Towards Sustainable Agriculture in China: Assessing the Robust Role of Green Public Investment

Abstract

This paper reviews the long-term impact of public investments on irrigation and agricultural research and development along with other control variables, including physical capital, irrigated area, fertilizer consumption, level of mechanization, and CO₂ emissions on China’s agricultural output from 1986 to 2017. This study applied various econometric methods such as the ARDL bound-testing approach and Johansen co-integration procedure to determine the long-term co-integrating connection amid the variables. The empirical outcomes from the ARDL bound-testing method confirm a long-term co-integrating connection among the variables. The long-run results demonstrated that public investment in agricultural research and development and irrigation have a substantial positive effect on agricultural productivity. Furthermore, results revealed that physical capital and fertilizer consumption also have a significant positive effect on agricultural output; however, CO₂ emissions have a substantial negative effect on agricultural production. These findings therefore suggest that the policy makers of China should initiate more effective policies to increase irrigation and agricultural research and development investments. Increasing irrigation and agricultural research and development investments will enhance agricultural productivity by ensuring food security in the country.

1. Introduction

Government spending leads to the accumulation of capital and promotes long-term economic development [1]. Since capital is injected in rural regions, it contributes to general employment and wages and also adds to overall economic development by providing affordable food to the growing population in urban areas [2]. Many empirical studies have employed several econometric methods to explore the interrelationship between agriculture and economic progression. The law of wager emphasizes economic progression as the fundamental determinant of government spending [3], while Keynesian procedure states that government spending is a key factor in economic development [4]. Several studies including one from Guandong & Muturi [5] in east Sudan admire Keynesian principles, while Wanger’s law is supported by Lingxiao et al. [6] in Romania, Palamalai [7] in India, and Salih [8] in Sudan. Furthermore, Liu et al. [9] and Loizides and Vamvoukas [10] concluded that public spending affects economic development in the short-term and long-term. Liu et al. [9] explored the idea that Keynesian theory is more powerful than the law of Wanger. Agricultural potential leading to economic development has been controversial among development economists [11]. Much of the previous work on this discussion corresponds to the debate on agriculture-caused development in many developing countries [12,13,14,15,16]. Agricultural growth advocates argue that innovation in agricultural development requires improving infrastructure and institutional growth [16,17]. These studies reported that the development of agriculture is the driver for global economic progression, and its influence on rural incomes and structural transformation is considerable [18,19,20]. Several studies have demonstrated a positive interrelationship between public spending on agriculture and economic progress [21,22,23]. A few studies have also analyzed the impacts of agricultural sub-sectors on economic development. Shuaib et al. [24] argued that research and development investment has the greatest impact on agricultural growth. However, this is an initial study and it contributes to the literature in several ways. First, this study investigates the long-term impact of public investments in irrigation and agricultural research and development on agricultural value added in the case of China over the 1986 to 2017 period by engaging the ARDL approach. Second, the present study also incorporated other important factors, including: physical capital, irrigated area, fertilizer used, mechanization, and CO₂ emissions. The dynamic relationship between considered variables is displayed in Figure 1. Finally, the current study employed various econometric procedures on the time-series data, including the ADF and KPSS unit root tests to check the stationarity of the variables, while the Johansen co-integration method was used to check the robustness results of the ARDL bound-testing approach. The study also answers the following questions.

• How does public spending on agriculture research and development impact China’s agricultural output?
• Does a rise in public spending on irrigation practices improve China’s agricultural output?
• Does agricultural technology (fertilizer usage and mechanization) adoption improve China’s agricultural output?

2. Public Spending and Chinese Agricultural Output: An Overview

Government spending in developing countries is the elementary tool used by policymakers to stimulate economic growth, which is a fundamental ingredient to boost overall economic development such as: providing better infrastructure, education, housing, and health and increasing agricultural output [23,25]. Government investment in all sectors of the economy is very important. In the agricultural sector, government spending essentially depends on the materials required by farmers to support annual increase, such as fine-quality insecticide and a proper water irrigation system to promote better crop production [2,26,27,28,29]. Government agricultural spending has thus been incurred in the public sector, relating to government expenditure at different levels, including the national and local levels [30]. Since 1978, China’s agricultural research program has rapidly expanded and is now ranked as the world’s largest system. The Chinese government has increased public spending and has implemented many pro-agricultural policies, aiming to enhancing agricultural productivity and farmers’ earning capacity [31]. The Chinese national government spent around 5000 billion Yuan on agricultural irrigation, subsidies, and research and development in 2004, which accounts for 16 percent of the total national government spending. Over the period from 2005 to 2007, agricultural subsidies supplied by the central government increased quickly; the rice subsidy was raised 12 billion Yuan in 2008, but in 2010, it exceeded 83.5 billion Yuan. In 2011, gross subsidies amounted to 133.49 billion Yuan. As per the national product cost survey, subsidies for agriculture have rapidly been augmented from 2004 to 2008. The rice subsidy was augmented from approximately 200 Yuan per hectare in 2004 to over 850 Yuan per hectare in 2008. In 2004, major cereal crops (i.e., wheat and maize) subsidies were approximately 100 Yuan per hectare and both increased to approximately 700 Yuan per hectare in 2008. From 2004 to 2008, about 378 and 327 Yuan per hectare were the subsidies for oil seeds and cotton crops, respectively [32]. In addition, government investment in irrigation accelerated much more rapidly from the 1980s to 2010; the average annual growth rate increased by 13 percent and reached approximately 68 billion Yuan in 2010 [33].

3. Literature Review

Many researchers have worked in various parts of the world to see the impacts of government spending on agricultural productivity by employing different econometric techniques during different periods. Empirical evidence from various emerging nations on the impacts of education demonstrated high returns to poverty obliteration and showed a positive interrelationship. For example, [34,35,36] and Yasmeen et al. [37] inspected the effect of several variables on agricultural output. Fan et al. [38] found that education and infrastructure positively and significantly affected agricultural productivity and reduced regional inequalities and poverty as well. Furthermore, they reported that government expenditure had a greater effect on agricultural productivity and on the rural economy. Benin et al. [39] attempted to reveal the influence of government expenditure on agricultural output in Ghana by utilizing district- and regional-level data from the period of 2001 to 2006. Findings exhibited that public spending on agriculture, rural infrastructures, health, and education played an important role in enhancing agricultural output. Likewise, an empirical study conducted in Tamil Nadu, India by Ashok and Balasubramanian [40] found that public investment in irrigation, markets, education, and rural infrastructure increased TFP growth of agriculture. In addition, the positive impact of farmers’ education on farm production is well documented.

Ahmed et al. [26] conducted an empirical study in Pakistan to investigate the influence of government expenditure on value-added agriculture during the time period of 1972–2014. Several estimation procedures containing the Johansen co-integration approach and the ARDL method were applied in this investigation. The estimated results showed that agriculture value addition was positively affected by government expenditure on health, education, and rural infrastructure in Pakistan. Similarly, Ewubare and Eyitope [41] conducted a research study in Nigeria to explore the impact of public spending on agricultural output by employing the ECM. Research findings pointed out that public expenditure, commercial bank loans to agriculture, and gross capital formation positively affected agricultural output in Nigeria. The authors proposed an increase in funding for the agricultural sector in Nigeria based on the above findings. Additionally, Xu et al. [32] examined the dynamic economic effects of different types of public agricultural expenditure on food production in China by employing the DCGE method. The research showed that public agricultural spending has a significant influence on food production, price, and trade. Furthermore, augmented government agricultural expenditures on research and development, irrigation, and agricultural subsidies also have certain effects on new sectors, including industry, service, and GDP progression.

Meanwhile, for Nigeria, Kolawole [42] attempted to analyze the long-term casual nexus among government expenditure and inclusive development over the period from 1995 to 2014. This research used the ARDL bounds test to check both the long- and short-term interrelationships among the studied variables. The empirical results demonstrated that public expenditures on health, financial freedom, common resources, and real GDP growth have a positive significant effect on inclusive progression in the long-term, whereas in the short-term, only real GDP growth has a significant effect on inclusive progression. Matthew and Mordecai [43] explored the causal association between public agricultural expenditure and agricultural production for the time period of 1981 to 2014 in Nigeria by using the Johansen co-integration technique, the ARDL approach, and the Granger causality test. Findings revealed that agricultural public spending has a significant interrelationship with agricultural production and that the rate of interest has an inappreciable positive association with agricultural production in Nigeria.

For Meghalaya, India, Dkhar and Kumar [44] analyzed the effects of public spending on agricultural and allied activities on economic progression over the period of 1984–1985 to 2013–2014. Results from the study revealed that the influence of public expenditure on forestry, dairy and, irrigation proved to be negative. Further results showed a positive significant effect of agricultural expenditure on crop cultivation on GSDP growth. Al-Bataineh [45] empirically discussed the influence of public spending on economic progression in Jordan in a period of time from 1990 to 2010. Findings of the research showed that aggregate public spending had a positive effect on economic progression in the country. Likewise, Alshahrani and Alsadiq [28] inspected the heterogeneous influence of various components of public expenditure on economic progress for Saudi Arabia using annual data for the period of 1969 to 2010. The empirical results demonstrated that healthcare spending and public and private funds boosted economic progress. The authors recommended public expenditure on real estate to increase short-term production.

Nurudeen and Usman [46] reported that total recurrent, total capital, and education expenditures had negative effects on economic progression. In another view, public expenditures on health, transference, and communication had a positive impact on economic progression in Nigeria. In addition, Adeniyi and Bashir [47] found that economic progression was positively affected by public expenditure on value-added agriculture, education, and structural adjustment programs in Nigeria. In their work, Salim and Islam [48] attempted to assess the dynamic interrelationship between research and development spending, rainfall, agricultural production, and productivity growth in Western Australia by utilizing time series data from 1977 to 2005. Findings indicated that government research and development spending resulted in agricultural output growth in the long term. Furthermore, based on the approach of co-integration testing and the VEC method, the estimated results indicated a unidirectional relationship between government research and development spending and TFP growth in both the short and long term.

In their pioneering research, Hall and Scobie [49] explored the role of the productivity of research and development input growth in New Zealand from the period of 1927 to 2001. The empirical results exhibited that research and development spending effects on agricultural productivity increased in the country. In another pioneering study, Saengchai et al. [50] employed the ARDL approach to examine the impact of agricultural government spending, agricultural loans, and extension services on the agricultural GDP. Findings pointed out that for long-term agricultural public spending, expansion services had a significant influence, while in the short term, only extension services had a great effect on real agricultural GDP in Asian countries.

4. Data and Methods

4.1. Data

The present empirical study attempted to explore the dynamic linkages between agricultural research and development expenditure, irrigation investment, gross capital formation, irrigated area, fertilizer consumption, level of mechanization, carbon dioxide (CO₂) emissions, and agricultural value added, using annual data from China covering the period from 1986 to 2017. The selected study variables in the multivariate framework comprise agricultural value added (% of GDP), agricultural research and development expenditure (million Yuan), irrigation investment (billion Yuan), gross capital formation (% of GDP), irrigated area (million hectares), fertilizer consumption (thousand tons), level of mechanization (number of tractors), and carbon dioxide (CO₂) emissions (million tons). The data of agricultural value added and gross capital formation come from the WDI [51], the data of agricultural research and development expenditure, irrigation investment, irrigated area, fertilizer consumption, and level of mechanization are from China Statistical Yearbooks, and the data of carbon dioxide (CO₂) emissions is come from the SRWE [52]. The overview of descriptive statistics is exhibited in the box plots in Figure 2. The tendencies of all selected variables are demonstrated in Figure 3.

4.2. Model Specifications and Estimation Method

This study empirically examines the interaction between agricultural research and development expenditure, irrigation investment, gross capital formation, irrigated area, fertilizer consumption, level of mechanization, carbon dioxide (CO₂) emissions, and agricultural output in China. The multivariate framework was established as follows:

$$AV{A}_t=f\left(R{D}_t,IRI{N}_t,PH{C}_t,IR{A}_t,F{C}_t,MEC{H}_t,C{O}_{2t}\right)$$

(1)

After the natural logarithm, Equation (1) is expressed as follow:

$$\begin{gathered} lnAV{A}_t={\alpha }_0+{\alpha }_{1}lnR{D}_t+{\alpha }_{2}lnIRI{N}_t+{\alpha }_{3}lnPH{C}_t+{\alpha }_{4}lnIR{A}_t+{\alpha }_{5}lnF{C}_t \\ +{\alpha }_{6}lnMEC{H}_t+{\alpha }_{7}lnC{O}_{2t}+{\epsilon }_t \end{gathered}$$

(2)

The ARDL bounds method for co-integration of Pesaran et al. [53] was applied to investigate the interaction between agricultural research and development spending, irrigation investment, gross capital formation, irrigated area, usage of fertilizer, level of mechanization, CO₂ emissions, and agricultural output in China. The ARDL technique provides effective outcomes whether the series are integrated at I(0) or I(1) or mutually co-integrated (I(0) and I(1)). This approach enables the determination of simultaneous long- and short-term associations among the selected research variables in small sample sizes as well large sample sizes [53]. In recent years, this approach has been widely used in many previous empirical studies [26,54,55,56,57,58,59].

The equations below show the ARDL bounds technique for co-integration:

$$\begin{array}{ll}\hfill \Delta lnAV{A}_t={\alpha }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\alpha }_{1i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{2i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{3i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{4i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{5i}\Delta lnIA{R}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{7i}\Delta lnC{O}_{2t-i}+{\alpha }_{8}lnAV{A}_{t-1}+{\alpha }_{9}lnR{D}_{t-1}+{\alpha }_{10}lnIRI{N}_{t-1}+{\alpha }_{11}lnPH{C}_{t-1}\\ & +{\alpha }_{12}lnIA{R}_{t-1}+{\alpha }_{13}lnMEC{H}_{t-1}+{\alpha }_{14}lnC{O}_{2t-1}+{\epsilon }_t,\end{array}$$

(3)

$$\begin{array}{ll}\hfill \Delta lnR{D}_t={\beta }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\beta }_{1i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{2i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{3i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{4i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{5i}\Delta lnIA{R}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{7i}\Delta lnC{O}_{2t-i}+{\beta }_{8}lnR{D}_{t-1}+{\beta }_{9}lnAV{A}_{t-1}+{\beta }_{10}lnIRI{N}_{t-1}+{\beta }_{11}lnPH{C}_{t-1}\\ & +{\beta }_{12}lnIA{R}_{t-1}+{\beta }_{13}lnMEC{H}_{t-1}+{\beta }_{14}lnC{O}_{2t-1}+{\epsilon }_t,\end{array}$$

(4)

$$\begin{array}{ll}\hfill \Delta lnIRI{N}_t={\gamma }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\gamma }_{1i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{2i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{3i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{4i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{5i}\Delta lnIA{R}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{7i}\Delta lnC{O}_{2t-i}+{\gamma }_{8}lnIRI{N}_{t-1}+{\gamma }_{9}lnR{D}_{t-1}+{\gamma }_{10}lnAV{A}_{t-1}+{\gamma }_{11}lnPH{C}_{t-1}\\ & +{\gamma }_{12}lnIA{R}_{t-1}+{\gamma }_{13}lnMEC{H}_{t-1}+{\gamma }_{14}lnC{O}_{2t-1}+{\epsilon }_t,\end{array}$$

(5)

$$\begin{array}{ll}\hfill \Delta lnPH{C}_t={\delta }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\delta }_{1i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{2i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{3i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{4i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{5i}\Delta lnIA{R}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{7i}\Delta lnC{O}_{2t-i}+{\delta }_{8}lnPH{C}_{t-1}+{\delta }_{9}lnIRI{N}_{t-1}+{\delta }_{10}lnR{D}_{t-1}+{\delta }_{11}lnAV{A}_{t-1}\\ & +{\delta }_{12}lnIA{R}_{t-1}+{\delta }_{13}lnMEC{H}_{t-1}+{\delta }_{14}lnC{O}_{2t-1}+{\epsilon }_t,\end{array}$$

(6)

$$\begin{array}{ll}\hfill \Delta lnIA{R}_t={\phi }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\phi }_{1i}\Delta lnIA{R}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{2i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{3i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{4i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{5i}\Delta lnAV{A}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{7i}\Delta lnC{O}_{2t-i}+{\phi }_{8}lnIA{R}_{t-1}+{\phi }_{9}lnPH{C}_{t-1}+{\phi }_{10}lnIRI{N}_{t-1}+{\phi }_{11}lnR{D}_{t-1}\\ & +{\phi }_{12}lnAV{A}_{t-1}+{\phi }_{13}lnMEC{H}_{t-1}+{\phi }_{14}lnC{O}_{2t-1}+{\epsilon }_t,\end{array}$$

(7)

$$\begin{array}{ll}\hfill \Delta lnMEC{H}_t={\psi }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\psi }_{1i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{2i}\Delta lnIA{R}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{3i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{4i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{5i}\Delta lnR{D}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{6i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{7i}\Delta lnC{O}_{2t-i}+{\psi }_{8}lnMEC{H}_{t-1}+{\psi }_{9}lnIA{R}_{t-1}+{\psi }_{10}lnPH{C}_{t-1}\\ & +{\psi }_{11}lnIRI{N}_{t-1}+{\psi }_{12}lnR{D}_{t-1}+{\psi }_{13}lnAV{A}_{t-1}+{\psi }_{14}lnC{O}_{2t-1}+{\epsilon }_t,\end{array}$$

(8)

$$\begin{array}{ll}\hfill \Delta lnC{O}_{2t}={\lambda }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\lambda }_{1i}\Delta lnC{O}_{2t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{2i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{3i}\Delta lnIA{R}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{4i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{5i}\Delta lnIRI{N}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{6i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{7i}\Delta lnAV{A}_{t-i}+{\lambda }_{8}lnC{O}_{2t-1}+{\lambda }_{9}lnMEC{H}_{t-1}+{\lambda }_{10}lnIA{R}_{t-1}+{\lambda }_{11}lnPH{C}_{t-1}\\ & +{\lambda }_{12}lnIRI{N}_{t-1}+{\lambda }_{13}lnR{D}_{t-1}+{\lambda }_{14}lnAV{A}_{t-1}+{\epsilon }_t,\end{array}$$

(9)

where $p$ signifies the lag length, $\mathsf{\Delta }$ denotes the various operator, and ${\epsilon }_t$ shows the error term. The null hypothesis of no long-term cointegration among the study variables in Equation (3) H0: ${\alpha }_8={\alpha }_9={\alpha }_{10}={\alpha }_{11}={\alpha }_{12}={\alpha }_{13}={\alpha }_{14}=0$ is verified against the alternative hypothesis H1: ${\alpha }_8\ne {\alpha }_9\ne {\alpha }_{10}\ne {\alpha }_{11}\ne {\alpha }_{12}\ne {\alpha }_{13}\ne {\alpha }_{14}\ne 0$. Similarly, in Equation (4) the null hypothesis H0: ${\beta }_8={\beta }_9={\beta }_{10}={\beta }_{11}={\beta }_{12}={\beta }_{13}={\beta }_{14}=0$ is verified against the alternative hypothesis H1: ${\beta }_8\ne {\beta }_9\ne {\beta }_{10}\ne {\beta }_{11}\ne {\beta }_{12}\ne {\beta }_{13}\ne {\beta }_{14}\ne 0$. In addition, in Equation (5) the null hypothesis H0: ${\gamma }_8={\gamma }_9={\gamma }_{10}={\gamma }_{11}={\gamma }_{12}={\gamma }_{13}={\gamma }_{14}=0$ is tested against the alternative hypothesis H1: ${\gamma }_8\ne {\gamma }_9\ne {\gamma }_{10}\ne {\gamma }_{11}\ne {\gamma }_{12}\ne {\gamma }_{13}\ne {\gamma }_{14}\ne 0$. Likewise, in Equation (6) the null hypothesis H0: ${\delta }_8={\delta }_9={\delta }_{10}={\delta }_{11}={\delta }_{12}={\delta }_{13}={\delta }_{14}=0$ is tested against the alternative hypothesis H1: ${\delta }_8\ne {\delta }_9\ne {\delta }_{10}\ne {\delta }_{11}\ne {\delta }_{12}\ne {\delta }_{13}\ne {\delta }_{14}\ne 0$. Furthermore, in Equation (7) the null hypothesis H0: ${\phi }_8={\phi }_9={\phi }_{10}={\phi }_{11}={\phi }_{12}={\phi }_{13}={\phi }_{14}=0$ is inspected against the alternative hypothesis H1: ${\phi }_8\ne {\phi }_9\ne {\phi }_{10}\ne {\phi }_{11}\ne {\phi }_{12}\ne {\phi }_{13}\ne {\phi }_{14}\ne 0$. Moreover, in Equation (8) the null hypothesis H0: ${\psi }_8={\psi }_9={\psi }_{10}={\psi }_{11}={\psi }_{12}={\psi }_{13}={\psi }_{14}=0$ is checked against the alternative hypothesis H1: ${\psi }_8\ne {\psi }_9\ne {\psi }_{10}\ne {\psi }_{11}\ne {\psi }_{12}\ne {\psi }_{13}\ne {\psi }_{14}\ne 0$. Additionally, in Equation (9) the null hypothesis H0: ${\lambda }_8={\lambda }_9={\lambda }_{10}={\lambda }_{11}={\lambda }_{12}={\lambda }_{13}={\lambda }_{14}=0$ is checked against the alternative hypothesis H1: ${\lambda }_8\ne {\lambda }_9\ne {\lambda }_{10}\ne {\lambda }_{11}\ne {\lambda }_{12}\ne {\lambda }_{13}\ne {\lambda }_{14}\ne 0$.

The null hypothesis H0 (no long-term co-integration between the study variables) could be rejected or accepted depending on the following conditions:

If the ARDL F-stat is ≥the value of the UCB, then the null hypothesis H0 is rejected and the study variables are co-integrated.

If the ARDL F-stat is ≤the value of LCB, then the null hypothesis H0 is accepted, however, if the ARDL F-stat is remains between both the UCB and the LCB then the result is inconclusive [53].

Once the long-term co-integration interaction among the research variables was established, the long- as well short-run elasticities of the research variables were estimated. The short-run analysis was formulated by the equations below:

$$\begin{array}{ll}\hfill \Delta lnAV{A}_t={\alpha }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\alpha }_{1i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{2i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{3i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{4i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{5i}\Delta lnIA{R}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\alpha }_{7i}\Delta lnC{O}_{2t-i}+{\theta }_{1}EC{T}_{t-1}+{\epsilon }_t,\end{array}$$

(10)

$$\begin{array}{ll}\hfill \Delta lnR{D}_t={\beta }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\beta }_{1i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{2i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{3i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{4i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{5i}\Delta lnIA{R}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\beta }_{7i}\Delta lnC{O}_{2t-i}+{\theta }_{2}EC{T}_{t-1}+{\epsilon }_t,\end{array}$$

(11)

$$\begin{array}{ll}\hfill \Delta lnIRI{N}_t={\gamma }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\gamma }_{1i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{2i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{3i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{4i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{5i}\Delta lnIA{R}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\gamma }_{7i}\Delta lnC{O}_{2t-i}+{\theta }_{3}EC{T}_{t-1}+{\epsilon }_t,\end{array}$$

(12)

$$\begin{array}{ll}\hfill \Delta lnPH{C}_t={\delta }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\delta }_{1i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{2i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{3i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{4i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{5i}\Delta lnIA{R}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\delta }_{7i}\Delta lnC{O}_{2t-i}+{\theta }_{4}EC{T}_{t-1}+{\epsilon }_t,\end{array}$$

(13)

$$\begin{array}{ll}\hfill \Delta lnIA{R}_t={\phi }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\phi }_{1i}\Delta lnIA{R}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{2i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{3i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{4i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{5i}\Delta lnAV{A}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{6i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\phi }_{7i}\Delta lnC{O}_{2t-i}+{\theta }_{5}EC{T}_{t-1}+{\epsilon }_t,\end{array}$$

(14)

$$\begin{array}{ll}\hfill \Delta lnMEC{H}_t={\psi }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\psi }_{1i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{2i}\Delta lnIA{R}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{3i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{4i}\Delta lnIRI{N}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{5i}\Delta lnR{D}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{6i}\Delta lnAV{A}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\psi }_{7i}\Delta lnC{O}_{2t-i}+{\theta }_{6}EC{T}_{t-1}+{\epsilon }_t,\end{array}$$

(15)

$$\begin{array}{ll}\hfill \Delta lnC{O}_{2t}={\lambda }_0+{\displaystyle {\displaystyle \sum }_{i=1}^p}& {\lambda }_{1i}\Delta lnC{O}_{2t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{2i}\Delta lnMEC{H}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{3i}\Delta lnIA{R}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{4i}\Delta lnPH{C}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{5i}\Delta lnIRI{N}_{t-i}\\ & +{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{6i}\Delta lnR{D}_{t-i}+{\displaystyle {\displaystyle \sum }_{i=1}^p}{\lambda }_{7i}\Delta lnAV{A}_{t-i}+{\theta }_{7}EC{T}_{t-1}+{\epsilon }_t.\end{array}$$

(16)

To evaluate the soundness of the estimated equations, we used various diagnostic tests, including the JBNT test, the BGSC test, and the ARCH test. In addition, the constancy of long- and short-run estimations was checked by the test of cumulative sum of squares (CUSUMSQ).

5. Outcomes and Discussion

5.1. Unit Root Outcomes

First, to test the stationary of the selected variables, we used the augmented Dickey–Fuller [60] and KPSS tests. The outcomes of both the ADF and the KPSS unit root testing are reported in

Table 1. Unit root outcomes.

Variables	ADF Statistic
	Level		$\Delta$
	$\tau$	$\tau \mathrm{and}\zeta$	$\tau$	$\tau \mathrm{and}\zeta$
LnAVA	−0.484316	0.134286 **	−1.981175	−4.887512 ***
LnRD	−0.120868	0.073065	−3.491348 **	−7.459985 ***
LnIRIN	−3.079714 **	0.073388	3.084590	−6.475249 ***
LnPHC	−1.928445	0.122525 *	−2.002726	−3.713071 **
LnIAR	0.042885	0.114068	−2.224613	−4.461708 ***
LnFC	−2.120654	0.183015 **	0.016106	−3.422516 *
LnMECH	−0.433694	0.182718 **	−2.412200	−3.594115 **
LnCO2	−0.611570	0.091016	−5.718108 ***	−9.881977 ***
KPSS Statistic
LnAVA	0.738800 **	0.134286 **	0.149350	0.157442 **
LnRD	0.721094 **	0.073065	0.180886	0.134135 *
LnIRIN	0.140297	0.073388	0.119252	0.119101 *
LnPHC	0.486199 **	0.122525 *	0.109638	0.106846
LnIAR	0.747263 ***	0.114068	0.107143	0.063000
LnFC	0.715913 **	0.183015 **	0.566202 **	0.085802
LnMECH	0.546279 **	0.182718 **	0.389839 *	0.119869 *
LnCO2	0.632433 **	0.091016	0.848046 ***	0.953457 ***

Note: LnAVA, LnRD, LnIRIN, LnPHC, LnIAR, LnFC, LnMECH, and LnCO₂ represent the natural log of agricultural value added, agricultural research and development expenditure, irrigation investment, physical capital, irrigated area, fertilizer used, level of mechanization, and carbon dioxide (CO₂) emissions, respectively. ∆ denotes the first difference, τ represents the trend, and ζ shows the intercept. *** 1%, ** 5%, and * 10% are the significance indicators.

, indicating that agricultural value added (lnAVA), agricultural research and development spending (lnRD), irrigation investment (lnIRIN), gross capital formation (lnPHC), and level of mechanization (lnMECH) are stationary at the level while irrigated area (lnIRA) and CO₂ emissions (lnCO₂) are not stationary at the level I(0). However, both variables become stationary at first difference I(1), so we can employ the ARDL method to explore the long-term interrelationship among agricultural research and development spending, irrigation investment, gross capital formation, irrigated area, fertilizer consumption, level of mechanization, CO₂ emissions, and agricultural productivity in China during the period of 1986 to 2017.

5.2. ARDL Bound-Testing and Johansen Cointegration Results

Next, after checking that selected variables are I(0) and I(1), for the ARDL-bounds method, Equations (3)–(9) can be employed to scrutinize the long-run symmetry interrelationship among the selected variables of the study.

Table 2. ARDL-bounds test results.

Estimated Models	Opt. Lags	F-Stat.
FLnAVA (LnAVA|LnRD, LnIRIN, LnPHC, LnIAR, LnFC, LnMECH, LnCO2)	ARDL (1, 2, 2, 1, 2, 0, 1, 1)	4.92 ***
FLnRD (LnRD/LnAVA, LnIRIN, LnPHC, LnIAR, LnFC, LnMECH, LnCO2)	ARDL(1, 0, 0, 1, 2, 2, 2, 0)	4.21 **
FLnIRIN (LnIRIN|LnRD, LnAVA, LnPHC, LnIAR, LnFC, LnMECH, LnCO2)	ARDL(1, 1, 2, 1, 2, 2, 2, 2)	21.42 ***
FLnPHC (LnPHC|LnIRIN, LnRD, LnAVA, LnIAR, LnFC, LnMECH, LnCO2)	ARDL(1, 1, 2, 2, 2, 2, 2, 2)	16.26 ***
FLnIAR (LnIAR|LnPHC, LnIRIN, LnRD, LnAVA, LnFC, LnMECH, LnCO2)	ARDL(1, 0, 1, 1, 2, 0, 0, 0)	7.10 ***
FLnFC (LnFC|LnIAR, LnPHC, LnIRIN, LnRD, LnAVA, LnMECH, LnCO2)	ARDL(1, 0, 2, 0, 2, 0, 1, 0)	2.27
FLnMECH (LnMECH|LnFC, LnIAR, LnPHC, LnIRIN, LnRD, LnAVA, LnCO2)	ARDL(1, 1, 0, 1, 0, 0, 2, 0)	3.55 *
FLnCO2 (LnCO2|LnMECH, LnFC, LnIAR, LnPHC, LnIRIN, LnRD, LnAVA)	ARDL(1, 0, 2, 1, 0, 2, 2, 2)	5.19 ***
Significance	I(0) Bound	I(1) Bound
1%	3.31	4.63
5%	2.69	3.83
10%	2.38	3.45

Note: LnAVA, LnRD, LnIRIN, LnPHC, LnIAR, LnFC, LnMECH, and LnCO₂ signify the natural log of agricultural value added, agricultural research and development expenditure, irrigation investment, physical capital, irrigated area, fertilizer used, level of mechanization, and carbon dioxide (CO₂) emissions, respectively. *** 1%,** 5%, and * 10% are the significance indicators.

presents the estimated results of the F-statistic of the ARDL-bounds test for Equations (3)–(9), indicating that there are long-run interrelationships amongst agricultural research and development spending, irrigation investment, gross capital formation, irrigated area, fertilizer consumption, level of mechanization, CO₂ emissions, and agricultural production in China from 1986 to 2017. For instance, the computed F-statistic for the model of FLnAVA (LnAVA|LnRD, LnIRIN, LnPHC, LnIAR, LnFC, LnMECH, LnCO₂) is 4.92, and above the UBC values at 1%. Likewise, in the second model of FLnRD (LnRD/LnAVA, LnIRIN, LnPHC, LnIAR, LnFC, LnMECH, LnCO₂), the estimated statistic is 4.21 and lies above the UCB values of 5%. According to the third, fourth, and fifth models of FLnIRIN (LnIRIN|LnRD, LnAVA, LnPHC, LnIAR, LnFC, LnMECH, LnCO₂), FLnPHC (LnPHC|LnIRIN, LnRD, LnAVA, LnIAR, LnFC, LnMECH, LnCO₂), and FLnIAR (LnIAR|LnPHC, LnIRIN, LnRD, LnAVA, LnFC, LnMECH, LnCO₂), the estimated statistics are 21.42, 16.26, and 7.10, respectively, and lie above the UCB values at the 1% significance level. Furthermore, the computed F-statistics for the models of FLnMECH (LnMECH|LnFC, LnIAR, LnPHC, LnIRIN, LnRD, LnAVA, LnCO₂) and FLnCO₂ (LnCO₂|LnMECH, LnFC, LnIAR, LnPHC, LnIRIN, LnRD, LnAVA) are 3.55 and 5.19, respectively, and are above the UBC values at 5% and 1%. Moreover, the outcomes of several diagnostic tests are reported in

Table 3. Diagnostic tests for ARDL.

Estimated Models	χ2SERIAL	χ2NORMAL	χ2ARCH	χ2RESET
FLnAVA (LnAVA|LnRD, LnIRIN, LnPHC, LnIAR, LnFC, LnMECH, LnCO2)	2.3322	1.8012	0.3384	0.0320
	(0.1526)	(0.7210)	(0.7159)	(0.9685)
FLnRD (LnRD/LnAVA, LnIRIN, LnPHC, LnIAR, LnFC, LnMECH, LnCO2)	3.1231	1.1199	1.4159	2.3854
	(0.1103)	(0.5712)	(0.2437)	(0.0317)
FLnIRIN (LnIRIN|LnRD, LnAVA, LnPHC, LnIAR, LnFC, LnMECH, LnCO2)	2.7855	5.6284	0.6835	0.8903
	(0.1194)	(0.1599)	(0.4151)	(0.3964)
FLnPHC (LnPHC|LnIRIN, LnRD, LnAVA, LnIAR, LnFC, LnMECH, LnCO2)	0.7718	1.7409	1.3270	1.1466
	(0.4978)	(0.4187)	(0.2587)	(0.2847)
FLnIAR (LnIAR|LnPHC, LnIRIN, LnRD, LnAVA, LnFC, LnMECH, LnCO2)	1.9709	0.1126	1.1546	2.3653
	(0.1717)	(0.9452)	(0.3637)	(0.1259)
FLnFC (LnFC|LnIAR, LnPHC, LnIRIN, LnRD, LnAVA, LnMECH, LnCO2)	1.2817	0.2564	0.4976	1.5105
	(0.3063)	(0.8796)	(0.4861)	(0.1504)
FLnMECH (LnMECH|LnFC, LnIAR, LnPHC, LnIRIN, LnRD, LnAVA, LnCO2)	2.6397	5.4319	2.1176	0.1143
	(0.1213)	(0.0661)	(0.1112)	(0.9103)
FLnCO2 (LnCO2|LnMECH, LnFC, LnIAR, LnPHC, LnIRIN, LnRD, LnAVA)	2.5635	2.3025	0.1827	2.8148
	(0.1219)	(0.3162)	(0.6722)	(0.1030)

Note: LnAVA, LnRD, LnIRIN, LnPHC, LnIAR, LnFC, LnMECH, and LnCO₂ indicate the natural log of agricultural value added, agricultural research and development expenditure, irrigation investment, physical capital, irrigated area, fertilizer used, level of mechanization, and carbon dioxide (CO₂) emissions, respectively.

. The study used Jarque–Bera (χ²NORMAL), serial correlation (χ²SERIAL), heteroscedasticity (χ²ARCH) and functional form (χ²RESET) tests for the models.

In addition,

Table 4. Results of the Johansen cointegration method.

Hypothesized		Trace	Critical
No. of CE(s)	Eigen Values	Statistics	Value	Probs.
None	0.956043	310.2850 ***	187.4701	0.0000
At most 1	0.867916	216.5484 ***	150.5585	0.0000
2	0.761807	155.8189 ***	117.7082	0.0000
3	0.731065	112.7787 ***	88.80380	0.0003
4	0.627933	73.38014 ***	63.87610	0.0064
5	0.521083	43.71967 **	42.91525	0.0414
6	0.352501	21.63285	25.87211	0.1541
7	0.249081	8.593710	12.51798	0.2069
Maximum Eigenvalue
None	0.956043	93.73652 ***	56.70519	0.0000
At most 1	0.867916	60.72951 ***	50.59985	0.0033
2	0.761807	43.04023 *	44.49720	0.0714
3	0.731065	39.39856 **	38.33101	0.0376
4	0.627933	29.66047 *	32.11832	0.0969
5	0.521083	22.08682	25.82321	0.1444
6	0.352501	13.03914	19.38704	0.3249
7	0.249081	8.593710	12.51798	0.2069

Note: *** 1%, ** 5%, and * 10% are the significance indicators.

reports the estimated results of Trace and Max-Eigen statistics. The Johansen cointegration approach outcomes validate a long-run cointegration linkage between agricultural research and development spending, irrigation investment, gross capital formation, irrigated area, fertilizer consumption, level of mechanization, CO₂ emissions, and agricultural output, which confirm the outcomes of the ARDL-bounds technique.

5.3. Long-Run and Short-Run Estimations

The present study engaged the ARDL method to find the long and short run interrelationships amongst the selected variables. The observed effects of the long- and short term of the ARDL model are demonstrated in

Table 5. Results of the ARDL method.

Dependent Variable: LnAVA; ARDL(1, 2, 2, 1, 2, 0, 1, 1); Akaike Info Criterion (AIC)
Long run estimate
Variables	Coef.	SE	t-stat	Probs.
LnRD	0.293681 ***	0.099256	2.958838	0.0111
LnIRIN	0.033690	0.027297	1.234201	0.2390
LnPHC	0.940082 ***	0.277009	3.393690	0.0048
LnIAR	−1.909159 *	1.064175	−1.794027	0.0961
LnFC	0.915927 **	0.444282	2.061589	0.0598
LnMECH	0.283658 *	0.147067	1.928765	0.0759
LnCO2	−1.029829 ***	0.251161	−4.100279	0.0013
Constant	1.588575	4.586389	0.346367	0.7346
Short-run dynamic
ΔLnAVA(−1)	0.030371	0.216681	0.140166	0.8907
ΔLnRD	−0.072848	0.065320	−1.115245	0.2849
ΔLnRD(−1)	0.220340 ***	0.066243	3.326229	0.0055
ΔLnRD(−2)	0.137269 **	0.062338	2.202005	0.0463
ΔLnIRIN	0.078054 ***	0.026428	2.953486	0.0112
ΔLnIRIN(−1)	−0.018239	0.023714	−0.769126	0.4556
ΔLnIRIN(−2)	−0.027147	0.015346	−1.769083	0.1003
ΔLnPHC	0.417243	0.291362	1.432042	0.1757
ΔLnPHC(−1)	0.494288 **	0.236850	2.086918	0.0572
ΔLnIAR	1.504587	0.942086	1.597080	0.1343
ΔLnIAR(−1)	−0.811116	1.154053	−0.702841	0.4945
ΔLnIAR(−2)	−2.544647 **	1.229052	−2.070415	0.0589
ΔLnFC	0.888109 **	0.394035	2.253881	0.0421
ΔLnMECH	0.125511	0.119220	1.052768	0.3116
ΔLnMECH(−1)	0.149532	0.094013	1.590540	0.1357
ΔLnCO2	−0.379810	0.364030	−1.043347	0.3158
ΔLnCO2(−1)	−0.618742 *	0.339683	−1.821530	0.0916
CointEq(−1)	−0.969629 ***	0.216681	−4.474920	0.0006
Diagnostic tests
R2	0.998663
Adj-R2	0.996811
F-stat	539.3900
Prob(F-stat)	0.000000
χ2SERIAL	2.5759 (0.1209)
χ2NORMAL	1.1204 (0.5710)
χ2ARCH	0.0989 (0.7553)
χ2RESET	0.0609 (0.8091)
CUSUM	Stable
CUSUMSQ	Stable

Note: *** 1%, ** 5%, and * 10% are the significance indicators.

. Figure 4 demonstrates the long-run connection amid the variables.

The estimated coefficients of public research and development spending are positive and substantial at 1%, demonstrating that a 1% upsurge in public research and development spending significantly increase agricultural output in China by 0.293% in the long and short run (see Table 5). Similarly, the coefficient of irrigation investment is positive but statistically insignificant in the long term while in the short run, the coefficient of irrigation investment is progressive and substantial at the 1% significance level. It illustrates that a 1% surge in irrigation investment indicates a 0.033% increase in agricultural yield in the long and short term (see Table 5). The results of this work are similar to the previous work of [22,25,26,32,48,49,61,62,63,64].

The estimated coefficients of physical capital are positive and substantial at 1% and 5%; this shows a 1% growth in the physical capital increase of agricultural output by 0.940% and 0.494% in both periods, respectively. This finding is consistent with the work of [54,65], who found that capital has a positive significant influence on agricultural output.

Similarly, estimated coefficients of fertilizer used are positive and noteworthy at 5%, showing that a 1% growth in fertilizer used enhances agricultural output by 0.915% and 0.888% in both periods, respectively. In addition, the long-run influence of level of mechanization on agricultural output is positive and major at 10% significance; this means a 1% rise is likely to boost agricultural output by about 0.283%. This is consistent with the previous studies of [66,67,68,69,70]. Ahmad [66] reported that level of mechanization has a positive and noteworthy effect on agricultural productivity in Pakistan.

The estimated coefficients of CO₂ emissions are negative and significant at 1% and 10%. These results suggest that CO₂ emissions reduce agricultural production. A 1% growth in the level of CO₂ emissions decreases agricultural production in China by 1.029% in both periods (see Table 5). In China, GHG emissions are the prime reason for climate change as well as other human actions such as urbanization, transportation, and non-renewable energy consumption; these are the basic sources that contribute considerably towards CO₂ emissions in China. This effect is similar to the results of [65,71], who revealed that CO₂ emissions have a negative impact on agricultural output. In contrast, Chandio et al. [72] revealed that CO₂ positively affected rice yield in the long and short run in the case of Pakistan. Likewise, using ARDL modelling, Ahmad et al. (2020) found that CO₂ has adverse effects on agricultural growth.

This study applied both the CUSUM and the CUSUMsq tests to check the stability of the ARDL model. The graphical depictions of both tests are shown in Figure 5 and Figure 6. The plots of both stability tests lie within the acceptance bound; this means the estimated parameters of the model are stable.

6. Conclusions

By using the ARDL-bounds testing of co-integration and the Johansen co-integration procedures, the current study inspected the long-term association among agricultural public expenditures and agricultural output in the case of China over the period of 1986 to 2017. Agricultural productivity was used as dependent variable while agricultural public spending, including agricultural research and development and irrigation, physical capital, irrigated area, fertilizer consumption, level of mechanization, and CO₂ emissions have been incorporated as independent variables. The study checked the stationarity of the selected variables through ADF and the KPSS unit root testing. The results of both the ARDL and the Johansen co-integration techniques demonstrated a long-run connection between considered variables. The estimated regression results showed that agricultural public expenditures, such as on agricultural research and development and irrigation, had a positive and substantial impact on agricultural productivity in both periods. Otherwise, physical capital and fertilizer consumption had progressive and substantial influences on agricultural productivity in the short and long run. In addition, in both cases, CO₂ emissions negatively and significantly affected agricultural productivity in China. Sustained agricultural development is an important goal for every country, and public funds for irrigation and agricultural research and development should increase. The Chinese government should give this top priority as it would significantly enhance agricultural productivity in China. This study found that CO₂ emissions had an adverse influence on agricultural productivity in both periods. This means that carbon concentration was higher than the desired levels in China. Hence, this paper recommends that the Chinese government launch a climate/carbon financing scheme through financial institutions and allocate funds to farmers to counter the impact of CO₂ emissions on agricultural productivity in the short and long run. In addition, more planting is also necessary, as it can reduce the adverse effects of CO₂ emissions.

This study has a limitation which can be explored by future researchers. Employing ARDL-ECM modelling, this study examined the long-term impact of public agricultural spending along with other important factors, including physical capital, irrigated area, fertilizer consumption, level of mechanization, and CO₂ emissions on agricultural productivity in China. However, there could be several other factors such as agricultural credit, improved seeds, agricultural extension services, rural roads, and rural education that might potentially impact agricultural output. Therefore, future studies should consider these factors and assess the effect of public research and development spending and climate change (i.e., extreme temperatures and low precipitation) on grain crop yields by using panel data on monthly, quarterly and annual bases.