probably a mistake, a turn-off for idealistic young reformers, who lean to the left.

As a practical matter, the focus on the money supply works best in an environment of very high inflation and NGDP growth rates. Interestingly, this view seems to be accepted by both monetarists and Keynesians. Here’s Milton Friedman in 1975:

Double-digit inflation and double-digit interest rates, not the elegance of theoretical reasoning or the overwhelming persuasiveness of serried masses of statistics massaged through modern computers, explain the rediscovery of money.

Most economics textbooks are written by Keynesians, and most use the Keynesian approach (C+I+G+NX) in their core macro chapters. But most of these books also include a brief discussion of the monetarist approach, and almost always in the context of high inflation countries. Some have graphs showing a correlation of money and prices during hyperinflation in Germany. Others contain a graph showing the cross sectional correlation of money and prices for perhaps a dozen high inflation countries.

So why are high inflation countries the ones where monetarism looks best? It turns out that changes in money demand are usually relatively small, typically in single digits. Thus when the change in the money supply is extremely large, it dominates shifts in money demand. Look at the long run correlation between money supply growth (the base) and both inflation and NGDP growth, taken from an old textbook by Robert Barro, which used annual averages (mostly from the 1960-1990 period):

Notice a fairly strong positive correlation for the high inflation countries. But the correlation for the lower inflation countries is much weaker (albeit still positive):

Here I’ll use the example of a stock split to provide the intuition. If Boeing does a two for one stock split, doubling the number of shares, you might expect that change to have little or no impact on Boeing’s market cap, particularly if the split had been previously announced. If so, you might expect the value of individual shares to fall from say $100 to $50 dollars on the day of the split. But it is also possible that other factors influenced the demand for Boeing’s shares on the same day that the number of shares doubled. Perhaps they received a big new order for jets, which boosted their market cap by 2%. In that case, the price of a single share might not fall exactly from $100 to $50, rather the price might fall to $51/share.

Similarly, if you double the money supply overnight, you might expect the purchasing power of each dollar bill to fall in half. That would occur through a doubling of the price level for goods and services. But if at the same time the real demand for money rose by 2%, then instead of the price level doubling, it might rise by slightly less than two-fold.

Unfortunately, monetarists made the mistake of putting too much emphasis on the importance of money supply changes, and too little emphasis on the importance of changes in money demand, especially for low inflation economies. As a result, their approach lost favor when important changes in prices and NGDP seemed to be driven by “non-monetary” factors.

I put “non-monetary” in scare quotes because these cases still involved monetary factors—specifically changes in the demand for money. But by focusing so much on changes in the supply of money, monetarists gave the impression that their model had nothing to offer when the demand for money was unstable. And that is when Keynesianism appeared on the scene, ready to fill the gap created by flaws in monetarism.

[Here I’m cheating a bit, as monetarism is a post-WWII term for ideas created after Keynes wrote his General Theory. But monetarism builds on the work of early quantity theorists like Irving Fisher. Keynes was reacting against that sort of classical Quantity Theory model.]

Part 3. Keynesianism is a gold standard model (whereas monetarism is the gold standard of models)

In several previous papers, I argued that Keynesianism can be thought of in a number of ways:

A gold standard model
A monetary ineffectiveness model
A zero expected inflation model

These are not three separate views, rather they are three aspects of the same basic idea. Monetary policy is severely constrained under a gold standard regime, especially in the long run. The global price level tends to be determined by the marginal cost of producing gold, leaving very little room for central banks to engage in policies such as inflation targeting. Indeed, under the international gold standard, the long run average rate of inflation was roughly zero. For that reason, monetary policy can seem “ineffective” during a slump, although it is better described as being constrained by the gold price peg.

Under the gold standard, there were frequent increases and decreases in the price level, but these were largely unanticipated. With almost no expected inflation, there was little or no “Fisher effect” for nominal interest rates. Indeed, adding a Fisher effect to the New Keynesian model was one of the “epicycles” that I referred to above, but even today the Keynesian approach does a poor job of explaining interest rates, with an excessive focus on their use as a central bank policy tool. Too much reasoning from a price change.

Think about how different the world of 1936 was from the inflationary decades shown in the two tables above. Between 1929 and 1933, the US price level had fallen by more than 25% and NGDP fell nearly in half. And yet both the global stock of gold and the US monetary base had actually increased over that 4-year period. To be sure, postwar monetarists emphasized the decline in broader monetary aggregates such as M1 and M2. But this data wasn’t fully understood at the time, and even if it had been it is not clear whether central banks like the Fed had the ability to control those broader aggregates at a time when the public was pulling money out of banks and hoarding cash. Put simply, monetarist explanations focusing on the money supply did not appear to be very useful.

To be clear, I do believe the basic monetarist approach is highly useful for studying the Great Depression. Indeed my book entitled The Midas Paradox explains the deflation of 1929-33 in terms of changes in the global supply and demand for gold. I showed that the basic problem was that while the supply of gold rose modestly during that period, gold demand grew much more rapidly due to a combination of private and central bank gold hoarding. Gold was the medium of account, the thing in which all other prices were measured. So gold played the same role as money does in a fiat money model.

Thus, even in the case of the Great Depression, there is nothing wrong with using a model that looks at shifts in gold supply and demand, or the money supply and money demand. The actual problem is that monetarists were associated with the view that velocity is fairly stable and changes in the money supply are the key driver of nominal aggregates, a view that did not seem to be applicable to a world where prices and NGDP fell sharply despite a large increase in the monetary base.

It’s not enough to be correct, your model must also seem correct. During the 1920s, Hayek had argued that central banks should stabilize NGDP. But when NGDP plunged sharply after 1929, Hayek refused to support monetary stimulus. His underlying model was correct, but his policy advice was so misguided that people stopped listening to him.

Keynes produced an explanation for depressions that was consistent with most people’s common-sense intuition about the economy and was also able to account for movements in NGDP that the quantity theorists struggled to explain. Keynes suggested that “aggregate demand” was determined by factors such as deficit spending, business confidence, and the consumer propensity to save. Here is AI Overview listing 5 factors that shift the IS curve to the left in. an IS-LM model:

Decreased Government Spending (G): A reduction in public works, defense, or government services directly lowers aggregate demand.
Increased Taxes (T): Higher taxes lower consumers’ disposable income, which dampens both consumer spending and overall demand
Decreased Investment (I): A drop in business confidence or tighter credit conditions reduces business investment regardless of the interest rate.
Decreased Consumer Spending (C): If households decide to save more of their income and consume less, the IS curve shifts left.
Decreased Net Exports (NX): A drop in foreign demand for domestic goods (or an increase in domestic demand for foreign imports) reduces the net export component of aggregate demand.

These are factors that the Keynesian model views as contractionary—fiscal austerity, lower confidence among businessmen, the paradox of thrift, trade deficits, etc. And when I read the output from Fable, it is quite clear that this is the framework it uses to evaluate macroeconomic debates. But this approach only makes sense in a world where monetary policy is somehow constrained, that is, where central banks are unable to control inflation and/or NGDP. In an unconstrained fiat money world, a central bank could and should offset any of these expenditure shocks in such a way as to prevent them from causing macroeconomic instability.

Here is the famous Keynesian IS-LM diagram, showing the effect of a negative shock such as increased business pessimism (less I), a higher propensity to save (less C), and/or fiscal austerity (less G):

Let me be clear: I hate IS-LM more than almost anything in the world, with the possible exception of MMT. But if I’m going to explain how Fable became Keynesian, I need to give the model the benefit of the doubt, at least for the moment. In addition, I need to make a few simplifying assumptions that are not strictly true but are reasonable approximations of the model’s implications.

The first approximation is to view the LM curve as representing monetary policy. Thus, an increase in the money supply (in the Keynesian model) causes the LM curve to shift right, leading to lower interest rates and higher output. Does it work this way in the real world? Probably not, and least not usually. But this is a widely held way of thinking about macro so let’s accept it for the moment.

If the LM curve is in some sense ”monetary policy”, then a world where monetary policy is constrained by the gold standard can be viewed as a world where the LM curve is stable and the business cycle is caused by left and right shifts in the IS curve. And that’s basically the world described by Keynes in the General Theory. Recessions occurred when businessmen lacked “animal spirits”, when consumers tried to save too much, when foreigners stole our jobs through trade deficits. And the solution was to shift the IS curve back to the right with expansionary fiscal policy (more G and/or lower taxes.)

Can this Keynesian view be reconciled with the monetarist approach? Since both are based on identities, there must be some way to reconcile the two views. Consider the case where the money supply is fixed—why would negative expenditure shocks affect NGDP? If Ms is fixed, a fall in NGDP would be caused by an increase in the demand for money (as a share of GDP). Or lower velocity, if you prefer that framing. And all of the bearish factors cited by Keynes do indeed tend to raise money demand and lower velocity, but not in the way that you might assume.

For instance, the “paradox of thrift” is actually the paradox of money hoarding. More saving by itself would not be contractionary, as saving equals investment. The actual problem is that an attempt to save more depresses interest rates, which increases the demand for money. Recall that interest rates are the opportunity cost of holding (non-interest bearing) cash. That increased demand for money is what causes NGDP to fall.

The logical solution to this problem would be to increase the money supply to match the rise in money demand, in which case an increased propensity to save would not have any contractionary effect. By the time I studied economics in the (inflationary) 1970s, people had mostly stopped worrying about the paradox of thrift, for exactly this reason.

From a monetarist perspective, you might say that Keynesians believe that recessions are caused by low interest rates. A contractionary shock impacts saving and investment in such a way as to reduce interest rates, which then increases real money demand. As people hoard more money, NGDP declines if the money supply doesn’t rise to match the increase in money demand.

Keynesians, however, would be horrified by the claim that their model implies that low interest rates cause recessions, because they don’t look at things using the monetarist (money supply/money demand) framework. They see these contractionary shocks in a very mechanical way, as directly impacting aggregate spending. In contrast, when Keynesians think about the causal impact of changes in interest rates, they think in terms of monetary policy. A Keynesian sees a fall in interest rates as an expansionary monetary policy, that is, a rightward shift in the LM curve, where you move down and to the right along a given IS curve.

And yet, the IS-LM model clearly suggests that we need to avoid this sort of “reasoning from a price change”. A fall in interest rates could be caused either by an expansionary monetary policy (LM shifts to the right) or a contractionary expenditure shock (IS shifts to the left.). And if we hold the money supply constant, then it is more likely that low interest rates would reflect a leftward shift in the IS curve, which would be contractionary. (Technically, it could also reflect money dishoarding.) I wish that Keynesians paid more attention to this possibility.

Part 4: Keynesianism is the QWERTY of macroeconomics

[Apparently, the inefficiency of QWERTY keyboards is a myth. But as they say, "When the legend becomes fact, print the legend,"]

After WWII, the US gradually evolved from a quasi-gold standard, where the price of gold was pegged at $35/ounce, to a 100% fiat money system (achieved in January 1968.) Under this new policy regime, the monetarist approach was far superior to the Keynesian approach. At that time, the Keynesian C+I+G+NX framework for aggregate expenditure should have been discarded and replaced with a monetarist approach where NGDP is determined by the monetary authority. The optimal policy was to adjust the money supply to accommodate any change in money demand, keeping M*V = NGDP growing at a steady rate of roughly 4 percent.

In this ideal world, there would be no worries about “animal spirits”, consumer pessimism and trade deficits. No call to run fiscal deficits to “create jobs”. Just keep NGDP growing at 4 percent by adjusting the money supply to accommodate shifts in money demand. After all, the Keynesian model was built to address a world where monetary policy was constrained by the gold standard, and that world no longer exists.

So why didn’t that happen? Why did the Keynesian model continue to dominate the profession, even at the highest level?

If you are a rational reader, your most sensible conclusion would probably be that “Sumner is wrong”. I have enough self-awareness to understand that. Indeed, that’s my “outside view”. But today, I’ll give you my inside view.

Like fish that don’t know they are wet, I believe the economists of the 1930s had no understanding of how much of their worldview implicitly reflected a commodity price peg for the monetary system. Many people know that Keynes opposed the gold standard. Far fewer know that he opposed a fiat money system even more strongly. Keynes favored what today might be called a quasi-gold standard, something like the Bretton Woods system (where the dollar was pegged to gold at $35/ounce and other currencies were pegged to the dollar.)

Keynes (on the right) participated at the 1944 conference that created the Bretton Woods System, although the final version was closer to the American proposal of Harry Dexter White (on the left). This picture was taken in 1946, and Keynes died a few weeks later.

If Keynes had favored fiat money, then presumably he would have advocated something like a 2% inflation target, or more likely an NGDP target. (In the General Theory, Keynes suggests that NGDP is more meaningful than inflation.). But Keynes was horrified by the thought of fiat money, and never envisioned an unconstrained fiat regime as a practical solution. This is why Keynes denied the existence of the Fisher effect, which is only important under a fiat money regime. And this is why he didn’t anticipate stagflation, which is much more likely to occur under a fiat money regime.

A second important factor explaining the durability of Keynesianism is the way that it matches the common sense of most people. To the average person (and even to Fable), it seems like common sense that if consumers try to save more and spend less, it would “hurt the economy”. This false intuition about the economy comes from inappropriately applying a true microeconomic concept to a macroeconomic environment where it is no longer true. If consumers wish to spend less on cars, it really does cost jobs in the auto industry. But if consumers wish to spend less on all goods, it does not cost jobs in the overall economy. Rather (assuming appropriate monetary offset) any loss of jobs in consumer goods industries is fully offset by gains in investment goods industries.

A third problem is that postwar monetarists started advocating money supply targeting, which only works during periods when money demand is stable. When money demand became unstable, the entire monetarist approach was (wrongly) seen as being discredited.

And finally, when countries like Japan (and later the US) hit the zero lower bound for interest rates, many economists wrongly assumed that central banks were “out of ammunition”, which made the Keynesian approach look more attractive. In Alternative Approaches to Monetary Policy I explain why this pessimistic view is incorrect. No fiat money central bank is ever out of ammunition, unless they run out of paper and green ink.

So instead of discarding the Keynesian model and replacing it with a better monetarist model, the profession stuck with the basic Keynesian framework, but kept adding epicycles to address the flaws that people like Milton Friedman kept discovering. The Phillips Curve shifts? Fine, we’ll add inflation expectations. Fisher effect? Fine, we’ll account for that. Rational expectations? Fine, we’ll add that to New Keynesian models.

All these fixes were appropriate, and New Keynesianism was clearly an advance over the older “vulgar” Keynesianism. But these fixes were attached to a decaying carcass. No matter how many epicycles were attached, the Keynesian approach was rotten at its core. This is why the mainstream did such a poor job during 2008, when a few market monetarists saw the underlying problem before the rest of the profession. Money may not have looked “tight” in 2008, but if NGDP is falling then it is tight.

If I knew nothing about a particular field of science, it would be rational for me to put more weight on the consensus view of experts, rather than the views of a tiny minority of heterodox thinkers. Thus, it is completely rational for Fable (and other AIs?) to prefer the Keynesian approach over the monetarist approach. In some ways I’d be frightened if they adopted market monetarism, as that would suggest they were open to be influenced by fringe theorists, which might be dangerous. If they like monetarism, what’s to stop them from being influenced by MMTers? The mainstream isn’t always correct, but it is the safest position in a world of great uncertainty.

Going forward, it is possible that AIs will become smart enough to adjudicate the dispute between Keynesians and monetarists and produce a verdict that is so intelligent that it is convincing to both sides. If and when that occurs, we will have achieved what I view as artificial super intelligence. I doubt I’ll live that long.

In the next post, I will refer to this post when explaining why I disagree with the Keynesian framing used in some responses by Fable.

PS. Because I hate IS-LM, I am not good at using the model. I never taught it. Please correct any mistakes in the comment section.

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

The empire strikes back

Scott Sumner — Sat, 11 Jul 2026 23:25:18 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

Vaidas Urba is one of my most well-informed commenters. In recent weeks, he has been helping me to better understand how LLMs can provide useful feedback on my interpretation of stabilization policy during the 2008-09 recession. Today, I’ll present an evaluation of market monetarist claims about 2008 generated by the Fable LLM, which Vaidas sent me by email. I prefer to rely on Vaidas for this sort of exercise, as he has a much better idea than I do about how best to utilize LLMs.

Beyond the paywall, I will respond to the Fable output, offering my own point to view. To be clear, I view Fable’s evaluation of market monetarism to be of very high quality, better than what you’d get from most economists. In the end, I have substantial disagreement on a few points, but in every single case it is based on subtle questions of interpretation—none of the Fable responses are “dumb”. Do not interpret my responses as an implied critique of LLMs.

In the prompt below, think of the HANK model as a mainstream macro model that puts more weight on fiscal policy than does the market monetarist model. In a follow-up, I’ll show the results when Vaidas asked the same question from a HANK perspective, and not surprisingly the result was slightly less favorable. (Recall that LLMs are somewhat obsequious.) Future posts will also look at what happens when Fable is asked to look at both responses, and see which one it liked best. We will also examine what happens when Fable is asked to predict where I would disagree with its analysis.

This was Vaidas’s prompt:

You are a discussant preparing for macroeconomics seminar. The topic of the seminar is the thesis:
Market monetarist policy prescriptions would have been more impactful mitigating the severity of Great Recession compared to the policy prescriptions of HANK [Heterogeneous Agent New Keynesian] models.
Identify and explain the top 5 arguments in favor of the thesis. Identify and explain the top 5 arguments opposing the thesis. Based on the arguments, provide your ultimate verdict with a reasoned justification.

Because the Fable response is a bit long, I will not indent the output:

Arguments for the thesis

1. Timing: the severity was locked in before fiscal tools could arrive. The catastrophic quarters were 2008Q3–2009Q2. The Fed held rates at 2% on September 16, 2008 — two days after Lehman — citing inflation risk even as TIPS breakevens collapsed. An NGDP level target would have mandated aggressive easing by mid-2008, while 200bp of conventional space remained. ARRA wasn’t signed until February 2009, after the worst of the collapse; expectations-driven monetary policy acts through markets in days, while fiscal transfers face legislative and disbursement lags.

2. Level targeting mobilizes expectations automatically. Under NGDPLT, any shortfall implies faster expected catch-up growth, which raises current spending — bygones aren’t bygones. The proof of concept is FDR’s 1933 devaluation (Temin–Wigmore; Eggertsson 2008), where a regime shift reversed deflationary expectations almost overnight. A credible NGDP commitment in 2008 could have prevented the collapse in nominal spending expectations that converted a housing correction into the Great Recession.

3. Monetary offset makes fiscal-first prescriptions second-order. The 2013 test case: sequestration plus the payroll-tax expiry (~2.5% of GDP consolidation) coincided with steady growth because QE3 offset it — as Sumner and Beckworth predicted against mainstream Keynesian forecasts of contraction. If the monetary regime pins down aggregate nominal spending, HANK-style transfers are hostage to the Fed’s reaction function; fixing the regime is logically prior.

4. NGDP stabilization treats the debt problem HANK identifies — at the root. Debts are nominal, so the ~8–10% NGDP shortfall from trend crushed borrowers’ incomes against fixed obligations: exactly the Fisher/Mian–Sufi amplification mechanism. Sheedy (2014) shows NGDP targeting functions as optimal risk-sharing between debtors and creditors in incomplete markets. Preventing the nominal income collapse beats compensating households after balance sheets implode.

5. Scale and feasibility. The nominal shortfall exceeded $1 trillion; transfers of that size were politically impossible (the ~$800bn ARRA was a bruising fight, and stimulus appetite died after 2010), whereas the Fed’s balance sheet requires no appropriation. And as real-time advice, market monetarism actually existed in 2008–09; HANK models (Kaplan–Moll–Violante, 2018) were formalized only years later.

Arguments against the thesis

1. HANK microfounds away the market monetarist transmission mechanism. Once at the ZLB (December 2008), NGDPLT works almost entirely through expectations of future policy — precisely the channel HANK discounts. McKay–Nakamura–Steinsson (2016) show forward guidance loses most of its power when constrained households respond to current cash flow rather than promises. The thesis leans on the one channel the rival framework shows is weak.

2. MPC heterogeneity makes targeted transfers mechanically potent. Roughly a third of households are hand-to-mouth, including the “wealthy hand-to-mouth” (Kaplan–Violante 2014), and the 2008 rebate studies (Parker et al.) confirm high spending responses among them. Transfers and UI extensions deliver purchasing power directly, while monetary easing routed through channels that were broken in 2008–09: refinancing blocked for underwater borrowers, impaired bank lending, and wealth effects accruing to low-MPC households (Auclert’s redistribution channel).

3. The shock was a balance-sheet/financial shock, not merely nominal. Mian and Sufi show consumption collapsed first in high-leverage counties — before Lehman and before NGDP fell — implying causation ran from household balance sheets to spending to NGDP. A promised NGDP path cannot recapitalize a bank in October 2008 or fix an underwater mortgage; debt relief and targeted transfers address the actual shock rather than its nominal symptom.

4. Credibility: the “concrete steps” problem. At zero rates, hitting an NGDP level path requires the public to believe in instruments whose measured effects were modest (Woodford 2012; Eggertsson–Woodford irrelevance results) and whose promises face time inconsistency — future FOMCs must willingly tolerate overshooting. Japan’s persistent undershooting even after 2013’s dramatic regime shift illustrates the limits. A transfer check requires no beliefs; it clears.

5. The 2020–21 natural experiment vindicates the HANK playbook. COVID policy was HANK at scale — direct transfers and expanded UI — and nominal demand fully recovered (and overshot) within 18 months, versus NGDP never regaining its trend after 2008 despite QE1–3 and guidance. Cross-sectional multiplier evidence (Nakamura–Steinsson 2014; Chodorow-Reich) supports multipliers above one at the ZLB, which also weakens the 2013 monetary-offset story: with the Fed pinned at zero, it wasn’t offsetting so much as independently easing.

Verdict

I narrowly accept the thesis — but on more limited grounds than market monetarists claim. The decisive consideration is what “severity” means: it was determined in roughly three quarters, during which (a) the Fed still had conventional ammunition and committed a documented, avoidable expectations error, and (b) no politically feasible fiscal package could have arrived fast enough or large enough to offset a trillion-dollar nominal collapse. An NGDP level target operating in mid-2008 plausibly truncates the collapse itself, and by stabilizing nominal incomes it would have blunted the very debt-deflation amplification HANK emphasizes. Prevention dominates treatment for this particular counterfactual.

But the margin is thin and conditional. The strongest opposing argument — that HANK undermines the expectations channel the whole market monetarist edifice rests on — bites hard once the ZLB was reached, and it’s why the verdict flips if the question were instead about engineering the fastest recovery from 2009 onward, where the 2020–21 evidence favors the HANK playbook decisively. The honest synthesis is that the frameworks answer different questions: market monetarism specifies the right nominal regime; HANK specifies realistic transmission and incidence. The optimal counterfactual policy is NGDPLT as the anchor plus HANK-informed automatic fiscal stabilizers as insurance for exactly the moments when expectations channels fail — which suggests the thesis is right about 2008, but for reasons that make it a poor guide as a general rule.

* * *

OK, this is me again. Although Fable suggested that it narrowly accepted the market monetarist thesis, I don’t believe it accepted the core claims of market monetarism. Below the paywall, I’ll respond to a number of specific points.

Kakistocracy

Scott Sumner — Sun, 05 Jul 2026 17:34:14 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

[I wanted to put this out before the release of Hanania’s new book, but the post ended up falling on the weekend of America’s 250th birthday. To be clear, I still think this is a great country in most ways, despite the recent decline in our political environment. There’s more to life than politics. So don’t take this post as America bashing, it’s American politics circa 2026 bashing. BTW, here’s how I’d rewrite Jefferson’s famous remarks:

In the more than 300 years since the invention of the printing press, the world’s best minds have produced innovations in philosophy, history, literature and political economy that have led thoughtful people to conclude that the only appropriate role of government is to secure the people’s right to life, liberty and the pursuit of happiness.

Yeah, not exactly as literary as Jefferson, but more informative.]

Photo by Danny Burke on Unsplash

Richard Hanania’s new book entitled Kakistocracy: Why Populism Ends in Disaster is the definitive account of populism in the 21st century. I’ll begin this post explaining why Hanania was the ideal person to write a book on modern populism as well as the reasons why I would not be able to do this sort of project. Then I’ll briefly discuss the contents of the book, focusing on the difficulty of making generalizations about politics.

The term kakistocracy means “rule by the worst”. The term “worst” can have multiple meanings, including the least competent and the most unethical. Long time readers know that I believe these two negative attributes often go hand in hand, and Hanania seems to hold a similar view. Hanania focuses on a style of populism that combines a rejection of expertise with a cavalier disregard for the truth.

Previous attempts to grapple with this topic often struggle to avoid “Trump Derangement Syndrome”, which means expressing outrage at Trump’s various misdeeds. (Future historians will be bemused that the term “TDS” was not a reference to the brainwashed members of the Trump cult who lived in an epistemic bubble where they were showered with misinformation and conspiracy theorizing-and instead referred to thoughtful pundits who accurately described this phenomenon.) So why is it so hard to avoid TDS, and why am I not well suited for this sort of project?

Recall my recent post entitled “The world is bigger than you can imagine”. Donald Trump is a near perfect example of this general phenomenon. I am vaguely aware that Trump has been producing almost nonstop scandals over the last decade, including everything from petty stylistic outrages to personal corruption to bad public policy to abuses of power. The examples number in the hundreds. But at any given point in time, I can only recall a few. The forefront of my mind simply refuses to devote a lot of space to Trump outrages. (Perhaps they are somewhere in my mind’s basement, to be recalled if reminded of some long-forgotten scandal.)

Because I can only recall a few examples at a time, any blog post I write on the subject would be woefully inadequate. Readers might reasonably wonder what’s so special about Trump; don’t other politicians also lie, issue corrupt pardons, engage in abuse of power, etc? Yes, on occasion they do. In the end, my TDS post would make Trump look better—more like a normal politician. So that wouldn’t work, I’d just be preaching to the choir. It would take a future Robert Caro doing a three-volume set of 1200-page books to properly document the scale of Trump’s awfulness, and that’s not something I can do. Sorry.

Hanania mentions a few Trump outrages but wisely goes at the subject from a different angle, focusing on the deeper causes behind the recent rise of anti-elite populism. Even better, he works from the vantage point of someone that voted for Trump in 2024 for policy reasons but has recently become disgusted with the entire MAGA movement. Those of us who have long been critical of Trump might be tempted to suggest that Hanania has finally seen the light, but in fairness the second term Trump really is far more populist and far more corrupt and incompetent than the first term version. Hanania always recognized Trump’s flaws but is no longer willing to put up with them.

Hanania’s personal history gives him insight into the MAGA movement that most of us do not have. Some of the best parts of the book describe his personal journey from the far right to a position closer to . . . here I need to be careful with labels . . . classical liberal? Neoliberal? Libertarian? In any case, closer to my own views on politics, however you wish to describe them.

In the introduction, Hanania dissects the populist movement with precision, showing how a segment of the right has recently become deranged over politics, partly due to an avalanche of misinformation in social media:

All of these facts could easily be found in countless media outlets by anyone with an interest in what happened on January 6 and the events leading up to that day. Yet I saw that conservative-leaning friends either didn’t know these facts or dismissed them in absurd ways, for example, by making false equivalencies to things Democrats had done in the past or, as we have seen, passing along a news media article with a headline that contained some magic words they thought proved a conspiracy or excused Trump’s behavior. Friendships ended, not over political disagreements; I could have respected those who said that Trump tried to overturn the election and that he was right to do so. Such a person might have different moral standards, but they would at least be connected to reality.
To deny or ignore widely known facts, however, about a major recent historical event while continuing to strongly believe that Trump had been treated unfairly by the American establishment was too much to take. . . .
[T]he same intellectual independence that attracted me to fringe ideas and political movements in the first place would not allow me to claim victory and lean in to cultivating existing friendships and my natural audience. Right-wing concerns about issues like DEI and the handling of Covid were completely justified. Yet I could not ignore that the Trump movement over time came to be dominated by epistemological nihilism, open bigotry, authoritarianism, and an embrace of conspiracy theories. Some of my values stayed the same, while others changed. I realized that being obsessed with race and gender issues is the other side of the coin of wokeness, and leads to a zero-sum view of the world and policy ideas that make society as a whole worse off. I came to see my embrace of white identitarianism in my early years as a kind of mental derangement, yet it was one that had largely taken over the American right.

Philosophers say that words have fuzzy meanings. Thus, of all the things that you can put on your feet, exactly which objects are called “shoes”? And no words are fuzzier than political terms, including socialism, fascism, liberalism, capitalism, populism and nationalism. Socialism has generally been defined a system that combined statist economic policies and egalitarian social insurance. But those are two very different things. As a result, the term ‘socialism’ gets applies to both highly statist regimes like Cuba and free market economies with extensive social insurance, such as Denmark. Similarly, the term liberalism gets applied to everything from classical liberal to neoliberal to mild socialist.

Defining a term like populism can be like nailing jello to the wall, as it means different things to different people. Hanania links modern populism with opposition to the elites, a term that is also up for grabs, no longer equating to wealth and power. Indeed, some of the world’s richest and most powerful individuals are known for their strong opposition to elite opinion:

Until the last few decades, then, identifying who elites and populists were was a straightforward task. Elites were people with influence and power, and populists were those with less power and influence who challenged elites through direct appeals to the masses. Such a framework, however, makes little sense today. The most obvious problem with these definitions is that, as previously mentioned, they would make Donald Trump an elite, given that he is a second-term president and has established complete domination over a major political party like perhaps no other figure in American history.
Trump and his supporters continue to consider themselves at war with the establishment no matter how much power they acquire. . . .
The migration of Sili con Valley billionaires, decabillionaires, and even centibillionaires—most notably Elon Musk—into the MAGA coalition in recent years has not caused it to tamp down on populist rhetoric. This just adds to the mys tery. Elite status is not wealth, and it is not power, or at least not direct political power. Rather, elites are defined by their social capital and the kinds of communities they form. Once we understand this, we can ana lyze populism. As we will learn, populist rhetoric on the right continues to exist in part because actors who are truly self-interested in the sense of seeking tangible resources, like Trump and Musk, along with many of their fans and supporters, struggle to get into the heads of traditional elites. This is one of the sources of tension between the two sides of the modern political spectrum. (pp. 2-3)

This creates an odd paradox:

We have this strange situation, then, where liberals like current elites (except perhaps for some tech billionaires and other plutocrats) in practice but dislike the idea of elites in theory, while conservatives are fine with elites in theory but don’t like the ones we have. (p. 11)

In a recent podcast, someone asked me for an example of where I had completely changed my view on a topic. In my own field of macro, the evolution of my views tends to be incremental. Instead, I cited the growth of diversity in the media, which I once thought would lead to an improved information environment. As Hanania points out, it did not turn out that way:

The left-leaning columnist Matthew Yglesias suggests that the tradition of objective journalism that we all take for granted came about not from a commitment to neutrality as such but from the incentives built into the business model of mid-twentieth-century American newspapers. These outlets sought mass appeal, drawing in both advertisers and subscribers by offering a well-rounded package: weather, sports, stock quotes, and straightforward reporting.
This meant that there was a need to avoid offending any part of the reading public. In this model, objectivity was less about a supply-driven desire for balance and more about reflecting a broadly acceptable centrist consensus. As Yglesias notes, the fragmentation of the media landscape, driven by digital platforms and tailored content, has undercut this approach. Stock prices, sports updates, and weather reports are now delivered instantly to our phones, which for many people eliminates any reason to buy a newspaper. Media outlets therefore compete over niche audiences rather than trying to reach everyone through maintaining a neutral and elevated tone. (p. 16)

Today, the media largely tells people what they wish to hear, which means the news is heavily skewed toward misinformation and wild conspiracy theories.

The traditional left/right dichotomy is increasingly being replaced by a split between the well informed and the poorly informed, with the latter drifting toward authoritarian populism. Hanania emphasizes the fact that this change is particularly pronounced in the GOP:

Only Democrats engaged with serious publications in substantial numbers. Among that group, 31 percent read The New York Times, and 26 percent read The Washington Post. The equivalent numbers for Republicans were 9 percent and 8 percent. One might think that Republicans read conservative papers instead. But even The Wall Street Journal was relied upon by more Democrats than Republicans, 15 percent to 11 percent. Meanwhile, 60 percent of Republicans watched Fox News. Democrats also relied to a large extent on TV, but they had many more people who read serious newspapers and websites in their coalition. (p. 36)

Because of the time lag in publishing, Hanania’s book was mostly written in 2025. Much of what Hanania describes seems even more relevant today, with “kakistocracy” an even more appropriate term for the 2026 version of the Trump administration than the 2025 version. On the other hand, I wonder if Hanania might have spent a bit more time of the insurgent left in the Democratic Party if he had written the book today. Until recently, I had assumed that the excesses of the left were beginning to recede, as 2020-style wokism was going out of style. Perhaps Hanania was working under the same assumption, which seemed plausible in 2025.

Toward the end of the book his analysis drifts over into a subject that I believe is better described as “authoritarianism”, and for the most part he does an excellent job of defining the concept, explaining why it is on the increase, and showing the various ways that it expresses itself, especially in Europe and Latin America:

In Brazil, Jair Bolsonaro’s presidency followed a similar trajectory, built on nostalgia for military dictatorship, climate denial, antivaccine activism, and attacks on electoral institutions. After losing the 2022 election, he refused to accept the results and ultimately inspired a violent attack on government buildings by his supporters, an event that was reminiscent of the January 6 Capitol riot. This demonstrates not only that populists around the world converge on similar ideas and ways of seeking power but also that they learn from one another, as Trump’s attempted coup of 2020/2021 seemingly inspired his Brazilian counterpart to try a similar gambit.

On a visit to Mexico in 1971 (at age 16), I had my first exposure to what was then called the “third world”. I was struck by the widespread acceptance of what I viewed as wild conspiracy theories, such as the idea that the CIA was behind all sorts of mysterious events in Latin America, or the view that the Holocaust never happened. Little did I know that I’d live to see America adopt Mexican-style epistemics.

If the first two thirds of the book is pretty great, the final portion is merely good. That’s not because Hanania gets things wrong, rather the topic of populism throughout the world becomes so sprawling that generalizations become increasingly difficult to make. The following comments are not so much areas I where disagree as they are areas where I believe some qualifications are important.

Hanania has a chapter entitled Giving the Devil his Due, which discusses some areas where the populists got things right, including some (but not all) of their views on Covid. He also suggests that authoritarian populists may have been a necessary evil in certain Latin American countries, citing El Salvador and Peru.

While his individual observations are reasonable, mixing so many examples can lead readers astray unless they pay close attention to his qualifiers. Consider these facts about the world (my words, not Hanania’s):

Argentina’s Milei is a democratically elected leader who uses a populist style and enacts neoliberal policies.

Peru’s Fujimori is a democratically elected authoritarian leader with a populist style that enacted neoliberal policies.

Singapore has authoritarian leaders that use a non-populist style to enact neoliberal policies.

Venezuela’s Chavez and Maduro were authoritarian leaders that employed a populist style and enacted socialist policies.

In other words, it’s complicated. In the final portion of the book, Hanania discusses authoritarian populism in a wide variety of situations. While he does draw many relevant distinctions in the cases that he discusses, I wish he had emphasized even a bit more that a populist style, populist policies and authoritarianism are three very distinct concepts, albeit frequently linked. I worry that less than careful readers might conclude that authoritarianism or populism are appropriate in many cases because of a few individual examples that employed some but not all of the MAGA playbook.

Here is Hanania discussing democracy on page 107:

As long as Western elites maintain the democratic versus authoritarian framework to understand contemporary trends, they will remain vulnerable to challenges of hypocrisy and bad faith. Figures like Donald Trump and Viktor Orbán, for all their flaws, have large numbers of sup porters who agree with them on important issues and have won free and fair elections. When their critics attack them for being undemocratic, their backers can legitimately point to polls and the authority of the ballot box. The mathematician and author Nassim Nicholas Taleb, for instance, mocks the kind of Western intellectual who praises “democracy” when it returns a result he likes but denounces “populism” when things go the other way.
These critics have a point. Even when it comes to spin dictators such as Putin, defenders of democracy have to contend with the fact that he is genuinely popular. Opponents can claim that that popularity is exaggerated by censorship and authoritarian aspects of the system. Nonetheless, it is still real, and those who point to democracy as a fundamental value must admit that illiberal policies themselves often have a large degree of support. If the majority of the public is fine with oppressing minority movements and a strong leader exercising arbitrary power to some degree, and then a leader becomes even more popular as a result when he responds to popular will, has the principle of democracy been violated, or has it reached its logical endpoint?

Nothing here is necessarily wrong, but I’d emphasize a few other points. Whether a country is democratic does not depend on whether the leader is popular. Most European leaders are currently unpopular, but European countries are almost all democratic. If Putin is popular, that fact doesn’t imply that Russia is democratic.

Nor does the fact that Putin was democratically elected make Russia a democracy. Russia certainly was a democracy for an extended period after Putin took office in 2000 but has recently become a repressive dictatorship. Democratically elected leaders occasionally turn their countries into non-democratic nations. (The Hitler analogy actually fits here.)

I also worry that people on both the right and the left underestimate the importance of process. Thus, some left wingers want to pack the Supreme Court and some right wingers want to ride roughshod over civil liberties to achieve goals like crime reduction and mass expulsion of illegals. It is important to avoid the temptation to support policies that achieve your own policy goals at the cost of giving up on checks and balances in political power. In the long run, we would all be worse off. Think about what will happen when those same powers are used by the opposition party.

At any given point in time, for any given political issue, it’s close to a coin flip as to whether a democracy or an authoritarian country does better. But over time, the informational and incentive advantages of liberal democracy generally dominate more authoritarian forms of government.

Hanania may agree with that view (I share most of his current political opinions), but I worry that his readers might read too much into the cases where authoritarian governments had limited successes, such as economic policy reforms in Pinochet’s Chile, or crime reduction in Bukele’s El Salvador. There was a time when Putin seemed like an improvement over the (more democratic) Yeltsin, and there was even a brief period when many Germans viewed Hitler as an improvement over the Weimar government. Eventually, authoritarian governments usually (not always) end up making things worse.

It is also important to recall that while voters often have illiberal views, democratic countries generally enact more classically liberal policies than do autocracies. Hanania cites the fact that majorities don’t always support the rights of minorities, but it’s worth emphasizing that, on average, minority rights are preserved far better in democracies than in authoritarian countries. Similarly, despite the fact that voters often favor brain dead economic policies such as rent control, democracies tend to have far better economic policies than authoritarian countries, on average. Ironically, the country most often cited for the opposing view—Singapore—does have contested elections and is thus better described as a quasi-democratic/soft authoritarian hybrid. I’d be thrilled if Russia and China liberalized even to the level of Singapore.

Consider this quotation from page 193:

In short, Latin America presents an unsettling paradox: It is the most democratic region in the developing world, and yet among the most violent, least stable, and slowest growing. Compared to regions like East Asia, where many states have modernized under authoritarian regimes with stronger central governments, those in Latin America appear to have institutionalized electoral competition and principles like free speech without ever establishing capable governance.
This offers a sobering counterpoint to the assumption that democratization and human rights norms necessarily lead to development and order.

Is this the relevant comparison? How about the following:

Most East Asian and most Latin American countries are democratic
Most African countries are non-democratic

Does that framing change your view?

Or how about the frequently heard claim (not made by Hanania) that China’s success relative to India supports authoritarian models of development. Do the following facts change your interpretation:

China is much poorer than less authoritarian ethnic Chinese places like Taiwan, Singapore and Hong Kong.
India has grown far faster than more authoritarian portions of South Asia, notably Pakistan.

Nationalism is another term with a fuzzy meaning:

Nationalism has something to be said for it in modern Ukraine, given that it is the victim of an aggressive war launched by an enemy that seeks to wipe out its existence as an independent country and replace the current government with something that would be much worse. (p. 137)

Again, this isn’t wrong, but I prefer a different definition of nationalism. To me, Hanania is describing Ukrainian patriotism, not nationalism. (I’ll do a future post on the distinction.) I believe that most Ukrainians would like to see a world where Ukraine becomes a member of the EU and Nato, and the rights of both Ukrainian and Russian speakers are protected.

For the same reason that I’m far more interested in real world communism than in anything Marx may or may not have favored, I focus on real world nationalism, which tends to be anti-EU, anti-Nato, authoritarian, and often defines nationality in narrow ethnic terms, not in terms of citizenship. Ukraine’s so-called “nationalists” seem to want their country to become more like a cosmopolitan Western European nation. They literally elected a Jewish president.

So how depressed should we be by the rise of anti-elite populism? Consider these three claims:

Politics is getting worse.
Voters are becoming less well informed.
The world is getting better.

Are all three claims true? I believe the answer is yes. There’s more to life than politics.

America is doing well despite our bad politics. What can we learn from that fact? Consider that Congress is willing to promote clowns to seemingly important positions like head of the Defense Department or HHS but balks at promoting clowns to the Fed or the Supreme Court. That suggests that Congress views most of the government as being essentially on automatic pilot—where it makes little difference who heads a department—and only the Fed and the Supreme Court make important decisions that actually move the needle.

I do worry that a continued slide toward banana republic status would eventually begin to corrode the broader American country, but for the moment we can take some solace in the fact that there’s a great deal of ruin in a nation.

For a more pessimistic take, consider the UK’s recent doom loop:

Populists in Britain force a referendum on the EU. It passes by a very narrow margin

The UK economy becomes increasingly stagnant, a problem widely attributed to Brexit.

The British public turns against Brexit, with polls showing nearly 2 to 1 now view it as a mistake.

The British public becomes very pessimistic about current conditions, which leads to increasing support for the (populist) political party that is responsible for Brexit.

Rinse and repeat.

What’s the solution? I don’t see one. When I was young, a candidate like San Jose mayor Matt Mahan would have done well in our recent primary election for governor. It seems like all the elite liked him; center, left and right. But he got only 3.5% of the votes, while a bunch of inferior candidates did much better. How can anyone have any optimism about our political system? The most optimistic take today is “Maybe politics don’t matter anymore.” Perhaps AI will be the most important trend of the 21st century, not populism. Or perhaps populism is just a passing fad, like 1960s-era liberalism and 1990s-era neoliberlaism.

Overall, I was extremely impressed with Hanania’s new book. I have strong and often contrarian opinions on politics, so naturally I would quibble about a few points when he extended his analysis to the international scene. But Hanania has a done a brilliant job of analyzing the causes and consequences of 21st century populism. Highly recommended.

PS. A few months ago I read Gombrowicz’s Diary, which is now one of my all-time favorite books. Here’s how (1959, pp. 417-18) he describes the cluelessness of uninformed Argentine populists:

Sound familiar?

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

Films of 2026:Q2

Scott Sumner — Wed, 01 Jul 2026 15:38:33 GMT

Today’s post is brought to you by my sponsor, Mechanize. They’re hiring junior software engineers at $300K/year base salary. Apply now!

* * *

Before getting into last quarter’s films, a few important announcements:

~~While LLMs have achieved impressive success, I’ve noticed that many firms have trouble fully utilizing this new technology.~~ The firm Mechanize sells evaluations and training data to frontier AI labs to help them make their AIs better at coding ~~specializes in helping companies use AI to boost productivity~~. They have agreed to sponsor The Pursuit of Happiness because they hope to recruit some of the highly talented readers who subscribe to this blog. (And that’s not just flattery; I was pleasantly surprised by the number of brilliant young people I met up at Lighthaven who read my blog.) Click on this link if you believe you are a good fit for Mechanize.
I originally intended not to have any ads in this blog, but Mechanize’s offer was too good to pass up. The good news is that it is a 100% exclusive contract, which does not allow for any other ads. So you’ll merely see a one sentence ad at the top of each post, and another small ad at the very bottom. But I’d also like to share at least a small portion of my financial gain with readers. Thus I will commit to never raising subscription prices as long as they continue sponsoring the blog, and may even cut prices next year when subscriptions come up for renewal.
As you’ll see in the next item, the unpaid subscribers are much more numerous than paid subscribers. For that group, I’ll move from my current system of one half free posts to a new system of two thirds free posts and one third paid posts. I hope this increases my total number of subscribers, even if it eventually costs me a few paid subscriptions (by making the free option modestly less disadvantageous.)
I’m a midwesterner at heart and always look for Pareto optimal solutions. :)
Thanks to you guys this blog has finally reached an important milestone, crossing the 10,000 subscriber line:

In other news, I read a bunch of novels by people like Modiano and Krasznahorkai, as well as two African travelogues (by Teju Cole and Paul Theroux) and an unconventional guide to Japan by Pico Iyer. But for me the standout book in Q2 was Witold Gombrowicz’s Diary. I’ll probably do a post on it at some point. Another standout was Richard Hanania’s Kakistocracy, which I’ll discuss in my next post.

Daniel Frank has a beautiful post discussing the Taiwanese film Yi Yi.

And here are two Youtube videos that I enjoyed. The first on one of my favorite British films:

And one on the difference between modernism and mid-century modern:

This interview of Tyler and Nabeel has some fascinating observations on music and film.

And finally, some astute comments about art by Janan Ganesh:

This obsession with the floor is even more extreme in other fields. In modern entertainment, almost nothing is total rubbish. A song or TV series will have a minimum of polish and a recognisable structure. Some of this is down to technological progress: the worst studio kit now is quite good. The rest is the result of corporate risk-aversion in an ultra-competitive market. If the audience can stream things from around the world, a content platform can’t afford to empower the kind of eccentric artist who might make a dud (or a masterwork). The captive audiences of the pre-streaming age perversely allowed creators to shoot for the ceiling.

This is also my view. Modern films have higher floors and lower ceilings.

2026: Q2 films

Newer films:

Lumière, le Cinéma! (France) 3.6 Not for everyone, but I was fascinated by this documentary composed of 120 clips from the dawn of cinema. Oddly, I’ve gone almost my entire life without seeing what life in the 19^th century actually looked like. Now I know. It’s even stranger than I imagined, partly because the people being portrayed regard their world as perfectly “normal”, just modern life. And yet, other than trains there’s almost no signs of modern technology. Even though the second industrial revolution began a couple decades earlier, the world portrayed in this film doesn’t yet seem to be affected by the newest technology. You don’t see any automobiles or electrical appliances—everything is done by hand.

If you are too lazy to sit through the entire 105 minutes, check out the scenes of women working around the 46 to 48-minute mark—a real eye-opener. Especially the French women mining coal. People talk about the possibility of AI replacing all jobs, but if the people of 1895 were transported to today’s world, they’d say that almost no one in the year 2026 is doing any “work” at all—just a bunch of people sitting in offices. Today, almost everything difficult and dangerous is done by machine.

Backrooms (US) 3.6 I don’t like horror films but I do like architecture. This is architectural horror, which is OK. The 20-something director shows a great deal of promise in a film that is slightly reminiscent of The Shining. A good horror film for people that don’t like horror.

In a recent post, I discussed the fact that we forget most of our past. This film reminded me of something I’d completely forgotten—that I once worked on a job installing an acoustic tile drop ceiling. The film also got me thinking about how younger viewers view dilapidated mid-century modern commercial spaces. To me, the 20^th century architecture in the film seemed modern, but I suspect that many younger viewers will see it as a bit antique, much like the way I viewed old Victorian houses in horror films that I saw as a teenager. I need to accept the fact that the 20^th century (which I’ll always view as “my century”) is over. It’s history.

Obsession (US) 3.6 This is naturally paired with Backrooms as both are low budget horror films made by very young directors that have become smash hits. This one focuses on the social anxieties of Gen Z people in the dating market, with some implied references to gender politics. As with Backrooms, I’m less interested in the film itself than in the potential for interesting futures films from these directors. It’s not an easy challenge for a contemporary director just starting out, as it is becoming increasingly difficult to find fresh styles in a medium that has already seen an enormous amount of innovation. Think of someone trying to create a great guitar-oriented rock band in the 2000s---not easy to do.

The Love That Remains (Iceland) 3.4 This exercise in magical realism has a relatively slight story but is enjoyable as a quasi-documentary about life in rural Iceland.

Eno (UK) 3.4 The avant-garde style of this documentary nicely matches the subject matter.

Tuner (US) 3.4 Hollywood is getting pretty good at producing this sort of film—too good in a sense. If the director pushes all the right buttons, the film starts to become predictable. Thus, when this film calls for the tuner’s labor to be cheap, a character says they haven’t raised the price in 30 years. Later on, when the plot calls for the service to be expensive, he says a tune-up costs $5000. You feel like you are in movie-world, not real life. The film is similar to Crime 101 in the way it gets you to root for the thief by including a heart tugging back story. I suspect that younger viewers will like this more than I did, as they haven’t seen the formula as often. Dustin Hoffman? Sorry, I’m not buying the shtick.

The Christophers (US/UK) 3.3 Soderberg is better when he does genre pictures revolving around crime or transgressive sex. This earnest middlebrow film is a bit clichéd in its view of art and has disappointingly bland visuals for a film about painting. Critics raved about Ian McKellen’s performance. It’s fine, but I’m bored by curmudgeonly old men. I much preferred the lead actress (Michaela Coel), who has a very distinctive face.

Köln 75 (Germany) 3.3 In one respect, this reminds me of the Dylan biopic A Complete Unknown. In both cases, my reaction to the film was heavily influenced by the fact that this music had a big effect on me in the mid-1970s. Thus, I cannot really judge the film purely on its merits; it’s more like visiting a shrine. If you are a jazz purist that thinks Jarrett is overrated, your mileage may vary.

I read that Jarrett himself doesn’t care for the Köln performance. I wonder if that because he realizes that dunces like me appreciate the performance for the wrong reasons. In some ways I prefer the Bremen/Lausanne CDs.

Yearend (Japan) 3.3 A pleasant film about rebellious high school students in Japan. I’ve seen that sort of thing before, but this has the added wrinkle that many of the students are foreigners—a sign of changing times even in immigration averse Japan.

The True Story of Tamara de Lempicka & The Art of Survival (US) 3.2 A very interesting story, but at times the style of the documentary is a bit irritating. Too many pictures flying by too fast, and in some cases the paintings being discussed are not those that are pictured—like one of those AI-created Youtube videos. The artist combines elements of mannerism, cubism and art deco in a very effective way. Some of her paintings have become quite iconic, which suggests that her reputation (and the value of her paintings) will continue to rise.

Forge (US) 3.1 Yet another film about art forgery. Passable entertainment but there’s really nothing new and the young woman’s rapid production of high-quality forgeries doesn’t seem plausible.

Older films:

Scenes From a Marriage (Sweden, 1973, CC, TV version) 3.9 This is the sort of film I generally don’t like. If you like this sort of thing then it’s a 4.0 film, as it’s basically perfect. It reminded me a bit of Knausgaard’s My Struggle: A very long Scandinavian drama with philosophical overtones that seems both hyper-realistic and partly autobiographical. By the way, “realism” in art is a tricky concept. I don’t mean to suggest that Bergman’s dialogue is likely to occur in the real world, rather I see realism as something quirky and highly specific—partly by avoiding clichés. This story does not play out the way you’d expect, or the way that Hollywood would handle things. I wouldn’t say it is realistic, rather it feels real.

I’m glad I waited until age 70, as I would have missed much of the nuance when I was younger.

The Devil’s Eye (Sweden, 1960, CC) 3.8 This Bergman comedy is similar to some of Woody Allen’s films, but with a somewhat better screenplay and much better cinematography.

The Hidden Fortress (Japan, 1958, CC) 3.8 I saw this many years ago but liked it much better this time around. When I was young, I would have rated Star Wars much higher. (The Lucas film (specifically A New Hope) was loosely based on this story.) Today, I rate them nearly equal--as I am now less impressed by the once dazzling special effects. Other Kurasawa films might rightly be regarded as artistically superior, but I don’t know of any that are more enjoyable.

El (Mexico, 1953, CC) 3.8 This Buñuel melodrama was clearly influenced by the Hitchcock films of the late 1940s and early 1950s. Conversely, it almost certainly influenced Vertigo and Marnie.

The Green Ray (France, 1986, CC) 3.8 More somber than other Rohmer films, it provides a convincing view of a woman suffering from depression due to circumstance. Fortunately, Rohmer avoids the predictability that generally drags down this sort of psychological film, always keeping things a bit ambivalent. Marie Rivière is excellent in the lead role.

Kaili Blues (China, 2015, CC) 3.7 I liked this more than in 2015, when it first came out. At that time, I had not yet visited the town of Kaili, in Guizhou province. More importantly, I knew nothing of the director Bi Gan. Now I see flashes of the visual style that would fully blossom in Long Day’s Journey into Night and Resurrection. In his first film, Bi Gan seems heavily influenced by Hou Hsiao-hsien, especially the train sequences. The Chinese title of the film can be translated into English as “Roadside Picnic”, which is the name of the novel that Stalker was based on. You can see the influence of Tarkovsky on Bi Gan’s visual style. This is a very difficult film for non-Chinese viewers, and even the second time around I did not fully grasp the plot.

Side Street (US, 1950, CC) 3.5 Bullitt is often credited with being the first great car chase on film. While the one at the end of this film isn’t quite as exciting, the skillful use of downtown Manhattan architecture makes it equally interesting. Farley Granger might have earned the lead in Strangers on a Train with his performance here. Anthony Mann directed.

Michael Clayton (US, 2007, CC) 3.5 This is the sort of film that Hollywood is very good at. My only reservation is that I wish an anti-corporate film didn’t feel like it was made by a giant Hollywood corporation.

The Dark Glow of the Mountains (Germany, 1985, CC) 3.5 Reinhold Messner’s insane drive to engage in ever more difficult climbs is a perfect subject for Herzog, and this documentary does not disappoint. It shows that even without modern technology such as drones a skilled director can create riveting images. Mountain climbers die young, but they experience things beyond the comprehension of most people. I suspect that normal people overrate the advantages of normalcy. Only 46 minutes long.

Princess Yang (Japan, 1955, CC) 3.5 This late Mizoguchi color film is a Chinese story done in the Japanese language.

Osaka Elegy (Japan, 1936, CC) 3.4 It took quite a while for sound to catch on in Japan—this is Mizoguchi’s first non-silent film.

A Master Builder (US, 2014, CC) 3.4 I don’t care for theatre and I would not enjoy this Ibsen story as a play. But as a film it works, mostly due to the actress who plays Hilde. That sort of acting does not work on the stage. The Andre Gregory/Wallace Shawn production is entertaining, but well below films like My Dinner with Andre and Vanya on 42^nd Street.

Return to Reason (France, 1923-2023, CC) 3.3 This is a collection of 4 silent films directed by Man Ray back in the 1920s, with a soundtrack added in 2023. The two earliest films (which appear second and third) are the most interesting---done in a Dadaist style. Jim Jarmusch’s band Squrl provided the soundtrack. It is interesting to watch films that would have seem quite avant-garde in 1923 but now seem rather antique.

2 Minutes Too Late (Sweden, 1952, CC) 3.3 Good Swedish noir that shows how different Stockholm looked in the early 50s. Everyone on the streets looks Nordic and there is still some fairly shocking urban poverty. Clever ending.

Street of Shame (Japan, 1956, CC) 3.2 In the 1950s, Japanese directors did a number of films looking at the effects of banning prostitution, which suffer from being a bit too didactic. (Women of the Night and Girls of the Night are other examples.) Not one of Mizoguchi’s better films but has a nice final shot.

Man Wanted (US, 1932, CC) 3.2 Hollywood has always liked to do films that touch on contemporary cultural trends—in this case the changing role of women in society. Kay Francis lights up an otherwise pedestrian film and is the main reason to see this blithe pre-code romantic comedy.

The International (US, 2009, CC) 3.1 Tom Tykwer is a good director and there are some really nice scenes in this film. Unfortunately, it has two big problems. First, the plot requires you to put your brain on hold—the conspiracy is so far-ranging as to be completely implausible. Even worse, at the halfway point the film completely “jumps the shark”, with a shootout at the New York Guggenheim that is like something out of a John Wyck film.

Berlin Express (US, 1948, CC) 3.0 A subpar noir that is filmed in location in Frankfort and Berlin. Of interest partly because it shows the devastation of German cities after WWII and also because it shows how for a brief period there was optimism (on the left?) that the Soviet Union could work together with the allied powers.

The World’s Greatest Sinner (US, 1962, CC) 3.0 Like most cult films, this one is pretty bad. But there are all sorts of interesting quirks, including a soundtrack by a very young Frank Zappa, a rock concert where an all-black audience goes crazy for a white guitarist, and a protagonist that thinks he’s God and looks sort of like Zappa (partly due to his “soul patch”.) “God Hilliard” then runs for president. Unfortunately, the plot is extremely far-fetched. It’s asking way too much of a modern audience to believe that millions of Americans would vote for a candidate that saw himself as a God. Oh wait . . .

The Ceremony (Japan, 1971, CC) 3.0 Some striking images but overall I didn’t enjoy this Ōshima film. The satire is much too heavy handed.

The Thing (US, 1982, CC) 3.0 A good horror film needs more than just a series of shocks. Indeed, the best horror films (The Shining, The Birds, Nosferatu, Cat People, etc.) are not even particularly scary.

Painters Painting (US, 1972, CC) 2.9 So-so documentary on contemporary American artists during the heyday of Abstract Expressionism and Pop Art. Strictly for art lovers.

~~No Way Out~~ Against All Odds (US, 1984, CC) 2.7 A nice illustration of why the 1980s sucked. Hollywood treated YIMBYs as if they were the bad guys, everyone looked plastic, and the band that the film showcased was almost a parody of bad 1980s music.

The Game (US, 1997, CC) 2.3 One of the silliest films I’ve ever seen. The only reason I stuck with it is because it was directed by David Fincher. But even the final joke didn’t justify 2 hours of boredom.

A message from my sponsor, Mechanize:

We’re hiring software engineers to build environments and evals that frontier AI labs use to train coding agents.

To get a better sense of the work we do, you can check out GBA Eval, where we had models build Game Boy Advance emulators from scratch and scored their performance.

Base pay starts at $300K/year for junior software engineers, with more for senior roles, plus equity and performance bonuses. Apply here.

Alan Greenspan, RIP

Scott Sumner — Fri, 26 Jun 2026 23:02:44 GMT

Before getting into links, a few comments on recent events:

Alan Greenspan was clearly one of the greatest Fed chairs, perhaps the greatest. But I evaluate his tenure a bit differently than many other pundits. Here are a few contrarian views:

While Greenspan did a great job producing stable NGDP growth during his tenure (1987-2006), there is no reason to believe that he would have done better than Bernanke during 2008, indeed his public comments about monetary policy during 2008-09 suggest that he was just as off base as the Fed.
On the other hand, Greenspan gets way too much blame for the 2008 financial crisis, which was not caused by “deregulation”, whatever that means. Indeed the regulators were pressing banks to do more high risk lending. If you want more regulation, does that mean you wanted the regulators to press banks harder to make subprime loans? Or are you assuming disinterested, omniscient regulators? The financial crisis was caused by a combination of moral hazard (FDIC and Too-Big-To-Fail) plus tight money. Greenspan’s biggest mistake in this area was the 1998 LTCM rescue, which slightly worsened moral hazard.
Don’t think of 1987-2006 as the “Greenspan era”, think of it as the “NeoKeynesian consensus era.” It wasn’t just the Fed, almost all the major central banks figured out how to do inflation targeting using a Taylor principle-type approach, and this largely explains why other developed countries had the same sort of inflation targeting success as America during this period.
In 1996, Greenspan spoke of “irrational exuberance” in the stock market. Years later, that warning was viewed as a prescient call, even though he was pretty clearly wrong. Stocks were not overvalued in 1996. This is just one of many examples of how human thinking is biased toward finding “bubbles” where they do not in fact exist.
Greenspan’s strongest attribute was his willingness to take seriously the information contained in asset prices.

Posts by Marcus Nunes, Lars Christensen, and Stephen Kirchner have good discussions of Greenspan’s legacy.

As a Milwaukee Bucks fan I cannot help commenting on the recent trade of Giannis Antetokounmpo to the Miami Heat. Casual fans have begun underrating Giannis in a couple ways. First, he gets blamed for the recent decline of the Bucks, which is actually due to the fact that 10 years of horrible drafting, bad trades, and bad injury luck produced a team lacking in talent. Second, people overestimate how often he’s injured. He only completely missed one playoff due to injury, plus a few isolated playoff games in other years, including the title year (2021.) In his 13 regular seasons, last year was the first time Giannis missed a lot of games, and that was partly due to the team holding him out when healthy in order to tank. He is an injury risk due to his age, size and style of play, but his injury history is not that bad. The amazing 2021 recovery from a severe knee injury right before the finals shows that he’s a quick healer. The betting market agrees with me—pushing Miami up to #5 in title odds (ahead of Denver), and disagrees with casual fans who view him as overrated. He’s clearly still a top 5 player, closer to #1 than #6. This will be the standard view by late November. (Boston understood, offering Jaylen Brown plus two firsts for Giannis.)

Last year, I discussed the fact that mixed race players were increasingly common in the NBA. I noticed that 5 of the top 13 picks in this year’s draft had one black parent and one white parent, including Milwaukee’s two picks (at #10 and #13.)

And now for my links:

The media won’t report this:

Oops, the Economist is part of the media, so I guess the media will report it.

Andrew McCarthy has a good piece on how to think about abuse of power. Hard to excerpt, but the key point is that illegal acts aren’t necessarily impeachable, and impeachable offenses aren’t necessarily illegal.
The US isn’t the only country shooting itself in the foot. Here’s Bloomberg:

China is restricting overseas travel for top AI professionals in private firms, requiring them to get approval from relevant authorities before embarking on overseas travel.
The government is targeting talent within the AI sphere, including startup founders, researchers, and executives, and adding individuals to the list based on assessments of their critical importance to the country.
The restrictions risk undermining the ability of AI firms in China to recruit and retain talent, and may force engineers with global ambitions to choose between staying home or going abroad earlier in their careers.

What sort of society would actually allow freedom of choice?

Walk into three different bars in Tokyo, and you may have three completely different experiences: one bar thick with cigarette smoke, one with a sealed glass smoking room humming in the corner, and one entirely smoke-free.
Within regulatory boundaries, the choice often lies with both the owner and the customer: Proprietors decide what kind of space they want to run, and patrons decide which environment they prefer.

When I’ve suggested that the US adopt this sort of system, I’m invariably told that it “won’t work”.

Words of wisdom from Matt Yglesias:

I wish labor would reflect more on the reality that almost all of the growth in the United States is happening in right-to-work states.
Some of that is because businesses prefer to invest in right-to-work states. But that’s not all of it. If it were the case that union-friendly labor laws had just devastated the economies of the Pacific Coast and the Northeast, then those states would be cheap. But demand to live in New York and New Jersey and Connecticut and Massachusetts and Maryland and California remains robust. If it were easier to build houses and apartments in those states, more people would live in them. If more people lived in them, a non-zero number of jobs would move with them. More jobs in states with union-friendly labor laws would create more opportunities to organize.
It would also mean much healthier pension situations for the large public-sector unions in those states.

The Bloomberg headline below caught my eye:

US Airlines See Robust Demand Even as Consumer Confidence Falls

Headline should read: US Airlines See Robust Demand Even as Surveyed Consumer Confidence Falls. I see no evidence that actual consumer confidence is falling—shoppers certainly are not acting as if they lack confidence.

Noah Smith reports that Claude is a neoliberal:

AI investor and founder Arram Sabeti recently asked Claude what policies it would enact in order to “fix everything” in America. . . . Claude’s answers were:
1.YIMBYism (upzoning, pro-housing deregulation)
Land Value Tax
Permitting/NEPA reform
Carbon tax and dividend
Repeal the Jones Act
Paying people to donate kidneys
High-skilled immigration
Reciprocal FDA approval agreements between rich countries
Reduce occupational licensing
Ranked-choice voting
This is pretty much just a list of neoliberal hobbyhorses.

A few comments. First, I entirely agree with Claude. Second, at least some of those views (selling kidneys, carbon tax/dividend, etc.) are unpopular.

Have you ever traveled to a strange foreign country where you know nothing about the culture? That’s what happens to everyone if you live long enough. In my email box I saw this headline and subhead from The Free Press:

Tough Love: Will I Get Canceled for Dating a Freshman?
A 22-year-old wants to ask out his 18-year-old classmate but fears campus backlash: ‘They’d label it predatory, and I’d be staring down the barrel of cancellation.’ Our advice columnist weighs in.

I did not read the article. I’d rather go to my grave not knowing what this is all about.

So blue states seem less suicidal:

And yet studies suggest that leftists are less happy. The left reminds me a bit of this quip from Annie Hall:

Life is full of loneliness, and misery, and suffering, and unhappiness, and it's all over much too quickly.

If Matt Yglesias is correct, then he and I comprise fully one third of all humans who have read Capital in the Twenty-First Century, cover to cover:

If you are one of the six people who actually read Thomas Piketty’s “Capital in the Twenty-First Century,” then you’ll know that the primary concern he raises in the book is specifically about this heir question.

Of course Piketty is French, and if a European Bill Gates tried to give $100 billion to charities fighting disease and poverty in Africa, instead of giving the money to his spoiled kids, they’d probably put him in jail. Speaking of which, this item in The Economist caught my attention:

Employment in the sector rose about 6% in 2024, the last year measured, according to Confindustria Nautica, an Italian yachting-industry association. It may yet stay afloat: one broker notes that America, the biggest market, is growing in spite of the oil crisis. Since last July, Uncle Sam considers vessels majority-owned by American firms tax deductible.

Tax deductible? Long time readers know that I have an unhealthy obsession with the idea of using a progressive consumption tax to shrink the size of superyachts by 25%. But even if I’m wrong, should superyacht buyers pay a 0% tax rate on purchases made with before tax income while rowboat purchasers must pay an 8% sales tax with purchases made using after-tax income?

Here’s The Economist:

This symbiosis between life insurers, reinsurers and private-asset managers (which these days often own or work closely with the other two) is dubbed the “Bermuda triangle”, after the Caribbean jurisdiction where it has blossomed.

I suppose “Bermuda” does sound a bit Caribbean. How about renaming the island Prospero, to avoid confusion? Rich countries should have rich sounding names (think Luxembourg and the UAE), and of course there’s a theory that The Tempest was loosely based on a shipwreck in Bermuda.

Or perhaps East Carolina?

12. This tweet caught my eye:

It’s worth mentioning that Orange County (southeastern area on the map) is far better governed than LA County. You notice the difference immediately when crossing into LA County (Long Beach) on the 405.

Utopia is living in a politically purple part of coastal California. (Unfortunately, the OC is more boring than LA—not for young people.)

The Economist has an eye-opening graph, especially given that Texas is a conservative oil and gas state:

A new paper by Xiwen Bai, Jesús Fernández-Villaverde, Yiliang Li, and Francesco Zanetti has this abstract:

We study how global supply chain disruptions affect monetary policy transmission. Post-pandemic evidence indicates surging transportation costs, goods-market imbalances, and rising prices. We develop a model in which logistical bottlenecks (upstream slack coexisting with downstream shortages) steepen the aggregate supply curve. This convexity amplifies price responses to monetary policy while dampening output effects. Threshold VAR and Local Projection estimates are consistent with this mechanism: during disruptions, contractionary policy reduces prices more at smaller output cost, easing the stabilization trade-off.

This makes sense, although I’d focus on the role of labor markets. The fact that there was a labor shortage in 2022-23 made it easier to bring down inflation with a more contractionary monetary policy. If a labor market goes from normal to weak you tend to get a recession. Going from overheated to normal doesn’t necessarily trigger a recession.

Matt Yglesias has a good analysis of the horseshoe theory of politics:
One is temperamental. Most people who work in politics have a kind of mainstream disposition. They probably have a few opinions that are way outside the consensus, but those opinions aren’t strongly held or aren’t about things that they feel passionately about. Their general approach to life is to try to make incremental progress on relatively mainstream issues. Extremists aren’t like that — they have an oppositional attitude toward authority and end up having that in common with each other.
Another is negative polarization. There are two kinds of people who really, deeply, and profoundly hate Hillary Clinton, Barack Obama, Chuck Schumer, and Hakeem Jeffries: Republicans, and people on the far-left. This ends up being something that they have in common with each other. Pair a far-leftist up with someone from the far-right who hates mainstream Republican Party leaders and they end up having enemies in common, which is a good way to start making friends.
A third is ideological. There is a “liberal center” to American politics that believes, fundamentally, in a positive-sum cooperative world. Then there are illiberal zero-sum tendencies floating around that tend to sort themselves on the basis of whether they blame foreigners for everything or whether they blame rich people for everything, but they have that zero-sumness in common. And of course many things — like trade — implicate both foreigners and rich people.
We frequently read about China’s massive military build-up and Europe’s pathetically low levels of spending. Thus, you may be surprised to learn that China’s military spending has fallen below 2% of GDP at a time when the EU’s military spending is up to 2.1% of GDP, and is still rising.

The two regions have roughly equal size economies, although China’s is larger in PPP terms.

Very good news from the FT:

Swiss voters have rejected a proposal to cap the country’s population at 10mn people, delivering a surprise defeat to a rightwing initiative that had appeared neck and neck in opinion polls until just days before the vote.
Projections on Sunday showed the initiative losing by roughly 54 per cent to 46 per cent, a significantly stronger result for opponents than suggested by recent polls, which had indicated a tight race in the final days of the campaign.
The proposal, backed by the rightwing Swiss People’s Party (SVP), would have made Switzerland the first country to impose a formal cap on its population. It sought to limit the number of residents to 10mn, from about 9.1mn today, and would have required the government to take measures to curb population growth once the population reached 9.5mn.

Iran War analysis from Noah Smith:

Before the war, Iran didn’t control the strait, simply because it didn’t realize it could. Drone technology had advanced to the point where Iran was able to shut down Hormuz, but Iran didn’t know that until the U.S. attack forced it to try the risky and desperate move of actually shutting down the strait. The gambit paid off spectacularly, and now Iran knows that modern drone weaponry gives it an advantage it didn’t have in previous decades. So it controls Hormuz.
It’s kind of wild to step back and consider how good of a position Iran’s leaders are in now, compared to the situation before the war. Iran had lost most of its proxy armies in the Middle East — Hezbollah, Assad, most of Hamas. The regime had been rocked by massive nationwide protests, which it only managed to quell by murdering tens of thousands of innocent Iranian citizens. The country’s economy was slowly dying. Now the leaders are firmly entrenched in power, their economy will be revived, and they find themselves the masters of Hormuz for the first time.
. . . A final silver lining is that the U.S. may step back from the Middle East in general, as I’ve long been urging.

And from Matt Yglesias, you can’t make this up.

Jessica Riedl has an excellent piece explaining why AI is unlikely to solve our budget problems:

Unfortunately, like all the other "easy solutions" to budget deficits, AI is highly unlikely to produce the trillions of dollars in annual fiscal savings necessary to avert the hard decisions. Betting America's fiscal future on AI is wishful thinking.

The job impact of AI is overrated:

Two main economic concepts apply to the rise of AI in the legal industry: the lump of labor fallacy and the Jevons paradox.
The lump of labor fallacy is the mistaken belief that there is a fixed amount of work to be done in the economy. Under this view, if AI eliminates a category of tasks, the people who did those tasks are simply left without work. History tells a different story. Many of the ten most common jobs from the early 20th century no longer exist in any recognizable form. Yet employment has expanded dramatically. Many of the jobs that exist now, including roles like AI data scientist or prompt engineer, didn’t exist then because the conditions that created them hadn’t yet arrived. Every major technological shift ultimately results in a larger total number of jobs, even when it disrupts specific roles.
Meanwhile, the Jevons paradox states that when something becomes cheaper and more efficient to use, total consumption of it tends to increase, not decrease. The original example involved coal. As steam engines became more efficient, requiring less coal to operate, there was a widespread expectation that coal consumption would decrease. Instead, it multiplied. Cheaper engines led to more engines being used in more places.

No idea what this means, but FWIW:

Works in Progress is full of great articles. One of them describes Spain’s successful urbanism (until recently):

For the price of one mile of the New York’s Second Avenue Subway extension, Spanish builders covered the entire 35-mile 1995–1999 expansion of the Madrid Metro.

Adrian Wooldridge on Swedish reforms:

A succession of neoliberal governments re-engineered the model from top to bottom. Sweden reduced its public spending as a proportion of GDP, cut the top marginal tax rate radically and dismantled taxes on property, gifts, wealth and inheritance. It introduced a universal system of school vouchers, invited private schools to compete with public, allowed private companies to provide state-funded health services and care for the elderly and donned the golden straitjacket of fiscal orthodoxy, keeping a downward pressure on debt and deficit, and swapping a retirement system with defined benefits for one with defined contributions while making automatic adjustments for longer life expectancy. All more Milton Friedman than Bernie Sanders.

A sentence I never expected to read:

Corporate officials who “falsely represent” their business as gay face up to a year in county jail.

Oscar Wilde would have had something clever to say about that.

From the FT:

Russia has accused the US of abandoning efforts to broker an end to the war in Ukraine after Donald Trump appeared to be shifting again in favour of Kyiv.
Russian foreign minister Sergei Lavrov on Tuesday said the US was “seemingly stepping back from the role of an objective mediator” in the war and had “forgotten” about Trump’s own statements last year inching towards Moscow’s position.
Instead, Trump was “hugely impressed and enthusiastic” about Ukraine’s recent campaign of long-range strikes on targets deep inside Russia at last week’s G7 summit, said two people briefed on the private discussions among the leaders. Trump at that summit also agreed to increase sanctions on Russian energy.

Trump values strength and the Ukraine is getting stronger. Ethics? Not so much.

Most people I speak with believe that saving Social Security will require a combination of tax increases and benefit cuts. It would seem that those “most people” come from a rather unrepresentative group, perhaps 5% of the population:

It is not just Washington inertia that makes reforms tough. In polling conducted by YouGov for The Economist, 71% of respondents believe Social Security spending should be increased, more than for any other category of government spending. In the same poll, 45% say it should be increased a lot. The proportion who wish it to be reduced is just 5%, slightly less than the share of Americans who believe that covid-19 vaccinations were used to microchip the population.

Only 5%? Now suppose you asked people if they favored raising taxes. How many would say yes? Suppose you asked people how many favored allowing Social Security to go bankrupt. How many would say yes? Would the three options add to 100%?

Or is a better interpretation of the data that there is no such a thing as public opinion.

Why is Matt Yglesias my favorite center left pundit? In back-to-back tweets he nails the essence of what’s going on. First this:

Chicago’s left already staged a successful insurgency, but the incumbent left-wing mayor who led the successful insurgency is an unpopular failure so people don’t talk about him anymore.

Question is whether the reformist counterinsurgency will spread from SF to the Windy City.

And then this:

There’s been an interesting trajectory since Bernie’s 2015 era pitch that “democratic socialism” meant “policy like Denmark” to the current situation where left policy proposals look nothing like actual Danish tax policy.

We’ve gone from “be more like Denmark” to “be more like Venezuela”. At the same time, the right has gone from “be more like Switzerland” to “be more like Hungary”.

I generally agree with “The Unpopulist”, but not this time:

The Unfortunate Necessity of Court Packing to Stop America’s Authoritarian Drift

Court packing is authoritarianism.

I refuse to insult the Turkish people by calling their country “Türkiye”. Only banana republics insist on foreigners calling their country by the local name. It’s a sign of weakness and insecurity. You don’t see residents of places like Germany, Spain and China insisting that we call their country Deutschland, España and Zhōngguó.
Just one more reason (out of dozens) why JD Vance should never be president.
What would you say if someone invited you to this Orange County restaurant?

No . . . ?

America's non-violent future

Scott Sumner — Fri, 19 Jun 2026 18:10:50 GMT

[Decompressing after travel—I hope to get back to regular blogging before long. A belated congratulations to Knicks fans, as their 1970 championship was the first one I remember watching. As a Bucks fan I feel a connection—we won in 1971 and then also had a 50-year wait for another title.]

Part 1: Apart from the 2020 surge in murders, a period when police sort of gave up on crime for a brief period, the US murder rate has been declining since 1991. I expect this decline will continue, and within about 20 years the US will have Canadian rates of murder. More specifically, we’ll have a murder rate close to the current Canadian rate of roughly 2 per 100,000 people. I anticipate that the Canadian rate will also decline further and thus remain below US levels in 2046.

In the US, the murder rate peaked at 10.7 per 100,000 in 1991 and fell to only 4.29 per 100,000 in 2025. Early figures suggest that 2026 is on pace for a rate of roughly 3.8 per 100,000. (And no, it’s not due to better medical care.)

While the data from early American history is rather fuzzy, I suspect this is the lowest rate in US history (based on some highly scientific analysis on my part—watching lots of gunslinger movies as a kid.) Seriously, I expect the rate will go much lower, for at least four reasons:

The level of births was 4.32 million in 2007, and has been falling ever since. That means the number of 19-year old Americans is virtually certain to fall sharply over the next 20 years. And young men are especially likely to be involved in violent crime. Also note that the fertility rate of African Americans is falling especially rapidly and is now below the rate for white Americans.
Declining crime tends to feed on itself. With fewer murders, each crime stands out more and attracts more police attention. That leads to a greater chance of being apprehended, further discouraging crime.
The sort of surveillance technology that has radically reduced crime in Chinese cities like Shenzhen is likely to eventually become a feature of American cities. Artificial intelligence will play a role, as will ubiquitous cameras in public spaces, the tracing of location via cell phones, automobile computers, and related technologies, as well as better DNA testing. Robocops will patrol our streets. The world of 1984 is arriving faster than I expected. Big Brother will be watching us all in 2046.
The sort of lifestyle changes that have reduced fertility rates are also likely to reduce crime rates. When I was young, bored young men stood around on street corners, which is a recipe for crime. Now they can enjoy all sorts of electronic entertainment in the comfort of their home. These forms of entertainment also increase the cost of incarceration, where Netflix may not be available. Alcohol use is declining and being partially replaced by pot use, which is less likely to provoke violence.

If you believe that my claim is far-fetched, consider that just a single factor—demographics—is likely to reduce crime committed by young men by roughly 15% to 20% by 2046. That alone would get the murder rate down from the 2026 estimated 3.8 to roughly 3.2 per 100,000. And as recently as 2021, the rate was 7.75 per 100,000. Much of the progress toward Canada’s (current) murder rate has already occurred, even though America still has about hundred gazillion guns. (By the way, I’m not a gun nut—some restrictions may be sensible—just saying that guns are not the primary factor.)

I haven’t even discussed immigration, which clearly plays a role in the especially rapid decline in murder rates in places like NYC.

To be clear, I’m not suggesting that a rate of even 2 per 100,000 is ideal—many European countries have rates of 1 per 100,000, or even less. But it would represent a dramatic reduction from the rate of nearly 10 per 100,000 during much of the 1970s and 1980s. Speaking of 2046, who can forget this gorgeous Wong Kar Wai film:

Part 2: I cannot prove this, but I strongly suspect that drug addiction in America is on the way down. Here are my claims; you tell me if I’m mistaken:

There is less alcoholism than in the past
Smoking is declining fairly dramatically
Opioid drug abuse is declining
Marijuana abuse is increasing

Now I’d like you to consider two different social science hypotheses and tell me which one better fits the data:

Pot is a “gateway drug”, which leads to even more harmful forms of drug abuse.
Pot is a substitute for other types of drugs, and legalizing pot would tend to reduce other (more serious) forms of drug abuse.

Behind the paywall, I’ll offer a few thoughts on the War on Drug Using Americans:

The silent revolutionaries

Scott Sumner — Wed, 10 Jun 2026 13:38:11 GMT

I am traveling this week and don’t have time to write a serious post. Instead, I’ll provide a take that is probably stating the obvious, but I don’t recall reading elsewhere. And since it’s fairly short I’ll skip my normal practice and deliver two free posts in a row. First, a few facts:

The decades after WWII saw many cultural, political and economic changes. The birth control pill and the loosening of divorce laws. Feminism and civil rights. Rock and roll and drugs. Soft on crime policies and a sharply rising crime rate. The end of the gold price peg and soaring inflation. The welfare state and affirmative action. A dramatic expansion of (unfunded) Social Security benefits. Environmentalism and NIMBYism. The most important changes began during the 1960s.
The decades after WWII saw a notable baby boom, generally dated from 1946 to 1964. I was born in 1955, smack dab in the middle of that period. I am a typical baby boomer.

Now I’m going to say something that might be controversial but is obvious when you think about it. I am not personally to blame for all of the cultural, political and economic policy changes of the 1960s.

I say this because I frequently see boomers being blamed for every single ill in modern society. The peak period of change was roughly 1965, sometimes called “the liberal hour”. I was ten years old. Not a single baby boomer was out of their teens. If you wish to blame a generation for all the ills of modern society, please blame the Silent Generation and the Greatest Generation. They got rid of traditional morality and pushed divorce rates much higher. They put Social Security on an unsustainable path. They ended the gold price peg for the dollar. They created affirmative action and NIMBYism. The ended the death penalty. Heck, they even invented rock and roll.

I also see people suggest that boomers are the lucky generation. No, it is smaller generations that are lucky. Big generations face a highly competitive job market. In 1982, I was paid $19,700/year as an assistant professor, at a time when the unemployment rate was 10%. Even in real terms, starting salaries for young Gen X professors had moved far higher by the late 1990s and early 2000s. It was mostly the silent generation and perhaps a very few early boomers that left college and entered a strong job market during the 1960s.

Even worse, my generation entered the job market at a time when the feminist revolution shifted many women from being housewives to career women, so labor force growth was even higher than working age population growth:

Real wages stagnated in the 1970s and 1980s.

The Supreme Court’s “soft on crime” rulings in the 1960s and early 1970s were not made by boomer justices. The feminist revolution was led by silent generation types like Gloria Steinem. The population control movement was led by Paul Ehrlich, also a member of the silent generation. Martin Luther King was a member of the silent generation. Please don’t assume that things that happened during the baby boom (and even soon after) were caused by baby boomers. Last time I looked, Congress is not composed of young children:

The 1976 election was the first one where I could vote. I recall pundits being surprised by the substantial support for Gerald Ford in college areas, as younger voters were already starting to edge back toward the center. The student radicals of the 1960s were often silent generation members. In my hometown of Madison, the antiwar movement was led by Paul Soglin, who later become mayor. I thought of him as a boomer, but he was from the final year of the silent generation.

In my mind, rock music is associated with boomers. But rock and roll was developed in the mid-1950s, and even the second wave (Beatles, Beach Boys, the Stones, The Who, Dylan, Hendrix, Paul Simon, Lou Reed, etc.) were silent generation people. The film Bob & Carol & Ted & Alice nicely shows how silent generation members led the cultural change of the 1960s. The boomers just followed along.

I’m not saying that younger people should not despise the boomers, but please do so for the right reasons. Hate them for taking credit for innovative sixties pop music that was created by silent generation musicians. Hate them for their current selfishness on issues like NIMBYism and Social Security. Hate them for their hypocrisy—romanticizing their youthful adventures while opposing drug legalization. But please don’t hate us for all the changes that occurred during the 1960s (some of which were good, BTW.)

As an aside, in Where the Music Had To Go, Jim Windolf points out that teenagers at Beatles concerts in 1964 screamed so hysterically that the music could not be heard, while older Dylan fans sat silently listening to acoustic folk music. So the terms “silent” and “boomer” fit these two generations in more ways than one. Windolf says that Dylan fans despised the “teenyboppers”, but does not say what Beatles fans thought of the folkies. Perhaps it was one of those “I don’t think about you at all” situations.

Time is continuous and the relationships between people and events are often not what they seem. Lines between generations are arbitrary. My mother was born June 7, 1926 (six days after another pretty lady), near the end of the Greatest Generation. But she was mostly in high school during WWII, so culturally she’s perhaps a bit like the Silent Generation. When she was a few days old, she was closer in time to the last days of Thomas Jefferson and John Adams (who both died on July 4, 1826) than to her current life. She’s experienced more than 40% of American history. For some reason that fact boggles my mind, as to me she’s sort of timeless, an eternal “mom”.

Happy 100th birthday.

PS. My favorite poster advertises a festival in Granada, Spain. Decades after I bought the poster, I noticed the date. Ever since, 1926 has been my favorite year:

The world is bigger than you can imagine

Scott Sumner — Thu, 04 Jun 2026 16:44:20 GMT

This post will begin with some silly dorm room philosophy, but eventually I’ll extend the idea to something that I’m actually qualified to discuss—the macroeconomy.

It has become an overused cliche that the universe is so vast that the puny human mind is unable to even imagine its scale. So let me begin with a claim that might be more controversial; a human life is so vast that a person cannot reasonably evaluate their own life. Try to sum up your life? It’s probably impossible.

It sounds a bit arrogant for me to tell other people that they cannot evaluate their own life, so I’ll stick to my own example. If it doesn’t apply to you, that’s fine. I won’t try to press the point. Perhaps I’m different.

In your vision, there is a blind spot that you don’t notice. The brain sort of fills in the gap, making it seem like you see the entire field of vision. I suspect that life is sort of like that. My mind gives me the impression that I can look back on my entire life, but all I’m seeing are a few tiny fragments.

When I was nine years old, time seemed to go by very slowly. My age nine life must have been vastly richer than my age 70 life, with lots more happiness and pain, boredom and drama. It was one of the most important parts of my life, at least in “utility” terms. And yet I recall almost nothing of that year, perhaps just one or two dozen events. But that’s not even a single day for a nine-year old!

Here’s another way of looking at things. I sometimes try to categorize my life by type of event. But the more I look, the more I see. You have school, and work, and time spent with friends. Then there are the days where you are sick. There are days spent traveling. That includes both lots of vacations and lots of travel for work. And time spent at work is itself quite diverse, featuring lectures, research, office hours, committee meetings, lunch conversations, commuting time, etc. And those are just the tip of the iceberg.

At age 70, I will occasionally recall some important past event that I had completely forgotten. Clearly, I had not completely forgotten it, but the memory was inaccessible most of the time, including those times when I try to look back and evaluate my life. The sudden recollection of one of those deeply buried memories triggers a feeling analogous to deja vu, but different. It triggers the feeling that life is much bigger than I had thought, that much more has happened to me than I can recall at any one time. The Dictionary of Obscure Sorrows coined the term “sonder” for this feeling.

There’s also a vast amount of time spent on things that are more passive than play, work and school. Thus, I’ve read many books and seen even more films. How should I think about the part of my life immersed in alternative worlds through the narrative arts? This is from the Taiwanese film YiYi, which I’ve seen twice:

Pangzi: My uncle says we live three times as long since man invented movies.
Ting-Ting: How can that be?
Pangzi: It means movies give us twice what we get from daily life.

I occasionally think about those medieval peasants that were not able to go into a dark theatre at age 12 and have their minds blown by the film 2001: A Space Odyssey. (Substitute Star Wars if you are a decade younger, or The Matrix if you are Gen X.) I’ve spent a lot of my life immersed in films, many of which are far more engrossing than my “real life”. How should I think about that time?

When most people evaluate their lives, they probably think about career success or failure and their family life. That’s fine, but is it possible that we underrate the importance of all the time we spent on hobbies like collecting stuff or watching sports or listening to our favorite music in college or using drugs or hiking in the wilderness or lots of other “trivial” pursuits? Are we actually evaluating our life, or are we evaluating the official version of the most important parts of life? In some ways, artists like Borges, Conrad, Velazquez, Cezanne, Kubrick and Dylan seem more real to me, a bigger part of my life, than many casual friends I’ve had over the years.

The film It’s a Wonderful Life probably describes the way that many people visualize looking back on their lives. But an honest appraisal of my life would have to include It’s a Wonderful Life. The two hours spent watching that film (big screen of course) was much more engrossing than the two hours right before of after I watched the film, when I probably did something boring and eminently forgettable. If someone asked me to “look back on my life”, I would not start by saying “Well let’s see, I watched It’s a Wonderful Life”, as that film portrays someone else’s life. But it’s also a part of my life!

After writing this, I started to recall all sorts of minor things I’d overlooked. For instance, initially I thought of “work life” as my 35 years at Bentley University. But then I recalled that I also spent lots of time doing physical labor, especially when young but to some extent throughout my life. I suspect that I did at least an order of magnitude more physical labor than most other college professors—so that’s also an important part of my life.

Some stuff that I recall seemed normal at the time, but weird in retrospect. At age 13, I would see my (divorced) dad every few weeks. We’d sit at the bar, and I’d drink soda (then called “pop”) while he consumed six or eight scotch and waters, and we’d talk about all sorts of stuff. Then he’d drive me home! (Blood alcohol level? Don’t ask.) He’d occasionally give me 40 cents to buy him a pack of Kents from the cigarette vending machine. Or at the same age, hanging around the rail yards at night with my friends, dodging the police while collecting flares to use in making fireworks. That’s part of my life too.

I honestly don’t know what sort of life I’ve had. If I look back on it when I’m feeling down, I see a vastly different life from when I look back on it when I’m happy. So, which is true? What if both are false?

Again, perhaps this agnosticism about life is just me, maybe I’m weird. But now let’s look at two issues where I do have some expertise, evaluating economic problems and evaluating economic policy regimes. Most of the bad analysis that I see in these two areas ultimately derives from a single mistake—vastly underestimating the size of the economy. Not in the sense that GDP is larger than measured due to the underground economy, rather in the sense that the economy is far more complex and diversified that it seems at first glance. I’ll illustrate the two claims, one at a time:

Part 1: There’s a great deal of ruin in a nation

I received an email from The Economist magazine, which had the following admission:

For the past year America has been running a giant experiment. What happens when you take one of the world’s most successful economies and put in place two big growth-busting changes: tariffs and a sudden stop in immigration? Many observers—including us—thought the impact would be unambiguously negative. Yet America’s economy is still the envy of the rich world. The country’s GDP grew faster than that of any other G7 country last year. Its stockmarkets have repeatedly hit record highs over the past 15 months. What is going on?

In this week’s cover story, which will be published tomorrow, we offer an answer. We have crunched the numbers. An artificial-intelligence boom, soaring stockmarkets and some of Donald Trump’s own measures have boosted the economy. But by our calculations, the president’s chaotic policymaking, sweeping tariffs and mass deportations are, in fact, hurting growth—a drag we call the “MAGA tax”. Our calculations suggest that without the MAGA tax, America’s economy would have grown more than half a percentage point faster.

I suspect that many people overestimated the negative impact of Trump’s policies because they forget how complex the economy is. If Trump’s policies reduced real output by $300 billion, they would be enormously destructive. But that’s only one percent of GDP and could be easily obscured by offsetting effects related to monetary policy changes, an AI boom, and other factors. We saw this in 2007, when the US economy continued to expand despite a dramatic collapse in home construction. Even at its peak in early 2006, homebuilding was only 6% of GDP.

In general, people tend to overestimate the impact of any single real factor on the US economy. But there is one type of shock that people underestimate—nominal (monetary) shocks. Because wages and prices are sticky, nominal shocks affect not just a single sector like homebuilding, rather they impact all markets. That’s why almost all recent US recessions (apart from Covid, which only lasted 2 months) have been caused by contractionary monetary shocks.

This also explains why there are so many false alarms, forecasts of recession that don’t pan out. People wrongly assume that a problem facing one particular sector will cause a decline in the overall economy. In a large and diversified economy, that is rarely the case. Again, Covid being a notable exception.

The fact that there is a great deal of ruin in a nation has implications for a wide range of issues. In a post entitled “Supply is elastic, installment number 1637”, Tyler Cowen cites the fact that our policy of murdering suspected drug smugglers in the Caribbean does not seem to have slowed the supply of cocaine, and the cutoff of oil from the Persian Gulf did not cause the European airline industry to shut down in 6 weeks (as had been predicted). Numerous other examples of surprisingly elastic supply could be cited, such as the almost universal failure of “sanctions” to have their desired effect. Doomsday resource depletion or overpopulation scenarios are often the result of an excessively simplistic view of the economy.

Do you recall the maxim that the best forecast is usually more of the same (which has allowed Bryan Caplan to win almost all his bets)? This is because the vast size and complexity of the economy results in no single factor having much effect.

Part 2: Policy regimes are more complex than they appear

It isn’t just the overall economy that is really big, economic policy regimes are also very large and complex. In previous blog posts, I’d occasionally point out that Singapore has one of the freest market economies in the world. They are number one in the Heritage ranking and number two (slightly behind Hong Kong) in the Fraser ranking. Commenters occasionally responded that Singapore’s government is highly interventionist, citing numerous examples. So, who’s right? Queue the “Why not both?” meme.

I asked AI Overview to estimate the number of regulations in the US:

Because there is no single agreed-upon definition of what constitutes an "economic regulation," researchers estimate there are well over 1.1 million distinct federal regulatory restrictions, filling roughly 130,000 to 190,000 pages in the Code of Federal Regulations (CFR). [1, 2, 3]

And that’s not even accounting for state and local regulations such as rent controls, zoning rules, minimum wage laws, occupational licensing laws, building codes, etc., etc.

I wonder how many regulations there are in North Korea? If the answer were 800,000, would you conclude that North Korea has a less regulated economy than the US? I hope you see my point. Modern economies are very difficult to summarize as “free market” or “socialist”. Not all regulations are equally important.

Most economies are very large and complex, and even relatively free market economies have extensive government intervention. This complexity allows people to find evidence to support almost any position on the question of optimal policy regimes. In a country that is viewed as relatively successful, it is almost always possible to find a few examples of policies that you favor. Then you merely need to claim that these policies explain the success of your chosen example.

Similarly, in a country that is widely viewed as economically unsuccessful, it is almost always possible to find specific policies that you disagree with. Then you can claim that these bad policies explain why the country is doing poorly. Due to motivated reasoning, I suspect that we usually overestimate the impact of any single policy on a country’s overall policy regime.

You can think of the wealth of a country as being due to three factors:

Natural resources.
Human capital.
Economic system (taxes, property rights, regulations plus cultural traits such as corruption, interpersonal trust, etc.)

Beyond oil and gas, natural resources don’t seem to have a dramatic impact on per capita GDP. Student scores on international exams provide a rough estimate of the average level of human capital. The third factor is the hardest to estimate. How in the world do we measure the extent to what a policy regime encourages wealth creation? If the US has 1.1 million regulations, how can we possibly make sense of such a complex picture?

I don’t know whether I have any persuasive answer to this question but FWIW, I rely on two primary pieces of evidence, natural experiments and basic economic theory. I look at what happens when countries undertake major changes in their policy regime, and I look at what economic theory predicts about the relationship between economic regimes and wealth creation. For instance, I notice that:

The sluggish West German economy took off after ending price controls in 1948.
The sluggish South Korean economy took off after removing some barriers to trade in 1964-65.
The sluggish Chinese economy took off after improving property rights in rural areas after 1979.
The sluggish Polish economy took off after extensive privatization in 1990.
The sluggish Indian economy took off after significant deregulation in 1991.

And there are many other similar examples.

I also look at the predictions of economic theory. In most cases, theory predicts that efficiency results from having prices freely move to a competitive equilibrium. Theory predicts that low marginal income tax rates on labor and capital encourage wealth creation. Theory predicts that free trade is more efficient than protectionism. Theory predicts that wealth creation will improve if people have secure property rights over their land and physical assets. This is roughly the package of policies that are labeled “economic freedom” by Heritage and Fraser.

So, we have a set of economic models that mostly (not always) predict that freer markets are more conducive to wealth creation, and we have lots of historical examples of economies taking off after reforms increased “economic freedom”. But it’s complicated, as a country like Denmark scores extremely high on free markets but also has a large welfare state. Intervention is multidimensional.

It is also important to avoid motivated reasoning. In some cases (such as Chile), it took a long time for free market reforms to produce higher output. As Marcus Nunes recently discussed in an excellent post, Milei’s reforms in Argentina have thus far produced only very small gains. Economies are quite complex and it is not always clear whether any single reform, or set of reforms, will prove to be truly transformative. There’s a lot we don’t know.

Let’s conclude with a few other examples of things being bigger than we imagine:

Physical size of Earth: I few years ago, I visited “Africa” and went on a safari, staying at four different hotels. Well, actually just one of Africa’s 54 countries—that’s more countries than America has states. After the trip, I looked on the map and saw that the seemingly vast area I covered was just a tiny part of Tanzania, up near the Kenyan border. I know nothing about the rest of Tanzania. And did I even see that small region? Land travel is basically two dimensional, along the narrow thread of a road. To the right and left you could travel 10 miles, 20 miles, or 50 miles, and there would be thousands of villages that I never saw, untouched by tourism.

But hey, I’ve been to “Africa”!

[I traveled in the rectangular area northwest of Arusha.]

People occasionally do European vacations. (If It’s Tuesday, This Must be Belgium.) More sophisticated travelers will visit a single country, say Italy. I once knew a couple that vacationed in Italy once a year, seeing just one province on each trip. But can you even do Tuscany in one trip? Isn’t it better to do Florence on one trip, then Siena & Arezzo, then Lucca & Pisa, then the hill towns, then Etruscan sites, etc. And even Florence might be too much. Spend one trip exploring the buildings and statues designed by Brunellesci and Michelangelo, another closely examining paintings in the Uffizi and Pitti palaces.

Life: The diversity of plant and animal life is almost beyond comprehension. Not just in big countries like the US and China; even medium size countries like Colombia have an amazing variety of local ecosystems.

Culture: People often say, “The Japanese are basically like this” or “The Italians are basically like that.” No, they are like 100 different things, often seemingly contradictory.

History: History is full of stuff to support almost any perspective. At one time, it was believed there was a Dark Age from 500 to 1000. Then revisionists found evidence suggesting there was no Dark Age. Then even more evidence was found, suggesting that yes, there was a Dark Age after all. Genetic research into man’s past keeps making the picture look more and more complex. You could spend your entire life reading only history books and never run out of surprising new information.

Literature: Read all the classics? The deeper you look, the more great writers you find.

Film: There are at least 50 to 100 films deserving of being in the top 10.

At least 1000 films deserving of being in the top 250. In other words, try as hard as you might in developing your list of the top 250, and I’ll easily find another 750 that are better than many films on your list.

The internet: How many great substackers have I yet to discover?

Here’s a recent Bloomberg headline:

Odd Lots: How the Invention of Rope Gave us Modern Civilization

How many words can you think of to replace “rope”, which leave the claim roughly equally plausible? The wheel? Farming? Writing? Printing press? Gunpowder? Steam engines? Banks? Roads? Sailing ships? Cement? Religion? Steel?

How many more? A hundred more? A thousand more? I can only estimate the number that immediately come to mind, not the vastly larger number that don’t immediately come to mind.

We only know what we know and vastly underestimate the universe of things that we don’t know.

PS. Here’s Scott Alexander:

I acknowledge that my inability to marvel at a live Caruso opera in Naples has cost me something deep and beautiful. But I cannot wish that the phonograph was never invented. Does the increased variety and quantity of music compensate for the decreased profundity of each musical experience? Surely this is part of it, but I would never accept this excuse in other areas that have not yet been cheapened. A thousand moderately pleasant one-night-stands cannot equal one passionate love affair.

I certainly would agree that the thought of a thousand moderately pleasant one-night-stands cannot equal the thought of one passionate love affair.

But the reality? Again, we only know what we know, and vastly underestimate the universe of things that we don’t know.

How about a thousand and one night stands, each accompanied by a clever story?

The South China shock and the world's biggest rustbelt

Scott Sumner — Mon, 01 Jun 2026 17:41:30 GMT

During the 1990s, China adopted a policy of “shock therapy”. If you get your information from fashionable pundits, you may not know that. You might have read that China avoided the “tragedy” of places like 1990s Poland by adopting a policy of gradual reform that avoided radical changes. Not true, China adopted a policy of shock therapy that led to one of the most dramatic examples of creative destruction in human history.

During the 1990s and early 2000s, China suddenly privatized thousands of bloated state-owned enterprises (SOEs), laying off tens of millions of workers. In aggregate, the policy was a big success. The restructuring made China’s economy much more efficient, contributing to rapid economic growth. But as is always the case with creative destruction, there were losers. China’s economic reforms cost lots of jobs in their “rustbelt”.

When most people hear the term rustbelt, they often think of an influential paper by Autor, Dorn and Hanson:

China’s emergence as a great economic power has induced an epochal shift in patterns of world trade. Simultaneously, it has challenged much of the received empirical wisdom about how labor markets adjust to trade shocks. Alongside the heralded consumer benefits of expanded trade are substantial adjustment costs and distributional consequences. These impacts are most visible in the local labor markets in which the industries exposed to foreign competition are concentrated. Adjustment in local labor markets is remarkably slow, with wages and labor-force participation rates remaining depressed and unemployment rates remaining elevated for at least a full decade after the China trade shock commences. Exposed workers experience greater job churning and reduced lifetime income. At the national level, employment has fallen in U.S. industries more exposed to import competition, as expected, but offsetting employment gains in other industries have yet to materialize.

The first half of that final sentence is true, but the second half has not held up well. Subsequent research does not support the claim that China reduced aggregate employment in the US, and indeed the US job market improved during the period they studied (1991-2007).

More importantly, the China shock literature misses the bigger story. The hardest hit area was China’s “rustbelt”, concentrated in the northeast region of the country. This region is known as the Dongbei, or Manchuria, and contains 100 million residents. Jordan Schneider directed me to a post by Zilan Qian that describes what happened:

The country’s enterprises, built for a planned economy, were suddenly exposed to market competition — and consequently began hemorrhaging money, especially in industries like steel and textiles. By 1997, the state had decided to consolidate the strategic enterprises and let the rest restructure, merge, or collapse. The slogan it coined was 减员增效 (jianyuan zengxiao) — “reduce headcount, increase efficiency.”
The consequences of this transformation depended on where you lived. Over 24 million workers in China lost their jobs in the state sector by the end of 1999. The layoffs were concentrated in the northeast — Liaoning, Heilongjiang, Jilin — once the industrial heartland of socialist China and now called China’s rust belt. In 1957, the [Manchurian] city of Shenyang’s Tiexi district produced the nation’s entire output of lathes, rock drills, gliders, rubber boats, and tower cranes, earning it the nickname “the Eastern Ruhr.”
By the late 1990s, 80% of the companies responsible for this output had gone out of production, and half of the district’s 300,000 industrial workers had been laid off. Between 1998 and 2000, nearly every year saw 7 to 9 million workers laid off nationally.

That’s right, the world’s biggest rustbelt was in China, and the majority of victims of the “Chinese shock” were Chinese workers. That’s how creative destruction works. At a time when Ohio had 5.6% unemployment (in 2007), Manchuria was in the midst of an economic depression.

If that’s the destruction, where was the creation part of creative destruction? Here:

Yet while the transition led northern China into economic crisis, the Pearl River Delta — geographically proximate to Hong Kong and Macau, home to China’s first Special Economic Zones, and the ancestral homeland of much of the Chinese diaspora in Southeast Asia and beyond — embraced rapid modernization and internationalization. The historical “land of fish and rice” became the “world factory.” Hong Kong investors established over 65,000 factories, employing about six million workers in the Delta. From 1991 to 2001, the Pearl River Delta’s regional GDP grew almost eightfold, and its population increased from 20 to 43 million.

Today, the Pearl River Delta has 86 million people, more than double the population of Jakarta, which is the world’s most populous metro area. The PRD is a bit too spread out to be viewed as a single metro area, but it’s a close call. The region has over 70 million people even if you remove Zhaoqing, Huizhou and Jiangmen from the light grey area on the map, and focus on the core cities:

(I plan to visit the area later this year.)

Recently, I’ve noticed an increasing number of people worrying that China is causing de-industrialization in other parts of the world. I believe these fears are exaggerated. Below the paywall, I’ll comment on arguments made by Matt Yglesias, the Peterson Institute, and Soumaya Keynes.

The Pursuit of Happiness

What is an undervalued currency?

August 2026 links

Trump-backed candidate in battleground Michigan district loses to opponent who dropped out

Trump Supporters Imperil $4 Billion Aluminum Plant Central to His Industrial Push

Donald Trump wants Washington museums to be less ‘woke’ on US slavery

Say’s Law in China

On the bigness of China

Our monetary Rube Goldberg device

John Cochrane on monetarism

Basil Halperin on sticky wage models

Thoughts on inflation and monetary policy

July 2026 links

Not just a cheap date

Fable responds

Arguments in favor of the thesis (HANK > market monetarism)

Arguments opposing the thesis (market monetarism > HANK)

Verdict

Why is Fable a Keynesian?

The empire strikes back

Arguments for the thesis

Arguments against the thesis

Verdict

Kakistocracy

Films of 2026:Q2

Alan Greenspan, RIP

The Unfortunate Necessity of Court Packing to Stop America’s Authoritarian Drift

America's non-violent future

The silent revolutionaries

The world is bigger than you can imagine

Odd Lots: How the Invention of Rope Gave us Modern Civilization

The South China shock and the world's biggest rustbelt