America finds herself in a peculiar position. At the same time that she is in an increasingly tense geopolitical standoff with China, her economy remains deeply dependent on China for everything from pharmaceuticals to high-tech computing.

This disconnect between political and economic realms has led to increasing calls for “decoupling”: disentangling the economic ties between the US and China. Classical liberals and economists tend to instinctively recoil at such suggestions. Not only is the economic logic of free trade unassailable, they claim, but economic entanglement helps prevent war as well.

But as the prospect of US-China war becomes ever-less remote, it is worth taking seriously the security vulnerability that comes with America’s dependence on China. If war were to erupt, could China cut the US off from essential resources and goods, suffocating not only America’s economy but her war effort itself? The US’s inability to produce N95 masks during the pandemic—because most of the world’s polypropylene production is in China—provides just a taste of what this could be like.

A liberalism that clings too unreservedly to the ideals of free trade will be a liberalism unable to defend itself. The few percentage points of GDP we gain by offshoring production to China is dwarfed by the catastrophe that would be losing WWIII. And as demonstrated by Japan’s dependence on the US before WWII, economic dependence doesn’t even necessarily deter war—in fact, dependence could spur war.

But “decoupling” is no more than a buzzword. The US’s economic entanglement with China is decades in the making and would be impossible to unravel overnight. Nor is simply reducing the gross volume of imports the right goal. Rather, we should systematically address the specific security vulnerability our economic dependence on China creates. Should it come to war, America’s supply chains must be robust enough to do without Chinese imports—even if the war is long and general.

Addressing this security vulnerability would require long-run investments in the capabilities and physical factory capacity to produce many vital goods domestically. (Or in allied countries, but I will omit this caveat henceforth for simplicity.) The question is, how to incentivize this adjustment process?

Ideally, we could directly target the “security externality” of trade. But we run into a fundamental informational problem. In the complex web of modern supply chains, how do we know which imports create vulnerability? While for some imports this may be obvious, there might also be “Achilles’ heel” imports: random, $5 components that nobody thinks about, but that turn out to be essential, and for which the production expertise and facilities do not exist in the United States.

The “decouplers” seem to retreat to the canonical approach to trade restriction: tariffs. The problem with tariffs, however, is that they are royally ineffective at reducing the security vulnerability we are concerned about. A general tariff incentivizes onshoring the production that is cheapest and easiest to onshore—but it is likely the imports for which onshoring would be the most expensive and difficult that present the greatest security vulnerability, as I will explain. In the language of economics, I argue that the imports that present the greatest security vulnerability are those with the most inelastic import demand—while a general tariff most reduces those imports with the most elastic import demand.

I propose an alternative approach: general per-product quotas. These would better target vulnerability than a general tariff. Perhaps even more importantly, this mechanism would reveal critical information about which imports are associated with the greatest vulnerability. Using this information, the government could then directly tax this vulnerability or subsidize domestic production alternatives. In that sense, the per-product quotas could function as a “stress test” of the robustness of American supply chains to disruptions in the flow of Chinese imports.

Note that even if you disagree with the premise that we ought to “decouple,” I invite you to consider this economic question. Whether you like it or not, decoupling may become the political mandate—and so, as economists, we ought to figure out how to best execute on that political mandate (rather than hiding behind the dogma “but free trade is good!”).

Conceptualizing the Security Vulnerability

Suppose the imports of a certain good were completely cut off. How bad would that be? That is what I shall term the “vulnerability” associated with importing that good.

Vulnerability depends on two factors:

  • 1) How essential the good is, and
  • 2) How difficult replacing the import with domestic production would be.

If suddenly our supply of nonessential goods—say, random gizmos like Rubik’s cubes—were cut off, that would be a bummer. But losing access to Rubik’s cubes would not be of greater relevance for the US economy or a potential war effort. By contrast, a critical component in microchip production would be more essential, for example.

Some imports may be essential, but the imports may be easily substituted by domestic production if necessary. For these goods, we might as well harvest the cost savings of offshoring during peacetime, and if necessary we could switch to emergency domestic production.

The imports we really should be worried about on security vulnerability grounds are those that are both essential and difficult to replace with domestic production. These might be imports like rare earths. If our rare earth supply were cut off, that could arrest our ability to procure new computer systems—and it would likely also be expensive, time-consuming, and difficult to establish alternative domestic production.

Critically, the degree of vulnerability associated with an import—whether it is nonessential or essential, whether it easily substituted by domestic production or not—is not reflected in its price. The dollar volume of rare earth imports from China is a fraction of the dollar volume of toy imports from China, for example. So we cannot measure the success of “decoupling” by simply looking the dollar volume of trade.

The Security Vulnerability in Economic Terms

We can put our conception of vulnerability in the language of economics by thinking about elasticities. In particular, vulnerability will correspond to inelastic import demand.

Consider first the two component factors of vulnerability mentioned above.

Essentiality: Elasticity of General Demand

The first factor, the essentiality of a good, corresponds to the elasticity of general demand (i.e. demand for the good regardless of source, whether foreign or domestic). The more inelastic general demand, the more buyers are willing to pay to continue procuring the good, so the more essential it is. If a good is nonessential, the more elastic general demand will be, as buyers would just forgo using that good altogether if the price got too high.

Substitutability with Domestic Production: Elasticity of Domestic Supply

The second factor, the difficulty of replacing the import with domestic production, corresponds to the elasticity of domestic supply of that good. The more elastic domestic supply, the easier and cheaper it would be to purchase more of this good domestically. Whereas inelastic domestic supply corresponds to it being expensive and difficult to procure this good domestically.

To be sure, some imports may be expensive to onshore (i.e. have inelastic domestic supply) simply because of labor costs, not because the US doesn’t have the capacity to make these goods in an emergency. But with respect to China, the view that offshoring is primarily about labor costs is outdated. As Steve Jobs once told Obama, iPhones are made in China not because of labor costs but because only China has the requisite manufacturing expertise and the vast industrial scale.

Vulnerability: Elasticity of Import Demand

It turns out that the combination of these factors—i.e., vulnerability itself—is captured by the elasticity of import demand for the good (i.e., the demand for the good solely from foreign, not domestic, sources). The more inelastic the import demand for a good, the higher the vulnerability associated with importing that good.

To see this, note that the elasticity of import demand captures both the elasticities of general demand and domestic supply. The elasticity of import demand will depend on the elasticity of general demand; if general demand is very elastic, then, straightforwardly, import demand will be very elastic too.

Critically, the elasticity of import demand will also depend on the elasticity of domestic supply. If domestic supply is very elastic, import demand will be very elastic as well, since buyers would just switch to acquiring the good domestically if the import got too expensive. Whereas if domestic supply is inelastic, there is little alternative to continuing to import the good, so import demand will be inelastic as well.

Thus, the import demand will be inelastic exactly when vulnerability is high—when the good is both essential (inelastic general demand) and the import is not easily substituted by domestic production (inelastic domestic supply).

Thus, to reduce the security vulnerability, we want to onshore those goods where the import demand is inelastic, while we care much less about those goods where the import demand is more elastic. Considering this elasticity helps clarify the effect of trade interventions.


A tariff most reduces purchases for goods with elastic import demand—while barely touching the quantity of purchases of goods with inelastic import demand. A general tariff on imports from China (~what the Trump Administration did) thus most reduces the imports that present the least security vulnerability, while barely touching the imports that present the greatest security vulnerability.

To put it in non-technical language: If tariffs make imports from China, say, 25% more expensive, you might stop importing those goods for which there is a cheap and easy domestic substitute, or those for which you don’t care much anyway. But would you onshore, say, rare earth production? Probably not—you’d have to acquire special expertise, make expensive capital investments, and the whole process would take years. Instead, you would just continue importing from China, now paying the slightly higher cost. But it is exactly these imports that are like rare earths that we are most concerned about.

What tariffs do accomplish is getting the most “bang for your buck” in terms of reducing the gross dollar volume of imports. That might make for a good political talking points (and achieve other protectionist aims), but does not further strategic decoupling—in fact, it masks the failure to reduce the security vulnerability.

Note that various “Buy American” provisions work similarly. These provisions typically mandate that, say, 75% of the value of a good purchased by the government must be made in America. But the last 25% of a good still made abroad is going to be the 25% that is hardest and most expensive to onshore, and thus the 25% that creates the greatest vulnerability. The marginal onshoring such provisions induce is going to be the onshoring that does the least to reduce the security vulnerability.

The Informational Problem

Ideally, then, we’d directly target the imports associated with the greatest vulnerability. A typical proposal along these lines might be some sort of government board that examines supply chains, identifies vulnerabilities, and then subsidizes domestic production of the critical imports.

How, though, to identify the goods that create the vulnerability? To be sure, for some goods this may be obvious; I mention rare earths earlier, and the vulnerability associated with them appears to be well-known.

But for all I know, in the intricate web of suppliers and sub-suppliers and sub-sub-suppliers that make up modern supply chains, there might be some $5 microchip component that we import from China and nobody thinks about, but that is absolutely essential to microchip production—and which requires specialized knowledge and manufacturing systems that we don’t currently have in the US. I would have very little confidence that a government board could identify these “Achilles’ heel” components.

Rather, we need a mechanism that directly reveals the security vulnerability associated with each import—that directly reveals the price elasticity of import demand.

Per-Product Quotas

What I call “general per-product quotas” would be a mandate that the imports of every good must be reduced by a certain percentage. The imports of Rubik’s cubes would be reduced just as the imports of rare earths and the imports of that unknown yet critical $5 microchip component. This mechanism would uniformly reduce the quantity of imports of all goods, regardless of how elastic their import demand.

These per-product quotas have an equivalent effect to a system of tariffs where the tariff rate scales inversely with the elasticity of import demand, i.e., a higher tariff rate on goods associated with greater vulnerability.

In terms of reducing security vulnerability, this would already be a much better mechanism than a general tariff. While the quotas wouldn’t exclusively target vulnerability, they also wouldn’t entirely exempt the imports with the greatest vulnerability (like a general tariff). In particular, this would generate demand for domestic production of every kind of good, ensuring that we have at least some base of expertise and capacity from which we could scale up in an emergency—compared to the situation today, where the US has likely lost the basic capability to produce many vital goods.

Technically, these per-product quotas could be implemented with auctions for import licenses of each good. In turn, the price of the import license would reveal the vulnerability associated with each good! The higher the price of the import license (relative to the original price of the import), the more inelastic import demand, and the higher the security vulnerability associated with that import. This mechanism harnesses the price mechanism to identify vulnerability—no imperfect government board necessary.

Using this revealed information on vulnerability, the government could then directly address the security vulnerability. For example, the government could tax the imports that present the highest vulnerability, or it could subsidize and support the domestic production of these goods. These targeted efforts could be complemented by continuing to ramp up the per-product quotas over time.

One way to think of these per-product quotas is a “stress test” of our trade with China. Our concern is about what would happen if our flow of imports from China were disrupted, e.g. a trade embargo during a war. One way to gauge that is to artificially restrict the flow of imports by a small amount. If you can suddenly only import, say, 95% of what you imported last year, how bad is that? If you are able to find an easy domestic substitute, or don’t really need the good much anyway, then the price of import licenses will be low; alternatively, if you are desperate to continue importing because the good is essential and there are no good domestic alternatives, the price of the import licenses will be high. This “stress test” can thus identify the imports whose loss in a trade embargo or other disruption would be most catastrophic. Of course, this “stress test” is imperfect; nonetheless, it should be revealing, and critically could help identify “Achilles heel” components.

One could even take this “stress test” idea further. In software engineering, there is an approach called “Chaos Monkey,” which involves randomly shutting down component servers to test the resiliency of the overall system to outages. Perhaps the government should try a similar approach to test the resiliency of supply chains. The government could randomly restrict different imports from China at different times. Seeing how supply chains cope with the loss of different imports would help reveal which imports present vulnerability; at the same time, this would incentivize firms to make their supply chains robust to disruptions in the flow of Chinese imports.


The “security vulnerability externality” of trade is real. But the goods with the most inelastic import demand are likely the goods where relying on China presents the greatest risk. Thus, simply restricting trade via a general tariff and reducing the gross flow of imports won’t be effective at mitigating the security vulnerability.

Figuring out a mechanism that addresses the vulnerability created by trade and incentivizes the long-run investments to reduce dependence on China seems like a central economic question. In particular, we need to figure out the “Achilles heel” informational problem—which imports present the greatest vulnerability?

If you have thoughts, please email me!

I am grateful to Basil Halperin, Stephen Malina, and Alexey Guzey for their invaluable feedback.