The impact of energy structure on agricultural green productivity in China

Abstract

The agricultural development in China is currently grappling with the challenges of limited water and soil resources, environmental pollution, and ecosystem degradation. Therefore, improving agricultural green productivity has become the focus of academic attention. Based on the panel data of 30 provinces in China from 2008 to 2020, this paper constructs an index system of agricultural productivity. Subsequently, the super-efficiency SBM model is used to calculate the green productivity of agriculture, and its characteristics are analyzed from the perspectives of temporal and spatial.Through the empirical test of threshold model, it is found that China’s agricultural green productivity shows a gradual growth trend from 2008 to 2020, and the growth rate has accelerated significantly in recent years. There is a non-linear relationship between energy structure and agricultural green productivity. Especially in areas with reasonable energy structure, the increase in the proportion of coal consumption will inhibit the improvement of green agricultural productivity. This paper suggests reducing dependence on fossil energy, improving market mechanism and optimizing resource utilization in order to promote green agricultural production in China. This study expands the research perspective of agricultural green productivity, provides new research methods and ideas for related fields, and provides a reasonable energy allocation plan for agricultural production, which helps to improve production efficiency and reduce costs.

Introduction

Green production is not only related to environmental protection and sustainable utilization of resources, but also to many aspects of economic, social and global development. Green growth in most developed countries is positive, while efficiency change is negative. Technological progress is the main driver of economic and environmental development¹. With the acceleration of China’s economic transformation and modernization, China’s agricultural green production has gradually attracted attention and made some progress. Green agricultural production can reduce the negative impact of agricultural production on the environment and reduce the pollution and destruction of soil, water sources and ecosystems. At the same time, through scientific management and resource utilization, we can improve the productivity of farmland, reduce the excessive development and consumption of natural resources, and create good conditions for the long-term and stable development of agriculture.

However, in the process of promoting green agricultural development, in addition to paying attention to economic benefits and social impacts, we must also fully consider the restrictive factors of energy structure. With the increasing challenges brought about by global climate change and the increasing requirements for emission reduction targets at the national level, how to rationally plan and optimize energy use in agricultural production has become one of the urgent problems to be solved.China’s energy mix is particularly special. From Figure 1, China’s energy structure is relatively special. As a country with coal as its main energy supply, although abundant coal resources provide strong support for economic development and agricultural production, they also bring great pressure to environmental protection. Therefore, it is particularly important to explore low-carbon or even zero carbon agricultural models. This not only requires the government to introduce relevant policies to support the research and development and application of clean energy technologies, and encourage farmers to use renewable energy such as solar energy and wind energy for irrigation or greenhouse heating; It is also necessary to strengthen the requirements for energy efficiency standards of existing machinery and equipment, promote energy-saving agricultural machinery and reduce fossil fuel consumption. Only by organically combining economy, society and environment can we truly achieve the long-term healthy and stable development of agriculture.

Fig. 1

Energy consumption structure of China from 1980–2022 (unit:%). Note: For different energy types, the calculation method of coal consumption for power generation is used to unify the energy unit—data sources: China Statistical Yearbook and China Energy Statistical Yearbook from 1980 to 2021. The consumption of clean energy, such as nuclear, photovoltaic, and wind energy, is not separately provided in the yearbook. Except for coal, oil, and natural gas, almost all energy sources are clean, so they are classified under the same category.

Theoretically, the energy structure will affect agricultural green productivity. Is the relationship between the energy structure and agricultural green productivity linear or nonlinear? How much does the energy structure affect the green productivity of agriculture? How can we further promote green agricultural production through energy policies? In order to answer these questions, based on the calculation of agricultural green productivity in 30 regions of China, this paper uses a double threshold model to verify the relationship between energy structure and agricultural green productivity.

Figure 2 shows the technical roadmap studied in this article. The remainder of this paper is organized as follows: Section 2 provides a literature review and theoretical hypotheses on related topics. Section 3 explains the research methods and variables, and makes a descriptive statistical analysis. Section 4 reports the level and spatial characteristics of agricultural green productivity; and establishes the threshold model, describes the variables and data sources and empirically tests the impact of the energy structure on agricultural green productivity. Section 5 provides the conclusion, policy recommendations and future research directions of this article.

Fig. 2

Technical Roadmap.

Literature review and theoretical hypothesis

Literature review

Literature review on the energy structure

Han et al.² used gray correlation and panel regression models to explore the evolution and influencing factors of the energy consumption structure in major scientific and technological powers over the past 30 years. The evolution of the energy consumption structure showed the characteristics of decarbonization, cleanliness, and security. Many factors affect the energy structure, such as urbanization³,industrial structure⁴, green finance⁵, and household energy consumption⁶, which are highly correlated with the proportion of fossil energy consumption in various countries, but the most significant correlation factors in different countries are different. Furthermore, Nghiem⁷ used the autoregressive distribution lag model (ARDL model) to analyze data from 1985 to 2013, and the results showed that carbon dioxide emissions, GDP growth, energy consumption, and industrialization were cointegrated and had a long-term equilibrium relationship. Chen et al.⁸ conducted an empirical study on the impact of energy regulation on the energy intensity and energy structure of manufacturing enterprises from 2003 to 2009. Stricter energy regulation led to a significant reduction in energy intensity. Doggart et al.⁹ examined how national energy policies affected household energy use for cooking from 1990 to 2018 and how energy policies made further progress in achieving national and global energy reduction goals. Wan et al.¹⁰ used the entropy weight method to calculate the level of green development. Green development promoted an increase in the proportion of clean energy consumption and further optimized the energy consumption structure. Magazzino et al.¹¹ argued that accelerated urbanization and economic activity can produce greenhouse gas emissions. They also discussed the new meaning and role of agricultural land use in sustainable development goals.

Literature review on agriculture green productivity

Measuring agricultural green total factor productivity is essential for evaluating agrarian production efficiency and environmental sustainability. In recent years, many scholars have conducted in-depth research on the measurement and decomposition of agricultural green total factor productivity. Andersen et al.¹² subdivided the traditional DEA model and proposed a superefficiency DEA model for comparative analysis of various decision-making units. Guo et al.¹³ used the global GML index to calculate the growth rate of agricultural green total factor productivity. This paper used the global Moran index to analyze spatial correlation and absolute correlation β convergence, conditions β convergence, and dynamic space convergence test space-time convergence. Chung et al.¹⁴ introduced the directional distance function and proposed the Malmquist‒Luenberger (ML) model, which can simulate multiple input and output processes while considering environmental factors and has become an ideal model for analyzing the low-carbon economy. Li et al.¹⁵ used the global ML index of the mixed distance function to account for agricultural green total factor productivity. On this basis, the Dagum Gini coefficient, kernel density estimation, and spatial Markov chain methods were used to examine the evolution of the regional gap and distribution dynamics of its growth. Based on indicators constructed from the perspectives of competitiveness, rural viability, environment, and expenditure equality, Uthes et al.¹⁶ assessed the implementation of rural development measures. The agricultural productivity increased significantly in the more competitive experimental group. Du et al.¹⁷ used the inventory analysis method to account for agricultural nonpoint source pollution in Dongting Lake. Then, they incorporated it into the agricultural total factor productivity (TFP) analysis framework to measure the temporal and spatial changes in agricultural green total factor productivity (GTFP) in Dongting Lake. Najafabadi et al.¹⁸ proposed the interval meta goal programming model (IMGPM) to resolve the conflict between agricultural water use and sustainable water use. This model can handle the imprecise problems of interval coefficients, target intervals, and interval boundaries of meta-targets.

Many factors affect agricultural green productivity. Through a comparative study of agricultural green fertilizer and traditional fertilizer, Adnan et al.¹⁹ found that green fertilizer can increase farm product output and improve agricultural green production’s efficiency. Czyzewski et al.²⁰ found that agricultural policy can not only reduce the adverse impact of agriculture on the ecological environment but also link agricultural policy with local environmental policy through spatial effects to promote green agrarian production. In addition, Jin et al.²¹ used a fixed effect model to study the impact of population aging on agricultural green total factor productivity. The study revealed that population aging can significantly improve agricultural green total factor productivity, and there are noticeable regional differences. Yang et al.²² established a spatial panel model based on new geographic economics. They found that agricultural trade has a significant inverted U-shaped impact on agricultural green total factor productivity and technological progress. Magazzho et al.²³ found that higher output levels mainly stimulate the economic development of the agricultural sector through productivity channels. In developed economies, credit is also used to promote economic development in the farming sector. Magazzino et al.²⁴ highlight different drivers of GHG emissions under broader categories such as population, economic development, forest density, and agricultural practices. Chopra et al.²⁵ confirmed a bidirectional causal relationship between agricultural productivity and renewable energy consumption through causal testing.

Most existing literature focuses on the measurement and influencing factors of agricultural green total factor productivity, lacking research on the effectiveness of energy policy implementation and analyzing the heterogeneity of the impact of different energy consumption structures on agricultural green productivity.Compared with existing literature, the potential contributions of this article are mainly reflected in two aspects: firstly, starting from the perspective of land energy consumption structure, this article innovatively analyzes the nonlinear relationship between energy structure and agricultural green productivity, providing new empirical evidence for promoting agricultural green production. Secondly, this article uses a threshold model to empirically test the U-shaped relationship between energy structure and agricultural green productivity.

Theoretical hypothesis

Combined with the historical changes in agricultural production and the changes in the energy structure in China, this paper discusses the impact of the energy structure on agricultural green productivity by combining the research results in related fields. The theoretical framework is shown in Fig. 3.

Fig. 3

Mechanism of the impact of the energy structure on agricultural green productivity.

The impact of the energy structure on agricultural green productivity may have promoting, inhibiting, and cost effects. Based on Figure 3, the specific analysis is as follows.

Promotion effect

The extensive use of coal and oil has increased the penetration of mechanical force in agricultural production activities, promoted the efficiency of agricultural production, and thus improved the green productivity of agriculture. In the past, China’s agricultural production mainly depended on human resources, and the popularization of electricity and diesel oil significantly improved the production efficiency of bulk grain, such as through wheat sowing and harvesting, which were entirely dependent on mechanical force.Agricultural mechanization will lead to the shift of middle-aged and elderly labor towards jobs that use machinery such as irrigation²⁶. Farmers who hold agricultural machinery have higher production technology efficiency than those who do not²⁷. Using mechanical force can also reduce the use of pesticides and fertilizers. Farm machinery and equipment can accurately fertilize and spray chemicals to reduce waste and environmental pollution. Currently, China’s precision agriculture technology can use satellite navigation and sensor technology to accurately fertilize and spray fertilizers according to soil and plant needs, avoid excessive use of fertilizers and pesticides, and reduce negative environmental impacts.

Inhibiting effect

In optimizing the energy structure, the proportion of coal, oil, and other energy consumption has declined, and the proportion of clean energy consumption has increased. The energy consumption structure in agriculture will become more diversified and present regional characteristics, and the increase in overall energy input in agriculture can further improve the development of agriculture as a whole²⁸.Renewable energy (such as solar, wind, and bioenergy) can provide a cleaner and environmentally friendly energy supply for agriculture, reduce the dependence on traditional fossil energy, reduce greenhouse gas emissions, and reduce environmental pollution. The optimization of the energy consumption structure can promote technological innovation, improve energy efficiency, reduce energy consumption in agricultural production, and improve farm productivity and resource utilization efficiency by introducing more energy-saving and efficient agricultural technologies and equipment.

Cost effect

Different energy consumption structures will directly affect the energy supply and cost of agricultural production, and the reliability and cost of the energy supply will directly affect energy use in agricultural production activities. Environmental regulations such as energy pricing policies force energy costs to rise and drive technological innovation²⁹, while excessively high energy costs lead to capital flowing towards clean technologies³⁰.The energy supply is unstable or the cost is too high. Therefore, farmers may face energy shortages and high energy costs, which limits the application and implementation of green agricultural technologies. For example, Shanxi and its surrounding areas, which rely more on coal as an energy source than coastal areas, may also face supply instability and price fluctuations. High energy costs will also pressure agricultural production, such as in Gansu, Xinjiang, and other regions where high energy prices and unbalanced natural endowments significantly limit green agrarian output. Therefore, the impacts of different energy structures may be different.

Therefore, this paper proposes two hypotheses to be tested:

Hypothesis 1

A nonlinear relationship exists between the energy structure and agricultural green productivity.

Hypothesis 2

Different energy structures have different effects on agricultural green productivity. The more reasonable the energy structure is, the lower the proportion of coal consumption, which can improve agricultural green productivity.

Research methods and variable description

Through mechanism analysis, this paper theoretically shows that there may be a nonlinear relationship between the energy structure and agricultural green productivity against the background of China’s Rural Revitalization policy. When coal consumption accounts for a relatively low proportion of the energy structure, an increase in the energy structure will promote green agricultural productivity. When coal consumption accounts for a relatively high proportion, an increase in the energy structure will restrain the green productivity of agriculture. The threshold model was first proposed by Hansen³¹ and is mainly used to analyze nonlinear relationships between variables. The main characteristic of this model is that when an independent variable reaches a specific value (i.e., threshold value or threshold value), its impact on the dependent variable will suddenly change. In other words, the threshold model can reveal the discontinuous relationships between variables. That is, under certain specific conditions, the response of the dependent variable will undergo significant changes. Wang et al.³² estimated the dynamic distribution characteristics of agricultural production efficiency and agricultural carbon emissions through nuclear density. The impact of agricultural production efficiency on agricultural carbon emissions was examined using the threshold panel regression model method. Lei et al.³³ used human capital and the level of agricultural economic development as threshold variables. The impact of agricultural technological progress on agricultural carbon emission efficiency has a dual threshold effect.

To verify the nonlinear relationship between the energy structure and agricultural green productivity, this paper uses Hansen’s³¹ threshold panel model to construct a threshold regression model with the energy structure as a threshold variable.

The first step is to verify the number of thresholds. Therefore, assuming that there are n thresholds in the threshold model, taking the energy structure as the threshold, the threshold regression model of the energy structure affecting agricultural green productivity is as follows:

$$\begin{gathered} AGP_{i,t} = \beta_0 + \beta_1 eng_{i,t} \cdot I\left( {eng < \lambda_1 } \right) + \beta_2 eng_{i,t} \cdot I\left( {\lambda_1 \leq eng < \lambda_2 } \right) + \cdots \hfill \\ + \beta_n eng_{i,t} \cdot I\left( {\lambda_{n - 1} \leq eng < \lambda_n } \right) + \beta_{n + 1} B_{i,t} + \omega_i + \varsigma_t + \mu_{i,t} \hfill \\ \end{gathered}$$

(1)

where $AGP_{i,t}$ is the explained variable of agricultural green productivity in region $i$ and year $t$ and eng is the energy structure, which is not only an explanatory variable but also a threshold variable.$I\left( \cdot \right)$ is the threshold indicator function. If the expression in brackets is true, then $I\left( \cdot \right) = 1$; if false, then $I\left( \cdot \right) = 0$. $\lambda_1, \lambda_2 \ldots \lambda_n$ is the threshold for dividing each section, which is determined by the endogenous nature of the sample data. $\beta_0$ is a constant term, and $\beta_1, \beta_2 \ldots \beta_n$ are the core explanatory variable coefficients in different sections. B is the control variable, and $\beta_{n + 1}$ is the coefficient of the control variable. $\omega$, $\varsigma$ and $\mu$ represent the regional fixed effect, time fixed effect, and random interference term of the threshold regression model, respectively. Next, this paper lists the explanatory variables, explained variables, and selected control variables of the threshold regression model.

Explained variable: agricultural green productivity (AGP)

Agricultural green productivity aims to improve the output efficiency of the farming system and reduce the negative impact on the environment. This paper studies agricultural green production as an explanatory variable.

(1) Method for measuring agricultural green productivity.

Data envelopment analysis (DEA) is a nonparametric efficiency evaluation method used to evaluate relative and boundary efficiency. According to the input and output indicators in panel data or cross-sectional data, the relative efficiency of each decision-making unit (such as enterprises and institutions) is measured by building linear programming models, mainly including the CCR, BCC, and SBM models. However, the traditional DEA model cannot address unexpected outputs. To solve this problem, Chung et al. (1997)[18] proposed the Malmquist Luenberger (ML) index based on the traditional DEA model. Compared with the conventional DEA method, the Malmquist‒Luenberger index considers improving environmental efficiency and contributing technological progress to efficiency changes. It measures the improvement of relative efficiency, technological progress and changes in environmental efficiency. On this basis, the SBM-DEA model was proposed by Tone³⁴ to calculate production efficiency from the input and output perspective. Its results include relaxation variables for ineffective rate calculation with less error. The innovation of this method lies in directly placing relaxation variables into the objective function, which not only changes the input-output relaxation, but also provides efficiency evaluation after adding unexpected outputs.Therefore, to estimate agricultural green productivity in China more accurately, this paper takes provinces as decision-making units and uses the SBM-DEA model to calculate farm production efficiency.

Assuming that the agricultural production system has $n$ decision-making unit with a total of 5 input–output quantities, namely input $x \in R^m$, expected output $y^a \in R^{s1}$, and unexpected output $y^b \in R^{s2}$, matrices $X = [x_1 ,x_2 ,..., x_n ] \in R^{mxn}$, $Y^a = [y_1^a ,y_2^a ,..., y_n^a ]$, and $Y^b = [y_1^b ,y_2^b ,..., y_n^b ]$ are defined as a relaxed indicator system. The efficiency model is shown in (2).

$$\rho = \min \frac{{1 - \frac{1}{m}\sum_{i = 1}^m {\frac{s_i^- }{{x_n }}} }}{{1 + \frac{1}{s_1 + s_2 }\left( {\sum_{i = 1}^{s_1 } {\frac{s_i^a }{{y_n }} + \sum_{i = 1}^{s_2 } {\frac{s_i^b }{{y_n }}} } } \right)}}$$

(2)

$$s.t$$

$$\left\{ \begin{gathered} x_n = \lambda X + s^- \hfill \\ y_n^a = \lambda Y^a - s^a \hfill \\ y_n^b = \lambda Y^b - s^b \hfill \\ \lambda \geq 0,s^- \geq 0,s^a \geq 0,s^b \geq 0 \hfill \\ \end{gathered} \right.$$

(3)

When the measurement result of the objective function is greater than 1, it indicates that the decision-making unit is effective; When the measurement result of the objective function is within the interval [0,1], it indicates that the decision-making unit has a loss of production efficiency and needs to improve input and output.

(2) Index system and data source.

By accurately calculating China’s agricultural green productivity and studying the relationship between the energy structure and agricultural green productivity, we can provide a scientific basis for formulating agricultural green development strategies and policies, promoting improvements in agricultural green productivity, and promoting sustainable agricultural development. The above equation calculates agricultural green productivity and narrow agricultural green production efficiency measures the green utilization efficiency of the planting industry for the invested resources. Therefore, this paper takes provinces as decision-making units to establish an agricultural green productivity index system. The results are shown in Table 1.

Table 1 Index system of agricultural green productivity.

From Table 1, this paper selects land, labor, machinery, and natural resources for agricultural production activities as the input variables of agricultural green productivity. We take the total agricultural output value as the positive input variable and the overall agricultural carbon emissions as the negative input variable.

Reason for selecting input variables: Land is one of the fundamental elements of agricultural production and has a significant impact on the growth and yield of crops. Therefore, using land as an input variable can help understand the contribution of land resources to agricultural production efficiency. Labor force is another key element of agricultural production, including farmers, technical personnel, etc. The quantity and quality of labor directly affect the scale and efficiency of agricultural production. Therefore, using labor as an input variable helps to evaluate the impact of human resources on agricultural production.Machinery: In modern agricultural production, the application of mechanical equipment is becoming increasingly widespread, such as tractors, harvesters, etc. The use of mechanical equipment can improve the efficiency and scale of agricultural production. Therefore, using machinery as an input variable can reflect the impact of mechanization on agricultural production efficiency; Natural resources. Agricultural production relies on natural resources such as water resources and climate conditions. The quality and quantity of these resources have a significant impact on agricultural production. Using natural resources as input variables helps to evaluate their contribution to agricultural production.Reason for selecting output variables: Output value is an important indicator for measuring agricultural production results, reflecting the scale and efficiency of agricultural production. Using output value as an output variable can intuitively reflect the overall performance of agricultural production; Carbon emissions are a negative impact of agricultural production on the environment, and treating them as unexpected outputs can measure the environmental benefits of agricultural production.

There is no direct statistical caliber for measuring agricultural carbon emissions. This paper indirectly estimates the use of fossil fuels through production factors such as chemical fertilizers (anthracite is the primary raw material), pesticides, plastic films needed for agricultural planting, and the consumption of electricity by agrarian machinery in agricultural production.The data used to calculate agricultural green productivity were obtained from the China Statistical Yearbook (2008–2020), China Agricultural Statistical Yearbook (2008–2020), and the Statistical Yearbook (2008–2020) of various provinces. Due to the availability of data, excluding the data of Tibet, Hong Kong, Macao, and Taiwan, the agricultural green productivity of 30 regions in China from 2008 to 2020 is estimated.

Explanatory variables and threshold variables

The energy structure ($eng$) refers to the proportion of different energy types a region or country uses in total energy consumption, such as fossil fuels and renewable energy. To date, coal consumption is still the highest and is used as an energy source in many industrial areas, such as steel, chemicals, and building materials. Therefore, the proportion of coal consumption in total energy consumption is used to measure the energy structure here. Due to the inconsistency of different types of energy units, the coal consumption calculation method for power generation is used to unify the energy units here. The higher the calculated value is, the more unreasonable the energy structure. To better verify the nonlinear relationship between the energy structure and agricultural green productivity, this paper takes the energy structure as a threshold variable to test the impact of the energy structure on agricultural green productivity under different energy structure zones.

Stata is a complete and integrated statistical software that provides users with data analysis, management, and professional charting. This article uses Stata 16 software to empirically test the impact of the energy structure on agricultural green productivity.

The control variables

With respect to the rural per capita disposable income ($r\_inc$), generally, the mode of agricultural production is also related to the income level of farmers. High-income people are more likely to choose cleaner ways to produce agriculture. The rural per capita disposable income is used to measure farmers’ income. The unit is yuan (RMB), and the empirical analysis takes the logarithmic treatment of this variable.

Forest cover ($for$). The ratio of the area covered by forest to the total area reflects the state of forest resources. To a certain extent, forests can affect the water cycle, promote surface runoff and groundwater recharge, protect the stability of water sources and water purification, and reduce the cost of achieving green agricultural productivity.

With respect to the degree of damage caused by natural disasters ($dis$), agricultural production relies on climate conditions, and the damage caused by natural disasters to the ecological environment may lead to long-term recovery and repair. Compared with other industries, the impact on agriculture is more profound, and the adverse effects are more difficult to digest. This paper measures the damage caused by natural disasters by the ratio of agricultural disaster area to total cultivated land area. The unit is %.

The average length of education in rural areas ($r\_edu$). The level of education of farmers will also affect the way farmers produce. Highly educated farmers know more about green agricultural production methods and are more willing to choose green agricultural production. This paper takes the average length of education in rural areas as a measure of farmers’ education level. The unit is %. Due to the lack of a direct statistical caliber for years of education in the statistical yearbook, this article uses farmers’ educational qualifications to convert them into years of education.

Industrial structure ($is$), the upgrading of industrial structure, is the process of changing the status and relationships of various industries in the industrial structure in a more coordinated direction. Generally, the greater the output of industry and agriculture is, the greater the carbon intensity of a region. This paper uses the ratio of the tertiary industry’s added value to the secondary industry’s added value to measure the industrial structure.

The urbanization level ($urb$), which is closely related to environmental protection, represents the city’s civilization process and socialization level. This paper uses the proportion of the urban population to the total population at the year’s end to calculate the urbanization level. The unit is %.

The explanatory variable of the threshold regression model is agricultural green productivity, which has been calculated previously. The explanatory variable is the ratio of coal consumption to total energy consumption. Other control variables were derived from China’s 2008–2020 China Statistical Yearbook, China Agricultural Yearbook, China Rural Statistical Yearbook, China Energy Statistical Yearbook, and the website of the National Bureau of Statistics and China’s Economic and Social Development Statistical Database. Some indicators, such as the degree of damage caused by natural disasters and industrial structures, are calculated according to statistics. Calculating the average number of years of education in rural areas is more complex. It is necessary to multiply the number of years of education of each rural educated group (such as primary school, junior high school, high school, university, etc.) by the population ratio of the corresponding group and then add up the results of all groups to obtain the total number of years of education of each group. Then, the total number of years of education is used to obtain the average level of education in rural areas. In addition, considering the particularity and data availability of Tibet, Hong Kong, Macao, and Taiwan, the data of these four regions were excluded from the data collation, and the panel data of 30 provinces in China were finally obtained. Table 2 shows the descriptive statistical analysis of each variable.

Table 2 Descriptive statistical analysis of variables.

Calculation results and empirical analysis

Calculation results and the spatial evolution characteristics of agricultural green productivity

(1) Calculation results.

Figure 4 shows the average value and growth rate of China’s agricultural green productivity from 2008 to 2020.

Fig. 4

The average value and growth rate of agricultural green productivity in China from 2008 to 2020. Note: The above data are calculated based on the results of agricultural green productivity in different regions of China from 2008 to 2020. The average agricultural green productivity is calculated as the simple arithmetic mean of the annual agricultural green productivity in 30 areas of China, which means that all the weights are equal.

Figure 4 shows that over time, the average value of agricultural green productivity gradually increased. From 0.4040 in 2008 to 0.6893 in 2020, the average increased. The growth rate of agricultural green productivity fluctuates between different years, with positive and negative values alternating. However, 2009 was the only year that had negative growth, and agricultural green productivity showed an upward trend in other years. The overall growth rate shows an upward trend, especially in recent years (from 2017 to 2020), and the growth rate has increased significantly. We can determine the difference between different years by comparing the years with the average growth rates. Figure 5 shows that China’s agricultural green productivity growth rate was higher in 2018 and 2019, at 9.25% and 8.97%, respectively, while the growth rate in other years was relatively low.

Figure 5 shows the average and growth rate of agricultural green productivity in different regions of China from 2008 to 2020.

Fig. 5

The average agricultural green productivity in different regions of China from 2008 to 2020.

From Figure 5, it can be seen that the average agricultural green productivity in the western region is the highest, followed by the eastern region, and the central region is the lowest. The agricultural green productivity in the eastern region has been increasing year by year since 2008, and has slightly decreased after reaching a peak in 2019. It has gone through a period of rapid growth, but the growth rate has since slowed down. The agricultural green productivity in the central region fluctuated less during this period and showed a stable trend overall.The agricultural green productivity in the western region has also shown an upward trend during this period, but the growth rate is relatively slow. This indicates that the western region has gradually improved agricultural green productivity during this period, but the growth rate is slower compared to the eastern region.

Overall, China’s agricultural green productivity showed a gradually increasing trend from 2008 to 2020, and the growth rate in recent years has been significantly higher than in previous years. With the continuous development of technology, agricultural production methods have been improved and enhanced. The introduction of new technologies, varieties, breeding methods, and modern agricultural facilities has improved production efficiency and agricultural product quality.

(2) Spatial evolution characteristics.

To further observe the spatial evolution characteristics of agricultural green productivity, Fig. 6 shows the regional agricultural green productivity in 2008, 2012, 2016, and 2020.

Fig. 6

Agricultural green productivity by province in China in 2008, 2012, 2016, and 2020. Note: Draw a graph based on the calculation results of agricultural green productivity in different regions of China from 2008 to 2020. The drawing software used was Excel.

From the perspective of provinces, in Fig. 6, the development trend of agricultural green productivity in different regions of China from 2008 to 2020 shows three characteristics. First, the growth rate of agricultural green productivity in coastal areas was relatively fast, and the agricultural green productivity in Tianjin, Guangdong, and Hainan reached 0.3856, 0.4757 and 0.3220, respectively, in 2008. These values are much greater than those in Ningxia, Inner Mongolia, Xinjiang, and other regions. The coastal climate is mild and humid and conducive to crop growth and development. At the same time, coastal areas are relatively rich in land resources and have high soil fertility, which is conducive to crop planting and production. Agricultural production in coastal regions tends to accept advanced farm technology and management methods earlier. After more than ten years of development, the green productivity of agriculture in some coastal areas has reached more than 0.7. Second, the development of agricultural green productivity in the central region, represented by Guizhou, Shandong, and Sichuan, started early and developed rapidly. After rapid development, agricultural green productivity increased. By 2020, the green productivity of agriculture in these three regions will reach 1.0529, 1.0607, and 1.1049, respectively. The central region pays more attention to the coordinated development of agriculture and other related industries. Agricultural product production and marketing docking has been realized through the development of the agricultural product processing industry, the agrarian product circulation industry, and other agricultural industrial chains, and the added value and market competitiveness of farming products have improved. Third, the green productivity of agriculture in some areas, such as Xinjiang, Shaanxi, Jilin, and Anhui, has remained relatively low, starting late and developing slowly. By 2020, the green agricultural productivity in these areas will reach 0.4721, 0.5175, 0.5280, and 0.5302, respectively. Farm technology and management in these areas are relatively limited; farmers generally adopt traditional agricultural production methods and lack modern agrarian technology and facilities, which limits the improvement of agricultural production efficiency and the promotion of green production modes. In addition, some areas lack specific policies and financial support for the development of green agriculture. The government’s investment in agricultural scientific and technological innovation, environmental governance, and green production is relatively limited, which limits the speed of green agrarian output.

Scatter plot

Before threshold regression, this paper uses a scatter plot to determine the relationship between the energy structure and agricultural green productivity and whether the two variables are positively correlated, negatively correlated, or unrelated according to the distribution of the data points. By visualizing the points on the two-dimensional plane, the scatter plot can show the distribution, correlation, and trend between the energy structure and agricultural green productivity to understand the data characteristics and patterns. Figure 7 shows a scatter plot of the energy structure and agricultural green productivity.

Fig. 7

Scatter plot of the energy structure and agricultural green productivity. Note: Using Stata 16 software for drawing, the drawing data include variable energy structure and green productivity.

The scatter plot shows a negative correlation between the energy structure and agricultural green productivity. Overall, the more unreasonable the energy structure is, the lower the green productivity of agriculture. Most of the data are concentrated in the 0.6 to 0.8 energy structure range. We can quickly identify outliers by observing outliers or abnormally distributed data points in the scatter plot. After removing outliers, there is a simple U-shaped relationship between the energy structure and agricultural green productivity. The negative relationship between the energy structure and agricultural green productivity is more evident in areas with low energy structure values. Scatter plots can be used to predict future trends. With the gradual rationalization of the energy structure, by observing the distribution of data points in the scatter plot, it can be found that the greater the green productivity of agriculture is, the greater the value.

Threshold regression results and deeper analysis

(1)Result.

According to the results of the scatter plot in the previous section, the more reasonable the energy structure is, the greater the inhibitory effect on agricultural green productivity. According to the theoretical analysis, the impact of different energy structures on agricultural green productivity will be different in regions with other energy structures. Next, the energy structure is used as a threshold variable to empirically test the nonlinear relationship between China’s energy structure and agricultural green productivity. However, before empirical research, this paper needs to determine the number of thresholds to determine the form of the threshold model. According to the “self-sampling” method of the threshold model of Lian et al.³⁵, the model is estimated under no threshold, one threshold, and two thresholds. Table 3 shows the regression results of the threshold effect test.

Table 3 Regression results of the threshold effect test.

From Table 3, the F value corresponding to the triple threshold test is 7.023, which is insignificant at the 10% level, indicating that there is no triple threshold in the model. In the single threshold test, the F value of the model was 84.580 and passed the 1% significance test, while the F value of the double threshold was 49.415, which was also significant at the 1% level. In the model with the energy structure as the threshold variable, the symbol of the impact coefficient of the energy structure on agricultural green productivity changed once. That is, at different intervals of the energy structure, its impact on agricultural green productivity can be promoted or restrained. In addition to the change in the symbol of the influence coefficient, there is a significant change in the size of the coefficient and no change in the symbol of the influence coefficient in the double threshold. This paper chooses the dual threshold model for threshold estimation to more accurately discuss the nonlinear relationship between the energy structure and agricultural green productivity. Table 4 shows the threshold estimation results with the energy structure as the threshold.

Table 4 Threshold estimation result.

According to Tables 3 and 4, there is a significant double threshold effect between the energy structure and agricultural green productivity, and the estimated values of the threshold variables are 0.578 and 0.743, respectively. The double threshold estimate divides the energy structure into three intervals, namely, the area with a reasonable energy structure ($eng < 0.578$), the area with a medium suitable energy structure ($0.578 \leq eng < 0.743$), and the area with an unreasonable energy structure ($eng \geq 0.743$). Next, this paper uses the least squares likelihood ratio (LR) to identify the threshold, and the value when the LR is zero is the threshold estimate. The likelihood ratio function of the double threshold estimate is shown in Fig. 8, where the horizontal red dotted line is the 95% confidence interval. According to the threshold estimation results in Table 4, the 95% confidence interval of the first threshold estimate of the energy structure is [0.737, 0.748], and the 95% confidence interval of the second threshold estimate is [0.578, 0.579].

Fig. 8

Threshold estimation results and confidence intervals. Note: Abscissa x is the energy structure (eng), and ordinate LR is the likelihood ratio of the threshold. The blue curve describes the trajectory of different threshold collection points, and each point on the curve’s ordinate represents the corresponding threshold’s likelihood ratio. The dotted line represents the 95% confidence interval, and the interval formed by the intersection is the 95% confidence interval. The smaller the blue area under the 95% confidence interval is, the smaller the influence of the unobservable factors of the threshold model and the more accurate the threshold regression results.

According to Figure 8, at the 95% confidence level, the confidence intervals corresponding to the first and second threshold estimates are small. The results show that the model with the energy structure as the threshold estimates more accurately, verifying a nonlinear relationship between the energy structure and agricultural green productivity. To further confirm the specific situation of this nonlinear relationship, we estimate the parameters of the dual threshold model, and the threshold regression results are shown in Table 5.

Table 5 Threshold regression results.

According to the threshold regression results in Table 5, taking the energy structure as the threshold, China is divided into areas with reasonable, moderately reasonable, and unreasonable energy structures. When the energy structure is in the third range, the impact coefficient of the coal consumption ratio on agricultural green productivity is 0.0413, which passes the 1% significance test. That is, with the gradual increase in the coal consumption ratio, agricultural green productivity has improved. With the gradual rationalization of the energy structure, the increase in coal consumption will still promote agricultural green productivity after entering the second interval. Nevertheless, this promotion effect is 0.0209, which is significantly lower than the impact coefficient of the first interval. Once the energy structure is lower than the first threshold of 0.578, an increase in the proportion of coal consumption will significantly inhibit agricultural green productivity, and this coefficient is -0.1365, which passes the significance test at the 1% level. During the transition process from the second interval to the first interval, the impact of the energy structure on agricultural green productivity changes from positive to negative, and an increase in the proportion of coal consumption inhibits the development of regional agricultural green productivity to a certain extent.

(2) Deeper analysis.

Generally, the threshold regression results show a nonlinear relationship between the energy structure and agricultural green productivity. Expanding coal consumption will restrain agricultural green productivity in areas with relatively reasonable energy structures. Expanding coal consumption in areas with unreasonable energy structures can promote agricultural green productivity to a certain extent. Agrarian production requires energy to drive agricultural machinery, irrigation facilities, greenhouses, and other equipment and process, store, and transport agricultural products. First, the energy structure will directly affect the level of energy consumption in agricultural production. Using renewable energy can reduce environmental pollution and carbon emissions and improve the efficiency of green agrarian production. However, in the early stage of agricultural development, especially in developing and underdeveloped countries, there is a lack of resources to create clean energy, and the use of nonclean energy can significantly promote the efficiency of agricultural planting, irrigation, processing and transportation, thus increasing the efficiency of agricultural green production. Second, with the acceleration of economic development, the supply of clean energy is becoming increasingly stable, and the energy structure has improved, which can promote the electrification process in rural areas and provide a steady and reliable energy supply. This will help to enhance agricultural production conditions, improve the application level of agricultural green production technology, and reduce the impact of traditional farming methods on the environment. In this process, the increasing proportion of coal consumption will restrain the green productivity of agriculture. Third, improving the energy structure can promote the development and application of renewable energy. In agriculture, the use of renewable energy may present some economic challenges. Clean energy technology is becoming increasingly advanced, significantly reducing the cost of clean energy production. Agricultural waste and byproducts can be converted into renewable energy through biomass energy and bioenergy technology, such as biomass pellet fuel and biogas. This method of energy utilization reduces the emission of agricultural waste and environmental pollution. It provides clean energy to replace traditional energy sources and reduce the scale of carbon emissions caused by agricultural production. With the maturity of renewable energy technology and the decline in cost, renewable energy has great potential to improve agricultural green productivity.

Specifically, the source structure has an essential impact on agricultural green productivity. At present, the energy structure in some areas of China is less than 0.578, and improving the energy structure can promote green agricultural productivity. Most places are still in the range of unreasonable energy structures, so we need to adjust the energy structure further to ensure food security and maintain green agricultural productivity growth.

Robustness test

The threshold model regression results show a nonlinear relationship between the energy structure and agricultural green productivity. This article uses fixed and random effects models to test the robustness of the threshold regression results using quadratic terms. This verifies the reliability of the results. The test results are shown in Table 6.

Table 6 Results of the robustness test.

Table 6 shows that both the fixed and random effects models have negative coefficients for the primary term of the energy structure. The results passed the 1% significance test. This indicates that the greater the proportion of coal consumption is, the lower the green productivity of agriculture. Models (2) and (4) are the regression results after incorporating the quadratic term of the energy structure. This shows that the symbols for the primary and secondary terms of the energy structure are different. The regression results of the quadratic term passed the 1% significance level test. This indicates a turning point in the relationship between the energy structure and agricultural green productivity. At the same time, the nonlinear relationships between the threshold model and energy structure and agricultural green productivity were also verified. Therefore, the threshold model is robust.

Conclusions, policy recommendations, and research prospects

Conclusion

Agricultural green productivity is of great significance to the development of contemporary society. On the one hand, it helps to ensure food security and the effective supply of agricultural products. In the context of growing population and intensified resource constraints, improving the output efficiency of unit resources through green production can meet people’s demand for high-quality agricultural products and stabilize the situation of agricultural production. On the other hand, agricultural green productivity plays a key role in ecological environment protection. Reducing the use of chemical fertilizers and pesticides and reducing agricultural non-point source pollution are conducive to protecting soil, water sources and biodiversity, and realizing the harmonious coexistence between agriculture and nature. Based on the panel data of 30 provinces in China from 2008 to 2020, this paper constructs the index system of agricultural green productivity, uses the super efficiency SBM model to calculate the agricultural green productivity, and analyzes its characteristics from the perspectives of time and space.

Firstly, China’s agricultural green productivity showed a gradual increase from 2008 to 2020, and the growth rate in recent years was significantly higher than that in previous years. Secondly, there is a nonlinear relationship between energy structure and agricultural green productivity. In areas with relatively reasonable energy structure, the expansion of coal consumption will restrain agricultural green productivity. Energy structure has an important impact on agricultural green productivity. At present, the energy structure in some areas of China is less than 0.578, and it is in the stage that improving the energy structure can promote agricultural green productivity. Most areas are still in the range of unreasonable energy structure, so we need to further adjust the energy structure to ensure food security on the basis of maintaining the growth of green agricultural productivity.

This conclusion has some inspiration for other developing countries to promote green agricultural development, especially how to balance the difficult problem of green agricultural productivity and energy structure upgrading. From the perspective of technological progress, other developing countries should encourage and support the research and development and application of green agricultural technologies, such as precision agriculture and intelligent irrigation systems, in order to improve agricultural production efficiency and reduce dependence on fossil fuels. At the same time, promote the popularization and application of clean energy technologies in rural areas, such as solar, wind and biomass energy. From the perspective of cross sectoral cooperation: a coordination mechanism among agriculture, energy, environment and other relevant departments can be established to ensure synergy in the process of policy formulation and implementation. Through cross sectoral cooperation, resources can be integrated more effectively to promote the mutual promotion of agricultural green productivity and energy structure upgrading.

Policy recommendations

Based on the conclusion drawn from the paper, China’s energy structure is not yet reasonable enough. Suggestions are proposed: on the one hand, establish large wind farms and solar photovoltaic power stations in suitable areas, and gradually phase out old coal-fired power stations. Promote distributed energy systems, such as rooftop solar panels and small wind turbines, especially in remote and rural areas. Develop other forms of renewable energy technologies such as geothermal energy, tidal energy, and wave energy. On the other hand, increasing investment in basic and applied research, especially in the fields of efficient solar cells, energy storage technology, and smart grids. Upgrade power grid facilities to enhance their adaptability to the volatility and uncertainty of renewable energy.

Based on the conclusion of the paper, the overall level of green productivity in China’s agriculture is relatively low. Suggestions two are proposed: on the one hand, adjust the crop planting structure according to market demand and increase the proportion of high value-added crops. Develop facility agriculture, such as greenhouses, to extend the growing season and increase yield. Promote intercropping and mixed cropping to improve land use efficiency and ecological diversity. On the other hand, intelligent agricultural equipment such as drone spraying, intelligent monitoring cameras, and automated control systems will be introduced. Utilize big data and artificial intelligence to analyze crop growth data and achieve precision agricultural management. Develop mobile applications to provide farmers with real-time weather information, market trends, and agricultural consulting services.

Based on the conclusion of the paper, there is a non-linear relationship between energy structure and agricultural green productivity. In areas with relatively reasonable energy structure, an increase in the proportion of coal consumption will suppress agricultural green productivity. Suggestions three are proposed: on the one hand, support small-scale wind and solar power projects in rural areas to provide stable clean energy supply for agricultural production. Using biomass energy technology to convert agricultural waste into energy not only reduces environmental pollution but also improves energy utilization efficiency. Promote organic and ecological agriculture practices, reduce the use of fertilizers and pesticides, and protect soil health and biodiversity. On the other hand, providing technical support and financial incentives to encourage farmers to adopt energy-saving and emission reducing agricultural production methods. Develop preferential policies, such as tax reductions, subsidies, etc., to encourage farmers and enterprises to use clean energy. Establishing green credit and insurance products to reduce the number of farmers.

Research outlook

This paper empirically tests the nonlinear relationship between China’s energy structure and agricultural green productivity but is limited by the availability and applicability of the data. This study has certain limitations and shortcomings. On the one hand, this paper only uses provincial-level data from 2008 to 2020, and the sample size is small, which limits this paper’s research depth and breadth. In future research, we will continue to pay attention to the development of China’s energy industry, look for prefecture-level city-level data, and discuss the heterogeneity of energy structure and agricultural green productivity from the perspective of prominent grain-producing areas and main grain-producing areas. On the other hand, a comprehensive understanding and scientific measurement of agricultural green productivity is the premise of promoting agricultural green development. The index system in this paper is not extensive enough, and some factors that are difficult to measure numerically have not been included in the index system. In the future, we will continue looking for and improving the agricultural green productivity index system.

In addition, most of the research on agricultural green productivity in China is carried out at the macro level, and there is a lack of microlevel literature. Therefore, future research can consider studying agricultural green productivity and its influencing factors from the perspective of farmers’ behavior. Through in-depth analysis of farmers’ decision-making behavior, resource utilization, and technology application, we can better understand the formation mechanism of agricultural green productivity. By combining macro- and microresearch results, we can more comprehensively grasp the development trend and influencing factors of agricultural green productivity. Connecting China’s experience with international experience can provide different paths and lessons for China to improve agricultural green productivity. In practice, we can formulate corresponding policies and measures to encourage farmers to adopt green planting and breeding methods, introduce new technologies and management models, and promote the transformation of agriculture toward a green and sustainable direction.