While the global chocolate industry's revenue reached US$ 186 billion in 2022 and is expected to grow further, about 90% of cocoa farmers in the main cocoa-growing countries in West Africa still do not earn a living income (Manu et al., 2024; Waarts et al., 2024). Some analysts attribute this persistent poverty partly to the concentration of market power in agri-food value chains (IPES-Food, 2017; Clapp, 2019; Oxfam België/Belgique, 2024). In a similar vein, retailers, chocolate manufacturers, and cocoa traders/grinders are increasingly called upon to adopt “good purchasing practices,” that is, long-term buying contracts, direct supply chains, price increases, and shorter payment terms (Fountain, 2023). However, scientific evidence on the influence of good purchasing practices on income remains underdeveloped.

Despite decades of corporate and governmental programs for cocoa farmers in West Africa, their income situation has not improved at scale (Ruf et al., 2020; Waarts et al., 2024). On the contrary, the exhaustion of the forest rent in Ivory Coast and Ghana—i.e., the lack of pristine forest land for slash-and-burn cocoa cultivation—may have increased production costs, with adverse effects on net household income (Odijie, 2016, 2018, 2019; Amanor, 2021). Even under soaring global cocoa commodity prices since 2023, farm-gate prices in Ivory Coast and Ghana are not yet sufficient to enable all farmers to earn a living income (Manu et al., 2024).

Several factors contribute to the persistence of living income gaps among cocoa farmers in West Africa. These include large household sizes and high dependency ratios (Habraken et al., 2022; Van Vliet et al., 2021), small farm sizes (Kuit et al., 2021; Habraken et al., 2022), land tenure insecurity (Asaaga et al., 2020; Gneiting and Arhin, 2023; Bros et al., 2019), and a high dependency of household incomes on cocoa (Impact Institute, 2021). These factors are likely to persist even if higher global cocoa prices were fully translated into exceptionally high farm-gate prices in the long term. Households that are more vulnerable due to small farm sizes, many dependents, or insecure land tenure are therefore unlikely to reach a living income from cocoa alone, even at high prices. Moreover, prices are unlikely to remain high in the long run, given the structural oversupply in the global cocoa market (Oomes et al., 2016; Bensch et al., 2023).

Therefore, income diversification has become a crucial component of strategies to lift cocoa-farming households out of poverty. It may not be in the immediate self-interest of cocoa traders/grinders and chocolate manufacturers to promote income diversification among the farmers who supply them. This reluctance is not only due to the costs involved, but also because diversification may undermine productivity gains associated with household specialization in cocoa farming (Bymolt et al., 2018). Nonetheless, traders and manufacturers are increasingly recognizing the need to invest in income diversification as part of their living income and sustainability strategies, particularly for more vulnerable farming households (Waarts and Kiewisch, 2021). However, it remains unclear to what extent, and through which mechanisms, “good purchasing practices” by downstream actors in the cocoa supply chain can effectively contribute to generating and expanding alternative sources of income for more vulnerable cocoa-farming households.

In this context, the aim of this paper is to analyze the effects of purchasing practices on income diversification. We investigate the mechanisms through which principles of good purchasing practices influence the income diversification by particularly vulnerable cocoa farmers. To this end, we conducted a single-case study of a cocoa cooperative in Ivory Coast that supplies European and US chocolate manufacturers. We use process-tracing methodology to explore the multiple hypothesized causal mechanisms that explain how purchasing practices shape income diversification.

The article is organized as follows. Section 2 provides background on living income and purchasing practices and presents our hypothesized mechanisms; section 3 describes our methodology, followed by context and case study description in Section 4. Our results are elaborated in Section 5 and discussed in Section 6.

We adopted an in-depth qualitative case study research design (Yin, 2009) using process-tracing methodology (Beach and Pedersen, 2019), because this approach allows analysts to assess a set of hypotheses about causal mechanisms of interest (ibid.). Process-tracing is based on a mechanismic understanding of causal effects and is therefore interested in the processes whereby X causes Y (Beach and Pedersen, 2019). This contrasts with a regularity approach to effects, which defines causality as the statistically regular association between X and Y controlled for other relevant possible causes, thereby largely black-boxing the underlying causal processes (ibid.). The purpose of process tracing is not to measure impact, but rather to increase confidence about whether, how and why an intervention had an impact (Befani and Mayne, 2014). It serves to identify likely pathways of change of interventions, especially in complex settings where multiple mechanisms could potentially influence the outcome (Aston, 2019). In the case of good purchasing practices, these interventions comprise several measures that could influence income diversification through various mechanisms (see Section 2.3). Process tracing is therefore well suited to identifying and empirically assessing these mechanisms in detail. The single case study generates knowledge claims that hold true for the specific case under study; more comparative or large-N research designs would be needed to generate representative or externally valid knowledge claims.

We selected the case of a supply chain between a medium-sized European chocolate manufacturer and a cocoa cooperative in Ivory Coast because their business partnership explicitly aims to promote income diversification among cooperative members through good purchasing practices. This setting facilitates the identification and empirical testing of the hypothesized mechanisms. We focused on one cooperative with which the manufacturer had collaborated for seven years, as the effects of good purchasing practices may require time to materialize. Among the cooperative's members, we selected farmers who (a) had supplied cocoa to the manufacturer for at least three harvest seasons (to ensure reliable accounts of experiences while still allowing comparison with earlier buyers and thus reducing recall bias) and (b) cultivated relatively small landholdings, in order to prioritize economically more vulnerable farmers who face larger living income gaps. The research included a consultation phase with cooperative representatives, field technicians, and the manufacturer to validate the research questions and align the design with local circumstances at both ends of the supply chain. Because many steps in the hypothesized mechanisms unfold within households and are not routinely documented (e.g., how premia are allocated, how planning security shapes investment decisions, or how training is applied), we relied primarily on semi-structured interviews (Helfferich, 2011) as our main source of causal-process observations on these mechanisms. We complemented these accounts with document analysis (e.g., the manufacturer's sustainability reports, the cooperative's project documentation and relevant gray and scientific literature) and triangulated farmer narratives with cooperative staff accounts to increase confidence in the inferred sequences. For each mechanism linking purchasing practices and income diversification (Section 2.3), we specified expected observable implications and used targeted interview questions to elicit temporally sequenced accounts of decision points and intermediate outcomes. Moreover, we asked the interviewees counterfactual questions and prompted them to consider alternative explanations (see the interview guides for the cooperative representatives and for the farmers in the Supplementary material). We also report negative evidence where predicted implications were absent.

Data collection at the cooperative took place between 9 October and 4 November 2023. Interviews were conducted in French or in local languages, depending on interviewees' language skills. In line with process-tracing methodology, our design prioritized high information power to achieve saturation: given our narrow aim, specific sample, and theoretically anchored hypotheses, a relatively small N can be sufficient (Malterud et al., 2016). Empirical work suggests that thematic/code saturation often occurs around 9–12 interviews, while meaning saturation is typically reached at around 16–24 interviews in focused designs (Guest et al., 2006; Hennink et al., 2017). For each hypothesized mechanism, we considered saturation reached when additional interviews no longer added information or changed our confidence in that mechanism. In total, we interviewed 25 cocoa farmers as well as the cooperative's board director, managing director, and sustainability manager. Wherever possible, farmers were interviewed outside the presence of family members to reduce social desirability bias. These interviews were complemented by informal exchanges with representatives of the chocolate manufacturer and a cocoa trader who provides trade support in the supply chain.

The interviews were transcribed manually and analyzed using qualitative content analysis with MaxQDA 2024 software (Mayring, 2000). We initially developed a coding frame based on the main categories derived from our hypotheses. We iteratively refined the coding frame inductively by regrouping and splitting up categories to arrive at a sufficiently granular yet uncluttered frame that appropriately captured both evidence related to the hypothesized mechanisms and additional contextual factors (Kuckartz and Rädiker, 2019). This allowed us to systematically compare the empirical material with the expected “fingerprints” of the GPP-based mechanisms, while also remaining open to indications of alternative or complementary mechanisms. Where the empirical material pointed to non-GPP influences on specific observations (for example diversification projects initiated by NGOs or public programs), we considered these as alternative mechanisms and assessed whether they provided a better fit to the evidence than the GPP-based mechanisms.

Ivory Coast is the world's largest producer of cocoa, accounting for over 40% of global cocoa exports and supporting nearly 8 million people (FAO and BASIC (Bureau d'analyse sociétale pour une information citoyenne), 2020). However, only 13% of Ivorian cocoa farmers earn a living income, with an average living income gap of 43% (Impact Institute, 2025). Farmers cultivate an average (mean) of 5.3 hectares of land, but many own less than 3 hectares, which makes it difficult to achieve financial sustainability (Bymolt et al., 2018). Gender disparities further shape the sector, as women have more limited access to land, credit, and farming inputs than men. The social structure within cocoa-growing regions includes indigenous people, internal migrants, and foreign migrants, particularly from Burkina Faso (Ruf et al., 2020). Land tenure is influenced by historical migration patterns and the tutorat system, which facilitated land access for migrants in exchange for labor obligations.

The cocoa sector underwent liberalization in the 1990s but was re-regulated in 2011 with the establishment of the Conseil du Café-Cacao (CCC), a governmental body responsible for regulating, stabilizing, and developing Ivory Coast's cocoa sector. The CCC sets a minimum farm-gate price through a forward auction system, ensuring price stability within a season but exposing farmers to inter-seasonal volatility. The introduction of the Living Income Differential (LID) in 2019 by the Ghanaian and Ivorian governments aimed to improve farmer incomes, although external pressures from cocoa buyers and fluctuating global prices pose challenges (Boysen et al., 2023; Adams and Carodenuto, 2023). Despite efforts to organize farmers into cooperatives, only 21–37% are members of such groups, and many cooperatives are affected by financial mismanagement (Bymolt et al., 2018; Ruf et al., 2021). The sector is dominated by multinational traders, with six major companies controlling 72% of cocoa exports in 2022 (Trase Earth, 2024). Although the CCC fixes farm-gate prices, cocoa buyers are allowed to pay certification or other price premia to producers on top.

This study focuses on the purchasing practices of a European chocolate manufacturer and one of its partner cooperatives in Ivory Coast. The set of interviewed member farmers of the cooperative consists of 17 men and 8 women (32%). In line with our focus on economically more vulnerable farmers, women are overrepresented compared with their share in the cooperative's total membership (18% women). Furthermore, the interviewees' total farm size (median 2.8 ha) is below the national average (median 4.1 ha) (Bymolt et al., 2018). The interviewees' median household size is eight members, of whom six contribute to the household's income. It is therefore reasonable to assume that these interviewees depend on non-cocoa income sources to reach a living income for their families, even under favorable cocoa-buying conditions. At the time of fieldwork, interviewees' ages ranged from 24 to 77 years (median 41), and they had been working as cocoa growers for between 4 and 49 years (median 15). The interviewees also displayed considerable experience with other buyers. Before joining the cooperative, 11 of them had sold exclusively to pisteurs (informal local cocoa traders), 10 reported having sold exclusively to one or several other cooperatives (sequentially), and four had sold to both a cooperative and pisteurs.

Cocoa dependency is relatively high among our farmer interviewees: 10 farmers stated that all their income came from cocoa, five derived 70–90% of their revenue from cocoa, seven derived 50–60%, and only two reported 30–40%. The most common non-cocoa income-generating activities are the cultivation of rubber and staple crops (mostly rice, yam, maize, and cassava), animal husbandry (mostly chickens), and small-scale trade. On average, 44% of the cocoa delivered to the cooperative by the interviewed farmers was sold to the case-study manufacturer and yielded the corresponding premia (compared with a cooperative average of 47%). The interviewed farmers live in seven out of the cooperative's 24 geographic and organizational sections, and the average road distance between their land plots and the cooperative's headquarters in the nearest city is 19 km.

The results show that good purchasing practices influence income diversification in partly unexpected ways at both the farmer level (Section 5.1) and the cooperative level (Section 5.2).

Our results highlight the nuanced ways in which good purchasing practices can both enable and constrain farmers' strategies for achieving a living income through income diversification, drawing on a single case-study design. In our case, premiums provide useful liquidity but often primarily benefit better-off farmers who already have the capacity to invest in new income-generating activities. Limited communication about multi-year purchase arrangements makes it difficult for farmers to leverage buyer–cooperative relationships when seeking credit. At the cooperative level, good purchasing practices provide some planning security and enable targeted investments in services such as village savings and loan associations and training programmes. However, these benefits are frequently undermined by weak market linkages for non-cocoa products and by broader structural constraints, including inadequate infrastructure and land tenure insecurity.

Several limitations need to be considered when interpreting these findings. First, process tracing does not confirm or reject hypothesized mechanisms with complete certainty; instead, it increases confidence in their presence or absence by closely examining processes within a single case (Beach and Pedersen, 2019). Much of our evidence consists of semi-structured interviews with farmers and cooperative staff. Such accounts can provide highly probative evidence when they yield specific, temporally ordered “fingerprints” that are unlikely under alternative mechanisms, but they can also be affected by imperfect recall. We therefore treat different pieces of evidence as having different inferential weight—ranging from weaker “straw-in-the-wind” indications to stronger necessary-condition (“hoop”) tests (ibid.). A lack of secondary data on the hypothesised causal mechanisms in the selected contexts constrained our ability to triangulate self-reported interview data. Second, a case-study design inherently limits external validity, as the inferences are made within one specific context. Finally, the set of interviewees is not intended to be statistically representative, including in terms of social difference dimensions such as gender. Rather, the results provide in-depth qualitative evidence that reached saturation in relation to the hypothesized causal mechanisms linking purchasing practices and income diversification.

These findings can inform future research that uses comparative designs across multiple cooperatives, regions, and buyer models to examine why specific good purchasing practices resonate more strongly with certain farmer segments. Rigorous quantitative methods—potentially including randomised or quasi-experimental designs—could complement our comprehensive process-tracing approach and strengthen selected insights. Further research could also investigate how farmers' demographic factors (e.g., age, gender, or education level) shape the uptake and success of non-cocoa income-generating activities, thereby informing more inclusive strategies for achieving living incomes. In addition, analyses of the interaction between good purchasing practices and complementary interventions—such as land tenure reforms, village savings and loan associations, and government subsidies—would clarify the broader institutional frameworks required for sustainable rural transformation.

Overall, the evidence suggests that good purchasing practices alone cannot deliver living incomes for the most vulnerable cocoa-farming households. Instead, such practices function as one lever among many in a broader “smart mix” of solutions that must involve supply management to avert price collapses, government interventions for rural development, and social support targeting the poorest households. Rather than displacing cocoa as a principal source of income, good purchasing practices can facilitate a more flexible livelihood portfolio—especially when premiums and technical assistance explicitly support farmers' non-cocoa ventures. The study thus underscores both the value and the limitations of market-based interventions. Deeper stakeholder collaboration, transparent communication channels, and strong national policies remain pivotal if good purchasing practices are to tangibly increase the incomes of cocoa-producing communities in Ivory Coast.