Loading research piece
Fetching primary parquet sources and computing exhibits.
Loading chart...
Fetching primary parquet sources and computing exhibits.
Autor, Dorn & Hanson (2013, AER) dated the first China shock to the 1991-2007 window and China's 2001 WTO accession. Autor, Dorn & Hanson's 2021 update (BPEA Fall 2021) and Bown (2023, PIIE PB 23-2) argue a distinct second wave has been building since the mid-2010s, concentrated in the green-tech industrial policy frontier. We cannot observe labour-market outcomes from BACI, but the trade side of the shock is unambiguous: across 2013-2024 China went from 35% to 59% of world lithium-ion battery exports, from 35% to 51% of world solar PV exports, and built a dominant EV export position that did not exist at the beginning of the window.
Figure 1 plots China's share of world exports in the three green-tech product groups from 2000 through 2024. Unlike the first China shock, whose signature moment was WTO accession, the green-tech wave has a staggered onset: solar PV reached saturation first (mid-2000s), lithium-ion batteries pulled away around 2018, and battery-electric vehicles (HS 8703.80, introduced as a separate line in the 2017 HS revision) went from near zero in 2017 to roughly a quarter of world exports within seven years. The aggregate pattern matches the policy framing in Bown (2023, PIIE PB 23-2): Chinese industrial policy, led by the 'Made in China 2025' initiative announced in 2015, targeted exactly these product classes, and the trade-flow response followed within the next decade.
-- China share of world exports, per HS6, per year
SELECT year, product_code,
SUM(CASE WHEN country_code = 156 THEN export_value ELSE 0 END) /
NULLIF(SUM(export_value), 0) AS chn_share
FROM country_year_product -- for HS96 codes 870390, 854140
UNION ALL SELECT year, product_code, ... FROM country_year_product_ext
WHERE product_code IN ('870380','850760','854141','854142','854143')
GROUP BY year, product_code;Figure 2 divides China's worldwide exports in the selected basket by each importer's total imports in the same basket, then caps the ratio at 100%. It uses country-product totals, without bilateral product flows. A value of 100% means the global Chinese supply is at least as large as that importer's basket imports. It leaves the actual China-origin share anywhere between 0% and 100%. The EU27 denominator includes trade between member states, so it is not the EU's extra-EU import base. This figure does not measure employment exposure or identify how much a market actually imports from China.
The Herfindahl-Hirschman index (Hirschman 1945; Herfindahl 1950) sums squared exporter shares of world trade at the HS6 level. The US Department of Justice and Federal Trade Commission Horizontal Merger Guidelines (DOJ/FTC 2010, §5.3) treat markets with HHI above 2,500 as 'highly concentrated' on the 0-10,000 scale. Figure 3 tracks HHI for each group from 2000 through 2024.
The 1,500/2,500 bands here follow the historical 2010 DOJ/FTC guidelines, withdrawn in 2023. The 2023 guidelines use a highly concentrated threshold above 1,800. These figures measure concentration across exporting countries, not firms in a defined market; the bands are descriptive benchmarks, not antitrust findings.
It is tempting to read the green-tech concentration as a universal law, China steadily capturing every product. The data refutes that. Over the full 1996-2024 span, the exporter-share HHI of solar PV (8541.40) rose from 1,352 to 2,351, a clear concentration. But passenger cars moved the opposite way: HHI for mid-size cars (8703.23) fell from 1,301 to 1,057, and for large cars (8703.24) from 2,231 to 1,148. The global auto industry deconcentratedas new producers entered, even as solar funneled toward a single supplier. Concentration is product-specific and structural, not a blanket “China rises everywhere.”
-- Endpoint HHI for solar vs cars
WITH p AS (SELECT product_code, year, country_code, SUM(export_value) v
FROM country_year_product
WHERE product_code IN ('854140','870323','870324') AND year IN (1996,2024) AND export_value>0
GROUP BY 1,2,3),
w AS (SELECT product_code, year, SUM(v) tot FROM p GROUP BY 1,2)
SELECT p.product_code, p.year, 10000.0*SUM(POWER(p.v/w.tot,2)) hhi
FROM p JOIN w USING(product_code,year) GROUP BY 1,2 ORDER BY 1,2;For each HS6, we compare the CAGR of Chinese exports in the pre-2013 window (from the code's first available year) to the 2013-2024 post-window CAGR. The 'shock index' below is absolute post-2013 dollar acceleration: the gain in CHN exports from 2013 to 2024 (or from first-available year if later than 2013). Post-2013 dollar gain is the quantity that actually lands on foreign producers; CAGR is relative and can be misleading when base-year values are tiny (as for EVs pre-2020).
Supporting table, per-HS6 CAGR decomposition:
| HS6 | Group | Description | Pre-2013 CAGR | Post-2013 CAGR | Acceleration | Gain ($B) |
|---|---|---|---|---|---|---|
| 870380 | EV | Electric vehicles (BEV) | n/a | 148.3% | n/a | $32.6B |
| 870390 | EV | Cars n.e.s. (legacy, incl. pre-2017 EVs) | 31.8% | 84.4% | 40.0% | $43.0B |
| 850760 | Battery | Lithium-ion batteries | 4.3% | 30.4% | 24.9% | $65.7B |
| 854140 | Solar | PV cells & LEDs (legacy aggregate) | 36.1% | 6.5% | -21.7% | $17.9B |
| 854141 | Solar |
Figure 5 maps 2024 basket imports from all origins. It applies the same non-overlapping code selection as Figure 1: 870380, 850760 and 854141/42/43. The legacy aggregates 870390 and 854140 are excluded in 2024. The map shows import values, not import shares, domestic demand or Chinese content.
The China-shock-2.0 story so far is about CHN supplying world markets. The mirror channel runs the other way: the Belt and Road Initiative, announced by Xi Jinping in September 2013, was pitched as a mechanism for partner countries to sell raw materials, intermediates, and agricultural produce into China on preferential infrastructure-backed terms (Du & Zhang 2018, Journal of Comparative Economics46(1): 189-205; Bluhm, Dreher, Fuchs, Parks, Strange & Tierney 2023, AEJ: Economic Policy 15(1): 302-332 on AidData BRI geocoding). Figure 6 compares each BRI-partner country's average annual exports to China in the five-year pre-BRI window (2008-2012) against the post-BRI window (2018-2022). 2023-2024 are omitted to suppress the Covid rebound; a five-year average on each side nets out commodity-price-cycle spikes. Only partners with USD 100M+ pre-BRI average exports to CHN are ranked.
Figure 1 plotted the full 2000-2024 line; Figure 7 isolates the Made-in-China-2025 window. Xi Jinping's State Council unveiled Made-in-China-2025 on 19 May 2015 (SC GuoFa [2015] No.28), designating new-energy vehicles, electrical equipment, and advanced power equipment as priority sectors. The 2015 vs 2024 delta isolates the policy-relevant decade-long arc without the long pre-period noise. Each group shows a paired bar: CHN share of world exports in 2015 (lighter) and 2024 (darker); the absolute delta is labelled above each pair.
If the China shock 2.0 has a symmetric partner, it is the US-Mexico nearshoring trade. Alfaro & Chor (2023, NBER WP 31661) and Fajgelbaum, Goldberg, Kennedy, Khandelwal & Taglioni (2024,American Economic Review: Insights 6(2): 295-312) find that the 2018 Section 301 tariffs reallocated US imports away from CHN toward third-country suppliers with low tariff exposure. USMCA (replacing NAFTA on 1 July 2020) locked in Mexico's duty-free access, and US Census FT900 reports Mexico overtook China as the largest single source of US goods imports during 2023. Figure 8 crosschecks the mirror direction using BACI-reconciled exporter-side values: MEX-to-USA and CHN-to-USA bilateral flows divided by total world-to-USA imports, 2015-2024.
The share story is widely told and widely misread. China's share of US merchandise imports fell from 21.9% in 2017 to 14.2% in 2024, a 35percent relative decline that the wire copy reads as China losing the US market. But China's dollar sales to the United States barely moved: $508.4B in 2017 against $449.8B in 2024, down only 12 percent. The share collapsed mostly because the pie grew: total US merchandise imports rose 36 percent over the same window (from $2.32T to $3.16T). The genuine reallocation went to others, above all Vietnam, whose exports to the US grew 2.9x (from $48.4B to $140.5B), and Mexico, which overtook China as the single largest US supplier in 2024.
-- China share vs dollar level of US imports, per year
WITH usa AS (SELECT MIN(code) c FROM countries WHERE iso3='USA'),
chn AS (SELECT MIN(code) c FROM countries WHERE iso3='CHN')
SELECT year,
SUM(CASE WHEN exporter_code=(SELECT c FROM chn) THEN total_value ELSE 0 END)*1000.0 AS chn_usd,
SUM(CASE WHEN exporter_code=(SELECT c FROM chn) THEN total_value ELSE 0 END)
/ NULLIF(SUM(total_value),0) AS chn_share
FROM bilateral_year WHERE importer_code=(SELECT c FROM usa) AND year BETWEEN 2015 AND 2024
GROUP BY year ORDER BY year;The share-of-world-exports panels above measure China's supply-side push. The complementary demand-side leg is whether China itself stopped importing the same products as domestic capacity ramped. Autor, Dorn & Hanson (2021, BPEA Fall 2021: 381-447) frame the persistence of the original China shock partly through this combination: a country that simultaneously expands its global export share and closes out its imports converts the trade balance into a one-sided flow that is harder for the rest of the world to offset. Figure 9 plots China's annual net trade balance (exports minus imports, in current USD) for each green-tech group on the same per-group HS6 selection rule as Figure 1.
The trade-flow signature of a China shock 2.0 is present in the data for all three green-tech product groups, with different onset years but the same basic pattern: double-digit CHN shares of world exports in batteries and solar by 2013, with EVs following a decade later, and dominant-to-near-monopoly shares across all three by 2024. This is what the original Autor-Dorn-Hanson signature looked like for apparel and electronics in the 2000s, compressed into a shorter window. The labour-market reading requires the ADH commuting-zone-by-industry mapping, which is not replicable from BACI alone; Autor, Dorn & Hanson (2021) argue the green-tech wave should produce smaller US employment losses than the 2001-2013 episode because the pre-shock domestic industry in EVs, batteries and solar was much smaller in the US than apparel or metal-working were in 2001. The subsequent policy response (IRA subsidies, Aug 2022; Section 301 extensions; CHIPS Act, Aug 2022) postdates the 2021 paper and cannot be attributed to its authors.
Units: trade values current USD, BACI stored in thousands and multiplied by 1,000 for display. HS6 codes are TEXT to preserve leading zeros. ISO3 codes uppercase. Country numerics: CHN=156, USA=842, EU27 uses 27 BACI M.49 codes. HS revisions: 870380 HS17, 870390 HS96, 850760 HS12, 854140 HS96, 854141/42/43 HS22.
-- Ceiling ratio per block × group × year
SELECT year,
LEAST(100.0, 100.0 * SUM(chn_export_value) /
NULLIF(SUM(importer_import_value_block), 0)) AS chn_share_ceiling
FROM (country_year_product ∪ country_year_product_ext)
WHERE product_code IN (basket) AND country_code IN (block_member_codes)
GROUP BY year, block;| LEDs |
| n/a |
| -1.4% |
| n/a |
| $-0.1B |
| 854142 | Solar | PV cells, not in modules | n/a | -12.0% | n/a | $-0.8B |
| 854143 | Solar | PV cells, in modules | n/a | -5.4% | n/a | $-3.3B |
-- 5-year pre/post averages of partner->CHN exports
WITH bi AS (
SELECT c.iso3, y.year, y.total_value * 1000 AS v
FROM bilateral_year y
JOIN countries c ON c.code = y.exporter_code
WHERE y.importer_code = (CHN code) AND c.iso3 IN (BRI_ISO3)
AND y.total_value > 0
)
SELECT iso3,
AVG(CASE WHEN year BETWEEN 2008 AND 2012 THEN v END) AS pre_avg,
AVG(CASE WHEN year BETWEEN 2018 AND 2022 THEN v END) AS post_avg
FROM bi GROUP BY iso3;-- Mexico and China shares of total US merchandise imports, per year
WITH usa AS (SELECT MIN(code) AS code FROM countries WHERE iso3='USA'),
mex AS (SELECT MIN(code) AS code FROM countries WHERE iso3='MEX'),
chn AS (SELECT MIN(code) AS code FROM countries WHERE iso3='CHN')
SELECT year,
SUM(CASE WHEN exporter_code=(SELECT code FROM mex) THEN total_value ELSE 0 END) /
NULLIF(SUM(total_value), 0) AS mex_share,
SUM(CASE WHEN exporter_code=(SELECT code FROM chn) THEN total_value ELSE 0 END) /
NULLIF(SUM(total_value), 0) AS chn_share
FROM bilateral_year
WHERE importer_code = (SELECT code FROM usa) AND year BETWEEN 2015 AND 2024
GROUP BY year ORDER BY year;-- CHN green-tech net trade balance, USD, per year x group
WITH hs92 AS (
SELECT CAST(year AS INT) AS year, product_code, 'HS96' AS revision,
SUM(export_value)*1000.0 AS exp,
SUM(import_value)*1000.0 AS imp
FROM read_parquet('data/parquet/country_year_product/**/*.parquet')
WHERE country_code = 156
AND product_code IN ('870390','854140')
GROUP BY year, product_code
),
ext AS (
SELECT CAST(year AS INT) AS year, product_code, revision,
SUM(export_value)*1000.0 AS exp,
SUM(import_value)*1000.0 AS imp
FROM read_parquet('data/parquet/country_year_product_ext/**/*.parquet')
WHERE country_code = 156
AND product_code IN ('870380','850760','854141','854142','854143')
GROUP BY year, product_code, revision
)
SELECT * FROM hs92 UNION ALL SELECT * FROM ext;