1 Introduction

Small-scale farming in developing countries is highly vulnerable to weather variability, droughts, and unpredictable precipitation, influencing agricultural production and many households’ welfare (Olagunju et al. 2020; Dhanya et al. 2022). To address the climate change-induced challenges, policymakers and advisory communities are promoting various technologies such as climate-smart agricultural (CSA) technologies. CSA technologies include a variety of technologies, practices, and services that have the potential to reduce the effects of climate change on agricultural systems (Khatri-Chhetri et al. 2017). They include water-smart technologies (e.g., rainwater harvesting), energy-smart technologies (e.g., zero tillage), nutrient-smart technologies (e.g., intercropping with legumes), weather-smart technologies (e.g., climate-smart housing for livestock), and knowledge-smart technologies (e.g., improved crop varieties) (Khatri-Chhetri et al. 2017; Jones et al. 2023; Tadesse and Ahmed 2023; Li et al. 2024; Vishnoi and Kumar 2024).

CSA technologies are linked to many outcomes, including enhancing farm productivity (Mossie 2022; Balasundram et al. 2023), boosting rural incomes (Dhaoui et al. 2020; Khoza et al. 2021; Awotide et al. 2022), alleviating poverty (Azzarri and Signorelli 2020; Kilombele et al. 2023), enhancing soil fertility (Ogieriakhi and Woodward 2022), and mitigating the negative environmental impacts of agriculture (Food and Agriculture Organization [FAO] 2018). Several empirical studies have analyzed the effects of CSA technologies on various economic and environmental outcomes. For example, Tabe-Ojong et al. (2023) found that CSA adoption increases crop yields and food security for farmers in Ghana, Mali, and Nigeria. Tadesse and Ahmed (2023) found that CSA adoption improves soil fertility and dietary diversity in Ethiopia. Similarly, the use of drought-tolerant seeds has been found to enhance maize productivity (Simtowe et al. 2019; Olagunju et al. 2020), increase farm income (Fentie and Beyene 2019; Abebe et al. 2023), and improve food security (Khoza et al. 2021; Awotide et al. 2022). Adopting row planting and zero tillage technologies has increased farm output and income (Fentie and Beyene 2019; Tamirat 2020; Mossie 2022) and lowered global warming in Africa (Keil et al. 2020).

While significant progress has been made in Sub-Saharan Africa (SSA) towards the adoption of CSA technologies such as drought-tolerant seeds, row planting, and zero tillage, there is little information on how CSA technologies should be promoted and disseminated in African countries such as Ghana (Ehiakpor et al. 2021; Balasundram et al. 2023). One of the impediments to the effective dissemination and adoption of CSA technologies is a lack of communication infrastructure. Only 30% of the SSA population is estimated to be active users of information and communication technologies (ICTs) (Goedde et al. 2021; Onyeneke et al. 2023). This low ICT usage is attributed to a lack of institutional support, communication infrastructure development, high technology costs, and a lack of digital skills (Cariolle 2021; Klerkx et al. 2019; Li et al. 2021). Nonetheless, the growth of mobile phones in the last decade has been a positive story in ICT, particularly in SSA (Adenubi et al. 2021).

Leveraging the growth of mobile phones and other ICTs, research and policy on digital advisory services (DAS) are gaining traction in SSA to reduce information asymmetry and provide climate-smart information to smallholder farmers (Adenubi et al. 2021; Asongu et al. 2019; Tchamyou et al. 2018). DASs are tools and platforms that integrate climate information into agricultural decision-making processes, including mobile applications, radios, online platforms, and extension services to promote knowledge exchange and aid farmers in adopting CSA technologies (FAO 2023). Thus, encouraging farmers to adopt new agricultural technologies, including CSA, through DAS remains a top priority for the government and non-governmental organizations in many developing countries, including Ghana (Khoza et al. 2021; Ngigi and Muange 2022). For example, policies promoting DAS include the IFAD’s Rural Poverty Stimulus Facility, which provided personalized agricultural advice to 1.7 million small-scale producers in Kenya, Nigeria, and Pakistan, including women and youth, to improve productivity, profitability, and environmental sustainability during COVID-19. Several studies have shown that adopting information technologies can improve crop performance (Ogutu et al. 2014; Khan et al. 2022), increase farm income (Ma et al. 2020; Khan et al. 2022), and aid rural advancement (Ma et al. 2022; Spielman et al. 2021). Mobile phones and emails have also improved peasant crop production (Otter and Theuvsen 2014; Deng et al. 2019; Spielman et al. 2021). Implementing ICT-based market information systems has also increased resource utilization (Ogutu et al. 2014).

However, empirical evidence linking DAS use and CSA technology adoption by smallholder farmers is scarce. Only a few studies have focused on this area; these include Singh et al. (2019, who found that Agro-Advisory Services, a type of DAS, aided farmers’ adoption of climate-smart practices in India, whereas Amith et al. (2022) found that Agromet Advisory Services aided farmers’ adoption of climate-smart practices in India. Aside from the studies in India, there is a lack of studies in other countries, particularly Ghana, that assess the linkage between DSA and CSA.

This study, therefore, aims to contribute to the literature by (i) investigating the impacts of DAS use on CSA technology adoption in Ghana and (ii) estimating the disaggregated impacts of DAS use, taking gender and geographical location into consideration. The research specifically contributes to the literature in three ways. First, it examines the impact of DAS use on CSA technology adoption in Ghana, the first to do so. We use a recursive bivariate probit (RBP) model to address the self-selection bias issues when farmers choose to use DAS. Second, we investigate the average treatment effect of DAS use on CSA technology adoption and estimate the disaggregated impacts of DAS use, taking gender and geographical location into account. As shown in the previous studies (Taylor and Silver 2019; Leng et al. 2020; Nikam et al. 2022), gender gaps and geographical differences exist in the use of DAS and are likely to impact the adoption of CSA technology adoption differently. Given the significant differences in demographic and institutional conditions and information technology infrastructure across Ghanaian regions, we hypothesize that these differences will influence DAS use and CSA technology adoption. The findings of this study can provide decision-makers and policymakers with information and insights into how farmers in various regions use DAS to adopt CSA technologies. It can inform the development of strategies to improve farmers’ use of DAS, encouraging the adoption of CSA technologies and ensuring food security.

We focus on maize production because it is one of the most important cereal crops globally, alongside rice and wheat (Ranum et al. 2014; Pauw 2022; Ankrah et al. 2023). In Ghana, maize is important as a staple food, being widely consumed across various regions (Ankrah et al. 2023; Prah et al. 2023). Maize cultivation covers over 14% of the total cultivated land in the country, playing a vital role in improving rural livelihoods (Pauw 2022; Ankrah et al. 2023). In 2021, the country achieved a record high production of 3.5 million metric tons, marking the highest output since 2010 (Ankrah et al. 2023; Prah et al. 2023). Ghana’s maize yield is among the lowest globally, with estimates ranging from 1.2 to 1.8 tonnes per hectare (Asante et al. 2019; Obour et al. 2022; Prah et al. 2023). However, it is estimated that Ghana could yield 4 to 6 tonnes per hectare (MoFA 2015; Wongnaa et al. 2019; Obour et al. 2022). Adopting improved practices and techniques, such as CSA technologies (drought-tolerant seeds, row planting, and zero tillage) via DAS, could contribute to such yield targets.

This paper is organized in the following ways: Conceptual framework is the next section. Section 3 presents the methodology comprised of the study sites, data and descriptive statistics, and analytical strategy. Section 4 presents empirical results and discussion, whereas Section 5 encompasses the conclusion and policy implications.

2 Conceptual framework

Figure 1 illustrates the relationship between the use of DAS and CSA technologies, highlighting how farmers’ decisions can conceptually influence the adoption of CSA technologies. DAS use can reduce the information asymmetry associated with CSA technologies and encourage farmers to adopt climate-smart practices and technologies (Amith et al. 2022; Fernando 2021; Kumar et al. 2022). Given the vulnerability of farming systems to climate change, such as changes in temperature and rainfall patterns, as well as the occurrence of droughts, farmers frequently seek advice to mitigate these risks, as these factors can have a negative impact on crop yields (Makate et al. 2019a; Antwi-Agyei and Stringer 2021). This shapes farmers’ information needs and information-seeking behavior, motivating them to seek advisory services to implement certain climate-smart technologies (Amadu et al. 2020; Khoza et al. 2021). Such information is commonly facilitated through traditional agricultural extension agencies and diverse information technologies, including television, radio, smartphones, computers, and the internet (Ma et al. 2020; Ngigi and Muange 2022).

Fig. 1
figure 1

 Conceptual framework

Full size image

DAS can significantly increase CSA technology adoption by providing critical information, fostering social capital, and promoting stakeholder communication, thereby improving learning opportunities and the adoption process (Eakin et al. 2015; Aldosari et al. 2019). Thus, access to appropriate advisory services, enabled by smartphones and computers, is crucial for farmers to incorporate CSA technologies and improve crop yields effectively.

Farmers’ utilization of DAS is determined by various factors, including farmers’ characteristics, farm level, institutional and location factors, and communication infrastructure (Spielman et al. 2021; Dhanya et al. 2022; Khan et al. 2022). Empirical studies indicate that farmers with higher levels of literacy, income, farming experience, family size, credit access, asset value, farm size, and membership in farmer-based groups are more likely to access farming-related information through ICTs (Raza et al. 2020; Nikam et al. 2022). Digital literacy is crucial for farmers to use ICT tools in agriculture effectively (Khan et al. 2022). Farmers who lack essential reading and writing skills may struggle to access and utilize information provided through advanced DAS, such as smartphone applications and social media (Khan et al. 2022; Leng et al. 2020; Singh et al. 2019). As shown by studies in Africa, access to information through DAS can also improve smallholder farmers’ awareness of weather and production shocks, leading to increased adoption of CSA technologies (Weniga et al. 2019; Antwi-Agyei and Stringer 2021; Kumar et al. 2022).

In addition to the factors influencing farmers’ decisions to use DAS, farmers’ socioeconomic, institutional, and location attributes can also affect their adoption of CSA technologies. Farmers’ characteristics such as age, gender, household size, plot size, income, experience, and education level have been found to influence their decision to adopt CSA technologies (Makate et al. 2019a; Simtowe et al. 2019; Weniga et al. 2019). Furthermore, institutional factors (i.e., access to credit, farmer-based groups, extension services, and road accessibility) play a crucial role in the adoption of CSA technologies (Makate et al. 2019a; Amadu et al. 2020; Khoza et al. 2021; Ma et al. 2022; Ngigi and Muange 2022). For instance, access to credit can ease the financial burden associated with adopting CSA technologies, as farmers may use the obtained credit to purchase ICTs such as smartphones and computers (Ma et al. 2022). As a network, farmer groups may facilitate known externalities, such as interactions among network members that can influence individual behavior to adopt CSA technologies (Khoza et al. 2021; Addai et al. 2021). Moreover, farmer groups may assist farmers in making informed decisions concerning crop management, technology choice, and marketing (Gangopadhyay et al. 2019; Nikam et al. 2022).

3 Methodology

3.1 Study area and sampling methods

The data used for analysis in this study were collected between August and December 2021, focusing on maize farmers in Ghana’s Brong Ahafo, Ashanti, and Northern regions. Figure 2 shows the map of the study area. The study considered three agroecological zones: Transition (Nkoranza, Ejura-Sekyeredumasi, and Kintampo South districts), Guinea Savannah (Zabzugu and East Gonja districts), and semi-deciduous forest zones (Ejisu-Juaben district). Farmers were selected using a multistage sampling technique, starting with a purposive selection of the three regions with high maize production in Ghana. Two high-producing maize districts were selected from each region, including Nkoranza and Kintampo South from Bono East, Ejisu-Juaben and Ejura-Sekyeredumasi from Ashanti, Zabzugu and East Gonja from the Northern region (MoFA 2015), and eight purposefully selected communities from each district. Between 60 and 70 maize farmers were randomly selected from each community, resulting in a sample size of 3197 maize farmers, with 2765 male-headed households and 432 female-headed households.

Fig. 2
figure 2

Map of the study area

Full size image

We employed positivism and quantitative research design. We used a structured questionnaire to gather data on farmer and farm-level characteristics, institutional and CSA technologies, production and weather shocks, and location variables in the study area. Before the formal survey, 50 maize farmers were interviewed in two selected communities, Ejura and Onwe (see Appendix Table 5). Based on the feedback from the pre-test survey, we improved the questionnaire. Enumerators fluent in both English and regional dialects were hired to assist with data collection.

3.2 Analytical strategy

The study estimates the impact of DAS use on the adoption of CSA technologies while accounting for personal and farm-level factors. DAS is not a random assignment but a self-selection case (Rajkhowa and Qaim 2021; Bonou-Zin et al. 2022). Various personal and household characteristics and socioeconomic and institutional factors influence maize farmers’ decisions. The non-randomness generates the potential endogeneity issue of the DAS use variable. Failing to address the endogeneity issue when estimating the impact of DAS use on adopting CSA technologies would generate biased estimates.

Earlier studies have suggested various methods for analyzing the impact of a binary endogenous variable (e.g., DAS use in the present study) on farmers’ binary decisions about technology adoption. These include the propensity score matching (PSM) model (Garcia Iglesias 2022; Zwane et al. 2022; Abebe et al. 2023), endogenous switching probit (ESP) model (Lokshin and Sajaia 2011; Li et al. 2020; Wu et al. 2023), and recursive bivariate probit (RBP) model (Li et al. 2019; Addai et al. 2021; Ma and Zhu 2021; Ngigi and Muange 2022). The PSM technique fails to account for endogeneity issues emanating from unobserved factors. At the same time, the ESP model cannot estimate the direct effect of DAS use on the adoption of CSA technologies. In comparison, the RBP model is an effective approach that addresses endogeneity issues from both observed and unobserved factors and can estimate a direct marginal effect of DAS use on the adoption of CSA technologies (i.e., row planting, drought-tolerant seed, and zero tillage). In addition, it is appropriate for such estimations where both treatment and outcome variables are binary. Therefore, the RBP model is employed.

Following previous studies (Thuo et al. 2014; Ma et al. 2018; Li et al. 2019; Ma and Zhu 2021), the two empirical specifications of the RBP model can be written as follows:

$$DAS\left(\varphi\right)_i^\ast=\beta_iW_i+\eta_iI_i+\varepsilon_i,\;DAS_i=\left\{\begin{array}{ccc}1&if&DAS\left(\varphi\right)_i^\ast>0\\0&if&DAS\left(\varphi\right)_i^\ast\leq0\end{array}\right.$$
(1)
$$CSA\left(\varphi\right)_i^\ast=\iota_i DAS_i+ W_i \mu_i+\varpi_i,{CSA}_i=\left\{\begin{array}{ccc}1&if&CSA\left(\varphi\right)_i^\ast>0\\0&if&CSA\left(\varphi\right)_i^\ast\leq0\end{array}\right.$$
(2)

where \({DAS\left(\varphi \right)}_{i}^{*}\) and \({CSAT\left(\varphi \right)}_{i}^{*}\) are the latent variables that ith farmer uses DAS through a mobile phone or computer and adopts the CSA technologies such as row planting, respectively. Also, the latent variables are observed by \({DAS}_{i}\) (1 if a farmer uses DAS and 0 for not using DAS) and \({CSA}_{i}\) (1 if the farmer adopts an identified CSA technology and 0 for not adopting any). \({W}_{i}\) denotes a set of explanatory factors such as socioeconomic variables (gender, age, education, marital status, household size and asset value), farm-level factors (farm size, perceived drought stress, perceived pest and disease), institutional factors (farmer-based organization and farm distance) and location variables. \({I}_{i}\) is the instrumental variable for the identification of the RBP model. The parameters to be estimated are \(\beta_i,\;\eta_i,\;\iota_i,\;\mathrm{and}\;\mu_i\;\;\)\(\;\varepsilon_i\;\mathrm{and}\;\varpi_i\) are the disturbance terms. Our explanatory variables are selected based on the theoretical and empirical literature of previous studies (Makate et al. 2019b; Weniga et al. 2019; Oyetunde Usman et al. 2020; Addai et al. 2021; Awotide et al. 2022; Damota et al. 2022; Mossie 2022).

In Eq. (1), we used the perceived high costs of DAS in the community as an instrumental variable (IV). The employed IV is measured as a dummy, which equals 1 if farmers perceive DAS in the community as a high cost and 0 for those who perceive it as a low cost. We expect that the IV influences the farmers’ decisions to use DAS directly; however, we do not expect the adoption of CSA technologies. Following Ma and Zhu (2021), we estimated individual probit models for Eqs. (1) and (2) and verified that the employed IV was statistically significant only in the DAS use equation and not in the CAS adoption equation.

We estimated Eqs. (1) and (2) simultaneously using the full information maximum likelihood estimator (FIMLE). This estimation procedure generates a correlation term between the two disturbance terms, \(\rho_{\varepsilon\varpi}=\mathrm{corr}\left(\varepsilon_i,\varpi_i\right)\). Based on Ma and Zhu (2021), the DAS variable is endogenous when the coefficient of \({\rho }_{\varepsilon \varpi }\) is statistically significant. The significance of \({\rho }_{\varepsilon \varpi }\) also suggests that farmers’ decisions to use of DAS and their decisions to adopt CSA technologies are simultaneously affected by the same unobserved factors (e.g., innate ability, motivations and aspirations) captured by the error terms.

Subsequently, we estimate the average treatment effect on the treated (ATT) to illustrate further how using DAS influences the adoption of CSA technologies (i.e., row planting, drought-tolerant seeds, and zero tillage). We specify the ATT as follows:

$$ATT=\frac{1}{{N}_{{DAS}_{i}}}\sum _{i=1}^{{N}_{{DAS}_{i}}}\left[{Pr}\left({H}_{ik}=1\right)\left|{DAS}_{i}=1\right)-Pr({H}_{ik}=0|{DAS}_{i}=1)\right]$$
(5)

where \({N}_{{DAS}_{i}}\) is the treated sample size. \({Pr}\left({H}_{ik}=1\right)\left|{DAS}_{i}=1\right)\) is the predicted CSA adoption probability for CSA technologies users in an observed context, and \(Pr({H}_{ik}=0|{DAS}_{i}=1)\) is the predicted probability that a farmer uses a CSA technology in a counterfactual context. Furthermore, the disaggregated impacts were obtained through a post-estimation from the RBP model.

4 Results and discussion

4.1 Descriptive results

Table 1 presents the measurements and summary statistics of the variables used in the analysis. It can be observed that 64% of the farmers in our sample used DAS. Adoption rates of drought-tolerant seeds, zero tillage, and row planting were 61.3%, 63.9%, and 58.3%, respectively. The average age of farmers was 47.49 years, with most of them being males (86.5%). On average, farmers spent 8.91 years of schooling. About 86.6% of farmers were married, while 55% belonged to farmer-based organizations (FBOs). The average family size was 6.67. Farmers cultivated less than 4.87 acres of land on average, and the distance from residence to the nearest farm was 6.74 km. The average asset value was 4963.78 Ghanaian cedi (GHS). Furthermore, 48.5% of the farmers perceived drought stress, while 56.2% perceived pest and disease occurrence during maize production.

Table 1 Definitions and descriptive statistics of variables
Full size table

Table 2 presents the mean differences in the observed characteristics between DAS users and non-users. There is a statistically significant difference between the two groups. Compared to non-users, DAS users were more likely to adopt drought-tolerant seeds, zero tillage, and row planting. The average age of DAS users was 46.18 years, significantly lower than the 48.23 years of non-users. Most non-users of DAS (89.3%) were males, and 86.7% of DAS users were married. Compared to non-users, DAS users operated on a smaller farmland acreage. Regarding education, DAS users have 0.318 years more of schooling than non-users. The difference in asset value between the two groups was statistically significant, with DAS users having a higher asset value than non-users. Drought stress was perceived by less than half of DAS users (49.6%), which is insignificant when compared to non-users (44.6%). However, non-users (71.7%) perceived higher pest and disease incidence than DAS users (47.5%). Most DAS users (59.8%) belonged to farmer-based organizations and perceived DAS in their communities to be expensive compared to non-users. According to the regional dummies, most farmers were from Brong Ahafo and Ashanti and primarily used the DAS.

Table 2 Mean difference in the selected variables between DAS users and non-users
Full size table

4.2 Empirical results

Table 3 shows the determinants of DAS use and CSA technology adoption, estimated using the RBP model. The significance of \({\rho }_{\epsilon \varpi }\) presented in the lower parts of Table 3 verify the appropriateness of using the RBP model. Because the estimation of the coefficients in the RBP model (see Table 6 in the Appendix for reference) is not straightforward in interpretation, we calculate and present the marginal effects results in Table 3 to improve our understanding. In the next section, we first discuss the determinants of DAS use and CSA technology adoption. Finally, we explore disaggregated results regarding the impact of DAS use on CSA technology adoption by gender and location.

Table 3 Marginal effects of DAS use and control variables on the adoption of CSA technologies
Full size table

4.2.1 Determinants of DAS use

Columns 2, 4, and 6 of Table 3 present the results reporting the factors influencing farmers’ decisions to use DAS. The age of the farmers has a negative and significant effect on the likelihood of using DAS. The marginal effects estimate suggests that a 1-year increase in age would reduce the probability of using DAS by 6.9–9.1%. Compared with their younger counterparts, older farmers are more conservative regarding the adoption of innovative technologies such as digital services. This is consistent with the findings of Onyeneke et al. (2023). Education significantly increases the likelihood of using DAS by 4.1–4.5%. Education improves farmers’ understanding of the benefits of new technologies such as DAS, motivating them to adopt it. This is consistent with the findings of Ma and Zhu (2021), who found a positive relationship between education and internet use in China.

The size of the household had a positive and significant impact on the likelihood of using DAS. The estimates show that an extra increase in household members would increase the probability of adopting DAS by 7.4–8.6%. A larger household also means a richer labor endowment and better income gains, allowing more members to be exposed to smartphone or computer use than small households. Besides, larger households have more diverse information needs and can benefit from the tailored advice provided by DAS (Ainembabazi and Mugisha 2014). A 1-km increase in the distance from farmers’ residences to the nearest farm reduces the probability of their DAS use by 3.2–4.2%. This is because longer distances are associated with higher transaction costs, making them less likely to use DAS (Fryer Jr and Levitt 2004; Fernando 2021). Farmers’ asset value positively and significantly influences their decision to use DAS. Assets serve as a proxy for resource endowment and wealth. This finding is consistent with Meier zu Selhausen (2016), who discovered that farmers with more resources, such as land and livestock, could easily convert them into cash to obtain DAS. This is typical in rural areas of African countries, where assets serve as a means to an end in the event of production failures (Addai et al. 2021). This finding is consistent with the findings of Wossen et al. (2017), who claimed that assets allow farmers to adopt new agricultural technologies.

Furthermore, the results suggest that farmers who perceive the occurrence of pests and diseases are more likely to use DAS. These results are similar to the findings of Teklewold et al. (2013), who found that the presence of pests and diseases increases the adoption of agricultural technologies in Ethiopia. Furthermore, weather shocks such as drought stress increase the likelihood of utilizing DAS. Specifically, farmers who experience drought stress during maize production are more likely to seek advisory services. This supports previous research that droughts positively influence the adoption of agricultural technologies (Teklewold et al. 2013; Wainaina et al. 2016; Makate et al. 2019a; Jha et al. 2020). Being a member of a farmer-based organization has a significant and positive influence on the likelihood of using DAS. FBOs, which are regarded as crucial institutional advancements, have the potential to alleviate the constraints that prevent smallholder farmers from accessing novel agricultural technologies (Ma and Abdulai 2016; Zhou et al. 2023). Compared to the base Northern region, farmers in Ghana’s Brong Ahafo and Ashanti regions have a higher tendency to utilize DAS. However, the perceived high cost associated with such services negatively and significantly affects their utilization. This suggests that when DAS is more affordable and accessible, farmers are more likely to use it. Similar results have been observed in Nigeria (Wossen et al. 2017) and India (Rajkhowa and Qaim 2021).

4.2.2 Determinants of CSA technology adoption

The estimated impacts of DAS use and control variables on the adoption of CSA technology are presented in columns 3, 5, and 7 of Table 3. DAS use positively impacts farmers’ decisions to adopt all three CSA technologies. Specifically, using DAS increases the probabilities of adopting row planting, zero tillage, and drought-tolerant seeds by 12.4%, 4.2%, and 4.6%, respectively. This confirms the findings of Singh et al. (2019) and Jha et al. (2020). Furthermore, because of the ease of access to information and resources through the internet, people in the digital era are more aware of the benefits and importance of using DAS, increasing the likelihood of CSA technology adoption. Prior studies have also highlighted the numerous benefits of using DAS for agricultural development (Amith et al. 2022; Kumar et al. 2022; Singh et al. 2019, 2020; Vrain et al. 2022). For instance, Singh et al. (2020) opined that DAS offers data-driven insights and analysis that support farmers in making more informed decisions. By utilizing DAS, farmers become aware of CAS technologies to respond proactively to climate variability and extreme weather events. Farmers can better anticipate and mitigate potential risks, thereby increasing the likelihood of successfully adopting and implementing CSA technologies. Vrain et al. (2022) indicated that DAS enhances farmers’ knowledge and skills related to CSA technologies. It allows them to access expert advice and peer-to-peer learning platforms and facilitates knowledge transfer and capacity building in CSA practices.

The results also suggest that the adoption of drought-tolerant seeds, zero tillage, and row planting by farmers is positively influenced by a variety of factors, including age, education, household size, FBO membership, farm size, perceived drought stress, perceived pest and disease incidence, and location. For instance, the last column of Table 3 shows that farmers with higher levels of education are 9% more likely to adopt drought-tolerant seeds. This result is consistent with the findings of Makate et al. (2019a) and Amadu et al. (2020). Furthermore, membership in a farmer-based organization increases the probability of adopting zero tillage and drought-tolerant seeds by 10.2% and 8.6%, respectively, which is consistent with the results of previous studies (Addai et al. 2021; Manda et al. 2020; Ma et al. 2018). The farm size significantly and positively impacts the adoption of CSA technologies during maize production. Specifically, farmers cultivating larger farms are 3.5% and 3.3% more likely to adopt row planting and drought-tolerant seeds. However, the adoption of zero tillage decreases as the size of the farmland increases. This finding is consistent with the results of a study by Ma and Abdulai (2019), which suggested that the adoption of new technology involves risks that may result in productivity loss in the absence of technical assistance. As a result, the possibility of adoption decreases as farm size increases. Additionally, the cost of adopting new technologies may be a factor, as farmers may struggle to invest the necessary capital to develop new farm acres and acquire new technologies (Meena et al. 2016; Brown et al. 2019).

Moreover, the results reveal that farmers who face greater weather-related challenges, such as drought stress and pest and disease outbreaks, are more likely to adopt climate-smart technologies such as row planting, zero tillage, and drought-tolerant seeds. Specifically, farmers are 24.1% more likely to adopt drought-tolerant seeds when under severe drought stress. In addition, when there are high pest and disease outbreaks, farmers are 7.2% and 9.1% more likely to adopt drought-tolerant seeds and row planting, respectively. The distance between the farmer’s homestead and the nearest farm negatively and significantly influenced the adoption of only zero-tollage technologies. This could be attributed to the labor-intensive nature of land preparation, which could increase production costs and the likelihood of adopting zero tillage. Farmers in the Ashanti and Brong Ahafo regions are more likely to adopt row planting and drought-tolerant seeds relative to the Northern region. On the other hand, farmers in the Brong Ahafo region were found to be less likely to adopt zero tillage, possibly because they in Brong Ahafo prefer no-tillage of their lands for crop production.

The marginal effects in Table 3 only show how farmers’ decisions to use DAS affect the adoption of drought-tolerant seeds, zero tillage, and row planting if they were previously non-users but became users. We used Chiburis et al. (2012) approach to estimate the impact of DAS on the adoption of CSA technologies and gain a better understanding. The approach uses bootstrapping to reduce sampling noise in the sample, resulting in a more accurate average treatment effect on the treated (ATT) estimates. It also considers selection bias between users and non-users of DAS, as they differ significantly in observed and unobserved factors. Based on the results in Table 3, farmers who use DAS are more likely to adopt row planting, zero tillage, and drought-tolerant seeds by 38.8%, 24.9%, and 47.2%, respectively. The marginal effects, which examine the likelihood of farmers adopting these practices upon using DAS, are distinct from the ATT estimates, which measure the causal relationship between adopting and utilizing CSA technologies. Similar findings were made by Lanfranchi and Pekovic (2014) and Ma et al. (2018).

Our findings demonstrate the significant influence of DAS in the adoption of CSA technologies, specifically row planting, zero tillage, and drought-tolerant seeds. These adaptation options are highlighted in the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report (AR6) in 2023 (IPCC 2023). The report proposes that smallholder farmers can enhance their agricultural output and welfare by adopting improved cultivars, implementing on-farm water management and storage systems, conserving soil moisture, utilizing irrigation techniques, practising agroforestry, implementing community-based adaptation strategies, diversifying agriculture at the farm and landscape level, and adopting sustainable land management approaches, among other options (IPCC 2023). Hence, our study adds to the increasing body of evidence regarding the significance of these adaptation strategies in addressing climate change.

4.3 Disaggregated analyses by gender and location

Previous studies have shown that males and females have different decision-making when adopting improved agricultural technologies (Doss 2018; Paudel et al. 2020; Tambo et al. 2021). In addition, location-based spatial effects exist in technology adoption (Fang and Richards 2018; Zheng et al. 2021). Therefore, we further looked at the impact of DAS use on the adoption of CSA technologies, disaggregated by gender and location. This helps enrich our understanding of the relationship between DAS use and CSA technology adoption.

Columns 2 and 3 of Table 4  present the gendered differentials in the impact of DAS use on CSA technology adoption. Compared to previous studies (Ma and Zhu 2021; Vatsa et al. 2022), our results show a minimal difference in the marginal effect of both male and female farmers. However, these differences are significant. Hence, we need to appreciate these differences to deepen our understanding of the relationship between DAS use and CSA technology adoption. Subsequently, the results reveal that DAS use by male farmers has a greater impact on the adoption of zero tillage and drought-tolerant seeds, increasing the probabilities of their adoption by 2.5 and 3.6%, respectively. However, DAS use by female farmers has a greater impact on the adoption of row planting, with a probability value of 2.4%, against the male counterparts, with a probability of 1.5%. In Ghana, men are more involved in agricultural activities than women, which may explain why they are more likely to adopt CSA technologies after using DAS.

Table 4 Disaggregated analyses by gender and locations
Full size table

The last three columns of Table 4 show the location-based effects of DSA use on CSA technology adoption. The results show that DAS use has the largest impact on row planting adoption and tolerant seed adoption for farmers in the Brong Ahafo region, while DAS use’s impact on zero tillage adoption is the largest for farmers in the Northern region. Specifically, farmers in the Brong Ahafo region are 26.1% and 27% more likely to adopt row planting and tolerant seeds, respectively. Farmers in the Northern region are 34% more likely to adopt zero tillage.

5 Conclusions and policy implications

This paper evaluates the impact of DAS use on the adoption of CSA technologies using a random sample of 3197 maize farmers in Ghana. Since maize farmers self-select and their decision to use DAS may be influenced by both observed and unobserved variables, an RBP model was used to mitigate selection bias and to obtain unbiased estimates.

The adoption of CSA technologies (i.e., drought-tolerant seeds, zero tillage, and row planting) was significant between DAS users and non-users without accounting for other factors. The empirical RBP model results show an inverse selection bias due to unobserved factors. The results show that the main factors influencing the decision to use DAS are age, gender, education, family size, asset value, distance to farm, the perceived incidence of pests and disease, drought stress, farmer-based organization, and locations. After controlling for selection bias, the results show that DAS increases the likelihood of adopting drought-tolerant seeds, zero tillage, and row planting by 4.6, 4.2, and 12.4%, respectively. Furthermore, age, education, household size, FBO membership, farm size, perceived drought stress, perceived pest and disease incidence, and location significantly impacted the adoption of row planting, drought-tolerant seeds, and zero tillage. The average treatment effect on the treated confirms that maize farmers who use DAS are 38.8, 24.9, and 47.2% more likely to adopt row planting, zero tillage, and drought-tolerant seeds, respectively. The disaggregated estimates confirm that the impacts of DAS use on adopting row planting, drought-tolerant seeds, and zero tillage are heterogeneous between male and female maize farmers and geographical locations. Furthermore, the estimates show that the use of DAS encourages the adoption of CSA technologies among rural maize farmers in Ghana.

Our research highlights the significance of using DAS via smartphones in agriculture. Our findings suggest that agricultural stakeholders, including government and non-governmental organizations, should encourage and promote the use of DAS among maize farmers. This would contribute positively to the development of rural agriculture and improve the livelihoods of maize farmers by enabling them to access up-to-date information on maize production, as well as adopt CSA technologies such as drought-tolerant seeds, row planting, and zero tillage, all of which can enhance maize productivity. Furthermore, our study shows that the use of DAS substantially impacts the adoption of selected CSA technologies. Therefore, extension officers and farmer-based groups should encourage farmers to use DAS and help identify the challenges that impede farmers from using this service. They should also educate farmers on the benefits of using DAS as a reliable source of information on climate-smart technologies, which can help to spur adoption. The government should set up and improve existing digital hubs/infrastructure and demonstration centers in rural areas where farmers can access and experience DAS technologies firsthand.

The limitation of the study is that empirical analyses are based on 1-year cross-sectional data. This precludes us from investigating the dynamic relationships between DAS use and adoption of CSA technologies over time. Furthermore, the study does not examine the impact of DAS use on the intensity of CSA technology adoption. Little is known about the cost and revenue of DAS users in adopting CSA technologies compared to non-users. Finally, the RBP model does not give the determinants of the impact of DAS on the adoption of CSA technologies among users and non-users.