Appendix C: Characteristics of monitoring agents

Granular demographic information on monitoring agents was available for this review only for ICI implemented CLMRS in Côte d’Ivoire. This information was merged with the interview data collected in order to understand how the profiles of monitoring agents relate to the interview outcomes. In the ICI CLMRS data base, we have basic demographic information, with varying detail, of 2’246 community-based agents in Côte d’Ivoire who have been hired and trained for child labour data collection under the CLMRS. This section presents and discusses basic summary statistics of the available agent characteristics which enter the regression analysis in section A


Gender information is available for all agents in the data base. The share of female agents in the sample is very low at 4% (the total number of visits by female agents is nevertheless sufficiently large to draw conclusions on the role of agents’ gender).

Education level

We have information on education level for only about 12% of the agents in the sample 57 . Since sound reading and writing skills are an essential requirement for data collection, the minimum level of education held by agents in the sample is primary school completed. Table 1 provides an overview of education levels acquired by those agents for whom this information is available. In the multivariate analysis of how agents’ characteristics correlate with child labour identification, we include only the sub-sample of agents for whom education information is available.

Table 1: Education levels amongst agents in the sample (for agents from who this information is available).

Level of education# of agents# of child interviews completed
Lower secondary693,886
Upper secondary583,199

Place of residence

The information on the agent’s place of residence allows us to differentiate monitoring visits by whether the monitoring agent who holds an interview is a member of the same community as the producer interviewed, or whether he is living a different community. Amongst the child interviews where the agent’s place of residence is known (approx. 60% of the interviews), 59% were held by an agent living within the same community. If an agent is living in the farmer’s community, chances for a personal relationship between members of the farming household and the agent existing before the first monitoring visit are obviously higher, even though some communities are too large for all community members to know each other personally. Vice versa, agents and farmers or members of their household may know each other through networks beyond the community, such as through cooperative membership. We nevertheless interpret the indicator of whether the agent lives in the farmer’s community as a proxy for personal acquaintance.

Level of experience

To measure an agent’s experience in conducting monitoring visits, we measure the number of months elapsed between the agent’s first interview in the CLMRS records, and the interview in question. At the time of the first-time household visit, agents had on average approximately 7 months of experience in this role. Around 80% of these interviews were held when the agent had less than one year of experience as a monitoring agent. (Within a sample of follow-up interviews held with children, or a sample of second round monitoring visits, the agents’ average duration of service would obviously be much higher.)

Number of households covered

Even though in general within ICI’s CLMRS, the number of farmers assigned to each monitoring agent ranges between 30 and 35 for agents hired on a part-time basis (which is the case for the large majority of agents), the actual number of households covered by each agent may deviate from this range. This is because some agents devote more time than others to this part-time occupation; because the farmers assigned to some agents are spread out across different communities, and travel time between communities can be significant; or because some agents have better modes of transport available than others, making it easier to cover many households within shorter periods of time. When we compute for each agent the number of cocoa producers where he or she recorded at least one child interview, the average in our sample is 28 households by agent. Approximately 12% of agents have covered less than 10 households (note that the sample also includes interviews by newly hired agents); and approx. 8% of the agents have covered more than 50 households.

Correlations between these characteristics

When we examine how the basic characteristics of monitoring agents in our sample are correlated with each other, we find the following patterns:

  • female agents, and agents with higher than primary education level, are more likely to hold interviews outside their own community
  • female agents cover fewer farming households on average (23) than their male colleagues (30)
  • agents with secondary education cover on average more farmers than agents with primary education
  • agents tend to first visit farmers within their own community, so that on average agents have more months of interview experience when they visit farmers in other communities
  • female agents in the sample have on average a higher education level
  • the average experience agents have at the time they hold an interview is higher for agents with secondary education, suggesting that these remain in service for longer