I believe that the fundamental problem with agricultural yields in Africa (see Low yields 'due to wary farmers, not climate change') is that farm input technologies that increase yields result in lower prices, and therefore lower profit margins.

In the United States, this phenomenon was a driving force for the depopulation of rural areas as farm size increased to compensate for lower margins. Smaller rural populations can maintain a viable income, but the clear beneficiaries of agricultural technologies are the urban consumers who enjoy lower food prices.

The resultant greater discretionary income is not spent on increased consumption of rural products but on other urban products, so the farmer does not share in this benefit. African farmers have learned this, and also understand that they cannot increase their scale of production without land tenure to allow this.

The only way that rural people can equitably share in the benefits of farm technologies is through subsidies derived from taxation of urban sales of all items.

The problem is how to redistribute this revenue. Using the revenue as 'price support' for farmers — usually by directly buying the crop — is discredited because it causes market distortions. But we should consider other approaches to ensure greater equity in the benefits of technology.

Nations could agree to provide their farmers with fertilizer paid for by a tax on urban wealth, greatly increasing the efficiency of water use and helping to conserve the natural resources of soil, forests and associated biodiversity for all people.

Such a subsidy could be applied everywhere without really distorting the competitive position of any farm community at the expense of others. Productivity responses to fertilizer are such that the amount of subsidy would be self-regulating — application of fertilizers beyond a certain point provides no additional response, so the subsidy level will be regulated by a physical environmental factor rather than political need to maintain profitability of the enterprise.