What Could Ghanaian Cocoa Farmers Be Paid Between Now and 2035?

Every cocoa season, often around September or October, one announcement is watched closely across Ghana’s cocoa-growing communities: the producer price.

For many farmers, that figure is not just another government announcement. It affects household income, school fees, farm investment, hired labour, debt repayment and whether cocoa farming still feels worth the effort.

But must everybody wait until the official announcement before having any sense of where producer prices may be heading?

I wanted to find out.
So I brought together roughly two decades of data on Ghana’s cocoa producer prices, international cocoa prices, exchange rates and domestic cocoa production, then tested several forecasting methods to see what they could tell us about the road ahead.

The most important result was not a single number for 2035.

It was a year-by-year picture of what Ghana’s producer price could look like from 2026 through 2035 under three different economic conditions.

What the model projects
The figures below are average annual producer-price forecasts in Ghana cedis per metric tonne.

Those numbers immediately tell three very different stories.

Under the baseline scenario, producer prices remain relatively stable for a few years before beginning a stronger climb. The model moves from roughly GHS 56,957 per metric tonne in 2026 to about GHS 79,625 by 2035.

Under the high-price scenario, the path is much steeper. The producer price approaches GHS 100,000 per metric tonne by 2030 and crosses that level in 2031, eventually reaching about GHS 140,433 by 2035.

The low-price/stress scenario is perhaps the most sobering. Instead of continuing to rise, producer prices fall gradually for several years, bottoming out at around GHS 45,421 in 2031 before beginning to recover.

That is important because the extraordinary cocoa-price increases of recent years can easily create the impression that prices will simply keep going up.

The data suggest otherwise.
These are not government announcements

This point is important.
The numbers above are not predictions of what government will certainly announce in any particular cocoa season.

Ghana’s producer price is ultimately a policy decision. It depends on international cocoa revenues, exchange rates, forward sales, marketing costs, industry expenses, farmer-welfare considerations and the producer-pricing framework in place at the time.

The figures are better understood as conditional forecasts.

In other words, if international cocoa prices fall substantially and production recovers, one path becomes more likely. If the market stabilizes somewhere in the middle, another path emerges. If global supply remains tight, cocoa prices remain elevated and exchange-rate pressures continue, the high-price path becomes more plausible.

That is why there is no single “correct” line to 2035.

Why I decided to test this
The idea behind the exercise was fairly simple.

Ghana announces cocoa producer prices every season. Farmers wait. The public waits. Policymakers weigh international conditions and domestic realities.

But we now have enough historical data to ask whether Data Science can provide an evidence-based picture of where producer prices might plausibly be heading.

To test that, I compared several forecasting approaches.

They included:

• a simple 12-month benchmark;
• Linear Regression;
• ARIMA;
• Exponential Smoothing;
• Ridge Regression;
• Random Forest;
• and XGBoost.

I expected the more sophisticated machine-learning models to perform very strongly.

What happened instead surprised me.
Machine learning did not win
The best-performing model was Ridge Regression.

A simple 12-month benchmark came second.
Random Forest and XGBoost — two powerful machine-learning models — performed much worse.

The reason became clearer when I looked at what the models had actually seen during training.

The highest producer price available to them in the training period was about:

GHS 6,800 per metric tonne.
But in the final test period, the actual producer price eventually rose to GHS 33,120 per metric tonne.

Random Forest’s highest prediction remained around GHS 6,800.

XGBoost’s highest prediction was also around GHS 6,810.

In simple terms, the models had never seen anything remotely like the 2024 price regime.

They were good at recognizing patterns within the world they already knew.

They struggled when reality moved far beyond it.

That may be one of the most important lessons from the entire exercise.

A sophisticated algorithm is not automatically a superior forecasting tool.

Sometimes the biggest test of a model is what happens when the future no longer looks like the past.

Another surprising finding: about 59 per cent

There was another result that caught my attention.

During the project, I had a long conversation with a former university colleague who works within Ghana’s cocoa industry. He explained that, historically, the producer price paid to farmers could often be thought of as roughly 60 per cent of the international value of cocoa, with the rest supporting marketing, administrative and other industry costs.

Rather than simply accept that claim, I tested something close to it.

I converted the international cocoa price into Ghana cedis using the official exchange rate, then compared that value with Ghana’s producer price.

Across the historical period, the average implied farmer share came to approximately 59 per cent. That does not mean Ghana followed a rigid 60/40 rule every year.

It did not.
But the fact that the historical average landed so close to that industry rule of thumb was striking.

It also helps explain why international cocoa prices and exchange-rate movements matter so much for Ghanaian farmers.

Cocoa’s future will depend on more than cocoa prices

One of the clearest lessons from the forecasting exercise is that Ghana’s producer price cannot be understood by looking at international cocoa prices alone.

The exchange rate matters enormously.
Cocoa is sold internationally in US dollars, but farmers are paid in Ghana cedis.

So even if the international cocoa price remains unchanged, a weaker cedi can significantly increase the domestic value of cocoa.

Domestic production also matters.
When production falls sharply, supply tightens. When Ghana and Côte d’Ivoire both face production difficulties, the effect can ripple across the global cocoa market.

That is exactly what made the 2023–2024 period so extraordinary.

So when we ask what farmers might receive in 2028, 2030 or 2035, we are really asking several questions at once:

What will happen to global cocoa supply?
What will happen to international prices?
What will happen to the cedi?
Will Ghana’s production recover?
And how will government balance farmer incomes against the wider costs of running the cocoa sector?

What should policymakers take from this?

I do not believe Data Science should replace the judgment of policymakers.

But I do believe it can strengthen it.
Imagine a forecasting system that continuously updates as new information arrives.

If international cocoa prices fall sharply, the model updates.

If the cedi depreciates, it updates.
If production expectations deteriorate, it updates.

Instead of waiting for one final number at the beginning of a cocoa season, policymakers could work with a range of plausible producer-price outcomes months earlier.

They could ask:
What happens if cocoa falls to US$4,000 per tonne?

What happens if the cedi weakens further?
What happens if production returns above 850,000 tonnes?

What if West Africa experiences another major supply shock?

Those are not abstract questions.
They have direct implications for farmer incomes, government finances, cocoa-sector sustainability and national export earnings.

For farmers, the message is equally important.

The recent surge in producer prices has been extraordinary.

But extraordinary does not mean permanent.
The low-price scenario in the forecast shows that producer prices could actually decline for several years if world cocoa prices normalize and production recovers strongly.

The baseline scenario suggests a more moderate future.

And the high-price scenario shows just how far prices could rise if today’s supply problems persist.

That wide range should caution us against assuming that the future will simply be an extension of the last two years.

The future is not one number
That may be the most important conclusion from this entire exercise.

Forecasting is not valuable because it gives us certainty.

It is valuable because it helps us understand uncertainty.

By 2035, Ghana’s cocoa producer price could plausibly be around GHS 51,600 per metric tonne under one set of conditions, around GHS 79,600 under another, or above GHS 140,000 in a sustained high-price environment.

But perhaps the more interesting story lies in the years between now and then.

Under the baseline scenario, the model suggests relative stability through the late 2020s before prices begin climbing more strongly.

Under the high-price scenario, the producer price crosses GHS 100,000 per metric tonne around 2031.

Under the stress scenario, prices fall before recovering.

Those are very different futures.
And that is precisely why farmers, policymakers and the cocoa industry should be thinking not about one forecast, but about a range of possible outcomes.

The model has not discovered what Ghana’s producer price will be in 2035.

No model can.
What it has done is something more useful: it has given us a way to see several possible futures before we arrive there.

About the author
Author’s contact: [email protected]

Stephen Sarpong Lartey is a doctoral candidate and a Data Scientist whose work focuses on the intersection of Data Science, economics, agriculture and public policy. A separate, more detailed technical version of this analysis-including methodology, feature engineering, model diagnostics and the full forecasting framework - is available on Medium at https://medium.com/@millano2895/forecasting-ghanas-cocoa-producer-prices-to-2035-can-machine-learning-outperform-traditional-7bccd651abec