Calendar Year or Marketing Year? The Question Beijing and Washington Never Answered on China’s Soybean Pledge

The White House fact sheets read like calendar-year commitments, but the administration’s own scorekeeping has drifted to marketing-year math on the 25 MMT soybean pledge — while the “prorated” $17 billion commitment only adds up on a calendar-year basis. With no signed text ever published, the yardstick may end up being chosen to fit the verdict.

Nine months after the Busan handshake and two months after the Beijing follow-up, one of the most consequential questions in U.S. agricultural trade remains officially unanswered: what clock is running on China’s purchase commitments? The question sounds technical. It is not. Whether China’s pledge to buy at least 25 million metric tons (MMT) of U.S. soybeans “in each of 2026, 2027, and 2028” is measured against calendar years or against the September/August marketing year determines whether Beijing is currently on pace or badly behind, when the make-or-break buying window arrives, and — ultimately — whether the administration can credibly declare the deal honored or breached.

The confusion is not the ag sector’s fault. No signed agreement text has ever been published — a sharp departure from the January 2020 Phase One accord, which was released in full with annexed purchase schedules keyed explicitly to calendar years. 

Everything the trade knows about the current commitments comes from two White House fact sheets and a running series of official characterizations that have not stayed consistent. When USTR Jamieson Greer told Congress in December there was a “discrepancy” between how the White House had described the first soybean deadline and “the actual deadline,” he confirmed what the grain trade had suspected: the public description and the private understanding are not the same document.

What the official documents actually say

Two texts carry the load. The Nov. 1, 2025, White House fact sheet, issued after the Busan summit, states: “China will purchase at least 12 million metric tons (MMT) of U.S. soybeans during the last two months of 2025 and also purchase at least 25 MMT of U.S. soybeans in each of 2026, 2027, and 2028.” The words “calendar year” never appear, but “the last two months of 2025” and “each of 2026, 2027, and 2028” read naturally as calendar years — and that is how the American Farm Bureau Federation and most general press initially scored the deal.

The May 17, 2026, fact sheet, issued after the Beijing summit, adds the second commitment: “China will purchase at least $17 billion per year of U.S. agricultural products in 2026 (prorated), 2027, and 2028, in addition to the soybean purchase commitments that it made in October 2025.” Greer has emphasized the $17 billion is “on top of” the soybean baseline and can include “soybeans… beef… grains… dairy products” — meaning soybean purchases above the 25 MMT floor could count toward the dollar target, but the baseline tonnage cannot.

Table 1. The two commitments side by side: what was written versus how officials have behaved.

The 12 MMT tranche already broke the calendar-year reading

The strongest single piece of evidence on the soybean question is not anything an official said — it is what the administration did with the first tranche. The November fact sheet was unambiguous: 12 MMT “during the last two months of 2025.” By early December, with only about 3 MMT booked, the deadline began to move. Greer told Congress the real deadline was the end of the “growing season,” citing the “discrepancy” between the White House description and the actual terms. Treasury Secretary Scott Bessent said China was reaching the “correct cadence” and would finish “by the end of February.” The White House eventually settled on end-February 2026, and an administration official explained that “soybean sales for a growing season typically happen between September to March of the next year.”

China completed the tranche — roughly 12.4 MMT — by late February. But note what that required: January and February 2026 purchases were credited backward to the 2025 commitment. Under strict calendar-year accounting, those same cargoes would have counted toward 2026’s 25 MMT. Once the administration allowed that, pure calendar-year scoring was dead. You cannot credit the same beans to 2025 in February and then insist the 25 MMT clock runs January through December.

USDA has moved to marketing-year language — and so has the trade

USDA Deputy Secretary Stephen Vaden supplied the closest thing to an official answer in June, telling the Wall Street Journal’s Global Food Forum that China has “the entire marketing period to meet the 25 million metric ton commitment for this year.” That is marketing-year framing from the department that will actually keep score — and it aligns with the plumbing: FAS export sales data, the only public real-time compliance instrument, is structured by marketing year, and the 2020 Phase One experience taught the trade that enforcement math inevitably migrates to the data that exists.

Much of the commercial trade has followed. StoneX analysts treat the 25 MMT as applying to the 2026/27 marketing year. Purchase trackers in the trade score the 12 MMT tranche against MY 2025/26 (complete, at roughly 103%) and the 25 MMT against MY 2026/27, where only about 1.2 MMT of new-crop bookings were on the books as of mid-July — a pace that looks alarming under a calendar-year reading and unremarkable under a marketing-year one, since the heavy Chinese buying window has always been October through February. Meanwhile USDA’s own balance sheet hedges: WASDE carries only about 15 MMT of Chinese demand, a 10 MMT gap between the political commitment and what the department treats as commercially bankable.

Three possible answers

With no controlling text, three interpretations circulate in the ag sector and grain trade. Each has evidence behind it, and each produces a different verdict on China’s current pace.
 

Table 2. Three interpretations of the purchase-commitment clock, and the pace verdict each produces.

Why the $17 billion almost certainly runs on calendar years

The dollar commitment is the easier call. “Prorated” is the operative word: the deal took effect in mid-May 2026, leaving roughly seven and a half months in the calendar year, which implies a 2026 floor somewhere near $10.5–$11 billion. Pro-rating a marketing year makes no sense for a basket that Greer says spans “wheat, feed grains, meat and non-food agricultural goods such as cotton and timber” — commodities with different and overlapping crop years. Dollar-value targets also follow the Phase One template, which was explicitly calendar-year (2020 and 2021) and scored against Census export data. If the two commitments end up on different clocks — tonnage on the marketing year, dollars on the calendar — that is not incoherence so much as an artifact of how each is measured: tonnage lives in FAS export sales data, dollars live in Census trade data.

Combined, the two commitments imply roughly $28–$30 billion in annual U.S. agricultural shipments to China in 2027 and 2028 — below the $38 billion peak of 2022 but a massive step up from 2025’s roughly $8 billion.

The economics argue the distinction will matter

If China were buying at a comfortable surplus to its commitments, the calendar/marketing-year question would be academic. It is not. U.S. soybeans still carry a 13% duty into China (a 10% retaliatory tariff stacked on the 3% MFN rate) against 3% for Brazil, and analysts have put the landed-cost gap near $48 per metric ton — roughly $500/mt CIF for U.S. origin versus $452/mt for Brazilian. That gap has made every documented Chinese purchase of U.S. beans a state-directed one: retired USDA economist Fred Gale’s analysis found that all 8.3 MMT of U.S. soybeans imported through May moved through Beijing-based state-owned firms filling government commitments, while private and provincial crushers bought Brazilian and Argentine origin on price. StoneX’s Arlan Suderman put it plainly: “Chinese buying is not for economical reasons of U.S. soybeans, but it’s for political reasons.”

Political buying is lumpy buying. It arrives when directed, not when the market signals — which is exactly why the scoring window matters. Under a marketing-year clock, Beijing can defer the bulk of its 25 MMT obligation to the October/February window, when Brazilian supplies are seasonally thin and state reserves logically restock, and still claim full compliance. Under a calendar-year clock, a thin first half puts Beijing visibly behind by summer and hands Washington a grievance. The choice of clock is, in effect, the choice of how much leverage each side holds in the interim — one more reason neither capital has rushed to nail it down.

The second ambiguity: purchases versus shipments

Layered on top of the calendar question is an unresolved measurement question. Officials consistently say “purchases” — which in export-sales terms means bookings. Phase One, by contrast, was scored on shipments, using Census data. The difference is not trivial: a cargo booked in November 2026 for March 2027 shipment counts toward 2026 under a purchases standard and toward 2027 under a shipments standard. The 12 MMT episode leaned toward the purchases interpretation (sales contracted by late February, with loadings stretching into March), but no official has ever said so in terms. Any serious compliance accounting will have to specify both the clock and the meter, and so far the administration has specified neither.

Bottom line

The weight of evidence favors the hybrid reading: the 25 MMT soybean commitment will be scored, in practice, on something close to a September/August season basis — whatever the November fact sheet’s calendar framing implied — while the $17 billion commitment runs on calendar years with 2026 prorated from mid-May. The strongest evidence is the administration’s own conduct: once January–February 2026 purchases were credited to the 2025 tranche, calendar-year accounting was abandoned in practice, and Vaden’s “entire marketing period” remark is as close to confirmation as anyone has offered on the record.

But the honest answer is that nobody outside the two governments has seen the controlling language, and the scoring rules may not be revealed until the administration wants to declare a verdict — at which point the yardstick may be chosen to fit the conclusion rather than the other way around. 

The tell to watch: how officials characterize China’s pace between now and October. If a thin summer draws no complaint from USTR or USDA, the marketing-year clock is the operative one. If the administration starts publicizing a 2026 shortfall before harvest, the calendar-year reading is back in play — as leverage, if nothing else.