1 Introduction

The impulse towards a larger introduction of Information and Communication Technology (ICT) in the agricultural field is currently experiencing its momentum, digitization has large potentialities to provide benefits for both producers and consumers; on the other hand, pushing technological solutions into a rural context encounters several challenges [1]. Digitalization, the sociotechnical process of implementing digital advancements, is expected to lead to the next agricultural revolution [2]. Various terms have emerged to indicate different forms of digitalization, including precision agriculture, digital agriculture, and smart farming. While there is no consistent term representing such a revolution, it is commonly characterized by a fusion of emerging digital technologies such as the Internet of Things, big data, robotics, remote sensors, and artificial intelligence [2]. Therefore, agricultural digitalization or precision agriculture can be defined as the application of modern information technologies such as GPS, sensors, drones, Internet of Things (IoT), artificial intelligence (AI), and data analytics in the management of crop production [3].

However, digitalization has some limitations and challenges such as high initial investment and maintenance costs because it needs expensive and complex equipment, such as sensors, drones, satellites, computers, and software, to collect and analyze data and control the farming operations [4,5,6]. Therefore, it seems that digitalization is not applicable especially in developing countries due to the presence of poor farmers, subsistence farming systems, small farmlands, lack of technical and software knowledge among farmers, and the high cost of application of its technologies [7]. A study conducted by the Technical Center for Agricultural Cooperation (CTA) and Dalberg Advisors noted that 33 million small producers on the African continent are registered in nearly 400 digital solutions, with an annual growth of 45% in the number of registrations since 2012 [8]. The same study, however, noted that more than 90% of the digital services market for African farmers remains untapped.

The use of information and communication technologies (ICT) has become obvious with the widespread use of mobile communication and various internet services for financing agricultural activities. It makes it possible to communicate and make available agricultural information on production, marketing and better access to State services and partners on agricultural value chains. Mobile telephony and the Internet, followed by the use of mobile financial services by the agricultural world, constitute the basis of agricultural digitalization in developing countries. Sub-Saharan Africa is adopting and adapting the latest digital technologies massively and at increasing speed [9, 10].

In Benin, the Internet is increasingly widespread [11] and its penetration has been growing in recent years. It increased from 48% in 2018 to 53% in 2019 then to 69% in 2020 [12]. As for the penetration rate of mobile financial services, it also increased from 23% in 2018 to 31% in 2019 then to 43% in 2020 [12]. Despite the wide Global System for Mobile communication (GSM) coverage over the country, family farmers in Benin are poorly connected to the Internet, i.e., 5% of social networks (WhatsApp, Facebook, Instagram), and 3% for information on agriculture [13]. For example, the use of drones by family farmers is currently absent, particularly in the Plateau department in South-East Benin [13].

In this context, for a better penetration policy of these technologies in the agricultural sector, it is important to identify the factors which contribute to their adoption. Roussy et al. [14], reported that research has focused on two groups of adoption factors. For these authors, work on the adoption of innovations in agriculture initially focused on the role of individual determinants directly observable by the researcher in the farmer’s decision-making process. Today, other determinants, not directly observable, such as perceptions and preferences are also highlighted. The work of Adesina and Baidu-Forson [15] in West Africa provided insight into the role that women may play in the adoption processes of new technologies. Alene and Manyong [16] reported that very old farmers may be less able to use some new technologies efficiently. Thus, they may be more reluctant to accept products from new technologies compared to young and adult farmers [17]. Household size, which also expresses the level of available family labor, could affect the decision to adopt a new technology in agriculture [16]. Teno et al. [18] showed that the adoption of new technologies in agriculture in sub-Saharan Africa could be influenced by a number of factors such as factors relating to the socio-economic characteristics of agricultural households, factors linked to the mode of operation and management of production, factors determined by the technological package, awareness and the role of social networks, the implementation approach and monitoring and support. Li et al. [19] using Deconstructive Theory of Planned Behavior (DTPB) as the analytical framework, showed that under the behavioral belief dimension, cotton farmers value the positive effect of perceived usefulness even though the risk of the technology itself has a dampening effect on adoption intentions. Under the normative belief dimension, superior influence influenced the willingness to adopt technologies to a greater extent than peer influence. Under the control belief dimension, factors such as self-efficacy and information channels influence willingness to adopt technology and behaviour. In addition, behavioural attitudes, subjective norms, and perceived behavioural control all contribute to cotton farmers’ willingness to adopt smart agriculture technologies, and can also influence behaviour directly or indirectly through willingness to adopt.

Thus, understanding the essential variables that could accelerate or hinder the adoption of these technologies has become an important concern. This study will focus on the use of digital agricultural machines, digital platforms and social networks, and mobile money in cotton production in Benin. This production represents the country’s main export product. Indeed, due to its contribution in terms of the total value of exports of agricultural products (85%) and the value of exports of all products combined (61%) in 2021 [20], the cotton sector constitutes the basis of the economy in Benin. As such, cotton occupies a special place in agricultural policy, making it the most organized of all agricultural sectors. With this support, production has experienced an exponential evolution in recent years, particularly during the 2017–2018 campaign, even reaching a growth of 122% in volume and an increase of 74% in the areas sown over the last two seasons [21]. Cotton farming therefore has increased its added value in the transformation value chain.

Thus, this study aims to explore and understand the determinants other than those of a socio-economic nature household, which can influence the adoption of digital technology among cotton producers for better policy penetration of new information and communication technologies. This study is based on the Unified Model of Acceptance and Use of Technology (UTAUT) developed by Venkatesh et al. [22]. This model combines the factors that can influence individuals in their intentions to adopt and use technologies, in various environments [23]. It arises from the evolution of the Technological Acceptance Model (TAM) developed by Davis et al. [24], with the consideration of new factors.

2 Material and methods

2.1 Study area

The municipality of Banikoara was selected for this study, because it was the first cotton production municipality in Benin and its economy is mainly based on agriculture. Banikoara is located in the Northwest of Benin with a Sudano-Sahelian climate which covers an area of 4,397.2 km2 including approximately 49% arable land and 50% protected areas [25]. This municipality is in the Alibori department, between 2°05’ and 2°46’ east longitude and between 11°02’ and 11°34’ north latitude. It is bounded to the north by the commune of Karimama, to the south by the communes of Kérou and Gogounou, to the east by the commune of Kandi and to the west by Burkina-Faso. Its population was estimated at 246,575 inhabitants including 124,130 women (50.3%) in 2013 [26]. It has ten districts: Banikoara, Founougo, Gomparou, Goumori, Kokey, Kokiborou, Ounet, Somperoukou, Soroko, Toura.

2.2 Study framework

The acceptance and use of information systems (IS) and information technology (IT) innovations has been a major concern for research and practice [27]. Over the last several decades, a plethora of theoretical models have been proposed and used to examine IS/IT acceptance and usage [27]. These include the Theory of Reasoned Action, the Technology Acceptance Model, the Theory of Planned Behavior, and Model of Personal Computer Utilization [24, 28,29,30,31]. Many of these theoretical models, developed to explain and predict the behavior of individuals with regard to the use of information and communication technologies, have referred to theories based on research in social psychology [32]. Based on a comprehensive review and synthesis of several theoretical models, Venkatesh et al. [22] proposed the Unified Theory of Acceptance and Use of Technology (UTAUT), which has since been used extensively by researchers in their quest to explain IS/IT acceptance and use.

The choice of the UTAUT model developed by Venkatesh et al. [22] in the context of this research is justified to the extent that the latter has the advantage of being a general model of all the theoretical models that have been developed to explain the adoption behavior of humans. In addition, taking into account the moderating variables (age, gender, level of education, marital status, banking, income, variety of cotton cultivated and the main speculation practiced by the cotton farmer) further justifies this choice.

Four basic variables define the Venkatesh et al. [22] model, namely, perceived ease of use, perceived usefulness, social influences and facilitating conditions. Perceived ease of use refers to the degree of ease associated with using the system. Perceived usefulness is defined as the degree to which a person believes that using the system will help them achieve gains in job performance. Social influence is the degree to which a person perceives that significant others believe that he or she should use the new system. Facilitating conditions are defined as the degree to which an individual believes that an organizational and technical infrastructure exists to support the use of the system.

In their model, Venkatesh et al. [22] integrated moderating variables (gender, age, experience and willingness to use) which have an influence on the explanatory variables. However, this model has evolved. Venkatesh et al. [33] proposed improvements to the first model (UTAUT 1) by adding other explanatory variables such as hedonic motivation, price value and habit.

The determinants of the adoption of the unified theory of acceptance and use of technology (UTAUT 1) are summarized in Fig. 1. As for the research analysis model, inspired by UTAUT, it is presented in Fig. 2. The variables retained in the analysis model used are linked to the variables of the theoretical model of Venkatesh et al. [22]. Thus, variables such as production activity, income, bank account holding, gender and age are linked to moderating variables; Satisfaction of using mobile money, advantages of using mobile money, perception, security and utility variables are linked to performance expectations and effort expectations; Membership in a social group and social reason for adoption are linked to social influence.

Fig. 1
figure 1

Source: Venkatesh et al. [22]

Unified Theory of Technology Acceptance.

Full size image
Fig. 2
figure 2

Structural analysis model used

Full size image

2.3 Data collection

The research combined a quantitative and qualitative dataset. The data were collected from producers using a questionnaire administered to each participant during February and March 2023. This data collection followed the obtaining of various authorizations from the structures involved. Thus, the Doctoral School of Agronomic and Water Sciences of the University of Parakou gave its authorization as part of the involvement of this collection for a doctoral thesis in the said School. The Applied Statistics and Computer Science Unit attached to the Forestry Study and Research Laboratory (LERF) of the University of Parakou authorized the collection following the validation of the questionnaire and the analysis methods to be used. Finally, Municipal Union of Village Cooperatives of Cotton Producers (UCOM-PC Banikoara) gave its authorization and informed consent for the collection of data from its members (cotton producers).

The questions consider the analysis model used (Fig. 2) and are categorized into seven (7) categories, namely: 1- Identification of the respondent; 2- Knowledge and determinants of adoption of e-agriculture and digitalization and its services; 3- Use of Mobile Money; 4- Perception of agricultural producers of digitalization; 5- Comparison of expenses and income linked to each itinerary before the advent of digitalization to those obtained with the use of digitalization; 6- Security of digitalization according to producers; 7- Suggestions. Except for a few open-ended questions, most of the questions were closed questions with proposed answers that the respondent must choose. The quantitative responses were specified by the producer by giving the exact values, such as the area sown, the workforce, the workforce hired, etc. The questionnaire was downloaded and deployed via the KoboCollect platform. An initial exploratory phase in the study environment allowed testing the questionnaire and making adjustments to certain questions and answers.

The target population is represented by the 23,150 conventional cotton producers (i.e. 99% of the total population of cotton producers) and the 260 organic cotton producers (i.e. 1% of the total) identified in the municipality of Banikoara for the 2021–2022 agricultural campaign by Municipal Union of Village Cooperatives of Cotton Producers (UCOM-PC Banikoara). The population of conventional cotton producers was 22,581 men (97.5%) and 569 women (2.5%). Organic cotton producers, were represented by 175 men (67%) and 85 women (33%). The choice of producers was made based on a reasoned sampling taking into account, among other things, conventional cotton producers, organic cotton producers, those combining the two types of production, gender, level of education, ethnicity and other criteria (Table 1). Thus, a total of three hundred and fourteen (314) cotton producers spread across the ten districts of the municipality of Banikoara were selected. The sample consists of 294 conventional cotton producers, 18 organic cotton producers and 4 producers combining the two types of cotton. The numbers selected for each type of cotton and by district took, were based on the proportions obtained during the 2021–2022 cotton producers census. Producers were randomly selected in each district while taking into account representativeness, gender, area sown (large, medium and small producers) and also age (young people, adults and elderly people). Digitalization in this study was assessed by the use of digital agricultural machinery, information and exchange platforms and the use of a mobile money account. Users of at least one of these three digital technologies numbered 240 (76%) and those not using any of the three forms of defined digitalization numbered 74 (24%). Among these numbers, 66 producers were users of digital agricultural machines in association or not with the other two forms of digitalization, 134 producers were users of digital platforms and 207 were holders of mobile money accounts.

Table 1 Description of the variables and their modality used in the study
Full size table

2.4 Data analysis

The data collected were analyzed using R software. A synthesis of the data was carried out using descriptive statistics methods. To identify the explanatory factors for the adoption of digitalization or a mobile account by cotton producers in the municipality of Banikoara, the procedure of generalized linear models (GLM) with the binomial distribution and the logit function as a link was used to perform a binary logistic regression model (logit model). A backward selection of variables starting from the full model was also carried out in order to identify the explanatory factors of adoption having a significant effect on the adoption of digitalization. The selection of variables was made according to the Akaike Information Criterion or AIC [34]. The goodness of fit of each model was assessed by residual deviance and AIC. In case of overdispersion, the quasibinomial distribution was replaced by the binomial distribution for parameter estimation.

According to Akueson et al. [35], binary logistic regression is a technique used to analyze the relationship between a dependence variable y, qualitative, nominal with two modalities (coded y = 0 and y = 1 for example), and one or more explanatory variables \({X}_{i}\) (i = 1, …, k) quantitative and / or qualitative ordinal or nominal, assumed to be perfectly known. In this case, it is a question of modeling the probability \(\pi\) of adoption of digitalization or possession of a mobile account by a cotton producer according to different factors. To achieve this, a transformation of the success probabilities is carried out by the link functions, denoted by \(g\). There are several link functions, but the most commonly used is the logit function (Eq. 1):

$$g\, = \,\log it\left( \pi \right)\, = \,\ln \left[ {\pi /\left( {1 - \pi } \right)} \right]\, = \,x^{\prime}_{i} \beta \,\,\,\,\,\,\,\,\left( {i\, = \,1,\, \cdots ,\,n} \right)$$
(1)

where \(x^{\prime}_{i} = \left( {1,x_{i1} , \cdots ,x_{ik} } \right)\)  is the 1 × (k + 1) vector corresponding to the k covariates associated with a producer i and \({\varvec{\beta}} = \left( {\beta_{0} ,\beta_{1} , \cdots ,\beta_{k} } \right)^{\prime }\)  is the (k + 1) × 1 vector of associated coefficients. \({{\varvec{\beta}}}\) are parameters to be estimated, most often by the maximum likelihood method. The inverse transformation makes it possible to find the estimated probabilities as a function of the \({\varvec{x}}\)  (Eq. 2):

$$\pi =\text{exp}\left(g\right)/\left[1+\text{exp}(g)\right]$$
(2)

For a given producer i, with characteristics \({{\varvec{x}}}_{{\varvec{i}}}\), the ratio between the probability \(\pi\) of adopting digitalization (or of having a mobile account) and the probability \((1-\pi )\) of not adopting it (or of not have a mobile account) represents the odds i.e. a ratio of chances. For example, if an individual has an odds of 3, this means that there is three times more chance of adopting digitalization (or having a mobile account).

3 Results

3.1 Descriptive statistics of variables used

The summary of the quantitative explanatory variables by modality of the dependent variables, presented in the Table 2, showed that the total area available to the producer, that devoted to cotton and the age of the producer are slightly higher among producers who use digital technologies than among those who are not users. In the case of Use of digital machines in agriculture, for users of technologies, the total area available, that devoted to cotton and the age of the producer were on average 10.6 ha, 6.3 ha and 39 years respectively against 8.4 ha, 4.8 ha and 40 years for producers who do not use digitalization tools. When considering Membership of a social network for information exchange between producers, the total area available, that devoted to cotton and the age of the producer were on average 10.7 ha, 6.3 ha and 38 years respectively against 7.5 ha, 4.3 ha and 42 years for non-user producers. For Use of a Mobile Money account, the total available area was 9.2 ha, that devoted to cotton was 5.4 ha and the age of the producer was 38 years on average, against 8.2 ha, 4.7 ha and 43 years respectively for non-user producers. A relatively high variability was noted, which could exceed 100%, at the level of the observations relating to each variable studied.

Table 2 Statistics parameters (mean, standard error, minimum, maximum and coefficient of variation) of each quantitative explanatory variable according to the modalities of the dependent variables
Full size table

The distribution of the qualitative explanatory variables by modality of the dependent variables (Table 3) showed that the majority of the producers surveyed were uneducated men of the Islamic religion who cultivate cotton conventionally and do not use labor, at least for non-users of digital technologies in the latter case. They were also many to have a second activity, to not have a bank account and to find the Internet and digitalization very useful, as well as the improvement of social relations that these tools can induce. The majority of producers also thought that the use of mobile money is satisfactory and advantageous. In terms of security, risk, perception and reason for using digitalization, producers who are not users of digital technology are many to not express their opinion. On the other hand, producers who use the technology think that security is very good, have a good perception and use it for profit.

Table 3 Distribution of producers for each qualitative explanatory variable according to the modalities of the dependent variables
Full size table

3.2 Factors influencing the use of digital machines in agriculture

The full logistic regression model (Table 4) showed a significant link between a number of factors and the adoption of use of digital machines in agriculture by cotton producers in the municipality of Banikoara. The large number of explanatory variables can hide the significant effect of a certain number of variables. This fact can come from the collinearity between the variables. To a certain extent, the selection of variables allows to remedy this fact. The backward selection method made it possible to retain the model presented in Table 4. This model represents the one which provided the lowest value of the AIC criterion. In addition, all the variables present in the model are significant. Thus, the backward selection method led to the selection of two variables in the category of variables linked to production activity (use of labor and the area sown for cotton). In the category of variables linked to income, only the holding of a bank account was selected. Knowledge of the information and exchange platform at the agricultural sector level in Benin is the only variable that was selected in the category of variables linked to social influence. Examination of the odds ratios showed that cotton producers using labor for production have a 3.48 greater chance of adopting digital technology in cotton production. Having an account in a commercial bank or in a decentralized financial system (microfinance) gives the producer gives the latter 8.24 more chances of adopting digital technologies in cotton production than those who do not have one. The coefficient − 0.15 of the quantitative variable “area sown with cotton” indicates that, all things being equal, the logarithm of the probability π/(1-π) decreases by 0.15 for each additional unit of area. Thus, increasing the surface area of a unit implies that the probability π that a producer adopts digital technology is exp(− 0.15) = 0.86 times the probability (1-π) that a producer does not adopt it. Otherwise, the more the surface area increases (because the sign of the coefficient is negative, and therefore the odds ratio is less than 1). The value of 0.86 is however close to 1. This could suggest the indifference of the producers with respect to this variable. This result may also be linked either to a sampling error or to the reality in the study area. Likewise, the lack of knowledge of the information and exchange platform at the level of the agricultural sector gives the producer less chance of adopting digital agricultural machines, i.e., 0.55 times the probability of a producer who have knowledge of the platform, because the sign of the coefficient is negative (− 0.6). In other words, those who were aware of digital platforms and social networks used more digital agricultural machines. In relation to the UTAUT theoretical model, only social influence and exogenous variables have an effect on the adoption of digital agricultural machines.

Table 4 Parameters estimated (coefficient and p-value) of the full model and selected model of use of digital machines in agriculture by cotton producers
Full size table

3.3 Factors influencing membership of a social network for information exchange between producers

The backward search procedure from the complete model made it possible to select the following variables: holding a bank account, strengthening relationships with others, risk linked to the use of digitalization, security linked to the use of digitalization and type of cotton used (Table 5). Holding an account in a banking institution and when both types of cotton were practiced presented coefficients with negative signs. This indicates that these variables were less likely to influence adoption as indicated by their odds ratios, which were all less than one. Examining the odds ratios of variables whose coefficients have positive signs showed that these variables were more chance to adopt digitalization. Thus, to the question related to the influence of digitalization on social relations, producers who answered no were more chance to adopt digitalization than those who did not give an answer. The chance was higher among those who gave a negative answer (around 28 time’s chance). Compared to the assessment of the risk linked to the use of digitalization, producers who think that the risk is medium were more chance to adopt digitalization (i.e. 234.34 times) than those who think that there is no risk. They were followed by those who think the risk is high with 49.40 times the chance of adoption and low risk with 1.55 times. As for the security presented by digitalization, producers who think it is very good were more likely to adopt it than those who gave no answer with 5.56 times the chance of the latter. Thus, producers who attributed a medium or high level of risk or a security failure to digital platforms were therefore more likely to adopt digitalization in the cotton sector than those who thought that the risk is low and that the platforms provide security. Furthermore, producers who practice conventional cultivation were more chance to adopt digitalization than those who practice organic cultivation or both types of cultivation with 19.83 more chances. Social influence, facilitating conditions and exogenous variables were the determinants of the unified theory of acceptance and use of technology that influenced the acceptance of belonging to a social network by cotton producers in the municipality.

Table 5 Parameters estimated (coefficient and p-value) of the full model and selected model of use of membership in a social network by cotton producers
Full size table

3.4 Factors influencing use of a mobile money account

Analysis of the results of the full model showed that no variable was significant (Table 6). The backward search procedure from the full model made it possible to select the following variables: age, holding a bank account, carrying out a secondary activity and using a hand of labor (Table 6). These variables, all significant at the 5% threshold in the selected model, represent the factors that influence the adoption of mobile money among cotton farmers in the municipality of Banikoara. The analysis of odds ratios showed that an increase of one unit in age implied that the probability π that a producer adopts digital technology was 1.03 times the probability (1-π) that a producer doesn’t adopt it. This means practically the same probabilities. Thus, the adoption of mobile money was independent of age, because the sign of the coefficient was positive and approximately equal to 1. Holding an account in a banking or microfinance institution was associated with a reduction in the risk of adoption of mobile money (because the sign of the coefficient was negative, so the odds were less than 1). Likewise, the engagement in a secondary activity and the use of labor were associated with a reduction in the risk of adoption, because the signs of the coefficients of these two variables were also negative in the selected model, and consequently their odds were less than 1. Having an account in a bank or a microfinance institution therefore hinders the adoption of a mobile money account. Producers with a secondary activity and/or using labor are more likely to prefer using banks over mobile money accounts in order to seek support from banking institutions, which is not possible with a mobile money account. The acceptance of using mobile money by cotton farmers in the municipality was influenced by the exogenous variables of the UTAUT theoretical model.

Table 6 Parameters estimated (coefficient and p-value) of the full model and selected model of use of mobile money by cotton producers
Full size table

4 Discussion

The use of new technologies in agriculture in Benin in the early 2000s or even 2010s has considerably changed habits in the farming community. Indeed, before this period, although some producers had tractors (but very rare), most agricultural work was done with rudimentary tools such as hoe and animal traction. As for communications between the different actors in the agricultural sector, they were done by community radios or by town criers or by word of mouth. The advent of digital technologies has the potential to enhance the efficiency of input usage, increase crop productivity, and reduce environmental harm, thereby benefiting both farmers and consumers [36, 37]. The best hope for achieving sustainable agricultural development lies in the innovative digital technologies to enhance agricultural productivity while balancing economic, environmental, and social outcomes associated with agricultural systems [38]. However, the adoption of digital tools by farmers, especially smallholder farmers in developing countries, is slow and low [2]. This raises concerns related to digital divides between large and small farms, as well as between farmers in industrialized and developing countries [39]. In this study the case of cotton producers in Benin was examined.

The overview of the literature review used in this research on the adoption of digitalization focused on the evolution of online services since its emergence, and served as a fundamental basis for proposing the research model. Also, the continued growth of research based on UTAUT, as the model chosen in this study, is essentially due to the proliferation and diffusion of new information technologies [23, 40, 41] and the consideration of moderating variables such as gender, age, experience and willingness to use technology in this model. The factors for adoption of digitalization in cotton production were grouped within the framework of this study into factors linked to production activities, sociological aspects, income, appreciation of digitalization and social influence as shown in Fig. 2. The analysis of digitalization is carried out through the use of agricultural machines and other digital technological tools including mobile phones and tablets, the use of agricultural platforms and social networks and the possession of a mobile money account.

4.1 Use of machines and other digital technological tools

The adoption of machines and other digital technological tools was influenced by two factors (labor and area of cotton sown) linked to production activities, a factor linked to income (possession of a bank account) and a factor linked to social influence. The use of labor (odd ratio equal to 3.48) and the possession of an account in a bank or in a decentralized financial system (microfinance institution) with an odds ratio equal to 8.24 had a positive effect on adoption. This is explained by the fact that banked cotton producers have an easier time obtaining loans from these financial institutions for investment in the acquisition and use of agricultural machinery and other digital technological tools. Although the use of machines and other digital technological tools leads to a reduction or disappearance of the use of labor, it is nevertheless important to have a qualified workforce for their use. This explains the positive influence of the use of labor on the adoption of digitalization. Thus, bank loans are used not only in the acquisition of machinery and other digital tools but also for the maintenance or recruitment of labor, hence the positive influence of these two factors on the adoption of machines and other digital tools. Conversely, difficulties in accessing credit could therefore impact the decision to adopt new technologies [18, 42, 43]. In general, most farmers face liquidity constraints during non-harvest periods [44] and their access to credit would reinforce the use of certain inputs [16, 45] and certain digital technologies for future production. According to Alene and Manyong [16], the size of the household, which also expresses the level of available family labor, could affect the decision to adopt a new technology in agriculture. Contrary to the results of Teno et al. [18] who stated that a labor-intensive technology will certainly be more within the reach of large families who, as a result, will be more favorable to its adoption; which would not be the case for smaller families, the variable number of dependents did not have a significant effect on adoption in the present study.

The results also showed that a marginal increase in the sown area negatively impacts the risk of using machines and other digital technological equipment. This could be linked to the cost that the purchase of these machines would generate. The additional increase in the sown area requires the use of more complex and very expensive machines over very large areas (drones, robots, etc.); which is not yet a reality in the study area. Any marginal increase in the sown area therefore does not lead to the adoption of digital machines. Anderson et al. [46] also obtained a negative effect of increasing farm size on the adoption of organic farming. However, Carrer et al. [47] explained that large farms are more complex to manage and that new technologies have demonstrated their effectiveness in optimizing production and reducing costs. Similarly, Yatribi [48] showed that farm size is often associated with farm income and that large farms are likely to adopt new precision agriculture technologies thanks to their financial capacity.

Also, cotton farmers in the municipality of Banikoara who were not aware of agricultural information and exchange platforms and who therefore do not use them were less likely to adopt the use of agricultural machines and tools. Indeed, they were under-informed about the functioning and usefulness of these machines and digital tools and find, in turn, less interest in their use; which is not the case for those who know about it through agricultural exchange platforms.

As part of this research, sociological variables such as age, gender, religion, marital status, level of education and the variable linked to the type of cotton produced did not have a significant influence on the use of agricultural machines and other digital tools. These results are consistent with those of Oulbaz et al. [49] who highlighted that the use of digital technologies in agricultural projects is not influenced by internal factors within the farm, whether the type of agricultural product marketed, the educational level, habits in the use of decision-making tools and the cost of digital technology. Knowler and Bradshaw [50] also did not find significant relationships between education and adoption. However, Yatribi [47] affirmed that the level of education is highlighted by many authors as a determinant of the adoption of new technologies [46, 51,52,53,54]. Roussy et al. [14] also showed through their study that age has an influence on the adoption of a new technology by farmers.

4.2 Use of agricultural platforms and social networks

In the literature, studies have demonstrated that a person can believe that the use of a certain technology will influence their professional image and status [22, 55]. In the present study, the use of agricultural platforms and social networks is influenced by the strengthening of relationships with others, the level of risk and security. Indeed, whatever the modality of the security level factor and that of the risk perceived by the producer, he is more likely to use platforms and social networks. The cotton farmers surveyed are therefore partly risk-adverse (37% producers opted for a medium and high level of risk). Thus, it is the producers who think about the existence of a risk who are more likely to use professional platforms. This could be explained by the fact it is quite difficult for the cotton farmer to understand the risk at the level of digital platforms and the majority of individuals interviewed (63% of the sample, or 197 individuals) have no idea about the degree of risk or think that the risk does not exist or is rather low. These results partly corroborate with those of Teno et al. [18] who believe that the degree of risk aversion would play a role in the decision to adopt a new technology. For these authors, risk-loving agricultural households will be more willing to accept the new technology unlike risk-phobic households [56]. But, Roussy et al. [14] showed that the level of risk aversion has been highlighted as a barrier to the adoption of innovations in agriculture [57, 58]. However, risk aversion alone cannot explain the behavior of farmers adopting innovations [59].

Conventional cotton producers also have more chance to use the platforms than those combining both types of cotton (conventional and organic). This category of cotton farmers is the most numerous in the sample and the oldest in the sector. They know each other better and have an easier time exchanging experiences through a platform.

Having a bank account makes it less chance to use agricultural exchange platforms. Banked cotton farmer’s exchange more with their bankers whom they prefer to contact physically instead of using a platform to communicate with the agricultural advisor. For the sale of their secondary crops (crops other than cotton) in order to repay their loans, they sometimes get help from the banker in finding customers rather than using a platform. Abid et al. [60], however, reported that farmers' resistance to digital platforms differs depending on the area targeted by this platform. According to these authors, perceived usefulness as well as perceived intrusion are the main factors of farmers' resistance to the adoption of resource platforms.

4.3 Having a mobile money account

Possession of a mobile money account by the cotton producer was influenced by age, possession of a bank account, secondary activity and use of labor. If age has practically no influence on having a mobile money account (odd ratio approximately equal to 1), this is not the case for having a bank account, carrying out a secondary activity and the use of labor which negatively influences the adoption of mobile money. In fact, cotton farmers with a bank account prefer to store their income in these accounts rather than using the mobile money account. To meet certain costs such as labor remuneration, cotton producers are sometimes forced to resort to banks and microfinance institutions to obtain loans, to the extent that they cannot access to this financing at the level of electronic money issuers. In addition, most cotton producers hold accounts with microfinance institutions. The latter do not offer digital solutions interfaced with “mobile money” (mobile banking) accounts unlike some traditional banks. It should also be noted that traditional banks have just set up in the study area (barely 3 years of activity for the first) and do not attract as many cotton farmers as microfinance institutions. Thus, cotton farmers prefer to carry out transactions from their bank account which also offers money transfer and other services. Furthermore, payment for cotton production is made through accounts opened by village cooperatives of cotton producers in these financial institutions, notably the “Caisses Locales de Crédits Agricoles Mutuels” (CLCAM). To avoid additional account management costs, cotton farmers do not opt to hold two accounts (bank and mobile money). This result corroborates with those of Akinyemi and Mushunje [61] who conclude, among other things, that it is unlikely that individuals who own or have access to bank accounts will adopt mobile money to send or receive payments. According to these authors, the reasons why mobile money is not adopted may be due to the fact that the services rendered by mobile money are also provided by commercial banks and the use of mobile money may result in duplication of services and costs. Fanta et al. [62] established that having a bank account, access to ATMs, mobile banking and internet banking are inversely related to having a mobile money account. These authors also found that mobile money adoption is lower among those who have a bank account as well as those who use ATMs, mobile banking and internet banking to access their bank account. Akinyemi and Mushunje [61] showed that apart from having a bank account, age, years of education, unemployment and mobile phone are the main determinants of the adoption of mobile money technology.

The results obtained also go in the same direction as those of Ndiaye and Weibigue [63] who showed, among other things, that the variables: having a job, age, gender and currently attending school do not seem to play a determining role in the adoption of M-Banking in Senegal and that the marginal effects of these variables are not significant. Contrary to the results obtained in the present study, these authors also reported that belonging to a banking network, that is to say being a customer of a bank, a microfinance institution or a postal check center promotes the adoption of M-banking. Fall and Birba [64] and Mbiti and Weil [65] also go in the same direction by mentioning that M-banking is a complementary service to traditional banking services. Other factors influencing the adoption of mobile money have been mentioned in the literature. Thus, Fall and Birba [64] found that gender, level of education, employment, knowing how to read and write and being banked positively influence the probability of adoption of mobile banking. Amegnanglo and Zounmenou [66] also showed that age, gender, turnover and education are the determinants of the use of electronic money account services by artisans in southern Benin. For Bidiasse and Mvogo [67], generally speaking, the advantages offered, the information available on how mobile money works and the proximity of the service are the variables for adoption of this service.

5 Conclusion

The purpose of this research was to analyze the determinants of the adoption of digitalization among cotton farmers in the municipality of Banikoara, a cotton basin in Benin. As part of this study, digitalization was subdivided into three (3) components, namely the use of machines and other digital tools (including mobile phones and tablets), the use of agricultural platforms and social networks, and possession of a mobile money account. The adoption of digitalization in these three (3) forms is influenced by holding an account in a bank or in a decentralized financial system (microfinance institution). This variable positively influences the use of machines and other digital tools and then negatively influences the use of platforms and the possession of a mobile money account. In addition to this factor, the use of labor, the assessment of risk and the level of safety, the exercise of a secondary activity, knowledge of agricultural information and exchange platforms and the area sown also have a significant influence either on the use of machines and digital tools, or on the use of platforms or the possession of a mobile money account. Contrary to the results of several previous studies, sociological factors, namely gender, age, level of education and marital status, did not have significant effects on the adoption of digitalization by cotton farmers in the municipality of Banikoara within the sample. Overall starting from the structured analysis model, the factors linked to production activity (surface area, workforce), income (possession of a bank account or in a decentralized financial system), the assessment of digitalization (perception of risk and security level) and social influence (knowledge of agricultural information and exchange platforms) have a significant impact on the adoption of digitalization in cotton farming in the municipality of Banikoara. Thus, any action in favor of promoting digitalization among this social layer essentially amounts to promoting their banking use while integrating and generalizing mobile banking services at the level of microfinance institutions and commercial banks, which are in contact with the producers.

Awareness and popularization sessions organized by the leaders of farmers' organizations and technical support for cotton producers on the use of digital tools, platforms and social networks (socio-professional WhatsApp group) and mobile money will undoubtedly contribute to the adoption of digitalization in cotton farming. The political and administrative authorities, by requiring each producer to open a bank account and the transfer of cotton funds to these accounts by the structure that manages the cotton sector, will undoubtedly promote the adoption and use of digitalization in the cotton sector in Benin. The promotion of digital agricultural machines (drones, tablets, etc.) and the subsidy on their purchase cost by the public authorities, the relaxation of the conditions for opening an account for this segment of the population are all decisions whose implementation will contribute to the adoption of digitalization. However, the adoption decision may also depend on exogenous factors (regulatory constraints for example) which were not taken into account in the study. The method of selection of cotton farmers (reasoned choice) can also influence the results obtained, which cannot be generalized. The size of the sample, the random method of choosing individuals, and the choice of a single production area will undoubtedly impact the results. Therefore, further research work is necessary to deepen and supplement the results obtained.