Public Investment, Environmental Regulation, and the Sustainable Development of Agriculture in China Based on the Decomposition of Green Total Factor Productivity

Abstract

This study aims to accurately assess the growth of agricultural total factor productivity and its driving components under the constraints of resources and environment, and provides reliable information for agricultural policy formulation and agricultural development practices. According to the input and output panel data of provincial agricultural planting in China and employing the Global Malmquist–Luenberger (GML) index method and the Bootstrap method, this paper measures China’s agricultural green total factor productivity (GTFP), technical efficiency change (EC), and technical best-practice gap change (BPC). In addition, the Tobit model is applied to analyze the impact of public investment and environmental regulation variables on China’s agricultural GTFP and its components. The results show that (1) China’s agricultural GTFP has steadily improved, and technological promotion is the main contributor; (2) agricultural GTFP and its components present significant spatial differences, which are overall manifested as agricultural priority development zone > agricultural moderate development zone > agricultural protection development zone; and (3) financing support of technical innovation and the intensity of environmental regulation have a significant positive impact on agricultural GTFP and its components. The combination of positive technical innovation support and appropriate environmental regulation helps to improve agricultural GTFP and achieve the sustainable development of China’s agriculture.

1. Introduction

Since the reform and opening up, China’s agricultural production capacity has continued to grow (the average annual real growth of agricultural output value is 10.83%). The level of food security has continued to increase (the average annual real growth of total grain output is 1.94%), and the level of farmers’ income has increased steadily (the average annual real growth of per capita rural income is 11.54%), and this provides sufficient production materials and consumption goods for the development of the national economy. The increases in traditional agricultural factors and total factor productivity (hereinafter referred to as TFP) are the main sources of economic growth. The scarcity of resources determines that the sustainable development of China’s agriculture cannot rely on the unlimited expansion of agricultural elements [1,2], and the economic growth of traditional extensive agriculture inevitably induces a series of endogenous agricultural pollutions. Overexploitation of agricultural resources, excessive use of agricultural inputs such as chemical fertilizers and pesticides, direct burning of crop straws, random dumping of livestock and poultry manure, and residues of agricultural plastic mulch film have, for many years, led to increasingly prominent agricultural non-point source pollution and serious threats to the sustainable development of agriculture [3]. With the continuous growth in population, food security challenges, and the continuous growth in consumer demand for high-quality agricultural products, the development model of “three highs” agriculture (high input, high output, and high waste) that relies on traditional inputs has made the effective supply of agricultural products apparently insufficient, and the momentum of agricultural development is declining [4]. “Implementation Opinions on Fighting the Tough Battle for the Prevention and Control of Agricultural Non-point Source Pollution (2015)” and “National Agricultural Sustainable Development Plan (2015–2030)” propose to accelerate the sustainable use of agricultural resources, agricultural environment protection, and agricultural non-point source pollution governance in order to achieve the practical needs of sustainable agricultural development. In recent years, Central Document No.1 has clearly proposed to promote agricultural supply-side structural reforms; improve agricultural innovation, competitiveness, and TFP; and promote the conversion of agricultural growth momentum and agricultural innovation-driven development. Therefore, the fundamental path of sustainable agricultural development lies in identifying the growth source of TFP and increasing the economic contribution of TFP.

Under the binding of resources and environment, the measurement and factor decomposition and influence factors analysis of China’s agricultural green total factor productivity (hereinafter referred to as “GTFP”) have become hot issues in current agricultural economic research [2,5,6,7,8,9,10,11,12]. Due to differences in research methods, input–output variable selection, and research period, the current GTFP measurement results show significant differences (as shown in

Table A1. Relevant research on China’s agricultural GTFP in recent years.

Literature Source	Research Time	Research Method	Desired Output Variable	Undesired Output Variable	GTFP
Du et al., 2016 [5]	1991–2013	DEA-GML	Total output value of planting industry	COD, TN, TP, Pesticide residue, Agricultural mulch film residue	0.56%
Ye and Hui, 2016 [3]	1995–2013	SBM-GML	Total output value of planting industry	COD, TN, TP, Pesticide residue, Agricultural mulch film residue	1.29%
Pan and Ying, 2013 [2]	1998–2009	DDF, ML index	Total output value of agriculture, forestry, animal husbandry, and fishery	Total agricultural non-point source pollution discharge	2.9%
Liang and Long, 2015 [6]	2000–2013	Based on optional technology DEA, ML index	Total output value of agriculture, forestry, animal husbandry, and fishery	COD, TN, TP	1.63%
Yue and Wang, 2013 [30]	2001–2010	Distance function, ML index	Total output value of planting industry	COD, TN, TP, Total agricultural carbon emissions	1.90%
Liu, 2014 [31]	2001–2010	DDF, ML index	Total output value of agriculture, forestry, animal husbandry, and fishery	COD, TN, TP, Total agricultural carbon emissions	7.6%
Ge et al., 2018 [28]	2001–2015	SBM, ML index	Added value of agricultural planting	Carbon emissions from plantation	1.56%
Lv and Zhu, 2019 [8]	2011–2015	Super-SBM model, M index	Total output value of agriculture, forestry, animal husbandry, and fishery	COD, TN, TP	9.5%
Meng et al., 2019 [9]	1997–2016	Super-SBM model, ML index	Total output value of agriculture and animal husbandry	Surplus intensity of agricultural land nitrogen	7.02%

of Appendix A), but general conclusions can also be drawn: China’s agricultural TFP is significantly overestimated due to the ignoring of resource and environmental factors. China’s agricultural GTFP shows a growing trend and greater volatility and regional differences. The driving force is mainly from green technology progress, and the improvement of green technology efficiency is relatively lagging, which presents as a dynamic coexistence of technological progress and deterioration of technological efficiency.

Accurately assessing the growth of TFP and its driving factors is essential for agricultural policy formulation and policy effectiveness evaluation [10]. Traditional TFP measurement often only considers the input constraints of production factors such as land, labor, and capital, and ignores resource and environmental consumption and environmental pollution, which makes it difficult to accurately evaluate economic performance and social welfare changes, and even leads to policy misdirection [13,14]. Therefore, recalculating China’s agricultural economic growth performance with TFP under environmental constraints, namely GTFP, can indicate the true level and influencing factors of agricultural economic growth, and provide scientific policy basis for promoting green, ecological, and sustainable development of the agricultural economy [8]. At present, China is in a critical period of evolving from traditional agriculture to green agriculture, ecological agriculture, and sustainable agriculture. Active agricultural environmental control, agricultural public investment, and agricultural technology progress have improved agricultural GTFP to a certain extent, and the coordination between agricultural economic growth and agricultural environmental quality has been continuously improved [8].

Multidimensional assessment of the evolution pattern of China’s agricultural GTFP and improvement of agricultural production efficiency under the constraints of resources and environmental carrying capacity are key to promoting sustainable agricultural development in China [15]. This study, based on China’s provincial agricultural planting input–output panel data, considers agricultural chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), pesticide loss, agricultural plastic mulch film residue, and other agricultural “undesired output” from agricultural production factors such as fertilizer, pesticide, agricultural plastic mulch film, straw, etc. The global Malmquist–Luenberger (hereinafter referred to as GML) index method [16] and the Bootstrap method [17] are used to measure the GTFP of China’s agricultural growth, technical efficiency change (hereinafter referred to as EC), and technical best-practice gap change (hereinafter referred to as BPC), and the Tobit model is used to analyze the impact of public investment and environmental regulation factors on China’s agricultural GTFP and its components.

The objectives of this paper are as follows: (1) Scientifically calculate the agricultural GTFP, and define the sustainable development capacity of each agricultural development zone (priority development zone, moderate development zone, protected development zone). (2) Based on the decomposition of agricultural GTFP, clarify the contribution of EC and BPC to agricultural GTFP, and discover the main driving force for the growth of agricultural GTFP in China. (3) Focus on analyzing the impact mechanism of public investment and environmental regulation on agricultural GTFP, and provide realistic basis for the formulation of sustainable agricultural development promotion policies for different regions.

2. Theoretical Framework

Sustainable agricultural development is a dynamic balance between economic sustainability, social sustainability, and environmental sustainability [18]. It is an agricultural development strategy that harmonizes population, economy, society, resources, and environment [8]. Based on the trend of agricultural development in the world and the status of agricultural development in China, the government put forward the Strategy of Sustainable Agricultural Development of China, and defined the strategic attributes of sustainable agricultural development, such as “not causing environmental degradation”, “appropriate application of technology”, “economically feasible”, and “socially acceptable”. The “National Sustainable Agriculture Development Plan (2015–2030)” marks that the sustainable development of agriculture has entered a new stage focusing on agricultural non-point source pollution control and agricultural green development. The calculation of agricultural GTFP under resource and environmental constraints is an important way to examine the coordinated relationship between agricultural development and resources and the environment, and to reflect the sustainable level of agricultural development [1,5,18,19,20].

DEA is a nonparametric statistical evaluation method that solves the efficiency measurement problem of multiple inputs and multiple outputs, but the Malmquist index of traditional DEA cannot measure TFP that contains undesirable output. Y.H. et al. (1997) [21] revised the Malmquist productivity index, introduced directional distance functions (hereinafter referred to as DDFs), and proposed the Malmquist–Luenberger productivity (hereinafter referred to as ML) index as the basic framework of GTFP research [21]. In order to solve the problem of measuring inefficiencies in the radial DEA model, only the input–output proportional changes are included, a non-radial and non-angular slack-based measure (SBM) model is proposed by introducing slack variables of input–output, and more importantly, undesired output is considered (Tone, 2002) [22]. Fukuyama & Weber (2009) [23] and Färe & Grosskopf (2010) [24] also proposed the combination of the directional distance functions and slack-based measures of efficiency to make the measured GTFP more accurate and comprehensive [23,24]. In order to solve the defects that the ML index method does not have transitivity and the linear programming has no solution, the ML index is combined with DDFs to construct the GML index [16,25]. At present, the GML index is an important tool for GTFP measurement with multiple inputs and multiple outputs, and considers environmental pollution and other undesired outputs. It provides an important method for accurately measuring agricultural GTFP and decomposition factors, and truly reflecting the regional differences and temporal characteristics of agricultural GTFP.

The key link of agricultural GTFP accounting is to rationally integrate resources and environmental factors into the TFP analysis framework. The current common approach is to introduce agricultural environmental pollution as an undesired output (such as agricultural non-point source pollution) and a desirable output (such as total agricultural output value) into the analysis framework together. Pan and Ying (2013) applied the unit survey method and inventory analysis method [2] to investigate and evaluate the pollutant emissions of TN, TP, and COD in the four pollution units of fertilizer application, livestock and poultry breeding, aquaculture, and agricultural solid waste [26,27], and incorporate agricultural non-point source pollution into an agricultural TFP analysis framework to measure China’s agricultural TFP under resource and environmental constraints. The current undesired output mainly includes five pollution units: chemical fertilizer, livestock and poultry breeding, aquaculture, agricultural solid waste, and rural life. The TN, TP, COD, and agricultural carbon emissions are taken as the total undesired output [6,7,8,28,29,30], and pesticide pollution and agricultural plastic mulch film residue pollution are also taken into account [3,5]. This study intends to use the unit survey and evaluation method to incorporate agricultural planting production pollutants such as chemical fertilizers, pesticides, straw, and agricultural plastic mulch film, and to measure TN, TP, COD, pesticide loss, agricultural plastic mulch film residue, and other undesired outputs of agricultural planting, so as to lay a foundation for accurate accounting of agricultural GTFP.

Since the reform and opening up, China’s agricultural GTFP has shown an increasing trend, and has shown greater volatility and regional differences. The level of rural economic development, agricultural industrial structure, and infrastructure investment have a significant impact on agricultural GTFP [31,32]. Agricultural tax reduction and exemption promote the increase in agricultural GTFP, and industrialization, urbanization, and agricultural trade hinder the growth of GTFP [6]. At the same time, the rate of agricultural disaster, agricultural financial support, income distribution, farmers’ income composition, urbanization rate, and other factors have different influences on agricultural GTFP, environmental technology efficiency, and technological change [5]. The growth of China’s agricultural economy has benefited from the comprehensive effect of factor accumulation and technological progress. Promoting technological innovation, increasing public investment in agriculture, and strengthening environmental regulations are of great significance for improving agricultural TFP [7,33].

Agricultural financial support, such as technology research and development, institutional arrangement, infrastructure construction, policy reform, and rural education promote sustained growth in agricultural productivity [34,35,36]. R&D investment promotes the improvement of agricultural productivity to a great extent [11,37], and local research and development (hereinafter referred to as R&D) investment is the leading factor in promoting the growth in agricultural TFP [38]. According to analysis, public investment has a significant role in promoting agricultural technical efficiency and agricultural growth. The differences in agricultural technical efficiency and agricultural growth in China’s provinces are mostly due to the spatial difference between the total agricultural financial support and the level of technical innovation support [39]. The Chinese government attaches great importance to R&D investment and financial support for agriculture in order to promote the transformation of agricultural technology, improve the efficiency of agricultural science and technology innovation, improve agricultural and rural infrastructure, and thereby increase agricultural TFP [33,40,41]. Therefore, public investment, such as agricultural financial support and technical innovation support, is the main driving force to regulate the relationship between economy, environment, and technology, and is an important factor for breaking through the “economy-environment” pressure of agricultural development, improving agricultural GTFP, and enhancing the ability of agricultural sustainable development [34].

To a certain extent, environmental regulations have led to major changes in agricultural production technology and production inputs [42], and have a deterrent effect on serious environmental pollution. They also urge the agricultural research institutions to increase technological innovation, improve agricultural production efficiency, achieve agricultural technology progress, and ultimately achieve a win–win scenario of ecological and economic benefits [43]. From a static perspective, environmental regulation is not conducive to agricultural technology progress, but from the perspective of long-term benefits, environmental regulation is conducive to agricultural technology progress, which demonstrates the “porter’s hypothesis” [44,45,46]. Environmental regulations have a significant positive impact on green technological innovation, and agricultural green technological innovation under environmental regulations is the driving force for improving agricultural GTFP [47]. The advancement of agricultural technology and the transformation of farmers’ production behavior induced by environmental regulations are the fundamental guarantees for enhancing agricultural GTFP [46,48]). The complementary coupling of environmental regulations and public investment is more conducive to the continuous growth in agricultural GTFP [49]. Therefore, analyzing the impact of public investment and environmental regulations on the growth of agricultural GTFP will provide a strong basis for the government to formulate different sustainable agricultural development policies. Therefore, the following hypotheses are proposed:

Hypothesis 1 (H1).

Public investment helps to promote the sustainable development of agriculture.

Hypothesis 2 (H2).

Environmental regulation will promote agricultural GTFP and enhance agricultural sustainable development ability.

3. Methods and Data

3.1. SBM-GML Model Considering Undesired Output

Oh (2010) [16] put forward the global index (GML), which takes all the inspection periods of each decision-making unit as the benchmark to construct the production frontier, which can not only satisfy the cyclicality but also avoid the situation of no solution, and allow the existence of technical retrogression when calculating the index [16,25]. The current benchmark constructs the reference set of production possibility set in period t, which is defined as follows:

$${\mathrm{P}}_{\mathrm{C}}^{\mathrm{T}}\left({\mathrm{x}}^{\mathrm{t}}\right)=\left\{\left({\mathrm{y}}^{\mathrm{t}},{\mathrm{b}}^{\mathrm{t}}\right)|{\mathrm{x}}^{\mathrm{t}} \mathrm{can} \mathrm{product} \left({\mathrm{y}}^{\mathrm{t}},{\mathrm{b}}^{\mathrm{t}}\right)\right\}$$

(1)

Different from the current benchmark, the global benchmark is defined as

$${\mathrm{P}}_{\mathrm{G}}={\mathrm{P}}_{\mathrm{C}}^1\cup {\mathrm{P}}_{\mathrm{C}}^2\cup \cdots \cup {\mathrm{P}}_{\mathrm{C}}^3$$

(2)

The subscript C represents the current benchmark, and the subscript G represents the global benchmark. All current benchmarks are enveloped to form a single reference set of global production possibility set, and other periods can be compared with it.

This study uses provinces as decision-making units (DMUs) to construct the frontier of agricultural production, introduces the period t and compares the actual production points of each DMU with the frontier mapping points to measure agricultural technical efficiency and technological progress. Fukuyama et al. (2009) [23] overcame the ignoring of nonzero relaxation terms, and proposed the combination of directional distance function (DDF) and slack-based model (SBM) to propose the concept of directional SBM. The global directional SBM function is defined as follows:

$${\mathrm{S}}_{\mathrm{G}}^{\mathrm{T}}\left({\mathrm{x}}_{\mathrm{k}}^{\mathrm{t}},{\mathrm{y}}_{\mathrm{k}}^{\mathrm{t}},{\mathrm{b}}_{\mathrm{k}}^{\mathrm{t}};{\mathrm{g}}^{\mathrm{x}},{\mathrm{g}}^{\mathrm{y}},{\mathrm{g}}^{\mathrm{b}}\right)=\begin{array}{c}\max \\ {\mathrm{s}}^{\mathrm{x}},{\mathrm{s}}^{\mathrm{y}},{\mathrm{s}}^{\mathrm{b}}\end{array}\left\{\frac{1}{\mathrm{n}}{\displaystyle \sum }_{\mathrm{n}=1}^{\mathrm{N}}\frac{{\mathrm{s}}_{\mathrm{n}}^{\mathrm{x}}}{\begin{array}{c}\mathrm{x}\\ {\mathrm{g}}_{\mathrm{n}}\end{array}}+\frac{1}{\mathrm{M}+1}\left({\displaystyle \sum }_{\mathrm{m}=1}^{\mathrm{M}}\frac{{\mathrm{s}}_{\mathrm{m}}^{\mathrm{y}}}{\begin{array}{c}\mathrm{y}\\ {\mathrm{g}}_{\mathrm{m}}\end{array}}+{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{J}}\frac{{\mathrm{s}}_{\mathrm{j}}^{\mathrm{b}}}{\begin{array}{c}\mathrm{b}\\ {\mathrm{g}}_{\mathrm{j}}\end{array}}\right)\right\}/2$$

(3)

$$\mathrm{s}.\mathrm{t}.\left\{\begin{array}{c}{\displaystyle {\displaystyle \sum }_{\mathrm{t}=1}^{\mathrm{T}}}{\displaystyle {\displaystyle \sum }_{\mathrm{k}=1}^{\mathrm{K}}}{\mathrm{z}}_{\mathrm{k}}^{\mathrm{t}}{\mathrm{x}}_{\mathrm{kn}}^{\mathrm{t}}+{\mathrm{s}}_{\mathrm{n}}^{\mathrm{x}}={\mathrm{x}}_{\mathrm{n}0}^{\mathrm{t}},\mathrm{n}=1,2,\cdots ,\mathrm{N}\\ {\displaystyle {\displaystyle \sum }_{\mathrm{t}=1}^{\mathrm{T}}}{\displaystyle {\displaystyle \sum }_{\mathrm{k}=1}^{\mathrm{K}}}{\mathrm{z}}_{\mathrm{k}}^{\mathrm{t}}{\mathrm{y}}_{\mathrm{km}}^{\mathrm{t}}+{\mathrm{s}}_{\mathrm{m}}^{\mathrm{y}}={\mathrm{x}}_{\mathrm{m}0}^{\mathrm{t}},\mathrm{m}=1,2,\cdots ,\mathrm{M}\\ {\displaystyle {\displaystyle \sum }_{\mathrm{t}=1}^{\mathrm{T}}}{\displaystyle {\displaystyle \sum }_{\mathrm{k}=1}^{\mathrm{K}}}{\mathrm{z}}_{\mathrm{k}}^{\mathrm{t}}{\mathrm{y}}_{\mathrm{kj}}^{\mathrm{t}}+{\mathrm{s}}_{\mathrm{j}}^{\mathrm{b}}={\mathrm{x}}_{\mathrm{j}0}^{\mathrm{t}}, \mathrm{j}=1,2,\cdots ,\mathrm{J}\\ {\displaystyle {\displaystyle \sum }_{\mathrm{k}=1}^{\mathrm{K}}}{\mathrm{z}}_{\mathrm{k}}^{\mathrm{t}}=1,{ \mathrm{z}}_{\mathrm{k}}^{\mathrm{t}} \ge 0, \mathrm{k}=1,2,\cdots ,\mathrm{K}\\ {\mathrm{s}}_{\mathrm{n}}^{\mathrm{x}}\ge 0,{ \mathrm{s}}_{\mathrm{m}}^{\mathrm{y}}\ge 0,{ \mathrm{s}}_{\mathrm{j}}^{\mathrm{b}}\ge 0\end{array}\right.$$

(4)

where $({\mathrm{x}}_{\mathrm{k}}^{\mathrm{t}},{\mathrm{y}}_{\mathrm{k}}^{\mathrm{t}},{\mathrm{b}}_{\mathrm{k}}^{\mathrm{t}})$ represents the input elements, desired output, and undesired output variables in the k region in period t, $({\mathrm{g}}^{\mathrm{x}},{\mathrm{g}}^{\mathrm{y}},{\mathrm{g}}^{\mathrm{b}})$ and $\left({\mathrm{s}}_{\mathrm{n}}^{\mathrm{x}},{ \mathrm{s}}_{\mathrm{m}}^{\mathrm{y}},{\mathrm{s}}_{\mathrm{j}}^{\mathrm{b}}\right)$ represent the direction vector and the slack vector, respectively, and $\left({\mathrm{s}}_{\mathrm{n}}^{\mathrm{x}},{ \mathrm{s}}_{\mathrm{m}}^{\mathrm{y}},{\mathrm{s}}_{\mathrm{j}}^{\mathrm{b}}\right)$ represents the excessive input, insufficient desired output, and undesired output redundancy, respectively. ${\displaystyle \sum }_{\mathrm{k}=1}^{\mathrm{K}}{\mathrm{z}}_{\mathrm{k}}^{\mathrm{t}}=1{ \mathrm{and} \mathrm{z}}_{\mathrm{k}}^{\mathrm{t}} \ge 0$, represents variable returns to scale (VRS).

Based on the SBM model, this paper measures the output-oriented GML productivity index under variable returns to scale.

$${\mathrm{GML}}^{\mathrm{t},\mathrm{t}+1}\left({\mathrm{x}}^{\mathrm{t}},{\mathrm{y}}^{\mathrm{t}},{\mathrm{b}}^{\mathrm{t}},{\mathrm{x}}^{\mathrm{t}+1},{\mathrm{y}}^{\mathrm{t}+1},{\mathrm{b}}^{\mathrm{t}+1}\right)=\frac{1+{\mathrm{S}}_{\mathrm{G}}^{\mathrm{T}} \left({\mathrm{x}}^{\mathrm{t}},{\mathrm{y}}^{\mathrm{t}},{\mathrm{b}}^{\mathrm{t}}\right)}{1+{\mathrm{S}}_{\mathrm{G}}^{\mathrm{T}} \left({\mathrm{x}}^{\mathrm{t}+1},{\mathrm{y}}^{\mathrm{t}+1},{\mathrm{b}}^{\mathrm{t}+1}\right)}$$

(5)

In Equation (5), ${\mathrm{S}}_{\mathrm{G}}^{\mathrm{T}}\left(\mathrm{x},\mathrm{y},\mathrm{b}\right)=\max \left\{\mathsf{\beta }| \left(\mathrm{y}+\mathsf{\beta }\mathrm{y},\mathrm{b}-\mathsf{\beta }\mathrm{b}\right)\in {\mathrm{P}}_{\mathrm{G}} \left(\mathrm{x}\right)\right\}$, given by the reference set P_G of the global production possibility set. If more desired output and less undesired output are produced, then GML_t,t+1 > 1, indicating that productivity is increased. If less desired output and more undesired output are produced, then GM_Lt,t+1 < 1, which means that productivity is reduced. The GML index is further decomposed as follows:

$$\begin{array}{l}{\mathrm{G}\mathrm{M}\mathrm{L}}^{\mathrm{t},\mathrm{t}+1}({\mathrm{x}}^{\mathrm{t}},{\mathrm{y}}^{\mathrm{t}},{\mathrm{b}}^{\mathrm{t}},{\mathrm{x}}^{\mathrm{t}+1},{\mathrm{y}}^{\mathrm{t}+1},{ \mathrm{b}}^{\mathrm{t}+1})=\frac{1+{\mathrm{S}}_{\mathrm{G}}^{\mathrm{T}}({\mathrm{x}}^{\mathrm{t}},{\mathrm{y}}^{\mathrm{t}},{\mathrm{b}}^{\mathrm{t}})}{1+{\mathrm{S}}_{\mathrm{G}}^{\mathrm{T}}({\mathrm{x}}^{\mathrm{t}+1},{\mathrm{y}}^{\mathrm{t}+1},{ \mathrm{b}}^{\mathrm{t}+1})}\\ =\frac{1+{\mathrm{S}}_{\mathrm{C}}^{\mathrm{T}}({\mathrm{x}}^{\mathrm{t}},{\mathrm{y}}^{\mathrm{t}},{\mathrm{b}}^{\mathrm{t}})}{1+{\mathrm{S}}_{\mathrm{C}}^{\mathrm{T}}({\mathrm{x}}^{\mathrm{t}+1},{\mathrm{y}}^{\mathrm{t}+1},{ \mathrm{b}}^{\mathrm{t}+1})}\\ \times \left[\frac{(1+{\mathrm{S}}_{\mathrm{G}}^{\mathrm{T}}({\mathrm{x}}^{\mathrm{t}},{\mathrm{y}}^{\mathrm{t}},{\mathrm{b}}^{\mathrm{t}}))/(1+{\mathrm{S}}_{\mathrm{C}}^{\mathrm{t}}({\mathrm{x}}^{\mathrm{t}},{\mathrm{y}}^{\mathrm{t}},{\mathrm{b}}^{\mathrm{t}}))}{(1+{\mathrm{S}}_{\mathrm{G}}^{\mathrm{T}}({\mathrm{x}}^{\mathrm{t}+1},{\mathrm{y}}^{\mathrm{t}+1},{ \mathrm{b}}^{\mathrm{t}+1}))/(1+{\mathrm{S}}_{\mathrm{C}}^{\mathrm{t}}({\mathrm{x}}^{\mathrm{t}+1},{\mathrm{y}}^{\mathrm{t}+1},{ \mathrm{b}}^{\mathrm{t}+1}))}\right]\\ =\frac{\mathrm{T}{\mathrm{E}}^{\mathrm{t}+1}}{{\mathrm{T}\mathrm{E}}^{\mathrm{t}}}\times \frac{{\mathrm{B}\mathrm{P}\mathrm{G}}_{\mathrm{t}+1}^{\mathrm{t}+\mathrm{t}}}{{\mathrm{B}\mathrm{P}\mathrm{G}}_{\mathrm{t}}^{\mathrm{t}+1}}={\mathrm{E}\mathrm{C}}^{\mathrm{t},\mathrm{t}+1}\times {\mathrm{B}\mathrm{P}\mathrm{C}}^{\mathrm{t},\mathrm{t}+1}\end{array}$$

(6)

TE and EC represent technical efficiency and efficiency changes, respectively. BPG_t,t+1 is the “Best practice gap” (BPG) between the current and the global technological frontiers. BPC_t,t+1 is a measure of the change (technical change) in the “best practice gap” between the two periods. BPC_t,t+1 > 1 indicates technological progress, and BPC_t,t+1 < 1 indicates technological retrogression.

Since the DEA model measures the limited predicted values, the estimated results will be disturbed by random factors and are easily affected by extreme values, so Bootstrap is chosen to solve this problem [50]. In this study, Bootstrap method was used to calculate the GML index, EC index, and BPC index.

3.2. Random Effects Tobit Model

Agricultural GTFP, EC, and BPC index have the characteristics of non-negative truncation. Using OLS method to estimate such restricted explained variables may yield biased results. This study intends to use the random effects Tobit model to analyze the influencing factors of GTFP, EC, and BPC to analyze the impact mechanism of public investment and policy regulations on China’s agricultural GTFP and its components.

$${\mathrm{Y}}_{\mathrm{k},\mathrm{it}}={\mathsf{\alpha }}_{\mathrm{it}}+{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{M}}{\mathsf{\phi }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{j},\mathrm{it}}+{\mathrm{u}}_{\mathrm{i}}+{\mathrm{e}}_{\mathrm{it}}$$

(7)

In Equation (7), Y is the explained variable. When k = 1, 2, 3, it represents agricultural GTFP, technical efficiency change (EC), and technical change (BPC), respectively. α_it represents the intercept term; φ_j represents the regression coefficient; x_j represents the explanatory variable; u_i represents the standard deviation of the individual effect (individual error); e_it represents the standard deviation of random interference items (random error); i = 1, 2, …, 31, represents 31 decision-making units (province, district, city); t represents the year; and M represents the total amount of explanatory variables—in other words, there are a total of M explanatory variables.

3.3. Variables and Descriptive Statistics

(1) Output variables of SBM-GML model

In order to increase the pertinence and specificity of the research, the “agriculture” in this study specifically refers to the plantation industry as defined by the Chinese National Bureau of Statistics. Agricultural output variables include desired output variables and undesired output variables. The desired output variable (Out_valu) is expressed in terms of total agricultural output value (constant prices in 1978).

Undesired output variables are represented by non-point source pollution such as chemical oxygen demand (Out_COD), total nitrogen (Out_TN), total phosphorus (Out_TP), pesticide loss (Out_pest), and agricultural mulch film residues (Out_film) generated by agricultural production. Applied chemical fertilizers (nitrogen fertilizer, phosphate fertilizer, compound fertilizer) enter the water body through surface runoff, farmland drainage, underground leaching, etc., to cause TN and TP pollution. Waste such as straw produced by planting crops (rice, wheat, corn, beans, potatoes, oil, and vegetables) produces COD, TN, and TP pollution if not handled properly. Applied pesticides enter the water body through underground leaching and surface runoff to cause pesticide loss pollution. After the agricultural mulch film is used, if the plastic mulch film recovery or harmless treatment is not thorough, residual mulch film is produced, which pollutes the ecological environment of the farmland. This study applies the unit survey and evaluation method, and calculates the parameters based on the agricultural non-point source unit [3,5,26,27], and according to the “Manual of Pesticide Loss Coefficient”, “Manual of Agricultural Mulch Film Residue Coefficient”, and “Manual of Fertilizer Loss Coefficient of Agricultural Pollution Sources” determined by the first national pollution survey, the total undesired output is calculated (see

Table 1. Variables in SBM-GML and their descriptions.

Variable	Observations	Mean	Std. Ev.
In_fert	Amount of fertilizer application (Ten thousand tons)	169.38	138.68
In_pest	Amount of pesticide application (Tons)	51,309.24	42,890.93
In_irri	The effective irrigation area (Thousands of hectares)	1931.382	1511.585
In_film	Amount of agricultural film application (Tons)	36,822.83	36,599.05
In_plan	Total planting area of crops (Thousands of hectares)	5145.351	3641.341
In_mach	Total power of agri-machinery (Ten thousand Kilowatts)	1435.969	1465.262
In_labo	Number of agricultural workers (Ten thousand)	513.2584	398.8217
Out_COD	COD loss (Ten thousand tons)	39.78574	32.79137
Out_TN	Total nitrogen loss (Ten thousand tons)	15.71015	12.61073
Out_TP	Total phosphorus loss (Ten thousand tons)	2.64157	2.627287
Out_film	Amount of residue of agricultural film (Ten thousand tons)	0.889549	1.135877
Out_pest	Pesticide loss (tons)	2058.776	2250.765
Out_valu	Total value of agricultural output (100 million Yuan RMB)	1052.414	1064.602

Note: In_fert represents fertilizer input; In_pest represents pesticide input; In_irri represents irrigation input; In_film represents agricultural film input; In_plan represents land investment; In_mach represents mechanical inputs; In_labo represents labor input; Out_COD represents chemical oxygen demand; Out_TN represents total nitrogen; Out_TP represents total phosphorus; Out_film represents mulch film residues; Out_pest represents pesticide loss; Out_valu represents total value of agricultural output.

).

(2) Input variables for the SBM-GML model

According to the current situation of agricultural planting production factor input, this study selects land, labor, machinery, irrigation, fertilizers, pesticides, and agricultural mulch film as agricultural input variables. The farmland input (In_plan) variable is expressed by the total sown area of crops. The labor input (In_labo) variable is expressed by the number of laborers in the plantation industry. The mechanical input (In_mach) variable is expressed by planting machinery power input [3,5]. The irrigation input (In_irri) variable is calculated with the actual effective irrigation area per year. The variable of fertilizer input (In_fert) is expressed as the consumption of chemical fertilizers (computation based on purity deduction) actually used in agricultural production each year. The pesticide input (In_pest) variable is expressed in terms of annual pesticide usage. The agricultural mulch film input (In_film) variable is expressed by the total amount of agricultural plastic mulch film used each year.

(3) Variables of Tobit Model with Random Effects

Combining current research results and selected research perspectives (see

Table 2. Variables in RE Tobit and their descriptions.

Variable	Observations	Description	Mean	Std. Ev.
PI_Agricu	Agricultural financial support	The proportion of agricultural fiscal expenditure in total fiscal expenditure	0.1010	0.0352
PI_R&D	Technological innovation input	R&D investment as a proportion of GDP	0.0132	0.0110
ER_Policy	Agricultural non-point source pollution remediation policy	0 = before 2014, 1 = 2015 and beyond	0.2222	0.4161
ER_Energy	Environmental regulatory intensity	Energy consumption per 10,000-yuan GDP (tons of standard coal)	1.7184	3.0455
Industry	Industrialization level	The added value of the secondary industry as a proportion of GDP	0.4540	0.0835
Urbanization	Urbanization rate	The proportion of urban population in total population	0.5031	0.1526
Agristructure	Agricultural planting structure	The proportion of grain sown area to the total sown area of crops	0.6555	0.1244
Distribution	Income distribution	Ratio of rural residents’ per capita net income to urban residents’ per capita disposable income	0.3565	0.6658
Environment	Natural environment conditions	The proportion of the affected area to the total sown area of crops	0.2305	0.1543
Salary	Farmers’ income structure	Farmers’ wage income accounts for the total household disposable income	0.3786	0.1430

), this research focuses on the impact mechanism of public investment and environmental regulations on China’s agricultural GTFP and its components. It reflects public investment factors through agricultural financial support variables (PI_Agricu) and technological innovation input support variables (PI_R&D). It reflects environmental regulatory elements through agricultural non-point source pollution remediation policy variables (ER_Policy) and environmental regulatory intensity variables (ER_Energy). In addition, it comprehensively considers control variables such as industrialization level, urbanization rate, agricultural planting structure, income distribution, natural environment conditions, and farmers’ income structure [5,6]. The variables included in the random effects Tobit model are as follows.

3.4. Data Sources

Research data mainly come from “China Statistical Yearbook”, “China Agriculture Yearbook”, “China Rural Statistics Yearbook”, and “China Regional Economic Database”, “China Macroeconomic Database”, “China agriculture, rural areas and farmers Database”, and “ China Environmental Database” on Eps data platform; some data are supplemented by statistical yearbooks of various provinces. This research constructs agricultural input–output panel data and influencing factor analysis panel data of 31 provinces and regions in China from 2003 to 2021.

“National Agriculture Sustainable Development Plan (2015–2030)”aims at the problems faced by sustainable agricultural development in various regions, and comprehensively considers the carrying capacity of agricultural resources, environmental capacity, ecological types and development foundations of various regions, and divides the country into optimized development zone, moderate development zone, and protected development zone. Specific regional divisions include the following: The priority agricultural development zone includes the Northeast Region (Heilongjiang, Jilin, Liaoning, and eastern Inner Mongolia), the Huang-huai-hai Region (Beijing, Tianjin, central and southern Hebei, Henan, Shandong, Anhui, and northern Jiangsu), and the middle and lower reaches of the Yangtze River (Jiangxi, Zhejiang, Shanghai, Jiangsu, south-central Anhui, Hubei, and most of Hunan), South China (Fujian, Guangdong, and Hainan). Moderate agricultural development areas include the northwest and areas along the Great Wall (Xinjiang, Ningxia, most of Gansu, Shanxi, central and northern Shaanxi, central and western Inner Mongolia, and northern Hebei), and southwestern regions (Guangxi, Guizhou, Chongqing, southern Shaanxi, eastern Sichuan, most of Yunnan, Hubei, and western Hunan). The protected development zone includes mainly Qinghai–Tibet areas (Tibet, Qinghai, Gansu Tibetan areas, western Sichuan, and northwest Yunnan).

4. Agricultural GTFP Measurement and Composition/Decomposition

4.1. Analysis of the Temporal Characteristics of Agricultural GTFP and Its Components

When considering agricultural production pollution, from 2003 to 2021, China’s agricultural GTFP grew at an average annual rate of 3.11%; BPC, the technology change index, averaged an annual increase of 3.80%; and the technical efficiency change index decreased by 0.42%. The main driving force of GTFP growth is technological progress (see

Table 3. Summary statistics of GTFP for 31 provinces over 2003–2021.

Region	GTFP		EC		BPC
	Mean	SD	Mean	SD	Mean	SD
Beijing	1.0277	0.0246	0.9936	0.0608	1.0345	0.0704
Tianjin	1.0361	0.0196	0.9992	0.0268	1.0370	0.0359
Hebei	1.0366	0.0186	1.0017	0.0180	1.0356	0.0269
Shanxi	1.0276	0.0101	0.9848	0.0114	1.0490	0.0189
Neimeng	1.0281	0.0050	0.9791	0.0168	1.0547	0.0218
Liaoning	1.0391	0.0259	0.9883	0.0177	1.0488	0.0410
Jilin	1.0155	0.0038	0.9777	0.0195	1.0463	0.0216
Heilongj	1.0453	0.0136	0.9992	0.0079	1.0462	0.0177
Shanghai	1.0675	0.0838	0.9970	0.0598	1.0742	0.1069
Jiangsu	1.0390	0.0153	1.0020	0.0208	1.0467	0.0328
Zhejiang	1.0484	0.0183	0.9993	0.0096	1.0496	0.0213
Anhui	1.0283	0.0033	0.9882	0.0248	1.0418	0.0267
Fujian	1.0339	0.0171	0.9993	0.0080	1.0346	0.0209
Jiangxi	1.0178	0.0030	0.9860	0.0252	1.0437	0.0277
Shandong	1.0396	0.0234	0.9996	0.0082	1.0400	0.0272
Henan	1.0390	0.0154	0.9995	0.0090	1.0396	0.0197
Hubei	1.0294	0.0091	1.0059	0.0320	1.0268	0.0365
Hunan	1.0246	0.0064	0.9991	0.0219	1.0327	0.0247
Guangdo	1.0320	0.0246	0.9995	0.0099	1.0325	0.0288
Guangxi	1.0399	0.0146	1.0004	0.0250	1.0403	0.0342
Hainan	1.0063	0.0454	1.0001	0.0209	1.0062	0.0512
Chongqin	1.0279	0.0106	1.0023	0.0340	1.0332	0.0393
Sichuan	1.0386	0.0190	0.9992	0.0064	1.0396	0.0222
Guizhou	1.0411	0.0157	1.0074	0.0385	1.0479	0.0503
Yunnan	1.0188	0.0048	0.9855	0.0170	1.0342	0.0193
Xizang	1.0046	0.0487	0.9928	0.0602	1.0133	0.0654
Shanxii	1.0416	0.0154	0.9989	0.0124	1.0426	0.0209
Gansu	1.0310	0.0169	0.9991	0.0121	1.0319	0.0227
Qinghai	1.0186	0.0249	0.9903	0.0601	1.0296	0.0766
Ningxia	1.0100	0.0190	0.9966	0.0363	1.0136	0.0425
Xinjiang	1.0292	0.0189	0.9989	0.0118	1.0305	0.0246
Northeast Region	1.0333	0.0144	0.9884	0.0150	1.0471	0.0268
Huanghuaihai Region	1.0346	0.0175	0.9970	0.0246	1.0381	0.0345
The middle and lower reaches of the Yangtze River	1.0378	0.0227	0.9982	0.0282	1.0456	0.0417
Southern China	1.0241	0.0290	0.9996	0.0129	1.0244	0.0336
Northwest and areas along the Great Wall	1.0279	0.0142	0.9929	0.0168	1.0371	0.0252
Southwest Region	1.0333	0.0129	0.9990	0.0242	1.0390	0.0331
Qinghai–Tibet area	1.0116	0.0368	0.9916	0.0602	1.0215	0.0710

Note: GTFP represents green total factor productivity; EC represents efficiency change; BPC represents best-practice gap change; SD represents standard deviation.

). When ignoring agricultural production pollution, the average annual growth rate of China’s agricultural TFP from 2003 to 2021 was 10.42%, the technology change index increased by 10.81%, and the technical efficiency change index decreased by 0.44% (see attached

Table A2. Summary statistics of TFP from 2003 to 2021.

Period	TFP		EC		BPC
	Mean	SD	Mean	SD	Mean	SD
2003–2004	1.0015	0.0143	0.9755	0.0247	1.0273	0.0293
2004–2005	0.9972	0.0137	1.0217	0.0538	0.9790	0.0499
2005–2006	1.0472	0.0252	0.9950	0.0517	1.0527	0.0593
2006–2007	1.8213	0.0330	1.0384	0.0534	1.7540	0.0920
2007–2008	1.1095	0.0154	1.0047	0.0317	1.1052	0.0379
2008–2009	1.0370	0.0130	0.9841	0.0305	1.0542	0.0369
2009–2010	1.0211	0.0129	0.9843	0.0319	1.0381	0.0364
2010–2011	1.1155	0.0146	1.0234	0.0377	1.0908	0.0420
2011–2012	1.1017	0.0139	0.9984	0.0387	1.1030	0.0424
2012–2013	1.0285	0.0164	0.9921	0.0332	1.0364	0.0403
2013–2014	1.1871	0.0212	0.9852	0.0451	1.2049	0.0573
2014–2015	1.1165	0.0184	0.9962	0.0361	1.1215	0.0437
2015–2016	1.0805	0.0160	1.0183	0.0371	1.0612	0.0386
2016–2017	1.0855	0.0106	1.0012	0.0300	1.0845	0.0332
2017–2018	1.0555	0.0123	0.9882	0.0287	1.0684	0.0330
2018–2019	1.0468	0.0166	0.9699	0.0362	1.0805	0.0405
2019–2020	0.9849	0.0194	0.9442	0.0576	1.0435	0.0563
2020–2021	1.0386	0.0262	0.9990	0.0449	1.0397	0.0411
Mean	1.1042	0.0174	0.9956	0.0390	1.1081	0.0450

). Therefore, this study reached a consistent conclusion that the agricultural GTFP under resource and environmental constraints is significantly lower than the traditional agricultural TFP that does not consider resource and environmental constraints. Traditional TFP calculations significantly overestimate the role of China’s agricultural development performance and technological progress, and may trigger policies that misjudge revelations [2,3,5,6,8,9,30,31].

Judging from the temporal characteristics of agricultural GTFP, the EC index, and the BPC index (see Figure 1, Figure 2 and Figure 3), China’s comprehensive agricultural production capacity has been continuously improved under resource and environmental constraints, and it has shown a typical technology-driven growth model [1]. The Chinese government has always attached importance to agricultural technological innovation; actively promoted independent innovation of key agricultural technologies such as biological seed, heavy agricultural machinery, smart agriculture, and green inputs; accelerated the establishment of scientific and technological innovation platform bases such as the State Key Laboratory in the agricultural field; and focused on strengthening enterprise technological innovation’s dominant status. The advancement of agricultural technology has become a key measure to reduce agricultural resource consumption and environmental damage and to improve the level of agricultural industry, and has become a key driving force for the growth of agricultural GTFP. However, due to the imperfect agricultural technology research system, insufficient commercialization of agricultural research findings, and imperfect agricultural technology extension systems and socialized service systems, the rate of application of agricultural technology in China is significantly lower than the speed of technological progress, making lower agricultural technology efficient [5]. At the same time, as China’s agricultural reform enters the “deep water zone”, deep-seated problems such as agricultural management and systems have not been fundamentally alleviated, which has restricted the efficiency of agricultural technology to a certain extent.

4.2. Regional Difference Analysis of Agricultural GTFP and Its Components

Table 3 lists the average agricultural GTFP, BPC index, and EC index of each agricultural sustainable development planning zone in China from 2003 to 2021. The results show that there are significant differences in average agricultural GTFP, BPC index, and EC index among provinces and regions in China from 2003 to 2021. The GTFP levels of Shanghai (6.75%), Zhejiang (4.84%), Heilongjiang (4.53%), Shaanxi (4.16%), and Guizhou (4.11%) were relatively high, indicating that the coordination of agricultural economic development and resource environment in these five provinces and regions has significantly improved, and the ability of sustainable agricultural development has been continuously enhanced. The agricultural GTFP of Tibet (0.46%), Hainan (0.63%), and Ningxia (1.00%) is relatively low, and the agricultural technological progress of these three provinces and regions is also at a low level. As a result, the impetus for agricultural growth driven solely by technological progress is insufficient, and agricultural development still requires relatively large resource consumption and high environmental cost. It is worth noting that the source of agricultural GTFP growth in Hainan, Guangxi, Hebei, Jiangsu, Chongqing, Hubei, Guizhou, and other provinces is the two-wheel drive of technological progress and technical efficiency, presenting a typical intensive agricultural growth mode [1].

Table 3 shows that there are significant differences in the agricultural GTFP, BPC index, and EC index in each district. The agricultural priority development zone had the highest GTFP, with an average annual growth rate of 3.24%; the agricultural moderate development zone had the second place, with an average annual growth rate of 3.06%; and the agricultural protection development zone had the lowest, with an average annual growth rate of 1.16%. In addition, the GTFP in South China, the northwest and areas along the Great Wall, and the Qinghai–Tibet area is also significantly lower than the national average GTFP level (3.11%). To some extent, this reflects the grim situation of resource conservation and environmental protection in the agricultural production of these three regions.

From the perspective of GTFP growth sources, the growth of China’s agricultural GTFP from 2003 to 2021 mainly relied on technological progress. The BPC index of the agricultural priority development zone was 3.88%, that of the agricultural moderate development zone was 3.80%, and that of the agricultural protection development zone was 2.15%. Among these areas, South China (2.44%) and Qinghai–Tibet (2.15%) have relatively low levels of technological progress, which may be due to insufficient agricultural economic contributions or weak economic foundation, which makes their agricultural technological progress relatively insufficient. From the perspective of the EC index, China’s agricultural technological efficiency was low from 2003 to 2021. The EC index of the agricultural priority development zone was −0.42%, that of the agricultural moderate development zone was −0.41%, and that of the agricultural protection development zone was −0.84%. Among these areas, the areas along the Northwest and the Great Wall (−0.71%) and the Qinghai–Tibet area (−0.84%) have low levels of technical efficiency. This may be due to a one-sided pursuit of agricultural technology advancement and neglect of agricultural technology promotion and application, which make agriculture technically inefficient. The Northeast Region (−1.16%) has the lowest level of agricultural technology efficiency. It may be difficult to improve technical efficiency due to the high level of agricultural organization, advanced agricultural technology, and high degree of marketization in the Northeast. As a result, the speed of catching up in agricultural technology (technical efficiency) is lower than the speed of technological progress, thus leading to the degradation of agricultural technical efficiency [5].

5. Public Investment, Environmental Regulations, and Agricultural Sustainability Development

5.1. Influencing Factors of Agricultural GTFP and Its Components

Using the random effects Tobit model, this paper analyzes the effects of public investment, environmental regulations, and control variables on agricultural GTFP and its components (

Table 4. Analysis of influencing factors of agricultural GTFP and its components.

	Model I (GTFP)	Model II (EC)	Model III (BPC)
PI_Agricu	0.0748 (0.1194)	−0.0900 (0.0579)	0.1319 (0.1292)
PI_R&D	3.6163 *** (0.8913)	1.6372 *** (0.3369)	3.0475 *** (0.9563)
ER_Policy	−0.0105 (0.0095)	−0.0018 (0.0049)	−0.0076 (0.0106)
ER_Energy	0.0085 (0.0093)	−0.0111 ** (0.005)	0.0210 ** (0.0103)
Industry	0.3014 *** (0.0558)	0.1335 *** (0.0270)	0.3445 *** (0.0589)
Urbanization	0.2745 *** (0.0419)	0.0941 *** (0.0179)	0.2606 *** (0.0399)
Agristructure	0.4422 *** (0.0436)	0.2302 *** (0.0194)	0.4622 *** (0.0449)
Distribution	0.7285 *** (0.0878)	0.2933 *** (0.0416)	0.7764 *** (0.0962)
Environment	−0.0245 (0.0208)	0.0095 (0.0118)	0.0010 (0.0235)
Salary	0.1931 *** (0.0488)	0.1341 *** (0.0234)	0.1314 ** (0.0538)
σu	0.1056 *** (0.0311)	0.5815 *** (0.0741)	0.0859 *** (0.0147)
σe	0.0529 *** (0.0024)	0.0306 *** (0.001)	0.0597 *** (0.0025)
ρ	0.7996	0.9972	0.6746
Prob > χ2	0.0000	0.0000	0.0000
Log likelihood	439.447	817.363	320.670

Note: σ_u is the estimated value of the individual effect; σ_e is the estimated value of the random interference item; the standard error is in the brackets; ***, ** indicate the significance levels of 1% and 5%, respectively.

). The individual error and random error of the model are both small, the variance ratio ρ is greater than 0.6, the individual effect variance accounts for a large proportion, the likelihood ratio (LR) is large, and the null hypothesis that the individual effect is zero is strongly rejected. The model fits well, and it is reasonable to use the random effects panel Tobit model to regress.

From the perspective of public investment, the support variable of technology innovation input (PI_R&D) has a significant positive impact on agricultural GTFP and the BPC index. Investment in technological innovation has a leverage effect, which can reduce the risk of enterprise technological innovation, enhance the vitality of agricultural technological innovation, and enhance the level of agricultural technological progress. However, the agricultural financial support variable (PI_Agricu) did not show statistical significance for agricultural GTFP and its components.

From the perspective of environmental regulations, the agricultural non-point source pollution remediation policy variable (ER_Policy) has no significant impact on agricultural GTFP and its components. The environmental regulation intensity variable (ER_Energy) has significantly increased the level of the BPC index, but to a certain extent resisted technological efficiency changes.

The variable industrialization level (Industry) and urbanization rate (Urbanization) have a significant positive impact on agricultural GTFP, the EC index, and the BPC index. The external impact of agricultural productivity affects the scale and efficiency of the regional industrial sector [51]. The rapid advancement of industrialization can also accumulate strong capital, cutting-edge technology, and advanced management concepts for agricultural development, and promote an increase in agricultural productivity. As the process of urbanization continues to accelerate, the role of urban economic agglomeration continues to expand, and the consumption scale of green agricultural products continues to increase, which can improve the supply quality and application efficiency of “eco-friendly” agricultural technologies, and contribute to the improvement of agricultural GTFP and its components.

Agricultural planting structure variables (Agristructure) have a significant positive effect on GTFP, the EC index, and the BPC index. Adjusting the structure of grain, economic, and forage planting is an important measure to adapt to changes in market demand, optimize the structure of the agricultural industry, and promote agricultural quality and efficiency. In recent years, China has been focusing on determining the grain planting area, optimizing the regional distribution of cash crops, expanding the area of forage crop planting, and continuing to carry out pilot projects for reforming grain-to-feed and grain-to-bean subsidies. Under the food security strategy of stabilizing food production and ensuring food security, China has accelerated the advancement of high-quality food projects, deeply explored the potential for stable production and increased production of varieties and technologies, accelerated the promotion and application of fertilizer and drug reduction technologies and the utilization of food straw resources, and promoted agricultural production combined with quality improvement; the agricultural environmental performance has effectively improved.

The variable income distribution (Distribution) and variable farmers’ income structure (Salary) have significant positive effects on agricultural GTFP, the EC index, and the BPC index. Fairness in income distribution can help improve the quality of the environment [52]. When the per capita income level increases and the degree of inequality decreases, the state will increase public expenditure on the ecological environment, thereby improving the quality of the agricultural environment [53]. In recent years, the level of per capita disposable income of Chinese urban and rural residents has continued to increase, the income gap between urban and rural residents has been further narrowed, and fairness in income distribution has been significantly enhanced. This has enabled the government to increase public investment in agricultural non-point source pollution control, and effectively improved agricultural GTFP and its components, helping to enhance the environmental quality of agriculture. This study reached a conclusion consistent with Du et al. (2016) [5].

Agriculture is a typical weak industry which is extremely vulnerable to meteorological disasters, diseases, pests, and weeds. The variables of the natural environment (Environment) and agricultural GTFP, EC index, and BPC index do not show a significant impact, but this does not absolutely deny the negative correlation with agricultural GTFP.

5.2. Public Investment and Sustainable Agricultural Development

Agriculture is an industry where natural reproduction and economic reproduction are intertwined. As a typical weak industry, agriculture is always faced with natural risks and market risks at the same time. Agricultural financial support mainly focuses on agricultural infrastructure construction, agricultural science and technology research and technology popularization, agricultural education and training, and other fields. It has a relatively high economic pull effect and income spillover, and can greatly overcome the influence of production factors, resource endowment, and ecological conditions on agricultural production, so as to improve agricultural development performance [39]. Since 2003, the Chinese government has continued to increase public investment in agriculture. It is sufficient to improve agricultural infrastructure, enhance the vitality of agricultural technology innovation, and improve the quality of agricultural technology extension services. However, the impact of the agricultural financial support variable (PI_Agricu) on agricultural GTFP and its components in this study does not meet this expectation, and this may involve the implementation efficiency and input structure of fiscal support for agriculture [5]. It is a rational choice of the Chinese government to implement an active agricultural financial support policy to provide public goods necessary for agricultural growth. However, due to the lack of information on farmers’ preferences for agricultural public products by policy makers, the government-determined agricultural public investment structure deviates from the demand structure of agricultural development, which weakens the efficiency and effectiveness of agricultural public investment policies [5,39,54]. Although China’s total agricultural fiscal expenditures have grown steadily, the investment scale of agricultural fiscal support in improving the agricultural technology extension system, the cultivation of new agricultural business entities, and the education and training of farmers’ professional skills is still limited, which means the adoption of green agriculture production elements and the publicity of agricultural green management and environment-friendly technology have not obtained remarkable achievement; this greatly restricts agricultural sustainable development potential. At the same time, due to differences in geographical conditions, resource endowments, economic foundations, social development, etc., there are significant spatial differences in the impact of agricultural financial support on agricultural performance [54]. The overall scale of agricultural financial support and the rate of return on investment in the priority agricultural development zone (developed areas such as Northeast and South China) are higher than those of moderately developed areas and the protected development zone (underdeveloped areas such as the Southwest, Northwest, and Qinghai–Tibet areas), which is consistent with the distribution of agricultural GTFP (Figure 1), EC (Figure 2), and BPC (Figure 3) indexes in each region.

Technological innovation and technological progress are key forces that promote high-quality social and economic development. Government R&D investment and other technological innovation support are manifestations of Pigovian tax, which is a free transfer payment for the government to correct market failure, and provides an important financial foundation for technological innovation and technological progress. With the in-depth implementation of the innovation-driven development strategy and the continuous advancement of the construction of an innovative country, the R&D funding support of various research entities has continuously increased. This initiative actively promotes key industrial generic technology innovation and effectively kick-started measured soil fertilizer, subsidies for low-toxicity biological pesticides, subsidies for unified control of diseases and insect pests, subsidies for recycling of residual mulch film, protection and improvement of cultivated land quality, demonstration of agricultural clean production, combined planting and breeding with circular agriculture, and resource utilization of crop stalks and other agricultural technology innovations, thereby providing technical support for the sustainable development of agriculture. This study also verified the significant positive impact of technological innovation input support variables (PI_R&D) on agricultural GTFP and its components.

5.3. Environmental Regulation and Sustainable Development of Agriculture

After long-term exploration and improvement, China has formed an environmental regulatory system with command-control environmental regulations, market-based incentive environmental regulations, and voluntary environmental regulations to ease resource and environmental constraints, responded to global climate change, and enhanced the capability for sustainable social and economic development. The applicability and effectiveness of environmental regulations are important factors for the growth of agricultural GTFP [15]. The enhancement of agricultural non-point source pollution control can trigger farmers to use advanced technologies to improve soil, prevent and control diseases and insect pests, and achieve precision fertilization, etc., to reduce environmental pollution of farmland [46]. From the perspective of the evolution of China’s agricultural pollution prevention and control system, the process of agricultural environmental regulation is slow and passive, the content of policy regulation is relatively scattered, agricultural pollution control is marginalized, and it even triggers adverse selection behaviors by farmers, enterprises, governments, and other stakeholders [55]. “Implementation Opinions on Fighting the Tough Battle for the Prevention and Control of Agricultural Non-point Source Pollution (2015)”, “National Agricultural Sustainable Development Plan (2015–2030)”, and other policy documents issued clarify the detailed targets and time points for the treatment of agricultural non-point source pollution, such as fertilizers, pesticides, and residual mulch film. Environmental regulation in the agricultural sector has officially shifted from advocacy and advice to practical operation, and from decentralization and marginalization to centralized governance. However, due to the decentralized nature of the main agricultural production, the path dependence of the traditional agricultural development model, the lower quality of the agricultural labor force, and other elements, a series of agricultural non-point source pollution control policies since 2015 have not been seen in the short term; that is, the agricultural non-point source pollution control policy variable (ER_Policy) has not shown a significant impact on agricultural GTFP and its components.

Strict environmental regulations force technology companies to upgrade agricultural technology and equipment, increase investment in green technology research, actively turn to green technology innovation, induce farmers’ green production behaviors to promote agricultural technology progress and improve agricultural technology efficiency, and then enhance agricultural GTFP [15]. From a dynamic point of view, the impact of the intensity of environmental regulations on the progress of agricultural green technology presents a U-shaped trend of first restraining and then promoting. However, when the intensity of environmental regulations is low, it is often difficult to effectively promote the technological innovation and technological progress of relevant stakeholders, and it is difficult to effectively improve agricultural GTFP. In recent years, the Chinese government has attached great importance to the governance of the agro-ecological environment, and the regulatory quality and control of the agro-ecological environment have been continuously strengthened. To a certain extent, this has prompted agricultural enterprises and farmers, especially large farmers, to choose advanced technologies to achieve cleaner production, energy saving, and emission reduction. In addition, it forces technology companies to increase investment in green technology innovation, realize agricultural green technology innovation, and seize market share [49]. An appropriate intensity of environmental regulation will help to achieve the dual effects of end-of-point management of farmers and enterprise technological innovation in non-point source pollution control, and achieve a win–win situation of ensuring farmers’ income and business performance, and improving the ecological environment of farmland.

Agricultural green technology innovation has the “dual externalities” of negative externalities related to environmental pollution and positive externalities related to knowledge spillover. It is difficult to rely on market mechanisms to complete technological innovation. Environmental regulatory policies and agricultural technology research and development funding have offset the dual externalities and funding shortages of agricultural green technology innovation to a certain extent.

6. Conclusions and Policy Implications

The accounting of GTFP and its components under environmental constraints is a scientific method for grasping the true performance of China’s agricultural economic growth and identifying the driving factors of China’s agricultural economic growth, which helps to provide a scientific basis for the policy formulation and policy evaluation of sustainable agricultural development in China. This study is based on the input–output panel data of the agricultural planting industry in various provinces and regions in China from 2003 to 2021, taking into account the loss of pesticides, residues of agricultural mulch film, and other undesired outputs, and calculating the GTFP and its components of China’s agricultural growth, focusing on the analysis of the impact mechanism of public investment and policy regulations on agricultural GTFP. The main conclusions of this study are as follows: (1) The agricultural GTFP under resource and environmental constraints is significantly lower than the traditional agricultural TFP, and the traditional TFP calculation significantly overestimates the role of China’s agricultural development performance and technological progress. (2) Judging from the time series characteristics of agricultural GTFP, the EC index, and the BPC index, China’s comprehensive agricultural production capacity has continuously improved under resource and environmental constraints, and it has shown a typical technology-driven growth model. (3) The agricultural GTFP, EC index, and BPC index show significant spatial differences in each subregion, and from the overall perspective, agricultural priority development zone > agricultural moderate development zone > agricultural protection and development zone. (4) Technical innovation input support variables (PI_R&D), environmental regulation intensity variables (ER_Energy), industrialization level variables (Industry), urbanization rate variables (Urbanization), agricultural planting structure variables (Agristructure), income distribution variables (Distribution), and farmers’ income structure variables (Salary) have different degrees of significant influence on agricultural GTFP and its components. (5) The combination of active technological innovation support and appropriate environmental regulation will help to improve agricultural GTFP and accelerate the sustainable development of agriculture.

It is worth noting that technological innovation, technological achievement transformation, and technological application are the three basic activities of the science and technology chain. All regions should adhere to the dominant position of agricultural technological innovation, accelerate the transformation and application of agricultural technological achievements, improve the agricultural technology extension system, and rely on agricultural technology innovation to build a new engine of modern agricultural development. At the same time, an excessively high intensity of environmental regulations would lead to excessive burdens on enterprises, making it difficult for technological innovation effects to make up for cost losses, and excessive government R&D funding may make enterprises inert in innovation and reduce the efficiency of agricultural technological innovation. At the same time, high-strength command-control environmental regulations may trigger negative behaviors in farmers. Therefore, the Chinese government should carefully design regulatory tools, reasonably determine the intensity of environmental regulations, accelerate the establishment of a balanced “environmental regulatory portfolio” system, and enhance the applicability and operability of agriculture environmental regulatory policies.

Based on the above analysis, this research has drawn some policy suggestions. (1) Continue to increase investment in agricultural technology innovation, support the key enterprises in researching agricultural green technology, and increase the contribution rate of agricultural technology progress in order to stabilize the advantage of agricultural technology. At the same time, pay attention to the optimal allocation of agricultural resources, accelerate the improvement of the agricultural technology extension system, and increase the publicity of agricultural non-point source pollution hazards, so as to effectively improve the driving force of agricultural technology efficiency on agricultural GTFP. (2) Fully respect the agricultural GTFP spatial differences in the agricultural priority development zone, moderate development zone, and protected development zone, promote the transition from extensive management to intensive management in the moderately developed zone and protected development zone, and focus on the advancement of their agricultural technology and the improvement of agricultural technology efficiency. (3) Continue to increase agricultural financial support, focus on improving the use efficiency and input structure of agricultural financial support funds, and pay attention to the benefits of agricultural support in the Northwest, Southwest, and Qinghai–Tibet areas in the moderate agricultural development zone and protected development zone. (4) Combine the differences in the technical backgrounds, institutional backgrounds, and economic foundations of each region, and establish a diversified and differentiated environmental regulatory policy system that effectively integrates command-control environmental regulations, incentive environmental regulations, and voluntary environmental regulations. At the same time, pay attention to the synergy of agricultural environmental regulation and technological innovation input support to provide an effective driving force for sustainable agricultural development.

There are some deficiencies in this study. Due to the dispersion, concealment, hysteresis, and spatial heterogeneity of agricultural non-point source pollution, agricultural pollution units and total pollution are difficult to accurately estimate, as is the case for industrial pollution. Although this study included pesticides, fertilizers, agricultural mulch films, and straws as much as possible, and applied the unit survey evaluation method to estimate agricultural undesired output, there is still a certain gap from an accurate estimate. This research will continue to track the latest scientific methods of accounting for agricultural undesired output, further optimize the results of agricultural GTFP accounting, and provide a more scientific basis for the formulation of sustainable agricultural development policies.