Examining the Threshold Effect of Environmental Regulation: The Impact of Agricultural Product Trade Openness on Agricultural Carbon Emissions

Abstract

Agricultural product trade openness is a crucial factor impacting agricultural carbon emissions. This study analyzes the relationship between trade openness and carbon emissions in China. Using panel data from 2002 to 2020 for 31 provinces, a threshold panel model is employed to examine the non-linear relationship from an environmental regulation perspective. The findings reveal a significant single-threshold effect of environmental regulation. When environmental regulation intensity is low, trade openness is positively associated with agricultural carbon emissions. However, when environmental regulation intensity is high, trade openness significantly reduces carbon emissions. Regional regression analysis indicates that this single-threshold effect holds true in major grain-producing, grain-selling, and grain- producing- selling balance areas, although regional differences exist. This study emphasizes the need to coordinate environmental regulations and trade policies, strengthen environmental control, and leverage the threshold effect of environmental regulation in reducing carbon emissions through agricultural product trade openness.

1. Introduction

In recent years, the issue of climate warming has attracted global attention. According to the IPCC’s latest report, “Climate Change 2022: Mitigating Climate Change” released in April 2022, GHG emissions have now hit a record high, and the concentrations of carbon dioxide, methane, and nitrous oxide, the planet’s main GHG emissions, continue to increase. If effective actions are not taken immediately, irreversible global ecological disasters and huge economic losses will result. Agriculture is an important source of carbon emissions [1,2]. The IPCC report demonstrates that the agriculture, forestry, and other land use (AFOLU) sector accounts for 23% of global GHG emissions, totaling 12 billion carbon-equivalent tons per year (https://www.ipcc.ch/srccl/ (accessed on 4 April 2022)). Moreover, agriculture will be the largest threat to carbon emissions in 2050 due to the complexity of its sources [3]. Therefore, reducing agricultural carbon emissions has been an important challenge for the world. With China being a country with a large population, its agriculture has always been important, and the country has a high dependence on it. Agricultural carbon emissions have become one of the main sources of carbon emissions in China, causing great pressure for China to reduce these agricultural emissions [4]. In September 2020, China put forward the goal of reaching a peak for its carbon dioxide emissions by 2030 and striving to achieve carbon neutrality by 2060. Under the “double carbon” goal, it has become urgent to reduce China’s agricultural carbon emissions.

There are many factors affecting agricultural carbon emissions. The agricultural product trade openness has been widely regarded as one of the important factors [5,6,7]. To ensure food security in China, it is essential to utilize international resources and foreign markets. With the development of globalization, China has become more integrated into the global agricultural market. Since joining the WTO, the average tariff level for agricultural products in China has decreased to 15.2%, which is about a quarter of the world average. As shown in Figure 1, China’s agricultural trade has grown dramatically, from USD 28 billion to USD 304.2 billion in 2002 and 2021, with an average annual growth rate of 13%. China has become a major agricultural trading country, ranking the second in the world. Meanwhile, the openness of agricultural product trade has also improved a lot, -from 12.1% to 23.8% in 2002 and 2021. Can agricultural product trade openness reduce China’s agricultural carbon emissions? Is this impact influenced by other factors? Based on the perspective of environmental regulation, this paper discusses the impact of agricultural product trade openness on agricultural carbon emissions, which has very important research value. When exploring the relationship between trade and carbon emissions, the existing literature often ignores the role of institutional factors such as environmental regulation and the possible nonlinear characteristics of the effect of trade on carbon emissions, which may lead to biased conclusions. Therefore, this paper attempts to examine the threshold effect of agricultural product trade openness on agricultural carbon emissions using a threshold panel model with environmental regulation as the threshold variable.

Many scholars have studied the impact of international trade on environmental pollution. Based on these studies, some research has begun to further explore the impact of agricultural trade on carbon emissions. Due to the complexity of agricultural carbon emission sources [8] and the special nature of China’s long-standing agricultural trade deficit, there is still much work to be done to delve into the impact of agricultural product trade on agricultural carbon emissions in China.

The present research findings have proven that international trade can have an impact on the environment through scale effects, structure effects, and technological effects [9,10,11]. In terms of scale effects, agricultural product trade causes changes in carbon emissions by changing agricultural production scale. Export and import may have opposite impact on carbon emissions. Generally speaking, the export of agricultural products can lead to an increase in agricultural production activities and the input of polluting factors such as fertilizers, which will contribute to agricultural carbon emissions. Agricultural imports, on the other hand, help to reduce carbon emissions caused by the overuse of polluting factors, with an offsetting effect brought by agricultural exports. China’s agricultural product trade has long been in a state of trade deficit, so it is possible that the agricultural carbon emissions generated by agricultural exports are smaller than those reduced by agricultural imports, which means that in terms of scale effect, China’s agricultural product trade deficit may be conducive to reducing agricultural carbon emissions.

In terms of structure effect, agricultural product trade may change the structure and resource allocation in agricultural industry, which can change the level of carbon emissions since different subsectors may vary in emission intensities. First, China’s main export destination countries of agricultural products include developed countries such as Japan, South Korea, and the U.S. When participating in international competition, enterprises exporting agricultural products tend to improve their product standards by scientifically adjusting input factors in order to meet the requirements of developed countries in terms of food safety and green health [12], which forces resources to flow to relatively cleaner sectors. Second, the import structure of agricultural products in China is also undergoing changes. In 2022, the proportion of livestock product imports in agricultural imports was as high as 21.89%. According to the data published by the Food and Agriculture Organization of the United Nations (FAO), ruminant livestock intestinal fermentation and livestock manure account for 32% and 7% of the agricultural carbon emission sources, respectively. Therefore, the increasing imports of livestock products can reallocate China’s agricultural resources, which is conducive to reducing agricultural carbon emissions.

In terms of technology effect, agricultural product trade may improve carbon emission efficiency through technology spillover. China’s agricultural trade is active in the international market, and developed countries such as Japan, South Korea, and the U.S are the main trading partners. China can often exchange experiences with its trading partners concerning new agricultural varieties, new agricultural technologies, and new agricultural production management systems, and learn the advanced ideas and experiences of its trading partners so that China can promote the low-carbon transformation of the domestic agricultural industry and develop green agriculture. At the same time, encountering green trade barriers (GTBs) in the exports of agricultural products may also force Chinese enterprises to develop environmentally friendly technologies, thereby reducing agricultural carbon emissions.

Can agricultural product trade conclusively reduce agricultural carbon emissions? Obviously, the answer is uncertain. Many scholars have come to different conclusions about this question [13,14,15]. The reason is that the carbon emission reduction effect of agricultural product trade may be restricted by many factors. Environmental regulation, as an important policy to restrain the pollution behavior, is one of the important constraints. In recent years, China’s agricultural pollution has become an increasingly significant problem, which has brought the construction of related environmental regulations into a substantial stage [16]. China has successively revised and issued some environmental regulations and policies. “The Environmental Protection Law”, revised in 2014, added provisions for the prevention and control of agricultural pollution. “The Regulations on Prevention and Control of Pollution from Large-scale Livestock and Poultry Breeding”, which came into effect in 2014, was the first specific law and regulation on agricultural environmental protection in China. In May 2022, the Ministry of Agriculture and Rural Affairs of the People’s Republic of China and National Development and Reform Commission jointly issued the “Implementation Plan for Emission Reduction and Carbon Sequestration in Agriculture and Rural Areas”. China’s environmental regulations and policies have been continuously improved. Can the continuous improvement in environmental regulations further promote the carbon emission reduction effect of agricultural product trade on agricultural carbon emissions? The answer to this question is very important for enhancing China’s green agriculture development. At the same time, it is worth noticing that there is significant heterogeneity in the intensity of environmental regulations in different provinces due to the great differences in economic development and the environmental pressure faced by various provinces in China. On the one hand, in order to develop agricultural trade and increase farmers’ income, some provinces may adopt loose environmental regulations to reduce agricultural production costs, increase agricultural exports, and attract the transfer of the production and processing of pollution-intensive products to the region, creating a “pollution paradise” effect. On the other hand, some provinces with more environmental pressure may adopt strict environmental regulations to force agricultural producers to develop green technologies and reduce polluting-factor inputs, which makes it easier for agricultural product trade to achieve agricultural carbon emission reduction through the technology effect.

Overall, agricultural carbon emissions have become one of the main sources of carbon emissions in China, causing enormous pressure for China to reduce agricultural carbon emissions. As trade liberalization progresses, agricultural product trade openness is widely regarded as playing an important role in affecting agricultural carbon emissions. Environmental regulations, as an important policy tool to constrain pollution behavior, may bring about different conclusions for how agricultural product trade openness affects agricultural carbon emissions. Therefore, it is necessary to examine the effects of agricultural product trade openness on agricultural carbon emissions from the perspective of environmental regulation. Clarifying the relationship between agricultural product trade openness, environmental regulations and agricultural carbon emissions is important for developing a green and low-carbon agricultural industry and exploring effective carbon emission reduction paths in agriculture.

The marginal contributions of this paper are mainly in the following aspects:

First, most of the existing studies have focused on the impact of international trade on industrial pollution emissions. Ever since the proposal of China’s “double carbon” goal, agricultural carbon emission reduction has been under great pressure. Based on this consideration, this paper provides a comprehensive estimation of the impact of agricultural product trade openness on agricultural carbon emissions, enriching the existing research.

Second, when examining the impact of international trade on carbon emissions, prior studies have mainly failed to take into account the non-linear relationship between variables and have not reached consistent conclusions. We used the threshold panel model to determine the non-linear relationship between agricultural product trade openness and agricultural carbon emissions through the perspective of environmental regulation, which avoided the possible bias in the estimation results caused by ignoring the impact of institutional factors, thus adding new evidence to the current debate.

Third, this paper compares and analyzes the differences in the environmental regulation threshold effects among the three functional grain areas in China, namely, the major grain-producing area, the major grain-selling area, and the grain producing-selling balance area, and proposes effective agricultural emission reduction policies for the different functional areas.

The remainder of this paper is structured as follows: Section 2 reviews relevant domestic and foreign literature; Section 3 describes the model setting, variable selection, and data source; Section 4 details the empirical test and result analysis, including the threshold effect test, threshold value estimation, parameter estimation, and robustness test; and Section 5 is the conclusion and countermeasure suggestion part of this paper.

2. Literature Review

Since the 1990s, research on trade and the environment has received much attention from domestic and foreign scholars. Grossman and Krueger (1991) first studied this issue. When estimating the environmental Kuznets curve (EKC), they proposed a three-effect model of the impact of international trade on the environment, i.e., the scale effect, the structure effect, and the technology effect, which was the main source for the subsequent theoretical development on this issue [9]. With the increasingly serious problem of global warming, many scholars have started to pay attention to the issue of international trade and carbon emissions. The present research findings about the impact of international trade on carbon emissions can be divided into three categories:

First, some scholars believe that international trade can significantly reduce carbon emissions. Guo and Liu (2013) argued that a country’s import trade tends to produce technology spillover effects on the host country in an open economy, which could reduce carbon emissions. The empirical results also showed that for each increase of 1% in the technology spillover stock of import trade, carbon emissions will be reduced by 0.513% accordingly [17]. Based on provincial panel data in China, Jiao et al. (2017) analyzed the technology spillover effect of import trade, with the results showing that import trade could significantly reduce the carbon emission intensity in China [18]. Pan (2017) constructed an environmental input output model from the perspective of agricultural product trade to test the carbon emission reduction effect of trade, and the study shows that agricultural product trade is beneficial for reducing the level of domestic carbon emissions [6]. Chen et al. (2022) used a spatial panel model to analyze the impact of agricultural product trade on agricultural carbon emissions in China and its spatial spillover effects; the results showed that agricultural product trade can significantly reduce agricultural carbon emissions, and it is more significant in the central and western regions, major grain-selling areas, and trade-deficit areas [15].

Second, some other scholars believe that the “pollution paradise hypothesis” (PPH) is valid, suggesting that international trade will increase carbon emissions. Sun et al. (2015) used the three-effect hypothesis as a theoretical basis and conducted an empirical test using provincial panel data in China, showing that trade openness generally increases carbon emissions [19]. Using a GVAR model from 33 countries from 1980 to 2010, Wang and Xu (2015) studied the dynamic relationship between trade openness, economic growth, and China’s carbon dioxide emissions, believing that China’s trade openness has increased China’s carbon dioxide emissions [20]. Ma and Shi (2016) studied the EKC curve of trade openness and carbon emissions, and their findings showed that trade openness has a positive effect on the per capita carbon emissions in China, with each 1% increase in trade openness increasing the per capita carbon emissions by 0.3% [21]. Yang and Qi (2021) used the panel data of 48 countries along the Belt and Road to verify the relationship between agricultural product trade and carbon emissions using a stochastic frontier gravity model, with the results showing that a higher openness of agricultural product trade in countries along the Belt and Road corresponded with more carbon emissions [13]. Using carbon emission data for France from 1980 to 2020, Omri and Saadaoui (2022) concluded that trade openness increases carbon emissions, and they believe that France should reduce fossil fuel imports in order to reduce carbon emissions [22].

The third category supports the “uncertainty theory”, which holds that the impact of international trade on carbon emissions can be non-linear. Gao and Chen (2014) explored the relationship between agricultural trade openness and agricultural carbon emission performance using the EBM model. The study calculated the agricultural carbon emissions of Chinese provinces and concluded that the import dependence ratio of agricultural products is positively related to the agricultural carbon emission performance, while the export dependence ratio of agricultural products has the reverse effect [23]. Based on the data from oil-exporting developing countries, Hasanov et al. (2018) examined the effects of exports and imports on carbon dioxide emissions separately and found that exports and imports have opposite effects on carbon dioxide emissions [24]. Zhang and Sun (2019) built an “international trade carbon emission” matrix based on international input output data and used a network analysis to capture the relationship between international trade and carbon emissions among countries. The results showed that developed economies transfer carbon emissions by importing large amounts of intermediate products, while BRICS and emerging economies receive carbon input from developed countries [25]. Appiah et al. (2022) tested the impact of international trade and carbon emissions using data from emerging economies. The empirical results showed that a 1% increase in imports was associated with a 0.471% increase in carbon emissions, while exports can improve the environment of emerging economies, but the result was not significant [26]. Ali et al. (2021) examined the extent to which trade openness influence environmental quality using the data of OIC countries. They confirmed that inverted U-shaped EKC curve exists in all groups of OIC countries [27]. Hu et al. (2021) investigated the impact of China’s agricultural trade shocks on carbon emissions of countries along the Belt and Road. The result showed that China’s agricultural trade supply shocks have a significant inhibitory effect on the intensity and share of agricultural carbon emissions in B&R economies, while the impact of demand shocks is not significant [28]. The effect of international trade on carbon emissions is constrained by multiple factors, which not only affect the size of the carbon emission reduction effect, but also may change the direction of this effect. Many scholars have studied this from different perspectives [29,30]. Some of them have also tried to conduct an analysis of the interactions among international trade, environmental regulation and environmental pollution, believing that environmental regulation has a threshold effect on carbon emissions from trade. On the one hand, environmental regulations can affect the direction and cleanliness of industry choices in international trade. When the intensity of regional environmental regulation is low, thus the cost of carbon emissions is accordingly low, the export of high polluting products has a distinct cost advantage. With the increase in environmental regulation intensity, the pollution content of major trade products decreases, and the scale of trade in clean products expands, which is likely for trade to have carbon emission reduction effects. On the other hand, environmental regulations can also affect the emission reduction efficiency of trading enterprises. The strict environmental regulations will force trading enterprises to improve green technology innovation by increasing independent R&D investment or utilizing technology spillovers from international trade, in order to achieve emission reduction goals. Zhong (2016) holds that there is a long-term stable equilibrium relationship between environmental regulation, foreign trade and carbon emissions [31]. He (2016) believes that the imposition of carbon tariffs could change the structure of export products and reduce the embodied carbon of export trade to a certain extent [32]. Zhan (2017) argued that there is a threshold effect of environmental regulation intensity on the impact of China’s foreign trade openness on carbon emissions, and that a moderate environmental regulation intensity should be adopted in order to reduce the carbon emissions through trade [33]. Huang (2020) found that in the global value chain (GVC), an increase in the level of trade liberalization of intermediate goods is conducive to reducing industrial pollution emissions when the environmental regulation intensity increases [34]. By using data on the chemical oxygen demand (COD) and major atmospheric pollutants such as sulfur dioxide, smoke, and dust, Wang et al. (2022) argued that with the growing implementation of environmental regulations, Chinese export enterprises are effectively improving their production processes and promoting the upgrading of the export trade structure, which has accordingly reduced the environmental pollution caused by exports [35].

To summarize, although there have been relatively rich studies about the impact of international trade on carbon emissions, most of them have focused on studying of the impact of international trade on industrial carbon emissions, and relatively little attention has been paid to how agricultural product trade affects agricultural carbon emissions. Moreover, in empirical studies, scholars have not reached consistent conclusions about whether international trade can reduce carbon emissions due to differences in research perspectives and research methods, etc. The failure to fully consider the influencing factors that may affect the effect of international trade on carbon emissions is also one of the reasons why the existing literature has failed to reach consistent conclusions. Many studies have failed to consider the non-linear relationship between variables and have ignored the influence of institutional factors, so the estimation results may be biased. Environmental regulation, as an important institutional factor to constrain pollution, has a non-negligible impact on how international trade affects the environmental pollution in developing countries [35]. However, few scholars have studied the impact of international trade on carbon emissions from the perspective of environmental regulation, which makes it necessary to further explored this topic.

3. Methodology and Data

3.1. Model Construction

From the analysis above, it can be seen that the relationship between agricultural product trade openness and agricultural carbon emissions may not be a simple linear relationship due to the influence of environmental regulation and other factors. There might be a threshold effect of environmental regulation in the impact of agricultural product trade openness on agricultural carbon emissions. The essence of the threshold model is to divide the total sample data into groups using the threshold values, and the threshold regression model is only needed when the estimated parameters of different groups of samples are significantly different. In this paper, based on the threshold model proposed by Hansen (1999) [36], the samples were endogenously grouped, which avoided the disadvantage of manual grouping. The traditional threshold model often uses manual grouping method, which may have a certain degree of subjectivity in the grouping criteria, and cannot specifically estimate the threshold value, nor can it perform significance testing on the differences in results between different sample groups. The threshold model proposed by Hansen (1999) [36] introduces the threshold value as an unknown variable into the model to construct a segmented function, endogenously grouping the samples. The threshold value and its number are entirely determined endogenously by the sample data. This method provides an asymptotic distribution theory to establish the confidence interval of the parameter to be estimated, and the bootstrap method can also be used to estimate the statistical significance of the threshold value, making up for the shortcomings of manual grouping methods. Hansen’s (1999) [36] panel threshold model has been widely used in empirical research. The basic equation is as follows:

$${\mathrm{Y}}_{\mathrm{it}}={\mathrm{u}}_{\mathrm{i}}+{\mathsf{\beta }}_1^{\prime }{\mathrm{X}}_{\mathrm{it}}\mathrm{I}\left({\mathrm{q}}_{\mathrm{it}}\le \mathsf{\gamma }\right)+{\mathsf{\beta }}_2^{\prime }{\mathrm{X}}_{\mathrm{it}}\mathrm{I}\left({\mathrm{q}}_{\mathrm{it}}>\mathsf{\gamma }\right)+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(1)

where i denotes the province, t denotes the year, q_it denotes the threshold variable, γ is the threshold value to be estimated, and ε_it is the random error term. The individual intercept term u_i denotes the fixed effect and I(∙) is the indicative function. If the expression in parentheses is true, it takes the value of 1 otherwise, it takes the value of 0.

By drawing on Equation (1) and referring to the existing literature, this study constructed the threshold panel model of agricultural product trade openness affecting agricultural carbon emissions with environmental regulation as the threshold variable:

$$\begin{array}{ll}{\mathrm{CARBON}}_{\mathrm{it}}={\mathrm{u}}_{\mathrm{i}}& + {\mathsf{\beta }}_1{\mathrm{TRADE}}_{\mathrm{it}}\mathrm{I}\left({\mathrm{ER}}_{\mathrm{it}}\le \mathsf{\gamma }\right)+{\mathsf{\beta }}_2{\mathrm{TRADE}}_{\mathrm{it}}\mathrm{I}\left({\mathrm{ER}}_{\mathrm{it}}>\mathsf{\gamma }\right)+{\mathsf{\theta }}_1{\mathrm{SUPPORT}}_{\mathrm{it}}+{\mathsf{\theta }}_2{\mathrm{MACHINE}}_{\mathrm{it}}\\ & + {\mathsf{\theta }}_3{\mathrm{STRUCTURE}}_{\mathrm{it}}+{\mathsf{\theta }}_4{\mathrm{LABOR}}_{\mathrm{it}}+{\mathsf{\theta }}_5{\mathrm{ENERGY}}_{\mathrm{it}}+{\mathsf{\theta }}_6{\mathrm{PRODUCTION}}_{\mathrm{it}}+{\mathsf{\theta }}_7{\mathrm{LAND}}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}}\end{array}$$

(2)

Equation (2) represents the single-threshold panel model, with the data separated into two regimes, in which CARBON denotes agricultural carbon emissions; TRADE denotes agricultural product trade openness, which is the core explanatory variable; ER denotes environmental regulation, which is the threshold variable; and the control variables include the level of fiscal support for agriculture (SUPPORT), the agricultural mechanization degree (MACHINE), the planting structure (STRUCTURE), labor input (LABOR), energy input (ENERGY), agricultural production (PRODUCTION) and farming land (LAND). The rest is the same as the explanation for Equation (1).

For the existence of a threshold effect, the original hypothesis is β₁ = β₂. If the original hypothesis holds, there is no threshold effect. In the case of a significant threshold effect, Hansen (1999) [36] further constructed the likelihood ratio statistic (LR) to calculate the confidence interval, and further tested the truthfulness of the threshold using the likelihood ratio test (LRT), that is, to test whether the threshold estimate is equal to its true value.

The threshold model of Hansen (1999) [36] can be extended to double or triple-threshold models. Equation (3) represents the double-threshold model, dividing the total sample into three groups, where γ₁ < γ₂.

$$\begin{array}{ll}{\mathrm{CARBON}}_{\mathrm{it}}={\mathrm{u}}_{\mathrm{i}}& + {\mathsf{\beta }}_1{\mathrm{TRADE}}_{\mathrm{it}}\mathrm{I}\left({\mathrm{ER}}_{\mathrm{it}}\le {\mathsf{\gamma }}_1\right)+{\mathsf{\beta }}_2{\mathrm{TRADE}}_{\mathrm{it}}\mathrm{I}\left({\mathsf{\gamma }}_1<{\mathrm{ER}}_{\mathrm{it}}\le {\mathsf{\gamma }}_2\right)+{\mathsf{\beta }}_3{\mathrm{TRADE}}_{\mathrm{it}}\mathrm{I}\left({\mathrm{ER}}_{\mathrm{it}}>{\mathsf{\gamma }}_2\right)\\ & + {\mathsf{\theta }}_1{\mathrm{SUPPORT}}_{\mathrm{it}}+{\mathsf{\theta }}_2{\mathrm{MACHINE}}_{\mathrm{it}}+{\mathsf{\theta }}_3{\mathrm{STRUCTURE}}_{\mathrm{it}}+{\mathsf{\theta }}_4{\mathrm{LABOR}}_{\mathrm{it}}+{\mathsf{\theta }}_5{\mathrm{ENERGY}}_{\mathrm{it}}\\ & + {\mathsf{\theta }}_6{\mathrm{PRODUCTION}}_{\mathrm{it}}+{\mathsf{\theta }}_7{\mathrm{LAND}}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}}\end{array}$$

(3)

3.2. Data Measurement and Description

3.2.1. Explained Variable

The explained variable is the agricultural carbon emissions (CARBON). According to the Vol. 4 (AFOLU) of IPCC guideline (2006), the agricultural emissions cover a lot of aspects [37,38]. In this paper, we calculated the agricultural emissions at the provincial level of China [39]. Due to limitations in data availability, we divided agricultural carbon emisssions into four parts [40,41,42]: carbon emissions generated by agricultural land use (${\mathrm{C}}_{\mathrm{land}}$), carbon emissions caused by rice cultivation (${\mathrm{C}}_{\mathrm{rice}}$), carbon emissions caused by rasing livestock and poultry (${\mathrm{C}}_{\mathrm{livestock}}$) and carbon emissions caused by biomass combustion (${\mathrm{C}}_{\mathrm{combustion}}$).

$$\mathrm{CARBON}={\mathrm{C}}_{\mathrm{land}}+{\mathrm{C}}_{\mathrm{rice}}+{\mathrm{C}}_{\mathrm{livestock}}+{\mathrm{C}}_{\mathrm{combustion}}$$

(4)

We first estimated the carbon emissions caused by agricultural land use. There are six main sources of carbon emissions in agricultural land use: the use of agricultural fertilizers, pesticides and agricultural films; the consumption of diesel fuel by agricultural machinery and equipment; the consumption of electricity in agricultural irrigation; and the carbon emissions from tilling the land. The total carbon emissions from agricultural land use are calculated by multiplying the six carbon sources by their corresponding carbon emission factors.

Table 1. Carbon emission factors of different sources in agricultural land use.

Sources of Carbon Emissions	Carbon Emission Factors	Reference Source
Fertilizers	0.8956 kg(C)/kg	Xiong et al. [43]
Pesticides	4.9341 kg(C)/kg	Xiong et al. [43]
Agricultural film	5.18 kg(C)/kg	Institute of Agricultural Resources and Ecological Environment, Nanjing Agricultural University
Diesel	0.5927 kg(C)/kg	United Nations Intergovernmental Committee of Experts on Climate Change
Irrigation	25 kg(C)/Cha	Dubey and Lal [44]
Tillage	312.6 kg(C)/km2	College of Biology and Technology, China Agricultural University

Data source: based on the literature by Xiong et al. (2016) [43] and Li et al. (2011) [45].

shows the corresponding carbon emission factors of different sources in agricultural land use.

It has been proven that planting rice produces a large amount of methane (CH₄). As the world’s second largest greenhouse gas (GHG) after carbon dioxide (CO₂), is also an important component of agricultural carbon emissions. The latest IPCC report shows that reducing CH₄ emissions is essential for achieving carbon emission reduction targets. We calculated the CH₄ emissions caused by rice cultivation mainly using the method of Min et al. (2012) [46], which uses the CH₄ emission factors of early-, middle-, and late-season rice in different provinces to weigh the sum of the local rice-planting area (

Table 2. Carbon emission factors of rice cultivation.

Province	Emission Factors (g/m2)			Province	Emission Factors (g/m2)
	Early-Season Rice	Middle-Season Rice	Late-Season Rice		Early-Season Rice	Middle-Season Rice	Late-Season Rice
Beijing	0	0	13.23	Henan	0	0	17.85
Tianjin	0	0	11.34	Hubei	17.51	39	58.17
Hebei	0	0	15.33	Hunan	14.71	34.1	56.28
Shanxi	0	0	6.62	Guangdong	15.05	51.6	57.02
Inner Mongolia	0	0	8.93	Guangxi	12.41	49.1	47.78
Liaoning	0	0	9.24	Hainan	13.43	49.4	52.29
Jilin	0	0	5.57	Sichuan/	6.55	18.5	25.73
				Chongqing
Heilongjiang	0	0	8.31	Guizhou	5.1	21	22.05
Shanghai	12.41	27.5	53.87	Yunnan	2.38	7.6	7.25
Jiangsu	16.07	27.6	53.55	Tibet	0	0	6.83
Zhejiang	14.37	34.5	57.96	Shaanxi	0	0	12.51
Anhui	16.75	27.6	51.24	Gansu	0	0	6.83
Fujian	7.74	52.6	43.47	Qinghai	0	0	0
Jiangxi	15.47	45.8	65.42	Ningxia	0	0	7.35
Shandong	0	0	21	Xinjiang	0	0	10.5

Data source: based on the literature by Min et al. (2012) [46].

). In order to make the various carbon emission measurement standards consistent, this study converted CH₄ into standard carbon uniformly, with a conversion coefficient of 6.8182 according to the IPCC report.

According to IPCC guideline (2006) [38], emissions from livestock and poultry and manure management are one of the important sources of agricultural GHGs, mainly including CH₄ and N₂O emissions. The intestinal fermentation of livestock and poultry produces CH₄, and animal manure management produces CH₄ and N₂O at the same time. In this study, we referred to the method used by Hu et al. (2010) [47] to calculate the CH₄ and N₂O produced during the raising process of cows, mules, horses, donkeys, camels, goats and sheep, pigs, and poultry. We used different factors (

Table 3. Carbon emission factors for raising livestock and poultry and manure management.

Livestock and Poultry	CH₄ Emission Factors (kg/head/year)		N₂O Emission Factors (kg/head/year)
	Gastrointestinal Fermentation	Manure Management	Manure Management
Dairy cows	68	16	1
Non-dairy cows	47	1	1.365
Mules	10	0.9	1.39
Horses	18	1.64	1.39
Donkeys	10	0.9	1.39
Camels	46	1.92	1.39
Goats and Sheep	5	0.16	0.33
Pigs	1	3.5	0.53
Poultry	/	0.02	0.02

Data source: based on the literature by Hu et al. (2010) [47].

) to obtain a weighted sum of the CH₄ and N₂O emitted by various animals. Since the statistical caliber of cattle changed in 2004, this paper divided cattle into dairy cows and non-dairy cows for separate calculations. According to the IPCC report, the conversion coefficients of CH₄ and N₂O to carbon are 6.8182 and 81.2727, respectively.

China is a major country in biomass burning, especially in rural areas, where the phenomenon of biomass burning is more common. Firewood and crop residue such as straw are the main sources of biomass. They produce CH₄ and N₂O during burning. In this study, we estimated the carbon emissions caused by biomass combustion from these two aspects.

Table 4. Carbon emission factors of biomass combustion.

Sources of Carbon Emissions	CH₄ Emission Factors	N₂O Emission Factors
Firewood burning	2.58 g/kg	0.075 g/kg
Straw burning	4.00 g/kg	0.13 g/kg

Data source: based on the literature by Climate Change Response Department of the National Development and Reform Commission (2014) [39].

shows the corresponding emission factors.

We calculated China’s agricultural carbon emissions from 2002 to 2020, and the results are shown in Figure 2. On the whole, agricultural carbon emissions have experienced two major twists and turns during this period. Since 2016, China’s agricultural carbon emissions have shown a continuous decreasing trend. Table 4 presents the agricultural carbon emissions of 31 provinces in China in 2020. It can be seen from the table that there is a wide gap in the agricultural carbon emissions among different provinces. The top 10 provinces in terms of total agricultural carbon emissions in 2020 were Hunan, Hubei, Anhui, Sichuan, Henan, Jiangxi, Jiangsu, Shandong, Inner Mongolia, and Guangxi (

Table 5. Agricultural carbon emissions of different provinces in 2020 (10,000 tons).

Province	Carbon Emissions	Province	Carbon Emissions	Province	Carbon Emissions
Beijing	18.39	Anhui	1405.96	Sichuan	1389.77
Tianjin	50.70	Fujian	456.39	Guizhou	527.69
Hebei	830.05	Jiangxi	1349.52	Yunnan	936.96
Shanxi	302.92	Shandong	1097.84	Tibet	418.66
Inner Mongolia	1062.20	Henan	1350.34	Shaanxi	453.57
Liaoning	570.31	Hubei	1468.85	Gansu	574.85
Jilin	557.98	Hunan	1749.35	Qinghai	365.69
Heilongjiang	948.06	Guangdong	940.96	Ningxia	160.23
Shanghai	69.60	Guangxi	992.48	Xinjiang	943.89
Jiangsu	1343.28	Hainan	164.29
Zhejiang	482.19	Chongqing	369.28

).

3.2.2. Core Explanatory Variable

The core explanatory variable of this study is agricultural product trade openness (TRADE). Trade openness can be measured by tariff rate, implementation of non-tariff barriers, nominal protection rate, trade dependence, etc., but no unified standard has been formed. In this paper, we used agricultural product trade dependence to measure trade openness, which has been used by many scholars. The agricultural products refer to the primary products from Plantation, forestry, animal husbandry and fishery, including cereal, vegetables, fruit, animal and aquatic products and so on. Based on the method of Chen et al. (2022) [15], we measured the agricultural product trade openness in each province using the ratio of the total import and export value of agricultural products to the value added of the domestic primary industry. The total import and export volume of agricultural products in each province was converted to CNY using the annual average exchange rate of CNY to USD per year.

3.2.3. Threshold Variable

The threshold variable is environmental regulation (ER). There is no unified approach to measure environmental regulation. Most of the existing literature has focused on the industrial field, which has very different characteristics from the agricultural industry, so methods for measuring environmental regulation in the agricultural field are rarely mentioned. According to the existing literature, there are three main methods to measure agricultural environmental regulation: First, the intensity of environmental regulation is presented by indicators such as the investment in agricultural environmental pollution control and the pollution emission intensity. Second, it is measured by agricultural environmental protection policy indicators, which may include the number and depth of related policies. Third, the adjusted coefficient of the GDP is used as a proxy variable of agricultural environmental regulation. The GDP related measurement of environmental regulation has been used by many studies [48,49,50]. Considering the availability of data and the universality of coverage, we referred to the method of Zhou et al. (2018) [51] and Zeng et al. (2021) [52] and used the adjusted coefficient of the GDP to measure the agricultural environmental regulation. As shown in Equation (5), GDP represents the gross domestic product of each province, “area” in the equation denotes the administrative area of each province, and the adjusted coefficient represents the reciprocal of the internal distance from the center of each province to the boundary.

$$\mathrm{ER}=\mathrm{GDP}\times \frac{1}{2/3\times \surd \mathrm{area}/\mathsf{\pi }}$$

(5)

In China, each province implements its own environmental policies, so provinces have heterogeneous environmental regulation. On the one hand, provinces with a higher level of economic development have more requirements for their environment quality due to a higher per capita income. These provinces are likely to invest more in environmental pollution management, pay more attention on environmental protection, implement more policies and regulations, allowing for a greater the intensity of environmental supervision. On the other hand, economically backward provinces may still have the phenomenon of sacrificing the environment for higher GDP, thus the intensities of environmental regulation may be relatively low.

In the case of the same GDP, the environmental regulation intensities of the provinces with smaller areas are greater than that of the larger provinces. The larger the administrative area, the larger the coverage required for environmental supervision. However, the edge areas, which are far from the center of province, are often rural areas, facing greater difficulty in environmental law enforcement and supervision. Therefore, the intensities of environmental regulation will be smaller.

3.2.4. Control Variables

When considering other factors affecting agricultural carbon emissions, this paper chose the following control variables:

(1)

Level of fiscal support for agriculture (SUPPORT): This was measured as the ratio of the local fiscal expenditure on agriculture, forestry, and water affairs to the local fiscal general budget expenditure, which can affect agricultural carbon emissions by changing agricultural factor inputs. For one thing, a higher level of fiscal support for agriculture may improve agricultural production conditions, including more advanced agricultural infrastructure, technologies and so on, which can be conducive to promoting agricultural carbon reduction. For another, agricultural fiscal support is also likely to induce more production that may increase emissions, if the same technology is used.

(2)

Agricultural mechanization degree (MACHINE): With the further acceleration of China’s agricultural modernization process, the level of agricultural mechanization in China has also improved. As an important input factor of agricultural production, agricultural machinery will have an impact on agricultural carbon emissions. A higher level of agricultural mechanization degree may reduce carbon emission by improving technologies, but it can also increase carbon emission because of more energy use. We used the ratio of the total mechanical power to the number of employees in the primary industry to represent the agricultural modernization degree.

(3)

Planting structure (STRUCTURE): there are different carbon emission intensities for planting various crops, so the planting structure will also have an impact on agricultural carbon emissions. In this paper, the ratio of the grain-planting area to the total crop-planting area was used to measure this index.

(4)

Labor input (LABOR): The labor force is the necessary input factor in agricultural production, so it will have a certain impact on agricultural carbon emissions. Basically, the more labor input there is, the greater the agricultural carbon emissions. We used the number of employees in the primary industry to represent this index.

(5)

Energy input (ENERGY): In China’s agricultural production, the main energy inputs are diesel and electricity. The greater the energy input in agricultural activities, the higher the carbon emissions generated. This paper used rural electricity consumption as the proxy variable for energy input.

(6)

Agricultural production (PRODUCTION): It is closed related to the agricultural carbon emissions because every link in agricultural production may generate carbon emissions. We used the value added of the primary industry as the proxy variable for agricultural production.

(7)

Farming land (LAND): It is measured by the cultivated land area. The farming land is a field that can be used to grow crops, vegetables, fruit, tea, shrubs and other trees, and is often cultivated and hoed. As an important part of agricultural land use, carbon emitted from land use can be significant and changes can occur in conjunction with land area.

3.2.5. Data Sources

This paper used balanced panel data from 31 provinces (municipalities and autonomous regions) in China (excluding Hong Kong, Macao, and Taiwan) from 2002 to 2020 as the samples. Agricultural import and export data came from the China Agricultural Products Import and Export Statistics Monthly and the China Agricultural Yearbook; the original data on agricultural carbon emissions mainly came from the China Rural Statistical Yearbook; the relevant data on environmental regulation came from the National Bureau of Statistics and official provincial websites; the number of employees in the primary industry came from the statistical yearbook of each province; and the relevant data on the fiscal support for agriculture, forestry, and water affairs, the total mechanical power, the rural electricity consumption, the planting structure, the value added of the primary industry and the cultivated area mainly came from the website of the National Bureau of Statistics and Compilation of Statistics for Sixty Years of New China. The descriptive statistics of the variables are shown in

Table 6. Descriptive statistics of variables.

Variables	Mean	Std. Dev.	Min	Max
CARBON	817.042	523.308	18.389	1937.524
TRADE	0.642	2.024	0.005	17.85
ER	1.375	1.867	0.00393	13.05
SUPPORT	0.100	0.039	0.009	0.204
MACHINE	3.670	2.093	0.412	12.59
STRUCTURE	0.660	0.134	0.354	0.971
LABOR	885.4	676.2	27	3398
ENERGY	224.1	355.0	0.360	2011
PRODUCTION	1403.319	1216.834	38.7	5559.9
LAND	4089.612	3097.325	93.547	17,750.44

.

4. Empirical Analysis

4.1. Threshold Effect Test

To determine the existence of the threshold effect, this study used agricultural carbon emissions as the dependent variable, agricultural product trade openness as the core explanatory variable, and environmental regulation as the threshold variable. From the significance test of the threshold effect, it was found that the single threshold of environmental regulation was significant at the 1% level with an F-statistic of 76.72 and the double-and triple-threshold effects did not pass the significance test. The F-statistics, p-values, and critical values corresponding to the significance test of the threshold effect are shown in

Table 7. Threshold effect significance test.

Threshold Type	F-Value	p-Value	Critical Values
			1%	5%	10%
Single	76.72 ***	0.000	45.945	32.277	28.213
Double	22.75	0.233	50.009	36.257	30.210
Triple	19.98	0.540	83.582	61.400	47.486

Note: *** denotes significance at the level of 1%.

.

Since the single threshold effect was significant, the estimated threshold values and confidence intervals in the single-threshold effect are reported in

Table 8. Threshold estimates and confidence intervals.

Threshold Type	Threshold Value	95% Confidence Interval
Single	1.603	[1.485, 1.649]

. The threshold value for the single-threshold effect was 1.603, and the 95% confidence interval was [1.485, 1.649]. Figure 3 shows the variation in the LR statistic for the threshold estimates; the value corresponding to the lowest point of the likelihood function is the threshold value, and the interval corresponding to the part of the likelihood function below the dashed line is the 95% confidence interval for the threshold value.

4.2. Analysis of the Threshold Model Regression

According to the threshold effect significance test, it was found that there was a single-threshold effect of environmental regulation in the impact of agricultural product trade openness on agricultural carbon emissions.

Table 9. Regression results of threshold model.

Variables	Coef.	Std. Err.	t	p > |t|	95% Conf. Interval
TRADE-1	0.051	0.011	4.59	0.000	0.029	0.073
TRADE-2	−0.033	0.005	−7.01	0.000	−0.042	−0.024
SURPPORT	−1.220	0.203	−6.01	0.000	−1.619	−0.821
MACHINE	0.017	0.005	3.46	0.001	0.008	0.027
STRUCTURE	0.158	0.088	1.79	0.074	−0.015	0.331
LABOR	0.318	0.033	9.51	0.000	0.252	0.384
ENERGY	0.051	0.019	2.63	0.009	0.013	0.089
PRODUCTION	0.047	0.018	2.56	0.011	0.011	0.082
LAND	0.193	0.035	5.56	0.000	0.125	0.262

shows the results of the threshold regressions based on the single threshold. TRADE-1 and TRADE-2 in the following table denote TRADE·I ($ER\le \gamma$) and TRADE·I ($ER>\gamma$), respectively.

A higher value in the environmental regulation index indicates a higher intensity of environmental regulation. According to the estimation results shown in Table 9, it can be seen that when the intensity of environmental regulation was lower than the threshold value of 1.603, the estimated coefficient of agricultural product trade openness was 0.051, which was significant at the 1% level. With an improvement in environmental regulation, the estimated coefficient became negative. When the value of environmental regulation was higher than the threshold value of 1.603, the estimated coefficient of agricultural product trade openness was −0.033 and passed the significance test at the 1% level. The results indicated that there are large differences in the impacts of agricultural product trade openness on agricultural carbon emissions at different environmental regulation intensities. When the environmental regulation was weak, the agricultural product trade openness led to an increase in agricultural carbon emissions, while high-intensity environmental regulation promoted the agricultural carbon reduction effect of agricultural product trade openness, indicating that agricultural product trade openness can only reduce agricultural carbon emissions under high-intensity environmental regulation. The possible reason is that under low-intensity environmental regulation, people tend to pay less attention to agricultural ecology, and the focus of agricultural development mainly lies in agricultural production and scale in order to obtain higher economic returns. Therefore, an improvement in agricultural product trade openness will promote more agricultural production inputs, thus causing more agricultural carbon emissions. As the intensity of environmental regulation increases, more attention will be given to agricultural green development and production and operation modernization in order to meet the requirement of strict environmental regulation, which can make resource allocation and utilization more reasonable. When environmental regulation exceeds the threshold value, agricultural product trade openness can significantly reduce agricultural carbon emissions. It can also be seen from the estimated coefficients of the control variables that the level of fiscal support for agriculture can significantly reduce agricultural carbon emissions, while an increase in the labor input or energy input significantly increases agricultural carbon emissions. And the estimated coefficient of the planting structure, the degree of agricultural mechanization, agricultural production and farming land are all positive, indicating that they were conducive to the agricultural carbon emissions.

According to the “National Food Security Medium and Long-Term Planning Outline (2008–2020)”, China has divided 31 provinces into three functional grain areas: the major grain-producing area (13 provinces), the major grain-selling area (7 provinces), and the grain producing-selling balance area (11 provinces). With good geography, soil, climate, and other natural conditions, the major grain-producing area is suitable for growing grain crops; it has a large planting proportion and high grain production. With a relatively higher economic development level and less arable land area, the major grain-selling area attracts more people to live there, resulting in a large gap between grain production and consumption. The grain producing-selling balance area can basically maintain self-sufficiency in grain consumption, but it makes limited contributions to national grain consumption. The specific provinces included in each functional grain area are shown in

Table 10. Division of functional grain areas.

Functional Areas	Provinces
Major grain-producing areas	Henan, Inner Mongolia, Hunan, Hebei, Sichuan, Jilin, Liaoning, Jiangxi, Shandong, Jiangsu, Anhui, Hubei, Heilongjiang
Major grain-selling areas	Zhejiang, Beijing, Fujian, Shanghai, Guangdong, Tianjin, Hainan
Grain producing-selling balance areas	Shaanxi, Yunnan, Guangxi, Xinjiang, Chongqing, Gansu, Shanxi, Qinghai, Guizhou, Ningxia, Tibet

. In

Table 11. Regression results of threshold models in different functional grain areas.

Variable	(1)	(2)	(3)
TRADE (ER ≤ 0.126)	0.093 ***
	(0.025)
TRADE (ER > 0.126)	0.024
	(0.021)
TRADE (ER ≤ 1.658)		−0.075 **
		(0.037)
TRADE (ER > 1.658)		−0.254 ***
		(0.037)
TRADE (ER ≤ 0.087)			0.043 **
			(0.017)
TRADE (ER > 0.087)			0.001
			(0.017)
Control variables	YES	YES	YES
Threshold type	single	single	single
F-statistic	30.68	49.75	48.86
p-value	0.046	0.003	0.033

Note: the standard deviation gathered at the provincial level is shown in parentheses; *** and ** denote significance at 1% and 5% respectively.

, columns (1), (2), and (3) show the regression results of the threshold effect for the major grain-producing area, the major grain-selling area, and the grain producing-selling balance area, respectively. As can be seen from the table, the single-threshold effects of environmental regulation held in all three functional grain areas, but there were obvious regional differences in the impact of agricultural product trade openness on agricultural carbon emissions. In the major grain-producing area, the estimated coefficients of agricultural product trade openness were positive, indicating that agricultural product trade openness had a positive impact on the agricultural carbon emissions, but the effect was not significant when the intensity of environmental regulation was high. The estimated coefficient decreased from 0.093 to 0.024, with an increase in the environmental regulation intensity, indicating that the contribution of agricultural product trade openness to the increase in agricultural carbon emissions was weakening. In the major grain-selling area, the estimated coefficients of agricultural product trade openness were all negative, indicating that agricultural product trade openness was conducive to reducing the agricultural carbon emissions. With an increase in the environmental regulation intensity, the coefficient of agricultural product trade openness became −0.254 and passed the significance test at the 1% level, showing that the impact of agricultural product trade openness on agricultural carbon emissions had further increased. In the grain producing-selling balance area, the estimated coefficients of agricultural product trade openness ranged from 0.043 to 0.001, with all positive signs, indicating that agricultural product trade openness can increase the agricultural carbon emissions. The estimated coefficients gradually decreased as the intensity of environmental regulation increased in the grain producing-selling balance area, but this was not significant when the environmental regulation was greater than the threshold value. To summarize, under high-intensity environmental regulation, agricultural product trade openness can significantly reduce the agricultural carbon emissions in the major grain-selling area. Additionally, it is worth noting that the threshold value of the major grain-selling area was slightly higher than the threshold value in Table 8, and the threshold values of the other two functional grain areas wre much smaller. The possible reason is that the major grain-selling area mainly includes more economically developed provinces such as Beijing and Shanghai, resulting in higher requirements for the ecological environment, so the intensity of environmental regulation is more likely to be greater. Nevertheless, most provinces in the grain producing-selling balance area belong to the western region of China, where the gap in the economic development level with other regions is big, so the intensity of environmental regulation is generally lower. Therefore, the threshold value of environmental regulation in the regression of the grain producing-selling balance area was the lowest. Overall, it seems that agricultural product trade openness is more likely to perform its carbon emission reduction effect in the major grain-selling area.

4.3. Regional Distribution Analysis

According to the threshold effect test, the single-threshold effect was significant, meaning that there is a single-threshold value of environmental regulation that allows the sample to be divided into two groups: provinces with a low environmental regulation intensity (ER ≤ 1.603) and provinces with a high environmental regulation intensity (ER > 1.603). Figure 4 shows the changes in the number of provinces in the two intervals from 2002 to 2020. During this period, the numbers of provinces with low and high environmental regulation intensities were 157 and 432 respectively, accounting for 26.66% and 73.34% of the total sample. This means that the intensity of environmental regulation in most provinces is relatively low, which fails to promote the impact of agricultural product trade openness on agricultural carbon emission reduction. As shown in this figure, before 2019, the number of provinces in the low-intensity environmental regulation interval was larger than the number of provinces in the high-intensity environmental regulation interval, but the gap between the two has shrunk year by year. Since 2019, the environmental regulation intensities of 18 provinces in total have been greater than 1.603, the threshold value, and the number of provinces in the high-intensity environmental regulation interval ranks first. This shows that more and more provinces in China have begun to strengthen their environmental regulation, which will undoubtedly play an important role in promoting the harmonious development of agricultural product trade and the environment.

Table 12. Provinces with different intensities of environmental regulation (2020).

Environmental Regulation	Provinces
ER ≤ 1.603	Liaoning, Shanxi, Inner Mongolia, Jilin, Heilongjiang, Guangxi, Hainan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang
ER > 1.603	Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Anhui, Fujian, Jiangxi, Shandong, Henan, Hubei, Hunan, Guangdong, Chongqing, Sichuan

shows the specific provinces located in different environmental regulation intervals in 2020. Among them, the provinces with a higher environmental regulation intensity mainly include Beijing, Shanghai, Jiangsu, Zhejiang, etc., and the provinces with a lower environmental regulation intensity mainly include Tibet, Qinghai, Gansu, Xinjiang, etc. It can be seen that the provinces with high-intensity environmental regulation are mainly concentrated in the eastern coastal areas, while the provinces with low-intensity environmental regulation are mostly distributed in the western areas. Six provinces out of seven in the major grain-selling area are within the scope of the high-intensity environmental regulation interval, which might be the main reason why agricultural product trade openness is more likely to perform its carbon emission reduction effect in the major grain-selling area.

4.4. Robustness Test

As shown in

Table 13. Robust test.

Variable	(1)	(2)	(3)
TRADE (ER ≤ 1.884)	0.082 ***
	(0.011)
TRADE (ER > 1.884)	−0.036 ***
	(0.005)
TRADE (ER ≤ 1.603)		0.045 ***
		(0.010)
TRADE (ER > 1.603)		−0.021 ***
		(0.005)
TRADE (ER ≤ 1.603)			0.048 ***
			(0.011)
TRADE (ER > 1.603)			−0.014
			(0.005)
Control variables	YES	YES	YES
Threshold type	single	single	single
F-statistic	163.90	53.35	50.32
p-value	0.000	0.000	0.003

Note: *** denotes significance at the level of 1%.

, robustness tests are conducted by replacing the explained variables (column 1), adding control variables (column 2), and performing a tailing process (column 3). Firstly, the explained variable was changed from the total agricultural carbon emissions to the per capita agricultural carbon emissions, which was measured by calculating the ratio of agricultural carbon emissions to the regional resident population. The results show that there was still a significant single-threshold effect of environmental regulation in the impact of agricultural product trade openness on the per capita agricultural carbon emissions. With an improvement in environmental regulation, the estimated coefficient of agricultural product changed from 0.082 to −0.036, with both being significant at the 1% level, meaning that agricultural product trade openness can reduce the per capita agricultural carbon emissions only when the intensity of environmental regulation is high, which is basically consistent with the results in Table 8. This suggests that the main regression results are robust. In column (2), the three control variables of urbanization level, rural residents’ consumption expenditure and level of agricultural land scale management were added to the original regression. From Table 13, we can see that the single-threshold effect was still robust. In light of the estimated coefficients of agricultural product trade openness, we also reached the conclusion that high-intensity environmental regulation is beneficial for enabling agricultural product trade openness to reduce agricultural carbon emissions, indicating the robustness of our main conclusions. In addition, considering that there may be abnormal data in the large data sample, this paper conducted a robustness test by performing a 1% tailing process to eliminate the extreme values of each variable, and the results are shown in column (3). This showed that there is a significant single-threshold effect of environmental regulation in the impact of agricultural product trade openness on agricultural carbon emissions, and the impact coefficient turns from positive to negative when the intensity of environmental regulation increases, which also indicates that the estimated results of our regression are robust.

5. Conclusions and Policy Implications

Based on the provincial panel data of China from 2002 to 2020, this paper first measured the agricultural carbon emissions and then used the panel threshold model to empirically test the impact of agricultural product trade openness on the agricultural carbon emissions from the perspective of environmental regulation by examining the overall effect and regional heterogeneity. The main conclusions are as follows:

First, there is a significant environmental regulation threshold effect on the impact of agricultural product trade openness on agricultural carbon emissions. When the intensity of environmental regulation was lower than the threshold value (1.603), the estimated coefficient of agricultural product trade openness was 0.051, meaning that a higher agricultural product trade openness would bring about an increase in carbon emissions. Agricultural product trade openness can significantly reduce agricultural carbon emissions only with the high-intensity environmental regulation.

Second, there is a single-threshold effect of agricultural product trade openness on agricultural carbon emissions in all three functional grain areas, namely the major grain-producing area, the major grain-selling area, and the grain producing-selling balance area, but obvious regional heterogeneity exists. Agricultural product trade openness was conducive to the agricultural carbon emissions in the major grain-producing area and the grain producing-selling balance area, but the contribution was weakening with an increase in the environmental regulation intensity. Conversely, agricultural product trade openness had a negative impact on the agricultural carbon emissions in the major grain-selling area. Overall, it seems that agricultural product trade openness has a greater tendency to perform a carbon emission reduction effect in the major grain-selling area.

The results presented in this paper has important implications. First, due to the significant threshold effect of environmental regulation on the impact of agricultural product trade openness on carbon emissions, it is crucial to leverage environmental regulation to promote positive environmental outcomes through trade openness. Empirical results indicate that only higher levels of environmental regulation can effectively reduce carbon emissions through trade openness. Therefore, the government should prioritize the coordination of trade and environmental regulation policies, while strengthening environmental management in the agricultural industry.

Second, agricultural development strategies should consider factors such as economic development level, trade intensity, environmental regulation intensity, and carbon emission reduction pressure in different functional grain areas. The major grain-producing area faces higher pressure to reduce carbon emissions, requiring increased support for green agricultural development through subsidies, technological assistance, and other means to enhance the use of environmentally friendly production factors. In the major grain-selling area, where trade openness is more likely to have a carbon emission reduction effect, efforts should be made to further improve agricultural product trade openness. Moreover, since the grain-producing-selling balance area generally has lower environmental regulation levels, it is essential to strengthen agricultural environmental regulations while ensuring basic grain self-sufficiency to ensure sustainable agricultural development.