Import dependence rose sharply for tech and pharma inputs while it fell overall on the shale-driven collapse in petroleum imports, 2008-2024

Summary

Over the 16 years this warehouse's trade data actually covers (2008-2024 — see the coverage note below on why not the full 20), American reliance on imports moved in sharply different directions by category. Semiconductor imports rose 119% in real terms and pharmaceutical imports rose 115% — the two largest increases of any category measured — while petroleum and product imports fell 63% and industrial supplies fell 43%, driven by the U.S. shale boom that turned the country into a net petroleum exporter around 2020. Computers (+47%), non-automotive capital goods (+45%), and automotive (+40%) also grew well above the pace of the total. Netting these moves against each other, total real goods imports grew only about 4% over 16 years while the U.S. economy grew roughly 35% in real terms — so in aggregate, goods-import dependence (imports relative to GDP) actually fell, even as dependence on a handful of high-tech and pharmaceutical inputs rose sharply. This is a genuine divergence, not a single trend: some industries became dramatically more import-dependent while the aggregate number moved the other way almost entirely because of one commodity (oil).

Why not a full 20-year, true industry-level answer

Before querying, this analysis checked a pre-built methodology note (an AskAmerica "recipe") for exactly this question shape, which confirmed: there is no single, freely queryable, industry-level import-penetration series spanning a full 20-year window in U.S. government data. The two natural sources are each blocked differently — Census's international trade API requires a registration key this connector doesn't hold, and BEA's Import Matrix / apparent-consumption tables (the correct denominator for a true "import penetration" ratio = imports / (domestic production + imports − exports)) are published only as an interactive web application, not a downloadable bulk series. This connector's own BEA foreign-trade detail table (econ.trade_statistics, sourced from NIPA Table 4.2.5B) is loaded starting in 2008, not 2005 — confirmed by a row-level coverage scan, not just the schema's declared window. So the quantitative analysis below covers 2008-2024 (16 years), and is built from import VALUE growth by BEA end-use goods category — the finest breakdown available — rather than a true production-relative import-penetration ratio. That is a related but different measurement: a category's import value can grow briskly even if domestic production grows just as fast (penetration flat), or fall even as penetration rises (if domestic output falls faster). The BEA categories used here (semiconductors, pharmaceuticals, computers, automotive, etc.) also correspond to goods-producing sub-industries, not the full universe of services industries, which BEA does not break out at this level of geographic/commodity detail in the tables available here.

A genuine defect was also found and worked around in the data: this connector's trade_statistics/trade_balance_summary tables mislabel most detailed BEA line items (e.g. "Chemicals," "Automotive vehicles, engines, and parts," "Semiconductors and related devices") as generic "trade_balance" rows rather than tagging them import or export, because the automatic classifier only recognizes lines whose text explicitly starts with "Imports of" or "Exports of." The real import and export values are present in the underlying value column; the fix used here was to identify the import-side line by its higher BEA line number (line numbers ≥100 in Table 4.2.5B are the imports section, versus <100 for exports) rather than trusting the connector's own trade_type field. This is worth flagging as a data-quality defect in this table's transform logic.

What the numbers show, category by category

Methodology: nominal category values pulled from econ.trade_statistics (BEA NIPA Table 4.2.5B, annual frequency), identified as imports by BEA line number ≥100; deflated to 2024 dollars with the adjust_inflation tool (CPI-U, BLS series CUUR0000SA0). Nominal (unadjusted) growth rates are higher across the board — e.g. semiconductors +219% and pharmaceuticals +213% nominal — but real terms are the right comparison for "dependence," since they strip out the roughly 46% cumulative CPI inflation over the period.

For scale: nominal U.S. GDP grew from $14.77 trillion (2008) to $29.30 trillion (2024), +98%, and real GDP grew on the order of 35% over the same period (BEA). Total real goods imports growing only 4.4% against that backdrop means the goods-import share of GDP fell from about 14.5% to 11.2% — an aggregate DEcline in imports-relative-to-GDP that is almost entirely a petroleum story, not a broad-based retreat from imports.

Which industries actually moved most, and why

Biggest increases in import dependence: semiconductors and pharmaceuticals. Both roughly doubled or more in real import value over the 16-year window. This lines up with well-documented industry structure: U.S. fabless chip design and foundry offshoring (Taiwan, South Korea) deepened over exactly this period, and pharmaceutical manufacturing (especially active ingredients and finished generics) shifted substantially to Ireland, India, and the EU. A September 2025 cross-sectional study by the Economic Innovation Group (using BEA's 2023 input-output import matrix, not a time series) corroborates the STRUCTURE of this dependence today: it finds computer & electronic products manufacturers import 37% of their intermediate inputs and capital equipment, and chemical/pharmaceutical manufacturers import 33% — the two most import-reliant categories of U.S. manufacturing in the current data, consistent with the growth trend measured here. Computers, non-auto capital goods, and automotive parts also grew well above the total-import pace, consistent with continued offshoring of electronics assembly and globalized auto-parts supply chains.

Biggest decrease: petroleum and products, -63% real. This is the single largest driver of the aggregate numbers moving the "wrong" way (i.e., overall import dependence falling): the U.S. shale oil and gas boom (roughly 2010-2020) cut U.S. crude oil import needs dramatically and the country became a net petroleum exporter around 2019-2020, a structural, well-documented energy-independence shift. Because petroleum was such a large share of goods imports in 2008 (over $693 billion in real terms, more than triple the pharmaceutical or semiconductor totals), its collapse mechanically drags down the whole "total goods imports" figure even though many other categories rose sharply — the total is not representative of what happened industry by industry.

Earlier years (pre-2008), for context only: This connector cannot query 2001-2008 trade detail, but the well-known academic "China Shock" literature (Autor, Dorn, and Hanson, NBER working papers, not re-verified against this warehouse) documents that import penetration from China in the most exposed U.S. manufacturing industries rose especially sharply following China's 2001 WTO accession, with import exposure growing roughly 0.5 percentage points per year on an employment-weighted basis from 1991-2011 and rising fastest 1999-2007. That means the sharpest rise in import PENETRATION specifically attributable to China happened disproportionately in the years just before this warehouse's data begins — a genuine limitation on how much of "the last twenty years" this report's own computed figures can speak to directly.

What This Report Does Not Answer

• 'Industries' framing vs. available goods-category data: The connector has no clean industry-level (NAICS) import-penetration series spanning 20 years — confirmed via find_recipe before querying. Genuine service-sector and true production-relative import dependence (e.g., BEA's Import Matrix / apparent consumption) exists only as an interactive web tool, not a bulk series, so it could not be pulled and joined here.

Every query behind this report

20 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT DISTINCT line_description FROM econ.trade_balance_summary ORDER BY 1

SELECT year, line_description, total_imports, total_exports
FROM econ.trade_balance_summary
WHERE quarter IS NULL AND line_description IN (
 'Imports of goods','Semiconductors and related devices','Telecommunications equipment',
 'Computers, peripherals, and parts','Pharmaceutical preparations, including vitamins',
 'Medicinal, dental, and pharmaceutical preparations, including vitamins',
 'Automotive vehicles, engines, and parts','Petroleum and products',
 'Apparel, footwear, and household goods','Industrial supplies and materials',
 'Capital goods and related parts, except automotive','Consumer goods, except food and automotive',
 'Household and kitchen appliances','Furniture and household items','Metals and products',
 'Chemicals','Fuels, fuel materials, and electric energy','Foods, feeds, and beverages'
) AND year IN ('2008','2024')
ORDER BY line_description, year

SELECT DISTINCT trade_type FROM econ.trade_statistics

SELECT line_number, line_description, "year", value FROM econ.trade_statistics
WHERE trade_type='imports' AND quarter IS NULL AND "year" IN ('2008','2024')
AND line_description ILIKE '%semiconductor%' OR line_description ILIKE '%pharmaceutical%'
ORDER BY line_description, "year" LIMIT 50

SELECT DISTINCT frequency, series_code FROM econ.trade_statistics
WHERE trade_type='imports' AND line_number='144' AND "year"='2008'

SELECT frequency, quarter, series_code, "value" FROM econ.trade_statistics
WHERE trade_type='imports' AND line_number='144' AND "year"='2008' AND quarter IS NULL

SELECT frequency, quarter, series_code, "value" FROM econ.trade_statistics
WHERE trade_type='imports' AND line_number='144' AND "year"='2008'

SELECT DISTINCT trade_type, line_number, line_description, frequency FROM econ.trade_statistics
WHERE line_description ILIKE '%pharmaceutical%'

SELECT DISTINCT line_description FROM econ.trade_statistics WHERE trade_type='imports' ORDER BY 1

SELECT DISTINCT line_description, trade_type FROM econ.trade_statistics WHERE line_description ILIKE '%Imports of%' LIMIT 30

SELECT DISTINCT frequency, series_code FROM econ.trade_statistics WHERE line_description='Imports of durable goods'

SELECT "year", line_description, "value" FROM econ.trade_statistics
WHERE frequency='A' AND line_description IN ('Imports of goods','Imports of durable goods','Imports of nondurable goods','Imports of capital goods less selected parts','Imports of services','Imports of nonpetroleum goods')
AND "year" IN ('2008','2024')
ORDER BY line_description, "year"

SELECT line_number, line_description, COUNT(DISTINCT "year") n_years, MIN("year") min_yr, MAX("year") max_yr
FROM econ.trade_statistics
WHERE frequency='A' AND line_description IN (
 'Semiconductors and related devices','Automotive vehicles, engines, and parts',
 'Apparel, footwear, and household goods','Computers, peripherals, and parts',
 'Chemicals','Petroleum and products','Metals and products',
 'Medicinal, dental, and pharmaceutical preparations, including vitamins',
 'Telecommunications equipment','Industrial supplies and materials',
 'Capital goods and related parts, except automotive','Consumer goods, except food and automotive'
)
GROUP BY line_number, line_description ORDER BY line_description, line_number

SELECT line_description, "year", CAST("value" AS DOUBLE) AS val
FROM econ.trade_statistics
WHERE frequency='A' AND line_number IN ('143','131','114','112','118','134','101','144','103','107','125','126')
AND "year" IN ('2008','2024')
ORDER BY line_description, "year"

SELECT "year", CAST("value" AS DOUBLE) AS gdp
FROM econ.national_accounts_enriched
WHERE table_id='T10105' AND line_description ILIKE 'Gross domestic product' AND frequency='A' AND "year" IN ('2008','2024')

SELECT * FROM (VALUES
 ('Semiconductors',25795,2008),('Semiconductors',82366,2024),
 ('Pharmaceuticals',78941,2008),('Pharmaceuticals',246894,2024),
 ('Computers',101222,2008),('Computers',217417,2024),
 ('Capital goods ex-auto',458698,2008),('Capital goods ex-auto',969077,2024),
 ('Automotive',233204,2008),('Automotive',475525,2024),
 ('Telecom equipment',45147,2008),('Telecom equipment',84525,2024),
 ('Consumer goods',485679,2008),('Consumer goods',806134,2024),
 ('Chemicals',54994,2008),('Chemicals',73576,2024),
 ('Metals and products',106927,2008),('Metals and products',137062,2024),
 ('Apparel/footwear/household',113671,2008),('Apparel/footwear/household',141381,2024),
 ('Industrial supplies',794578,2008),('Industrial supplies',654918,2024),
 ('Petroleum and products',476129,2008),('Petroleum and products',254080,2024),
 ('Total goods imports',2148678,2008),('Total goods imports',3267309,2024)
) AS t(category, amount, yr)

SELECT "year", CAST("value" AS DOUBLE) AS real_gdp
FROM econ.national_accounts_enriched
WHERE table_id='T10106' AND line_description ILIKE 'Gross domestic product' AND frequency='A' AND "year" IN ('2008','2024')

SELECT "year", metric, value FROM econ.gdp_statistics WHERE quarter IS NULL AND metric IN ('Nominal GDP','Real GDP') AND "year" IN (2008,2024) ORDER BY metric,"year"

SELECT DISTINCT metric FROM econ.gdp_statistics

SELECT DISTINCT type FROM econ.gdp_statistics