Decoupling is not deglobalization
"Protectionism vs. free trade" is not a good way of thinking about the changes facing the world economy.

As the push for decoupling with China gains steam, a number of voices in the financial press and international economic organizations are sounding a note of concern — or even panic. For example, here’s James Bacchus, a former WTO official and current Cato Institute adjunct:
Inescapable nowadays is the cascade of commentary announcing (and often celebrating) the “death” of globalization. Economic globalization is variously described as in “decline,” in “retreat,” in “reversal” and at an “end.”
To hasten the fait accompli of this “deglobalization,” we are “decoupling,” “reshoring” and “friend-shoring” in ways that will further fragment the global economy and undermine global institutions. The…demise of globalization is increasingly treated as a foregone conclusion…
[But] two recent studies...cast considerable doubt on the certainty of this conclusion. One, by researchers for the McKinsey Global Institute, reminds us of how interconnected the trade of the world truly is, and, thus, how hard it will be to disconnect it. The other, by staffers of the International Monetary Fund, warns us of the significant negative economic consequences that could result from full-blown global economic fragmentation.
Martin Wolf of the Financial Times takes a more measured tone, but worries about many of the same things, and cites a 2021 paper by Cerdeiro et al. on the costs of decoupling. And Adam Tooze declares that “the folks in Washington are cooking up something weird”, and identifies deglobalization with his concept of “polycrisis”.
Now, I do think there are plenty of dangers associated with decoupling, and there are bound to be costs as well (though perhaps also some under-appreciated benefits). But the way that many of decoupling’s critics are thinking about the issue just seems kind of confused; they talk as if decoupling, deglobalization, industrial policy, friend-shoring, “Buy American”, reshoring, and every other alternative to the trade patterns of the 2000s and 2010s are all one and the same. They are not one and the same. We are faced with a large menu of possible replacements for the old order, and conflating these options is unhelpful. Instead of wringing our hands and wishing for the world of 2015 to come back, we need to think productively about what comes next.
“Globalization” does not mean “Made in China”
One mistake I see the critics of decoupling make is that they almost instinctively equate globalization with the sourcing of production in China. Even Tooze seems to conflate these two things:
And trade trends between China and the rest of the world and between Emerging Markets hardly suggest a sudden stop to globalization. That dynamic of integration continues…
You can cleave to the old religion that economics always wins. In which case you dismiss the talk of deglobalization as journalistic hype…You could, for instance, cite the recent FT analysis which shows just how hard it is proving for Apple to disentangle itself from its Chinese supply chains.
Other commentators often do the same.
But although we might have gotten used to mentally equating globalization with “made in China” over the past two decades, those are obviously different things. If Apple moves an iPhone factory from China to India, and keeps selling the phones in the U.S., has the world “deglobalized” at all? No, not one bit. You still have the same amount of foreign direct investment and the same amount of trade. The only thing that changed is that now the factory is in India instead of China.
Decoupling doesn’t have to mean deglobalization, nor should it. In a recent post, I argued that some people in the Biden administration are making a mistake by trying to reshore production that should be moved to friendly countries instead:
In fact, decoupling could even make the world more globalized. The McKinsey report that Bacchus cites is all about concentration of imports — i.e., how some countries import some products almost entirely from one other country. A key example the authors give is how the U.S. imports most of its electronics from China:
The United States’ concentration fingerprint has remained stable in recent years; electronics, the largest import sector, remains relatively concentrated. US electronics imports are particularly concentrated in comparison with other large economies such as China and Germany. This mainly reflects strong trading ties with China for laptops (more than 90 percent of imports) and mobile phones (about 80 percent of imports).
Diversifying our sources of electronics imports could actually increase the volume of trade, through at least three mechanisms. First, if Chinese component makers sell their wares to Indian phone assemblers instead of to Chinese phone assemblers, that will represent an increase in trade flows. Second, if the economies of places like India and Vietnam and Bangladesh grow, they’ll probably trade more between each other — again, giving globalization a boost. And third, if Chinese consumers buy iPhones from Indian factories instead of from Chinese factories, that’s also an increase in trade.
Now, it’s possible that decoupling might decrease globalization. We don’t know yet. In my big post about decoupling last October, I predicted that China and the developed democracies would form into two economic blocs, with some trade between the two, but less trade in high-tech or other strategic industries. The breaking of high-tech trade links between the blocks will certainly mean a reduction in trade flows. But that could be compensated for — or more than compensated for — by a deepening of trade of all kinds within the blocs.
One commentator that gets this exactly right is Tyler Cowen. In a recent Bloomberg article, he writes:
Even the most successful “nationalistic” industrial policies rely on a highly globalized world…
Consider the semiconductor policies being pursued by China and the US. Currently, the highest quality semiconductor chips rely on an engraving technique from one company, ASML Holding of the Netherlands. US sanctions have made it difficult for that company to trade with China, while the US is glad to diversify its sources of supply. The result is likely to be a further expansion of trade networks in the chip industry, with…more suppliers in more countries…
Today’s industrial policy is not an alternative to globalization. It is preparing the world for the next round of it.
Deglobalization is old, decoupling is new
A second important point here is that actual deglobalization, if it does happen, is not necessarily the result of deliberate policies by national governments.
The 2001 IMF report by Aiyar et al. that Bacchus and others cite shows that anti-globalization rhetoric spiked around 2016, and trade restrictions around 2018. No big surprise there; that was the Trump Effect.

But actual deglobalization started long before that; it started with the financial crisis of 2008, or slightly before that. For the world as a whole, as well as for the two biggest trading nations, trade as a percentage of GDP stopped growing about a decade before Trump:

Now, one caveat here is that this counts exports and imports in terms of finished goods, not in terms of value added. But in terms of the direction of change, those two tend to track each other fairly closely, and there’s plenty of evidence that supply chains stopped globalizing around the time of the financial crisis.
The point here is that even if globalization does continue trending downward over the next few years, it won’t automatically make sense to blame national industrial policies for the continuation of that pre-existing trend. The world seemed to hit some kind of natural limit of economic integration in 2006 or 2007. That suggests that a bit of deglobalization could be the economically efficient outcome; gains from trade aren’t infinite. (Also remember that China’s pivot to industrial policy and onshoring of supply chains happened earlier than the U.S.’, and probably isn’t contingent on whatever the U.S. does.)
What are the real costs of decoupling?
The next important question to ask is: What are the actual costs of economic decoupling? Wolf cites two studies, one by Cerdeiro et al. (2021) and the other by Goes & Bekkers (2022). Both predict fairly substantial losses from decoupling — anywhere from 2% to 12% of total world GDP. The midpoint of that range — say, 7% — wouldn’t be catastrophic, but it would be significant. It would basically be like shaving 2 years off of global GDP growth.
Now, one thing to remember here is that there’s more to human welfare than money. The U.S. isn’t trying to stop China from building advanced semiconductors so that Americans have more jobs in the chipmaking industry — it’s trying to stop China from being able to build a military strong enough to conquer East Asia and establish global hegemony. I’m not sure how likely the strategy is to succeed, and obviously how much you value that outcome is a matter of opinion. But the point here is that there are potential costs of 2015-style permissive trade policies that aren’t measured purely in dollars and yuan. It’s also worthwhile to remember that to people who care a lot about geopolitical hegemony, hurting a rival country’s economy might be considered a benefit rather than a cost.
But OK, setting national security concerns aside for the moment, how seriously should we take these numbers? Looking at the two papers that are being cited, I notice that both of them base their estimates on general equilibrium models of the macroeconomy. This class of models has a pretty poor empirical track record, even when the models are much simpler than the versions being used in these papers. Now, that’s not a knock against these authors at all; the global economy is a complex phenomenon with a lot of feedback mechanisms, so you basically have to use some sort of general equilibrium model here. And these authors didn’t go around claiming that their particular modeling choices were the only ones that made sense; the pundits are the ones who pulled the numbers from their papers and presented them as empirical truth.
But pundits should know that papers like this involve a whole lot of highly questionable assumptions. For example, Goes & Bekkers assume that each country’s trade balance is fixed as a percentage of GDP. That makes the math easier to work with, it’s very much not true in reality. They also assume fixed savings rates, when in fact countries’ savings rates change a lot.
Those sorts of simple-but-wrong assumptions are probably not too hard to relax, even if it requires doing a little extra math. But then we get to the weird stuff. The model relies on countries’ industries engaging in a modified form of something called Betrand competition, which means they basically compete away each other’s profits to zero. This is unrealistic, but it would be tough to replace this element of the model with some alternative, more realistic form of competition. The model also crucially depends on the idea that trade creates knowledge spillovers — probably a realistic assumption, but extremely hard to measure quantitatively in the data.
Anyway, these are only a few of the specific assumptions the paper makes. On top of that, there’s all the implicit assumptions contained in the elements that they don’t include in the model — for example, governments’ own research spending, and how that might be contingent on trade policy, or how decoupling might accelerate the development of countries like India. And I haven’t even mentioned the difficulty of “calibrating” the model so that the parameters are right — an exercise that only makes sense in the first place if the models’ assumptions hold true.
Of course, the Cerdeiro et al. paper relies on a model with its own large set of (very different) assumptions, and faces similar issues.
Again, all these things are not reasons to criticize the authors of these papers. They did their best with a very challenging problem. But the numbers that these kinds of models spit out should not be treated as any sort of reliable guide to quantitative reality.
So how do we evaluate the costs and benefits of decoupling? Do we just throw up our hands and say that it’s unknowable, so we might as well do whatever we feel like? Well, no, we can’t do that. But instead of making policy in advance based on highly speculative and empirically unvalidated models, we should, in the words of Deng Xiaoping, “cross the river by feeling the stones” — we should see how a bit of decoupling goes, both in terms of economics and strategic results, and adjust our direction as needed.
There’s no going back to 2015
Analyses of trade and industrial policy that treat all types of government interference as one and the same, or which equate trade with making things in China, cloud our decision-making more than they help. They contain the implicit assumption that there are only two directions for the global economy — back toward the old equilibrium of 2015 or forward into a dark, frightening world of protectionism and autarky. But in fact, the world of 2015 just isn’t coming back. China’s leaders have committed to capturing a dominant position in key strategic industries, and this will not change depending on whether the U.S. cancels or tightens export controls. And China’s recent geopolitical aggressiveness simply won’t be ignored or forgotten by either the publics or the militaries of other countries.
So instead of looking back to the time when “free trade” was a simple concept and an easy rallying cry, we should be thinking about how to shape the next wave of globalization in a way that encourages global economic growth while also providing security to ourselves and our allies. It won’t be perfectly optimal, but nothing ever is.
Update: For much more info on the pre-existing trend of mild deglobalization that started around the time of the financial crisis, see this highly informative thread by the excellent Hyun Song Shin:
As a Canadian, I have my doubts about American intensions. The U.S. under Trump slapped a tariff on Canadian aluminum and steel in the name of "National Security". No one in the U.S. Government challenged this. The U.S. under Trump and Biden prevented companies from selling Covid vaccines to Canada. No one in the U.S. Government challenged this.
So looking at this subject from Canada's point of view, "decoupling" means we can't depend on the U.S. to be a supplier of any strategically important items. It also means Canadian companies are foolish to expand and depend on the U.S. as a customer of any strategically important items.
It appears that because of "decoupling", Canada can't depend on access to the U.S. market. Thus, it seems we must "deglobalize" the production of any strategically important items, so we don't find ourselves left high and dry by the U.S. in the future.
Texas Instruments announced it will build an $11 billion semiconductor plant in Lehi, Utah.
And Salt Lake City residents are exposed to toxic heavy metals that precipitate out of the shrinking Great Salt Lake and are carried by the wind. Sure, why not stick another semiconductor straw in the rapidly declining water resources in the Southwest, wasting more than 100 million gallons of fresh water (this is assuming Texas Instruments will claim, like Intel, that it recycles 95% of the water used by the average Fab).
The government should have attached conditions to the billions it’s giving these corporations, requiring the new chip fabs be built in the Great Lakes Area/Rust Belt. There are a hell of a lot of unemployed blue-collar workers there than in sparsely populated Utah.