Estimating agriculture technologies’ impact on maize yield in rural South Africa

New technologies and digital infrastructures are enabling smart farming with high productivity levels across countries. Yet, there is dearth of evidence on how they are enabling smallholder farmers in Africa to increase their farm produce. By focussing on smallholder maize crop producers in South Africa and relying on a panel data for the period 2011–2021, this study applied a stochastic production frontier framework and Cobb–Douglas production function to estimate the relationship between variance of yield increase in relation to aggregate modern technologies adopted. The stochastic production frontier model was estimated using simple ordinary least square, maximum likelihood estimation and probit regressions. The analysis adopted a three-stage procedure which began with the specification of regression model, the calculation of least residuals squares of the explanatory variables and finally the estimation of the variances in the functional forms of output levels. The results revealed a positive relation between the application of digital agriculture technologies and increased crop performance for rural maize cultivators in South Africa. The policy inference from this study is that accelerating investment in digital agriculture infrastructure offers the promise of a quadruple return for South Africa’s agriculture sector.


Introduction
New technologies and digital infrastructures such as robotics, remote sensing technologies and unmanned aerial vehicles, artificial intelligence and big data analytics among others are enabling smart farming with high productivity levels across the world (Boakye and Babatunde 2021;Boakye et al. 2021). Thanks to advances in today's digital revolution, agriculture is now becoming more integrated in the global value chain. Yet, the interplay between technology adoption and increased crop yield is one that is not well explored in the literature. Given the significance of yield productivity as a determining factor of measuring a country's overall output, examining the drivers of high crop yields is crucial to inform better policy decisions (Edgerton et al. 2012;Evans 2018).
Crop yield is a principal unit of measuring agriculture's productivity. Crop yield also forms part of the basis to understand the efficiency of new techniques introduced in farming practice. Yet, few studies have explored how yield productivity is driven by the application of modern tools and techniques in small-scale farming. The lack of extensive research in this field suggests that data on how farm produce are influenced by technology application is limited. The study of crop yield and its relation to technology application is critical to understand the capacity of agriculture to contribute to overall welfare. As emerging technologies continue to be integrated in the agriculture sector, this study explores how they influence maize yield performance in rural South Africa.
Maize is one of the most important grain crops cultivated in South Africa. An average of 9 million metric tons of maize are produced annually in the country. In 2020, South Africa's total production of maize was estimated at 15 million metric tons, indicating a slight increase from 14 million metric tons in the previous year. Although large-scale maize production is generally carried out by commercial farmers, emerging and small-scale farmers equally contribute significantly to the country's overall production capacity (Aguera et al. 2020). However, in recent times, maize yields by small-scale farmers in rural areas have been declining. The decrease in the yields has been attributed to harsh climate conditions, yet it is difficult to rule out farmer's limited farming skills and the lack of climate smart technologies applied in crop production. There is generally a low uptake and application of modern technologies by these farmers because of capital constraints and limited government support. Moreover, many of the rural farmers have not been exposed to the benefits of adopting modern innovations in farming and as such do not fully appreciate their usage.
Given the growing levels of maize related food consumption in the country amidst the disruption of global food supply chain orchestrated by the Russia Ukraine conflict, South Africa famers may need to adopt new and better approaches to increase their maize production. Fortunately, there are several proven technologies and interventions that can support maize production sustainability. What is needed is to encourage their adoption by farmers including smallscale farmers in rural areas. To achieve this objective, however, requires having reliable data that demonstrate strong positive net effects of technology application on yield performance. As indicated earlier, the study of technological impact on yield performance is critical to understanding South Africa's agriculture sector capacity to provide sustainable food for long periods. This position forms the foundation of this study.
When it comes to estimating technological contribution to yields, most evidence from agriculture literature exclusively focus on highly industrialised countries (Ramirez and McDonald 2006;Tolhurst and Ker 2016). In contrast, there is limited evidence from South Africa even though technology integration in farming practices is widely adopted. This study, therefore, purports to contribute to the literature by drawing evidence from South Africa using the country's yield data for the period 2011-2021. The study draws on the overwhelming evidence which suggest that adopting digital technologies in small-scale farming is good for maximising productivity and increasing production levels. This study is novel in that it adopts a sophisticated methodology which relies on a stochastic production framework and Cobb-Douglas production function. The focus on smallholder farmers yield in rural South Africa also contributes to the study's uniqueness since the area has not attracted the needed attention from recent scholars. The overall objective is to encourage policy makers to design and implement policies that promote the fast adoption of smart technologies in rural farming. The structure of the article proceeds as follows: the next section provides an overview of crop production in South Africa. It then explores the potentials of modern technologies in crop production after which the literature on technology adoption and increase in crop yield is presented. The subsequent sections focus on methodology, results, discussion, and policy recommendation.

Overview of crop production in South Africa
The agriculture sector is one of the important sectors of South Africa's economy. The sector contributes around 10% to the country's total export earnings and remains a key source of livelihoods for most rural population (Aguera et al. 2020). Among the major agriculture practices in South Africa, crop production represents one of the key practices engaged by most farmers. Different kinds of crops are cultivated in South Africa. Meanwhile the predominant ones are maize, wheat, soybeans and oat (Christiansen and van den Brink 1994). Approximately, 9 million metric tons of maize are produced annually on a 2.5 million hectare of land. According to the Food and Agriculture Organisation, both hybrid and local maize are grown in South Africa, yet the local ones are often preferred for home consumption (Christiensen and Demery 2018). The local maize which is mostly cultivated by farmers in rural areas has recently been experiencing low yields. This phenomenon has been attributed to the lack of modern techniques and poor farming practices adopted in crop production by small-scale maize growers. Traditionally, their farming operations are characterised by aggressive practices such as ploughing, burning and the application of toxic chemicals which decrease soil fertility. In addition, majority of crop cultivators in rural communities engage in land intensification, a practice which has negative consequence on land productivity. All these factors contribute to low crop production. Within the past decade, for instance, production levels have suffered fluctuations (Lowder et al. 2016;Gray et al. 2018). As indicated in Fig. 1, in 2016, South Africa experienced a decline in the production of maize to about 8 million metric tons from a previous 15 million metric tons in 2014. Similarly, production dropped from approximately 17 million metric tons in 2021 to 15.3 million metric tons in 2022.
The cause of this reduction is not well known; however, experts attribute it to harsh climate conditions, inadequate land preparation, low access to quality seed varieties and lack of innovative farming practices (Dihel et al. 2018;Alvarez and Berg 2019). Some scholars do not also rule out the impact of COVID-19 pandemic on the low maize production recorded in 2021 and 2022. The restrictions on movement meant that farmers had limited access to important agriculture inputs such as fertilisers.
South Africa is one of the key producers and exporters of maize within the Africa region. The country does not only export maize to neighbouring countries but also beyond the African borders to Asia and European markets. In 2022, for instance South Africa exported about 2.47 million metric tons of yellow maize and 568,569 metric tons of white maize to Asia and European markets (Statista 2023). Amidst the increasing climate impact on agriculture productivity, maize producers may need to adopt new and improved practices if they are to sustain their production levels. To achieve this will, however, not be easy as the lack of technical expertise among maize growers especially those in rural areas continues to be a hindrance to efforts to achieving food security. To boost productivity in the coming years will mean applying innovative techniques and practices by farmers (Tolhurst and Ker 2016;Evans 2018). Against this backdrop, this study estimates the benefits of digital technology application on maize crop production with a focus on rural South Africa. In particular, the research sought to stimulate the idea that adopting digital technologies can have greater impact on agriculture productivity as well as serve as corner stone of both continues growth and poverty reduction. To achieve this underlining objective, the paper responds to the following research questions: What are the potentials of modern technologies in crop production? What are the parameters for estimating technological impact on farm productivity? How does the application of digital technologies drive crop productivity?
Answers to these questions will provide a better understanding on how modern digital technologies bring value to small-scale farmers in South Africa. The paper also examines the evidence regarding the impact of technology change on farming and poverty reduction and highlights areas of doubts, particularly relating to the role of digital technologies in agriculture growth.

The potentials of modern technologies in crop production
In view of the growing world population and the concerns about food security, the issue of improving crop yield has become more relevant in the discourse of many scholars and policy makers. From time immemorial, farmers have adopted diverse ways to boost their agriculture productivity. From the use of fertilisers to the planting of different seed species, increasing crop performance has been a fundamental objective of most farmers. Today, thanks to new innovations, the agriculture sector is also benefiting from the prospects of modern science and technology services (Duncombe 2016;Knierim et al. 2018;Rotz et al. 2019). For instance, novel digital technologies are enabling farmers the opportunity to monitor their crop growth, apply fertilisers as well as predict accurate weather (Aboh 2008;Rotz et al. 2019;Klerkx 2019).
Today's farming operations rely on sophisticated innovative tools such as sensors, robotics and aerial images and satellites to drive long-term productivity and increase profitability. The evolution of stress factors and plant diseases has given importance to the use of sensors to monitor crop growth and performance at every stage of development process (Trendov et al. 2019;McCampbell et al. 2021;Boakye et al. 2022). For instance, from the early stages of planting to harvesting, regularly monitoring of crop growth is an essential practice for timely detection of any potential threat which may affect crop yield. Digital technologies offer this possibility. For example, internet of things enabled sensors and drones do not only allow efficient monitoring of plant growth but also help gather valuable data on temperature and nutritional contents for optimal decision making. Drones in particular are considered useful in pest control and fertiliser applications in ways that do not expose plant soil to toxic chemicals.
As indicated earlier, weather has a profound influence on crop development and growth. Hence, ensuring accurate weather analysis is critical for crop performance. The good news is that farmers today have the opportunity to adopt modern tools and techniques to obtain accurate weather predictions. For instance, big data analytics as well as soil thermometer enables the collection and analysis of soil temperature, humidity, precipitation to inform decisions on when to plant, irrigate and apply pest control interventions (Thompson and Gyatso 2020;Nguimkeu, and Okou 2021). In essence, proper weather forecast and analysis allow farmers to avoid heat or frost damages thereby contributing to their crop yield.
Another important factor in the farming revolution is the invention of seed drills. Seed drills are devices used to plant seeds in an evenly and required depth to allow for proper germination. Seed drills contribute to labour efficiency in that it reduces the amount of time one spent on planting seeds manually. The use of seed drills improves crop yield significantly in that it correctly sows seeds in the right depth and distance to prevent pest invasion (Townsend 2016;Subramanian 2021). In addition, the high level of accuracy benefited from direct drilling means low cost as a result of less soil damage and wasted resources.
Finally, robotic technologies enable more reliable management and monitoring of water, air and soil quality. Robotic technologies equally give farmers greater control over crop production, processing, transportation and storage which results in high efficiencies, increase profitability and reduce environmental and ecological impact (Nakawuka et al. 2018;Ng and Ker 2019;Shah 2020). These technologies offer huge potentials for increased yield productivity if they are combined and used properly.

What does the literature reveal about technology adoption and increase in crop yield?
A growing number of studies have explored the impact of digital technologies in crop production. Most of these studies have revealed a strong linkage between technology adoption and increased agriculture productivity (Klerkx et al. 2019;Thompson and Gyatso 2020). Emerging evidence from Africa also shows how digital technologies are transforming small-scale farmers productivity and profitability by increasing efficiency, reducing vulnerability and improving access to farm inputs (Kansiime et al. 2014;Ogada and Nyangena 2015;Thomas 2020). In Kenya for example, limited access to quality fertiliser is often cited as one of the shortcomings of crop performance. To respond to this challenge, a group of innovators is using a software known as SafiOrganics to downsize and decentralise fertiliser production using locally available resources and labour. This platform is now allowing rural farmers to cut down logistical cost while producing high quality fertilisers capable of increasing their yields by 30%.
In other part of Kenya, start-up companies called UjiziKilimo and SunCulture are using big data analytics to enhance farmers' insights in irrigation, fertiliser application and pest management (Ogada and Nyangena 2015). These platforms are enabling farmers to increase their crop produce while achieving high level of efficiency.
In some rural areas in Nigeria, mobile phone applications, sensors, satellites, and radio-frequency identification are being adopted by local farmers to measure and analyse soil quality to inform planting and irrigation decisions (Aboh 2008). In Ghana, Farmerline and AgroCenta are leveraging data-driven farming practices for small-scale farmers to reduce import waste and improve crop yield (Thompson and Gyatso 2020).
Given that sustainable farming reflects the capacity to generate sufficient food in an economically efficient, socially responsible, and environmentally sound way, solutions for raising agricultural productivity are influenced by the shifting relationship between technology adoption and sustainable farming. In a study exploring the potential impact of innovation and adoption of digital agriculture within the Middle East and North Africa (MENA) region, Bahn et al. (2021) found that digital agriculture shows promise in generating high-value agricultural production as well as improving supply chain and logistics performance while ensuring optimise use of scarce natural resources in rural MENA countries.
It is well known that plant disease and pest invasion are some of the biggest problems local farmers in rural areas face. Crop damage from pests often have huge economic consequences, hence taking actions to reverse pest infestation is essential to combating this adverse impact (Edgerton et al. 2012;Trendov 2019). In East Africa, yields from smallholder crop production are typically only 20-30% of what could be produced if the best seeds, fertilisers, pest control, agronomic and water management practices are applied (Kansiime et al. 2014;Nakawuka et al. 2018). Interestingly, a start-up company called farm Africa has identified this problem and is combining farmer driven innovations with local knowledge to boost farm productivity and crop yield.
Evidence from other parts of Africa also reveal transformative results in the adoption of digital technologies in agriculture productivity. For example, in Ethiopia, a government-sponsored initiative know as Farmer Hotline is offering farmers free advisory services via interactive voice response (IVR)/short message service (SMS) on how to maintain crop health and boost crop production.
Rwanda government is also supporting smart agriculture by investing in largescale digital hardware and software systems that enable smart farming. The government has recently introduced a Crop Intensification Program (CIP) policy which is aimed at boosting agricultural productivity through an improvement of productive inputs use, irrigation coverage and soil quality.

Data sources
South Africa-Maize yield (kg per hectare) actual values, historical data, forecasts and projections were sourced from World Bank from 2011 to 2021. Maize yield, measured as kilograms per hectare of harvested land. Production data related to crops harvested for dry grain only. Crop harvested for hay or harvested green for food, feed, or silage and those used for grazing were excluded. The traditional production variables include yield (per hectare of farm), labour, fertiliser, seed, and irrigation.
The technology variables considered in this study included farmer-preferred adaptation technologies. Environmental variables were also included to capture effects of climate variability on the mean and variance of maize crop production. These included the climate satisfaction index of the preceding main agricultural season.

Estimating the contribution of technological progress in crop yield
To accurately estimate the contribution of technological progress in crop yield for small-scale farming is very challenging due to the highly heterogeneous performance of crop within a given plot. In most cases, the approach to measure the impact of technology is conditioned on observing yield trend relative to adoption of particular technology over a period of time. Several scholars have suggested that the best approach to use is the one based on change in total output over a given time period per the level of technological change (Adrian 2012;Tolhurst, and Ker 2015;Park et al. 2019). Their argument is that if a researcher chooses a base period (t o ) by which a farmer operates at a near equilibrium output and adopt new techniques to achieve a new level of output in the second year (t 1 ), then the difference represents the contribution which technological change has made to outputs between (t o ) and (t 1 ). This approach is based on the concept of Cobb-Douglas linear production function. For instance, if a production function is constructed for a base year and a technological progress is adopted in a given period, one can estimate the influence which technological change has made to the change in output between the two periods by the difference between the index of output actually produced in the given period and the index of output estimated from the base period production function (Ng and Ker 2019).
The Cobb-Douglas production function is given by where Q j,t is the output of the vessel j in period t and X j,i,t and X j,k,t are the variable and fixed vessel inputs (i,k) to the production process. As noted above, the error term is separated into two components, where v j,t is the stochastic error term and u j,t is an estimate of technical inefficiency. Other methodologies have also been suggested by different authors in estimating technological change in crop yield. One common approach usually mentioned is the deterministic approach. This approach measures output changes given capital stock inputs; nondiscretionary stock inputs and existing technology (Ramírez and McDonald 2006;Renard et al 2013). In this approach, the output level is estimated using linear programming procedures and is interpreted as the output that could be produced with full and efficient utilisation of the variable input(s), given the capacity base.
This is expressed as: ln Y t= λ j + σ 2 j + β 1 N it−1 + 2 H it−1 where the unknown parameters λ j , σ 2 j and functions hj (t) are estimated with a maximum likelihood approach using the heuristic EM algorithm for the j components of the mixture. A limitation of the deterministic approach is that it does not quantify uncertainties as part of its production decline evaluation.
Another approach is stochastic approach which uses probability simulator to perform multiple iterations. It conducts outputs estimation through modification of the inputs incorporated in the production (or distance) function. A potential advantage of the stochastic production frontier approach over DEA is that random variations in catch can be accommodated, so that the measure is more consistent with the potential harvest under "normal" working conditions. A disadvantage of the technique is that, although it can model multiple output technologies, doing so is somewhat more complicated, requires stochastic multiple output distance functions, and raises problems for outputs that take zero values. The stochastic model is given as where q j is the output produced by firm j, x is a vector of factor inputs, v j is the stochastic (white noise) error term and u j is a one-sided error representing the technical inefficiency of firm j. Both v j and u j are assumed to be independently and identically distributed with variance 2 v and 2 u , respectively. It should be noted, however, that, whatever method used for crop yield forecasting and estimation at any level of aggregation, it is important to distinguish between accuracy claims and a publicly available and independently verifiable track record between the forecast and the final yield estimate after harvest. Within the context of smallholder crop production, standardising methods for yield estimate is crucial to obtaining accurate date and determining the suitability of farming practices under different environment (Hanuschak 2013). This is because a standard methodology allows for identifying any trade-off in crop performance after adopting a particular technology.
It is equally important to stress that technology adoption may also contribute to the efficiency of other factors of production. It may be, therefore, inconsequential to attribute crop yield performance to only improved technology. For instance, how can one be certain that differences in crop yield is as a result of the sole application of a particular technology? To address this uncertainty or casual inference requires the use of experimental date. Experimental data are gathered through a process of active intervention to produce and measure change when a variable is altered. This data usually allows the researcher to determine a causal relationship and is typically projectable to a larger population. However, without experimental data, there are two problems a researcher may have to contend with. One is self-selection, which has an element of bias and the second is relativity, which does not account for unobserved characteristics. Hence, for better estimation of the impact of technology on crop performance, it is advisable to combine different methods (Ramírez andMcDonald 2006).

Research framework and model specification
This study employed the stochastic production frontier framework. This model is applied to study the technical efficiency of production inputs in relation to outputs. The model operated from the assumption that for any given bundles of inputs, the production process faces two random disturbances: either negative or positive with different characteristics. The stochastic production frontier model was selected for this analysis because of its consideration for technical inefficiencies and symmetric errors in explaining variance of output. The analysis specifies and adopts the Cobb-Douglas production function to estimate the relationship between variance of yield in relation to aggregate inputs. The model is given as where Y denotes the outputs, X represents the explanatory variable, μ denotes the estimated coefficients, f denotes deterministic variable μ the disturbance term with a zero mean, h the variance function, and ε denotes the error term. Table 1 shows how the parameters used vary from the average (mean).
The stochastic production frontier model is estimated using simple Ordinary Least Square (OLS), maximum likelihood estimation (MLE) and probit regressions. The analysis adopts a three-stage procedure. It begins with the specification of regression model for both the predicted efficiency in technology application and the level of production efficiency with respects to the explanatory variables and random disturbance. The parameters of the model at this stage employ the simple Ordinary Least Square (OLS). The second stage involves calculating least residuals square and in the third stage, we estimate the variances in the functional forms of output levels using linear, logarithms and exponential functions. The decision to add logarithm function in our estimation is based on the fact that it helps improve normality of dependent variable and residuals, thus controlling for any possible outliers. Throughout this procedure, we apply propensity score matching (PMS) to control for bias as a result of observable heterogeneity. The explanatory variables included farmer-preferred technologies which were correlated with the observed crop yield. The technology variables considered in this study are both mean precipitation and mean temperature.  Table 2 presents a summary of the set of new digital technologies used by farmers in South Africa. Majority of the maize farms and farmers were in areas where connectivity is low making full digital integration in real time challenging. Nevertheless, the analysis revealed that most farmers rely on drones, sensors and aerial images among others in their farming practices. For instance, drones help farmers to scout and monitor their crops while sensors, seed drills and aerial images are used to assess soil temperature, plant seeds and gather data, respectively. The output revealed that data are more clustered to the mean as low variation can be observed. Econometric results on regression coefficients of corn yield data are presented in Tables 3 and 4 for the probability and variance functions of corn production in general. Land size (acre), kernels per ears, kernels per rows and total corn harvested showed positive and significant impacts on the mean of crop output. In terms of estimating the average crop yield with respect to the rates of digital technology application as depicted in Table 4, land cultivation size showed the largest production elasticity among the technologies applied (p value of 0.003). Technology effects on yield variability also differed with regard to the frequency of technology usage, duration and crop quantity, all indicating significant positive coefficients.

Estimating the impact of technologies on mean and variance of crop yield
Results as shown in Table 5 depict the varying effects of the various technologies by variables. Assessing the effect of technological variables on the mean and variance of crop production indicated that plant health, soil nutrition, crop density, pest control significantly and positively affected the mean yield variability. The effects, however, varied across the value of crop yield. For instance, with respect to how technology adoption on improves plant health (p value = 0.024) and pest control (p value = 0.012) in relation to yield volumes, there was a significant positive correlation between adoption and increased crop yield. On the other hand, the results are not statistically significant for soil nutrition (p values 0.280 and crop density (p values = 0.41) at a 5% significance level. Crop density variation in relation to yield volumes was generally rated less effective than soil nutrition. These results suggest that technological adoption does not necessarily affects the factors that contribute to change in crop yield. Nevertheless, the analysis indicates an overall positive relation between technology application and crop yields. That is in the long run, an improvement in plant health, pest control and soil nutrition driven by technological adoption leads to an increase in crop yield.

Discussion and policy implications
Technology is considered a vital component in the production function by neoclassical theorists. The neoclassical theory which forms the framework of this study is associated with Robert Solow and Koopmans (1999). The theory posits that growth is dependent on labour force, technology, and the total capital stock of a country. To measure the growth and equilibrium of an economy, the theory adopts the concept of Cobb-Douglas production function. This concept was adopted in the current study to underscore the net positive effects that new technologies and digital infrastructures have on crop productivity. Several studies in the past have explored the potential benefits of the use of digital technologies in agriculture, yet few have focussed on how it relates to smallholder farmers in Africa and its contribution to poverty reduction. By concentrating on smallholder maize crop producers in South Africa, this study relied on panel data to examine the nexus between the application of digital technologies and improvement in crop performance.
While the results from this study confirm findings of other past studies, it distinctively reveals a more versatile and comprehensive approach to the analysis. For instance, it captures the effects of climate variability on the mean and variance of maize crop production by smallholder farmers in South Africa. The study revealed two critical findings in relation to farmer-preferred technologies. First, results show that drones, sensors, seed drills, smart irrigation and aerial images are among the top new technologies mostly deployed in farming practices in rural areas. Second, farmers adopt changes in cultivation practice with respect to the use of these technologies only when they have seen evidence of its positive net effect on production. This particular finding is in line with findings reported by Rotz et al. (2019) whose study explored the politics of digital agricultural technologies. It also reflects results published by Knierim et al. (2018) in their study on what drives adoption of smart farming technologies.
The statistical output also makes a strong case for technology adoption as having pronounced effects on aspects of yield. However, the study revealed that technology effects on yield variability may differ with regards to the frequency of technology usage and duration, all indicating significant positive coefficients. For instance, in places where seed drills and smart irrigation techniques are often used in cultivation practices, crop productivity appears to be high. In contrast, where these technologies are barely used, productivity seems low. This particular result is consistent with findings reported by Aboh (2008) in a study assessing the frequency of ICT tools usage by agricultural extension agents in Imo State, Nigeria. In general, this study has shown a positive correlation between the application of digital agriculture technologies and increased crop performance as also reported in other studies (Evans 2018;Klerkx et al. 2019;McCampbell et al. 2019;Nguimkeu and Okou 2021).
Despite these findings, the analysis has shown that adoption of digital agriculture technologies by itself does not necessarily contribute to increased productivity. This means there are other basic factors such as farm size, quality of seeds and better farming practices among others that influence yield performance. Further studies on assessing the contribution of digital technologies on crop yields will, therefore, require complex research technique.
Findings from this study are subject to the usual limitations of regression analysis which include the issue of attribution. Other limitations were related to scope and data availability. The study scope was limited due to time and resources. In addition, data on crop performance from farmers were difficult to come by. These limitations need critical consideration in future studies. These limitations notwithstanding, the outcome of this research is reflective of the general situation and condition in South Africa rural areas where maize cultivation is a predominant economic activity. Hence, the weaknesses identified did not undermine the study's relevance and outcome.
The policy inference from this study is that digital technology is increasingly becoming an important solution to the many of the challenges in the agriculture sector. They are allowing farmers to maximise production capacity. The evidence is also clear among commercial farmers in South Africa. Meanwhile among smallholder farmers, the adoption of technologies is still slow. To increase the uptake of smart farming technologies among local farmers will require the design of policies and programs that increase farmers ICT skills and ensure availability, accessibility and affordability of digital techniques. Policy makers should also create enabling environment that will allow the private sector to invest in agriculture smart technologies to increase its availability to small-scale farmers. Findings from this study also highlight the importance of increasing digital inclusion amongst farmers as a measure to increase their visibility in the wider agriculture value chain system.

Conclusion
This study sought to estimate the benefits of digital technology application on maize crop production with a focus on rural South Africa, using panel data from 2011 to 2021 from World Bank's open data source. The study employed stochastic production frontier framework which assesses technical efficiencies of production inputs in relation to outputs. The paper further adopted the Cobb-Douglas production function to estimate the relationship between variance of yield in relation to aggregate inputs. The study revealed two critical findings in relation to farmer-preferred technologies. First, results show that drones, sensors, seed drills, smart irrigation and aerial images are among the top new technologies mostly deployed in farming practices in rural areas. Second, farmers adopt changes in cultivation practice with respect to the use of these technologies only when they have seen evidence of its positive net effect on production. The statistical output further showed a strong positive relation between technological adoption and yield increase. Notwithstanding these findings, it was noted that the mere adoption of digital agriculture technologies by itself does not necessarily contribute to increased productivity. Therefore, to assess to holistic impact of digital technologies usage on crop performance, it is important to consider factors such as farm size, quality of seeds and better farming practices among others. Against this backdrop, it is recommended that future studies exploring the nexus between digital technology application and crop performance must use larger sample with complex research technique.