The data used in this study is from a field survey conducted by the research group in Honghu City and Qichun County, Hubei Province, China, in 2022, targeting the economic behaviour of farmers. The survey mainly includes basic information on the population of farmers; a family's natural, social, financial, and material capital; land use situation, production and operation; and household income and expenditure.

To ensure the representativeness of the survey samples, stratified and probabilistic random sampling methods were used to select the samples. Firstly, the sample cities and counties were identified. The research locations were Honghu City, Jingzhou City, Hubei Province, China and Qichun County, Huanggang City. The terrain of Qichun County is narrow and long, with high terrain in the north and low terrain in the south. The terrain is complex, with a mountainous area in the north, a hilly area in the middle, and a flat fan-enclosed area in the south. The altitude ranges from 12 metres to 1244.1 metres. The entire territory of Honghu is located in the Yangtze River floodplain, with a gently sloping terrain from northwest to southeast, forming a wide and flat terrain with high north-south and low middle. The altitude is mostly between 23 and 28 metres, and the agricultural land is 2.4872 million acres, accounting for 66.49 % of the land area of Honghu City. The agricultural planting area is wide. The terrain of the two regions includes plains, hills, and mountains, covering almost all the terrain of Hubei Province, which to some extent represents the agricultural planting status of Hubei Province. The second is to determine the sample towns. Based on the terrain distribution and economic development status of the sample cities and counties, three sample towns, Laowan Hui Township, Wulin Town, and Yanwo Town, were randomly selected in Honghu City. Six sample towns, Datong Town, Liuhe Town, Qizhou Town, Qingshi Town, Tanlin Town, and Zhulin Town, were randomly selected in Qichun County, and 9 sample towns were ultimately obtained. The third is to determine the sample village. Mainly considering the terrain conditions, economic development level, and distance from the town centre of each village, 2–9 sample villages were selected from each sample town, totalling 38 sample villages. Finally, sample farmers were selected. Each village randomly selected 30–40 households to conduct a survey, distributed 1100 questionnaires, and collected 1050 questionnaires. After removing missing data samples, 1046 valid samples were obtained.

Table 1 shows the distribution of samples. This paper also classifies rural households according to their income into three categories: low-income, middle-income, and high-income. Amongst them, farmers with per capita household income of less than 8000 yuan are low-income, those with per capita household income of 8000 yuan to 30,000 yuan are middle-income, and those with per capita household income of 30,000 yuan or more are high-income.

Table 2 shows the explained variables are whether farmers promoted technological innovation in grain production and degree of technological innovation in grain production. Digital Technology Adoption were taken as the core explanatory variables. In addition, 11 control variables, including agricultural machinery service expenditure, age of the head of household, education level of the head of household, health status of the head of household, per capita income of the family, family size, average age of family members, average education level of the family, labour force proportion, training labour force proportion, and per capita land area of the family, were introduced.

Table 3 shows descriptive statistics of the variables. Amongst the 1046 sample families, 29.45 % of farmers improve their grain production technology innovation. In the survey sample, most farmers spend less on agricultural machinery service, and the difference in agricultural machinery service expenditure is large. The average age of the head of household is 59 years old, and the overall health is relatively good. The average education level of household heads has not exceeded the nine-year compulsory education level. Most are concentrated in primary and junior middle schools, and their education level is still low. The per capita income of most of the rural households in the sample is at a moderate or high level. The average family size of the sample was 4.61, and the average age of the family members was 42.64 years old. The average educational level of households is 3.26, which does not exceed the nine-year compulsory education level, indicating that the educational level of rural households in the study area is low. The proportion of the labour force and the proportion of training labour force are 56.31 % and 42.51 % respectively, indicating that the coverage of skills training of the rural labour force in the study area is not high. The average per capita household land area was 1.98 mu, which is higher than the national per capita farmland area of 1.79 mu.

Technological innovation in grain production is introduced as the explained variable to verify the research hypothesis of this paper. A core explanatory variable, data technology application, is introduced to ‘Is the Internet of Things, artificial intelligence, drones and other digital technologies used in production to improve the production management process and achieve precision production?’ to invert the variable. Based on control variables, including individual characteristics of household head, basic characteristics of the household, and household livelihood capital (human capital, natural capital, physical capital, social capital and financial capital), an analytical framework and index system of the relationship between digital technology and grain production were constructed. The first explained variable in this study is whether to improve the technological innovation of grain production Y1, assuming that the technological innovation of grain production is 1, and the technological innovation of grain production is 0. The dependent variables are 0 and 1, and the model is binomial Logistic. Therefore, the Logistic model form for studying the impact of digital technology on grain production is:

In the above formula, Pi is the probability of technological innovation improvement of the I-th household grain production; X1i is the core variable, the rest is the control variable, α, β1, β2,… β18 is the variable to be estimated, and ɛi is the disturbance variable. The second explained variable in this study is grain production technological innovation Y2, and the multiple linear regression model is constructed as follows:

In the above formula, Y2 is the explained variable, X1 is the core variable, and the rest are the control variables. α, β1, β2,… β18 is the variable to be estimated, and ε is the disturbance variable.

Table 4 shows the results of binomial Logistic full sample regression and income-based sub-sample regression of digital technology and control variables on technological innovation improvement of grain production. In the full sample regression results, the regression coefficient of the core explanatory variable digital technology on the improvement of grain production technology innovation is 0. 8362, which is significantly positive at the 1 % statistical level, indicating that digital technology has a significant improvement effect on grain production technology innovation, and hypothesis 1 has been verified. Amongst the control variables of the whole sample, the regression coefficient of agricultural machinery service expenditure on grain production technology innovation is −0.0516, which is significantly negative at the 5 % level, indicating that farmers with more agricultural machinery service expenditure have less probability of improving grain production technology innovation, and the technical effect of agricultural machinery technology has been verified. The age of the head of household and the health status of the head of household do not have a significant effect on the improvement of technological innovation in grain production, indicating that the improvement of technological innovation in grain production has no strong correlation with the age and labour ability of the head of household. The education level of the household head, the average education level of the household and the proportion of trained labour force have a significant negative impact on the improvement of grain production technology innovation. The possible explanation is that farmers with higher education levels have higher insights and visions, a better ability to optimise the allocation of household production factors, and higher agricultural production efficiency. If farmers receive agricultural skills training, their agricultural production skills will be enhanced, positively impacting agricultural production efficiency. The per capita household land area positively impacts the improvement of grain production technology innovation, which may be due to the large and scattered land and the imperfect land transfer system, which can not be completely transferred and rationally utilised.

According to the regression results of the sub-samples of low-income, middle-income, and high-income households in the table, digital technology has a significant positive impact on the technological innovation of grain production in the three types of household samples. Amongst them, low-income households’ digital technology has the most significant impact on grain production technology innovation, and high-income households’ grain production technology innovation is least affected by digital technology. In low-income households, the impact of agricultural machinery service expenditure on grain production technology innovation is not significant, which may be due to the small production scale of low-income farmers, the small demand for agricultural machinery services, and the small technical effect of agricultural machinery technology on low-income households. In the sample of middle-income and high-income households, the expenditure on agricultural machinery services has a significant negative impact on the technological innovation of grain production at the level of 5 %. The resource endowment of middle-income and high-income farmers is better than that of low-income farmers, the scale of agricultural production is relatively large, and the demand for agricultural machinery services is also large. The use of agricultural machinery services improves agricultural production efficiency. In summary, if other conditions remain unchanged, farmers’ technological innovation in grain production is significantly different with the difference in the degree of adoption of digital technology, and the adoption of data technology can improve technological innovation in grain production. Simultaneously, the expenditure of household agricultural machinery services, the education level of the household head, the average education level of the household, the proportion of trained labour force, the per capita income of the household and the per capita land area all have an impact on the technological innovation of grain production to a certain extent.

Table 5 shows the results of multiple linear full sample regression and income-based sub-sample regression of digital technology, control variables and technological innovation in grain production. In the whole sample, the regression coefficient of digital technology on grain production technology innovation is 0.3167, which is statistically significant at the 10 % level, indicating that the larger the grain production technology innovation of farmers adopting digital technology. Thus, hypothesis 2 is verified. In the control variables of the whole sample, the regression coefficient of agricultural machinery service expenditure on grain production technology innovation is −0.0827, which is significantly negative at the level of 1 %, indicating that farmers with more agricultural machinery service expenditure will have less grain production technology innovation, and the technical effect of agricultural machinery technology has been verified. The age of the head of household and the health status of the head of household had no significant influence on the technological innovation of grain production, indicating that the technological innovation of grain production had no strong correlation with the age and labour ability of the head of household. There is a significant negative correlation between the education level of the head of household and the average education level of the family and the technological innovation of grain production. The possible explanation is that farmers with higher education levels have higher insights and visions, stronger ability to optimise the allocation of household production factors, higher agricultural production efficiencies, and greater use of land resources. The per capita household land area has a significant positive impact on the technological innovation of grain production, possibly due to the large and scattered land and the imperfect land transfer system.

According to the regression results of the sub-samples of low-income, middle-income, and high-income households in the table, digital technology has a significant positive impact on the technological innovation of grain production in the three types of household samples. Amongst low-income households, agricultural machinery service expenditure significantly correlates with technological innovation in grain production. The reason may be that low-income farmers have small production scale and poor natural resource endowments, and it is difficult for agricultural machinery technology to produce scale effects on their agricultural production, thus reducing technological innovation in grain production. In the samples of middle-income and high-income households, agricultural machinery service expenditure has a significant negative impact on the technological innovation of grain production. The possible explanation is that middle-income and high-income farmers have better resource endowments than low-income farmers, relatively large-scale agricultural production, and greater demand for agricultural machinery services. The use of agricultural machinery services improves agricultural production efficiency to encourage greater use of land.

In summary, other conditions remaining unchanged, the higher the degree of digital technology adoption, the higher the technological innovation in grain production. Simultaneously, the agricultural machinery service expenditure, the education level of the household head, the per capita income of the household, the average education level of the household, the per capita land area of the household and other factors will have an impact on the technological innovation of grain production to some extent.

To ensure the robustness of the research hypothesis in this paper, a Tobit regression model was constructed to analyse the index system mentioned above. The Tobit regression model is as follows:

In the above formula, Y represents the technological innovation of grain production and whether the technological innovation of grain production is improved, Xi represents the digital technology; control is a series of control variables; and a and ɛi represent constant and random disturbance terms, respectively.

The regression results are shown in Table 6 below. Digital technology has a significant positive impact on the technological innovation of grain production at the statistical level of 1 %; that is, farmers who adopt digital technology have improved the technological innovation of grain production, which again verifies the research hypothesis 1. There is a significant positive correlation between digital technology and grain production technology innovation at the level of 10 %, and the higher the adoption of digital technology, the higher the grain production technology innovation, which again verifies the research hypothesis 2. Tobit regression verifies the hypothesis again, and the conclusion is robust.

The above model parameter estimation results only test the impact of digital technology on technological innovation of grain production, and the specific contribution rate of digital technology on technological innovation of grain production, which factors are the key factors affecting technological innovation of grain production, and the degree of barrier effect, income effect and technological effect of digital technology on technological innovation of grain production. These need to be measured further. Based on the estimation results of models 1 and 2, the regression decomposition method improved by Wan (2002) was used to estimate the contribution rate of each variable to the technological innovation of grain production. Specific calculation methods and models are as follows:

Amongst them, Vulirepresents the grain production efficiency of farmers, α represents a constant term, Xi represents an explanatory variable, and ɛi represents a residual term. After estimating the parameters of Eq. (4), the estimated value of βi is obtained, and then the estimated value Vuli of Vi^ and the estimated value Vi^′ of Vuli without considering the constant term is calculated, namely:

Then, regression decomposition is performed in the following. Step 1: Use CV(*)to represent the coefficient of variation and calculate the contributions of residual and constant terms to Vuli, Cɛ and Cα:

• Step 2: Calculate the contribution of each variable to CV(V^). The contribution of each variable to CV(V^) can be calculated using the Sharply valued theoretical decomposition proposed by Shorrocks (2013). Generally, different farmers have different values for x, and replacing x with the sample mean of Xi can eliminate the difference in Xi. The calculated V value after replacement is denoted as Vi (excluding constant terms), thus obtaining the coefficient of variation CV(V), which depends on the difference after x eliminates Xi. Similarly, replacing Xi and Xj with the sample mean of Xi and XJ can eliminate the difference between Xi and Xj. The calculated V value after replacement is denoted as Vi, resulting in the coefficient of variation CV(V). Through analogy, more differences in X can be eliminated.

Using Cimm to represent the n contribution of variable i to the coefficient of variation in the m order, the calculation formulas for each time are as follows:

The contribution of variable i to the coefficient of variation at the m th time is:

In Eq. (27), Cim represents the contribution of variable i to the coefficient of variation at the m time. The contribution of variable i to the difference in the impact of technological innovation on grain production is as follows:

• Step 3: Calculate the contribution rate of each variable to the technological innovation of grain production.

  
  
  

Where CDɛ, CDα, and CDi represent the contribution rates of residual term, constant term, and variables to the difference of technological innovation in grain production, respectively.

Table 7 shows the contribution rates of residual term, constant term, and explanatory variables to technological innovation in grain production. Although the regression decomposition method cannot identify the contribution of unincluded explanatory variables, these unidentified variables can be reflected in the residual term. In this paper, the contribution rate of the residual term to the technological innovation of grain production is 27.89 %, and the contribution rate of explanatory variables to the technological innovation of grain production is 78.89 %. According to the sample data of different groups, the contribution rates of residuals and explanatory variables to the impact of technological innovation on grain production are roughly the same: the contribution rates of residuals and explanatory variables to the impact of technological innovation on grain production of low-income households are 23.85 % and 81.57 %, respectively. The contribution rates of residuals and explanatory variables to the impact of technological innovation on grain production were 28.39 % and 78.60 %, respectively. The contribution rates of residuals and explanatory variables to the impact of technological innovation in grain production of high-income households are 28.87 % and 77.43 %, respectively. Therefore, even if unobservable factors are not included in the regression model, the explanatory variables selected in this paper can still explain the impact of technological innovation on grain production to a large extent.

Further, the regression decomposition method is used to estimate the contribution rate of each explanatory variable to the impact of technological innovation on grain production, and the results are shown in Table 8. It is not difficult to identify in Table 8 that amongst all factors, digital technology contributes the most to the impact of technological innovation on grain production, reaching 21.54 %. The results of subgroup samples showed that the contribution rate of digital technology in low-income, middle-income and high-income households to the impact of technological innovation in rural grain production was 20.45 %, 22.69 %, and 21.65 %, respectively. Amongst all explanatory variables in each group of samples, the contribution rate of digital technology on the impact of technological innovation in grain production was the largest. This further validates the conclusion that digital technology is the main factor affecting technological innovation in grain production. In addition to digital technology, other factors have a more significant impact on grain production technology innovation, amongst which is the contribution rate of per capita household land area to the impact of grain production technology innovation reached 43.65 %, second only to digital technology. Simultaneously, the family's per capita net income, the family's size, the family's average education level, and the proportion of the labour force also affect the technological innovation of grain production to a certain extent. Agricultural machinery service expenditure, age of the household head, education level of the household head, average age of family members, average education level of the household, and the proportion of trained labour force have negative contribution rates to the impact of technological innovation in grain production, indicating that these factors hinder technological innovation in grain production and can promote the effective use of land. The results of grouped samples also support this finding.

The adoption of digital technology has a significant impact on technological innovation in grain production, and the adoption of digital technology can enhance the level of innovation in grain production. This article's descriptive statistical analysis and empirical analysis support this conclusion. It can be seen that the adoption of digital technology has a significant driving effect on the innovation of grain production technology. It is necessary to further promote the digital adoption level of grain planting farmers to enhance the innovation of grain production technology and ensure the stability of global grain production. Meanwhile, this research also found that different income types of households have differences, and the adoption of digital technology has the greatest impact on the innovation of grain production technology for low-income farmers, followed by middle-income households and a minor impact on high-income farmers. Therefore, how to promote the digital adoption of low-income farmers is the focus of future digital technology promotion. Essentially, innovation in grain production technology is the result of multiple factors such as digital technology, agricultural machinery and technology promotion, farmers’ income, and skill training. This article further demonstrates the impact of digital technology adoption on technological innovation in grain production. The results of the regression decomposition method show that digital technology is the main factor affecting technological innovation in grain production. The application of digital technology in grain production has significantly improved technological innovation. Additionally, the barriers and limitations of digital technology on innovation in grain production technology can be overcome by improving the level of agricultural machinery services and promoting agricultural technology training. It is necessary to attach importance to the adoption of digital technology and take diversified measures to enhance the innovation level of grain production technology and to improve the innovation level of grain production technology amongst farmers.

China is a major agricultural country, and grain production is an important part of agriculture. However, with the rapid development of China's economy and the acceleration of industrialisation and urbanisation, digital technology has promoted increased grain production. Innovation in grain production technology has become an important means to improve grain production. It is particularly important to analyse the impact and mechanism of digital technology on innovation in grain production technology to promote the effective utilisation of digital technology and improve the efficiency of grain production. This study empirically analyse the impact of digital technology on technological innovation in grain production, which can provide some beneficial insights for improving the application level of digital technology, improving grain production systems, and optimising agricultural policies, as shown below:

In recent years, the importance of data technology in various industries has become increasingly evident. One sector that significantly benefits from the advancements in data technology is grain production. With the increasing global population and the need to ensure food security, data technology has played a pivotal role in improving grain production efficiency, productivity, and sustainability.

One of the critical aspects where data technology significantly impacted is precision agriculture. Precision agriculture involves using data-driven tools and technologies to optimise crop production. Farmers can monitor the health and growth of their crops through satellite imagery, drones, and sensors in real-time. This data allows them to make informed decisions regarding irrigation, fertilisation, and pest control, leading to better yields and cost savings.

Data technology also enables farmers to utilise predictive analytics to anticipate potential issues and make proactive decisions. Farmers can predict crop diseases, yield fluctuations, and market demand by analysing historical data, weather patterns, and soil conditions. This helps them manage risks, plan for contingencies, and maximise profitability. Moreover, data technology allows for optimal resource allocation, ensuring that inputs such as water, fertilisers, and pesticides are used efficiently, reducing waste and environmental impact.

The importance of data technology in grain production is further highlighted by the concept of smart farming. Smart farming combines various data-driven technologies, such as artificial intelligence, Internet of Things (IoT), and robotics, to create an interconnected and automated farming system. This allows for real-time monitoring, decision-making, and control of farming operations, leading to increased efficiency, reduced labour costs, and improved productivity.

In conclusion, data technology has significantly enhanced grain production by enabling precision agriculture, predictive analytics, supply chain management, research and development, and smart farming. The use of data-driven tools and technologies has revolutionised how farmers approach grain production, leading to increased efficiency, improved productivity, and sustainable practices. As the global population continues to grow, the importance of data technology in ensuring food security and meeting the demands for grain production cannot be overstated.