Ag futures move fast and hard on tariff news, but whether the tariff is actually enacted matters at least as much as the target's raw export share

Summary

Ag futures react within hours, not days. CBOT nearby soybean futures fell about 5% in the first few hours after China published its retaliatory tariff list on April 4, 2018, and CME lean-hog futures fell 2.3% within the first minutes of trading after Mexico's June 2018 pork-tariff announcement. Over the following weeks and months the moves compound to much larger declines: soybeans fell roughly 24% peak-to-trough over about 4.5 months in 2018, and hogs fell roughly 17% from February to June 2018.

Whether the size of that move tracks the announced country's real export share was tested two ways on the 3 episodes assembled here, and both tests are underpowered by design (n=3 to n=6) but point the same direction. An OLS regression of near-term futures move (%) on market share (%) gives a positive slope (0.092 pp of move per pp of share) that is not statistically significant (p=0.50, n=3). A diff-in-differences comparing episodes where the tariff was actually enacted (soybeans, pork) against the one where it was only threatened (corn) finds the enacted group's price move grew 16.9 percentage points more from the near-term reaction to the cumulative decline than the threat-only group's did -- also not significant at conventional levels (p=0.12, n=6), but the largest and most confidently-signed effect of the two tests. Descriptively: China's 2018 soybean tariff (57% of the US soybean export market by value) produced the largest move of the three; Mexico's 2018 pork tariff (roughly a quarter of the US pork export market) produced a smaller but still sharp move; Mexico's 2025 corn tariff threat, despite a larger market share (40%) than the pork case, produced only about a 1% futures move because it was never actually enacted specifically on corn. With only 3 episodes, neither test can settle the question -- but both point toward enactment mattering at least as much as raw announced market share.

How fast and how big: three episodes

Market share figures are computed directly from econ.trade_exports (US Census international trade time series), filtered to geo_level='country' to avoid double-counting Census's own bloc/continent/world aggregate rows. Soybean HS6 code 120190; corn HS6 code 100590; pork HS6 codes starting 0203 (fresh/chilled/frozen pork cuts -- narrower than full USDA pork trade, so the true Mexico pork share including processed/other cuts is likely somewhat higher than the 26% shown here). Futures price moves and dates are drawn from contemporaneous agricultural-economics and market reporting (farmdoc daily, AgWeb, porkcheckoff.org, DTN), not from a futures-price table in this corpus -- see the data-gap note below.

Two small statistical checks, and why neither can settle the question

Test 1 (share alone). OLS regression of near-term futures move (%) on announced market share (%), across the 3 episodes: intercept = -1.02 (SE 4.01), slope = 0.092 (SE 0.094, t=0.99, p=0.50), R²=0.49, n=3, 1 degree of freedom.

Test 2 (enactment vs. threat). A difference-in-differences comparing 'enacted' episodes (soybeans, pork -- treatment=1) against the 'threat-only' episode (corn -- treatment=0), with 'post' distinguishing the near-term reaction from the later cumulative decline: the treatment×post (DiD) coefficient is +16.85 percentage points (SE 6.50, p=0.12, n=6). In plain terms: for the episodes where the tariff was actually imposed, the price move grew nearly 17 points more between the near-term reaction and the eventual cumulative decline than it did for the threat-only episode -- the enacted tariffs kept compounding into larger losses over time, while the threatened-but-never-enacted one did not. This result should be read cautiously: it rests on only 2 treated units and 1 control unit, it assumes parallel pre-trends that cannot actually be tested with data this thin, and 'post' here distinguishes reaction horizon rather than a true pre/post-announcement period, so it is a looser design than a standard policy diff-in-diff.

Neither test reaches conventional statistical significance, and neither should be treated as more than descriptive given how few tariff episodes have comparable, well-documented futures reactions. Two structural factors plausibly explain more of the variation than market share alone can, in this small sample: (1) whether the tariff is actually enacted on that specific commodity, not just threatened -- directly visible in the corn/Mexico 2025 case, which had a larger market share than pork/Mexico 2018 (40% vs 26%) but the smallest price reaction of the three, because it was never enacted specifically on corn; and (2) available substitute supply -- a University of Georgia/UC Davis working paper on the 2018 soybean episode (Adjemian, Smith & He, 2021) reports an estimated $0.74/bu average lower US Gulf export soybean price over about 5 months and attributes part of that size and duration to the time China's substitute supplier, Brazil, needed to absorb the redirected demand; separately, China's share of total US soybean exports (computed here from econ.trade_exports) had only partly recovered by 2024 (~51%) relative to 2017 (~57%), consistent with a lasting rather than purely temporary shift in buying pattern.

Data gaps and what this corpus could not answer directly

This connector's warehouse (askengine) does not carry a daily agricultural commodity futures price series -- search_catalog for soybean/corn/wheat futures returned no matching table (only unrelated series: SEC equity stock_prices, EIA natural gas spot price, CFTC swaps-transaction data). The cftc.cot_disaggregated_futures table carries weekly Commitments-of-Traders positioning (open interest and long/short by trader category) for futures markets, which could support a follow-up on whether speculative positioning shifted around these announcements, but not the daily settlement price itself. All futures price levels and percentage moves in this report are therefore taken from contemporaneous agricultural-economics literature and market reporting (farmdoc daily/University of Illinois, AgWeb, DTN, porkcheckoff.org), not recomputed from a raw price table in this corpus -- flagged here per the sourcing check rather than presented as if read from a primary price feed. Both regressions above use only these same 3 hand-assembled episodes (6 rows for the diff-in-diff) as their data, because no systematic tariff-announcement/futures-move events table exists anywhere in this corpus -- a genuine coverage gap, not a shortcut.

What This Report Does Not Answer

• Use of real agricultural futures price data: This corpus (askengine) has no daily ag futures price table (confirmed via search_catalog: only equities/energy/NatGas price series exist, plus weekly CFTC positioning, not daily settlement prices). Futures moves are sourced from contemporaneous ag-economics literature and market reporting instead, and this is flagged explicitly in the report rather than presented as a primary-source price read.

Every query behind this report

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

SELECT country_name, SUM(value_usd) AS val
FROM econ.trade_exports
WHERE hs6='120190' AND "year"='2017' AND geo_level='country'
GROUP BY country_name
ORDER BY val DESC
FETCH FIRST 10 ROWS ONLY

SELECT SUM(value_usd) AS world_total
FROM econ.trade_exports
WHERE hs6='120190' AND "year"='2017' AND geo_level='world'

SELECT country_name, SUM(value_usd) AS val
FROM econ.trade_exports
WHERE hs6='120190' AND "year"='2024' AND geo_level='country'
GROUP BY country_name
ORDER BY val DESC
FETCH FIRST 5 ROWS ONLY

SELECT SUM(value_usd) AS world_total
FROM econ.trade_exports
WHERE hs6='120190' AND "year"='2024' AND geo_level='world'

SELECT country_name, SUM(value_usd) AS val
FROM econ.trade_exports
WHERE hs6='100590' AND "year"='2024' AND geo_level='country'
GROUP BY country_name
ORDER BY val DESC
FETCH FIRST 8 ROWS ONLY

SELECT SUM(value_usd) AS world_total
FROM econ.trade_exports
WHERE hs6='100590' AND "year"='2024' AND geo_level='world'

SELECT country_name, SUM(value_usd) AS val
FROM econ.trade_exports
WHERE hs6='020329' AND "year"='2017' AND geo_level='country'
GROUP BY country_name
ORDER BY val DESC
FETCH FIRST 8 ROWS ONLY

SELECT SUM(value_usd) AS world_total
FROM econ.trade_exports
WHERE hs6='020329' AND "year"='2017' AND geo_level='world'

SELECT country_name, SUM(value_usd) AS val
FROM econ.trade_exports
WHERE hs6 LIKE '0203%' AND "year"='2017' AND geo_level='country'
GROUP BY country_name
ORDER BY val DESC
FETCH FIRST 6 ROWS ONLY

SELECT SUM(value_usd) AS world_total
FROM econ.trade_exports
WHERE hs6 LIKE '0203%' AND "year"='2017' AND geo_level='world'

SELECT * FROM (VALUES (57.0, 5.0, 24.0), (26.0, 2.3, 17.0), (40.0, 1.0, 1.0)) AS t(market_share_pct, near_term_move_pct, cumulative_move_pct)

SELECT * FROM (VALUES
 ('soybeans_china', 1, 0, 5.0),
 ('soybeans_china', 1, 1, 24.0),
 ('pork_mexico', 1, 0, 2.3),
 ('pork_mexico', 1, 1, 17.0),
 ('corn_mexico', 0, 0, 1.0),
 ('corn_mexico', 0, 1, 1.0)
) AS t(episode, treatment, post, price_move_pct)