Three takes on “Do imports subtract from GDP?”

The release of the first quarter U.S. GDP sparked a flood of comments that imports do not subtract from GDP. This is only partly true.

The U.S. GDP growth came negative for the first quarter of 2025. Not interesting on its own, but interesting for the reason: a surge in imports driven by the desire to beat tariffs has resulted in an extremely negative contribution to GDP, here is chart courtesy of Joe Politano:

The obvious newspaper commentary then states that a surge in imports leads to a drop in GDP, something one would expect from somebody who knows enough economics, such as those with a business school degree or business journalists. (FYI, I can make these potshots as I have econ degree from business school.)

More imports do not equal lower GDP

This then provoked a counter-push from people who understand economics at a higher level and know that imports do not subtract from GDP. Most prominent was Noah Smith, who even changed his Twitter name, but I have seen the argument from at least 20 people. It is very simple: GDP measures the amount of goods and services produced domestically, and whether you import a lot of pharma products in the first quarter, as opposed to the second quarter, should not affect how much you produce at all.

Sure, there is the GDP identity, GDP=C+I+G+Ex-Im, but that is really only a misleading identity due to things being canceled. It is GDP as defined from the perspective of how output has been used. The full identity should be something like this: GDP=C_d+I_d+G_d+Ex, where the “_d” indicates that it counts only goods and services produced domestically. But since it is hard to measure consumption or investment for only domestically produced goods and services, it is easier to add and subtract foreign-produced goods and services:

GDP= C_d+I_d+G_d+Ex +(C_f+I_f+G_f) –(C_f+I_f+G_f)= C_d+C_f +I_d+I_f+G_d+G_f+Ex-C_f-I_f-G_f

Simplifying this by noting that C=C_d+C_f, etc., and Im=C_f+I_f+G_f gives you the usual formula. But thinking about the expanded formula, it is clear that an increase in imports does not mean lower GDP at all!

Well, they kind of can

But this should not be the end of the story – otherwise, it is a bit of gaslighting. Why? Well, imagine that normally I get a haircut every month from my hairdresser. (I don’t because I cut my hair myself, which is not counted as GDP, but that is a whole separate story 😊.) Then during summer I travel to my wife’s parents' abroad and get a haircut there. As a result, my hairdresser has an empty slot instead of my regular slot, which he is not able to fill. Effectively, I have imported services – here, haircut. But importantly, this import was a direct replacement of domestic production, and as such, it means lower GDP!

What is going on? Well, whether imports lower GDP or not really depends on the context. If you say “I have bought and imported a German car”, then this does not directly imply lower GDP; as shown above, the result will be higher consumption/investment and higher imports, with no overall effect on GDP. But if I say “I have bought a German car instead of a domestically produced car, and as a result, the domestic producer decides to produce one less car”, then higher imports do mean lower GDP. My action has not changed; what has changed is the context/counterfactual, and specifically the behavior of somebody else as a result of my action.

Before we get carried away, it is important to put a qualifier on this story. Sure, it does make sense that if I decide to opt for a German car instead of the domestic car, then the domestic producer has one less buyer and hence produces one less car. In practice, this, of course, does not happen overnight – they might react like that down the road when they see lower demand, but surely not right away, unlike in the story of a hairdresser. And even in the story of car producers, things are not so clear cut once you start thinking about long-run adjustments: Most macro sees the economy operating at its full capacity in the long run, and hence not demand-determined. This would mean that even in the long run, imports do not lower GDP, albeit you can create a story based on economies of scale so that they partly do.

And this time, in a weird way, they might have

Ok, back to US GDP. What is the context there? It is much more the first story than the second story. So does this mean imports were not responsible for the lower GDP after all? Well, kind of, but maybe also in some weird way they were. The question is: where did the imports go? Joe provides a large part of the answer: inventories, especially pharma inventories, which saw an unprecedented surge. Another part of the answer was fixed investment, especially in computers, and there are some other smaller parts, like cars. But even adding all this up will leave you somewhat short: the surge in imports was bigger than the surge in inventories and other categories.

And here lies the aspect in which this time the surge in imports might have caused lowered GDP: I think it is more than likely that the BEA has undercounted some of this, even after artificially boosting pharma inventories. Simply, counting GDP is a hard exercise, and different parts are easier or harder to count. Here, imports are much easier to measure than inventories. If you fully account for a surge in imports, but not for a surge in inventories, you end up with a lower measured GDP. So imports might have lowered GDP, albeit only as reported, not the reality.

This has interesting implications. One possibility is that as BEA gets more data, the first quarter gets revised higher, because imports will remain the same, but inventories or investment get boosted. I would expect some of that. The other, funnier, if you like macroeconomics teasers, possible implication is that this gets reversed in Q2: then imports will be much weaker, while inventories built up in the first quarter will offset this and decrease. If BEA again underestimates the change in inventories, this time their decline, then Q2 GDP will be artificially boosted. Now tell me macro is not fun…