Spatial Analysis of Agriculture Ecological Efficiency and Its Influence on Fiscal Expenditures

Abstract

The faster development of the agriculture sector has led to the excessive waste of agriculture resources and that development causes serious damage to the ecological environment. So, ecological agriculture has become an important topic to understand the relationship between agriculture and the environment. This study tries to examine the impact of agricultural fiscal expenditures on agricultural ecological efficiency. In the first step, the researcher found the ecological efficiency of agriculture. The second step tries to examine the agricultural fiscal expenditure impact on agriculture’s ecological efficiency. The study used data from 30 provinces of China from 1998 to 2018. For the empirical results, the researcher used the super-efficiency DEA window analysis in the first step. Moreover, we used the spatial Durbin model in the second step. The results of the study revealed that China’s agricultural ecological efficiency shows a downward trend to an upward trend. The efficiency value of the eastern region is higher than the central region’s efficiency value, which is lower than the national average level. The western region’s efficiency value is equal to the national average level, and the differences between regions vary with the time period. The nuclear density of agricultural eco-efficiency in China showed a trend of “left-shift rising—double peaks—right-shifting decline”. In terms of space, China’s agricultural ecological efficiency has a significant positive spatial spillover effect, especially in the eastern and western regions. The results of the spatial Durbin model revealed that agricultural fiscal expenditures have a significant role in promoting China’s agricultural ecological efficiency. While agricultural governance has shown a significant spatial crowd-out effect. Therefore, it is proposed to promote the development of inter-provincial agricultural cooperation, and the formulation of agricultural financial support strategies should be green-oriented and should consider regional heterogeneity.

1. Introduction

Since the reform and opening up, agriculture has been driven by China’s agricultural subsidies and has achieved rapid development. While the output level has been continuously improved, the “reverse ecologicalization” brought about by agriculture has not. The problem has also become more prominent. The environmental protection of China started in the 1970s [1]. From the 1970s to the 1990s, when traditional agriculture transitioned to modern agriculture, the problem of agricultural pollution became prominent. During this period, China’s rural environmental protection policy system began to be established. In the 1990s, due to the excessive abuse of chemical fertilizers and pesticides, the problem of agricultural non-point source pollution in China became more and more serious, and surveys showed that more than 84% of farmers would use pesticides in excess of the prescribed standard dose [2]. Now, in this modern era of technology, agricultural activities are one of the main sources of methane and nitrous oxide production in China. According to statistics in the Third National Information Bulletin on Climate Change of the People’s Republic of China released in 2019, methane and nitrous oxide gases produced by agricultural activities in China accounted for 40.5% and 65.4% of the national total emissions, respectively, so the voice for the development of agriculture-ecology is getting louder. The proportion of agricultural non-point source pollution in the overall environmental pollution has gradually increased, but the prevention and control of agricultural non-point sources of pollution are still basically at a blank stage at this stage [3]. Agricultural non-point sources of pollution in some areas even exceeded industrial pollution [4].

Agricultural eco-efficiency is an extension of eco-efficiency, which refers to obtaining as much agricultural output with as little environmental cost and resource consumption as possible under a certain combination of input elements. The research on agricultural ecological efficiency in China mainly focuses on the following two levels: one is the national level, and the other is the provincial, city (county), and district levels [5]. However, the influencing factors of China’s agricultural ecological efficiency are still in a state of contention and there is no unified conclusion. These influencing factors include but are not limited to the income level of rural residents [6], production characteristics, technical conditions and social structure [7], financial support to agriculture strength and industrialization level [8], location [9], agricultural land ownership confirmation, agricultural production capacity, agricultural machinery density, agricultural scale level, agricultural disaster rate, industrialization level, urbanization level [10], multiple cropping index, and planting structure [11].

Ref. [12] found the agriculture fiscal expenditures efficiency by applying the three-stage DEA model in 11 cities of Pearl River Xijiang. The study results revealed that agriculture’s fiscal expenditure efficiency has declined. The fiscal agriculture expenditures have a different regional impact on the economic growth of China. The effect on the eastern province is smaller, but the effect on the western provinces is larger [13]. Taking the panel data of 16 prefecture-level cities in Anhui province, we found the efficiency of fiscal supporting agriculture expenditures. The study results revealed that the overall efficiency of agriculture fiscal expenditures is relatively high [14]. The study found the efficiency of agriculture fiscal expenditures by collecting the data from nine prefectures and cities in Fujian province. The result of the study explained that agriculture fiscal expenditure efficiency in various cities is not high [15]. Through the above discussion, this paper argues that the issue of China’s agricultural ecological efficiency has high research value in today’s increasingly tense environment and resource constraints. The study adopts the Window analysis method of the super-efficiency DEA model to calculate China’s agricultural ecological efficiency. Through the construction of a spatial econometric model, it explores the spatial spillover effects of China’s agricultural eco-efficiency. After estimating the spatial spillover effect of China’s agricultural ecological efficiency, we have tried to examine the agriculture ecological efficiency impact on agriculture fiscal expenditures.

2. Materials and Methods

2.1. Data Sources

This study selected data from 30 provinces in China for 21 years from 1998 to 2018. Due to the particularities of agricultural production in Tibet, Hong Kong, Macao, and Taiwan, these are not included in the scope of the study. The data involved in the research comes from the data published by the National Bureau of Statistics, the “First National Pollution Source Census Bulletin,” and other statistical data. Some missing data are imputed by various methods such as the EM method and linear regression method using SPSS 23.0 statistical software and choosing the best option.

Indicator selection for Input and Output

Agriculture, in a broad sense, includes agriculture, forestry, animal husbandry, and sideline fishing. Agriculture in a narrow sense refers to planting, which is the basis of agriculture and the most important part of agriculture’s ecological impact. This paper takes narrow agriculture (planting) as the research objective to measure agricultural ecological efficiency. The input-output indicators and variable selection are shown in

Table 1. Input–output indicators selection and variable description.

First-Level Indicator	Secondary Indicators	Variable Description
Input Indicator	Labor input	Number of employees in agriculture, forestry, animal husbandry, and fishery × (gross agricultural output value/total output value of agriculture, forestry, animal husbandry, and fishery) (ten thousand people)
	Land input	The total sown area of crops (thousand hectares)
	Irrigation input	Effective irrigated area (thousand hectares)
	Mechanical input	Total power of agricultural machinery (10,000 kilowatts)
	Diesel input	Agricultural diesel application amount (10,000 tons)
	Input of draft animals	Number of large livestock at the end of the year (10,000 heads)
	Fertilizer input	Agricultural chemical fertilizer application amount (10,000 tons)
	Pesticide input	Pesticide application amount (10,000 tons)
	Agricultural film input	Agricultural plastic film application amount (10,000 tons)
Output Indicator	Expected output	Gross agricultural output value (100 million yuan) (constant price in 1978)
	Undesired output	Agricultural non-point source pollution (cubic kilometers)
		Carbon emissions (tons)

.

Agricultural production includes a variety of input indicators. Referring to previous studies, this paper selects 8 main input indicators and selects 8 variables to characterize them. All variables have direct data except that labor input data is estimated based on agricultural, forestry, animal husbandry, and sideline fishery practitioners. To maintain a uniform statistical caliber, the expected agricultural output index is represented by the total agricultural output value. At the same time, to eliminate the impact of price factors, all data have been adjusted to take 1978 as the constant price.

Undesirable agricultural output includes agricultural non-point-source pollution and agricultural carbon emissions. Among them, agricultural non-point-source pollution involves three indicators. The loss of fertilizer’s nitrogen and phosphorus, the ineffective use of pesticides, and the residual amount of agricultural film characterize the pollution level. To improve the validity of the DEA model calculation, referring to the experience of the three indicators, they were combined into a comprehensive indicator of “agricultural non-point source pollution” through the entropy method [9]. The correlation coefficient mainly adopts the literature research method and the relevant data published by the National Bureau of Statistics and also refers to the “First National Pollution Census: Manual of Fertilizer Loss, Pesticide Loss, and Plastic Film Residue Coefficients”. Considers the impact of geographical gaps as much as possible in the accounting process [16,17]. Generally speaking, agricultural carbon emissions mainly come from the following aspects: emissions caused by the production and use of three major agricultural chemicals such as fertilizers, pesticides, and agricultural films; emissions caused by the consumption of fossil fuels (mainly diesel) by agricultural machinery; emissions indirectly caused by electricity consumption for irrigation (mainly thermal power generation); loss of organic carbon caused by agricultural tillage. Referring to previous studies, the emission coefficients of the six types of emission sources are respectively 0.8956 (kg/kg) for fertilizers, 4.9341 (kg/kg) for pesticides, 5.18 (kg/kg) for agricultural films, 0.5927 (kg/kg) for diesel, and 0.5927 (kg/kg) for agricultural irrigation, 20.476 (kg/km²), agricultural sowing 312.6 (kg/km²) [18,19].

2.2. Super-Efficient DEA Model

The measurement of agricultural ecological efficiency mainly includes the following two categories: the parametric method and the non-parametric method. These methods are represented by stochastic frontier analysis (SFA) and the data envelopment method (DEA). The stochastic frontier method calculates efficiency values by constructing the corresponding production function, which can avoid the effects of extreme values and is suitable for analysis of large samples. The data envelopment method does not need to set parameters, so it can avoid model setting errors. At the same time, the data envelopment method can handle multiple output indicators, which is an irreplaceable advantage of SFA. This study involves multiple output indicators, so the DEA model has been selected for analysis. Considering that there are many input indicators selected in this study, multiple indicators may be effective, and the super-efficiency model allows the efficiency value of the decision-making unit to be greater than or equal to zero, which avoids the calculation error of the efficiency value and can fully sort the efficiency value of each decision-making unit. The research model is constructed as follows:

$$\begin{array}{c}\begin{array}{c}min\theta -\epsilon ({\displaystyle {\displaystyle \sum }_{i=1}^m}{S}_i^{-}+{\displaystyle {\displaystyle \sum }_{r=1}^s}{S}_r^{+}),\\ s.t. {\displaystyle {\displaystyle \sum }_{j=1}^n}{x}_{ij}{\lambda }_j+S_i^{-}=\theta x_{ij}0 i=1,2\dots ,m\\ {\displaystyle {\displaystyle \sum }_{j=1}^n}{y}_{ij}{\lambda }_j-S_r^{+}=y_{rj0} r=1,2,\dots ,s\end{array}\\ {\lambda }_j,S_i^{-},S_r^{+}\ge 0,j=1,2,\dots j_0-1,j_0+1,\dots ,n\end{array}$$

(1)

Among them, θ is the super-efficiency value of the j₀ decision-making unit, ε is a non-Archimedes infinitesimal quantity. n is the number of decision-making units, each decision-making unit includes m input variables and s output variables, $S_i^{-}$, $S_r^{+}$ are the input and output slack variables, respectively x_ij represents the value of the jth decision-making unit at the ith input index, y_ij represents the value of the j decision unit on the rth output variable. λ_j is the input and output index. The weight coefficient of θ, λ_j, $S_i^{-}$, $S_r^{+}$ are unknown parameters, which can be solved by the model.

Window Analysis

It is necessary to conduct a comparative analysis between different provinces, and at the same time, it is necessary to study the annual differences caused by time changes. The DEA model is a non-parametric model, and its results have been affected by the input-output data of the decision-making unit. The calculated efficiency is relative efficiency, which only has internal comparison significance. If the calculation is performed with cross-sectional data, time information cannot be reflected. If the calculation is performed with time-series data, the comparability between individuals will be lost. Therefore, the study adopts DEA window analysis to avoid the trade-off between individuals and time. The principle of DEA window analysis is to classify the data of multiple periods into one window, and each window takes the same number of periods. The same decision-making unit in different periods has been regarded as different decision-making units for comparison. Different decision-making units in different periods form a reference set, which makes it meaningful for both horizontal and vertical comparison. However, there is no uniform standard for the window width. With reference to the experience of previous scholars and the length of the panel data used in this study. The following window parameter setting method has been adopted: If the window width m is an even number and w is the sample period, then m = (w + 1)/2 ± 1/2; if the window width m is odd, then m = (w + 1)/2. The interpretation of the results of the window analysis mainly includes the mean efficiency of all decision-making units in different windows. The standard deviation of all the efficiencies of decision-making units in different windows, the comprehensive column distance, and the full distance. The column distance is the difference between the maximum and minimum values of efficiency values in different windows of the same period. The comprehensive column distance represents the difference between the maximum column distance and the minimum column distance in the same decision-making unit. The full distance represents the difference between the maximum efficiency value and the minimum efficiency value under different windows of the same decision-making unit in different periods. The smaller the difference between each column distance, the more stable and reliable the calculation result is.

2.3. Spatial Spillover and Financial Expenditure Impact of Agriculture Ecological Efficiency

Due to the differences in resource endowment, agricultural economy, and geographical location between different regions, there are differences in agricultural ecological efficiency. The difference in regional spatial distribution may have spatial autocorrelation; that is, the change in agricultural ecological efficiency in a certain region will also affect the agricultural ecological efficiency of its neighboring areas through technological spillovers, factor flow, and other reasons [11]. In this case, it is necessary to study and analyze the spatial spillover effect of regional agriculture’s ecological efficiency.

2.3.1. Spatial Measurement Model

Referring to [20], this paper first assumes the following general spatial econometric model:

$$\{\begin{array}{c}AE{E}_{it}=\rho w_i^{\prime }{y}_t+x_{it}^{\prime }\beta +d_i^{\prime }{X}_t\delta +u_i+{\gamma }_t+{\epsilon }_{it}\\ {\epsilon }_{it}=\lambda m_i^{\prime }{\epsilon }_t+v_{it}\end{array}$$

(2)

AEE_it it is the agricultural ecological efficiency value in year t in region i, ρ is the spatial lag coefficient of the dependent variable, $w_i^{\prime }$ is the ith row of the spatial weight matrix y_t is the dependent variable in year t in region i, $x_{it}^{\prime }$ is the independent variable in year t in the i region, β is the independent variable influence coefficient, $d_i^{\prime }$ is the ith row of the spatial weight matrix D, and x is the independent variable matrix. The 12 × 4 matrix of the four indicators of “agriculture science and technology financial expenditure and agricultural education financial expenditure”, δ is the spatial lag coefficient of the explanatory variable, u_i is the individual effect, y_t is the time effect, and $m_i^{\prime }$ is the disturbance term of the spatial matrix M. In row i, λ is the spatial error term coefficient, where W, D, and M are equal in this paper. Regression is carried out according to the constructed general model, and then the model is tested as follows: when λ = 0, the model is a Durbin model (SDM), and when ρ ≠ 0 and δ = 0, the model is a spatial autoregressive model (SLM); When δ = −β λ, the model is a spatial error model (SEM).

2.3.2. Spatial Matrix Construction

Spatial matrices are indispensable in spatial econometric models, but the construction and selection of spatial matrices is still a highly controversial issue. There is no unified objective standard for the construction of spatial matrices in academia. Testing and so on all need to rely on the subjectively constructed spatial distance matrix. To avoid the bias caused by subjective construction, the study attempts to construct multiple matrices to screen out the ideal matrix.

Referring to previous research experience, for the first time, the author used ArcGIS 10.2 to calculate the centroid position of each province and city. Moreover, we used two distance formulas, spherical distance (using a geographic coordinate system map) and Euclidean distance (using a plane projection map) to construct an inverse distance matrix. As in Formula 3, Among them, d is the distance between the center positions of regions i and j, and k is the attenuation coefficient. The decay index is selected as 1, 2, and 3, respectively. Using these six space matrices, with Stata 16.0 as the computing platform, the Moran index I test was performed (For the sake of brevity, the formula of Moran index I is not described here. For details, please refer to Pages 578–579 of Advanced Econometrics and STATA Applications (second edition), edited by Chen Qiang, or Pages 77–78 of Spatial econometrics as follows: Application Analysis Based on MATLAB, edited by Xiao Guangen et al. [20,21]). After testing, except for a few years, there is no spatial autocorrelation relationship between the agricultural ecological efficiency values in most other years.

$$w{d}_{ij}=\{\begin{array}{c}\frac{1}{d^k}, i\ne j\\ 0, i=j\end{array} (k=1,2,\dots )$$

(3)

Then consider building an adjacency matrix, as in Equation (3).

$$w{d}_{ij}=\{\begin{array}{c}1, (if i and j are adjacent to the rear (queen Contiguity\left)\right)\\ 0, (i=j or i and j are not adjacent to the rear)\end{array}$$

(4)

Considering that the provincial capital city, as the political and cultural center of each province, usually has a greater influence on the surrounding provinces and cities, referring to Equation (4). (Queen Contiguity: Two adjacent regions share common edges or vertices.) (In order to avoid the existence of isolated islands, this paper assumes that Hainan Province and Guangdong Province have a post-adjacent relationship.)

Considering the provincial capital city distance matrix, further economic information has been added, such as Equations (5)–(7), to construct an economic distance matrix. in, wd is the geographic distance matrix, E or per capita GDP, the matrix takes into account both geographical factors and economic factors and contains more comprehensive information.

$$wed=wd\times diag(\frac{{\overline{E}}_1}{\overline{E}},\frac{{\overline{E}}_2}{\overline{E}},\dots ,\frac{{\overline{E}}_n}{\overline{E}})$$

(5)

$${\overline{E}}_i=\frac{1}{t_k-t_0+1}{{\displaystyle \sum }}_{t=t_0}^{t_k}{E}_{it}$$

(6)

$$\overline{E}=\frac{1}{n(t_k-t_0+1)}{{\displaystyle \sum }}_{t=t_0}^{t_k}{{\displaystyle \sum }}_{i=1}^n{E}_{it}=\frac{1}{n}{{\displaystyle \sum }}_{i=1}^n{\overline{E}}_i$$

(7)

2.4. Indicator Selection for Agriculture Fiscal Expenditure Impact on Agricultural Eco-Efficiency

In

Table 2. Index Selection for Agriculture Fiscal Expenditure Impact on Agricultural Eco-Efficiency.

Index		Meaning	Expected
Explained Variable	Agriculture ecological efficiency	Obtained by calculating the input and output of agricultural production in each province and city	-
Core Explanatory Variables	Agricultural financial support	Local financial expenditure on agriculture, forestry, and water affairs	Negative
Other Explanatory Variables	Financial support for rural environmental governance	Gross agricultural output/Gross regional product × local environmental protection expenditure	Positive
	Rural science and technology financial support	Gross Agricultural Output/Gross Regional Product × Local Financial Science and Technology Expenditure	Positive
	Financial support for rural education	Gross agricultural output/Gross regional product × local financial education expenditure	Uncertain

, indicator selection agricultural expenditure, agricultural environmental governance fiscal expenditure, agriculture fiscal spending on science and technology, and agriculture educational support financial expenditure are measured by four components. The effect of China’s fiscal expenditure on agricultural ecological efficiency, the specific impact mechanism is the following: Agricultural financial support can increase the productive capital of farmers so that farmers can increase agricultural production through irrigation input, machinery input, diesel input, draft animal input, chemical fertilizer input, pesticide input, agricultural film input, and other input elements. On the one hand, increasing the expected output of agriculture will, on the other hand, increase the consumption of resources and environmental pollution, and increase the undesired output of agriculture. Agriculture fiscal expenditures for environmental governance can control agricultural non-point source pollution caused by the use of chemical fertilizers, pesticides, and agricultural film, and reduce undesired agricultural output. Moreover, to promote the transformation and upgrading of the agricultural industry and improve agriculture’s ecological efficiency. At the same time, protecting the ecological environment by means such as returning farmland to forests may also reduce the expected output of agriculture. Financial support for science and technology can promote the progress of agricultural technology, which is the key to the development of modern agriculture [22]. The realization of agricultural modernization is conducive to improving the efficiency of agricultural production and increasing the expected agricultural output. In addition, the state’s scientific and technological support can promote the research and development of environmentally friendly products such as green pesticides and agricultural films, thereby alleviating the non-point source pollution of agriculture caused by the use of pesticides, chemical fertilizers, and agricultural films, and reducing undesired agricultural output. On the one hand, financial support for education can improve the quality of the agricultural labor force, transfer the agricultural labor force, and promote the realization of large-scale agricultural production, thereby improving agricultural production efficiency and increasing the expected agricultural output. Farmers may possibly increase the input of chemical fertilizers, pesticides, agricultural film, and other chemicals to replace labor, which will lead to an aggravation of agricultural non-point source pollution and carbon emissions, increasing undesired agricultural output. Moreover, educational financial support can improve farmers’ awareness of ecological environmental protection and reduce the use of chemical fertilizers, pesticides, agricultural film, etc., thereby reducing undesired output and improving agricultural ecological efficiency.

3. Results

3.1. Temporal and Spatial Changes in China’s Agricultural Eco-Efficiency Value

3.1.1. Analysis of Agriculture Ecological Efficiency at the National Level

Referring to the above formula for setting the window width, set the window width of the 21-year data of the research sample to 10. The first window selects the 10-year data of 30 provinces and cities from 1998 to 2007, the second window selects the data from 1999 to 2008, and so on. A total of 11 windows are formed, and Matlab R2016a is used as the computing platform, referring to the code provided by [23], to obtain the ecological efficiency value of each year under all windows in each region and calculate the average ecological efficiency value of each region in each year. Figure 1 and

Table 3. Statistics on the 21-year average value of agricultural ecological efficiency in each region.

Efficiency	0.9–1	0.8–0.9	0.7–0.8	0.6–0.7	0.5–0.6
East	Shanghai, Hainan, Jiangsu	Beijing, Zhejiang, Guangdong	Fujian and Tianjin	Liaoning, Hebei	Shandong
Central	-	-	Hubei and Hunan	Henan, Heilongjiang, Jiangxi	Jilin, Anhui, Shanxi
West	-	Guizhou, Chongqing, Xinjiang, Shaanxi	Sichuan, Qinghai, Guangxi	Inner Mongolia	Ningxia, Yunnan, Gansu

and

Table 4. Average value of agricultural ecological efficiency in China from 1998 to 2018.

Area	Mean	Full Distance	Growth Rate	Area	Mean	Full Distance	Growth Rate	Area	Mean	Full Distance	Growth Rate
Shanghai	0.992	1.445	6.20%	Sichuan	0.787	0.700	0.30%	Hebei	0.620	0.669	1.10%
Hainan	0.933	1.563	−2.30%	Hubei	0.769	0.666	1.60%	Ningxia	0.590	0.709	1.80%
Jiangsu	0.923	2.371	4.60%	Qinghai	0.766	0.745	2.00%	Shandong	0.583	0.616	3.00%
Guizhou	0.887	2.195	2.80%	Guangxi	0.764	0.598	2.10%	Jilin	0.578	0.515	−2.60%
Beijing	0.871	1.324	6.00%	Hunan	0.764	0.871	5.50%	Anhui	0.573	0.482	1.70%
Zhejiang	0.865	2.252	3.20%	Tianjin	0.736	0.641	0.20%	Yunnan	0.532	0.483	1.80%
Guangdong	0.860	1.185	3.10%	Henan	0.667	0.703	2.10%	Gansu	0.510	0.436	0.10%
Chongqing	0.847	2.249	6.80%	Heilongjiang	0.664	0.847	2.60%	Shanxi	0.502	0.586	1.00%
Xinjiang	0.809	1.005	2.30%	Inner Mongolia	0.658	0.726	−1.00%	National	0.762	-	2.20%
Shaanxi	0.806	1.832	4.60%	Jiangxi	0.650	0.499	0.30%
Fujian	0.789	1.142	4.60%	Liaoning	0.633	0.550	0.90%

Note: The regions are sorted according to the total average value of agricultural ecological efficiency from high to low, and the provinces with the same average value are ranked first with the smaller distance. Considering the full distance ≥ comprehensive column spacing, to ensure the brevity of the article, only the full spacing is listed here to measure the stability of each decision-making unit under different windows.

are obtained.

The study found that from 1998 to 2018, China’s agricultural ecological efficiency value showed an overall upward trend. The period from 1998 to 2003 had a decline stage of agricultural ecological efficiency. The period from 2003 to 2009 had a fluctuating stage, and the period from 2009 to 2018 had a rising stage. Since 2011, the agricultural ecological efficiency value of many provinces has been greater than 1. By 2018, the agricultural ecological efficiency value of Beijing, Tianjin, Shanghai, and other 15 provinces and cities had been greater than 1, which is at an effective level.

From the perspective of expected agricultural output, the value of China’s agricultural eco-efficiency showed a downward trend from 1998 to 2003. During this period, China has experienced everything from preparing for “WTO accession” to formally joining the World Trade Organization (WTO). During this period, China’s agricultural market-oriented reforms with the background that agricultural internationalization has not yet been completed. The conditions for agricultural internationalization are not yet available, and China’s agricultural sector will inevitably be challenged [24]. On the one hand, some agricultural products that do not have price advantages, such as wheat, corn, and soybeans, have been significantly impacted by the international market. Among them, the main producing area of soybeans and corn—the Northeast region—has been severely affected. On the other hand, under the constraints of the WTO agricultural framework, China needs to reduce agricultural trade subsidies. Therefore, at this stage, the expected output of China’s agriculture has been significantly impacted. In 2004, China fully liberalized the grain purchase market and purchase price, and the grain price was formed by the market. Since between 2004 and 2006, the state implemented the minimum purchase price policy for two key grain varieties, paddy, and wheat, respectively, in the main producing areas. During this period, the value of agricultural ecological efficiency in China showed an upward trend in fluctuation. In 2008, the financial crisis swept the world. The food prices in the international market fluctuated greatly, the domestic agricultural product market was damaged, and the agricultural ecological efficiency declined. In 2009, affected by the lag of the financial crisis, it continued to decline. Subsequently, the overall development of China’s agricultural industry was stable, and the value of agricultural ecological efficiency gradually increased.

From the perspective of undesired agricultural output, in the 1990s, China’s agricultural and rural life pollution problems became prominent. The use of chemical fertilizers, pesticides, agricultural films, and other agricultural production materials increased rapidly. In 1998, as the turning point, the rural environment’s pollution exceeded the environmental capacity and began to show signs of obvious deterioration [3]. The research results show that China’s agricultural ecological efficiency has shown a clear downward trend since 1998. In 1999, the former State Environmental Protection Administration issued China’s first policy for environmental protection in rural areas—“Several Opinions of the State Environmental Protection Administration on Strengthening Rural Ecological Environmental Protection”. However, due to the high utilization rate of chemical fertilizers and pesticides in China. This amount has been used much higher than the world average. So, China’s agricultural ecological environment is still becoming increasingly tense at this stage. Since the beginning of the 21st century, the level of agricultural production in China has continuously improved. The state has continuously increased its efforts in environmental protection. In the national environmental protection “10th Five-Year Plan”, it is clear that the control of agricultural non-point source pollution, rural life pollution, and improvement of rural environmental quality are regarded as environmental protection.

3.1.2. Analysis of Agriculture Ecological Efficiency at the Regional Level

The below Table 3 calculate the average efficiency value of each region in China during from 1998 to 2018.

The study found that the agricultural ecological efficiency value in the eastern region was generally higher and higher than the national average level. The agricultural ecological efficiency value in the western region was comparable to the national average level, and the efficiency value in the central region was relatively low. The eastern region has the following obvious inherent advantages: a strong agricultural foundation, a developed economy, and a culture. So, have a good ecological environment, which provides a high-quality backup force for agricultural development. The degree of agricultural mechanization in the western region is low and the natural foundation is poor. But the overall development has been good through the agricultural development method of “wide planting and low harvesting”. In comparison, although the central region is rich in agricultural resources, the abuse of pesticides and fertilizers, the barbaric exploitation of land resources. The lack of inter-provincial coordination and integration has been constrained by the inefficiency of ecosystems [20].

3.1.3. Analysis of Agriculture Ecological Efficiency at the Provincial Level

Calculate the average value of agricultural efficiency in each province of China during the 21 years from 1998 to 2018, and obtain the statistical average value as shown in Table 4.

The above Table 4 explains that Shanghai’s agricultural ecological efficiency value ranks first in the country and reaches an effective level. Hainan, Jiangsu, Guizhou, Beijing, Zhejiang, Guangdong, Chongqing, and other places also show high agricultural ecological efficiency. Except for Hainan Province, these provinces have maintained a high average annual growth rate, achieved agricultural economic growth, and a coordinated development of environmental protection. Hainan, Inner Mongolia, and Jilin provinces showed negative growth rates. The early agricultural eco-efficiency value of Hainan Province was relatively high, although it has declined. In Inner Mongolia and Jilin provinces, due to the low input level of agricultural capital factors and extensive management methods. Moreover, this shows the production characteristics of “low input and low output,” showing low ecological efficiency and a negative growth trend. Hubei, Hunan, Henan, Jiangxi, and other central regions have been located in the middle and lower reaches of the Yangtze River plain. As the main grain-producing areas in China, they have a certain agricultural production base, and the agricultural ecological efficiency is at an intermediate level. In Shandong, Anhui, Yunnan, Gansu, Shanxi, and other provinces, due to extensive agricultural production, unreasonable allocation of resources, serious agricultural non-point source pollution, and other factors, resulting in low agricultural ecological efficiency.

3.2. Kernel Density Analysis

To further discuss the agglomeration of agricultural ecological efficiency in China’s provinces over time, 1998, 2002, 2007, 2012, and 2018 have been selected as representative years for analysis using the kernel density function. The Stata 16.0 statistical software has been used to generate agroecological data.

The efficiency kernel density map is shown in Figure 2. The study found that in 1998, the distribution of agricultural eco-efficiency values in various provinces had been relatively scattered, indicating that the agricultural eco-efficiency values of various provinces and cities have been quite different during this period. In 2002, the kernel density increased significantly, and the gap between the agricultural ecological efficiency values of the provinces narrowed. But the peak value shifted to the left, indicating that compared with 1998, the efficiency values of the provinces have been clustered at a lower level. After that, in 2007 and 2012, there was a “double peak” cluster. The peak began to shift to the right, and the gap between regions with the growth of agricultural ecological efficiency value gradually became more prominent. In 2018, the core density of agricultural ecological efficiency values in each province dropped sharply, showing a scattered distribution and extending to the left, indicating that with the change of time. China’s agricultural ecological efficiency values continued to increase, but the inter-provincial differences have also gradually expanded. In general, the nuclear density of China’s agricultural eco-efficiency values showed a changing trend of “left-shift up—double peaks—right-shift down”.

3.3. Spatial Spillover and Agriculture Fiscal Expenditure Impact on Eco-Efficiency of Agriculture

3.3.1. Spatial Autocorrelation Test

The spatial autocorrelation test results given in

Table 5. Spatial autocorrelation test—part 2.

Years	1999	2000	2001	2002	2003	2005	2006
Adjacency Matrix	0.148 *	0.184 *	0.223 **	0.202 **	0.218 **	0.209 **	0.161
Capital Matrix	0.020 *	0.038 **	0.032 **	0.029 **	0.023	0.037 **	0.060 ***
Economic Matrix	0.029 **	0.044 **	0.039 **	0.039 **	0.041 **	0.055 ***	0.076 ***
Years	2007	2008	2009	2010	2011	2015	2018
Adjacency Matrix	0.254 **	0.081	0.265 **	0.159	0.121	0.105	0.208 **
Capital Matrix	0.030 *	−0.011	0.032 **	0.014	0.02	0.008	0.002
Economic Matrix	0.058 ***	0.034 **	0.044 **	0.039**	0.041**	0.028*	0.004

Note: In order to ensure the simplicity of the article, only the years in which the p-value of the spatial autocorrelation test is less than 0.1 are listed in the table, and the remaining years not listed indicate that there is no significant spatial autocorrelation relationship in their agricultural ecological efficiency; (2) *** p < 0.01, ** p < 0.05, * p < 0.1.

.

The above results show that after including economic factors, each matrix shows a higher spatial autocorrelation relationship. So, it can be recognized that there has indeed been a spatial correlation between the values of agricultural ecological efficiency among provinces, which has been embodied in the adjacency correlation, provincial capital city correlation, and per capita GDP. In terms of geographical distance alone, since 2009, there has been no spatial correlation between provinces, which is consistent with the results of the kernel density test above, but economic distance has extended the correlation between provinces over time.

Referring to the previous research results, it has been found that most studies believe that there is a significant spatial autocorrelation relationship between the agricultural ecological efficiency values in China over the years, and the p-value of most years is less than 0.01, the spatial correlation coefficient is high, and the spatial autocorrelation relationship is strong. However, this study believes that the reason for this may be that the author used the method of calculating the cross-sectional data of each year separately when calculating the value of agricultural ecological efficiency. As a non-parametric model, the DEA model calculates the relative value within the sample based on the input and output data of the sample. In addition, in this study, the efficiency value of each year has been calculated in the form of cross-sectional data, and the test results are similar to those of previous studies. In addition, many scholars have found that China’s agricultural eco-efficiency values generally show an α divergence trend [8,9], indicating that in terms of time trends, the differences in China’s agricultural eco-efficiency values are expanding. Therefore, the study believes that Table 5 may more truly reflect the spatial correlation of China’s agricultural ecological efficiency values.

3.3.2. Analysis of Spatial Durbin Model

Using the statistical software Stata 16.0, based on the general spatial model constructed above (using the economic distance matrix), the Wald test was firstly carried out. The test results rejected the null hypothesis at the 1% level, indicating that it cannot be simplified to a spatial error model (SEM) or a spatial autonomic model. A regression model (SAR). The LR test results show that compared with the spatial autoregressive model, the spatial error model, and the spatial autocorrelation model (SAC), the log-likelihood value of the spatial Durbin model is higher, and the AIC and BIC values are lower. The LR test results also point to the Durbin model. Finally, the Hausman test has been used to determine whether the fixed effect or random effect has been used. The Hausman test obtains a chi-square statistic of 46.9, and the null hypothesis is rejected at the 1% level; that is, from a purely technical point of view, the fixed-effect model should be selected. However, the study finally chooses to take the regression results of the random-effects model as the criterion. The specific reasons include the following points:

First, there is currently no consensus in the academic community on the selection, use of random-effects models, and fixed-effects models. For example, when dealing with sequence data with a large N, random effects are more effective [25]. This view is also confirmed in the papers of [26,27]. Second, in the regression, this paper first used ordinary least squares (OLS), random effect model, and fixed-effect model, and then carried out the regression of the spatial Durbin random effect model and the spatial Durbin fixed-effect model. The comparison found that the random effect and the regression results of the model and the ordinary least squares method have a close match. The robustness test indicates that the regression results of the random effect model are more stable.

From a national perspective, the agricultural ecological efficiency of neighboring provinces has a positive spatial spillover effect on the agriculture ecological efficiency of the region, which has also been reflected in the eastern and western regions, while the provinces in the central region have positive spatial spillover effects from neighboring provinces but it is not obvious.

In

Table 6. Regression results of spatial Durbin model.

		(1)	(2)	(3)	(4)	(5)
		Two-Way Fixed	Random Effects	East	Central	West
W × AEE (rho)		−0.448 ***	0.518 ***	0.321 ***	0.0884	0.248 ***
Agricultural expenditure	Direct effect (Main)	−1.423 ***	−1.123 **	−0.913	3.167 ***	−4.284 ***
		(0.489)	(0.485)	(1.071)	(0.746)	(0.723)
	Indirect effects (W × Agricultural Support)	10.11 ***	7.350 ***	1.229	−3.427 ***	9.246 ***
		(3.378)	(2.320)	(3.659)	(1.083)	(1.693)
	total effect	8.688 **	6.227 ***	0.317	-	4.962 ***
		(3.506)	(2.342)	(4.146)	-	(1.553)
Fiscal Expenditures for Agricultural Environmental Governance	Direct effect (Main)	3.064 ***	2.368 **	2.950	0.662	0.613
		(0.921)	(0.934)	(1.931)	(1.665)	(1.724)
	Indirect effects (W × Environmental Support)	−10.25 **	−16.62 **	2.095	−6.711 **	−4.418
		(4.534)	(6.967)	(7.319)	(2.860)	(4.173)
	total effect	−7.181 *	−14.25 **	5.045	-	−3.805
		(4.454)	(6.906)	(8.100)	-	(3.675)
Agricultural Science and Technology Fiscal Expenditure	Direct effect (Main)	2.510 **	2.759 ***	7.911 ***	4.385 ***	23.27 ***
		(1.031)	(1.072)	(1.733)	(1.526)	(4.971)
	Indirect effects (W × Tech Support)	6.599	21.33 ***	18.99 **	8.229 ***	43.87 ***
		(4.920)	(6.176)	(7.407)	(2.759)	(15.38)
	total effect	9.110 *	24.09 ***	26.90 ***	-	67.14 ***
		(5.039)	(6.322)	(8.200)	-	(16.34)
Agricultural Education Support	Direct effect (Main)	1.370 ***	0.622	1.521 **	−1.597 **	0.613
		(0.435)	(0.435)	(0.759)	(0.775)	(1.724)
	Indirect effects (W × Agricultural Support)	0.473	−4.334 ***	−6.343 ***	2.190 **	−4.418
		(2.493)	(1.028)	(1.650)	(0.880)	(4.173)
	total effect	1.843	−3.712 ***	−4.822 ***	-	−3.805
		(2.457)	(1.000)	(1.711)	-	(3.675)
sigma2_e		0.0175 ***	0.0219 ***	0.0276 ***	0.00725 ***	0.0204 ***
R-squared		0.133	0.270	0.294	0.358	0.468

Note: (1) Due to the change of statistical caliber, to ensure the accuracy of the results, the sample used in this table is the data of 30 provinces and cities from 2007 to 2018 (excluding Tibet, Hong Kong, Macao, and Taiwan regions); (2) Models 1 and 2 in the table are based on the total sample was subjected to two-way fixed-effect regression and random-effects regression to obtain the results. Models 3, 4, and 5 were obtained by random-effects regression for the eastern (11 provinces and cities), central (8 provinces), and western (11 provinces and cities) respectively. result. (3) The spatial autocorrelation coefficient of Model 4 fails the test, so the spatial spillover effect is not decomposed, and only the Main coefficient and the Wx coefficient are reported. The spatial autocorrelation coefficients of the rest of the models have passed the test, and their spatial spillover effects need to be decomposed into “direct effects, indirect effects, and total effects”; (4) The data in parentheses are standard errors; (5) *** p <0.01, ** p < 0.05, * p < 0.1.

the agricultural fiscal expenditure has a significant negative impact on the agricultural ecological efficiency of the region at a level of 5%, which is consistent with the research results of many scholars. However, the surrounding areas have a significant positive spatial spillover to the province’s AEE at the 1% level.

Agriculture’s environmental fiscal expenditures have a significant positive impact on improving the agricultural ecological efficiency in the region, but environmental governance in adjacent areas has a significant negative impact on the region.

The financial expenditure on agricultural science and technology has a significant positive impact on the region at the level of 1%, and the financial support for science and technology in the surrounding areas also has a significant positive impact on the finances of the region at the level of 1%, and the impact is greater than that of the region, indicating that the support for science and technology has a significant positive impact. Good positive externalities. At the same time, the marginal impact of scientific and technological expenditures on agricultural ecological efficiency is the largest among the various financial expenditures in the study, indicating that the scientific and technological development of agriculture advocated by China’s agricultural development strategy has good returns and is still in the stage of rising income. It has a large room for growth. When comparing the proportion of the financial expenditure on science and technology in each region to the total financial expenditure of the region, it is found that the ratio of the eastern region is 3.41%, the ratio of the three regions is the highest, and the lowest is 1.11% of the western region.

The financial expenditure on agricultural education has no significant impact on the agricultural ecological efficiency of the province, and the education financial expenditure in the surrounding provinces has a significant negative impact on the agricultural ecological efficiency of the province.

4. Conclusions and Recommendation

The study has tried to examine agricultural ecological efficiency and its influences on agriculture’s fiscal expenditures. The study used data from 30 provinces of China from 1998 to 2018. In the first step, researchers used super efficiency DEA window analysis for calculating agriculture’s ecological efficiency. In the second step, researchers used the Kernel density function to analyze the temporal and spatial changes in agriculture’s ecological efficiency. Based on this analysis, a spatial Durbin model has been constructed to analyze the spatial impact of agricultural fiscal expenditures on agricultural ecological efficiency. Further, this analysis is divided into the following three categories: eastern, middle, and western areas. The following conclusions are drawn:

China’s agricultural ecological efficiency first showed a downward trend and then an upward trend. The efficiency value of the eastern region has been higher than the national average level, the western region has been equal to the national average level, and the central region has been lower than the national average level. The nuclear density showed a trend of “rising leftward—double peaks—falling rightward”, and the inter-provincial agriculture ecological efficiency value gap gradually widened during the rise.

On the whole, China’s agricultural ecological efficiency has a positive spatial spillover effect, but the agricultural fiscal expenditure has a significant negative impact on the agricultural ecological efficiency of the region. But the agricultural finance in the surrounding areas has a significant positive spatial spillover effect on the region. Environmental governance has the effect of improving the agricultural ecology of the region. Science and technology investment not only has a significant role in promoting the development of agricultural ecology in the region but also benefits the surrounding areas. However, the overall impact of education investment in China’s agricultural ecological environment is space-crowding out.

The study found that the agricultural ecological efficiency in the western region has most obviously been affected by the national fiscal policy, especially the western region achieved the catch-up effect by increasing investment in science and technology. Due to the favorable natural and economic conditions in the early stages, the eastern region has entered a stage of diminishing marginal returns. However, due to the lack of benign inter-provincial cooperation in the central region, the crowding-out effect of environmental governance space is obvious.

The above analysis has the following implications for the central and local governments to improve their financial support strategies to improve agricultural ecological efficiency: First, a green and ecologically oriented financial support system should be established.

In terms of agricultural financial support, rationally reduce chemical fertilizers, pesticides, and other support. In terms of financial support for science and technology, support agricultural technology innovation to promote agricultural environmental protection and governance technologies. In terms of financial support for environmental governance, they increased the special funds for rural environmental governance. When controlling agricultural pollution, it is necessary to strengthen communication between regions, cooperate to control agricultural non-point source pollution, and jointly improve agricultural ecological efficiency. In terms of financial support for education, accelerate the process of agricultural modernization and prevent farmers from increasing the use of chemical fertilizers and pesticides. Finally, the state should actively guide inter-provincial exchanges and cooperation in the development of agricultural industries, and jointly promote China’s green-led rural revitalization.