The Tariff Study That Doesn’t Add Up
Just about the biggest question in trade economics today is who is paying for the tariffs.
We know that hundreds of billions of dollars of tariffs are being paid. The U.S. government is collecting tariff revenue and this has helped push down the budget deficit. These tariffs are directly paid by U.S. importers. The question is who ultimately bears the cost. Is it foreign manufacturers, perhaps because they have been forced to reduce prices to maintain market share? Is it the importers absorbing the import levies? Are the costs being passed-through to consumers in the form of higher retail prices?
A new study from Germany’s Kiel Institute claims to have settled the tariff debate with “unambiguous evidence” that its authors claim demonstrates that American importers and consumers bear 96 percent of the cost of the 2025 tariffs, while foreign exporters absorb a mere 4 percent. The tariffs are “an own goal,” the authors declare.
There’s just one problem: the study doesn’t actually prove what it claims.
(iStock/Getty Images)
The Illusion of Precision
The headline figure—96 percent pass-through to American buyers—sounds authoritative. But it rests on a statistical foundation considerably less solid than the authors let on.
Their key finding is a coefficient of −0.039, with a standard error of 0.024, significant only at the 10 percent level. For non-statisticians: this means the estimate is quite noisy. With 25.6 million observations in their dataset, achieving only marginal statistical significance suggests enormous underlying variation in the data.
The true absorption by foreign exporters could plausibly range anywhere from zero to nine percent based on their own results. Presenting this range as a precise “four percent” figure requires more confidence than the data supports. A 10 percent significance level doesn’t justify 96 percent certainty.
The Disappearing Budget Widgets
The deeper flaw lies in what happens when tariffs reduce imports, which is exactly what the study says occurred. The authors report import values and volumes fell by roughly 28-33 percent. This is where their methodology breaks down.
Consider a simplified example. Before tariffs, the U.S. imported widgets from China at three quality tiers: budget widgets at $10 per kilogram, mid-tier at $15, and premium at $25.
Now impose a 50 percent tariff. Budget suppliers face impossible math. Their $10 widget now costs American buyers $15 after tariff, competing directly against higher-quality mid-tier products. Unable to cut prices enough to survive, they exit the market. Mid-tier suppliers have more margin. They might cut their price to $13, but still lose market share. Premium suppliers can afford to cut from $25 to $22 and maintain sales to less price-sensitive customers.
After the dust settles, the import mix has shifted dramatically toward premium products, and the average unit value has risen from $15 to $20.
The Kiel authors observe this and conclude: “Prices rose—clear evidence the tariff was passed through!”
But that’s not what happened. Actual transaction prices could fall across all tiers while the average rises, because low-price suppliers disappeared from the data. You’re now comparing premium widgets to what used to be a mix of budget, mid-tier, and premium. When tariffs wipe out the low-margin segment, the remaining shipments aren’t the same goods at the same quality—they’re the survivors.
That enormous underlying variation in their statistical results? This is exactly what causes it—different products within categories responding completely differently as low-price varieties exit and high-price varieties persist.
Built-In Measurement Problems
The study has other issues that compound this central flaw. Tariffs are imposed at very detailed product levels—HS8 or HS10 classifications that distinguish, say, frozen boneless beef cuts from frozen beef with bone. (HS refers to the Harmonized System, the international standard for classifying traded products, where higher numbers mean more detailed categories.) But the Kiel data only goes to HS6, a much broader category that might simply be “frozen beef.” In short, they are measuring tariffs with a ruler and prices with a yardstick.
This means when the authors assign a tariff rate to their data, they’re using an average across products that actually face quite different tariffs. Some sub-products within “frozen beef” might face 50 percent tariffs while others face 10 percent. The mismatch means their tariff variable is measured with error.
The consequence is predictable: when your key independent variable is noisy, statistical results get biased toward finding no effect. And “no effect” of tariffs on unit values is exactly what the authors interpret as “full pass-through to Americans.” They may have built their regression in a way that naturally shrinks the very effect they’re trying to measure.
Their control group is contaminated too. Tariffs cause trade diversion—U.S. importers shift from China to Vietnam or Mexico. This surge in demand allows non-tariffed exporters to raise prices or shift toward higher-margin products. When both tariffed and non-tariffed unit values rise, the difference looks small not because exporters don’t absorb costs, but because the comparison is corrupted.
The Leap to Consumers
Even accepting their import price findings, the authors make an enormous additional leap: they assert that American consumers ultimately bear the burden. But they provide zero empirical analysis of retail prices, corporate profit margins, or pass-through along the supply chain.
The incidence could land on corporate profits, on wholesalers and retailers compressing margins, or be split among multiple parties. The claim that this is a “consumption tax” is speculation, not evidence.
The Efficient Baseline That Never Was
Beyond their incidence findings, the Kiel authors claim tariffs create “deadweight loss,” economic waste from distorted consumption patterns and disrupted supply chains. They describe these costs as “pure economic waste—costs borne by Americans with no offsetting benefit.” This sounds damning, but it rests entirely on the assumption that the pre-tariff world represented an efficient allocation of production and trade.
That assumption is obviously false. If decades of Chinese industrial subsidies, below-market loans from state-owned banks, forced technology transfer, and market access restrictions had already distorted global production, then tariffs might actually reduce total distortions rather than create them. Standard welfare economics calls this the “theory of the second best”: when multiple distortions exist, correcting one while ignoring others can make things worse, not better. Conversely, adding a tariff that offsets foreign subsidies could move the economy closer to efficiency.
The Kiel authors never address whether their “deadweight loss” might actually be the cost of unwinding previous distortions—which would make it an investment in a more efficient future, not pure waste.
Stripped of its rhetorical confidence, this study shows something much narrower: in ocean-freight shipments, average unit values didn’t fall much while trade volumes collapsed.
This is consistent with exporters maintaining prices. It’s also consistent with price cuts masked by composition shifts. Or measurement error obscuring true effects. Probably all three.
It does not prove 96 percent of costs fall on Americans. It does not prove consumers pay higher prices. And it certainly doesn’t prove tariffs are an “own goal.”
