Introduction

Coffee is one of the most traded commodities globally in terms of quantity and value (Vegro & de Almeida, 2020). The international coffee market has become complex, with more countries participating in the trade, and some re-exportation of processed coffee by non-producing countries (Utrilla-Catalan et al., 2022). The major coffee producers are located in developing countries, while the developed nations are the main consumers (Vegro & de Almeida, 2020). Approximately 70% of the global coffee production comes from Brazil, Vietnam, Colombia, and Indonesia (Vegro & de Almeida, 2020), while the European Union, United States, Brazil and Japan are the leading consumers (International Coffee Organization, 2020). The global coffee production has exhibited slight oscillations in productivity while consumption demand has steadily been increasing (International Coffee Organization, 2024). The global coffee production stood at 169.18 million 60 kg bags in 2023/24, representing a 3% growth rate in production; while the consumption reached 177 million bags, representing a demand deficit of 7 million bags (International Coffee Organization, 2024). The fluctuations in coffee production may be attributed largely to climate-related effects and the biennial bearing phenomenon (Bager & Lambin, 2020), while the ever-rising consumption demand could be linked to the growing world population, changes in lifestyles and the nutritional benefits associated with coffee.

In the 2023/2024 crop year, Africa’s total coffee production stood at 20.1 million 60 kg bags representing a 12.2% growth rate from the 2022/2023 crop year (International Coffee Organization, 2023). Kenya produces 5% of the African coffee and is ranked fifth in coffee production behind the leaders, Ethiopia and Uganda, with a production share of 39% and 23%, respectively. Cote d’Ivoire takes the third place with 13% production share while Tanzania accounts for 6% of African coffee production (International Coffee Organization, 2023). The African coffee consumption is highly concentrated in Algeria, South Africa, Egypt, Morocco, Ethiopia and Sudan (International Coffee Organization, 2023). In Kenya, coffee is produced in 33 counties spread in Western, Nyanza, Rift Valley, Central, Eastern and Coastal Regions (International Coffee Organization, 2019). The coffee sub-sector contributes about 1% to the GDP, 8% of the total agricultural export earnings and 30% of the total labor force in the agricultural sector (International Coffee Organization, 2024). The Kenyan coffee is produced by approximately 800,000 smallholder farmers and about 3000 large and medium-scale farmers, commonly referred to as estate farmers (International Coffee Organization, 2019). Kirinyaga is the second leading coffee producing county after Kiambu, while Embu is ranked a distance 8th (Agriculture and Food Authority AFA, 2024). Coffee productivity among the smallholder farmers is relatively low at an estimated average of 2 kg of cherry/tree/year compared to the potential productivity, which is estimated at 10 kg of cherry/tree/year (Wambua et al., 2024).

The smallholder coffee farmers in Kenya are organized into cooperative societies, which enables them to benefit from economies of scale in coffee processing and marketing. The farmers deliver harvested coffee cherries to the factories in their affiliated cooperative societies for primary processing (wet processing). The resultant parchment coffee is then sent to the dry mills for secondary processing, which comprises hulling, polishing, and grading of coffee into uniform green beans. The green beans are also referred to as clean coffee, and this is the form in which the bulk of Kenyan coffee is marketed. Insecurity during coffee transportation, handling and storage, information asymmetry, delayed payments, inadequate marketing skills, high deductions, and lack of capacity to enforce some agency obligations are some of the notable challenges along the coffee value chain (Takama et al., 2022). About 85% of Kenyan coffee is marketed through a centralized auction managed by the Nairobi Coffee Exchange (NCE), while the rest goes through direct sales facilitated by appointed commercial coffee marketing agents (Takama et al., 2022; GoK, 2024). In the central auction system, licensed coffee dealers competitively bid to purchase the coffee. Some buyers may request small samples of the coffee for sensory analysis to determine the beverage or cup quality, which they use to set their price bid offers. On the other hand, the direct sale system entails a direct negotiation between growers and buyers who may be marketing agents or international roasters or suppliers. The parties enter into a sales agreement, which is registered with the Coffee Directorate. The growers may be individual farmers or a group of farmers in an association or a cooperative (Takama et al., 2022).

The marketing infrastructure, including the trading floor at the NCE, is underdeveloped (GoK, 2020) with some leakages or gaps that tend to favor the traders at the expense of the producers. Lack of transparency and accountability in the marketing process does not protect the farmer, and this exposes them to opportunistic behavior of market actors such as traders or marketers, which is exacerbated by a lack of basic marketing skills among the farmers. The coffee growers are usually attracted to a certain marketing channel if the expected utility gained through that particular channel is greater than that derived from other available options (Jebesa, 2019). Coffee productivity is on the decline, and the reasons are not well documented. Smallholder coffee farmers’ involvement in the marketing process is usually limited to the channel of choice. Determinants of the farmer’s choice of a particular marketing channel, the channel characteristics and income levels are not well documented. Information gap and flow is a huge challenge to these smallholder farmers, leading to elite capture and price fixing in the coffee value chain. Previous studies have shown that before choosing a marketing channel, farmers consider the transaction costs, returns, market information, level of trust among the available channels, and familiarity with the markets (Mgale & Yunxian, 2020). This study provides empirical evidence on the determinants of coffee production at the farm level and the choice of a particular coffee marketing channel in Kenya. The findings would be immensely helpful to the smallholder coffee farmers, market actors and policy makers in policy formulation and review for coffee value chain inclusion and sustainability. The paper is organized into introduction, materials and methods, results, discussion, conclusion and recommendation.

Materials and methods

Study area

The study was undertaken in Kirinyaga (0.53° S, 36.98° E) and Embu (0.43° S, 37.48° E) Counties in Central Kenya. Coffee producing areas in these counties lie in the Upper Midlands 1 (UM 1) and Upper Midlands 2 (UM 2) agro-ecological zones. Majority of coffee farmers are smallholders who are organized into cooperative societies through which they process and market their coffee.

Sampling and sample size

Purposive sampling technique was used to identify the coffee-producing sub-counties in the two target counties. The three main coffee cooperative societies were sampled from each of the two sub-counties, and probability proportional to size technique was used to construct the sampling frame based on the membership of the sampled cooperative societies (Table 1). The resultant sample size was 385 comprising of 209 respondents from Embu County and 176 respondents from Kirinyaga County, as determined using the Cochran (1963) sampling formula as shown below.

$$n=\frac{{Z}^{2}pq}{{E}^{2}}$$
(1)
$$=\frac{{(1.96)}^{2}(0.5)(0.5)}{{(0.05)}^{2}}=385\;respondents$$
Table 1 Sampling frame.
Full size table

Data sources

The study deployed a cross-sectional survey design to collect primary data on the socio-economic characteristics of the sampled farmers and their coffee production data for the 2022/2023 crop year. The data collection period was between January and April 2023. The primary data consisted of household farm characteristics (farmer demographics), economic and institutional factors, production levels, market information, characteristics of the marketing channels, and the determinants of the choice of marketing channels. Data cleaning, coding and entry were done for descriptive and inferential analysis.

Empirical Models and Specifications

Descriptive statistics for farmer demographics, such as mean, proportions, frequencies, standard error, and chi-square tests, were used. The Shapiro-Wilk Test was deployed to test for normality of the distribution of the data (Gaussian distribution). A large p-value above 0.5 indicates a normal distribution of the data. The study adopted the Tobit regression model and Endogenous Switching Regression model to evaluate the determinants of produced volumes and the farmers’ choice for coffee marketing channels, respectively. According to Khonje et al. (2018) the Tobit model assumes that the observed dependent variable Yj for observations \(j=1,\mathrm{....}.n\) satisfies;

$${Y}_{j}=\,\max ({Y}_{j}^{\ast },0)$$
(2)

Where \({Y}_{j}^{\ast \text{'}}\) = latent variables generated by the linear regression model.

$${Y}_{j}^{\ast }={\beta }^{\text{'}}{X}_{j}+{U}_{j},{Y}_{j}=\sum \frac{{Y}_{j}^{\ast }if{Y}_{j}^{\ast }\ge 0}{0if{Y}_{j}^{\ast }\le 0}$$
(3)

Where Xj = corresponding explanatory variables, β = vector parameters, Uj = model errors assumed to be independently normally distributed.

On the other hand, the Endogenous Switching Regression model was preferred for evaluating the determinants of choice of marketing channels since it accounts for the selection bias and endogeneity originating from the observed and unobserved heterogeneity. It was hypothesized that farmer’s decision to choose a certain marketing channel was based on a random utility framework and the expected benefits associated with it. In the first stage, the farmer’s choice of a particular marketing channel was hypothesized to be a function of socio-economic and institutional factors observed ceteris paribus. The empirical model was specified as follows:

$${U}_{jit}^{\ast }={\alpha }_{j}{X}_{jit}+{\omega }_{j}{\overline{X}}_{ji}+{\varepsilon }_{jit}$$
(4)

where \({U}_{jit}^{\ast }=\) ith farmer’s choice, Xjit = vector of observed exogenous variables (socio-economic and institutional factors), \({\overline{X}}_{ji}=\) means of unobserved explanatory variables, α, ω = parameters to be estimated and εjit = random error term.

$${Y}_{i}={X}_{i}^{{\prime} }{\alpha }_{s}+{\varepsilon }_{i},s=0,1$$
(5)

Where Yi = latent variable for produced volumes, \({X}_{i}^{{\prime} }=\) observed explanatory variables, αs = parameters to be estimated and εI = random error term.

$${Y}_{1}={X}_{i}{\beta }_{i}+{\delta }_{\in i\in }{\lambda }_{i}+{\mu }_{i}\,{\bf{if}}\,{\bf{D}}={\bf{1}}$$
(6)
$${Y}_{o}={X}_{i}{\beta }_{o}+{\delta }_{\in o\in }{\lambda }_{o}+{\mu }_{o}\,{\bf{if}}\,{\bf{D}}={\bf{0}}$$
(7)

Where yj with j = 0, 1 latent variable for coffee returns.

Results

This section provides results on the data normality tests, coffee production among the farmers, coffee marketing channels and characteristics and determinants of choice of marketing channels. Both descriptive and inferential statistics are presented in this section for meaningful inferences.

Normality test for adequacy of the sample

The dependent variables fitted in the normality test were all significant at 5% level indicating a normal distribution of the data with Wilk coefficients (W) of above 0.5 which is the conventional threshold. The z-test (Z) statistics and the p-values (V) both confirmed normal distribution of the data. This confirmed adequacy of the sample in subjecting the data into inferential statistics and test of hypotheses.

Bayesian distribution (residual plot) for coffee output among the farmers

Figure 1 provides Bayesian Distribution (Residual plot) for the produced coffee volumes among the farmers along the production node of the value chain. The iteration plot (trace) indicates the values as entered in the distribution, indicating the randomness of the data to generate the histogram. The histogram indicates a normal and uniform distribution of produced coffee volumes at a 95% confidence level. The autocorrelation plot indicates that there was no excessive correlation in the model specification, hence its suitability. The kernel density plot of the produced volumes also indicates a normal distribution for the entire sample.

Fig. 1
figure 1

Bayesian distribution (residual plot) for produced coffee volumes.

Full size image

Coffee production in the study area

The trend of coffee production levels with respect to farm size under coffee in the study area is presented in Fig. 2. Majority of the farmers had less than one acre of land under coffee with production levels of less than 1000 kilograms per crop year.

Fig. 2
figure 2

Coffee production volumes (•) against different farm sizes.

Full size image

Effect of farmer demographics on coffee production (output) at farm level

Table 2 presents the influence of socio-economic characteristics on coffee output at the farm level as determined using the Tobit multiple regression model. Off-farm income, total land size, farm size under coffee production and market information had significant effects on the coffee output levels (Table 2). Off-farm income had a positive coefficient of 0.123, implying that farmers with diversified income sources increased their coffee output levels by 12%. Total household land size had a positive coefficient of 0.235, implying that a unit increment in total land size would increase coffee production by 23% ceteris paribus. Farm size under coffee production had a negative coefficient of 0.288, implying that a unit increase in farm size under production reduced coffee output per unit area by 28% and vice versa due to diseconomies of scale. Market information had a positive coefficient of 0.468, implying that farmers’ access to market information improved production levels by 46%.

Table 2 Effect of farmer demographics on coffee output at farm level.
Full size table

Availability of coffee marketing information

Table 3 provides descriptive statistics on the access and value of the coffee marketing information. The findings indicate that the majority (91%) of the farmers had access to market information on prevailing market prices. The type of market information accessed was mainly on market prices (63.7%) and market linkages (24.5%). The main source of the market information was extension officers (63.9%), other farmers (13.2%), radio/television (11.9%), as well as internet (9.3%). The majority (68%) of the respondents opined that the information was reliable and considerably impactful.

Table 3 Descriptive statistics for market information on coffee marketing.
Full size table

Coffee marketing channels

The coffee marketing in the study area was found to be dominated by one marketing channel, which is the Farmer—Cooperative Society—Miller/Marketer—Nairobi Coffee Exchange (NCE)—Export. This channel commanded 84% preference (Table 4). The second most popular channel was the Farmer—Cooperative Society—Miller/Marketer—Export with a popularity of 13%. The third channel was Farmer—Miller/Marketer—NCE—Export, which was being used by only 2% of the respondents (Table 4).

Table 4 Proportion statistics for marketing channels.
Full size table

Determinants of choice of marketing channels

The choice of a marketing channel was a function of various explanatory socio-economic and institutional variables (Table 5). The model specification results [Wald chi2 (25) = 154.59, Log likelihood = −48.3071 and Prob > chi2 = 0.0000] implied that the fitted model was statistically significant and suitable in explaining the variations observed. Membership in a cooperative society, production volumes, processing losses and price per kilogram of coffee sold significantly influenced the farmer’s choice of a particular coffee marketing channel. Membership to cooperative society had a positive and significant (p = 0.000) effect on choice of marketing channel with a coefficient of 0.64994, implying that farmer’s choice to be a member of a cooperative society influenced his or her marketing channel choice by 64% as majority of the smallholder farmers were using the cooperative societies to deliver and market their coffee.

Table 5 Determinants of choice of marketing channels.
Full size table

Produced volumes at the farm level had a negative and significant (p = 0.000) effect on choice of marketing channel with a coefficient of 0.27516, implying that reduced or low volumes at the farm level limited the farmer’s capacity to choose a particular channel by 27%, such as the direct marketing which would require high volumes for a farmer to independently process and source a buyer directly. Processing losses (accounted leakages) had a positive and significant (p = 0.000) effect on choice of a marketing channel with a coefficient of 0.27517, implying that higher leakages influenced a farmer’s choice of a marketing channel by 27% as a result of old and ineffcicent facilities and equipment at the cooperative societies. Price payable per kilogram of coffee by the buyers had a positive and significant (p = 0.009) effect on choice of marketing channels with a coefficient of 0.53488, implying that a unit increase in prices paid per kilogram, subject to coffee quality and grade, increased and the channel preference by 53% ceteris paribus.

Discussion

Sustainable production and marketing are critical elements of the coffee value chain. Economic factors are integral in the value chain development, as they are the drivers of economic growth and capital accumulation. In this study, factor endowment was hypothesized to play a key role in influencing farmers’ coffee output levels. Off-farm income was found to have a significant influence on the coffee production at the farm level. Apparently, the additional income was useful in absorbing the risks and random shocks in the coffee production and marketing. Possibly, the competitiveness of the less-endowed farmers is reduced by the increasing marketing costs and transactional friction caused by value chain complexity and the long coffee marketing channels. Similar findings were reported by Wambua et al. (2024); Ngare (2021) and Gebre et al. (2020). On one hand, the smallholding nature of most of the sampled farmers significantly influenced the coffee outputs at the farm level, with coffee farm size expected to be directly proportional to the available land size. The shrinking land sizes may have resulted from land subdivision due to increasing population, urbanization, changing agricultural production patterns, among others, as reported by Murimi et al. (2019). On the other hand, an increase in the coffee farm size reduced coffee output per unit area, possibly because intensive crop management is usually affected by large land sizes, thus reducing the output per unit area. This is based on the ceteris paribus assumption, that when farm size is the only factor under consideration, which in reality it is not the case, an increment in farm size may lead to diseconomies of scale, spread thin on management skills and crop management practices among others. These findings concur with those of Kudama (2019), Adane and Bewket (2021) and Wambua et al. (2021).

This study also identified market information as another critical factor influencing coffee production at the farm level. The plausible explanation is that the availability of reliable market information gives the farmers some production confidence and helps to offset the inherent production risks along the value chain. The majority of the sampled farmers confirmed that they had access to market information, particularly on the market prices and market linkages. A study by Mgale and Yunxian (2020) revealed that farmers’ share of produced output was one of the quantitative measures of production efficiency; the greater the farmers’ share, the higher the production efficiency. According to the sampled farmers, the most reliable source of coffee marketing information was the extension officers, probably because they usually have a close attachment with the farmers, thus building their trust. Market information was one of the factors affecting coffee productivity at the farm level as reported by Bhattarai et al. (2020) and Wambua et al. (2021). Access to reliable market information may be cushioning the farmers from the marketing risks, thus building their production confidence. Access to market information reportedly boosts the farmer’s production confidence thus enabling them to choose better farm inputs that would guarantee profit maximization based on the prevailing market dynamics. This argument supports the findings of Jebesa (2019) that inadequate market information is morally hazardous and may result in adverse selection of inputs and production strategies, which in turn increases the cost of production and consequently affects farm productivity.

Marketing channel is a pathway through which a commodity moves along the value chain to the ultimate consumer (Canwat, 2023). Farmer’s choice of a particular marketing channel plays a key role in coffee profitability. In Kenya, coffee is largely produced for the export market, with little value added and local domestic consumption. This is largely similar to the international framework of coffee production, whereby the production majorly happens in emerging economies while consumption is concentrated in developed economies (Utrilla-Catalan et al., 2022; Canwat, 2023). The millers and marketers provide the link between the producer (either the cooperative or individual farmers) and the overseas buyer, either directly or through the auction. The main marketing channel identified in this study, i.e., Farmer—Cooperative Society—Miller/Marketer—NCE—Export, was centered on cooperative societies as the primary entry level. The farmers deliver their coffee to the wet mill factories affiliated with the cooperative societies. After the primary processing in the wet mills, the coffee parchment is then subjected to secondary processing in the dry mills. Some cooperatives have their own dry mills while others have contractual agreements with millers for secondary processing. The processed coffee is then auctioned at the Nairobi Coffee Exchange (NCE) through competitive bidding, mainly by the exporters. The farmer's involvement in this marketing channel is minimal and ends at the cooperative society, with little or no involvement in the upstream decision-making.

On the choice of marketing channels, membership in a cooperative society was a key determinant. This is because most of the sampled coffee farmers are smallholders who are organized into cooperative societies through which they cost-share the coffee processing and marketing activities, and by so doing, they enjoy economies of scale and reduce their production cost. In addition, the farmers’ membership in the cooperatives enables them to accumulate higher coffee volumes that are adequate for joint marketing, as similarly reported by Gebre et al. (2020). Pham et al. (2019); Mwinyiheri et al. (2023) reported that harvested quantity largely influenced farmers' choice of a particular marketing channel, with higher quantities choosing industrial processing while mid-size farms choosing wholesale markets. Fon et al. (2019) and Abasimel (2020) reported that the financial and market capacity of the cooperatives would increase farmers’ choice towards cooperative outlets based on the category. A study by Orr et al. (2018) revealed that smallholder farmers join cooperative societies or other organizations to benefit from cheaper production inputs, collective marketing, credit and other services. The cooperative societies, through the contracted extension officers, also facilitate knowledge transfer and collective action among the farmers. Eshetu et al. (2021); Fon et al. (2019) found the education level of members, payment system, average price of coffee, access to inputs and extension services to cooperative members as the main determinants of the choice of marketing channel.

Produced volumes at the farm level were also a key determinant of the farmers’ choice of a particular marketing channel. The possible expectation is that individual farmers who are able to raise adequate volumes are the only ones who can independently choose their preferred marketing channel. Amrulloh et al. (2021) reported that coffee productivity was a key determinant for expanding coffee exports in Indonesia. On the other hand, a cooperative society attaining higher volumes of coffee from its members attracts many marketers and thus has better bargaining grounds and a bigger leeway to choose the most attractive marketing channel. The raw coffee quality was also a key determinant of the choice of marketing channel, as it mainly determines the processing losses and ultimately affects the marketable volumes. Better coffee quality also attracts many potential buyers, thus presenting the producers with a wider choice of their most desirable marketing channel and an opportunity to demand better prices. These findings concur with those of Maciejewski et al. (2023). Similarly, Feleke (2018) found a direct relationship between coffee quality and market performance and marketing channel selection in Ethiopia. Tikuneh et al. (2023) observed that coffee certification and the domestic demand for high-quality coffee as key explanations for coffee market performance and market channel selection among farmers in Ethiopia.

The price payable per kilogram of coffee sold influenced the choice of a particular marketing channel. Profit maximization under the firm theory is one of the primary objectives of any rational producer. Medium estate farmers who process and market their coffee directly to international markets have the confidence and bargaining power to negotiate prices directly with the international buyers or at the auction, where the bidders place bid offers based on the coffee quality. The findings by Lalzai et al. (2023) revealed varying profit margins for farmers and intermediaries with three channels having different market efficiency scores at 47% (channel III), 32% (channel II) and 29% (channel I). At the international markets, coffee quality and grade determine the payable prices. Good quality coffee usually fetches higher prices at the international markets, and thus farmers producing such quality are more confident in seeking a market channel that would offer them the highest utility (profit). A study by Bro et al. (2020) reported that coffee consumers/producers with heterogeneous preferences and tastes are willing to choose a coffee product that maximizes their utility. Asmare et al. (2024) found that the choice of spot markets for selling dry-processed coffee in Ethiopia was significantly influenced by the average selling price of the product. Similarly, Fon et al. (2019) found the payment system and average price of coffee as key determinants of the choice of coffee marketing channel among smallholder farmers.

Conclusion and recommendations

This study concluded that the produced coffee volumes among smallholder farmers are mainly influenced by socio-economic factors, particularly availability of off-farm income, total land size, coffee farm size and market information. On the other hand, the choice of the coffee marketing channel is mainly determined by the membership in a cooperative society, the production volumes at the farm level, raw coffee quality and the price payable per kilogram of coffee sold. To gain more market share and bargaining power, relevant stakeholders need to exploit opportunities in production and marketing functions for timely payments to farmers and possibly increase the average price payable to coffee farmers. Considering that the dominant coffee marketing channel among the farmers was the ‘Farmer—Cooperative Society—Miller/Marketer—NCE—Export’, it is apparent that farmer involvement in coffee marketing may be minimal. In addition, accurate, timely, reliable and evidence-based information should be shared among the key stakeholders for informed decisions, ultimately reducing the exposure to exploitation and beggar-thy policies. This underscores the need for some policy interventions to protect the interests of the farmers whose position in the value chain exposes them to more economic vulnerabilities.

Limitations of the study

Smallholder agriculture dominates the majority of the least developed countries’ economies, especially in Sub-Saharan Africa, characterized by low productivity, rural-based, rain-fed, low technology adoption, among others. The coffee value chain has become lengthy and complex, with many intermediaries affecting the gross incomes of these farmers. This study only focused on the smallholder coffee farmers in two counties. The study also focused on the marketing node of the value chain, although there are other critical nodes such as production, processing, and consumption, among others. The data used for the study were for one crop season.