Import Prices From China Exclude Duties, So the 2026 Value Swing Reads as Quantity

Two federal series describe U.S. goods imports from China every month, and neither contains a tariff. The Bureau of Labor Statistics prices what the foreign seller charges. The Census Bureau reports what U.S. Customs and Border Protection appraises the same goods at. Because the duty is absent from both, dividing one by the other strips the price term out of the trade numbers and leaves something neither agency publishes.

The July 2026 import price release, issued August 18, 2026, put the price index for imports from China 2.7 percent above its July 2025 level. Eight months earlier, in the November 2025 data released January 15, 2026, it sat 3.6 percent below its year-earlier level.

Both Measurements Stop Before the Duty Is Assessed

The BLS question-and-answer page states: "The prices for the items used to calculate the Import/Export Price Indexes exclude duties." The reason is a national-accounts requirement, stated in the Handbook of Methods: "For example, import prices exclude charges for duties because BEA excludes duty when aggregating the net exports component of GDP and the denominator (prices) needs to match the numerator (GDP) when calculating the real value of net exports."

The same page gives the price basis: "The majority of prices used in calculating import price indexes are quoted FOB (Free On Board) Foreign Port and the majority of prices used in calculating export price indexes are quoted FAS (Free Along Ship) U.S. Port; duty is not included." The word is majority, not all, and BLS repeats it a line later: "While the International Price Program prefers exit point price bases, point of origin or entry point price bases are used if they are the industry standard." Duty is excluded either way, but the observation point is not uniform across the sample.

The Census Bureau lands in the same place by a different route. Its Guide to Foreign Trade Statistics states: "The Customs value is the value of imports as appraised by the U.S. Customs and Border Protection in accordance with the legal requirements of the Tariff Act of 1930, as amended. This value is generally defined as the price actually paid or payable for merchandise when sold for exportation to the United States, excluding U.S. import duties, freight, insurance, and other charges incurred in bringing the merchandise to the United States." The value is appraised rather than taken as declared, and the price definition is generally the price paid, not always. Country detail, including monthly trade with China, uses that basis.

What the China Index Did, and When It Turned

Table 7 of the monthly release, by locality of origin, carries a separate index for China on a December 2003 = 100 base. In the July 2026 release the China row reads index 98.7 for June 2026, 99.5 for July 2026, and a twelve-month change of 2.7 percent, with monthly changes of minus 0.2 for March to April, 0.4 for April to May, 0.6 for May to June, and 0.8 for June to July.

The twelve-month figures as first published run as follows, with one reference month missing from the middle. November 2025: minus 3.6 percent, described in that release as "The 12-month decline in November was the largest over-the-year decrease since the index was first published in December 2003." December 2025 has no reading. January 2026: minus 2.6. February: minus 1.9. March: minus 1.1. April: plus 0.3, which the May 14, 2026 release called "the first year-over-year advance since December 2022." May: plus 1.1. June: plus 1.3. July: plus 2.7.

December 2025 is a publication gap, not a data gap. The February 10, 2026 release states: "Due to the compressed release schedule following the 2025 lapse in appropriations, the Bureau of Labor Statistics will update the Import and Export Price Indexes (MXP) website and database with December 2025 data but will forgo a detailed news release." It carries no China figure. The index exists: the March 5, 2026 release reports that "Import prices from China increased 0.3 percent in January, following no change the previous month." No release printed a December twelve-month change for China, so those eight readings span nine reference months.

That is a swing of 6.3 percentage points across those eight readings, all of them twelve-month changes. On that basis every reading from November 2025 through March 2026 is negative and every reading from April 2026 forward is positive. The one-month changes run separately: Table 7 puts March to April 2026 at minus 0.2 percent, in the month the twelve-month reading first turned positive.

The July 2026 release also reported: "From July 2025 to July 2026, nonfuel import prices rose 4.5 percent, the largest over-the-year advance since the index increased 4.6 percent for the year ended June 2022." China at 2.7 percent is rising more slowly than the nonfuel aggregate.

Dividing One Series by the Other

Because both are duty-exclusive, their ratio has a clean interpretation. Removing the twelve-month change in the China price index from the twelve-month change in the value of imports from China leaves an implied change in quantity.

This implied series is computed here. Neither BLS nor the Census Bureau publishes it. It is arithmetic on two published series, not an agency product.

Census customs-value imports from China, not seasonally adjusted, against the price change as first published:

• January 2026. 21,057.9 million against 41,770.4 million a year earlier: value minus 49.6 percent, price minus 2.6, implied quantity minus 48.2.
• February 2026. 18,955.6 against 31,654.6: value minus 40.1 percent, price minus 1.9, implied quantity minus 39.0.
• March 2026. 20,859.0 against 29,510.8: value minus 29.3 percent, price minus 1.1, implied quantity minus 28.5.
• April 2026. 19,789.2 against 25,332.4: value minus 21.9 percent, price plus 0.3, implied quantity minus 22.1.
• May 2026. 23,507.8 against 20,424.5: value plus 15.1 percent, price plus 1.1, implied quantity plus 13.8.
• June 2026. 25,150.0 against 18,898.7: value plus 33.1 percent, price plus 1.3, implied quantity plus 31.4.

The value change spans 82.7 percentage points across those six months, from minus 49.6 to plus 33.1 percent. The price term moves the answer by at most 1.7 percentage points in any single month, and by 1.1 on average. Across those six months the dollar value is a quantity story; the duty-free price is a rounding term against it.

The 2025 base months are extreme: imports from China fell from 41,770.4 million in January 2025 to 18,898.7 million in June 2025, a drop of 54.8 percent in five months, so twelve-month comparisons on that base produce large numbers in both directions during 2026. Across those same eight readings the extremes are minus 3.6 percent for November 2025 and plus 2.7 percent for July 2026. That bounds those eight readings, not every value the series has taken and not the months no release printed.

Reading the Rest of Table 7

The same table prices other origins over the twelve months ended July 2026, all of the following positive: Asia Near East 16.4 percent, Canada 15.0, Asian NICs 8.3, industrialized countries 7.6, Taiwan 6.7, Pacific Rim 5.1, European Union 3.2, Latin America 3.2, ASEAN 2.7, China 2.7, Mexico 2.1, Japan 1.7. The top of that list is energy rather than manufacturing: Canada splits into nonmanufacturing at 29.4 percent and manufacturing at 8.8. Table 5 of the same release, which groups imports by Harmonized System rather than by origin, puts mineral fuels up 26.8 percent over the same twelve months.

The rows also overlap. The table's footnote defines the Pacific Rim as "China, Japan, Australia, Brunei, Indonesia, Macao, Malaysia, New Zealand, Papua New Guinea, Philippines, and the Asian Newly Industrialized Countries." China sits inside it, so adding rows double counts.

The Appropriations Gap and the Three-Month Revision Window

Two documented features will bite anyone building a series from this. The first is suppression. The January 15, 2026 release states: "BLS did not collect survey data for October 2025 due to a lapse in appropriations." The consequence comes with a quantifier: "Because the import and export price indexes are calculated with both survey and nonsurvey data, many U.S. Import and Export Price Index values for October 2025 using primarily survey data will be suppressed for publication, including the all goods import and export index values, and will not be available in the BLS database." A footnote in the same release reaches past October: "Some Oct and Nov 2025 data values are not available due to the 2025 lapse in appropriations."

The operative words are many and some. The release names the all goods indexes among the suppressed values but does not say which locality-of-origin indexes survived, and nothing cited here settles the China row or says how BLS will handle the twelve-month change once October 2026 data appear. November 2025 did carry a China figure, which is where the minus 3.6 percent comes from.

The second is revision. The question-and-answer page states: "The monthly release includes first-published data for the reference month, as well as revised data from the previous three months." The May 14, 2026 release reported: "Import prices from China rose 0.8 percent in April, the largest 1-month advance since the index increased 0.8 percent in July 2008." The July 2026 release shows the March-to-April 2026 change as minus 0.2 percent: a gap of 1.0 percentage point, with the sign flipped. That comparison is made here by reading two releases side by side; BLS publishes no revision series for these indexes.

Where This Doesn't Apply

• Both universes have carve-outs, and they are not the same carve-outs. The BLS Handbook states the indexes "cover nearly all trade in merchandise goods, except for military goods, works of art, used items, charity donations, railroad equipment, items leased for less than a year, rebuilt and repaired items, and selected exports (custom-made capital equipment)." The Census guide has its own exclusions: shipments to the U.S. Armed Forces for their own use, merchandise shipped in transit through the United States from one foreign country to another, monetary gold and silver, and articles repaired under warranty. The overlap is partial: the Census entries are qualified, to Armed Forces shipments for their own use and to repairs under warranty, while the BLS entries are not. A military item imported commercially, or a repair outside warranty, is excluded from the BLS indexes while meeting neither Census exclusion. Census also compiles in full only shipments above $2,000, or $250 for certain quota items, among articles required to be reported on a formal entry; below that the figures are estimates. Dividing one series by the other assumes none of that swings the aggregate, which is untested here.
• Customs value is not strictly price times quantity. The Census guide requires the value of "assists" furnished to a foreign manufacturer to be included in the value reported for the merchandise, adding that "where it is not possible to isolate the value of 'assists', they are included. In these cases the unit values may be increased due to the inclusion of such 'assists'." An assist that could not be separated out is neither, and it lands in the residual this article calls quantity.
• Origin is a declared characteristic, not an observed one. The BLS Handbook treats origin as a price-determining characteristic evaluated in the sample. Where goods are routed through third countries and re-declared, both series follow the declaration, and neither separates a sourcing shift from a paperwork change.
• The ratio is not a quantity index. It is a value series deflated by a price index of different construction and weighting. It will not match a BEA chain-weighted real import measure and carries no within-month composition adjustment.
• Vintage mismatch. The Census values are current as of the August 4, 2026 release; the price changes are as first published each month. The mismatch adds error a same-vintage pull would not.
• Duty exclusion says nothing about who pays. A duty-exclusive price that rises says the seller's invoice went up. It does not allocate the tariff between exporter, importer, and final buyer; that needs domestic price data these indexes lack.

What to Watch Next Week

The Census FT-900 for July 2026, scheduled for Thursday, September 3, 2026, carries the July country detail. The July China price index is already published at plus 2.7 percent, so the July quantity figure can be computed the day the trade data appear. Imports from China in July 2025 were 26,383.6 million. For implied quantity to come in flat year over year, July 2026 imports would need roughly 27,096 million. Below that, quantity is still falling; above it, the May and June rebound reached a third month.

• September 16, 2026. BLS price indexes for August 2026, carrying revisions to May, June, and July, so the June and July China figures above can change.
• October 16, 2026. September 2026 data. The last twelve-month comparison whose base month sits outside the appropriations gap.
• November 17, 2026. October 2026 data, the release where the suppressed October 2025 base becomes a live publication question.
• Beyond four weeks: whether the China index closes the gap with the nonfuel aggregate or falls further behind.

Concrete Framework

1. Check the price basis before reading an import price series as a tariff measure. BLS excludes duties and quotes the majority of prices FOB foreign port. A series that excludes the tariff cannot measure it.
2. Confirm both sides of a ratio share a value basis. Census customs value excludes duties, freight, and insurance. C.I.F. value adds freight and insurance and still excludes duties. A C.I.F. numerator against an index quoted mostly at the foreign port puts a freight term inside what looks like quantity.
3. Record the vintage of every percent change. With a three-month revision window, a series assembled from monthly releases is not the series a database pull returns; April 2026 shows the difference can flip a sign.
4. Bound the price term before attributing anything to it, and say which readings the bound covers. The eight readings the news releases carried run from minus 3.6 to plus 2.7 percent, a spread of 6.3 percentage points. That bounds those eight, not the series.
5. Treat suppressed values as absent, not as zero. BLS says many October 2025 values are suppressed and some October and November 2025 values unavailable, without resolving the China row. Check whether the series you need survived before comparing across that stretch.
6. Do not add Table 7 rows. Pacific Rim contains China and Japan; industrialized countries contains Canada and the EU.