Effect of using a mobile phone on technical efficiency and productivity of climate-smart horticulture farmers in Taita-Taveta county, Kenya

Abstract

Horticulture is highly affected by climate variability. Various measures including climate-smart horticulture (CSH) practices are applied by farmers to curb the severity of climate change. Some farmers use mobile phones to access information and enhance their knowledge and skills related to CSH. However, the empirical effect of mobile phone usage on productivity of CSH farmers remains unclear. To address this, the study analyzed primary data from 403 farmers in Taita-Taveta County. Data envelopment analysis (DEA) technique was used to assess technical efficiency (TE), while a two-stage model incorporating partial factor productivity and a Tobit model was employed to examine the effect of mobile phone use on CSH productivity. Results showed that farmers’ TE scores were 24.9 %, 34.3 % and 54.3 % for green grams, tomatoes and both crops, respectively. Productivity levels obtained for green grams, tomatoes and both crops were 143.3 kg/acre, 4817.1 kg/acre and 2716 kg/acre, respectively. Mobile phone usage in CSH has the potential to enhance both TE and productivity. These findings demonstrate the need for horticulture stakeholders to develop an effective information management policy to enable delivery of credible, timely and simple CSH information to farmers.

1. Introduction

Climate change characterized by irregular rainfall, floods, rising temperatures and extreme weather events pose serious impacts on crop productivity and food quality globally [1,2]. Increasing temperatures lead to increased levels of crop pest infestations, low enzymatic activities in the soil and reduced crop duration [3]. Climate change affects agricultural productivity by changing crop suitability, inducing heat stress and reducing soil water, hence threatening food security [[4], [5], [6]].

Horticulture sector (comprising fruits, vegetables and flowers) is the second foreign exchange earner in Kenya, after tea. In 2022, the sector contributed Kshs. 152 billion in foreign exchange - equivalent to 20 % of all the country’s domestic exports [7]. Vegetable enterprises provide employment and nutrition to many smallholder farmers and traders along the value chain. However, the horticulture sector recorded negative growth in the year 2022 due to prolonged drought experienced in the country [8,9].

Many tomato and green gram farmers in Kenya are technically inefficient compared to their counterparts in other countries. For instance, while farmers in Cameroon exhibit a (TE) level of 68 %, those in Kenya have efficiency scores ranging from 39 % to 65 % [10,11]. The low efficiency level is attributed to a myriad of factors including farmers’ low education levels, low adoption of good agronomic practices and exposure to adverse changes in climatic conditions. In Taita-Taveta County, most green gram and tomato farmers sell up to 90 % of their produce. However, these crops are significantly affected by erratic climatic factors including persistent droughts, unpredictable rainfall and extreme temperatures [12]. These variabilities contribute to high prevalence of crop pests and diseases [13]. Such disturbances have potential implications on reducing productivity, hence the need for more timely and simplified information provision approaches [14].

To mitigate these impacts, the Government of Kenya has implemented climate-smart agriculture (CSA) practices to enhance efficiency and farm productivity. For example, the Kenya Climate-Smart Agriculture Project (KCSAP), financed in 2017 by the government of Kenya and the World Bank, aims to improve smallholder farmers' adaptive capacity and productivity [15]. In recent studies, there have been suggestions that CSA practices should be context-specific [[16], [17], [18]], hence a shift from general climate-smart agriculture to a more specific approach that will generate specific solutions is necessary. Recent studies have explored how market access, knowledge, social capital and technology affect upscaling of CSA practices [19]. However, none has delved into the specific role of mobile phones in enabling uptake of climate-smart horticulture (CSH) and the potential impacts on farm efficiency and productivity. This is the knowledge gap that the present study addresses.

The CSH concept is an approach that seeks to improve productivity by building resilience and adapting horticulture practices to changes in climatic conditions while reducing greenhouse gas emissions [20,21]. These approaches include environmental and agro-ecology systems conservation, adoption of proper water management techniques, appropriate integration of digital technologies, horticulture waste management and risk reduction techniques [[22], [23], [24]]. Integrating information communication and technology (ICT) in farming reduces knowledge gaps and creates awareness on best agronomic practices and prevailing market prices for agricultural products [25,26]. For example, the type of mobile phone owned by farmers influence the quality information they receive [27].

Previous studies show that some farmers use mobile phones to obtain information on agriculture and livestock while others use them to facilitate buying of inputs or selling output and receiving weather information alerts. Likewise [28,29], noted that mobile phones were useful tools for farmers in accessing real-time weather information. Precisely [29], argue that enabling access of agro-advisory services through a mobile phone helps to reduce information gaps and improve agricultural outcomes, such as increased production. This study builds on these findings by assessing the effect of using a mobile phone on productivity of CSH farmers using an econometric model.

Extant literature indicates that mobile phone use can enhance farmer productivity and stimulate agricultural development [30]. For instance, mobile phone ownership and use by maize farmers in Benin enhanced production [31]. Similarly, a study of Boro rice production in Bangladesh found that mobile phones significantly increased production [32]. These studies drew a distinction between mobile phones as a consumptive good and productive good, depending on how it was used by the farmers. Both [31,32] applied Cobb-Douglas model coupled with Poisson approach to measure the impact of mobile phone usage on agricultural productivity. These studies included mobile phone use (in the second step) as Hick’s neutral technology indicator in their estimations. While the current study applies the assertion that mobile phone use should be considered as a consumptive good or productive good when measuring its effect in agriculture, we conceptualize mobile phone use as a binary independent variable and estimate its effect on TE and productivity using partial factor productivity (PFP) coupled with the Tobit model.

Studies by Refs. [33,34] found that mobile phone use has a positive association with agricultural productivity in Uganda and Tanzania, respectively. Specifically [35], indicated that farmers who used mobile phones in maize production reported lower costs, more profits and less time investments. On the other hand [34], found that Ugandan banana farmers used mobile phones to seek information on weather forecasts, fertilizer usage, controlling various pests and diseases and markets. These studies highlight the value of mobile phone use in enhancing agricultural production. The current study builds on the aforementioned observations by examining the TE of farmers facing climate change challenges. It treats mobile phone as a production tool and assesses its influence on productivity using econometric models.

Retnaningsih et al. (2023) assessed the effect of using a mobile phone on agriculture production in Indonesia with a specific focus on rice farmers. Their study employed propensity score matching (PSM) technique where mobile phone used for agricultural purposes was the treatment variable. The current study is different in two ways: First, we sampled farmers from different agro-ecological zones engaged in horticulture production without considering their participation in any project. This is because concentrating on projects often ignores farmers’ decisions which are mostly dependent on self-innovation and information gathered from other farmers. Secondly, the PFP and Tobit models are used to determine the effect of mobile phone usage on horticulture productivity, treating TE and productivity as dependent variables.

Over the years, mobile phone technology has undergone tremendous shifts that improves its utility for farmers [22]. The development of user-friendly applications (herein referred to as app.) such as: Aeroview infield app. used to identify and track pests, disease, and irrigation problems; AGMRI app. used to identify yield suppressing weeds and nutrient deficiencies; Becrop app. that provide all the information about a crop of interest; CABI mobile phone app. that is used to diagnose crop pests and disease symptoms and offer recommendations; among other mobile apps. Other platforms include short messaging services, web portals (such as Kenya Agricultural Observatory Platform (KAOP)) and social media pages (including farmer WhatsApp and facebook groups). All these are pathways that farmers use to obtain farming information. However [36,37], noted that the use of such services and apps is influenced by gender, age, education and access to credit by farmers.

Mobile phones improve farming practices by facilitating real-time flow of information, easy access to agronomic information and cognitive assets, thus enabling farmers to respond to the challenges more swiftly and appropriately [33]. Some studies have also shown that mobile phone use enhances TE and agricultural production by increasing farmers' knowledge and awareness on farming matters [38,39]. Most of these studies applied logit, probit, PSM and Cobb-Douglas models in estimating the impact of using a mobile phone on TE and agricultural production. Logit and probit regression models are only suitable for limited dependent response variables [38] while PSM is extensively used to measure the impact of new technologies [40,41]. The application of Cobb-Douglas model limits the analysis of the effect of technology by assuming that it entirely encompasses a Hick’s neutral or an error term [32,33]. While many studies employed econometric methods for analysis, there is a notable lack of empirical research using PFP and Tobit models to evaluate the impact of mobile phone usage on the TE and productivity of horticulture farmers in the context of climate change. Furthermore, it remains unclear whether using a mobile phone in agriculture enhances TE and productivity for CSH farmers. This study aims to address this gap by exploring how mobile phone usage affects TE and productivity among CSH farmers. The study is based on the hypothesis that mobile phone usage in agriculture improves both TE and productivity for CSH farmers1

2. Methodology

2.1. Study location

The study was carried out in Taita-Taveta County (Fig. 1), because it hosted a CSA project since the year 2017. The area is also susceptible to erratic climate changes including unusually high temperatures and unpredictable rainfall [42]. The county is among the top horticulture producers in Kenya, contributing about 8 % of the national output [43]. However, it is characterized by high poverty level estimated at 57 % [42].

Fig. 1.

A map of the research sites in Taita-Taveta County.

Wundanyi, Taveta and Mwatate sub-counties were purposively chosen for the study because they have relatively higher acreage under tomato and green grams and output than the rest of the regions of Kenya [44]. The county has a land area of 843 hectares under green grams and contributes an average of 0.6 % of total output in Kenya [45]. Likewise, tomato production covers 2932 hectares of the county’s crop land and contributes, on average, 8 % of total value of tomatoes sold in Kenya [43]. About 90 % of green grams and tomatoes produced in Taita-Taveta are sold, thus providing income for other household needs.

2.2. Method of sampling and data collection

This study utilized primary survey data gathered from a random sample of 403 green gram and tomato farmers across Wundanyi, Mwatate, and Taveta sub-counties within Taita-Taveta County. These three sub-counties were chosen due to their high populations of green gram and tomato farmers, as well as the significance of these crops for both subsistence and income in the regions. The sample size was determined using the formula provided by Ref. [46] as shown in equation (1).

$${\mathrm{n}}_0=\frac{Z^{2}pq}{e^2}$$ (1)

where; ${\mathit{n}}_{\mathbf{0}}$ represents the sample size, Z is the Z-critical value corresponding to a specific confidence level, p denotes the maximum variance, q is (1-p) and e is the desired margin of error. For this study, a confidence level of 95.1 % and a margin of error of 0.049 were used to enhance the reliability of the results. The 95.1 % confidence level was chosen to ensure the sample size adequately represents the population, considering the variability among smallholder farmers, including differences in resource access and ownership. The p was set at 0.5 due to the unknown variance among CSH farmers and the fact that green grams, being more drought-resistant than tomatoes, introduces additional variability. Consequently, the sample size was calculated as detailed in equation (2);

$${\mathrm{n}}_{\mathrm{o}}=\frac{{1.96}^2\times 0.5\times 0.5}{{0.049}^2}=400$$ (2)

The sample size was proportionally allocated across the three sub-counties based on their respective populations [47]. Following standard practice in cross-sectional surveys, an additional 20 farmers (equivalent to 5 % of the calculated sample size) were included to account for incorrectly filled questionnaires and potential non-response. Subsequently, individual tomato and green gram farmers were selected randomly and semi-structured questionnaires were administered to collect information about the socio-economic characteristics, mobile phone usage and CSH awareness by farmers. A total of 415 farmers participated. During data cleaning, 12 incomplete questionnaires were excluded from the analysis. As a result, the study utilized a final sample of 403 valid respondents: 222 from Taveta, 122 from Mwatate and 59 from Wundanyi sub-counties. Among these respondents, 115 were tomato farmers, 259 were green gram farmers, and 29 grew both crops.

2.3. Conceptual framework

Fig. 2 provides the pathway(s) in which mobile phone use affects horticulture productivity. This conceptual framework is based on previous studies [see for example, 30, 32, 33, 34, 38]. These studies revealed that farmers utilize mobile phones to obtain agriculture and climate information in different contexts. The application of mobile phone in CSH is affected by various factors, including gender, education level, farm size, farm location, credit access and group membership [37,48]. An extensive review of the literature by Ref. [31] reported that using a mobile phone improves market systems for inputs and output between farmers and end-users by simplifying the communication process. They also showed that by providing information in advance, mobile phones reduce climate variability and vulnerability. A study on farmers’ perceptions about use of mobile phones in Tanzania by Ref. [35] also found a positive association between mobile phones and farm productivity due to reduced transaction time and costs.

Fig. 2.

An illustration of the pathways through which use of mobile phones affects farm efficiency and productivity.

Mobile phone use can be categorized as productive and consumptive purposes. Productive purpose involves the use of a mobile phone for economic gain – e.g. farmers using it for agricultural purposes, while consumptive purpose entails using a mobile phone for normal social interactions and entertainment which do not result to any economic gain [32,33]. In summary, effective application of mobile phone increases TE and agricultural production [35].

2.4. Theoretical framework

This study is based on a random utility theory (RUT). The theory models an individual’s preferences on alternatives by independently drawing a real-valued score on each alternative from a parameter distribution [49]. These alternatives are then ranked according to the identified scores. In this case, the farmer is assumed to be a rational decision maker who seeks to maximize farm output or minimize input cost relative to his/her choices subject to his/her socio-economic, institutional and infrastructural characteristics [48]. This theory has been widely used for modelling choices among discrete alternatives [50,51]. For example [51], used the RUT to conceptualize household’s participation in CSA interventions. He postulated that a farmer would be inclined to adopt CSA practices if the perceived benefits of adoption outweigh those of not adopting. Previous studies including [34,35,52] demonstrated a positive link between mobile phone usage and crop productivity. Farmers who believe that mobile phone usage enhance productivity are more likely to utilize them compared to those who do not hold this belief [35].

This study assumed that a decision by a farmer to use mobile phone in CSH (as an agricultural production good) can be modelled as a choice, given the farmer’s socio-economic, institutional and infrastructural factors. Therefore, farmers are expected to choose an alternative which has the greatest perceived utility – in this case perceived higher TE and increased horticulture productivity. The expected signs of various covariates are summarized in Table 1.

Table 1.

Variables incorporated in the Tobit model.

Variable	Description	Tobit model
Farmer’s gender	Male = 1, Female = 0	+/−
Age of the farmer	Number of completed years from birth	+/−
Farmer’s education level	Number of years spent in formal schooling	+
Size of the household	Number of dependants in a household at the time of survey	+
Experience in farming	Years of active farming	+
Mobile phone use in CSH	Yes = 1, No = 0	+
CSH participation	Yes = 1, No = 0	+
Farm size	Number of acres under crops covered in this study	–
Membership in a farmer group	Yes = 1, No = 0	+/−
Access to agricultural extension services	If the farmer received agricultural training within the last one year (Yes = 1, No = 0)	+
Distance from farm to the bank	Distance in kilometers	–
Climate change awareness	Yes = 1, No = 0	+/−
Climate-smart horticulture awareness	Yes = 1, No = 0	+

Age was expected to either positively or negatively influence TE and CSH productivity. A study by Ref. [53] found that young people used smartphone more on entertainment and social interactions while older people used it for getting agricultural-related information. Conversely [54], showed that the probability of implementing CSA practices increased with age but eventually drops with further age increase due to reduced capacity to work. On the other hand [27], found that age was not a significant determinant of mobile phone use in farming. This was because most farmers owned low-capacity phones, which limited their access to multiple agricultural services.

Owusu et al. (2017) showed that fewer female farmers owned mobile phones and used internet relative to their male colleagues. Gezimu et al. (2019) and [55] found that male farmers realized higher agricultural productivity relative to female counterparts. However, under similar environment the productivity for females was higher than males. Therefore, gender was expected to either positively or negatively influence CSH productivity based on the utilization of mobile phones in farming.

In this study, education was measured as the number of years completed in school. According to Ref. [56], increasing years spent in formal education increases the likelihood of mobile phone use. Likewise [54,57], indicated that literate household heads were likely to adopt CSA practices. Based on these findings, educated farmers were likely to use their mobile phones in farming, adopt CSH practices and have higher productivity due to their ability to access wide range of services [14,36].

Household size was assessed based on the number of dependents living in the household at the time of the survey. According to a study by Ref. [58], a larger family size was found to significantly boost the use of mobile phones in agriculture. Urgessa (2015) and [59] found that increased household size significantly improved agricultural productivity. Consequently, the expected effect of household size on CSH productivity was positive. This was based on the premise that a larger household would provide labor for CSH activities, which improve productivity.

Farming experience was estimated based on the number of years the respondent had engaged in active farming. According to Ref. [60], greater farming experience positively influences the adoption of CSA practices, suggesting a positive relationship between farming experience and the number of CSA practices adopted. This is anticipated to positively impact both technical efficiency (TE) and productivity.

The size of the farm was measured in acres under green grams and/or tomatoes. Akinola (2017) found that farm size positively affects mobile phone use in agriculture. Similarly, an increase in farm size is associated with a higher likelihood of adopting CSA practices [57,60]. However, [61], 112] established a negative relationship between farm size and productivity. The negative relationship arises because larger farms make crop management practices difficult due to the limited resources and labor. Therefore, the anticipated effect of farm size on TE and CSA productivity is expected to be negative.

In this study, mobile phone use is represented as a dummy variable, taking a value of 1 if the farmer utilizes a mobile phone for CSA and 0 otherwise. Aminou et al. (2018) and [33] demonstrated a positive correlation between mobile phone use and agricultural productivity. Therefore, it was hypothesized that mobile phone use would positively impact CSA productivity, as it provides convenient access to information that enhances the farmer's production skills.

Access to credit was represented as a dummy variable, with a value of 1 if the farmer obtained credit in the past twelve months and 0 otherwise. Anang (2019) showed that credit is positively related to productivity. Therefore, the expected effect of credit on productivity was assumed to be positive because it improves productive capacity of farmers – in terms of required agricultural assets and inputs. Okello et al. (2014) found that membership in a farmer organization increased the likelihood of utilizing market information and mobile phone-based money transfer services. However [62], reported mixed effects of farmer group membership on the adoption of agricultural technology and productivity. Therefore, the impact of group membership on CSA could be either positive or negative. This is because farmer groups make it possible to transfer information and technologies from one farmer to another, which means that if they get the right information they will do well while wrong information would affect their productivity negatively.

According to Refs. [57,63], access to agricultural extension services by farmers was positively associated with the number of Climate Smart Agriculture (CSA) practices adopted and agricultural productivity. Therefore, it was proposed that access to extension services would positively affect the TE and CSA productivity in the study area. This is because agricultural extension enhances the farmer’s skills thereby improving productivity.

Awareness on climate change and CSH was measured based on the understanding of farmers about the concept and practice. Mwikamba et al. (2024) found that general awareness of climate change did not lead to an increase in the adoption of CSH practices. However, awareness of CSH enhanced the probability of adopting more CSH practices. Additionally, studies by Refs. [[64], [65], [66]] demonstrated that CSA improves farmers’ soil fertility and crop yields and food security. Therefore, this study postulated that the TE and productivity effects of climate change awareness would be either positive or negative; while CSH awareness would be positive.

2.5. Data analysis

A paired t-test was employed to identify statistical differences in farm and farmer characteristics between tomato and green gram growers. The TE was assessed using input-oriented data envelopment analysis (DEA) [57]. This is because attempts in previous studies to compare different efficiency measures using cross-section data exhibited similar results. For instance, a study by Ref. [67] on comparison of different frontier approaches, was unable to conclude on the best frontier approach since stochastic frontier analysis (SFA) and DEA results were correlated. It has been shown that DEA provides more satisfactory productivity measurement than SFA by accounting for size and scale of operation [68,69], for example farm size.

The DEA model is a linear programming technique used to evaluate the efficiency and productivity of decision-making units (DMUs) – which in this case are CSH farmers [59]. Empirical evidence indicates that there is no significant difference in efficiency levels of decision making units between input and output-oriented DEA models [[70], [71], [72]]. Consequently, the input-oriented DEA was applied in this study since it fitted the data well as opposed to the output-oriented DEA. The TE scores were calculated using DEA variable returns to scale (VRS) input-oriented model shown in equations (3), (4), (5), (6), (7)) following Huguenin (2012):

$$Minimize{\theta }_k$$ (3)

Subject to:

$$y_{rk}-\sum _{j=1}^n{\lambda }_j{y}_{rj}\le 0r=1,\dots ,s$$ (4)

$${\theta }_k{x}_{ik}-\sum _{j=1}^n{\lambda }_j{x}_{ij}\ge 0i=1,\dots ,m$$ (5)

$$\sum _{j=1}^n{\lambda }_j=1$$ (6)

$${\lambda }_j\ge 0\forall j=1,\dots ,n$$ (7)

where: ${\theta }_k$ represent the technical efficiency of farm k; $y_{rk}$ is the quantity (in kilograms) of green grams and tomatoes generated by farm k; $x_{ik}$ is the amount of input - such as acreage of farm under crop, man days of labor and kilograms of fertilizer consumed by farm k; s is the number of yields (output); n is the number of firms to be evaluated (in this case 403 farms for pooled sample); m is the number of inputs; ${\lambda }_j$ indicates the weight assigned to the outputs and inputs of firm j; $\sum _{j=1}^n{\lambda }_j=1$ is the convexity constraint and ${\Sigma }_{j=1}^n$ is an n × 1 vector of ones. This method accounts for variations in returns to scale among DMUs and yields TE scores that are at least as high as those from a constant returns to scale model [57]. Further, the study was premised on the logical rationale of [32,33,73] that a mobile phone can be considered a production good if used for productive activities. A two-step approach was applied in measuring the effect of using a mobile phone on productivity of CSH; computation of productivity estimates followed by analysis of factors that influence the productivity levels.

The two main approaches used for measuring productivity include total factor productivity (TFP) and PFP. The PFP is used to evaluate the efficiency of the specific resource of interest (single input) and it is easy to measure and interpret with cross-sectional data. However, it does not consider other inputs involved in the production process [57]. Conversely, TFP assesses productivity by comparing the overall output to a combined index of all inputs. This method accounts for the productivity of all inputs applied in the production process and is often accounted for by changes in technology such as information and communication technology. This poses a challenge in empirical measurement with cross-sectional data that may not fully capture such dynamics [74]. In this study, productivity was measured using the PFP method due to data limitations. The choice of this method is also based on the premise that the PFP measure is suitable when the analyst’s objective is on a specific policy issue – in this case, mobile phone use. The PFP was computed as shown in equation (8).

$${\text{PFP}}_{\mathrm{i}}=\frac{{\mathrm{Q}}_{\mathrm{i}}}{{\mathrm{X}}_{\mathrm{i}}}$$ (8)

where; ${\mathit{P}\mathit{F}\mathit{P}}_{\mathit{i}}$ is the partial factor productivity score for farm i, ${\mathit{Q}}_{\mathit{i}}$ is the value of output from farm i and ${\mathit{X}}_{\mathit{i}}$ represents the size of the farm (under crop) i in acres.

The decision to focus on farm size was due to the fact that land was the commonly used input by all farmers to produce green grams and tomatoes in the study sites – other inputs such as pesticides were not commonly used by all farmers and hence it would be difficult to use them as the basis for computing productivity.

The productivity measurements obtained from equation (8) were then transformed into logarithms as shown in equation (9).

$${\text{PFP}}_{\mathrm{i}}=\ln \left(\frac{{\mathrm{Q}}_{\mathrm{i}}}{{\mathrm{X}}_{\mathrm{i}}}\right)$$ (9)

The logarithmic transformation was necessary to linearize the dependent variable and thus reduce skewness of PFP scores. The transformed PFP scores can be regressed against various independent variables using ordinary least squares (OLS) or Tobit model [75]. The OLS provides better results when qualitative data is used [62] but leads to biased estimates where the data is bounded between two limits [76,77]. On the other hand, the Tobit model provides unbiased estimates where the data is distributed within a range, since it uses maximum likelihood method [75,78,79]. Therefore, this study applied a Tobit model (censored from below) [49] to assess the determinants of CSH productivity and specifically to identify the effect of using a mobile phone on CSH productivity. This is because the Tobit model fitted the data better than OLS and the fact that productivity estimates cannot be negative. Following [78], the Tobit model was estimated as follows:

$${\mathrm{Y}}^{*}={\mathrm{\beta }}_0+\beta \mathrm{X}+\mathrm{\epsilon },\mathrm{\epsilon }|\mathrm{X}\sim \mathrm{N}\left(0,{\mathrm{\delta }}^2\right)$$ (10)

where, Y* is the latent variable (lnPFP_i) that satisfies classical linear model assumptions (the model is linear in parameters, regressors are uncorrelated with the error term, homoscedasticity, among others), β represents a vector of coefficients to be estimated, X denotes a vector of independent variables and ε is the error term, which is assumed to have a normal distribution. Since the dependent variable was log-transformed, interpretation and reporting of the Tobit model results were based on normalized values using the formula below:

$$100\times \left({\mathrm{e}}^{\hat{\mathrm{\beta }}}-1\right)$$ (11)

where; $\hat{\mathit{\beta }}$ is the coefficient of estimation obtained from Equation (10).

2.6. Model diagnostic tests

2.6.1. Multicollinearity test

Multicollinearity test was done using variance inflation factor (VIF) to identify the variance of the independent variables [80] as shown in equation (12).

$$VIF=\frac{1}{\left(1-R_j^2\right)}$$ (12)

where, R_j² is the coefficient of determination.

The average VIF obtained was 1.9, meaning that there were no serious issues of multicollinearity between the independent variables. All the variables had a VIF of less than 4 indicating that there was a low level of multicollinearity [66]. In addition, partial correlation analysis confirmed insignificant correlation between any two independent variables.

2.6.2. Test for heteroscedasticity

Heteroscedasticity is said to exist if the error term does not exhibit equal variance given the values of independent variables [75]. This causes the model to produce biased standard errors and test statistics which may result to inaccurate conclusion in hypothesis testing [67]. This study applied Breusch-Pagan test and concluded that there was presence of heteroscedasticity in the independent variables used to estimate the Tobit model. This issue was mitigated by employing robust standard errors, which are unbiased and offer a more precise estimate of the true standard error of a regression coefficient [66].

2.6.3. Test for endogeneity

Endogeneity problem arises when some of the independent variables are related with the residual term in the model [68]. This occurs under various circumstances including omitting important variables from the model (omitted variable bias) and when a regressor is a regressand (simultaneity bias) [75,81]. In the presence of simultaneity, OLS estimators are biased and inconsistent, which leads to overestimation of coefficients in the model. This was tested through the Hausman specification test, which confirmed that the regressors were exogenous.

3. Results and discussions

3.1. Farmer and farm attributes

Table 2 indicates farmer(s) and farm characteristics. The results indicate that more male farmers produced tomatoes while more female farmers produced green grams. This is attributed to the fact that tomatoes are high value horticultural crops and fetch higher returns per unit area relative to green grams; hence attracting more interest from men. This is in line with the extant literature which indicate that due to socio-cultural norms in African societies, male farmers tend to cultivate cash crops more than women when the monetary returns are higher. There were also significant differences in farmer’s age; younger farmers preferred tomato production compared to green grams. This is because most young farmers in the area do not own land. Thus, they resort to leasing land in which they have to grow short maturing but high value crops to recover the extra costs and make profit.

Table 2.

Mean differences in socio-economic and institutional characteristics between green grams and tomato farmers.

Variable	Green grams (n = 259)	Tomatoes (n = 115)	Mean difference	Pooled sample (n = 403)
Farmer‘s gender (male = 1)	0.47	0.84	−0.37***	0.60
Age of the farmer (years)	48.73	44.57	4.16***	47.08
Education level of the farmer (years)	7.63	9.73	−2.10***	8.32
Size of the Household (number of dependents)	5.47	5.24	0.23	5.35
Experience of the farmer (years)	9.04	10.61	−1.56	9.41
Mobile phone use in CSH (yes = 1)	0.47	0.70	−0.23***	0.56
CSH participation (yes = 1)	0.84	0.99	−0.15***	0.89
Farm size under crop (acres)	1.47	0.99	0.48***	1.39
Mobile phone ownership (yes = 1)	0.95	0.97	−0.02	0.96
Distance from farm to the nearest bank (km)	19.23	15.84	3.39*	18.97
Agricultural extension services (yes = 1)	0.57	0.73	−0.16***	0.63
Membership to a farmer group (yes = 1)	0.49	0.27	0.22***	0.43
Access to credit (yes =1)	0.10	0.10	0.00	0.10
Awareness on climate change (yes = 1)	0.92	0.94	−0.02	0.92
Awareness on CSH (yes = 1)	0.65	0.75	−0.10*	0.69

Note: * and *** indicate statistically significant differences at 10 % and 1 %, respectively. The pooled sample includes pure green grams and tomato farmers and farmers who produced both green grams and tomatoes.

The results also revealed that green gram farmers spent less years in school compared to tomato farmers. This is because green gram is a traditional crop and does not require specialized knowledge and skills to grow. The crop is not labor intensive and diseases affecting it are relatively easy to control hence is preferred by older farmers, the majority of whom did not attend formal education. On the other hand, tomato crops are labor intensive and require technical skills to plant, control pests and diseases, plan for harvesting and delivery to market. Educated farmers produced tomatoes because they understood that they have more returns per unit area than green grams.

Furthermore, more tomato producers used their mobile phones for CSH information than green gram producers. This is because tomato production needs specialized knowledge and skills, which can be accessed online or from other farmers and agricultural extension service providers (this also explains why more tomato farmers accessed agricultural extension services than green gram farmers).

Most tomato producers were young and could surf the internet to access any information of value to them while most green gram farmers were older and used their phones mainly for social interactions.

Interestingly, green gram farmers had larger amounts of land and most of them were members in farmer groups compared to tomato farmers. This is influenced by the distribution of agro-ecological zones and land ownership patterns in the area. Most green gram producers are located in low rainfall areas such as lower parts of Mwatate sub-county and Matta ward (in Taveta). In these areas, most farmers have large pieces of land compared to highland farmers. The group membership by green gram farmers was influenced by the requirements by agricultural projects, such as the KCSAP and ICRISAT/AVCD project, operating in the area. These projects require farmers to be in groups for them to benefit. These results imply that there is need to invest in programs that directly support the young farmers to access land and practice farming. Since the young farmers easily adopt digital technology, facilitating access to land will highly improve their productivity and development of new ways of dealing with climate change problems at farm level.

3.2. Technical efficiency scores for tomato and green gram farmers

Table 3 displays the input-oriented DEA results. It indicates that the average TE for green gram farmers and tomato farmers in Taita-Taveta County was 25 % and 34 %, respectively. This implies that, on average, green gram and tomato farmers could maintain their current output levels by reducing all inputs by 75 % and 66 %, respectively [70,82]. It also means that farmers who produced either green grams or tomatoes were highly inefficient compared to farmers elsewhere, for example those in Cameroon whose efficiency levels for these crops exceed 65 % [11]. This may be attributed to high infestation of crop pests (Tuta absoluta) for the case of tomato farmers and environmental factors such as adverse impacts climate change. Other contributing factors include; types of seeds used, lack of agricultural extension services and requisite information and skills on crop production. Similar findings were obtained by Ref. [83] in the study of factors influencing TE among tomato farmers in Ethiopia. On the other hand, farmers who produced both crops exhibited an average TE score of 54 %. This suggests that farmers who produced both crops could reduce all inputs by 46 % to be fully efficient. This is mainly due to proper utilization of land, where more total yield is realized under intercropping system compared to monocrop [70].

Table 3.

Technical efficiency scores for green grams and tomato farmers.

Variable	Green grams (n = 259)	Tomatoes (n = 115)	Both green grams and tomatoes (n = 29)	Pooled sample (n = 403)
Mean	0.249	0.343	0.543	0.297
Maximum	0.992	1.000	1.000	1.000
Minimum	0.000	0.000	0.027	0.000
Standard deviation	0.305	0.333	0.318	0.323

Fig. 3 illustrates the distribution of TE among all farmers in Taita-Taveta County, including those cultivating tomatoes, green grams, and both crops. The results reveal that 60 % of the farmers attained technical efficiency levels below 0.25, indicating that most farmers in the region have the potential to enhance their efficiency by up to 75 %. The farmers can achieve this by integrating mobile phone technology in their farming schedules to gain skills that will enhance their efficiency levels [71]. Further, encouraging farmers to attend agricultural training meetings and improving extension services (in terms of accessibility and affordability) will improve efficiency [72].

Fig. 3.

Distribution of technical efficiency for green grams and tomato farmers in Taita-Taveta County. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

3.3. Productivity levels among climate-smart horticulture farmers in Taita-Taveta County

Table 4 shows the results of the first-step analysis (partial factor productivity). The average productivity levels for tomato and green gram farmers in Taita-Taveta County were 143 kg/acre and 4817 kg/acre, respectively. These levels are much lower than previous figures of 280kg/acre and 12,667 kg/acre for green grams and tomatoes, respectively for the same county in 2017. Additionally, the estimated productivity levels for the study site are below the national average that ranges between 520 and 600 kg/acre for green grams and 8000 to 10,000 kg/acre for tomatoes [107, 108]. Therefore, the results of this study represent 62 % and 49 % drop in tomato and green gram productivity, respectively, between the period 2017 and 2021.

Table 4.

Productivity scores (Kgs per acre) for green grams and tomato farmers.

Variable	Mean	Std. deviation	Minimum	Maximum
Green grams (n = 259)	143.30	215.42	0.00	1800.00
Tomatoes (n = 115)	4817.12	5736.50	50.00	32,000.00
Both tomatoes and green grams (n = 29)	2716.03	2508.18	24.59	11,500.00
Region
Wundanyi (n = 59)	3591.97	3700.53	200.00	20,000.00
Mwatate (n = 122)	394.47	1203.81	0.00	11,000.00
Taveta (n = 222)	1843.84	4422.56	0.00	32,000.00
Pooled sample (n = 403)	1661.00	3770.55	0.00	32,000.00

The study area’s low productivity was attributed to high incidence of crop pests (Tuta absoluta) - for the case of tomato farmers, lack of agricultural information and agronomic skills, farm-specific challenges (such as topography, soil nutrient degradation, lack of water for irrigation), lack of capital resources including finance (most farmers had only land and lacked other quality enablers to realize better yields), prolonged drought and unpredictable rainfall distribution (seasons) experienced in the area due to climate change. Other factors that contributed to low yields included disruption of horticultural supply chain and markets by the Covid-19 pandemic and low purchasing power of inputs by farmers. Wundanyi exhibited higher productivity relative to Taveta and Mwatate regions because it is located on a high-altitude area with greater agricultural potential and all farmers were tomato producers, which has high output per unit area compared to green grams.

Development of programs that support the youth to take up agriculture will contribute to realization of higher output. This is because most youths in the area have basic education and can easily access relevant agricultural information through their mobile phones compared to older farmers. They can exploit opportunities presented by improved mobile technology to deal with complex climate change problems, pests and disease incidences and farm-specific challenges [22,27,33].

3.4. Effect of using a mobile phone on technical efficiency of climate-smart horticulture farmers in Taita-Taveta County

The pooled results in Table 5 indicate that social characteristics including gender, education and farming experience positively affected TE by 12 %, 1 % and 0.4 %, respectively. This implies that male farmers had higher TE compared to female farmers. This finding is consistent with studies by Ref. [55] and [110] which, reported that male households achieved higher TE and productivity than their female counterparts due to low access to inputs and information by female farmers.

Table 5.

Tobit regression analysis results on the effect of using a mobile phone on technical efficiency of climate-smart horticulture farmers.

Variable	Green grams (n = 259)		Tomatoes (n = 115)		Pooled sample (n = 403)
	Coefficient	Robust Std. Error	Coefficient	Robust Std. Error	Coefficient	Robust Std. Error
Independent variables
Gender (male = 1)	0.071	0.044	0.148***	0.056	0.117***	0.033
Age of the farmer (years)	−0.001	0.002	−0.001	0.002	−0.001	0.001
Education level of the farmer (years of schooling)	0.014**	0.006	−0.007	0.010	0.011**	0.005
Household size (count)	−0.001	0.009	−0.023*	0.013	−0.009	0.007
Farming experience (years of farming)	0.002	0.003	0.008*	0.004	0.004*	0.002
Farm size (acres under crop)	0.025	0.019	0.048**	0.021	0.047***	0.014
Access to agricultural extension services (yes = 1)	−0.121	0.073	−0.118	0.073	−0.066	0.047
Farmer group membership (yes = 1)	−0.066	0.061	0.036	0.058	−0.032	0.037
Distance between the farm and commercial bank (km)	0.002	0.001	0.009***	0.002	0.005***	0.001
Mobile phone use in CSH (yes = 1)	0.081*	0.045	0.100*	0.051	0.087***	0.034
Participation in CSH (yes = 1)	0.131	0.086	0.011	0.119	0.057	0.063
Awareness of climate change (yes = 1)	0.027	0.093	−0.145	0.107	−0.014	0.071
Awareness of CSH (yes = 1)	0.026	0.068	0.169***	0.061	0.093*	0.049
Constant	−0.083		0.198		−0.060
Log likelihood	−90.460		−5.720		−119.550
Pseudo R2	0.263		0.874		0.380
Prob > F	0.000		0.000		0.000

Note: *, ** and *** represent levels of statistical significance at 10 %, 5 % and 1 %, respectively. CSH means climate-smart horticulture. The dependent variable was the level of technical efficiency measured as a ratio ranging from 0.00 to 1.00.

Also, an increase in farm size by one acre resulted in an increase in TE by 5 %. This finding is similar to that of [111] who observed that TE was higher in larger farms than smaller ones. This can be attributed to the fact that large farms enjoy economies of scale and can access a wide range of technologies compared to smaller farms. In addition, most small-scale farmers treat their farms as a secondary source of income hence they may not focus all their efforts and investments on it [84].

Further, the results show that being located away from a commercial bank positively affected TE of CSH farmers. This contrasts with findings of [113], which suggested that being close to sources of credit improved access to finance hence positively affecting agricultural production. However, in climate change context, the results suggest that farmers located far away from commercial banks are more productive. This is likely due to the availability of mobile money transfer services (such as M-Pesa) and mobile banking, which allow farmers to conduct transactions via their phones, freeing up more time for productive activities like farming.

The results indicate that mobile use on CSH enhances the TE of farmers by 9 %. This result support our hypothesis that mobile use contributes to improved efficiency. The finding aligns with previous research, which demonstrates that mobile phone ownership and use boost TE in crop farming [105, 114]. This improvement is attributed to mobile phones bridging information gaps, enhancing agricultural extension services, facilitating the adoption of advanced technologies, and providing access to cognitive resources that improves both TE and productivity [115].

Likewise, being aware of CSH (what it means and the appropriate practices it entails) increased TE by 9 %. Thus, farmers who are aware about CSH (both in concept and practice) adopt practices that are best suited to their environments [116−118]. This leads to improved TE. Previous literature showed that awareness of CSH was largely influenced by farmer’s age, access to extension services, farming experience and indigenous knowledge [62, 116].

3.5. The effect of using a mobile phone on productivity of climate-smart horticulture farmers

Table 6, Table 7 show the Tobit model results on the effect of using a mobile phone on productivity of CSH farmers based on the crop and geographical location, respectively. The coefficients were normalized using equation (11) for easy reporting – since the dependent variable was log-transformed [85].

Table 6.

Tobit regression (censored from below) results for the effect of using a mobile phone on CSH productivity.

Variable	Green grams (n = 259)		Tomatoes (n = 115)		Pooled sample (n = 403)
	Coefficient	Robust Std. Error	Coefficient	Robust Std. Error	Coefficient	Robust Std. Error
Independent variables
Farmer’s gender (male = 1)	0.325	0.322	0.823***	0.287	1.492***	0.278
Age of the farmer (years)	0.001	0.014	−0.024**	0.010	−0.028**	0.012
Farmer’s education level (years spent in school)	0.122***	0.040	0.053	0.036	0.169***	0.037
Farmer’s household size (count)	0.014	0.068	0.045	0.051	0.027	0.060
Farming experience (years of farming)	0.011	0.020	0.026	0.018	0.050***	0.018
Farm size (acres under crop)	−0.279**	0.124	0.088	0.057	−0.233**	0.112
Access to agricultural extension services (access = 1)	−0.559	0.488	−0.433	0.307	0.126	0.371
Farmer group membership (yes = 1)	−0.874*	0.454	0.073	0.221	−1.376***	0.323
Distance between farm to the nearest commercial bank (km)	0.011	0.009	0.019**	0.008	0.010	0.008
Mobile phone use in CSH (yes = 1)	0.939***	0.318	0.073	0.248	0.903***	0.294
Participation in CSH (yes =1)	−0.234	0.594	−0.561	0.574	0.618	0.542
Awareness of climate change (yes = 1)	0.279	0.546	−0.482	0.406	−0.239	0.529
Awareness of CSH (yes = 1)	0.787**	0.469	0.490	0.306	0.808**	0.390
Constant	1.995**		7.670***		2.607***
Prob. >F	0.000		0.000		0.000
Pseudo R2	0.053		0.093		0.083
Log likelihood	−523.750		−173.520		−882.010

Note: *, ** and *** indicate levels of statistical significance at 10 %, 5 % and 1 %, respectively. Partial productivity (measured as kg/acre) was the regressand in the Tobit model.

Table 7.

Tobit regression (censored from below) results for the effect of using a mobile phone on CSH productivity in the three sub-counties.

Variable	Wundanyi (n = 59)		Mwatate (n = 122)		Taveta (n = 222)		Pooled sample (n = 403)
	Coef.	Robust SE	Coef.	Robust SE	Coef.	Robust SE	Coef.	Robust SE
Independent variables
Farmer’s gender (male = 1)	0.634*	0.256	−0.073	0.280	1.656***	0.269	1.492***	0.278
Age of the farmer (years)	−0.008	0.009	−0.014	0.012	−0.073	0.011	−0.028**	0.012
Farmer’s education level (years of schooling)	0.020	0.021	0.063**	0.030	0.086**	0.038	0.169***	0.037
Size of the household (count)	−0.003	0.037	−0.057	0.063	0.024	0.055	0.027	0.060
Farming experience (years of farming)	0.040**	0.018	0.038**	0.018	0.087	0.014	0.050***	0.018
Farm size (acres under crop)	−0.183	0.215	−0.220	0.257	−0.051	0.081	−0.233**	0.112
Access to extension services (access = 1)	−0.140	0.385	−0.528	0.502	0.027	0.285	0.126	0.371
Membership to a farmer group (yes = 1)	−0.030	0.144	−0.463*	0.368	−0.518**	0.312	−1.376***	0.323
Distance between the farm to commercial bank (km)	−0.267*	0.018	−0.004	0.019	0.024***	0.007	0.010	0.008
Mobile phone use in CSH (yes = 1)	0.508	0.251	0.253	0.238	1.204***	0.261	0.903***	0.294
Participation in CSH (yes =1)	0	omitted	1.874	0.801	−0.444	0.425	0.618	0.542
Awareness of climate change (yes =1)	−0.457	0.527	1.092	0.554	−0.598**	0.385	−0.239	0.529
CSH awareness (yes =1)	0.640*	0.261	2.281**	0.539	0.805*	0.346	0.808**	0.390
Constant	4.507***		4.507**		2.506***		2.607***
Pseudo R2	0.119		0.083		0.089		0.083
Prob. >F	0.013		0.004		0.000		0.000
Log likelihood	−65.340		−200.590		−402.050		−882.010

Note: *, ** and *** indicate levels of statistical significance at 10 %, 5 % and 1 %, respectively. Partial productivity (measured as kg/acre) was the regressand in the Tobit model.

The pooled results in Table 6 showed that being a male farmer, having education, and more farming experience positively affected productivity (kg/acre) by 150 %, 17 % and 5 %, respectively. This implies that male farmers attained higher productivity levels than the female farmers, both in terms of the crop enterprise and geographical location. Gebissa et al. (2019) observed that the male household heads achieved higher productivity compared to their female counterparts due to the limited access to inputs experienced by female farmers. Education and farming experience enhance the farmer’s knowledge and skills to deal with the changing environment and use a mix of production technologies for higher productivity. The findings regarding the effect of education are consistent with those of [81], which demonstrate that education significantly enhances productivity. However, a more significant impact of education on farmer productivity was noted on those who had more than 5 years of schooling.

An increase of one year in a farmer’s age resulted in a 3 % reduction in productivity, implying that older farmers tend to produce less output per unit area. This decline is attributed to a reduced capacity for production, including factors such as labor, access to credit, and production technologies, which are crucial for achieving higher productivity. Similar observations were made by [121] who noted that productivity was low among the aged population compared to middle-aged farmers. This imply that the capacity of young people to take up the role in farming should be developed through training, facilitating access to land and provision of affordable credit facilities.

According to the results (Table 6, Table 7), being in a farmer group was shown to negatively affect productivity by 1.4 times. This finding is in line with [62] who noted that farmer group membership affected productivity differently based on the type of crop enterprise. The negative effect of group farming was attributed to the type of crop enterprise and poor quality of agricultural information passed from one farmer to another. It was observed that most of the farmer group members (74 %) in the region were green gram farmers, majority of whom were in Mwatate sub-county. The lower parts of this sub-county usually experience longer dry seasons compared to Wundanyi and Taveta. It was also noted that group members did not have the right information on rainfall distribution and the right agro-chemicals. Likewise, the green gram farming groups depended on the seeds provided by KCSAP, which fetched lower market prices compared to non-members.

The results (Table 6, Table 7) revealed that using a mobile phone on CSH increases productivity by 90 %. This finding aligns with existing literature, which shows that mobile phone ownership and use enhance agricultural productivity [32,86]. This is due to the fact that using a mobile phone reduces agricultural information gaps, improve agricultural extension service delivery, adoption of improved technology and access to cognitive assets that lead to increase in horticultural productivity [30, 125]. This implies that agricultural extension agents should encourage farmers to use their phones to access information on farming. In addition, increased investment is needed to enhance mobile phone network coverage in rural areas where farming is the primary economic activity.

Being aware of CSH (what it means and practices involved) increased productivity by 81 %, ceteris paribus. Therefore, it implies that most farmers who are aware about CSH (both in concept and practice) adopt practices that are best suited to their environment [116–118]. Adoption of CSH practices such as soil management, use of improved seeds, water management and agronomic practices positively affected food crop yields and productivity. The finding is consistent with [65,66] who contend that CSA practices have a positive impact on productivity. The results indicate the need for extension agents to create more awareness and train farmers on climate smart horticulture techniques that are suitable in various contexts, to deal with climate change problems.

4. Conclusion and recommendations

This study found that TE and productivity levels of both green grams and tomatoes in Taita-Taveta County were much lower than previous levels as well as the national average potential. The findings showed that producing both crops resulted in higher TE scores compared to a single crop. Mobile phone use on CSH coupled with credible information were found to significantly improve TE and productivity. Other important determinants of TE and productivity in the context of CSH, include gender, farmer’s education level, age, farm size under the target crop, awareness of CSH and membership to a farmer group. In order to ensure optimal use of resources and safeguard household food security and income needs, we recommend joint cultivation of green grams and tomatoes.

The study also recommends a strong partnership between telecommunication service providers and agricultural service providers to provide credible, timely and simple/easier to understand CSH information accessible by mobile phone users. Efforts should be directed towards improving the skills of frontline agricultural extension workers and farmers in using mobile phones for CSH. Additionally, developing a policy for agricultural management information systems is essential to ensure the effective delivery of reliable agricultural content to farmers.

The county government and development partners, including organizations like the Food and Agriculture Organization of the United Nations, should work together to raise awareness about climate change and CSH among farmers to enhance their resilience and adaptability. Both the County and National governments, along with development partners, should offer incentives that encourage the youth and women to adopt technology in CSH.

Such incentives would take the form of youth and women concerts where participants are facilitated to attend and showcase their creativity in applying mobile phones to communicate CSH content; awards could further motivate exemplary performers in such events.

While this study has offered some insights on the value of using a mobile phone on efficiency and productivity, there were various limitations occasioned by the nature of data used, analysis method and crop enterprises selected. Future research could bridge this gap by providing more comprehensive information about the effect of using a mobile phone on CSH using time series data and total factor productivity models that accommodate multiple inputs in the productivity analysis and the impact of CSH awareness on climate change adaptive capacity and farmer welfare components such as food security and income using multi-period data.

Funding statement

This study was financed by the Kenya Climate Smart Agriculture Project (KCSAP) under scholarship grant number; KCSAP/New MSc/COMP.3/26.

Data availability statement

The authors have not deposited data associated with this study into a publicly available repository. However, the data will be availed upon reasonable request.

Ethical declaration

The authors confirm that they sought approval from the National Commission for Science, Technology and Innovation (NACOSTI) prior to conducting the survey. The approval was granted through license number: NACOSTI/P/21/9137. In addition, they orally sought consent from the respondents before they interviewed them and only those who were 18 years of age and above were allowed to participate in the exercise.

CRediT authorship contribution statement

Jimson Nyambu Mwikamba: Writing – original draft, Visualization, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. David Jakinda Otieno: Writing – review & editing, Validation, Supervision, Project administration, Methodology, Formal analysis, Data curation. Willis Oluoch-Kosura: Writing – review & editing, Validation, Supervision, Methodology, Data curation.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: