Data from children suggest that the prevalence of child labour in the cocoa industry in Côte d’Ivoire is a worrying 38%. However, while surveying children has the potential to reduce sensitivity biases such as social desirability bias, it presents significant ethical dilemmas and may also be associated with other reporting biases, making accurate estimates of child labour difficult. To address this, we collected data from 1741 cocoa producers to assess whether parents could provide more accurate estimates using indirect survey methods. We compared direct questioning with a list experiment and a novel non-verbal method (‘colorbox’). We found higher prevalence rates of child labour using indirect elicitation methods, but lower than those obtained from children’s surveys. Qualitative in-depth interviews revealed motivations for underreporting, including fear of legal consequences and mistrust of stakeholders. Indirect methods alone are not sufficient to correct for underreporting when child labour is collected from parents. Future research should prioritise direct data collection from children and address ethical concerns to obtain more accurate estimates of child labour.