How many tonnes of cocoa beans will the Ivory Coast produce in 2026?

TL;DR

My forecast is 1.98 million metric tonnes of cocoa beans in calendar year 2026. The median is 1.98 million tonnes, with an 80% interval of 1.77–2.18 million and a 90% interval of 1.69–2.24 million. A strong 2025/26 crop is being partly offset by a weaker and later September–December 2026 crop, putting the calendar-year result below the full old-campaign total.

Context

The target is January–December production, but most timely cocoa data use a crop year or count beans arriving at ports. ICCO's bulletin posted on August 31, 2026 withheld 2025/26 production data, while the FAO annual series currently ends in 2024; Côte d'Ivoire's national statistical bulletin supplies a 2025 calendar-year figure of 1,840,311.5 tonnes (ICCO; FAO via OWID; ANStat).

Côte d'Ivoire also changed its domestic marketing calendar in 2026: the main crop now runs September–February and the intermediate crop March–August (AIP). Calendar 2026 therefore combines January–August from the old campaign with September–December from the new one. September must be added separately when using the widely reported October–December arrival figures.

Evidence

I use 2010–2025 as the modern reference window. This is the full history for that window; the pre-2010 series sits in a lower production regime and is less comparable (FAO via OWID).

Across these 16 observations, the mean is 1.846 million tonnes and the sample standard deviation of the 15 annual changes is 209,000 tonnes. The last decade has included a 537,000-tonne fall from 2022 to 2023, so a narrow crop forecast would be false precision.

The cleanest direct calendar nowcast starts from the official 2025 total. Q1 2025 production was 334,221.3 tonnes, and ANStat says Q1 2026 was 43.6% higher (2025 economic report; ANStat Q1 2026). Replacing only Q1 2025 with the implied Q1 2026 amount gives about 1.986 million tonnes:

$$1.8403 + (0.3342 \times 1.436 - 0.3342) = 1.9860$$

All terms are millions of metric tonnes. This is a useful neutral baseline because it uses a calendar-year series and does not double-count the crop-year evidence.

The old campaign then pushes the estimate up, while the new campaign pulls it down. The regulator forecast 2.0–2.1 million tonnes for 2025/26 on May 21, 2026; an August 31 national report put the campaign near 2.2 million; port arrivals reached about 2.14 million by August 30; and the regulator reported 650,000 tonnes processed locally, equal to a rounded 30% of campaign production (Reuters; AIP campaign estimate; arrivals; AIP processing). I read this as a campaign total near 2.15 million, not as proof of 2.2 million tonnes of biological production: arrivals can include delayed stock and cross-border beans.

Late 2026 is weaker. Reuters reported an eight-to-ten-week delay and expected about 900,000 tonnes to reach ports in October–December 2026, versus a revised 1.1 million in October–December 2025 (Reuters). A July survey of four pod counters and five exporters put the September–February main crop at 1.35–1.45 million tonnes, down from about 1.6 million (Reuters). By September 20, the new traceability system had recorded 50,085 tonnes purchased, while 21,467 tonnes had physically reached exporters at the ports; these are different stages and must not be added (CCC data via AIP; port receipts). The data says delayed crop, not collapse.

My literal calendar bridge is:

$$2.15 - 1.10 + 0.04 + 0.90 = 1.99$$

The terms are the 2025/26 campaign total, October–December 2025 production proxied by revised arrivals, September 2026 production, and October–December 2026 production proxied by expected arrivals, all in millions of tonnes. A retrospective bridge using ICCO crop totals and Reuters/ICCO fourth-quarter arrivals had a mean absolute error of about 22,000 tonnes in calendar years 2021–2024, but it missed the official 2025 annual figure by about 113,000 tonnes when ICCO's 1.681-million-tonne 2024/25 crop estimate was used (ICCO historical production; ICCO recent production; ICCO 2024/25; 2020 arrivals; 2021 arrivals; 2022 arrivals; 2023–24 arrivals; 2025 revised arrivals). That failure is why the final distribution is much wider than the four-year backtest alone suggests.

I combine five alternative frames rather than treating the bridge as exact:

The mixture mean is 1.976 million tonnes and the median is 1.981 million. It assigns 13% below 1.8 million tonnes, 45% above 2.0 million, and 8% above 2.2 million.

What's non-obvious

The marketing-calendar change adds a September term. Using 2.15 − 1.10 + 0.90 gives 1.95 million tonnes, but that omits the first month of the new campaign. September purchases and port receipts support an increment near 40,000 tonnes, moving the literal bridge to about 1.99 million (AIP). This small accounting detail explains much of the gap between a 1.95-million and a 2.0-million forecast.

The official 1.8-million-tonne budget forecast is stale. The same planning table forecast 1.71 million tonnes for 2025, while the later official 2025 result was 1.840 million; it set 2026 at 1.80 million before the 43.6% Q1 rebound and the strong old-campaign flows were known (budget projection; ANStat 2025; ANStat Q1 2026). I give that number little weight.

Uncertainties

• No official Q2 or Q3 2026 calendar-production tonnage was available at the cutoff. An ANStat first-half or nine-month table, using the same DGE/DPPSE definition as the 2025 annual figure, would remove more uncertainty than another port-arrivals report.
• October–December 2026 is still a forecast. Below-average September rain, black-pod risk and the expected very strong El Niño raise downside risk, but El Niño intensity does not determine rainfall in Côte d'Ivoire one-for-one (Reuters weather report; NOAA; WMO).
• The resolving series may revise history or classify stocks and cross-border flows differently. ICCO has withheld 2025/26 production, FAOSTAT will arrive later, and the question permits a crop-year proxy if a calendar number is unavailable (ICCO; FAO metadata). I keep proxy risk small because a FAOSTAT 2026 observation should be available before the June 2029 deadline.