Trade & Sourcing
Averaging the 57 Tariff Rates Would Overstate the Recorded Duty Bill by $14.62 Billion
For 2025, the simple mean of the 57 family observed rates is 14.97%, versus a customs-value-weighted rate of 12.36%. Applying that unweighted mean to the panel value overstates recorded duties by $14.62 billion.
Editorial evidence cutoff: September 9, 2026. Published September 29, 2026. Observation periods are stated throughout; older figures are retrospective evidence.
A spreadsheet shortcut can distort a September 2026 planning model before any forecast assumption is added. The completed 2025 manufacturing records provide a retrospective test: does the chosen average reproduce the burden of the actual import portfolio? This investigation uses data available by September 9 and demonstrates an aggregation error, rather than alleging that a company incurred the resulting hypothetical loss.
Take the 57 observed family duty rates in the 2025 manufacturing panel, add them and divide by 57. The result is 14.97%. Take the panel’s actual calculated duties and divide by its actual customs value, and the result is 12.36%.
If the first percentage were applied to the panel’s $559.30 billion import value as though it represented the recorded portfolio, the implied duty bill would be overstated by $14.62 billion. The error would not come from a wrong rate in any row. It would come from treating unequal buying categories as if they carried equal dollars.
This is a concrete aggregation problem, not evidence that a company or agency actually made that mistake. It demonstrates how a plausible spreadsheet operation can produce a materially different answer when its weights do not match the question.
FIFTY-SEVEN RATES DO NOT REPRESENT FIFTY-SEVEN EQUAL PURCHASES
The simple mean gives every tracked family one vote. A small import category influences it as much as a category with many times the customs value. That is a legitimate way to describe the average rate among the named families, provided the reader knows what is being averaged.
The dollar-weighted rate asks a different question: what share of the recorded customs-value portfolio appears as calculated duty? Its weights follow the actual import values. Multiplying it back by total customs value reproduces the observed $69.13 billion duty total.
The equal-family mean instead implies about $83.75 billion when applied to that same dollar base. The difference is $14.62 billion. This is why the two averages should not be presented interchangeably even though they come from the same 57 USITC DataWeb family records.
THE EXPENSIVE TAIL CARRIES LESS OF THE VALUE
Thirteen families have observed rates above 20% in 2025. Together, they represent 10.11% of the panel’s customs value and 22.65% of its calculated duties. Their contribution to the duty bill is much larger than their share of the goods’ value.
That asymmetry explains why the tail deserves attention without implying it should receive equal weight in every aggregate. A dashboard organized only around the highest rates can make the panel appear more expensive than its actual value-weighted burden. A dashboard organized only around import dollars can understate the importance of the high-rate tail.
The published table should preserve both shares. Showing a high rate without the associated customs value leaves the reader unable to assess its contribution. Showing the value without the duty share leaves the burden concentration out of view. The relevant relationship is in their pairing.
CHANGE THE THRESHOLD AND THE PATTERN SURVIVES
The 20% cutoff is a useful illustration, but it is not an economic boundary. At a 15% threshold, 20 families account for 22.52% of customs value and 38.89% of calculated duties. At 25%, seven families account for 5.86% of value and 14.92% of duties.
The same comparison was checked at 10% and 30%. In each case, the group above the threshold has a larger share of duties than of customs value. The pattern is therefore not created by selecting one convenient cutoff after seeing the results.
That does not make every threshold equally useful for a particular buyer. A plant’s relevant groups depend on what it imports. The purpose of varying the cutoff is to expose the distribution and test the claim, rather than to manufacture a universal category called high tariff.
THE PREVIOUS YEAR SHOWS THE SAME ARITHMETIC PROBLEM
In 2024, the equal-family mean was 4.03%, compared with a value-weighted rate of 3.26%. The difference was smaller, but the direction was the same. The aggregation issue did not appear only after the larger changes recorded in 2025.
The median tells yet another story. It was 3.42% in 2024 and 12.27% in 2025. The median locates the middle family; it does not reconstruct the dollar bill. Its proximity to the weighted rate in 2025 is an observed feature of that distribution, not a rule that makes the two measures interchangeable.
A compact chart of mean, median and weighted rate can therefore be useful if each is labeled by the question it answers. Three percentages without those labels risk making an analytical improvement look like three conflicting estimates of the same thing.
A NATIONAL WEIGHT IS NOT A FACTORY’S WEIGHT
The correct national-panel average is not automatically the correct benchmark for a particular company. A manufacturer importing mostly high-rate hardware has a different exposure from one importing mostly low-rate equipment. Applying the panel rate to both would suppress that difference.
For a company portfolio, the relevant weights would come from its actual matched customs values, with verified entry treatment for the items concerned. The national family rates can inform context or explicitly labeled sensitivity scenarios. They do not replace classification, origin, eligibility or the other facts needed to determine a shipment’s liability.
Calculated duties also retain their statistical meaning. As the Census definitions explain, they need not equal duties finally paid. A technically correct weighting scheme cannot turn the source measure into a cash-payment audit or a legal quote.
MAKE THE WEIGHTS PART OF THE HEADLINE
The lesson is not to ban simple averages. If the question concerns the typical tracked family, an equal-family summary can be informative. If the question concerns the recorded dollar burden, customs-value weights are essential. If the question concerns the middle observation, the median belongs in the report.
The $14.62 billion overstatement is the cost of crossing those questions without noticing. Every input rate can be accurate and the resulting dollar estimate can still be wrong for the portfolio because the aggregation changed what each row represented.
A clearer report would state the measure before displaying the percentage, keep the denominator beside it and make the family table downloadable. That allows a reader to replace national weights with a relevant purchasing mix instead of mistaking one headline number for a universal rate.
Source window: calendar 2024 and 2025; 57 identical families. All calculations use unrounded duty and customs values. The demonstrated overstatement is an explicit arithmetic scenario, not an allegation of an observed financial loss.
Sources and evidence
Evidence period: 2025; 2024 as distribution comparison. The frozen evidence record lists the source files and verified hashes available September 9, 2026. Source revision: 5e4fb7726c3d40060c0151c086c903baae856cab. Later live-data updates do not alter the historical evidence in this article.
census.gov/foreign-trade/guide/sec2.html
Published 2026-09-29.