The Fed Gets ‘Who Pays for Tariffs’ Wrong Again
“You’re Paying 90% of Trump’s Tariffs,” the Wall Street Journal editorial board declared last week.
It’s a startling claim. The hook is obvious: you, the shopper, are eating the tax. To back it up, the WSJ’s editors pointed to the New York Fed’s latest look at import prices.
But the study—led by New York Fed economist Mary Amiti—is based on data and methodology known to be problematic for at least seven years. Ultimately, it does not tell us who is paying the tariffs.
What the New York Fed’s study shows is actually something narrower: customs unit values didn’t change much, so it doesn’t see evidence that foreign exporters cut invoice prices to absorb the duty. That observation is useful. It just doesn’t support the WSJ’s take or answer the question asked by the New York Fed’s Liberty Street Economics blog: “Who is Paying for the 2025 U.S. Tariffs?”
(iStock/Getty Images)
The Fed Study Has an Identification Problem
Here’s the reveal that makes the 90 percent claim buckle. About half of U.S. imports are related-party transactions, shipments between firms under common ownership. The Census Bureau puts related-party imports for consumption at 49.5 percent in 2024. Half the “price” data, in other words, comes from transactions where buyer and seller are part of the same corporate family. The invoice is often a transfer price: a compliance and accounting number, not a market price discovered in bargaining between strangers.
In economese, this is an identification problem. The Fed wants its data to reveal who bears the tariff, but the thing it observes—the customs invoice—doesn’t reliably behave like the variable its story requires.
Start with the Fed’s method. A tariff is remitted at the border by the importer of record. Nobody disputes this. It is true and trivial. The incidence question is what happens before and after that check gets written—whether the foreign side bears the burden by cutting the pre-tariff price, or whether the burden remains on the U.S. side, meaning the importer and then whoever the importer can squeeze or charge.
The New York Fed’s approach operationalizes “foreigners pay” as “foreign exporters cut their pre-tariff price.” Since we can’t see an exporter’s internal price book, the Fed proxies the pre-tariff price using customs unit values—import value divided by quantity at detailed product-country levels—and then looks for declines in unit values after tariff hikes. If unit values don’t fall, the Fed calls it “U.S. incidence.” If they do, it calls it “foreign incidence.” That is the entire machine.
And that machine runs into three deep problems, each large enough to spoil the “Americans paid ninety percent” conclusion, and together enough to destroy it.
First, we have the foreign-subsidiary problem. “U.S. importer” is a location, not a nationality. Many importers of record are U.S.-incorporated subsidiaries of foreign multinationals. If a tariff hit shows up as lower margins in the importing entity, the burden may land on the foreign parent’s consolidated profits and on foreign shareholders. Fed Governor Stephen Miran has made the definitional point directly: trade data can be misleading because “U.S. importers” can be U.S. subsidiaries of foreign companies. The Fed’s framework doesn’t observe who owns the importer’s residual profits. It observes that the remitter is on U.S. soil. That is not “Americans paid.” It is “a U.S.-registered entity wrote the check.”
This matters because it means the “U.S. side” bucket is not even cleanly “U.S.” in an economic sense. It is the legal remitter. The leap from remitter to “Americans” is a leap over ownership.
When Price Declines Look Like Tax Avoidance
Second, we have the regulatory constraint problem. The Fed’s test looks for a clean export-price response: if foreigners are paying, invoice prices should fall. But in related-party trade, the neat “price cut” the Fed is looking for can look like exactly what customs authorities worry about: shrinking the dutiable base right when tariffs rise.
This is not a conspiracy theory. It’s embedded in the customs valuation framework that governs related-party trade. WTO materials on customs valuation make clear that transaction value is the primary method—even for related parties—but when customs has doubt it inquires whether the relationship influenced the price, and importers seeking acceptance may have to demonstrate, among other things, that past imports were valued at about the same prices.
Translated into English: an affiliate can’t casually “discover” a lower invoice price the moment tariffs rise without inviting scrutiny because that looks less like bargaining and more like duty-base management. So, the absence of measured invoice-price cuts is not clean evidence about economic burden; it can be evidence about what is defensible to customs.
Once you admit this, the Fed’s key empirical “finding”—no big decline in unit values—becomes ambiguous. It may mean foreigners didn’t absorb. It may mean multinationals didn’t dare change the paperwork.
Third, we have the cost-shifting problem. Even if you set aside ownership and compliance constraints, the Fed’s inference doesn’t follow as a matter of logic in a multinational world.
A multinational hit by tariffs doesn’t have to respond by cutting the invoice price. It has other levers. It can keep the declared import price steady (especially if it fears customs scrutiny), keep the retail price steady (competition), and finance the tariff by squeezing costs in the foreign production platform: wages, supplier payments, overhead, quality, investment. None of that requires any change in the invoice price. None of it shows up as a lower customs unit value. Yet it can shift real burden abroad—onto foreign labor and suppliers—while the import paperwork remains unchanged.
The Fed’s framework can’t see this channel. It classifies the burden as “U.S. incidence” precisely because it defines foreign absorption as an invoice-price decline. When the adjustment happens through foreign factor payments rather than invoice prices, the regression dutifully records “no foreign price response” and prints “U.S. pays.”
There is another wrinkle that makes the invoice-price series a poor incidence instrument: tax incentives can run opposite tariff incentives. The U.S. International Trade Commission has noted the cross-pressure directly—understating import values to customs can reduce tariffs, while overstating those same values to tax authorities can reduce corporate profits and thus income tax. In other words, even when invoice values move—or don’t—they may be reflecting optimization across tax and customs regimes, not “who paid.”
Misreading Customs Data Led to Overreaching Analysis
Put these together and the “Americans paid 90 percent” headline isn’t merely overstated; it’s not identified. The Fed’s study doesn’t tell us anything about who is paying for the tariffs.
The real-world data is consistent with the identification failure. Since Liberation Day, commodities less food and energy—the CPI category most directly exposed to import price pass-through—has risen at an annualized rate of less than half a percent. That is not what 90 percent consumer incidence looks like.
What the Fed can responsibly say is narrower and more valuable: reported customs unit values didn’t fall much after tariff rates rose. That’s a fact worth investigating. Unfortunately, the Fed economists did not investigate it.
Instead, the Journal and the Fed came up with a number that sounds decisive—90 percent—and attach it to “you.” That’s rhetorical overreach invited by the New York Fed’s misreading of customs data.
We flagged this problem back in 2019 when similar claims were made about the China tariffs. Seven years later, the New York Fed economists are still making the same mistake.
Read the full article here
