What has the industry actually committed to on child labour, and did it work?

In 2001 the cocoa and chocolate industry signed a voluntary agreement with two US legislators, setting out a six-point approach and a timetable for addressing the worst forms of child labour in cocoa growing. It is universally known by the two legislators' names.

The agreement was voluntary, and that is the structural fact that explains most of what followed. Its original deadline passed without the standard being met; the target was extended, reframed as a percentage reduction, and extended again. The reduction target was not achieved.

What the protocol did produce is institutional. A multi-stakeholder foundation was established in 2002 and works in the two largest producing countries. Child labour monitoring and remediation systems — household-level identification of children in hazardous work, followed by a remediation response such as school enrolment support or household income measures — became the dominant operational mechanism, and the major certification schemes and company programmes now largely point at them.

Coverage has grown substantially and remains partial. These systems operate where a company has a traceable supply chain to point them at, which by construction means the organised, certified, cooperative-supplied share of production rather than the whole.

It does not show that child labour in cocoa has been eliminated or substantially reduced. Successive assessments have found prevalence high in the major producing countries, and a monitoring system that identifies cases is evidence of identification rather than of reduction.

It does not show that the monitoring is ineffective. Systems that find and remediate individual cases do something a commitment document does not, and the honest position is that the mechanism works at the scale it reaches while the scale it reaches is not the sector.

It does not show that the industry was insincere. A voluntary agreement between competitors, covering millions of independent households in two sovereign states, is a genuinely hard instrument to make bite — and that difficulty is the argument for regulation rather than evidence about anybody's motives.

And it does not distinguish, on its own, between hazardous child labour and the ordinary family farm work that most agricultural societies practise. That distinction is definitional, is what the surveys actually measure, and is where a great deal of confused reporting originates.

The agreement, the institutions and the mechanism are matters of public record. Prevalence estimates come from large household surveys whose methodology and comparability between rounds are themselves debated, and the industry-funded share of the evaluation literature is high. The foundation that operates much of the monitoring was established by the industry it monitors, which is material and does not by itself invalidate its reporting.