Factors influencing adoption of cocoa agroforestry in Ghana: Analysis based on tree density choice

ABSTRACT

Despite the beneficial effects of cocoa agroforestry on cocoa production and yield levels, adoption is below expectation while the factors driving adoption are not well known. This study therefore explored the typology of cocoa agroforestry (AF) in Ghana and the drivers of adoption using cross-sectional data from cocoa producers in Ghana’s Sefwi Wiawso Municipal. A 5-point Likert scale was used to analyze farmers’ perceptions of cocoa AF while an ordered probit model was applied to assess the factors influencing adoption of cocoa AF. The study revealed that the cocoa landscape in the study area consisted of 7.51% full sun, 31.62% low shade, 30.83% medium shade and 30.04% high shade. Moreover, adoption of cocoa AF was influenced by farmer characteristics (such as sex, farming experience, and household size), farm-specific factors (such as total farm size, number of cocoa farms), and institutional factors (such as involvement in off-farm work, participation in AF training, farmer association membership, and participation in cocoa hand pollination exercise. The most important constraint affecting cocoa AF adoption is poor communication of the benefits of shade trees on cocoa farming. We conclude that farmers are indifferent to the adoption of the AF systems, while the adoption determinants vary across the typology of AF systems. The authors recommend provision of effective training of farmers on cocoa AF to increase adoption. Also, the benefits of cocoa AF should be effectively communicated to farmers via appropriate communications channels such as cooperative groups and farmers’ fora to enhance adoption of the technology.

KEYWORDS:

• cocoa agroforestry
• adoption
• ordered probit model
• Ghana

Introduction

Cocoa (Theobroma cacao) is an economic crop with worldwide importance and grown in humid low land tropics (Scudder et al., 2022). The dominant players in global cocoa cultivation are Ghana and Cote d’Ivoire which contribute about three quarters of the world output (Wainaina et al., 2021). Ghana’s economy has historically been based on cocoa for more than a century and the crop contributes significantly to employment generation and government revenue (Anang, 2016; Nunoo & Owusu, 2017; Wainaina et al., 2021). However, the cocoa sector faces several challenges including deforestation, turning cocoa farms into arable crop production, and mining activities (Boateng et al., 2014; Osman et al., 2022; Snapir et al., 2017). These challenges have dire consequences for cocoa production and therefore calls for measures to safeguard cocoa cultivation using sustainable production practices. One of such sustainable practices that has gained prominence in recent times is cocoa agroforestry.

Agroforestry (AF) is an agricultural practice that involves the interaction of woody perennials, food crops and/or livestock on the same land unit (Nair, 1984). It also involves the practice of integrating forest trees and non-forest perennial trees into farming by preserving existing trees, vigorous planting, and tending natural tree regeneration (Ruf et al., 2004). The key components of AF systems include trees, shrubs, pasture, crops, and livestock as well as the environmental aspects of soil, climate, and landform (Nair et al., 2021).

Cocoa AF is the integration of cocoa with a variety of high-value tree species, as well as other crops (Essouma et al., 2020). Cocoa AF is analogous to systems in which cocoa trees are intercropped with fast-growing shade trees, commercial timber or fruit trees to varied degrees denoted as ‘planted shade’ to give farmers additional income and product. This method of cocoa cultivation is more ecologically friendly than other methods of producing cocoa or undertaking other farming practices in tropical woodland/forest zones.

Cocoa AF is one of the best examples of sustainable agriculture practices that protects the tropical forest ecosystem and promotes higher levels of biodiversity (Ruf et al., 2004). It also prolongs the productivity of cocoa farms due to its ability to enhance functions of the soil ecosystem (Amponsah-Doku et al., 2022), and has the ability to tackle global issues including climate change and land degradation (Kouassi et al., 2021). Planting cocoa alongside diverse food crops, shrubs, and trees provides temporal shade, weed reduction and soil improvement. Shade trees in cocoa AF systems minimize cocoa fruit abortion and decreases the effect of mirids on cocoa yield (Bos et al., 2007). Cocoa AF systems also have a large potential for carbon sequestration (Gockowski & Sonwa, 2008).

The cocoa AF systems in Ghana come in four different types according to Nunoo et al. (2014): no shade/full sun, low shade, medium shade, and heavy shade systems as indicated in Appendix 1. The number of shade trees per hectare, the proportions of canopy cover, and the intensity of management are used to classify these systems. The no shade system, also called full sun, has 0–2 shade trees and the canopy cover is less than 36%. It is distinguished by its extensive use of agrochemicals and labour, making it the most luxurious cocoa AF system to implement. With regards to the low shade system, the number of desirable shade trees ranges from 3–9 trees per hectare, with a canopy cover of 36–65%. Intensive management is a feature of this system but it is not as input-demanding as the no shade system. Low shade AF provides good balance between economic demands and ecological needs (Steffan-Dewenter et al., 2007). According to Nunoo (2015), the medium shade cocoa AF system is the most recommended of all the cocoa AF systems because it produces a good yield over an extended length of time. The number of shade trees per hectare under this approach varies from 10 to 15. The medium shade cocoa agroforestry is shaded by a variety of deliberately planted and naturally grown trees, with a shade cover of around 66–85%. The fourth cocoa AF system in Ghana is the heavy shade system. Under this approach, cocoa fields contain more than 15 desirable trees per hectare providing more than 85% shade cover. Compared to the other cocoa AF systems, this system requires less agricultural inputs such as labour and agrochemicals for managing the farm. The tree density level, however results in a highly humid environment in which some pests and diseases may thrive (Nunoo, 2015). The Cocoa Research Institute of Ghana (CRIG) suggested a tree density of 15–18 desirable trees per hectare but this might not be appropriate for all types of shade trees (Ashley et al., 2020).

Despite the fact that cocoa cultivated in full sun has greater mean harvests and gross margin figures than cocoa grown in the shade, it is ecologically unsustainable in the long run. The tree crops cultivated in shaded systems tend to sustain production over time and are less prone to pest and diseases than monocultures planted in full sun (Belsky & Siebert, 2003). In addition, cocoa cultivation is a primary driver of deforestation, hence it is imperative to adopt cocoa agroforestry as a possible solution (Wainaina et al., 2021).

The current study is motivated by the paucity of studies on the determinants of cocoa agroforestry with a direct focus on tree density choice. Available studies such as Kaba et al. (2020), Kouassi et al. (2021) and Nunoo et al. (2020) assessed the determinants of cocoa agroforestry adoption but they did not focus on the tree density choice. This is against the backdrop that cocoa AF systems are classified based on tree density. This implies that adopters of agroforestry can be classified into distinct adopter categories based on the tree density classification, which is ordered (no shade, low shade, medium shade and high shade). From a methodological point of view, an ordered probit/logit model is more appropriate for analyzing the determinants of cocoa AF adoption when the dependent variable is ordered, although it is possible to simply group the respondents into adopters and non-adopters. Even though Gockowski et al. (2004) estimated the determinants of adoption of cocoa agroforestry using an ordered probit model, they used a different shade classification system from that provided in Nunoo et al. (2014). Methodologically, this study differs from previous studies which measured cocoa AF adoption as a binary variable and instead, leans towards a regression model with an ordered dependent variable for tree density choice which is a more appropriate representation of the data.

The study has as its objectives the assessment of farmers’ perceptions about agroforestry, the factors driving tree density on cocoa farms, and the constraints associated with cocoa agroforestry adoption. The study will test the hypothesis that farmer, farm and institutional factors influence the adoption of cocoa AF. The study will further test the hypothesis that there is no agreement between the respondents in their ranking of the constraints affecting AF adoption. The study adds to the body of literature by examining the role of tree density classification systems in cocoa agroforestry adoption using empirical field data and an ordered probit model which deals with inherent tree density choice.

Literature review

The subject of cocoa agroforestry is topical due to its importance and the extant literature contains several studies on it. These studies vary in scope but shed light on the subject matter. Asare et al. (2014) assessed how cocoa agroforestry could be used to enhance forest connectivity in a fragmented landscape in Ghana. They noted that Ghanaian farmers view forest reserves as land banks for enhancing yield, leading to fragmentation of existing protected forests. The authors developed a multi-disciplinary approach to enhance forest connectivity using cocoa agroforest corridors. They observed that while planting lumber trees within agroforests reduces yield losses, the on-farm benefits are insufficient to justify adoption. Their conclusion was that paying cocoa producers a premium price and providing ecosystem services could encourage adoption.

Kaba et al. (2020) assessed the role that shade trees play in the uptake of cocoa AF systems in Ghana. Their study focused on the semi-deciduous rain forest agroecology. Their study analyzed farmers’ perceptions of shade trees and how this shapes uptake of cocoa AF systems. They noted that 87% of producers planted shade trees at some phases of setting up their farms, compared to 13% who already had trees. The majority of farmers connected the value of shade trees to healthy crop growth and fodder, while the potential for higher yield and lack of awareness of the benefits of shade trees were main reasons for removal of shade trees. They called for partnership between producers and stakeholders to avert the drive for no shade AF system.

Nunoo et al. (2020) assessed the factors that determine adoption of cocoa AF as a climate change mitigation strategy in Ghana. The study categorized the respondents into adopters and non-adopters and employed a logistic regression model to assess the factors affecting AF adoption. The study’s results indicated that farm size, farming experience, group membership, awareness of climate change, and access to agricultural extension influenced cocoa AF adoption.

Using a multi-dimensional meta-analysis, Niether et al. (2020) compared cocoa AFs with monocultures. Their findings indicated that cocoa AF systems performed better in terms of most of the performance indicators such as total system yield, potential for climate change mitigation and adaptation, and biodiversity conservation.

In a study conducted in Côte d’Ivoire, Atangana et al. (2021) investigated how the choice of tree species influence rebuilding tree cover in deforested cocoa landscapes. The study revealed that farmers’ age, tree-planting expertise, and anticipated benefits, such as cash and food, affect the introduction of new tree species through planting or ‘retention’ while clearing land for cocoa cultivation. The main reasons for choosing particular tree species were to provide shade for cocoa, provide valuable tree products (38%), and generate cash from the sale of these items (7%). When clearing land for cocoa production, fruit tree species were most commonly planted, whereas timber tree species were largely spared.

A study in Ghana by Yamoah et al. (2021) revealed that shaded cocoa AF systems give sustained yields in the medium- to long-term compared to unshaded systems, while farmers’ awareness of the implications of an unshaded cocoa AF system could encourage environmentally-friendly practices, such as soil fertility preservation.

Gyau et al. (2015) noted that, because of full-sun farming practices, cocoa yields in Côte d’Ivoire are low and decreasing yearly. Consequently, interest in creating sustainable cocoa agroforests by reintroducing shade trees is growing. In their study, they evaluated the factors influencing the existence and density of the alternative tree species of interest to farmers. Their findings revealed that social network effects, ethnic group, and geographic region were the most important drivers. The likelihood that poorer farmers and those living in remote areas would associate their cocoa with crops like oil palm was also higher. As a result, the authors recommended that agroforestry initiatives should target particular tree species depending on the area, ethnic group, market accessibility, and income level.

While there exist a number of studies on cocoa agroforestry, there is paucity of studies on the drivers of cocoa AF adoption from the perspective of the tree density choice. As shown in the literature, tree density on cocoa farms vary. This is reflected in the recommended measure of tree density, which is the number of trees per hectare. Thus, the classification system clearly identifies ranges for the number of trees per hectare, which results in an ordered categorization of the tree density on farms.

Materials and methods

Description of the study area

The research was conducted in Sefwi Wiawso Municipal in the Western North Region of Ghana. Western North Region is one of the leading cocoa-producing regions in Ghana, with Sefwi Wiawso Municipal as one of the dominant cocoa-producing districts in the Region. The study area is located in a forested area, where farming is the primary source of income. Cocoa cultivation is a dominant economic activity among farm households in the municipality. Food crops like maize, plantain, cassava, and cocoyam, are some of the most cultivated staples. Small-scale mining and trading are also important economic activities in the area. The activity of small-scale miners is perceived to have a negative influence on cocoa production as some farmers give out their cocoa farms to mining firms, resulting in deforestation and loss of biodiversity. The municipality has a population made up of 75,905 males (representing 50.2%) and 75,315 females (representing 49.8%), with a population density of 152.4 km² (Ghana Statistical Service, 2021).

Sampling and data collection

The sample size for the study was determined by applying Green’s (1991) sample size determination formula. According to the formula, if there are m explanatory variables in the model, the minimal sample size needed for this study is 50 + 8(m), or 50 + 8(19). This results in a minimum sample size of 202, which we subsequently increased to 254. The study employed a multi-stage random sampling technique to select the respondents for the study. Sefwi-Wiawso Municipal was purposively selected as the study site because it is one of the leading cocoa-producing districts/municipals in Ghana. In the subsequent stage, 11 communities in the municipal were selected at random. Next, 14 cocoa-producing farm households were selected from each of the 11 communities using simple random sampling to give a total sample of 254 respondents. The communities and cocoa-producing farm households were selected at random (using simple random sampling), to provide a random sample for the econometric analysis. Cocoa farmers in the municipal have been exposed to cocoa agroforestry through extension agents. Also, based on the cocoa AF system (tree density classification), every farmer is assumed to be implementing a particular system, hence the use of random sampling to select the respondents. Due to incomplete information, one respondent was dropped from the analysis.

Primary data was gathered from the respondents using a structured questionnaire. The primary data collected included farmers’ demographic and socioeconomic factors, information on cocoa production, output level, access to services, adoption of cocoa agroforestry and other technologies, challenges farmers face in adopting cocoa agroforestry, among others. Prior to the interviews, the participants’ informed consent was requested. The enumerators communicated the goal of the research to the participants who were given the choice to opt to participate or not. Thus, participation was voluntary and the information obtained was used only for the intended purpose of this research.

Method of data analysis

The first objective of the study assessed the drivers of tree density on cocoa farms. Econometrically, the choice model to analyze the data depends on the nature of the dependent variable, whether continuous, discrete, or ordered. The classification of cocoa agroforestry system in Ghana is categorized into four levels of tree density: full sun, low density, medium density, and high density. Ordered dependent variables are typically modelled using ordered probit or ordered logit regression models. Due to the ordered nature of the dependent variable, that is, tree density, the study employed an ordered probit model (OPM) to investigate the factors driving tree density choice on cocoa farms among the respondents. The study adopted the empirical construction and application of the model as applied by Gockowski et al. (2004). The OPM specification is as follows:

(1) $Y_i^{\ast }=X_i^1\beta +e_i,i=1,2,\dots N$(1)

where $Y_i^{\ast }$ is unobserved and referred to as the latent or index variable. The regressors are denoted by $X_i^1$ and the error term is $e_i$ such that $E\left(e_i/x_i\right)=0,Var\left(e_i/x_i\right)=1$. Considering $Y_i$ as the observed variable with J response choices or categories and serving as proxy for the unobserved variable $Y_i^{\ast }$, and defining $\omega ={\omega }_{-1},{\omega }_0,{\omega }_1\dots {\omega }_{J-1},{\omega }_J$ as parameters of unobservable cutpoint vectors. The observed and unobserved variables are related as follows:

(2) $Y_i=jif{\omega }_{j-1}<Y_i^{\ast }\le {\omega }_j,j=0,1,2,\dots J$(2)

Given that ${\omega }_{-1}=-\mathrm{\infty },\hspace{0.17em}{\omega }_0=0,\hspace{0.17em}{\omega }_J=\mathrm{\infty },\hspace{0.17em}$ and ${\omega }_{-1}<{\omega }_0<{\omega }_1\dots <{\omega }_J$

Thus the probabilities are specified as follows:

$Prob\left(Y_{i=j}\right)=Prob\left[w_{j-1}<Y_i^{\ast }\le {\omega }_j\right]$

$=Prob\left[{\omega }_{j-1}-x_i^1\beta <e_i\le {\omega }_j-x_i^1\beta \right]$

(3) $=\mathrm{\Psi }\left({\omega }_j-x_i^1\beta \right)-\mathrm{\Psi }\left({\omega }_{j-1}-x_i^1\beta \right)$(3)

where $\mathrm{\Psi }\left(.\right)$ represents the standard normal cumulative distribution function, and categorical variable J is ordered alternative response choices, in this case the four cocoa tree density categorizations 0, 1, 2 and 3.

The empirical model used is stated below:

(4) $T{D}_{ij}=\alpha +\beta H_i+\gamma F_i+\delta I_i+{\epsilon }_i$(4)

where H = household/individual characteristics, F = farm-specific factors, I = institutional factors as specified in Table 2 and TD depicts the tree density levels represented by choice of cocoa shade levels (0 = no shade, 1 = low shade, 2 = medium shade, 3 = high shade). The subscripts i and j (j = 0,1,2,3) represent a particular household’s cocoa farm and choice of cocoa shade levels respectively.

Farmers’ perception with regards to cocoa agroforestry was assessed using a five (5) point Likert scale with 1 = strongly agree, 2 = agree, 3 = undecided, 4 = disagree, and 5 = strongly disagree. During the field interview, farmers were asked to respond to specified perception statements captured by the Likert scale. The Likert scale efficiently records responses to qualitative knowledge assertions on an ordinal scale, hence appropriate for assessing the study’s objective (Likert, 1932).

The third and final objective assessed the constraints to cocoa agroforestry adoption and was analyzed using Kendall’s coefficient of concordance, which is a rank-based measurement that establishes agreement amongst raters (Legendre, 2005). It is a metric for measuring how well various raters agree on a collection of n elements they are rating. The Kendall’s coefficient of concordance (W) can be computed as shown in eqn. 5:

(5) $W=\frac{12\left[\sum T^2-\frac{{(\sum T)}^2}{n}\right]}{\left(n^3-n\right)m^2}$(5)

where n is the number of objects being ranked, m is the number of raters (respondents), and T is the sum of factors being ranked. W takes a range of values ranging between 0 and 1, where 0 indicates no agreement and 1 indicates complete agreement.

Results and discussion

Descriptive characteristics of the sampled farmers

Pertaining to the individual/household characteristics of the respondents, it was indicated that majority (60%) of the farmers are men (as shown in Table 1). Cocoa production is traditionally a male-dominated activity in Ghana, which reflects the fact that men are mostly considered to have control over factors of production in the farming households, hence more likely to have resources to manage perennial tree crops such as cocoa. The result agrees with that obtained by Danso-Abbeam et al. (2012) and Kyere (2018) which revealed male dominance in cocoa production in Ghana. Additionally, the respondents had 22 years cocoa farming experience, which is in conformity with the result of Danso-Abbeam et al. (2014) which showed that the farming experience of cocoa farmers in Ghana was 15 years and above. The years of cocoa farming experience shows that majority of the farmers have been involved in cocoa farming a very long time and there have considerable knowledge of cocoa production. This is expected to be influential in their decision to adopt farm technologies. The respondents also had an average of 8 years of formal education, while cocoa production was identified as the main economic activity for about 83% of the respondents. The mean years of education, even though low on paper, is significant for a farming population in a developing country such as Ghana. Eight years of education equates to junior high school education, which is reasonably good for a farming community. The average household size of the respondents was found to be six members. The relatively large family size argues in favor of the idea that family labour is a necessity for smallholder cocoa farmers’ farm operations even though a large family size can impose a limitation due to high financial burden hence farmers might not be able to afford enough farm inputs to boost production (Kyere, 2018).

Table 1. The descriptive characteristics of the respondents

Variable	Description	Mean	Std. Dev.
Individual/household characteristics
Sex	Sex of farmers: female = 1; 0 otherwise	0.399	0.491
Years of education	Number of years of formal education	8.186	4.602
Household size	Number of household members	6.237	3.997
Farming experience	Number of years in cocoa farming	21.99	13.17
Total cost of assets	Value of household assets	1000	1132
Main economic activity	Equals 1 if main activity is cocoa production; 0 otherwise	0.826	0.380
Sex*experience	Interaction of sex and farming experience; Female = 1	9.269	14.45
Farm-specific factors
Total farm size	Total farm size in acres	5.912	4.836
Number of farms	Number of coca farms	2.138	1.228
Own cultivation	Equals 1 if famer cultivated own land; 0 otherwise	0.822	0.383
Institutional factors
Off-farm work	Participation in off-farm work: 1 for participation, 0 otherwise	0.510	0.501
Access to extension	Access to extension services: 1 = yes, 0 otherwise	0.711	0.454
Group membership	Member of a farmer group: 1 = yes, 0 otherwise	0.680	0.467
Agroforestry training	Participation in agroforestry training: 1 = yes, 0 otherwise	0.542	0.499
CMSE	Participation in cocoa mass spraying exercise: 1 = yes, 0 otherwise	0.443	0.498
CHTFERT	Participation in cocoa hi-tech programme: 1 = yes, 0 otherwise	0.822	0.383
CHPE	Participation in cocoa hand pollination exercise: 1 = yes, 0 otherwise	0.221	0.416
CRP	Participation in cocoa rehabilitation programme: 1 = yes, 0 otherwise	0.273	0.446

In terms of the farm-specific factors, the respondents had on average 6 acres of farmland allocated to cocoa production, compared to an average farm size of 7.74 acres as reported by Danso-Abbeam and Baiyegunhi (2017) in their study that covered selected cocoa-producing regions in Ghana. The result shows that the farmers are operating relatively small farm holdings. Land fragmentation due to increasing population and family sizes is a factor that affects the size of cocoa farms, which may influence cocoa AF adoption decisions. The data also indicates that 82% of the farmers managed their own farms, as against those who were settler farmers or rented their farm lands. Averagely, each farmer operated two cocoa farms (or plots).

With regards to institutional factors, it was also observed that about 51% of those surveyed worked outside the farm as additional income source. Thus, apart from cocoa production, cocoa farmers also engage in other income generating activities to supplement the household income. Furthermore, 71% of the respondents received agriculture extension visits from extension workers during the crop year. The data further revealed that 68% of those surveyed were members of a farmer group, which is anticipated to encourage the uptake of cocoa agroforestry (Kassie et al., 2009). With regards to training in agroforestry, 54% participated. Also, 44.3% and 22% participated in cocoa mass spraying exercise and cocoa hand pollination exercise, respectively. In addition, 82% and 27% participated in cocoa hi-tech programme and cocoa rehabilitation programme, respectively. The results indicate that there is generally low participation in the Government of Ghana flagship cocoa programmes. For instance, cocoa mass spraying exercise is a state-sponsored programme to control cocoa pests and diseases, yet several farmers are unable to participate. There have been instances where the spraying gangs could not reach all farms due to logistical challenges. Also, not all farmers subscribe to the cocoa rehabilitation programme which involves rehabilitation of old cocoa farms through replanting, among others. Also, the cocoa hand pollination exercise is not well-known among farmers and this has resulted in low adoption.

Classification of tree density on cocoa farms in Sefwi Wiawso Municipal

The data reveals that a small percentage of the respondents (7.5%) practice no shade or full sun agroforestry system (see Table 2), suggesting that no shade cocoa agroforestry is not a preferred choice in Ghana. This suggests that Ghanaian cocoa farms are typically associated with some level of tree cover. The data further shows that the proportion of farmers having low shade, medium shade, and heavy shade are similar. It is likely that farmers may be indifferent between adopting low, medium or high shade agroforestry. Land clearing for cocoa production usually involves cutting down trees on the farm which may not be replanted and therefore can reduce the tree density. Farmers may also decide to vigorously plant trees to increase tree density on their farms. Nunoo (2015) posited that medium shade cocoa agroforestry system is the most recommended of all the cocoa AF systems in Ghana because it produces a good yield over an extended length of time. However, the result of the study does not show a higher preference among farmers for the medium shade AF system, which should be of concern to cocoa agroforestry practitioners and policy makers in Ghana. Possible reasons for the findings may include the following. Low shade cocoa AF may appeal to farmers because it is not as input-demanding as the no shade system and provides good balance between economic demands and ecological needs (Steffan-Dewenter et al., 2007). The heavy shade system on the other hand may appeal to farmers because it requires less agricultural inputs such as labour and agrochemicals, even though it may result in a highly humid environment that promotes pest and disease build-up (Nunoo, 2015).

Table 2. Classification of tree density on cocoa farms in Sefwi Wiawso Municipal

Shade class	Frequency	Percentage
Full sun/No shade	19	7.51
Low shade	80	31.62
Medium shade	78	30.83
High/Heavy shade	76	30.04
Total	253	100.00

Farmers perception of cocoa agroforestry

Farmers’ perceptions of the ecological advantages of cocoa agroforestry are essential for the implementation of environmentally-friendly production practices. Hence, farmers’ perception with regards to cocoa agroforestry was assessed using a five (5) point Likert scale with 1 = strongly agree, 2 = agree, 3 = undecided, 4 = disagree, and 5 = strongly disagree. The summary results are presented in Table 3. The analysis revealed a minimum and maximum mean values of 1.28 and 1.87 respectively.

Table 3. Farmers’ perception of cocoa agroforestry

Perception statements	Mean	Rank
Agroforestry technology extends the life of cocoa	1.28	1^st
Shade trees increase the humidity level of the farm	1.34	2^nd
Shade trees with medicinal benefits increase traditional knowledge of medicinal plants	1.39	3^rd
Shade trees protect cocoa trees from direct sunlight	1.40	4^th
Shade trees provide alternate source of wood for cooking	1.40	5^th
Shade trees improve soil nutrient content	1.42	6^th
Cocoa AF improves farm income	1.46	7^th
AF system conserves natural resources and maintains ecosystem	1.51	8^th
Cocoa AF helps to cultivate other beneficial crops	1.55	9^th
Cocoa AF provides sustainable yield than no shade system	1.59	10^th
Cocoa AF technologies reduce the cost of farm management	1.70	11^th
Cocoa under agroforestry technology requires less fertilizer	1.70	12^th
Cocoa under AF technology increases pest incidence than no shade cocoa technology	1.87	13^th

Generally, the results revealed that cocoa farmers strongly agree that cocoa under AF has longer life span. This may be attributed to the fact that cocoa AF was perceived by the farmers to conserve natural resources and maintain the ecosystem. Compared to monocultures planted in full sun, tree crops grown in shaded systems typically sustain production over time and are less vulnerable to pests and diseases (Belsky & Siebert, 2003). Farmers also strongly agree that shade trees increase the humidity of the farm and enhance traditional knowledge on the use of medicinal plants. Shade trees provide additional tree canopy which prevents the direct effect of sunlight thus improving humidity on the farm. Cocoa AF practice also enhances farmers’ knowledge of medicinal plants. In addition, farmers strongly agree that shade trees protect cocoa trees from direct sunlight, provide alternate source of wood for cooking, while cocoa AF helps to improve farm income. Owusu et al. (2021) in their study in Ghana found that more cocoa farmers strongly agreed that shade trees enhance farm net returns. Furthermore, the respondents strongly agree that shade trees enhance soil nutrient content. This is because shade trees improve soil moisture content and help to conserve the natural ecosystem. A similar result was obtained by Owusu et al. (2021) in their study in Ghana which showed that more farmers strongly agree that shade trees enhance soil nutrient and moisture content.

Additionally, cocoa farmers agree that AF technology conserves natural resources and maintains the ecosystem, helps farmers grow other beneficial crops on the same land, and reduces the cost of farm management. The no shade AF system, for example, is the most luxurious cocoa AF technique to adopt because of its heavy reliance on labour and agrochemicals. Furthermore, the respondents agree that cocoa under shade trees requires less fertilizer. This is because shade trees improve soil nutrient content. The majority of respondents in a related study by Owusu et al. (2021) agreed that cocoa under cocoa AF require less fertilizer.

The results further indicate that farmers agree that cocoa AF gives sustainable yield. This is because cocoa AF improves soil nutrient and moisture content, maintains the ecosystem, and extends the life of cocoa trees. The finding is supported by a study conducted in Ghana by Yamoah et al. (2021), which revealed that shaded cocoa AF systems give sustained yields in the medium- to long-term compared to unshaded systems. However, the results contradict those of Owusu et al. (2021), who noted that a greater proportion of the farmers in their research strongly disagreed with the notion that cocoa yields are increased by shade trees.

Factors influencing adoption of cocoa agroforestry

In this section, we present the ordered probit model estimates of the factors influencing adoption of cocoa agroforestry. The results can be seen in Table 4. We focus on the marginal effects of the explanatory variables on the ordered dependent variable (tree density). We reported the marginal effects because the coefficients do not indicate the magnitude of the effect of the independent variables.

Table 4. Marginal effects of the ordered probit model

Variable	Full sun		Low shade		Medium shade		High shade
	Marg. Eff.	S. E.	Marg. Eff.	S. E.	Marg. Eff.	S. E.	Marg. Eff.	S. E.
Farmer characteristics
Sex	0.079*	0.040	0.223***	0.084	−0.061*	0.036	−0.241***	0.088
Years of education	−0.001	0.001	−0.004	0.005	0.001	0.001	0.004	0.006
Household size	−0.003	0.002	−0.009*	0.005	0.002	0.001	0.010*	0.006
Experience	0.002***	0.001	0.007***	0.002	−0.001*	0.001	−0.008***	0.002
Asset value (normalized)	−0.009*	0.005	−0.033**	0.017	0.006	0.005	0.036**	0. 018
Main economic activity	−0.067*	0.035	−0.160***	0.056	0.064*	0.039	0.162***	0.051
Sex*experience	−0.002*	0.001	−0.007*	0.004	0.001	0.001	0.007**	0.004
Farm-specific factors
Total farm size	0.018***	0.004	0.066***	0.016	−0.012**	0.006	−0.072***	0.015
Total farm size squared	−0.002***	0.001	−0.007***	0.003	0.001*	0.001	0.008***	0.003
Number of farms	−0.024***	0.009	−0.086***	0.030	0.016*	0.009	0.094***	0.032
Own cultivation	−0.080*	0.042	−0.179***	0.059	0.078*	0.044	0.182***	0.055
Institutional factors
Extension visits	0.002	0.016	0.008	0.057	−0.001	0.010	−0.009	0.063
Farmer group member	−0.031	0.022	−0.097*	0.058	0.025	0.022	0.103*	0.058
Off-farm work	0.036**	0.017	0.127***	0.048	−0.023	0.014	−0.140***	0.052
Training in agroforestry	−0.029*	0.016	−0.100**	0.044	0.020	0.014	0.109**	0.047
Participation in CMSE	−0.003	0.012	−0.011	0.045	0.002	0.008	0.012	0.049
Participation in CHTP	−0.031	0.025	−0.091	0.060	0.028	0.026	0.094	0.059
Participation in CHPE	−0.026**	0.012	−0.113**	0.056	0.007	0.011	0.133*	0.068
Participation in CRP	0.028	0.020	0.087*	0.051	−0.023	0.019	−0.092*	0.051

***, ** and * are statistically significant at 1%, 5% and 10% respectively.

The results revealed that sex positively influenced adoption of no shade and low shade agroforestry, but reduced adoption of medium and high shade agroforestry. This indicates that female farmers have higher probability of practicing full sun and low shade agroforestry systems and lower probability of engaging in medium and high shade systems. A possible explanation for this outcome is that women are more constrained in terms of resource ownership and access, which may limit their capacity to adopt improved technologies. Also, women usually lack secure access to land (land tenure insecurity) which may constrain their adoption of technologies. In addition, it is a tedious job to plant and nurture trees and this could limit adoption of cocoa AF among women farmers. Finally, it is a common practice for women to focus more on food crops, while their male counterparts focus on cash crops. The result of this study confirms findings by Danso-Abbeam et al. (2012) and Nunoo et al. (2020) which revealed male domination in cocoa AF adoption in Ghana.

Additionally, the results indicate that adoption of low shade agroforestry system is inversely related to household size, while adoption of high shade cocoa AF increases with household size. Thus, larger farm households have lower likelihood to adopt low shade cocoa AF but more likely to adopt high shade agroforestry system. Intuitively, household size improves the possibility that family labour will be available to implement labour-intensive new farming management techniques and adaption measures. This finding is in parity with that of Kyere (2018) which discovered that because cocoa production requires a lot of labour, smallholder cocoa farmers who have big families can use family members to assist them with on-farm tasks. Namwata et al. (2010) also observed that producers with large household size had a higher likelihood to adopt improved technologies thereby increasing the density of trees from low to high shade.

The findings further revealed that adoption of full sun and low shade agroforestry correlated positively with farming experience while adoption of medium and high shade agroforestry had a negative correlation with farming experience. Thus, the more experienced the farmer, the less likelihood to adopt medium and high shade systems. This means that an increase in farming experience is associated with lower tree density and similar to the results obtained by Nunoo et al. (2020) which showed that farmers become more risk-averse of innovative practices such as agroforestry with age and experience. Farmers who are older and more experienced may lose interest in making long-term investments in the farm, which would shift the density of shade trees towards low/no shade.

We also found that farmers who own assets are less likely to adopt low or no shade agroforestry system but would rather increase adoption of medium and high shade agroforestry systems which agrees with the findings of Obadimu et al. (2020). Moreover, farmers who consider cocoa farming as their major source of livelihood are expected to practice medium or high shade agroforestry system as opposed to no/low shade agroforestry.

Furthermore, producer having larger farms are less likely to adopt medium and high shade system but more likely to adopt no shade or low shade agroforestry. That is to say that an increase in farm size is associated with lower tree density. While the result does not lend itself to easy interpretation, it may be argued that the larger the farm, the more difficult it is for farmers to populate it with trees, except for natural growth of trees. The cost of increasing tree populations on a farm may be prohibitive to farmers, who may thus rely on existing tree cover as a result. This result is, however, inconsistent with other past studies which reveal a positive association between farm size and tree density (Nunoo et al., 2020; Phiri et al., 2004).

The results also indicated that farmers who cultivate their own farms are likely to adopt medium and high shade agroforestry systems. The findings are plausible because personal ownership provides long-term land security, which may promote the planting of trees on farms. Cultivation and tending of trees take several years, hence without assurance of long-term land security, there is little likelihood and motivation to plant trees on a farm. The findings agree with Sebukyu and Mosango (2012) who asserted that producers who cultivated their own land had a higher likelihood to practice agroforestry because of the assurance of long-term land security.

It was further revealed that members of farmer groups are likely to adopt high shade agroforestry system with decreasing probability of practicing low shade tree density system. The results are justified because farmers who aggregate in groups through cooperative societies have a higher chance to receive information on agroforestry practices and the associated benefits. Thus, high social capital and participation in farmers’ organizations improve access to AF information such as the recommended shade tree density. It is also important to emphasize that farmer-to-farmer learning plays a critical role in technology adoption among farmers. As indicated by Lin et al. (2021), the possibility that the farmer would understand the significance of agroforestry technology and subsequently adopt it are increased through social connections and networks. The result aligns with that of Kassie et al. (2009) in their study in Ethiopia.

Additionally, the results revealed that engaging in off-farm employment reduces the likelihood of adoption of high shade agroforestry system as testified by Wijayanto et al. (2022). This result, however, contradicts the findings of Kassie (2017) which showed that off-farm activities promote cocoa AF and technology adoption. Farm households typically engage in off-farm activities to generate additional income to support household expenditure. This income may be used to support on-farm activities or finance household consumption. Adoption of cocoa agroforestry may not increase with off-farm participation if the income from off-farm work is spend on household consumption and not on farm work. Also, farmers who engage in off-farm work may not be full-time farmers, hence they may not devote as much time and resources into the farm business as full-timers would. These factors could explain the findings of this study.

Also, cocoa farmers who received training in agroforestry were found to have higher probability of engaging in high shade agroforestry and unsurprisingly had lower chance of adopting no shade and low shade agroforestry systems. Adoption of farm technologies is positively correlated with farmers’ knowledge of the technology, which is further enhanced by training received from agricultural extension agents and other stakeholders working with farmers to enhance their productivity. The result conforms to that of Pratiwi and Suzuki (2019) which found training to enhance knowledge and practice of AF among farmers in Indonesia.

The results also indicated that cocoa hand pollination programme participation encourages the probability of adopting high shade agroforestry (Jha et al., 2021; Pratiwi & Suzuki, 2019) while decreasing the probability of embracing no shade and low shade systems. Moreover, farmer participation in cocoa rehabilitation programme was found to decrease the probability of adopting high shade agroforestry system and counter intuitively favors adoption of low shade system. One explanation for the outcome is that the current Ghana cocoa rehabilitation programme has no AF promotion component as its objective.

Constraints to cocoa agroforestry adoption

In this section, we present the various constraints associated with adoption of cocoa AF by the respondents. The most important constraint is the improper communication of the benefits of shade trees to cocoa farmers as shown in Table 5. This affirms the finding by Nunoo (2015) which revealed that benefits of shade tree had not been communicated to farmers which limited the planting of shade trees on cocoa farms. Poor information dissemination about cocoa agroforestry technology ranked second whilst lack of training on cocoa agroforestry technology was the third most pressing challenge among the constraints to adopt cocoa agroforestry shade levels, which is similar to the findings of Shah (2023) on scaling up agroforestry in Fiji.

Table 5. Constraints to cocoa agroforestry adoption

Constraints	Mean rank	Rank
Benefits of cocoa AF not well communicated to farmers	3.01	1^st
Poor information dissemination about the technology	3.28	2^nd
Lack of training on cocoa AF	3.93	3^rd
Difficulty in managing shade trees	4.33	4^th
Lack of quality seeds and seedlings	4.51	5^th
Seedlings are costly and not affordable	4.96	6^th
Inadequate agricultural extension staff	5.92	7^th
Long distance from home to farm	6.06	8^th

Other constraints identified include difficulty in managing shade trees due to lack of training, inadequate access to quality tree seedlings to plant as shade trees, and the high cost of seedlings and their unaffordability. The challenge of inadequate agricultural extension staff and inadequate extension visits was identified as the penultimate constraint, while the least constraint is long distance from farmers’ residence to the farm, which increases the transaction cost in adopting new technologies.

Conclusion and recommendations

The study examined cocoa farmers’ perception of agroforestry, the determinants of cocoa agroforestry adoption as well as the challenges confronting cocoa agroforestry adoption in the Sefwi Wiawso Municipal, Ghana. The data was analyzed descriptively and econometrically. With regards to farmers’ perception of AF, the respondents either ‘agreed’ or ‘strongly agreed’ to most of the perception statements regarding the positive effect of cocoa AF on cocoa production. This shows that the farmers perceive AF to exert a positive influence on cocoa production. The results further indicated that the cocoa landscape in the study area consists of 7.51% full sun (no shade), 31.62% low shade, 30.83% medium shade and 30.04% heavy/high shade. Hence, we conclude that the farmers are somehow indifferent to the adoption of the different cocoa AF systems. This result is contrary to expectation taking into account the good perceptions farmers have about cocoa AF. Thus, apart from farmers’ perceptions, other factors could play a key role in cocoa AF adoption. The factors influencing the adoption of cocoa agroforestry technology in the study area include individual/household, farm-specific, and institutional factors. The results indicated that the adoption determinants varied across the different AF systems. The study also identified major challenges affecting cocoa AF adoption. The most critical challenges associated with cocoa agroforestry adoption were poor communication of the benefits of cocoa AF technology to farmers, lack of knowledge of cocoa AF, lack of training on the technology, and difficulty in managing shade trees.

We recommend that effective training should be given to cocoa farmers to promote cocoa agroforestry adoption. When farmers are provided with the appropriate training, it will enhance their knowledge and expertise and thereby promote agroforestry adoption. This will also help farmers to overcome the difficulties they face in managing shade trees on their farms. It is also recommended that the benefits of the cocoa agroforestry technology should be well communicated to cocoa farmers via appropriate communications channels such as farmers’ fora, farmer groups, electronic media, among others to promote adoption. In line with this, farmers should be encouraged to join farmer-based organizations and other social networks for peer learning to enhance adoption of cocoa agroforestry. The Ghana Cocoa Board (COCOBOD) under the Cocoa Research Institute of Ghana (CRIG) division should also ensure that tree seedlings are made available and affordable to cocoa farmers to enhance cocoa agroforestry adoption. We further propose that similar to the national tree planting campaign (Green Ghana Day celebration), a cocoa Tree Planting campaign should be instituted to promote cocoa agroforestry adoption. This will heighten awareness about the importance of agroforestry and help to reduce the loss of forest trees and vegetation cover due to small-scale mining activities, logging and clearing of cocoa farmlands for crop production.

Limitations of the study and suggestions for further research

The study used a small sample size and was localized in only one cocoa producing district in Ghana. It is recommended that future studies should focus on a wide geographical area and use a larger sample size. In addition, further empirical studies are required to ascertain the impact of tree density levels on the yield of cocoa.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Appendix 1.

The classification of cocoa agroforestry systems

Table

Characteristics	Full Sun	Low Shade	Medium Shade	High Shade
Number of shade trees/Ha	0–2	3–9	10–15	>15
Canopy cover (%)	<36	36–65	66–85	>85

Source: Adapted from Nunoo et al. (2014).