Loading workbench page
Fetching primary parquet sources and computing exhibits.
Loading chart...
Fetching primary parquet sources and computing exhibits.
When a single country or a small set of countries produces most of the world's supply of a particular good, a tariff, sanction, or natural disaster there becomes everyone's shock. This page measures supply-side concentration across the full HS6 catalog in 2024, identifies the products where one country commands a majority share, and tracks whether global trade is becoming more concentrated over thirty years.
For each HS6 line, we compute the Herfindahl-Hirschman index over exporter shares of world exports, on the 0-10,000 scale. HHI = 10,000 means a single country supplies everything; HHI near 0 means the product is produced everywhere. The US DOJ/FTC (2010) Horizontal Merger Guidelinesflag HHI > 2,500 as 'highly concentrated' and HHI from 1,500 through 2,500 as 'moderately concentrated'. Beck, Georgiadis & Gräb (2021), in their IMF Working Paper on critical inputs, argue that concentrated supply combined with low substitutability is the right operational definition of 'critical dependence'. The list below is restricted to products with more than US$500M of world trade in 2024 to suppress artefacts from ultra-thin markets.
The 1,500/2,500 bands here follow the historical 2010 DOJ/FTC guidelines, withdrawn in 2023.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.
WITH tot AS (SELECT product_code, SUM(export_value) AS t FROM 'country_year_product/year=2024/*.parquet' GROUP BY product_code) SELECT product_code, SUM(POWER(export_value/t.t,2)) AS hhi FROM cyp JOIN tot USING (product_code) GROUP BY product_code ORDER BY hhi DESC LIMIT 20;
Chowdhry & Felbermayr (2023) define strategic dependenceas a product-country pair where the importer sources a dominant share from a single supplier and no easy substitute exists. The supply-side counterpart is simpler: HS6 lines where one country controls more than half of world exports. We restrict to products with non-trivial world value (> US$500M) and rank by total trade value to emphasise where the stake is largest.
Goldberg & Pavcnik (2007), surveying the effects of trade liberalisation in developing countries, emphasise that import-side concentration reshapes domestic labour markets and prices through the intensity of foreign competition. The geodep database of Arjona, Connell & Pisu at the OECD (2023, ECO/WKP 1775 'An analysis of trade interdependencies') computes, for every importer × HS6 × year, the share of imports coming from the dominant supplier. The figure below reports the share of each country's HS6 import basket that is sourced from a single foreign supplier as defined by the OECD methodology (a high-C1 threshold, not a literal 100%), restricted to economies importing at least 4,000 distinct HS6 lines.
The IEA (2022) Critical Minerals Review and the USGS Mineral Commodity Summariesconsistently flag rare earths, lithium, and cobalt as the choke points of the energy transition. The BACI HS6 lines below map onto those products, but with an important distinction: these are export-market shares, not mining- or processing-production shares. For mined cobalt the USGS reports DRC at ~74% in 2023; for mined lithium (spodumene), Australia at roughly 46%. Those production shares will not equal the BACI shares below because producers may refine or consume domestically, ore may cross borders en route to refiners (e.g. DRC → China for cobalt), and spodumene, carbonate and hydroxide sit under different HS codes. Read this chart as 'where world buyers source from', not 'where this mineral is produced'.
Gaulier & Steingress (2020), tracking global value-chain unbundling through BACI, document that the median product became more, not less, concentrated after 2001 as China took over entire HS6 categories. Below, the world-average top-1 exporter share across every HS6 line traces that arc: a slow rise from 30% in 1996 to 34% in 2024. The drop in 2020 reflects a Covid-year reshuffling rather than a trend break, by 2024 the series has resumed its upward path.
Antràs & Chor (2013, Econometrica81:6) formalised upstreamness as the weighted mean steps-to-final-demand implied by the Leontief inverse; Antràs, Chor, Fally & Hillberry (2012, AER P&P) built the cross-country series. A coarser but globally comparable cut is the UN Broad Economic Categories (BEC) Rev 5 classification, whose SNA end-use dimension assigns every HS6 line to intermediate consumption, final consumption, gross fixed capital formation, or a mixed (dual-use) class. The stage table maps BACI HS96 trade through the official UNSD correspondence and plot world trade composition by end use, 1995-2024. The intermediate class is the layer where concentration propagates most: a shock to an input passes through every subsequent production step (Baqaee & Farhi 2019, Econometrica 87:4). Baldwin (2013, WTO/Fung) coined this the 'second unbundling': tasks, not goods, are what crosses borders.
HHI captures second-moment concentration; concentration ratios CRN are the top-N exporter cumulative share (Bain 1951, QJE; Saving 1970, IER; Miller-Pauly-Sobel 2000, International Economic Review 41:3). For supply-chain risk, CR4 is the share that a four-country blockade would take out and CR8 the share a G7+friends-only exclusion would still leave exposed. The DOJ/FTC (2010) Horizontal Merger Guidelines use HHI; Miller-Pauly-Sobel (2000) argue CRNtracks disruption pass-through more linearly because concentration ratio is invariant to small-tail reshuffling.
The Herfindahl, CRN, and single-source measures above are all snapshots. Baldwin's (2013, in Elms & Low eds., Global Value Chains in a Changing World, WTO/FGI) 'second unbundling' thesis is the time-series backbone behind why those snapshots look the way they do: once ICT and air cargo made it cheap to coordinate multi-country production, the share of world trade that is intermediate inputs rose relative to the share that is final goods. We measure the Baldwin index directly with the SNA end-use dimension of BEC Rev 5: trade in goods classed as intermediate consumption divided by total world trade, annually from 1995 to 2024 (dual-use 'mixed' lines excluded from the numerator). A rise means the world produces more jointly; a fall means re-shoring or regionalisation is winning. Pair this with the CR4/HHI data above: if the unbundling index holds high while CR4 on intermediate HS6 lines rises, the world is simultaneously interconnected andmore concentrated, which is the specific fragility pattern Baqaee & Farhi (2019, Econometrica 87:4) show propagates shocks super-linearly.
The cleanest descriptive proxy for the post-2018 'Great Reallocation' (Alfaro & Chor 2023, NBER 31661) is foreign direct investment into Mexico, the closest low-cost production base to the US market and the only USMCA partner with meaningful industrial capacity. US Census FT-900 reports that Mexico overtook China as the largest source of US goods imports in 2023 for the first time since 2002. The World Bank WDI series BX.KLT.DINV.CD.WD tracks net FDI inflows in current USD; we plot Mexico 2015-2024 below and compare the 2024reading against the 2015-2021 pre-IRA average. Freund, Mulabdic & Ruta (2022, World Bank Economic Policy) document that nearshoring reallocations follow FDI before they show up in customs data, so this chart leads Figures 1-8 by roughly two years.
This basket covers integrated-circuit processors and controllers (HS 854231), memories (854232), amplifiers (854233), other integrated circuits (854239), doped elements and compounds for electronics (381800), and semiconductor manufacturing equipment (848620). These broad customs categories do not isolate manufacturers, controlled technologies, fabrication capacity or the effects of industrial policy. World exports in the basket total $1.06T in 2024, compared with $1.05T in 2022 (+0.4%).
Figure 3 measured single-source dependence as a count over the entire HS6 import basket. The OECD geodep panel (Arjona, Connell & Pisu 2023, ECO/WKP 1775) goes further: each critically-dependent HS6 line is tagged to one of seven WTO-style strategic sectors (agri-food, chemicals, pharmaceuticals, steel, defence, transport, electronics) plus an 'other' bucket. This decomposition is what the European Commission's 2023 Economic Security Strategy (JOIN(2023) 20 final, 20 June 2023) and the US Department of Commerce's 100-day supply-chain review (E.O. 14017, 24 February 2021) both use to score 'strategic dependence': not the raw count of dependencies but the count weighted by how strategic the underlying sector is. The bars below report, for the 15 importers with the highest total critical-product count in 2022, how that count splits across the seven strategic sectors.
The four panels of Figures 2, 4, 6 and 7 together describe the fragility pattern that the post-2022 industrial-policy wave is pushing against. The US Inflation Reduction Act(H.R. 5376, 16 Aug 2022) and CHIPS Act (P.L. 117-167) use local-content rules and foreign-entity-of-concern exclusions to pull battery, semiconductor and clean-energy supply chains away from the concentrated-and-distant corner of Figure 7. The EU Critical Raw Materials Act (Regulation (EU) 2024/1252, 11 April 2024) sets a 65% ceiling per third-country origin per strategic raw material, which is a direct response to Figure 4's single-country export shares. China's dual-circulation strategy (14th Five-Year Plan, 2021) plus the 2023 MOFCOM export controls on gallium, germanium and graphite run in the opposite direction: convert Figure 4's concentration into leverage. The lesson is symmetric: the same HS6 lines are strategic from both ends.
WITH s AS (SELECT year, product_code, country_code, export_value / SUM(export_value) OVER (PARTITION BY year, product_code) AS share FROM 'country_year_product/**/*.parquet' WHERE export_value > 0), t AS (SELECT year, product_code, MAX(share) AS top1 FROM s GROUP BY year, product_code) SELECT year, AVG(top1) FROM t GROUP BY year ORDER BY year;
SELECT year, end_use, SUM(value_thousands) AS v FROM 'country_year_stage.parquet' WHERE flow = 'import' GROUP BY year, end_use ORDER BY year;
SELECT year,
SUM(CASE WHEN end_use = 'intermediate' THEN value_thousands ELSE 0 END)
/ NULLIF(SUM(value_thousands), 0) * 100 AS share
FROM 'country_year_stage.parquet'
WHERE flow = 'import' GROUP BY year ORDER BY year;SELECT year, value FROM 'wdi_data.parquet' WHERE iso3 = 'MEX' AND indicator = 'BX.KLT.DINV.CD.WD' AND year BETWEEN 2015 AND 2024 AND value IS NOT NULL ORDER BY year;
WITH selected AS (
SELECT product_code, year, export_value
FROM 'country_year_product/**/*.parquet' WHERE product_code = '381800'
UNION ALL
SELECT product_code, year, export_value
FROM 'country_year_product_ext/**/*.parquet'
WHERE revision = 'HS07'
AND product_code IN ('854231','854232','854233','854239','848620')
)
SELECT product_code, year, SUM(export_value) * 1000 AS usd
FROM selected WHERE year BETWEEN 2015 AND 2024 AND export_value > 0
GROUP BY product_code, year ORDER BY product_code, year;SELECT iso3_importer, crit_agrifood, crit_chemicals, crit_pharmaceuticals,
crit_steel, crit_defense, crit_transport, crit_electronics, crit_other
FROM 'geodep_country_summary.parquet'
WHERE year = 2022 AND total_products >= 4000
ORDER BY (crit_agrifood + crit_chemicals + crit_pharmaceuticals + crit_steel
+ crit_defense + crit_transport + crit_electronics + crit_other) DESC
LIMIT 15;