The increasing incidence of pests and insects' infestation in the tropical and sub-tropical region are due to the impact of climate change and poor adoption of modern farm practices, which is critical to the rising food insecurity globally, and particularly higher in many African countries (Aniah et al., 2021; Birch et al., 2011; Cai et al., 2016; Pretty and Pervez Bharucha, 2015). Pest and insect infestation can cause severe on-farm and post-harvest loss if not properly managed (Birch et al., 2011; Pretty and Pervez Bharucha, 2015). While available data show that pesticide usage rate appears comparatively lower in Africa than in other regions (FAO, 2024; Yami et al., 2025), the notion that the risks and impacts must also be correspondingly lower might ignore the effect of the toxicity of the pesticides used and poor pesticide handling practices (Williamson et al., 2008; Sharifzadeh et al., 2018; Kong et al., 2025), which have been largely reported in many studies from different African countries (Okoffo et al., 2016 in Ghana; Rahman and Chima, 2018; Yami et al., 2025 in Nigeria) and other developing countries (Zhang et al., 2011 in China; Mubushar et al., 2019 in Pakistan; Schreinemachers et al., 2016; Akter et al., 2018 in Bangladesh). Nevertheless, pesticides use is on the increase in many sub-Saharan African countries, particularly among smallholder farmers (Andersson and Isgren, 2021). Likewise, there is a growing informal market for pesticides and proliferation of unauthorized pesticides (Yami et al., 2025; Williamson et al., 2008; Staudacher et al., 2020). Despite that increasing pesticide use is associated with a wide range of risks to human health and the environment, most countries in the Global South do not have the capacity for residue testing (Dinham, 2003).

The industrial agrifood system appears to be ecologically and economically unsustainable in the quest to increasing productivity (Sharifzadeh et al., 2018; Zhang, 2024) due to the intensity of synthetic chemical inputs including herbicides and insecticides which cause loss of biodiversity, pollution and risk to human health (Mahmoud and Shively, 2004; Williamson et al., 2008; Staudacher et al., 2020). To address these issues, various alternative practices and models have emerged including organic production system, climate smart agricultural practices, and IPM to achieve better economic, social, and ecological sustainability (Zhang, 2024). IPM is gaining traction across diverse farming systems and agro-climatic zones, it offers safer approach to pest management (Sharifzadeh et al., 2018). IPM integrate multiple pest control methods to prevent pest infestations and ensure quality crop produce—as pests risk-management approach that incorporate several methods to address socioeconomic, health, and environmental threat posed by pests and pesticides use while maintaining acceptable productivity levels (Zhou et al., 2024).

Despite the benefit of pesticide use in crop production, the levels of farmers awareness pesticides effects on human health and the environment varied significantly across countries and regions (Abdollahzadeh et al., 2015). Although, the intensity of pesticides use across most of the African countries is relatively low (FAO, 2024; Yami et al., 2025) and the conventional wisdom that farmers use low agrochemicals in many African countries (Gianessi and Williams, 2011; Oerke and Dehne, 2004; Williamson et al., 2008; Zhang et al., 2011) holds with 16% of the farmers using chemical inputs (Sheahan and Barrett, 2017). While farmers are the primary users of pesticides, the level of awareness about the risk posed by pesticides use influences the methods of pest management (Abdollahzadeh et al., 2015; Akter et al., 2018; Okoffo et al., 2016; Mahmoud and Shively, 2004). As the positive and negative effect of pesticides use are the main drivers of other methods of pest management such as biological, cultural control organic pesticides and IPM (Abdollahzadeh et al., 2015; Kong et al., 2025). While numerous studies have considered the use of chemical inputs and pesticides among farmers across many developing countries, especially in Africa, little is known of the variation across different cropping systems and crop categories based on farmers specialization.

Additionally, farmers' knowledge of both the positive and negative effect of pesticides use may be influenced by various socio-economic, farm, and institutional factors which ultimately can affects farmers choice of pest control strategies and farmer's attitudes toward pesticide use (Aniah et al., 2021; Bagheri et al., 2021; Bandanaa et al., 2024; Yami et al., 2025; Lelamo et al., 2023; Mehmood et al., 2021; Mengistie et al., 2017; Rahman, 2021; Timprasert et al., 2014). Farmers' awareness is shaped by socioeconomic characteristics, such as formal education and level of technical knowledge regarding pesticide use (Bandanaa et al., 2024; Lelamo et al., 2023; Moumenihelali and Amooghli-Tabari, 2025). Similarly, farmers choice regarding other pest control methods such as organic pesticides and IPM are subjective and may varied based on individual farmers characteristics and knowledge (Abdollahzadeh et al., 2015; Rahman, 2021; Timprasert et al., 2014). While we consider the use of chemical pesticides as the use of either herbicides, fungicides, insecticides or nematicides, the use of organic pesticides includes the use bio-pesticides, ashes and neem waters among others. In contrast, we consider IPM adoption as those farmers who combines cultural e.g., crop rotation, early planting, clean seed, scarecrow, hand picking, chemical e.g., inorganic pesticides, insecticides, and weed management and biological controls methods such as bio pesticides, organic pesticides, resistant varieties among others (Angon et al., 2023; Zhou et al., 2024). However, we have no prior knowledge of any study that have considered the heterogeneity of various factors influencing the use of pesticides and IPM adoption among farmers specialized in a more market-oriented crop and staple crops.

Farmers' pesticide use, or organic fertilizer and IPM adoption choices, can be shaped by access to training, farmers' experience, institutional factors, and other specific socioeconomic characteristics (Sharifzadeh et al., 2018; Adeola et al., 2025). For instance, those trained in pesticide use or introduced to IPM may likely be aware of health risks associated with pesticide use or natural pest enemies and ultimately adopt pest control methods that are safe and have less impact on the environment and health (Damalas and Koutroubas, 2017; Jordan and Guerzoni, 2020). While farmers with limited farm and off-farm income or farther away from the pesticide shop may prioritize cost and accessibility factors (Wilson and Tisdell, 2001; Drall and Mandal, 2024). Likewise, access to formal education and income can shape farmers' preferences for pesticides and IPM. As educated farmers may prioritize health and environmental benefit (Angon et al., 2023; Zhou et al., 2024), while those with higher income, farm size, and experienced farmers may consider productivity and performance (Lelamo et al., 2023). Beliefs about pesticide hazards and the effectiveness of alternatives also influenced the weight given to environmental vs. performance criteria (Ahmad et al., 2024). Williamson et al. (2008) investigate pesticide use and handling practices among smallholder farmers in Benin, Ethiopia, Ghana, and Senegal with focus on cotton, vegetables, pineapple, cowpea, and mixed cereals and legumes, for export and local markets. They found that the level of crop susceptibility to insect attack, increased pest incidence, lack of access to information on alternative methods, access to credit, input support, and poor attention to the economics of pest control affect the use of pesticides. In Uganda, Andersson and Isgren (2021) found that most farmers rely heavily on pesticides, despite their health and environmental risks, which are unevenly distributed, especially across gender, and that the burden of pesticide risk falls disproportionately on already vulnerable smallholders. They also noted that structural barriers such as weak regulation, market liberalization, poor extension services, and lack of alternatives limit farmers' capacity to adopt sustainable practices like IPM. Thus, the adoption of pesticides and IPM can be influenced by a complex interplay of socioeconomic, demographic, institutional, agroecological, and economic factors (Adeola et al., 2025).

In Liberia, pre- and post-harvest crop losses due to pests and diseases are about 40 to 50% (MOARL, 2008), and most farmers employ physical and mechanical methods for pest control, but some continue to rely on chemical pesticides. However, growing demand for organic produce have intensified interest in non-chemical alternatives. Although, there has been increasing effort of the government to promote pesticide use and IPM adoption, and regulate pesticide market (Ministry of Agriculture Government Republic of Liberia, 2018; MoARL, 2019), but little is known regarding the different crop market or staple-oriented farmers behavior to the use of pesticides and IPM practices given their individual, farm and institution characteristics. Against this backdrop, this study examines the factors influencing pesticide use and IMP adoption for commodities marketed under formal contract arrangements and in informal market settings i.e., commodities with long and short value chain structures in Liberia.

The study was carried out in Liberia, a country with a population of about 5.5 million people in the West African region, which sits on a land area of 111,370 Km2 and lies on latitude and longitude of 6° 43′ N and 9° 43′ W. It borders Guinea to the north, Sierra Leone, to the northwest, Cote d'Ivoire to the northeast and east, and the Atlantic Ocean to the south and southwest. Its terrain includes plateau, low mountains, and flat to rolling coastal plains characterized by lagoons, mangrove swamps, and river-deposited sandbars (FAO, 2005). The land area comprises 19,230 km² of agricultural, 75,560 km² of forest land, land 15,000 km² of inland waters. As noted earlier, most of the people in Liberia like other sub-Saharan African countries, rely on agriculture as the main source of livelihood. The map of the study area is presented in Figure 1.

Using a systematic three-stage sampling procedure, we rely on unique cross-section dataset from 947 farming households selected fromthree counties, namely Lofa, Bong, and Nimba, based on their potential production to ensure the inclusion of the most relevant areas in the first stage (Lescuyer et al., 2024; Schroth et al., 2015). These counties represent about 51% of the 338,630 agricultural households in Liberia according to the 2022/23 Liberia agriculture census (Liberia Institute of Statistics and Geo-Information Services, 2024). The selection of these counties is also informed by the presence of other value chain actors such as processors, off-takers, and agro-dealers, which are more prevalent in these three counties. Experts from the Central Agricultural Research Institute CARI also consulted for the selection of other value chain actors for the three commodity value chains. We obtained a list of enumeration areas from the Liberia Institute of Statistics and Geo-Information Services. Then, we selected 10 enumeration areas with more growers in the second stage from each county while ensuring the sample was representative of regions with significant agricultural activity. Finally, within these selected enumeration areas, we randomly sampled 947 farming households comprising 336, 320, and 291 cocoa, coffee, and cassava farm households, respectively, selected for the study.

Specifically, the coffee growers were randomly selected from 10 enumeration areas EAs in Nimba and Lofa counties, with the selection made proportionate to the size of each area as the top two regions known for coffee production in Liberia (Lescuyer et al., 2024; Schroth et al., 2015). In contrast, the cocoa growers were selected from 10, 5, and 2 EAs in Bong, Lofa, and Nimba counties, respectively. For the cassava growers, the selection was made from 4 EAs in Bong and Nimba counties and 6 EAs in Lofa county. The household survey was supplemented by additional data from other actors involved in the agrifood supply chain to gather insights from both the markets and institutional perspectives that influence the use of pesticides and what quality aspects of pesticides are required in the marketplace. The actors involved were input suppliers n = 14, buyers/traders n = 36, processors n = 11, and consumers n = 11; selected using snowball sampling methods. Interviews with these stakeholders examined the ways in which pesticide use and pesticide residues are evaluated in their operations, the quality characteristics that are valued and transactionally important, and whether premium prices are paid to farmers who meet sustainability criteria for compliance with certain standards. Incorporating these perspectives allows us to triangulate farmer-reported practices with downstream quality assessment and procurement processes, and to situate pesticide use and IPM adoption within a broader agrifood system and sustainability context.

We deployed a structured questionnaire using open-data knowledge source ODK administered by trained enumerators, with experienced supervisors from the Central Agricultural Research Institute CARI experts. We conducted a two-day intensive in-house training between May 22 and 25, 2024, and the piloting on May 27. The actual data collection took place between May 27 and June 12, 2024. The survey gathered detailed information on farmers‘ and their households' characteristics, agriculture production and cropping pattern, pesticide use and IPM adoption, institutional factors like access to extension, insurance, and market, among others.

Adoption of improved technology or practices is often measured using mutually exclusive binary responses (Bandanaa et al., 2024; Lelamo et al., 2023; Rahman, 2021; Schreinemachers et al., 2016; Timprasert et al., 2014). We follow the underlying assumption of the random utility theory that if the utility gained by those who adopt the technology is higher than those who do not, farmers will adopt it (Fischer and Qaim, 2012; Rahman, 2021) as farmers are often faced with the choice to adopt or not adopt it. In the dichotomous choice model, logit and probit have been commonly used in many studies to examine the factors influencing farmers' decisions to adopt improved technology or practice (Rahman, 2021; Timprasert et al., 2014). We therefore expressed the logit model for our study as follows:

Where P_i is the probability of using pesticide or adopting IPM, X_i is a vector of household, farm, and institutional explanatory variables, α and β_k are estimated parameters. Similarly, the probability of not adopting IPM or using pesticides is given as:

By combining Equations 1, 2, we get Pi1-Pi=eZi and by taking the natural log of both sides, we are left with:

We also explore the effect of the interaction effects of some key institutional variables on the use of pesticides and IPM as stated in Equation 4.

Where β₀, and γ₀ are the intercept, β₁, β_k, γ₁, and γ_k are the correlates of the covariates included in the models which are described in Table 1, ψ is the correlates of the composite function of the interactive terms of some institutional factors A, and ε and u are the error terms. Notably, we can obtain the odds ratio by exponentiating the β_k's in Equation 3, 4 to capture how a change in our covariates changes the odds of adopting IPM or using pesticides. Following (Tambo and Liverpool-Tasie 2024), IPM adoption in this study is defined based on the combined use of practices across multiple IPM components rather than reliance on any single method. See Table 2 for the different components of IPM considered in this study. We estimate our model using the maximum likelihood method. The choice of our covariates includes age, gender and education of the household head, access to credit and extension services, premium price, membership of association, distance to market, pesticides dealer, number of trusted buyers, farm size, fertility, and land right. These are guided by similar previous studies (Bandanaa et al., 2024; Lelamo et al., 2023; Rahman, 2021; Schreinemachers et al., 2016; Timprasert et al., 2014).

However, three different channels for endogeneity issues may affect the fitness of our model. First, the distance to input dealers and the number of traders are often measured with substantial error in household surveys, which can attenuate coefficient estimates and bias standard errors. This is widely acknowledged in studies modeling adoption of chemical inputs (Zhang et al., 2019; Kadjo et al., 2020). Second, omitted variables such as farmers' risk preferences or environmental awareness can confound the estimated impact of factors like credit access or education on pesticide use (Riwthong et al., 2015; Kadjo et al., 2020; Drall and Mandal, 2024), and unobserved household wealth or managerial ability may bias estimates of credit access or education effects on pesticide use and IPM adoption, while both influences and is influenced by market participation or credit decisions, requiring careful temporal ordering to avoid reverse causality (Wilson and Tisdell, 2001; Kadjo et al., 2020). Several estimation strategies like instrumental variables, control function, and fixed effect have been used by several scholars (Ojo and Baiyegunhi, 2020; Kadjo et al., 2020) but due to the cross-sectional nature of our data and lack of appropriate and valid instruments, we do not account for reverse causality and endogeneity concerns that may bias our estimates. This is the caveat of our model estimate, and we do not assert causality.

The summary statistics of the socioeconomic factors of the sampled farmers included in our model are reported in Table 1. The average age of the farmers cultivating market-oriented crops primarily cocoa and coffee is approximately 48 years, and is slightly higher than the 45 years recorded for those cultivating predominantly consumption-oriented crops, such as cassava. This age gap may reflect a historical divide for cash crops, where older farmers are more likely to engage in perennial crops due to land tenure, traditional knowledge, less risky and less labor-intensive as previously observed in African contexts (Wheeler and Marning, 2019; Waldman and Richardson, 2018; Kassie et al., 2009, 2015). Similarly, gender differences are notable. More than 80% cocoa and coffee farmers are male, compared to 69.4% among cassava producers, suggesting a relatively higher participation of female-headed households in cassava production. This aligns with earlier findings that women in sub-Saharan Africa are more involved in food crop production, which is less capital-intensive and more oriented to household consumption (Baada et al., 2023; Palacios-Lopez et al., 2017).

Additionally, formal education is slightly higher among the cassava and cocoa farmers above 50%, than among the coffee farmers 42%. However, across the groups, access to productive resources and services remains limited. For instance, 15% and 14% of cocoa and coffee farmers and 12% of cassava farmers report receiving credit. Moreover, just one in four farmers in either group has access to market price information and reports receiving extension services except for coffee farmers, with only 7% having access to extension services. Traders are the predominant credit providers for cocoa and coffee producers, reflecting their stronger integration into commodity value chains, while cassava farmers more commonly rely on microfinance institutions and cooperatives (Figure 2). Non-governmental organizations NGOs are the leading providers of extension services for both farmer groups (Figure 3), highlighting the limited reach of state-led advisory services.

The household size across all farmer groups is at least 7 members on average, reflecting the relatively large family structure typical of rural agrarian households in sub-Saharan Africa. Landholding size varies considerably by crop orientation. Households primarily engaged in cocoa production operate an average of 7 acres, while those focused on coffee manage slightly smaller holdings of about 6.2 acres. In contrast, cassava-oriented households cultivate an average of only 2.3 acres. This disparity aligns with earlier findings that cash crop producers tend to control larger land areas due to the higher land and capital intensiveness of perennial crops like cocoa and coffee (Roessler et al., 2022; Donkor et al., 2021), as more than half 50% of them report having land right ownership.

Regarding agro-input accessibility, the average distance to pesticide dealer shops is about 17.4 km for cocoa farmers and is relatively higher than 15.6 km and 11 km for cassava and cocoa farmers respectively. This indicates a general challenge in physical access to inputs, which may constrain timely and effective pest management practices, particularly among smallholders in remote areas. Off-farm income is an important livelihood strategy, especially among coffee farmers, 48% of whom report earning additional income from non-agricultural activities, compared to cocoa 36% and cassava 37% farmers.

Over half of the farmers have access to mobile phone in all groups, facilitating communication and information exchange, including market updates and extension messages. On average, farmers in both categories maintain relationships with at least one trusted trader, a crucial component of market linkage and input/output exchange.

Figures 4–7 present the results that show the use of pesticides and IPM for both market and consumption-oriented crops. Despite the increasing advocacy for IPM as a sustainable alternative to chemical pest control, we find that its awareness and adoption remain distinctly low among smallholder farmers across cocoa, coffee, and cassava value chains in Liberia. Specifically, Figure 4 shows that while 24.7% of cocoa and 19.4% of coffee farmers reported that they have been introduced to IPM, only 8.3% of cassava farmers indicated that they have previously been introduced to IPM. We also find that NGOs were the predominant source of IPM introduction, particularly in cassava 83.3%, while government extension services and agro-dealers played negligible roles across all crops.

Figure 5 present the result of pesticide use and IPM adoption among market and consumption-oriented crops. The results shows that 14.3% of cocoa and 11.6% of coffee farmers used chemical pesticides, while very few 5.8% of the cassava farmers report the use of chemical pesticides. On the other hands, over 20% of farmers across all three crops reported the use of organic pesticides such as neem-based solutions and other biopesticides, suggesting a modest but notable awareness of environmentally friendly pest control practices. However, adoption of IPM methods was particularly limited, with only 7.7% of cocoa farmers and 7.9% of cassava farmers indicating IPM use.

Considering the IPM practices employed as presented in Table 2, we find that the most important practices adopted are cultural methods among cocoa and coffee growers as well as cassava growers. While physical and pest monitoring pest monitoring and physical control method are somewhat common >25%, only few of them < 20% applied various biological control i.e., ash, neem water, biopesticides and chemical control methods.

Figure 6 shows the result of the pesticides used by cocoa coffee and cassava growers and we find that 13% of them applied pesticides herbicide, insecticide, fungicide, or synthetic pesticide as the use of herbicides and spraying of insecticides are the most used pesticide categories. For non-adopters of IPM, Figure 7 reveals the reasons for not adopting. We find that the primary reason for not adopting is insufficient knowledge of IPM methods above 85% across all the crop category, followed by high cost approximately between 6 and 10%. The other reasons emphasized include lack of knowledge of IPM and high level of pest infestation.

Table 3 presents the regression estimates of the factors associated with pesticide use among cocoa, coffee, and cassava farmers (Equation 3), while extends the model to include key interaction effects (Equation 4). Results are disaggregated by crop type to capture heterogeneity across farming systems. We find that the model is statistically significant and a good fit for the coffee and cassava farmers group, unlike that of the cocoa grower, which is insignificant as the p > 0.1 with a very low pseudo-R² of 8.4%. For coffee and Cassava farmers, the pseudo-R² is relatively higher indicating that the explanatory variables explain about 36.7% and 16.3% of the variation in the pesticides used respectively.

The results of the correlation coefficient indicate heterogeneity across crops in the influence of socio-economic factors and institutional access. Among cocoa farmers, being male and a member of a farmer group significantly reduces the likelihood of pesticide use p < 0.05 with marginal effect of about 16%, while farm size positively predicts use by 7.7%. We find similar result in the extended model estimates. For coffee farmers, age negatively influences pesticide use, while access to land rights, owning a mobile phone, and larger farm size significantly increase the probability of pesticide use. Non-farm income is associated with reduced pesticide use, suggesting a substitution effect.

In the extended model Equation 4, interaction effects reveal deeper mechanisms. Notably, for coffee and cassava farmers, the combined effect of education and access to extension services significantly increases the likelihood of pesticide use p < 0.01 with marginal effect above 80%, highlighting the importance of knowledge pathways. Conversely, the interaction between market access and cooperative membership reduces pesticide use by 13.8% among cassava farmers, suggesting that collective marketing structures may support more cautious pesticide behavior. Access to extension alone is associated with a strong reduction in pesticide use among both coffee and cassava farmers, reinforcing its central role in promoting sustainable practices. Overall, these findings emphasize the differentiated effects of education, access to services, and institutional factors across crops and underscore the need for crop-specific policy targeting in pesticide risk mitigation.

This section presents the results of logistic regression estimates of the factors influencing the use of organic pesticides—biopesticides such as neem seed or leaf extracts—among cocoa, coffee, and cassava farmers. Table 4 reports the results from Equation 3, which considers direct effects, and the interaction terms to capture moderating effects (Equation 4). In Table 4, land rights and non-farm income consistently exhibit strong and significant positive effects across all three crops. Farmers with secure land tenure are significantly more likely to adopt organic pesticide practices, suggesting that tenure security encourages long-term investment in environmentally sustainable inputs.

Access to agricultural extension services positively and significantly influences organic pesticide adoption among cocoa farmers, but not for coffee or cassava. Interestingly, education has a negative effect on adoption for cocoa and cassava farmers, a counterintuitive result that may suggest educated farmers lean more toward conventional pesticide use, possibly due to their greater access to commercial products or skepticism about traditional methods.

Among cassava farmers, gender significantly influences organic pesticide use, with female farmers less likely to adopt. Additionally, distance to pesticide dealers significantly increases the likelihood of organic pesticide use among cocoa and coffee farmers, implying that remoteness may limit access to synthetic chemicals and encourage alternatives. The interaction models Equation 4 model provide little insights. The positive and significant coefficient of the education-extension interaction for cassava, although not statistically significant, suggests that when educated farmers have access to advisory services, they are more likely to consider organic options, hinting at complementarities between formal knowledge and contextual advice. Similarly, among cocoa farmers, the inclusion of interaction terms slightly improves model fit pseudo-R² = 23.9% vs. 23.2%, affirming the role of institutional interactions in shaping adoption.

Table 5 presents the logistic regression estimates for the determinants of IPM adoption among cocoa and cassava farmers. The results of the marginal effect are derived from Equations 3, 4. Given the very low rate of adoption of IPM among coffee farmers we do not estimate it determinants. The models appear significant for both cocoa and cassava farmers. Among cocoa farmers, access to extension services significantly increases the likelihood of IPM adoption, with a marginal effect of 16.6%. This suggests that farmers who receive extension advice are substantially more likely to adopt IPM practices.

However, when the interaction between education and extension is introduced Equation 4, the positive effect of extension is moderated by the negative and significant interaction term with a marginal effect of 68.3%, indicating that better-educated farmers may rely less on extension services for IPM knowledge or that the current extension messages may not be well-targeted to their needs.

Membership in farmer organizations negatively affects IPM adoption across both groups, particularly among cocoa farmers, where the marginal effect is a significant 10.3% reduction. This is counterintuitive and may reflect internal dynamics within these groups that discourage experimentation with non-conventional methods, or that IPM knowledge is not adequately disseminated through these networks. Farmers with secure land tenure are more likely to adopt IPM practices across both crops, with marginal effects of 10.9–12.6%.

Interestingly, farm size exhibits contrasting effects: larger cocoa farms are associated with a lower probability of IPM adoption, while cassava farmers with larger holdings are slightly more likely to adopt. Moreover, the distance to pesticide dealers negatively correlates with IPM adoption among cocoa farmers, suggesting that farmers far from pesticide outlets may opt for IPM as a more accessible or cost-effective alternative. Education is strongly associated with higher IPM adoption only in Equation 4 for cocoa farmers, where its interaction with extension services is significantly negative.

For cassava farmers, access to extension 8.3–8.5% and land tenure 11.9–12.6% are consistently significant drivers of IPM use. Non-farm income also positively affects adoption about 5%, indicating that households with additional income sources may be more financially capable of adopting improved pest management techniques. Lastly, trust in market intermediaries measured as the number of trusted traders shows divergent effects: it discourages IPM adoption among cocoa farmers but encourages it among cassava farmers. This contrast may stem from different value chain structures or variations in how traders influence input choices across crop systems.

This section synthesizes insights from key value chain actors including input suppliers, processors, and buyers on pesticide use and the adoption of IPM practices in Liberia. The findings reveal several systemic challenges and opportunities that shape pest control strategies along the cocoa, coffee, and cassava value chains. Among input suppliers, the pesticide market is dominated by imported products, with only one-third of respondents reporting trade in locally produced pesticides. Moreover, most suppliers 12 out of 14 source their stock from outside the county, indicating limited local access to agricultural inputs.

However, these efforts are constrained by several structural barriers. Input suppliers identified high input costs, limited sourcing options, poor road infrastructure, and concerns about product quality as the major impediments to their operations. Cassava processors reported sourcing roots either from local farmers or their own plots and typically processing them within one to seven days of harvest. They do not use pesticides, emphasizing sensory attributes such as taste, texture, and aroma in their product standards. This perspective was mirrored by cassava consumers, who noted varying shelf life across different cassava-based products. The limited relevance of pesticide use in cassava processing reflects the low incidence of postharvest pest concerns in root crops.

In contrast, cocoa and coffee buyers exhibit a different orientation. None of the interviewed buyers had organic certification, and only 19.4% reported labeling their products by origin. Quality assessment was primarily conducted through visual inspection or bean cracking. Although most buyers require adherence to postharvest protocols, particularly for fermentation and drying. As only 15.4% explicitly demanded low pesticide residues, and just 7.7% mentioned labor standards such as child labor restrictions. Notably, none required organic certification or pesticide-free production. Furthermore, only 8% of buyers stored cocoa or coffee beans without the application of chemicals for pest control. These findings suggest that while market actors recognize the importance of quality control, there is minimal formal demand for low-pesticide or organic products.

Prior to our estimation of the determinants of pesticides use, we assessed the potential for multicollinearity among explanatory variables by computing pairwise correlation coefficients. These results are reported in Appendix Tables 1a–1c. The analysis shows no evidence of strong multicollinearity, as all correlation coefficients are well below the conventional threshold of 0.5, indicating that the covariates used in the model are sufficiently independent. Additionally, we conducted a series of post-estimation diagnostic tests to ensure the robustness of our findings. First, we assessed multicollinearity among the explanatory variables using the Variance Inflation Factor VIF. As reported in Appendix Table 2, all VIF values are well below the commonly accepted threshold of 5, indicating no significant multicollinearity concerns. Second, we re-estimated our models using a probit specification as an alternative to the logistic regression framework. The probit estimates, presented in Appendix Tables 3–5, are consistent in sign, magnitude, and statistical significance with those obtained from the logistic models for both Equation 3 and Equation 4. This consistency across estimation techniques reinforces the reliability and stability of our main results.