A Multidimensional Exploration of the Factors Influencing Comprehensive Grain Production Capacity from a Spatial Perspective

Abstract

Improving comprehensive grain production capacity and ensuring food security is a critical step toward promoting national economic development and social stability. This study constructs an evaluation framework for comprehensive grain production capacity, examining five key dimensions: resource endowment, factor inputs, scientific and technological infrastructure, grain output, and sustainable development. The entropy method is employed to assess the national comprehensive grain production capacity. Subsequently, the study was performed based on the panel data from 31 provinces in China, covering the period from 2011 to 2022. This analysis was framed within the context of the Spatial Durbin Model and focused on four key aspects: the inputs of production factors, advancements in agricultural technology, national food policies, and the impact of natural disasters. The results indicate the following: (1) Grain production in the 31 provinces of the country demonstrates a significant positive correlation, with a clear space centralization effect. (2) The sown area of grain, labor inputs, fertilizer usage, and the advancement of the digital economy are the primary positive factors influencing regional grain production, but the disaster-affected area has a negative impact on grain production. As a result, this study makes policy recommendations to increase food security and sustainable agricultural development by increasing the sown area of grain, optimizing fertilizer use, and improving agricultural practice digitization. Special emphasis is being placed on reviving high-quality agricultural development by hastening the integration of the digital economy into agriculture.

1. Introduction

Grain production is the cornerstone of national security and social stability, and is of irreplaceable strategic significance in ensuring the basic needs of the people and promoting economic development. In the context of global climate change, increasing resource and environmental constraints and continued population growth, it is particularly important to recognize and enhance the comprehensive grain production capacity (hereinafter referred to as CGPC) within the region. CGPC refers to the stable output capacity of food that a country can achieve under specific technical conditions and inputs of production factors, and is an important indicator of a nation’s ability to ensure food security [1]. The concept encompasses not only the actual production of food, but also various dimensions, including the efficiency of land resource utilization, the advancement of agricultural science and technology, the development of agricultural infrastructure, and the capacity to withstand risks. Ensuring food security through stable grain production is fundamental to the development of the Chinese nation. Since 2003, when agricultural issues, along with those concerning farmers and rural areas, were first incorporated into the Government’s work report, the Central Committee of the Communist Party of China (CPC) and the State have placed significant emphasis on the matter of grain production. At the 2013 Central Rural Work Conference, General Secretary Xi Jinping emphasized that the rice bowls of the Chinese people must be kept firmly in their own hands at all times, emphasizing the strategic significance of food security. The central government’s No. 1 document for 2022 focuses on comprehensively promoting the rural vitalization strategy, and explicitly proposes to ensure that the sown area of grain is stable and output remains above 1.3 trillion jin, while emphasizing that the foundation of national food security is strengthened by reinforcing the basic support of modern agriculture and increasing CGPC. This gives crucial policy guidance and practical direction to in-depth analyses of the elements influencing grain production, as well as the promotion of high-quality agricultural development in the new era.

Overall, the country’s agricultural development has reached a new historical milestone. China, being a significant agricultural nation, has abundant arable land resources and a diverse range of agricultural producing zones. Notably, commercial grain bases such as the Songnen Plain and the Sanjiang Plain play an important role in national grain production. In the context of promoting high-quality agricultural development, it is essential to scientifically examine the key factors that influence the CGPC across different regions, while considering the specific conditions of grain production in each area. This assessment is critical to the long-term development of agriculture and food security throughout the country.

2. Literature Review

Based on existing research, the evaluation system for CGPC has gradually improved, and the development of a multi-dimensional indicator system can more comprehensively reflect grain production’s potential and constraints, while the use of inter-provincial panel data also provides a scientific basis for conducting inter-regional comparison [2]. However, national studies have found that, although China’s CGPC has continuously grown, it still confronts challenges such as unbalanced regional development and increasing resource and environmental limits [3,4]. With regard to the measurement and assessment of CGPC, scholars have conducted extensive research from different perspectives and using different methods. Using empirical data spanning from 1983 to 2005, He and colleague conducted a preliminary discussion on China’s CGPC by analyzing grain yields and the sown area of grain [5]. On this basis, Wang and colleagues further used a spatial regression model to focus on the spatial distribution of CGPC and its underlying causes in Hebei Province, revealing the joint mechanism of natural conditions such as land resources, precipitation, and socio-economic factors such as agricultural machinery power [6]. With the diverse development of research methodologies, Xin and colleagues used the Entropy Weight Method combined with panel data spanning from 2004 to 2015 to analyze the complete grain production capacity of China’s major grain-producing areas and their sustainable development capability [4].

The elements impacting CGPC involve natural conditions, agricultural input components and the level of science and technology. Shuai Chen et al. examined the impacts of temperature, irrigation, and other environmental factors on Chinese agriculture using maize and soybean as examples, offering an important reference for relevant policy decisions [7]. Abbas Ali Chandio and colleagues analyzed the long-term and short-term impacts of technological innovation, digitization level, and temperature on grain production in China based on a sustainable development perspective, revealing the complex effects of multidimensional factors on food security [8]. Wang and colleague discovered through spatial econometric modeling that multifunctional farmland use transition (MFFUT) has a significant positive effect in enhancing local grain production while inhibiting the effect on neighboring regions [9]. In addition, Abbas Ali Chandio and colleagues used the example of major rice-producing provinces in China to establish that public agricultural investments, fertilizers, and the usage of agricultural machinery make a significant contribution to grain production [10]. At the same time, grain production in various places is influenced by a wide range of causes, each with its own set of mechanisms. For instance, a study conducted by Li and colleagues in China identified that the sown area of grain, precipitation, and fertilizer application were the most important predictors of grain yield in dry farming areas. Notably, changes in mixed grain purchase prices had a substantial impact on grain production in Gansu Province [11]. Chen and colleagues discovered that the discounted amount of fertilizer applied per unit was the most critical element influencing the CGPC in the Ningxia region [12]. The Middle East and North Africa (MENA) region has long relied on food imports; however, the escalation of the Russian–Ukrainian conflict presents a significant challenge to grain production and food security in the area [13,14]. At the same time, one in three Ukrainians were food insecure at the time of the surveys [15]. The preceding analysis demonstrates that the influencing elements of grain production in different regions share both common and regional characteristics, and that the development of targeted solutions matched to local conditions is critical for increasing overall grain production capacity.

Among other things, fostering high-quality agricultural development is critical for achieving a sustained increase in CGPC. As a basic industry of the national economy, high-quality agricultural growth is not only linked to food security, but also has a direct impact on the long-term development of the economy. In the contemporary phase of development, the digital economy has emerged as a significant catalyst in transforming agricultural development models. It contributes to the enhancement of efficiency, quality, and sustainability in grain production through technological innovation and the optimization of industry chains [16,17]. In addition, agro-technological innovations essential for enhancing the quality of agriculture [18,19], such as artificial intelligence (AI), have emerged as powerful tools for grain production and food security [20], particularly in developing countries [21]. However, the problem of low productivity within the agricultural sector continues to be significant. Douglas Gollin examined the relationship between agricultural development and economic growth, highlighting that the development of non-agricultural industries may have an indirect impact on agricultural productivity [22]. Subsequent research conducted by Douglas Gollin and his colleagues on developing countries has shown that the disproportionate allocation of labor between the agricultural and non-agricultural sectors significantly contributes to the lower productivity observed in the agricultural sector [23]. Jin and colleagues found that public agricultural research and its extension have a significant role in increasing agricultural productivity by analyzing agricultural production data from all U.S. states [24]. However, studies in recent years have highlighted a concerning trend regarding the negative impact of climate change on agricultural productivity. This trend presents new challenges for the sustainable development of agriculture and food security, both globally and regionally [25,26,27].

Through reference existing literature, it is evident that, although there is no clear definition of CGPC in foreign contexts, most scholars tend to concentrate on the influence of specific factors or a combination of factors on grain production or agricultural productivity, including water resources, temperature, and soil fertility. Notably, research on grain production and the high-quality development of agriculture has been conducted previously in other countries than in China, providing a wealth of research methodologies and practical experiences that can be utilized. In China, existing studies have thoroughly examined the influencing factors and assessment methods related to CGPC from various perspectives, thereby offering a significant theoretical foundation for understanding the current state of grain production and guiding future development directions. Nevertheless, several limitations persist in the current body of research. Firstly, the majority of studies focus on specific regions or provinces, such as major grain-producing areas, the Three River Plain and Henan Province, resulting in a lack of macro-level assessments of CGPC at the national scale. Secondly, while there has been an emphasis on interregional disparities, the analysis of heterogeneity within different regions remains underexplored.

Based on previous research, the paper selects the entropy method as the evaluation approach for CGPC in the main grain producing areas, aiming to enhance the objectivity and comprehensiveness of evaluating CGPC. And using spatial econometric modeling, it analyzes the important factors influencing China’s CGPC. Food security is a critical component of national security. Therefore, clarifying the factors affecting grain production and thus enhancing CGPC has far-reaching significance for ensuring national food security, and promoting stable economic and social development.

3. Analysis of the Status of Grain Production

3.1. Production Conditions

China has a diverse range of land resources, but per capita availability is low, and distribution is highly uneven. The total area of arable land across the country is approximately 1.918 billion mu, resulting in a per capita arable land area of only 1.36 mu, which is significantly below the global average. Furthermore, the high proportion of medium- and low-yield fields, combined with serious issues of soil degradation and erosion, has negatively impacted the efficiency of land utilization. Secondly, China has a wide variety of climatic resources. The eastern region of the country is heavily influenced by a monsoon climate, which combines precipitation and warmth, giving ideal circumstances for agricultural development. In contrast, the west of China has arid and semi-arid climates with limited precipitation and high evaporation rates, which impede agricultural development. Nonetheless, the Tibetan Plateau and the northwestern region benefit from plentiful sunlight and heat, creating favorable circumstances for grain growth. In addition, although the total amount of water resources is comparatively rich, the per capita share is just one-quarter of the global average, and the distribution is exceedingly uneven, with more in the south and less in the north, more in the east and less in the west. Water resources south of the Yangtze River account for more than 80% of the country’s total water supply, whereas the northern regions, particularly North and Northwest China, suffer from severe water shortages. With the rapid development of the economy and society, the supply–demand gap for water resources is becoming more pronounced.

3.2. Inputs for Grain Production

Since the 21st century, China’s sown area of grain has steadily increased. In 2000, the national sown area for grain was approximately 103 million hectares, and from 2004 onwards, driven by the abolition of agricultural taxes and the implementation of supportive agricultural policies such as grain planting subsidies, the sown area began to rebound, reaching 106 million hectares by 2007 and increasing to 116.77 million hectares by 2020, an approximate 13.4% increase over 2000. In 2022, the sown area for grain was 118 million hectares, accounting for a higher proportion of the country’s total arable land.

Since the beginning of the 21st century, China’s agricultural machinery capacity increased from 769 million kilowatts in 2007 to 1.056 billion kilowatts in 2020. During the same period, the mechanization rate increased from 35% in 2000 to 73% in 2022. Furthermore, the use of chemical fertilizers peaked at 59.96 million tons in 2015 before falling to 52.507 million tons by 2020. This trend reflects a shift toward more efficient and sustainable agricultural practices, while maintaining grain production stability.

In terms of policy support, the government has reduced farmers’ production costs by eliminating agricultural taxes, subsidizing grain production, seeds, and machinery, and enacting arable land protection measures. The development of high-quality farmland and the implementation of minimum grain purchase price policies help to stabilize farmers’ incomes. Collectively, these measures improve grain production and ensure stable agricultural output, contributing to the growth of farmers’ incomes.

3.3. Food Output

China’s grain output steadily increased from 462 million tons in 2000 to 687 million tons in 2022. In terms of crop composition, the three primary staple grains—rice, wheat, and maize—predominate, with maize production demonstrating particularly significant growth. Additionally, per capita food availability rose from approximately 400 kg in 2000 to over 486.1 kg in 2022, surpassing the global average. Currently, China’s grain output remains stable at a high level, underscoring the country’s enhanced grain production capacity and the robust foundation of its food security.

4. Measurement of CGPC

4.1. Source of Data

The CGPC refers to the analysis of the relatively stable level of grain production that can be achieved over a specified period, within a particular region, and under defined economic and technological conditions. This capacity is the result of the combined inputs from various factors of production. Drawing upon existing research and the current circumstances of the 31 provinces in China, this study has established five levels for the analysis of CGPC: the level of resource endowment, the level of factor inputs, the level of scientific and technological equipment, the level of grain output, and the level of sustainable development. The data utilized in this research are sourced from the China Statistical Yearbook, the China Rural Statistical Yearbook, and statistical reports from grain production-related departments across various provinces. To address potential data gaps, this study encompasses 31 provinces and municipalities, measuring data from 2007 to 2022, with any missing values being filled in through linear interpolation.

4.2. Selection of Models

This study uses the Entropy Weight Method to measure the weight of each indicator and to measure the CGPC more comprehensively. Grain production capacity is influenced by multiple variables such as climate, soil, pests and diseases, policies, and market demand. Each of these variables itself is uncertain. In information theory, entropy is a measure of uncertainty. The more information there is, the less uncertainty there is and the less entropy there is, and vice versa [28]. Entropy Weight Method determines the degree of dispersion of an indicator by calculating the entropy value; the greater the degree of dispersion of the indicator, the greater the influence of the indicator on the comprehensive evaluation [29]. Therefore, the Entropy Weight Method is a relatively objective method of assignment, and this study analyzes the weights of the indicators based on the CGPC via the following process.

To eliminate the impact of the differing scales of the variables on the results, the data were normalized using the polar deviation approach.

Positive indicator:

$$x_{ij}^{\prime }={\displaystyle \frac{x_{ij}-x_{ij}^{min}}{x_{ij}^{max}-x_{ij}^{min}}}$$

(1)

$$x_{ij}^{\prime }={\displaystyle \frac{x_{ij}^{max}-x_{ij}}{x_{ij}^{max}-x_{ij}^{min}}}$$

(2)

In this context, positive indicators refer to variables that exert a beneficial influence on CGPC and negative indicators pertain to variables that have a negative impact on CGPC. x_ij is the data for indicator j in year i. The standardized variables are characterized by a mean of 0 and a variance of 1.

We then calculate the specific gravity, via

$$p_{ij}={\displaystyle \frac{x_{ij}^{’}}{{\sum }_{j=1}^m{x}_{ij}^{’}}}$$

(3)

followed by calculating entropy,

$$e_j=-\left({\displaystyle \frac{1}{\mathit{ln}n}}\right)\sum _{i=1}^n{p}_{ij}\mathit{ln}{p}_{ij}$$

(4)

where $e_j\in \left[\mathrm{0,1}\right]$, n = 496

We then calculate the coefficient of variation, as

$${\lambda }_j=1-e_j$$

(5)

A bigger value of ${\lambda }_j$ indicates that the indicator is more essential in the comprehensive evaluation.

Then, we calculate the weights of the indicators,

$${\omega }_j={\displaystyle \frac{{\lambda }_j}{{\sum }_{j=1}^m{\lambda }_j}}={\displaystyle \frac{1-e_j}{{\sum }_{j=1}^m\left(1-e_j\right)}}$$

(6)

and then calculate the value of CGPC,

$$u_i=\sum _{j=1}^m{\omega }_j{x}_{ij}^{’}$$

(7)

4.3. Analysis of Results

In the process of establishing an evaluation index system for CGPC, this study draws upon existing research while considering the general trend of grain production within China, as well as the specific regional characteristics. To ensure the objectivity, representativeness, and accessibility of the evaluation index system, the analysis encompasses five levels: resource endowment, factor inputs, scientific and technological equipment, grain output, and sustainable development. Furthermore, 14 secondary indicators have been selected to collaboratively construct a CGPC evaluation system. The results are shown in

Table 1. Indicator system for evaluating CGPC.

Primary Indicator	Weights	Secondary Indicators	Property	Weights
Level of resource endowment	0.473	Cultivated land per capita (hectares per 10,000 people)	P	0.064
		Area of grain sown per capita (hectares per 10,000 people)	P	0.046
		Water resources per capita (cubic meters per person)	P	0.363
Level of physical infrastructure	0.284	Total power of agricultural machinery (million kilowatts)	P	0.074
		Effective irrigated area (thousands of hectares)	P	0.071
		Electricity consumption in agriculture (billion kWh)	P	0.139
Grain output level	0.078	Total grain production (tons)	P	0.076
		Gross agricultural output index (previous year = 100)	P	0.002
Policy backing intensity	0.137	Share of expenditure on agriculture, forestry and water (%)	P	0.018
		Depth of agricultural insurance (%)	P	0.119
Sustainable development capacity	0.031	Agricultural disaster rate (%)	N	0.001
		Fertilizer application (tons)	N	0.009
		Pesticide usage (tons)	N	0.013
		Agricultural plastic film usage(ton)	N	0.008

Note: P for positive property, N for negative property.

.

Figure 1 and Figure 2 illustrate the trend in CGPC across 31 provinces and municipalities for the years 2007, 2012, 2017 and 2022. The data presented in Figure 1 and Figure 2 indicate a general upward trajectory in the CGPC among the 31 provinces analyzed; however, sig-nificant disparities exist in the levels of production capacity across different regions. For instance, provinces such as Heilongjiang, Inner Mongolia, and Tibet exhibit relatively high production capacity values over several years, which may be attributed to their fa-vorable natural conditions for grain cultivation and extensive arable land. Conversely, provinces like Tianjin demonstrate comparatively low production capacity, potentially constrained by limitations in land resources. Furthermore, notable changes in CGPC are observed in certain provinces in 2022 when compared to the years 2007, 2012 and 2017. For example, Beijing’s production capacity in 2022 is markedly higher than in the previ-ous years, likely due to factors such as progress in agricultural technology and supportive policies. In contrast, some provinces and municipalities, such as Tibet, show minimal fluctuations over the years, maintaining a relatively stable level of production capacity.

5. Empirical Analysis of Factors Influencing Grain Production Utilizing Spatial Econometric Modeling

5.1. Model Construction and Descriptive Statistics

In studies examining China’s CGPC, actual grain production is frequently utilized as an indicator of this capacity. In terms of determinants of actual grain yield, Wang Yuling (1999) posited that factors affecting CGPC include labor, land, agricultural water conservancy facilities, and chemical fertilizers [30]. Xiao Haifeng and Wang Jiao (2004) identified sown area and fertilizer input as primary determinants of China’s CGPC [31]. Wang Guomin and Zhou Qingyuan (2016) categorized influencing factors into modern and traditional dimensions: modern factors encompass fixed assets in agriculture, production investment, chemical fertilizer application, and total agricultural machinery power; traditional factors include disaster-affected area, grain sown area, and grain prices [32]. Fan (2000) and Zhu (2004) argued that technological R&D and irrigation serve as key drivers of agricultural growth in China [33,34]. Collectively, existing research categorizes the factors influencing actual grain yields into four primary groups. The first group encompasses the inputs of production factors, which include land, labor, capital, fertilizers, and equipment. The second group pertains to advancements in technology. The third group involves Government policy support. Finally, the fourth group addresses the effects of natural disasters. Due to data limitations, this study analyzes data from 31 provinces covering the period from 2011 to 2022.

An ordinary panel regression model is initially constructed, as

$$\begin{array}{c}{\left(\mathit{ln}Y\right)}_{it}=C+a_1{\left(\mathit{ln}{X}_1\right)}_{it}+a_2{\left(\mathit{ln}{X}_2\right)}_{it}+a_3{\left(\mathit{ln}{X}_3\right)}_{it}+a_4{\left(\mathit{ln}{X}_4\right)}_{it}+a_5{\left(\mathit{ln}{X}_5\right)}_{it}+a_6{\left(\mathit{ln}{X}_6\right)}_{it}{+a}_7{\left(\mathit{ln}{X}_7\right)}_{it}+{\epsilon }_{it}\end{array}$$

(8)

where $Y$ represents the total grain output (10,000 tons), $a_1,a_2,a_3,a_4,a_5,a_6,a_7$ serves as the dependent variable in this analysis, $X_1$ represents the sown area of grain (1000 hectares), $X_2$ represents the quantity of fertilizers utilized (10,000 tons), $X_3$ represents the labor force input (10,000 people), $X_4$ represents the local financial expenditure on agriculture, forestry, and water affairs (100 million yuan), $X_5$ represents the total machinery power (10,000 KW), $X_6$ represents the affected area (thousand hectares), $X_7$ represents the level of digital economy development, measured by Peking University’s digital financial inclusion index [35], and ${\epsilon }_{it}$ represents the residual value. Descriptive statistics for each variable are presented in

Table 2. Descriptive statistics of variables.

Variable	Obs	Mean	Std.	Min	Max
lny	372	7.036	1.336	3.360	8.971
lnx1	372	7.652	1.321	3.839	9.594
lnx2	372	4.696	1.272	1.030	6.574
lnx3	372	6.105	1.140	3.045	7.849
lnx4	372	6.186	0.589	4.519	7.215
lnx5	372	7.273	1.089	4.694	9.480
lnx6	372	5.711	1.757	0	8.348
lnx7	372	5.331	0.673	2.786	6.133

:

However, it has been demonstrated that actual grain yields display significant spatial correlation; thus, neglecting spatial heterogeneity in grain yields may result in biased measurement outcomes. To assess spatial autocorrelation, traditional econometric regression models are no longer suitable. Consequently, this study will employ a spatial econometric model to analyze the factors influencing grain production. Therefore, this study develops the subsequent Spatial Durbin Model,

$$\begin{array}{ll}{\left(\mathit{ln}Y\right)}_{it}=C& +\beta W{\left(\mathit{ln}Y\right)}_{it}+a_1{\left(\mathit{ln}{X}_1\right)}_{it}+a_2{\left(\mathit{ln}{X}_2\right)}_{it}+a_3{\left(\mathit{ln}{X}_3\right)}_{it}+a_4{\left(\mathit{ln}{X}_4\right)}_{it}+a_5{\left(ln{X}_5\right)}_{it}+a_6{\left(ln{X}_6\right)}_{it}\\ & +a_7{\left(ln{X}_7\right)}_{it}+{\theta }_{1}W{\left(\mathit{ln}{X}_1\right)}_{it}+{\theta }_{2}W{\left(\mathit{ln}{X}_2\right)}_{it}+{\theta }_{3}W{\left(\mathit{ln}{X}_3\right)}_{it}+{\theta }_{4}W{\left(\mathit{ln}{X}_4\right)}_{it}+{\theta }_{5}W{\left(\mathit{ln}{X}_5\right)}_{it}\\ & +{\theta }_{6}W{\left(\mathit{ln}{X}_6\right)}_{it}+{\theta }_{7}W{\left(\mathit{ln}{X}_7\right)}_{it}+{\epsilon }_{it}\end{array}$$

(9)

where $\beta$ represents the spatial regression coefficient of the explanatory variables, $\theta$ represents the spatial error coefficient of the explanatory variables, and $W$ represents the neighbor weight matrix.

Nonetheless, the particular model to be chosen demands further examination. As a result, this paper presents the following assumptions:

H0. 

The grain production of provinces and cities is independently distributed spatially (no spatial autocorrelation).

H1. 

There exists significant spatial autocorrelation in the grain production of provinces and cities.

5.2. Spatial Autocorrelation Test

5.2.1. Spatial Weights Matrix

The selection of the weight matrix is essential for examining the spatial distribution characteristics of grain yield. After applying different weight matrices, this study has found that the spatial adjacency weight matrix better reflects the actual grain production situation of provinces and cities in China. The spatial adjacency matrix is constructed based on geospatial adjacency, which is based on the principle of defining two spatial units as neighbors when they share a common boundary. During the construction process, the matrix element $W_{ij}$ is assigned a value of 1 if the space cells $i$ and $j$ are adjacent; otherwise, $W_{ij}$ is 0. Additionally, the diagonal elements are generally assigned a value of 0 to mitigate the potential interference of self-referential neighbor relationships in the analysis.

$$W=\left\{\begin{array}{l}1,\mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}i\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}j\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{d}\mathrm{j}\mathrm{a}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\\ 0,\mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}i\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}j\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{d}\mathrm{j}\mathrm{a}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\end{array}\right.$$

(10)

5.2.2. Global Moran’s I

Global spatial autocorrelation refers to the spatial representation of a geographic phenomenon or attribute value across an entire region. The primary objective of determining whether such a phenomenon or attribute value exhibits spatial aggregation properties is to establish this fact, with the key analytical tool being the Global Moran’s I statistic. The formula for calculating Moran’s I is as follows:

$$\begin{array}{c}I={\displaystyle \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{W}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{S^2{\sum }_{i=1}^n{\sum }_{j=1}^n{W}_{ij}}}\end{array}$$

(11)

In this study, n = 31, and represent the sample values in space i and space j, respectively, $\overline{x}$ represents the mean of x, $S^2$ represents the variance of x, and $W_{ij}$ represents the spatial weight matrix. Total grain production in 2011–2022 for the whole global Moran’s I is specifically shown in

Table 3. Global Moran’s I, 2011–2022.

Year	I	p-Value
2011	0.199	0.052
2012	0.209	0.042
2013	0.203	0.047
2014	0.199	0.051
2015	0.210	0.042
2016	0.219	0.034
2017	0.214	0.038
2018	0.210	0.040
2019	0.205	0.044
2020	0.200	0.049
2021	0.196	0.053
2022	0.196	0.054

.

The findings in Table 3 reveal that Moran’s I is positive in the global region between 2011 and 2022, fluctuating between 0.196 and 0.219, with all values being significant at approximately the 5% level. This indicates that grain output in the 31 provinces across the country exhibits significant positive spatial correlation, meaning provinces with high grain output tend to be adjacent to other provinces with high grain output, while those with low output are similarly clustered with neighboring provinces. This rationale further validates the choice of spatial econometric models to analyze the factors affecting grain yield.

5.2.3. Moran Scatterplot

The Global Moran’s I reveals the spatial correlation of the total grain output of 31 provinces in China as a whole. However, its limitation is that it fails to intuitively present the detailed spatial correlation characteristics among different provinces. In contrast, local spatial autocorrelation analysis effectively addresses this key limitation, providing a more comprehensive understanding of the spatial patterns.

Local spatial autocorrelation effectively reveals unique spatial correlation patterns within individual provinces. By employing the Moran scatter plot, we can clearly distinguish four distinct spatial correlation patterns:

Quadrant 1. High–high agglomeration (high grain production in the province and its neighboring provinces);

Quadrant 2. Low–high agglomeration (low grain production in the province but high production in neighboring provinces);

Quadrant 3: Low–low agglomeration (low grain production in the province and its neighboring provinces);

Quadrant 4: High–low agglomeration (high grain production in the province but low production in neighboring provinces).

From Figure 3 and Figure 4, it can be seen that the main grain-producing areas have continued to stabilize at high levels and exhibit high aggregation trends. With the advancement of agricultural modernization, such as the construction of high-standard farmland, the scope and intensity of aggregation may further strengthen, as seen in provinces like Heilongjiang and Henan, which are influenced by the radiation-driven role of surrounding areas. On the other hand, regions with poor natural conditions and weak agricultural bases may remain in the low–low aggregation quadrant for an extended period, such as Qinghai and Tibet. However, with the emphasis and support from the State on balanced agricultural development, there is a gradual improvement and potential shift to other quadrants. Provinces and cities in the low–high and high–low quadrants are fewer in number and may experience changes from year to year in response to their own agricultural policies, structural optimizations, and other factors.

5.3. LM, LR Tests and Model Selection

To select an appropriate spatial econometric model for subsequent regression analysis, this study performs the LM test (Lagrange multiplier test) and the LR test (Likelihood ratio test) on the variables defined in the previous section. The results of the LM test, as shown in

Table 4. Results of the LM test.

Lagrange Multiplier Test Value	Statistic	p-Value
LM-Error	8.254	0.004
Robust LM-Error	9.122	0.003
LM-Lag	12.857	0.000
Robust LM-Lag	13.725	0.000

, indicate that LM-Error, LM-Lag, Robust LM-Error, and Robust LM-Lag are all significant at the 1% level. These results clearly indicate that the spatial measurement equations contain both a spatial lag term and a spatial error term, thus the Spatial Durbin Model (SDM) is the most appropriate choice for regression analysis. Additionally, to examine whether the Spatial Durbin Model (SDM) may degenerate under the condition of a spatial neighbor weight matrix, further tests were conducted. The results of the LR test, as shown in

Table 5. Results of the LR test.

Likelihood Ratio Test Value	Statistic	p-Value
LR-SLM	53.080	0.000
LR-SEM	71.010	0.000

, reveal that both LR-SLM and LR-SEM are significant at the 1% level. Therefore, the SDM model does not degenerate into the SEM model.

Based on the above hypothesis tests, when exploring the factors affecting grain production, the Spatial Durbin Model should be selected for analysis, indicating that H1 is validated.

5.4. Results and Analysis

Table 6. The regression results of the Spatial Durbin Model.

	(1)	(2)	(3)
	ind	time	both
Main
lnx1	0.877 ***	0.871 ***	0.837 ***
	(35.55)	(35.87)	(31.76)
lnx2	0.107 ***	0.118 ***	0.118 ***
	(3.36)	(5.88)	(3.68)
lnx3	0.0224	−0.115 ***	0.0475
	(0.77)	(−4.63)	(1.61)
lnx4	−0.0441	0.0477	−0.0551 *
	(−1.95)	(1.61)	(−2.34)
lnx5	0.0617**	0.194 ***	0.0519 **
	(3.23)	(7.69)	(2.59)
lnx6	−0.00254	−0.0567 ***	−0.00247
	(−0.66)	(−7.11)	(−0.64)
lnx7	0.134 ***	0.198 **	0.136 ***
	(4.36)	(3.11)	(4.45)
Wx
lnx1	−0.472 ***	−0.179	−0.493 ***
	(−5.83)	(−1.13)	(−5.85)
lnx2	−0.122 *	−0.163 ***	−0.0410
	(−2.41)	(−3.45)	(−0.67)
lnx3	−0.0315	0.00607	0.0876
	(−0.71)	(0.11)	(1.54)
lnx4	0.0236	−0.00599	−0.0637
	(0.75)	(−0.09)	(−1.42)
lnx5	−0.0944 ***	0.0761	−0.133 **
	(−4.12)	(1.18)	(−3.17)
lnx6	−0.00718	0.0382 *	−0.00245
	(−0.98)	(2.22)	(−0.32)
lnx7	−0.0996 **	−0.0262	−0.0491
	(−3.16)	(−0.25)	(−1.09)
Spatial
rho	0.302 ***	0.289	0.192 **
	(4.48)	(1.75)	(2.59)
Variance
sigma2 e	0.00194 ***	0.0145 ***	0.00188 ***
	(13.50)	(8.42)	(13.46)
r2	0.946	0.973	0.934
N	372	372	372

Note: t statistics in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001.

reports the regression results of the Spatial Durbin Model (SDM) under the spatial neighbor weight matrix. The value of the spatial autoregressive coefficient (rho) is significantly positive in both individual fixed and double fixed effects, indicating a strong positive spatial spillover effect of the explanatory variable, total grain production. This suggests that changes in total grain production in one region affect neighboring regions with the same trend, further validating the spatial autocorrelation of grain production.

The coefficient associated with the sown area of grain (x₁) is 0.877, which is statistically significant at the 1% level, indicating that an increase in the sown area of grain has a substantial positive impact on the total grain production within the region. The coefficient for fertilizer use (x₂) is 0.107, which is statistically significant at the 1% level, indicating that increased fertilizer use is effective in enhancing total grain yield. Additionally, both total mechanical power (x₅) and the level of digital economy development (x₇) are significantly positive in all effects, indicating that advancements in mechanization and the development of the digital economy substantially contribute to grain production. The performance of local financial spending on agriculture, forestry, and water resources (x₄) is more particular, most likely because increases in local financial expenditures do not directly translate into increased grain production. Because of China’s vast territory and relatively decentralized spatial distribution, the impact of natural geographic conditions on agricultural production varies significantly across regions, and geographic differences may weaken the actual effects of local fiscal expenditures. For example, in places with low water supplies or poor soil quality, even if local financial inputs for agricultural infrastructure development are increased, their contribution to grain production may be limited due to these intrinsically inadequate natural conditions. Furthermore, the results of the spatial lag terms demonstrate the geographical spillover effects. The spatial lag terms for $x_1$ and $x_2$ were found to be significantly negative in both individual fixed effects and two-way fixed effects models. This suggests that increases in the sown area of grain and fertilizer use in neighboring regions may exert a dampening effect on local grain production.

5.5. Robustness Testing

To assess the dependability of the Spatial Durbin Model’s estimation results, this study performs a robustness test by substituting the weight matrix and removing explanatory variables. The results suggest that the model is highly resilient. In

Table 7. Robustness testing results.

	(1)	(2)	(3)
	W2	W1	W1
Main
lnx1	0.891 ***	0.871 ***	0.865 ***
	(43.27)	(35.87)	(37.26)
lnx2	0.0872 ***	0.118 ***	0.109 ***
	(4.12)	(5.88)	(5.60)
lnx3	−0.102 ***	−0.115 ***	−0.0946 ***
	(−4.59)	(−4.63)	(−4.34)
lnx4	0.0577	0.0477
	(1.74)	(1.61)
lnx5	0.199 ***	0.194 ***	0.210 ***
	(7.78)	(7.69)	(8.90)
lnx6	−0.0547 ***	−0.0567 ***	−0.0552 ***
	(−7.23)	(−7.11)	(−7.28)
lnx7	0.152 *	0.198 **	0.240 ***
	(2.52)	(3.11)	(4.05)
Wx
lnx1	0.692 *	−0.179	−0.160
	(2.47)	(−1.13)	(−1.05)
lnx2	−0.145	−0.163 ***	−0.171 ***
	(−0.69)	(−3.45)	(−3.71)
lnx3	−0.430 **	0.00607	−0.00453
	(−2.76)	(0.11)	(−0.10)
lnx4	0.267	−0.00599
	(1.24)	(−0.09)
lnx5	0.148	0.0761	0.0790
	(0.64)	(1.18)	(1.31)
lnx6	0.0420	0.0382 *	0.0290
	(0.78)	(2.22)	(1.77)
lnx7	0.202	−0.0262	−0.0441
	(0.56)	(−0.25)	(−0.44)
Spatial
rho	−0.114	0.289	0.289
	(−0.55)	(1.75)	(1.79)
Variance
sigma2 e	0.0146 ***	0.0145 ***	0.0146 ***
	(13.70)	(8.42)	(8.61)
r2	0.939	0.973	0.971
N	372	372	372

Note: t statistics in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.00.1.

, (1) represents the regression results of the time-fixed model after replacing the weight matrix with an inverse distance weight matrix, (2) represents the regression results of the time-fixed model with the original weight matrix, and (3) represents the regression results of the time-fixed model after eliminating the lower without replacing the weight matrix. The sign and significance of the explanatory variable coefficients remain virtually unchanged after altering the weight matrix, as compared to the original weight matrix regression findings. In addition, the sign and significance of the coefficients of the remaining explanatory variables remain stable after being removed without replacing the weight matrix, indicating that the removal of $x_4$ has no significant effect on the overall structure of the model or the effects of other variables, as well as reflecting that changes in individual explanatory variables have a lower impact on the model. In summary, the aforementioned robustness test has improved the reliability of this study’s estimation results, and the conclusions drawn about the impact of the factors on total grain output have a high level of confidence.

6. Conclusions and Policy Implications

Based on the Entropy Weight Method and the Spatial Durbin Model, this study investigates China’s total grain production capacity and the factors that influence it. The results suggest that the sown area of grain, fertilizer application, total mechanical power, and the level of digital economic development are the most important positive elements contributing to the rise of total grain output, with the sown area of grain having the greatest direct effect. These findings not only highlight each factor’s direct impact on total grain production, but also serve as a crucial foundation for developing scientific and reasonable agricultural policies. Overall, increasing the sown area of grain, optimizing chemical fertilizer use, and improving agricultural production digitization should be the primary goals for future agricultural output growth.

Based on the conclusions of the preceding study and analysis, this study proposes the following three policy proposals to enhance the region’s comprehensive grain production capability:

To effectively enhance the sown area of grain, a comprehensive three-pronged strategy should be implemented, encompassing guiding policies, the optimization of land resources, and innovation in production technologies. Firstly, the government and relevant agencies can incentivize farmers to expand the scale of food crop cultivation while ensuring their economic viability by establishing supportive policies. These may include providing subsidies for arable land, enhancing the agricultural insurance system, and refining the land transfer mechanism. Secondly, it is necessary to focus on the transformation of medium- and low-yield fields and the development of high-standard farmland. This can be achieved by improving the quality of arable land through remediation and soil enhancement techniques, as well as converting idle or underutilized land into high-quality arable land that is conducive to grain production. Furthermore, contemporary scientific and technological tools, including remote sensing monitoring and geographic information systems, should be comprehensively employed to accurately evaluate the potential of land resources and to strategically plan the distribution of planting areas. This approach aims to prevent the waste of resources and mitigate ecological risks associated with indiscriminate expansion. Concurrently, it is essential to focus on ecological protection; while increasing the sown area of grain, take measures such as crop rotation and fallowing. These measures will facilitate the harmonious development of grain production alongside the ecological environment, thereby promoting the sustainable production of regional grain.

To achieve sustainable agricultural development, it is essential to optimize fertilizer usage and increase the proportion of alternative materials. To begin, the promotion of precision fertilization technology is critical. This entails the establishment of a regional soil information database and the application of remote sensing technology, as well as unmanned aerial vehicles (UAVs), to gather real-time data, thereby providing scientific support for precision fertilization. Second, it is critical to increase the research and promotion of novel, high-efficiency fertilizers. This includes bio-fertilizers, organic fertilizers derived from agricultural waste, slow-release fertilizers, and novel biocontrol products. Implementing these measures will reduce reliance on conventional chemical fertilizers and insecticides. Furthermore, the government should set aside special funding in its budget to assist in the development of critical technologies such as soil evaluation and fertilizer development. The development of mid- to long-term strategic plans will clarify research aims and orientations, resulting in more effective resource allocation and policy advice. Moreover, the creation of green agricultural development demonstration zones will allow for pilot testing and the promotion of breakthrough technology. Finally, integrating resources from agricultural, scientific, and environmental departments is critical. Building a cross-domain collaboration platform will encourage information sharing, provide regular training for farmers and technical workers, and improve the practical implementation of emerging technology. These techniques not only increase grain production capacity but also minimize environmental pollution, resulting in a win–win situation with both economic and ecological benefits.

Last but not least, the development of the digital economy should be accelerated to empower smart agriculture. First, increasing investment in scientific research and the construction of digital infrastructure in rural areas—such as high-speed internet coverage and the widespread adoption of smart devices—provides essential hardware support for the digital transformation of agriculture. For instance, 5G networks are being deployed in rural areas to provide high-speed and reliable internet connections, which are critical for big data transmission and real-time monitoring. Simultaneously, the use of smart sensors, drones, and other equipment in agricultural production has improved, allowing for the more exact monitoring and management of farming environments. Secondly, the development of an agricultural big data platform is being promoted to integrate information resources such as meteorology, soil conditions, and market trends. The goal of this program is to help people make more informed decisions about precision agriculture production and to improve resource efficiency. By constructing a national agricultural big data center, the platform collects and analyzes meteorological data, soil quality metrics, and market demand patterns from across the country, resulting in a comprehensive collection of agricultural knowledge maps. Using these data, farmers may create more scientifically sound planting plans, irrigation techniques, and fertilization programs, reducing resource waste and optimizing production processes. Finally, enterprises are encouraged to collaborate with research institutions to develop digital solutions applicable to agriculture, such as smart irrigation systems and drone monitoring technologies. Incorporating artificial intelligence (AI) technologies can help reduce production costs and increase yields. For example, AI-based smart irrigation systems may automatically alter the quantity of water delivered based on real-time weather conditions and soil moisture levels, ensuring that crops grow in the best conditions. Meanwhile, drones equipped with high-resolution cameras and multi-spectral sensors may survey fields on a regular basis, identify weeds, pests, and illnesses, and provide farmers with fast feedback to help them apply pesticides or weed precisely.

These integrated digital technology applications can not only dramatically raise agricultural productivity, but also improve agriculture’s resilience to natural calamities. They encourage the development of sustainable and ecological farming practices, achieve sustainable growth in grain production, and provide a strong assurance of national food security.