Tony's Open Chain Cocoa Farmer Panel Dataset

Background During the years 2019-2023, Tony’s Chocolonely commissioned five annual household surveys among farmers in all of their partner cooperatives in Ghana and Côte d’Ivoire. In both countries, a research company was tasked with conducting the face-to-face interviews. The main aim of these surveys was to assess the multidimensional poverty rates, using the global Multidimensional Poverty Index (MPI). Besides the MPI-relevant items on health, education and standard of living, questionnaires also included a range of other questions. These covered various topics such as the use of external labour, livestock ownership or the ratio of cocoa Vs. non-cocoa income. The surveys were set up as a panel. In as much as possible, enumerators returned to the same respondents each year. Out of the initial 2019 sample of farmers, 45% participated in all five rounds of the panel. Sampling Simple random samples were drawn in each coop, using cooperative membership lists as sample frames. As new cooperatives in Côte d’Ivoire joined Tony’s Open Chain in 2021, 2022 and 2023, those were added to the panel. During the first two survey rounds, sample sizes were determined in proportion to the total number of farmers in the coop: in each coop, a sample of about 10% of total coop size was drawn. As Tony’s Chocolonely data needs changed in 2021, it was decided to switch to non-proportionate sampling: during the last three rounds of the panel, about 150 farmers per coop were interviewed, irrespective of coop size. For these rounds, a weight variable per country is included in the dataset for country-level analyses. MPI and HFIAS scales The provided dataset includes the ten MPI deprivation indicators, the sum of weighted MPI deprivations, as well as a binary variable to distinguish MPI-poor from MPI-non-poor households. Furthermore, the dataset includes a variable to categorize households as ‘food secure’, ‘mildly food insecure’, ‘moderately food insecure’ or ‘severely food insecure’, based on the nine items of the Household Food Insecurity Access Scale (HFIAS). Data anonymization All data have been anonymized by removing personally identifiable information, such as names, phone numbers, GPS data, detailed household composition and responses to open questions. Each respondent has been assigned a unique ID to be identifiable across rounds (famer_hash). To further increase confidentiality, coop names have been replaced by pseudonyms. Across time consistency of farmer codes Users of the dataset should note that some farmer codes may have different demographic information associated with them across different years. In most cases, this will mean that a different household member has been interviewed by the enumerator. In some cases it may be that the farmer code has been reassigned to a different household (likely due to farmers leaving the cooperative). This should not affect the representativeness of the sample per year but will introduce some noise when running time series analyses. Using the data Tony’s Open Chain encourage open use of this data and are excited to learn from how others use and analyse the data – if you carry out an analysis of this data, please do share it with us by contacting [e-mail].