Fetching primary parquet sources and computing exhibits.
research
The Great Rerouting
Since the 2018 tariffs, China's share of US goods imports fell from 21.9% to 14.1%. Tracing all 912 products where that retreat happened, 40% of the trade that moved to other suppliers went to countries that simultaneously deepened their own Chinese sourcing of the very same products.
The question
Decoupling has two very different endings. In one, American buyers find genuinely new suppliers: Taiwanese chipmakers, Irish pharma plants, Korean battery lines. In the other, the same Chinese content takes one extra hop: a connector economy imports the goods or their near-final forms from China and ships them onward to the United States. Both show up identically in the bilateral statistics Washington watches. This page separates them, product by product, using the full BACI bilateral matrix through 2024, two years beyond the window in most published studies of the reallocation (Alfaro and Chor 2023; Freund, Mattoo, Mulabdic and Ruta 2023).
Method in four steps
Decoupled products. All HS6 lines with US imports above $50M in 2017 where China's share of US imports fell by at least 5 percentage points by 2024: 912 products. At 2024 market size, the share China vacated is worth $238B.
Gainers. In each decoupled product, every other country that gained at least 1 point of US import share. Their combined capture: $333B (it exceeds the vacated value because these products grew and non-China suppliers also took share from each other).
The China-linked test.A gainer's capture of a product is flagged China-linked when that gainer's own China-sourcing share of its imports of the same HS6 rose by at least 2 points and reached at least 10% by 2024. Same-product matching is deliberately strict: it catches minimal-transformation rerouting, not all China content.
Benchmark.Outside the US, China's import share went the other way: from 13.3% to 16.0% of world-ex-US imports. The US retreat is policy-specific, not a global fading of Chinese competitiveness.
Figure 1
China's import share: United States vs the rest of the world, 1995-2024
The two lines rise together for two decades and split after 2018: the US line falls from 21.9% to 14.1% while the world-ex-US line keeps climbing to 16.0%. Whatever reduced China's presence in US statistics did not reduce China's presence in world trade.
Source: CEPII BACI 202501 (retrieved 2026-04-28), all bilateral flows aggregated by importer. Authors calcs.
Who captured the vacated ground
Table 1
Top 12 gainers in decoupled products, 2017 to 2024
gainer
capture
China-linked
linked share
products
Vietnam
$79.5B
$51.8B
65%
419
Chinese Taipei
$59.2B
$2.9B
5%
170
Mexico
$31.9B
$15.2B
48%
365
India
$21.3B
$5.3B
25%
335
Thailand
$19.4B
$12.4B
64%
200
Ireland
$16.5B
$0.0B
0%
19
Korea, Rep.
$15.1B
$9.6B
63%
201
Malaysia
$10.6B
$4.4B
42%
119
Cambodia
$8.3B
$6.7B
81%
127
Indonesia
$6.9B
$4.5B
65%
131
Canada
$6.5B
$1.7B
25%
207
Israel
$5.3B
$0.2B
4%
34
all gainers
$333B
$132B
40%
The league table splits cleanly. Vietnam captured $79.5B, of which 65% passed the China-linked test. At the other pole, Chinese Taipei and Ireland captured large gains with essentially no rise in same-product Chinese sourcing: semiconductors and pharmaceuticals are genuine reallocation, not rerouting.
Source: CEPII BACI 202501 (retrieved 2026-04-28), full bilateral HS6 flows, 2017 vs 2024. Authors calcs.
Figure 2
Connector imports from China within the decoupled-product set, 2017 = 100
Where the biggest moves happened
Table 2
Top 15 decoupled products by China-vacated value
HS6
product
vacated
China share 2017 → 2024
top gainer
linked
847330
Machines: parts and accessories of automatic data processing, magnetic…
$30.6B
57% → 19%
Chinese Taipei
no
852520
Transmission apparatus: for radio-telephony, radio-telegraphy, radio-b…
$22.2B
65% → 45%
India
no
847193
Data processing machinery: storage units, whether or not presented wit…
$13.5B
93% → 66%
Vietnam
yes
847191
Digital processing units: whether or not presented with the rest of a …
$10.7B
19% → 1%
Chinese Taipei
no
847199
Data processing machinery: n.e.s in heading no. 8471
$8.0B
58% → 5%
Chinese Taipei
no
852390
Media, unrecorded: prepared products n.e.s. in heading no. 8523, for s…
$8.0B
45% → 2%
Korea, Rep.
yes
850440
Electrical static converters
$5.7B
49% → 16%
Thailand
yes
854380
Electrical machines and apparatus: having individual functions, n.e.s.…
$4.4B
47% → 20%
Thailand
yes
852810
Television receivers: colour, whether or not combined, in the same hou…
$4.3B
40% → 15%
How sensitive is the headline?
Table 3
China-linked share of captured value under three test definitions
test: sourcing-share rise / 2024 level
linked value
share of capture
≥ 1pp rise, ≥ 5% level
$158B
47%
≥ 2pp rise, ≥ 10% level (headline)
$132B
40%
≥ 5pp rise, ≥ 20% level
$117B
35%
The headline is not an artifact of one threshold. Loosening the test to a 1-point rise reaching 5% lifts the linked share; demanding a 5-point rise reaching 20% still leaves roughly a third of all captured value flagged.
Source: CEPII BACI 202501 (retrieved 2026-04-28), full bilateral HS6 flows, 2017 vs 2024. Authors calcs.
What this shows, and what it cannot
This is descriptive accounting of gross bilateral flows, and three limits matter. First, flows are gross, not value-added: a Vietnamese phone with a Chinese screen counts fully as Vietnamese in US statistics, so the same-HS6 test here understates total Chinese content, which input-output work puts higher. Second, same-product matching catches minimal-transformation rerouting; when assembly genuinely migrates, components arrive under different HS headings and this test does not flag them, again making the headline a lower bound on China linkage. Third, correlation is not customs fraud: a connector can legitimately import Chinese goods for its own market while separately exporting its own production to the United States. What the pattern does establish is that the official bilateral decoupling number overstates the separation of the two economies, in line with the supply-chain evidence in Alfaro and Chor (2023) and Freund et al. (2023), here extended to 2024 at full product resolution.
Alfaro, L. and D. Chor (2023). Global supply chains: the looming great reallocation. NBER Working Paper 31661.
Freund, C., A. Mattoo, A. Mulabdic and M. Ruta (2023). Is US trade policy reshaping global supply chains? World Bank Policy Research Working Paper 10593.
Antras, P. (2020). De-globalisation? Global value chains in the post-COVID-19 age. NBER Working Paper 28115.
Gaulier, G. and S. Zignago (2010). BACI: international trade database at the product-level. CEPII Working Paper 2010-23. Data: BACI 202501 (retrieved 2026-04-28).
The other side of the triangle. While these economies gained US market share in decoupled products, their own imports of the same products from China grew well past their 2017 level. The pull is strongest exactly where the league table flags the highest linked shares.
Source: CEPII BACI 202501 (retrieved 2026-04-28); connector imports from China summed over the 912 decoupled HS6 lines, indexed to 2017. Authors calcs.
Mexico
no
940161
Seats: with wooden frames, upholstered, (excluding medical, surgical, …
$3.4B
69% → 27%
Vietnam
no
854140
Electrical apparatus: photosensitive, including photovoltaic cells, wh…
$2.9B
17% → 1%
Vietnam
yes
851830
Headphones, earphones and combined microphone/speaker sets
$2.6B
72% → 15%
Vietnam
yes
852110
Video recording or reproducing apparatus: magnetic tape-type
$2.6B
44% → 17%
Vietnam
yes
940360
Furniture: wooden, other than for office, kitchen or bedroom use
$2.2B
47% → 16%
Vietnam
yes
391810
Floor, wall or ceiling coverings: of polymers of vinyl chloride, wheth…
$2.2B
82% → 34%
Vietnam
yes
Electronics dominate the vacated value, and the biggest single gainers in those lines are mostly flagged China-linked. The exceptions are instructive: where the top gainer is unflagged, the product sits in a capability the gainer already owned.
Source: CEPII BACI 202501 (retrieved 2026-04-28), full bilateral HS6 flows, 2017 vs 2024. Authors calcs.