The USMCA review is being drafted against transshipment. Mexico’s trade with China shows a different and larger problem, and the rule that reaches it has a price.
Aluminum arrives in Mexico under one six-digit product code and leaves for Laredo under another, welded into a trailer. Tariff schedules sort goods by code, and one of the tests the United States-Mexico-Canada Agreement (USMCA) uses to decide whether processing in Mexico confers origin is whether the code changed. Chinese metal that changes code inside Mexico passes that test. The customs record both governments will argue over when the fourth round of the USMCA review opens in September can see that the code changed. How much of the trailer’s value came from China is invisible to it. Every rule now being drafted for the review is written about one of those two things, the code or the content.
The review is under way. On July 1 the United States did not confirm the agreement’s sixteen-year extension, which sends the agreement into annual reviews through 2036. On July 23, closing the third bilateral round in Mexico City, Ambassador Jamieson Greer and Secretary of Economy Marcelo Ebrard issued a joint statement that recorded agreement on the urgency of “addressing free-riding from non-parties.” That phrase covers two different kinds of trade. Chinese goods can enter Mexico, acquire Mexican origin and continue north, and an origin check catches them. Mexican plants can also build for the American market out of Chinese parts, and an origin check written about codes waves those goods through, because the parts changed code on the way to becoming a machine.
Washington’s own accounting separates those two kinds of trade, and it leans toward the plants. The United States Trade Representative (USTR) delivered an automotive report to Congress the same July 1, which tracks where the value in transport equipment imported from Mexico originates. Between 2017 and 2024 the Chinese share rose from 4.5 to 7.1 percent, the American share fell from 23.1 to 18.3 percent, and the Mexican share rose from 50.9 to 57.2 percent. Goods passing through carry almost no Mexican value, so a rising Mexican share is evidence of production. The Federal Reserve Board’s June decomposition of Mexico’s export gain agrees. It puts direct transshipment below one percentage point of the gain and calls it negligible. Chinese production located in Mexico accounts for fourteen points.
This article measures how much of the trade each kind of rule reaches. It reads the trade code by code, across 4,584 six-digit products, and runs the same tests on ten other economies so that Mexico’s figures can be read against others such as Vietnam’s and Canada’s. Mexico buys about 23 cents of Chinese goods for every dollar it sells to the United States. Vietnam buys $1.13, and across seven years that contain the whole trade war, Mexico’s figure moved by about a penny. Goods that leave Mexico under the code they entered, at close to the price they entered, account for about 2 percent of its export growth, $4.2 billion across sixty-six codes, and Mexico’s own tariff reform of last December already reaches $1.6 billion of that.
The content channel is larger. The rule that reaches it is a value threshold of the kind the agreement already applies to vehicles, and the kind Reuters reported in May as the American ask, at 82 percent regional content with half of a vehicle’s value made in the United States. A threshold set that high also raises costs for the Mexican plants that turn Chinese parts into goods bound for the United States, and the fourth round has to weigh one against the other.
What the published estimates find
The charge is built from three series that move together. Between 2017 and 2024 Mexico’s goods exports to the United States rose from $308 billion to $493 billion. China’s exports to Mexico rose from $67 billion to $113 billion. China’s direct exports to the United States fell by about 10 percent. Laid side by side, those lines are the case for a diversion reading.
Caroline Freund’s screen at the University of California San Diego finds that the transshipped share of Mexico’s exports to the United States has never exceeded 1.5 percent since the tariffs began, and was highest in 2022 at about $4.6 billion. Vietnam’s peaked at 7.5 percent in 2020. The Fed’s figure, below one point, sits on a different denominator, because its percentages divide up the growth in Mexico’s exports while Freund’s is a share of everything Mexico ships north. The largest single component of the growth in the Fed’s decomposition, 38 points, is diversion out of sources other than China, a category that in the Fed’s definition includes Mexican firms as well as American and other foreign multinationals expanding in Mexico. A further 35 points is trend. Neither rule under discussion is written about that trade.
Enforcement produces cases, and cases come one importer and one product at a time. When the Commerce Department examined whether refrigerant blended in Mexico from Chinese components was circumventing antidumping duties on Chinese product, it issued a negative determination in November 2024; other inquiries end in affirmative ones. Neither outcome adds up to a share of trade. The estimates of that share all come from researchers, the Fed’s staff among them, and the review is being drafted on those.
Eleven economies on one scale
The simplest test is coarser than any of those screens: everything an economy buys from China divided by everything it sells the United States, the ratio behind the 23 cents. On the eleven-economy scale every documented rerouting case sits well above Mexico. Malaysia buys $1.35 per export dollar, Indonesia $2.51, and Vietnam, the case the transshipment debate usually has in mind, $1.13. Only Canada, at about 15 cents, sits below Mexico.
The penny is the net of a rise and a fall. Mexico’s ratio stood at roughly 22 cents in 2017, climbed to 25 cents in 2021, and settled back to roughly 23 cents in 2024. A country turning into a back door for Chinese goods would show that ratio climbing through the tariff years and staying up.
The ratio counts everything Mexico buys from China, including goods absorbed at home that never enter an export, so it is a ceiling on Chinese content. At the margin, Mexico bought about 25 cents from China for every additional dollar it sold the United States between 2017 and 2024. Mexico’s customs data put 2024 imports from China at almost $130 billion; China reports shipping about $90 billion, and the figures used here reconcile the two. On Mexico’s own declaration the ratio is roughly 26 cents, and Vietnam’s is still more than four times as high.
Two companion measures rank the eleven the same way. One is the ratio computed on the growth alone. The other is the share of export growth sitting in codes where Chinese exports to the United States fell while both other legs rose, the pattern a diverted shipment leaves in the data. Mexico is second lowest of the eleven on both ratios and third lowest on that diversion triple, and the documented rerouting cases sit above it on all three. The triple is the noisier measure: Canada ranks above Mexico on it, with half of its export growth in such codes against a third of Mexico’s, and nothing in the case record suggests Canada is a rerouting hub. I lean on the two ratios and treat the triple as a check on them.
What 2 percent measures
The 2 percent rests on three conditions. A code is flagged when the rise in Chinese imports in the same six-digit code covers at least half of that code’s export growth to the United States, when Chinese imports cover at least half of its 2024 export level, and when the good leaves Mexico at between 0.6 and 1.6 times the price per tonne at which it entered. Sixty-six codes meet all three. The 2 percent is their share of $199 billion, the gross growth in codes that grew by at least $10 million. Of everything Mexico sold the United States in 2024, Freund’s denominator, the sixty-six codes come to 1.8 percent, and to 1.5 percent on her own count, which takes the smaller of the two legs in each code. That puts them at her ceiling.
Across sixty combinations of the thresholds the share of that growth ranges from 1.1 to 12.6 percent. A stricter definition cuts the set to twenty-six codes and $1.9 billion. The flagged codes cluster in machinery and electrical equipment: twenty-five of the sixty-six, carrying $2.4 billion of the $4.2 billion. Aluminum, in five codes, and truck tires account for a further half billion.
Running the same test on the other ten economies shows how far the 2 percent can be trusted. It puts Cambodia at 1.5 percent of its own export growth, below Mexico, which contradicts the case record, since Cambodia is one of the best-documented rerouting hubs in Asia. Cambodia’s score rests on seven codes, so one reclassification would move it. The screen only catches pass-through that keeps its code, and much of the documented rerouting in Vietnam and Cambodia changes the code in processing. The price condition is the weakest of the three, since about a fifth of the quantity data behind it are estimates. I use it to confirm the coverage conditions and would not rest a claim on it alone. Two percent is therefore a floor under same-code pass-through, and the question of rerouting through processing stays open.
Between the floor and the ceiling
How much larger the content channel is can be bracketed by widening the definition of a product. At six digits, 14 percent of Mexico’s export growth to the United States is matched by Chinese import growth in the identical code. At four digits the figure is 16 percent, at two digits 22 percent, and for the economy as a whole about 25 percent, which is the marginal 25 cents from the connector comparison read as a share. The top rung is that ceiling, and the rungs below it show how much of it is reached at each level of product detail. Vietnam’s series runs from 21 percent at six digits to 105 percent economy-wide. Chinese content enters Vietnam’s export growth under codes different from the ones it leaves under, at a scale Mexico’s trade does not show. For Mexico the content channel lies somewhere above the 2 percent that keeps its code and below the 25 percent ceiling.
Where the growth landed is the other half of the picture. About 77 percent of Mexico’s export growth, some $153 billion, went into codes where Mexico already held a comparative advantage in 2017, before the tariffs (79 percent on a pre-2016 base). Of the $4.2 billion flagged, $670 million billion sits in those codes and $3.49 billion in codes where Mexico held no advantage at all.
Mexico’s December tariff, matched against the flagged codes
The $1.6 billion comes from a statute. Mexico’s Congress passed a reform of the tariff schedule in December 2025, in force since January 1. It set duties of 5 to 50 percent on 1,463 tariff lines from countries without a free trade agreement, which in practice means principally China. Several law-firm summaries describe it as expiring at the end of 2026; the decree’s transitory articles set an entry-into-force date and no expiry, so the duties stand until repealed. A second executive decree in April 2026 added about 200 lines; the figures here cover the December law.
Textiles and apparel account for about half of the affected lines. Steel and steel articles, footwear, furniture, toys, paper, plastics, and vehicles carry most of the rest. Machinery and electrical equipment together account for roughly 3 percent of the lines, and those are refrigerators, washing machines, fans, small motors, and engine parts.
Twenty-one of the flagged codes fall inside the decree, worth $1.6 billion of the $4.2 billion, and that is an upper bound, since the decree’s eight-digit lines are matched at six. Forty-five codes and about $2.5 billion sit outside it. The largest flagged code is among them: portable computers, where Mexican exports to the United States grew by $538 million over the period. So is the third largest, lithium-ion cells, at $305 million. The second largest, fans, at $314 million, is covered. The same law, in its fourth transitory article, authorizes the Economy Ministry to open import channels from non-FTA (Free Trade Agreement) countries where Mexican plants have no other competitive source of an input.
The code and the content
The two rules operate on different parts of Mexico’s trade and at different scales. A classification test operates on the $4.2 billion, most of it in codes where Mexico had no comparative advantage before the tariffs. A value threshold audits the growth itself, including the three-quarters of it that went where Mexico was already competitive.
The American ask, as Reuters reported it, is a content rule of that kind. It would pass over almost all of the $4.2 billion, since only two of the sixty-six flagged codes sit in the vehicle chapter, trailer parts and goods vehicles, $259 million between them. It would operate on the trade USTR’s own report tracks, where the Mexican share of value has risen faster than the Chinese share. A higher regional floor targets the Chinese share and raises the cost of the Mexican share that grew alongside it.
Both rules pass over the 73 points of the gain that the Fed attributes to diversion from other sources and to trend. The published estimates, this one included, size the slice that keeps its code. What a content threshold at the reported levels would cost the plants that use Chinese parts has never been measured, because it needs the input shares of individual plants, which customs data do not hold.
A content rule written in Washington will need the clause Mexico’s Congress wrote into its own law, or the plants it is meant to protect will pay for its absence. The fourth round will have to settle which inputs get the clause. Mexico’s own Congress, writing its tariff, left lithium-ion cells and portable computers out of it entirely.
Sources
Trade figures throughout are computed from the Base for International Trade Analysis (BACI) database of the Centre for Prospective Studies and International Information (CEPII). HS 2012 vintage, release V202601, covering 2017 to 2024, and are cited on that basis alone. The data end before Mexico’s tariff reform and the current American tariffs took effect, so they describe the structure those measures were aimed at; their effect is outside the window. BACI runs about 2 percent below US Census figures for Mexican exports, so the two should not be mixed in a single comparison. The HS 2012 vintage is used deliberately: Mexico reported its 2017 trade under that nomenclature while China and the United States reported under HS 2017, and the later vintage carries no Mexican declaration for that year. The coverage ratio at the center of the diagnostic is borrowed: it is Caroline Freund’s fourth screening criterion, applied to growth and at a lower threshold, and the estimator of Iyoha, Malesky, Wen, and Wu bounded at one; the contribution here is the per-code classification, the capability anchor, and the aggregation ladder. Because BACI reconciles reporter and partner declarations, it cannot display mirror gaps, which is why the effect of Mexico’s own declaration is given in the text; the Mexican figure is as reported by the Federal Reserve Bank of Dallas, the Chinese one as filed to the United Nations. The decomposition of Mexico’s export gain is from the Federal Reserve Board’s FEDS Note of June 5, 2026, a different institution from the Dallas bank. Mexico’s reform is the decree published in the Diario Oficial de la Federación on December 29, 2025; the count of 1,463 tariff lines is the Secretaría de Economía’s, and the shares by sector are computed from the decree’s own table. The match against the flagged codes truncates the decree’s eight-digit lines to six digits; the executive decree of April 23, 2026, which added about 200 lines, has not been matched. The USMCA figures are from USTR’s 2026 biennial report to Congress on automotive trade and from the joint statement of July 23, 2026; USTR’s value-added shares and the agreement’s regional value content are computed on different bases and are not interchangeable. Full method note and the flagged product codes are available on request at gilberto@grippoint.co.
