Loading workbench page
Fetching primary parquet sources and computing exhibits.
Loading chart...
Fetching primary parquet sources and computing exhibits.
For HS Section XVI (machinery and electrical equipment), we combine OECD TiVA's foreign value-added share of gross exports (Koopman, Wang & Wei 2014; Johnson & Noguera 2012) with CEPII BACI bilateral import flows to size backward linkages and the concentration of foreign input suppliers. This is the backward-GVC view that Baldwin (2016, The Great Convergence) argued replaced traditional comparative-advantage analysis: a country's competitiveness in a sector is as much about who it buys from as what it sells.
The foreign value-added (FVA) share of gross exports is the share of every dollar of sector exports that originally came from imported inputs, traced through the full input-output matrix. Koopman, Wang & Wei (AER, 2014) formalised the decomposition; Johnson & Noguera (JIE, 2012) earlier showed that the gap between gross exports and value-added exports grew 10-15 percentage points across advanced economies between 1970 and 2009. We compare USAagainst income peers with GNI per capita PPP within ±30% of its latest reading (World Bank WDI, indicator NY.GNP.PCAP.PP.CD), using OECD TiVA EXGR_FVA (MAINSH release, % of gross exports) for 2020.
SELECT iso3, AVG(value) AS fva_share
FROM 'tiva.parquet'
WHERE measure = 'EXGR_FVA' AND unit = 'PT_EXGR' AND dataset = 'MAINSH'
AND year = 2020 AND activity IN ('C26','C27','C28')
GROUP BY iso3;Decomposition of USA's gross imports in HS Section XVI by source country in 2024. True foreign value-added by partner requires a bilateral input-output table with partner-resolved flows, which the OECD 2023 TiVA release does not publish at the country × partner × ISIC level we'd need here (partner dimension is collapsed to World in the EXGR_FVA series we load). The gross-import ranking below is the best proxy the BACI bilateral archive permits and is what the Hausmann-Klinger (2007, 'The Structure of the Product Space', CID WP 146) literature uses for dependency mapping in the absence of firm-level I-O tracing.
Within HS Section XVI, each two-digit chapter has its own supplier HHI on the 0-10,000 DOJ scale (Hirschman 1945; Herfindahl 1950; US DOJ/FTC 2010, Horizontal Merger Guidelines §5.3). Chapters above the 2,500 red-zone threshold indicate concentration among exporting countries. This can help prioritize supplier-diversification analysis, but does not by itself establish pricing power or quantify the effect of a disruption on output prices.
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 base AS (
SELECT SUBSTR(product_code,1,2) AS hs2, country_code, SUM(export_value) AS v
FROM 'country_year_product/year=2024/*.parquet'
WHERE SUBSTR(product_code,1,2) IN ('84','85') AND export_value > 0
GROUP BY hs2, country_code
),
tot AS (SELECT hs2, SUM(v) AS t FROM base GROUP BY hs2),
shares AS (SELECT b.hs2, b.v/t.t AS s FROM base b JOIN tot t ON t.hs2 = b.hs2)
SELECT hs2, SUM(s*s)*10000 AS hhi FROM shares GROUP BY hs2 ORDER BY hhi DESC;For each HS2 chapter in the section, the estimated share of USA's imports coming from its single largest bilateral supplier. Chapters above the 40% single-source threshold (Arjona, Connell & Pisu 2023, OECD ECO/WKP 1775) are where a procurement board should be asked the qualification-of-second-source question. Note the construction caveat: bilateral HS-level attribution uses each exporter's share of world exports in the chapter to apportion USA's total imports from that partner, because the BACI bilateral × HS6 raw archive is not part of this build. The proxy converges to the true bilateral only where partner composition at HS6 is roughly proportional to world composition; published comparisons (CEPII BACI 2024 methodology note) suggest this is an acceptable first pass at HS2 granularity.
85, about 67% of USA's imports come from CHN. Rows are sorted by top-supplier share descending; empty cells indicate USA has no recorded imports in that chapter in 2024.A backward linkage is only a concern if the input flows through further production steps before reaching final demand. Antràs & Chor (2013, Econometrica 81:6) formalise upstreamness as the weighted average steps-to-final-demand implied by the Leontief inverse of the input-output matrix. A coarser but internationally 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; we apply the official UNSD correspondence to BACI HS96 trade. Leontief (1941, Structure of the American Economy) introduced the input-output accounting behind both measures; WIOD and OECD TiVA extend it with global bilateral coverage. Baqaee & Farhi (2019, Econometrica 87:4) show that shocks propagate super-linearly through the input layers, which is why the end-use mix matters for procurement stress tests.
Figure 5 gives this section's end-use split straight from the official UNSD HS-BEC crosswalk; the growth trace below reads the same unbundling story in the time dimension, following Baldwin (2013, Global Supply Chains: Why They Emerged): chapters pulled deeper into cross-border production fragmentation grew world exports faster than chapters whose trade just tracked commodity prices or final-demand income. For HS Section XVI, we plot each HS2 chapter's compound multiple of world-export value between 2000 and 2024. The intermediate end-use share of world trade itself barely moved, from 52.8% in 2000 to 54.6% in 2024 (Figure 5), so a chapter multiple far above the all-goods world multiple signals chapter-specific unbundling, not a rising world intermediates tide.
The within-section HS2 exposure in Figure 4 is one face of dependence; the other face is the HS6 line where a single country controls most of world trade everywhere, not just for one importer. Arjona, Connell & Pisu (2023, OECD ECO/WKP 1775) operationalise single-source concentration at CR1> 50%; Chowdhry & Felbermayr (2023) set a stricter 70% cut to isolate the lines where second-source qualification is effectively impossible within a 2-3 year horizon. We apply the stricter cut to every HS6 line in BACI 2024 with world exports above US$50M and rank the top 20 by trade value. These are the lines where the EU CRMA's 65% single-origin ceiling (Regulation (EU) 2024/1252 Art. 5) would bite hardest if extended from strategic raw materials to the full HS6 catalogue.
Figure 7 reads the supply side. The mirror of that is the demand side: HS6 lines where a single country buys more than half of world imports. Chowdhry & Felbermayr (2023, Trade in Times of Uncertainty, Kiel Working Paper) define strategic dependence symmetrically: concentration on either side of the ledger is a fragility because the losing side cannot reallocate within a commercially relevant horizon. A single dominant buyer is also the leverage channel of record for import-side policy tools, the US Section 301 tariffs, the EU Carbon Border Adjustment Mechanism (Regulation (EU) 2023/956), and China's 2023 MOFCOM critical-mineral buying quotas all operate through the top-1-importer role.
Figure 1 ranks USA against narrow income peers; Figure 9 steps back and plots every TIVA-covered economy at (log GNI per capita PPP, FVA share). The expected shape is an inverted U: low-income economies trade primary goods (low FVA); middle-income economies specialise in assembly stages of global value chains (high FVA); advanced economies move into services and intangible content (FVA falls back). Baldwin (2016, The Great Convergence, ch. 5) and Johnson & Noguera (2012, JIE 86:2) document this for the manufacturing-trade FVA series. USA's position relative to the cross-sectional fit is the cleanest single test of whether the country is over-integrated, under-integrated, or on-trend in HS Section XVI for its income level.
These four figures assemble the sourcing-dependency case that the latest wave of industrial policy is explicitly responding to. The US Inflation Reduction Act (H.R. 5376, 16 Aug 2022) Section 45X advanced-manufacturing credit and domestic-content bonus codify a preference for FVA moving toward North America and free-trade partners; the CHIPS and Science Act(P.L. 117-167) adds foreign-entity-of-concern rules to deny the credit when a targeted input chapter is >40% from a concern entity. The EU Critical Raw Materials Act (Regulation (EU) 2024/1252, adopted 11 April 2024) directly targets Figure 4's red zone with a 65% ceiling on any single third-country origin for strategic raw materials. China's dual-circulation strategy (14th Five-Year Plan, 2021) and the parallel '2+N' import-substitution guidance work the same dependency matrix from the opposite direction, aiming to cap inward FVA share in strategic ISIC divisions. Used as a triptych, Figures 1 + 4 + 5 tell you whether a given sector in a given country is a candidate for any of these instruments.
WITH world_sec AS (
SELECT SUM(export_value) AS t
FROM 'country_year_product/year=2024/*.parquet'
WHERE SUBSTR(product_code, 1, 2) IN ('84','85') AND export_value > 0
)
SELECT c.iso3, SUM(cyp.export_value)/ws.t AS share
FROM 'country_year_product/year=2024/*.parquet' cyp, world_sec ws
JOIN 'countries.parquet' c ON c.code = cyp.country_code
WHERE SUBSTR(cyp.product_code, 1, 2) IN ('84','85') AND cyp.export_value > 0
AND c.iso3 != 'USA'
GROUP BY c.iso3, ws.t ORDER BY share DESC LIMIT 15;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;
-- section split
SELECT b.end_use, SUM(cyp.export_value) * 100.0 / SUM(SUM(cyp.export_value)) OVER () AS pct
FROM 'country_year_product/year=2024/*.parquet' cyp
JOIN 'hs_bec.parquet' b ON b.hs6 = cyp.product_code
WHERE SUBSTR(cyp.product_code, 1, 2) IN ('84','85') AND cyp.export_value > 0
GROUP BY b.end_use;WITH base AS (
SELECT SUBSTR(product_code,1,2) AS hs2, CAST(year AS INTEGER) AS year, SUM(export_value) AS v
FROM 'country_year_product/**/*.parquet'
WHERE SUBSTR(product_code,1,2) IN ('84','85') AND year IN (2000, 2024) AND export_value > 0
GROUP BY hs2, year
)
SELECT hs2, MAX(CASE WHEN year=2024 THEN v END)/MAX(CASE WHEN year=2000 THEN v END) AS growth
FROM base GROUP BY hs2 ORDER BY growth DESC;WITH tot AS (SELECT product_code, SUM(export_value) AS t FROM 'country_year_product/year=2024/*.parquet' WHERE export_value > 0 GROUP BY product_code),
shares AS (SELECT product_code, country_code, export_value/NULLIF(t,0) AS s, t AS world_v
FROM 'country_year_product/year=2024/*.parquet' JOIN tot USING(product_code)
WHERE export_value > 0),
ranked AS (SELECT *, ROW_NUMBER() OVER (PARTITION BY product_code ORDER BY s DESC) rn FROM shares)
SELECT product_code, country_code, s, world_v*1000 AS world_usd
FROM ranked WHERE rn=1 AND s > 0.70 AND world_v > 50000
ORDER BY world_v DESC LIMIT 20;WITH tot AS (SELECT product_code, SUM(import_value) AS t FROM 'country_year_product/year=2024/*.parquet' WHERE import_value > 0 GROUP BY product_code),
shares AS (SELECT product_code, country_code, import_value/NULLIF(t,0) AS s, t AS world_v
FROM 'country_year_product/year=2024/*.parquet' JOIN tot USING(product_code)
WHERE import_value > 0),
ranked AS (SELECT *, ROW_NUMBER() OVER (PARTITION BY product_code ORDER BY s DESC) rn FROM shares)
SELECT product_code, country_code, s, world_v*1000 AS world_usd
FROM ranked WHERE rn=1 AND s > 0.50 AND world_v > 50000
ORDER BY world_v DESC LIMIT 20;WITH gni_latest AS (
SELECT iso3, MAX(year) AS y FROM 'wdi_data.parquet'
WHERE indicator = 'NY.GNP.PCAP.PP.CD' AND value IS NOT NULL GROUP BY iso3),
gni AS (SELECT w.iso3, w.value AS gni FROM 'wdi_data.parquet' w
JOIN gni_latest gl ON gl.iso3 = w.iso3 AND gl.y = w.year
WHERE w.indicator = 'NY.GNP.PCAP.PP.CD'),
tiva_agg AS (SELECT iso3, AVG(value) AS fva FROM 'tiva.parquet'
WHERE measure = 'EXGR_FVA' AND unit = 'PT_EXGR' AND dataset = 'MAINSH'
AND year = 2020 AND activity IN ('C26','C27','C28')
GROUP BY iso3)
SELECT g.iso3, g.gni, t.fva FROM tiva_agg t JOIN gni g ON g.iso3 = t.iso3
WHERE g.gni > 0 AND t.fva IS NOT NULL;