Market Check · August 20, 2026

Cocoa Rebounded Sharply. That Still Isn’t Proof of an El Niño Crop Loss.

Three weeks ago this site reported cocoa down 21% from its July peak. It has since recovered about two-thirds of that decline and remains below the peak. What has not arrived is any measurement of a 2026/27 crop loss. The gap between a repriced forecast and a measured shortfall is the most useful thing in this market right now.

By Christopher W. Corwin · IAMElNino.com · 8 min read
← All Posts

What this is. Educational context on how weather information reaches commodity markets. Not investment advice, and no trade recommendation. Every price below has multiple drivers and El Niño is one input among many.

On 1 August, Market Watch Edition 2 led with an uncomfortable fact: over the second half of July, Ghana’s regulator forecast a 16% production fall, citing weather alongside disease and the trees’ natural production cycle, visible stocks drew down, and cocoa fell about 21% anyway. The lesson was that a price cannot be read off a single weather narrative.

Three weeks later the market has gone the other way, and the lesson holds in reverse.

~2/3
of the July slide recovered, indicative
+20%
2025/26 Ivory Coast port arrivals, y/y
surplus
2026/27 balance, still projected

The Cocoa Move, With Its Caveats

On the approximate benchmark readings used in this series, cocoa rose about 18% from 30 July to 19 August, standing near $6,031 per tonne against the $5,112 recorded in Edition 2. Because the underlying contract series could not be confirmed, that percentage should be treated as indicative rather than exact. On the same basis the level remained below the $6,455 peak of 9 July, and the recovery covered roughly two-thirds of the decline rather than all of it.

These are not exchange settlements, and the distinction is not cosmetic. Barchart shows the September ICE contract closing 2.39% higher on 19 August, while the aggregator series we used implies about 1.9%. Different continuous-contract constructions also produce different July highs and lows. We could not establish a single identified contract and settlement series covering all three dates, so the 18% figure should be read as a property of this benchmark series rather than an auditable move in a named contract.

DateLevel ($/tonne)Change from prior rowWhat this site said at the time
9 July 2026~6,455eight-month highpeak coincided with CPC’s 81% figure
30 July 2026~5,112~−21%“possible weather-premium unwind”
19 August 2026~6,031~+18%this article

Percentage changes are IAMElNino.com calculations from the levels shown and are indicative, not exact. Sourcing caveat below: these are reported levels from market data aggregators, not exchange settlement records, and the 9 July and 30 July figures were described as approximate when first published.

A sourcing limitation, stated plainly. Reuters is not accessible to this site, and we did not route around that. We were also unable to retrieve exchange-primary settlement data from ICE or CME for these dates. Every price in this article is a reported level from a market data aggregator or secondary market report, and should be treated as approximate and possibly intraday rather than as an audited settlement. Where we could not establish a contract month, we say so. This matters more than usual in an article about price precision.

The Physical Record Is From a Different Crop Year

Cumulative Ivory Coast port arrivals for the 2025/26 marketing year, from 1 October 2025 through late July 2026, were reported at about 2.11 million tonnes, roughly 20% above the same period a year earlier. Ivory Coast is the world’s largest producer, and this is a measurement of beans that physically reached ports rather than a forecast. It is also not a shipment or export figure.

Crucially, that is the wrong crop year to test the bullish case. The arrivals above belong to 2025/26. The surplus revision and the Ghanaian shortfall forecast both concern 2026/27, the crop now being set. Strong current-season arrivals do not disprove damage to the coming crop. They establish only that cumulative beans reaching Ivory Coast ports were running substantially ahead of the previous season.

The broker StoneX was reported in late July to have cut its 2026/27 global surplus estimate to about 25,000 tonnes, from roughly 149,000 tonnes in April. That is a large proportional cut. Note what it is not: it is still a surplus. The projected balance moved closer to zero without crossing it.

So the cocoa complex right now contains, simultaneously: a 2026/27 production forecast that has deteriorated, a 2025/26 physical flow that ran a fifth higher than the year before, and a projected 2026/27 balance that remains in surplus. A price well off its recent low is consistent with all three, because the price is trading a crop that has not been grown yet.

The Evidence Table

Classifications below are IAMElNino.com analysis, not agency or exchange designations.

CommodityWhat the market is doingImmediate driverPhysical loss confirmed?Observation windowAttribution
Cocoa~$6,031, up sharply from 30 Jul, still below 9 Jul peak2026/27 crop forecasts, surplus revision2026/27 loss not yet measurable; 2025/26 arrivals ~20% higherOct–Dec main crop pod fillProbable weather-risk component
Coffee, arabica (KCU26)closed down ~1.07% on 19 Aug (one session)not establishedNoSep–Oct 2026 Brazil flowering, for the crop harvested in 2027Insufficient evidence
Coffee, robusta (RMU26)closed down ~0.99% on 19 Aug (one session)not establishedNoOct–Dec Vietnam harvestPlausible future exposure
Wheat (CBOT)+17% since the start of JulyBlack Sea port attacks and export disruptionNot applicable: export logistics, not yieldAustralia Sep–NovPrimarily non-ENSO
Sugar (#11 SBV26)marginally higher on 19 Aug (one session)mixedNoIndia and Thailand, Oct–DecInsufficient evidence
Palm oilnot verified this weeknot establishedNofresh fruit bunch effects may emerge after several months and can persist beyond a yearInsufficient evidence

Three of six rows read “insufficient evidence.” That is the honest state of the record on 20 August, not a gap we intend to fill with narrative.

Coffee Fell on 19 August

Worth stating because the framing of this piece began with an assumption that coffee was rising. On 19 August, September arabica (KCU26) closed down about 1.07% and September robusta (RMU26) down about 0.99%. That is one session and establishes no trend in either direction; we did not obtain a multi-day series. Edition 2 recorded arabica up 9.2% year on year at the end of July, driven by physical certified stocks rather than drought.

The window that matters for an El Niño coffee story has not opened. Brazilian arabica flowering runs roughly September into October 2026 and sets the crop harvested in calendar 2027, marketed in 2027/28. Until flowering, rainfall and subsequent crop assessments are in, a loss in that crop is not something anyone can claim, and we are not going to claim it.

Wheat: The Counterexample, Restated Briefly

Chicago wheat has risen more than 17% since the start of July, trading around $6.70 a bushel in mid-August. This site already classified wheat as “global, not ENSO-attributable” on 1 August, so the classification is not new. What is new is how specific the non-ENSO driver has become.

The immediate port disruptions and export reductions are not ENSO effects, although weather remains another influence on the wheat market. Two further points cut against reading this as a global shortage: USDA raised projected world ending stocks slightly in its August estimates, which is what a trade-flow and logistics squeeze looks like rather than a supply collapse; and Australian production remains a genuine El Niño exposure whose September to November window has not yet run. Folding a Black Sea shipping crisis into an “El Niño food shock” would be the single easiest error to make this month.

Five Stages, Routinely Collapsed Into One

Almost every dispute about El Niño and food prices comes from treating these as the same event. They are sequential, and each transition can fail.

  1. A futures contract responds to a forecast. Cocoa is here. This can happen in an afternoon and can reverse just as fast, as July demonstrated in both directions.
  2. Weather deteriorates in a producing region. Requires observed rainfall and temperature anomalies, not a seasonal outlook.
  3. A crop loss is measured. Requires harvest data or a credible field assessment for the crop in question. The 2026/27 harvest has not happened.
  4. A supply deficit is confirmed. Requires the global balance to actually cross into deficit. Cocoa’s projected 2026/27 balance has not.
  5. Retail prices rise. Retail prices can lag commodity moves by months because of inventories, contracts, hedging and company pricing decisions. Cocoa is only one component of a finished product’s retail cost.

We are at stage one in cocoa. Stage three has not yet been reached for the 2026/27 crop, which is precisely why the price move cannot settle the question either way.

What Would Confirm the El Niño Trade

Ivory Coast arrivals falling behind the prior year, accompanied by field assessments or production data linking the decline to adverse weather. Arrivals alone would not settle attribution, because port logistics, farmer selling behaviour and inventory timing all move them. Ghana’s forecast shortfall appearing in actual purchases rather than projections. A published global balance crossing from surplus into deficit. Observed rainfall deficits in West Africa during pod fill. Brazilian flowering assessments in September and October showing damage.

What Would Weaken It

2026/27 arrivals running ahead of the prior year once the main crop begins. The surplus estimate stabilising or widening again. Normal rainfall across West Africa in the fourth quarter. Cocoa retracing toward the late-July level on unchanged weather, which would suggest the move was positioning rather than fundamentals. Any of these would say the premium was priced ahead of a crop that did not fail.

The market does not wait for a crop failure. It prices the possibility first. That makes futures useful as an early warning, and it makes them poor evidence for anyone trying to prove that El Niño has already damaged supply.

Method and limitations. Reuters is not accessible to this site and was not routed around; three Reuters reports of 18 and 20 August were the original starting points for this piece and none could be read. Exchange-primary settlement data from ICE and CME could not be retrieved. All price levels are reported figures from market data aggregators or secondary market reports, are treated as approximate, and may be intraday rather than settlement. Contract months are given where identifiable and omitted where not. The 9 July and 30 July cocoa levels are carried forward from this site’s 1 August edition, where they were also described as approximate, and we could not confirm that all three dates come from one identified contract and settlement series. Barchart’s September ICE contract shows a 2.39% gain on 19 August against roughly 1.9% on the aggregator series, which is direct evidence the series differ. The Ivory Coast arrivals cutoff was reported variously as 26 July and 2 August; we use “late July” rather than assert either. The StoneX surplus revision and Ghana’s production forecast are private and institutional figures reported secondhand; they are directional, not auditable, and this site’s June article referenced an earlier StoneX revision that may or may not be the same one. Ivory Coast arrival totals are cumulative marketing-year figures as reported, not an official government release retrieved directly. No El Niño attribution is asserted for any price move in this article. ENSO probability figures are RONI-based and are discussed in Edition 3.

Sources

← All Posts Live dashboard →