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

TL;DR

I forecast 1.92 million metric tonnes of cocoa beans for Côte d’Ivoire in calendar 2026. My 10th–90th percentile range is 1.72–2.11 million tonnes. The strong 2025/26 tail is being offset by a weaker crop starting in September, while headline port arrivals include delayed stocks and are not a clean measure of beans newly harvested in 2026 (CCC outlook, new-crop survey, stock-removal program).

Context

The cleanest historical target is an annual production figure. FAO defines the measure as the amount produced in the year; the available vintage contains 64 annual observations for 1961–2024, in metric tonnes, and was updated on February 25, 2026 (FAO data via Our World in Data). Côte d’Ivoire’s own economic statistics then reported 1,840,311.5 tonnes for calendar 2025, down from 1,890,774.9 tonnes in 2024 (ANStat/DGE December 2025 table).

Calendar 2026 spans two harvest cycles. The government moved the main campaign to September 1–February 28 and the intermediate campaign to March 1–August 31, so January–August 2026 mainly belongs to 2025/26 and September–December mainly belongs to 2026/27 (official calendar change). The regulator’s May estimate was 2.0–2.1 million tonnes for 2025/26, while a July Reuters poll put the two full crop years at 1.98 million and 1.80 million tonnes respectively (CCC estimate, Reuters poll).

Evidence

The historical backbone is the full FAOSTAT series below: Côte d’Ivoire, cocoa beans, production, tonnes, 1961–2024, N=64, retrieved February 25, 2026 (FAO/OWID dataset and metadata).

The relevant modern regime centers near two million tonnes but has wide annual swings. The 2015–2024 mean is 2,031,257 tonnes; the 2020–2024 mean is 2,100,067 tonnes; and the sample standard deviation of annual changes over 2010–2024 is about 215,000 tonnes. The fall from 2,358,991 tonnes in 2022 to 1,822,441 in 2023 shows why a narrow interval would be false precision (same FAO series).

The ending 2025/26 crop is stronger than early forecasts. The CCC said on May 21 that it expected 2.0–2.1 million tonnes and had recorded more than 1.7 million tonnes of port arrivals through May 11 (Reuters/MarketScreener). A market tracker put arrivals at 2.10 million tonnes through July 19, 21% above the prior year (CocoaIntel). But the government was moving 123,000 tonnes of residual cooperative stocks toward the ports from late January, and market reports recorded an upward arrival-data revision of more than 260,000 tonnes in June (government stock program, arrival revision report). I therefore center true 2025/26 production near 2.03 million tonnes rather than treating 2.10 million arrivals as a hard lower bound.

The crop beginning in September is the main downside force. Four pod counters and five exporters put the September–February main crop at 1.35–1.45 million tonnes, against about 1.6 million for the current main crop, after heavy rain, flower and cherelle losses, black-pod pressure, and weak crop treatment (Reuters/Business Recorder). The broader July Reuters poll put full 2026/27 output at 1.80 million tonnes, down from 1.98 million in 2025/26 (Reuters poll). NOAA assigned an 81% chance to a very strong El Niño during October–December 2026, but local evidence is less one-way: SODEXAM expected broadly normal rainy-season totals, and late-July dry spells and sunshine modestly improved the main-crop outlook (NOAA outlook, SODEXAM interview, late-July update). Swollen-shoot disease remains a structural drag, with Enveritas estimating 15% of national supply at risk; the April 2026 regional roadmap will not restore infected mature trees during this forecast window (Enveritas, FAO roadmap).

My central calendar bridge is an inference, not a published identity. I put 45% weight on the portion of 2025/26 harvested or attributed to January–August 2026 and 55% on the front-loaded September–December portion of 2026/27. The larger new-crop weight reflects the September start and the concentration of main-crop deliveries early in the campaign; official reporting already showed nearly one million tonnes delivered by December 29, 2025 (campaign timing, year-end government update). The central arithmetic is:

$$0.45 \times 2.03 + 0.55 \times 1.80 = 1.90\ \text{million tonnes}.$$

Here 2.03 is the estimated 2025/26 crop, 1.80 is the estimated 2026/27 crop, and the weights approximate the shares falling inside calendar 2026.

The final predictive distribution is a five-part mixture: 76% baseline calendar case centered at 1.91 million tonnes with a 120,000-tonne standard deviation; 8% severe downside centered at 1.68 million with a 130,000-tonne standard deviation; 10% allowed 2025/26 crop-year proxy centered at 2.06 million with a 90,000-tonne standard deviation; 4% upside production or revision case centered at 2.16 million with a 100,000-tonne standard deviation; and 2% heavy-tailed model risk. It has a mean of 1,916,400 tonnes and a median of 1,920,209 tonnes. The 5th–95th percentile range is 1.65–2.16 million tonnes; the chance of exceeding 2.0 million tonnes is 29%, and the chance of falling below 1.8 million is 21%.

What's non-obvious

Port arrivals are not production. The arrival total is useful evidence that 2025/26 was not another disastrous crop, but it mixes harvest timing, delayed cooperative stocks, reclassification, quality acceptance, and movement into the two ports. The fact that arrivals reached roughly the regulator’s full-season production range before the season ended is evidence of a measurement mismatch, not proof that fresh production must exceed that range (CCC range, July arrivals, official residual stocks).

The other trap is to turn “very strong El Niño” into a fixed production haircut. NOAA’s Pacific forecast is strong, but the crop responds to local rain timing, sunshine, humidity, farm treatment, and disease. Côte d’Ivoire moved from excess-rain and cloud concerns in June to a modest sunshine improvement in late July, so I use El Niño to widen and skew the distribution rather than mechanically subtracting 15% (NOAA discussion, June farmer report, late-July update). FAOSTAT’s current annual series ends in 2024 and is updated yearly; if that cadence continues, a 2026 annual observation should exist well before the June 30, 2029 deadline, which is why I give the permitted crop-year proxy only 10% weight (FAO update schedule).

Uncertainties

• No public source gives a clean monthly series of beans biologically harvested. Port arrivals are the available high-frequency proxy, but the 123,000-tonne stock-removal program and large arrival revisions show why that proxy is noisy (official stock program, arrival revision report).
• The public ICCO release dated May 29, 2026 revised the global 2024/25 balance but did not disclose a free country-level 2025/26 Côte d’Ivoire figure. A paid country table, the August or November bulletin, or a direct ICCO calendar-year series would reduce source uncertainty (ICCO May bulletin).
• The September–December 2026 crop was not fixed on the forecast date. August–September pod counts, sunshine, rainfall distribution, black-pod incidence, and the share of the 1.35–1.45 million-tonne main crop harvested before year-end could move the answer by more than 100,000 tonnes (Reuters pod survey).
• Final resolution may still depend on methodology. The national annual series, FAOSTAT, ICCO crop-year estimates, and USDA attaché figures can differ because of timing, revisions, stocks, and informal cross-border flows (national annual table, FAO annual series, ICCO statistics page, USDA attaché report). The most useful missing document would be an authoritative reconciliation of the 2026 calendar year to the 2025/26 and 2026/27 campaigns.