A Study on the Impact Mechanism of Agricultural Trade on Agricultural Carbon Emissions

Abstract

Controlling agricultural carbon emissions is crucial for addressing global environmental challenges. As a major player in agricultural trade, China needs to explore a specific pathway to reduce its agricultural carbon emissions. This study delves into the impact of China’s agricultural trade on agricultural carbon emissions: (1) The research demonstrates that agricultural trade (AT) significantly reduces China’s agricultural carbon emissions (ACEs), with robustness and endogeneity tests supporting these findings. (2) In the process of lowering agricultural carbon emissions, AT exerts both direct and indirect effects. The direct effect stems from the import substitution effect of agricultural trade, while the indirect effects include agricultural technological innovation (ATI) and the agricultural carbon emission intensity (ECI). (3) The reduction in ACEs is more pronounced in eastern regions, coastal areas, and non-major grain-producing regions. This study reveals the underlying mechanisms between AT and ACEs, suggesting that China has the potential to achieve mutual benefits in international trade and environmental protection. It also provides a trade-oriented perspective for formulating agricultural emission reduction policies.

1. Introduction

In recent years, the entrenched conflicts between human activities and natural ecosystems have become increasingly evident. Environmental pollution, rising atmospheric CO₂ levels, and the greenhouse effect have precipitated persistent climate change while amplifying the frequency of extreme weather events globally. Climate warming and carbon emission reductions have emerged as critical concerns for nations worldwide. Historically, carbon reduction efforts under accelerated urbanization and industrialization predominantly focused on high-pollution industrial sectors. However, agricultural activities often remain overlooked due to their inherent carbon sink characteristics. According to the Food and Agriculture Organization of the United Nations (FAO), over one-third of global anthropogenic greenhouse gas emissions originate from food systems, with absolute emissions continuing to rise. Controlling agricultural carbon emissions has become imperative for achieving long-term sustainable socio-economic development. On 24 October 2021, the Central Committee of the Communist Party of China and the State Council issued the “Guidelines on Comprehensively Implementing New Development Concepts for Carbon Peaking and Neutralization”, explicitly stating that “China needs to accelerate green agricultural development and enhance carbon sequestration capacity and efficiency in agriculture”, establishing new requirements for agricultural carbon emissions and sequestration. Addressing carbon reduction has progressively become a research priority for scholars domestically and internationally.

In the agricultural sector, numerous scholars have proposed carbon reduction pathways, including establishing carbon constraint indicators and developing agricultural carbon markets [1]; implementing multi-agent controls encompassing farmers, governments [2], and societal participation; and adopting concrete measures such as clean energy adoption, production scale management, productivity enhancement, and technological innovation [3]. Since the ratification of the Paris Agreement in 2016, carbon emissions have evolved into a globally coordinated challenge, shifting academic perspectives from domestic markets to international contexts. Transnational carbon emission transfers offer expanded options for agricultural production and emission management. Following China’s accession to the World Trade Organization (WTO) in 2001, the nation has established trade relations with dozens of countries, witnessing rapid growth in international agricultural trade. WTO statistics from 2011 revealed China surpassing the United States as the world’s largest agricultural importer and exporter. Regarding trade structures, both import and export volumes of Chinese agricultural products have increased, though with divergent growth rates—exports demonstrate a steady growth while imports exhibit an exponential expansion. According to China’s Ministry of Commerce statistics, agricultural imports in 2022 reached 2.4 times the value of exports. The export portfolio predominantly comprises labor-intensive, low-value-added primary processed products, such as vegetables, fruits, and aquatic products, whereas imports primarily consist of land-intensive commodities [4], including oil crops, industrial/medicinal plants, feedstocks, and grains. The production processes of these imported agricultural products involve multiple high-emission stages. Agriculturally advanced nations possess environmental technological advantages, such as precision irrigation and fertilization controls, enabling low-carbon agricultural production. Guided by comparative advantage theory, international agricultural trade could serve as an effective pathway for emission reductions due to varying carbon intensities in agricultural production. Therefore, whether China’s international agricultural trade can effectively mitigate agricultural carbon emissions, along with the mechanisms through which import–export activities influence agricultural emissions, has become a crucial practical issue for China’s high-quality agricultural development.

The current research on mechanisms linking agricultural trade to carbon emissions remains limited [5]. This study employs panel data from 30 Chinese provinces (2001–2022) to empirically examine the impact of agricultural trade on carbon emissions and its underlying mechanisms and is grounded in the theoretical analysis of China’s agricultural trade-carbon emission relationship. The marginal contributions of this paper lie in refining research perspectives and expanding theoretical insights. First, utilizing data from 2001 to 2022 and applying a two-way fixed effects model, this research comprehensively investigates how AT affects China’s ACEs, thereby broadening the scope of agricultural carbon research. Second, it empirically examines both direct and indirect effects of AT on ACEs, revealing critical pathways for agricultural carbon reduction and enriching related academic discourses. Third, leveraging China’s geographical diversity, this study conducts a cross-regional heterogeneity analysis, elucidating varying emission effects under different contextual characteristics. Furthermore, given China’s rapid economic and technological development with significant interprovincial disparities, these findings could inform the agricultural carbon reduction policy formulation in other emerging economies.

2. Literature Review

The theoretical foundation of transboundary carbon emissions traces back to the “Pollution Haven Hypothesis” proposed by Copeland in 1990, which posits that developed nations can improve their environmental quality by relocating polluting industries to developing countries, thereby turning the latter into pollution havens. Subsequently, American economists Grossman and Krueger [5] integrated environmental science into the Kuznets curve, formulating the “Environmental Kuznets Curve (EKC)”, demonstrating an inverted U-shaped relationship between a nation’s environmental quality and economic growth. In less developed economies, the rising per capita income exacerbates environmental pollution, whereas in developed economies, an increased income correlates with reduced pollution. Cole [6] synthesized these perspectives, empirically validating the existence of the “Pollution Haven Hypothesis” while assessing the impact of foreign trade levels and structural changes on the EKC.

An academic consensus remains elusive regarding the relationship between import-export trade and carbon emissions, with divergent perspectives persisting [7]. First, proponents argue that trade openness reduces carbon emissions. Li and Wang [8], analyzing carbon productivity drivers in Belt and Road countries, identified trade as a mitigating factor. Similarly, Ike et al. [9] demonstrated a significant CO₂ emission suppression through a trade volume analysis in G7 nations. Second, supporters of the “Pollution Haven Hypothesis” contend that trade openness amplifies emissions. For instance, Dou et al. [10] revealed that trade openness positively influences greenhouse effects, where imports increase emissions while exports reduce national carbon footprints. Zeng and Yue [11], investigating spatial carbon characteristics in the Yangtze River Delta, observed that trade expansion elevates local and adjacent regional emissions, indicating spatial spillover effects. Third, a reciprocal promotion between international trade and environmental impacts emerges, suggesting that a higher agricultural trade openness correlates with increased emissions, which in turn stimulates foreign trade [1]. Fourth, nonlinear trade-emission relationships are evident, with trade impacts varying directionally and in magnitude across nations at different income levels [12]. Furthermore, R&D investments exhibit threshold effects: below critical thresholds, trade expansion increases primary sector emissions, whereas surpassing thresholds activates emission reductions through trade [8].

In summary, regarding impact mechanisms and regional heterogeneity, the existing research exhibits unresolved debates and gaps in AT and ACEs. This study addresses these limitations through a theoretical analysis of AT and ACEs, employing panel data from 30 Chinese provinces (2001–2022) to empirically analyze the influence of AT on the China’s ACEs and underlying mechanisms. The analysis systematically explores direct and indirect pathways while conducting cross-regional heterogeneity tests to elucidate spatial variation patterns.

3. Theoretical Analysis and Research Hypothesis

Grossman and Krueger [13] pioneered the construction of a general equilibrium model of trade, systematically testing the environmental impacts of trade through three core mechanisms: scale effects (economic expansion driving emissions growth), structural effects (industrial relocation shaping emission patterns), and technological effects (innovation-driven emission mitigation). Building upon this theoretical foundation, this study extends the analytical framework to AT’s influence on ACEs into direct effects and indirect effects.

3.1. Direct Effect of AT on ACEs

Agricultural trade influences domestic agricultural carbon emissions through the import substitution of domestically produced commodities [14], thereby reshaping domestic agricultural production systems, optimizing factor input structures, and ultimately affecting national agricultural carbon footprints. From the export perspective, developing countries typically prioritize agricultural export expansion to capture an international market share and maximize economic returns. This strategic orientation often leads to production scale enlargement and yield intensification, potentially resulting in an irrational land use expansion and excessive deployment of pollution-intensive inputs (e.g., chemical fertilizers and pesticides). Such practices degrade the ecological foundation for sustainable agriculture, consequently elevating agricultural carbon emissions.

Regarding imports, nations can reduce domestic carbon emissions through substitution effects by importing agricultural products with high embedded carbon emissions. This strategy decreases pollution-intensive input utilization in the domestic production of equivalent goods, constituting the carbon substitution effect of agricultural trade. To determine whether agricultural import–export trade ultimately inhibits or promotes carbon emissions, the magnitude of these countervailing effects must be empirically evaluated. China’s AT exhibits three distinctive characteristics: (1) the sustained expansion of an AT deficit in recent years; (2) the export dominance of primary agricultural products with lower carbon footprints; and (3) the import prevalence of processed and resource-intensive products associated with higher emission levels.

Based on this structural pattern, this study hypothesizes that the ACEs generated through export production are outweighed by emission reductions achieved through import substitutions. This implies that AT can effectively reduce China’s ACEs via the import substitution (IS) effect. The following hypotheses are consequently proposed:

H1. 

AT can significantly diminish ACEs.

H2. 

AT can directly diminish ACEs by the IS effect.

3.2. Indirect Effects of AT on ACEs

AT can indirectly diminish ACEs by fostering technological innovation through competition and learning effects. First, when a country enters the global market, its domestic agricultural sector faces intensified competition, which drives green innovation [15]. Moreover, environmental regulations in developed countries compel exporters to adopt low-carbon technologies and optimize production processes, further reducing emissions.

Second, as environmental degradation worsens, economically advanced nations, like Japan, increasingly prioritize ecosystem protection and scrutinize the environmental impact of agricultural trade [16]. Their advanced green technologies and management practices generate spillover effects through trade, enabling importing countries like China to adopt sustainable models. Under its “dual-carbon” policy, China has leveraged agricultural trade to assimilate foreign innovations, accelerating its transition to low-carbon agriculture. Therefore, this study proposes the following hypothesis:

H3. 

AT can indirectly diminish ACEs by the agricultural technology innovation effect (ATI).

On the other hand, trade in agricultural products can also optimize the energy mix and reduce the intensity of energy consumption, thereby indirectly contributing to the reduction in carbon emissions in agriculture. The principal reason for China’s high carbon emissions in agriculture lies in the inefficient utilization of resources and crude agricultural production methods, the increased dependence on petroleum fuels, and the high energy consumption of agriculture. As of 2018, the proportion of carbon emissions from agricultural energy consumption has exceeded that of chemical fertilizers, becoming the top source of emissions [1]. The expansion of agricultural trade and the increase in marketization can notably optimize the energy consumption status in the domestic production of agricultural products and reduce agricultural carbon emissions. First of all, in the international trade market, the green barriers of various countries make China’s export of agricultural products more restrictive, increasing the cost of sales and the price. Further exploring the green and low-carbon agricultural production pattern is necessary to ensure a competitive advantage. In recent years, in rural revitalization, China’s agriculture has gradually improved its production efficiency and reduced its energy consumption by changing the traditional production model. Secondly, with the growing import scale, agricultural imports will further squeeze the domestic market, forcing the domestic agricultural industry to upgrade and to technologically progress. Finally, as China has maintained a trade deficit in agricultural products in the past two decades, domestic labor, land, and other factors of production are released so that the allocation of resources is further optimized in the direction conducive to reducing total carbon emissions [17]. Therefore, this study proposes the following hypothesis:

H4. 

AT can indirectly diminish ACEs by the energy consumption intensity effect (ECI).

Figure 1 shows the specific theoretical mechanism of agricultural trade affecting agricultural carbon emissions.

4. Research Design

4.1. Variable Selection

(1) Dependent variable

The ACE is the dependent variable of this article, which is measured according to the approach of Tian et al. [18].

$$C=\sum C_i\times {\sigma }_i$$

(1)

where $C_i$ stands for the level of activity of various sources of carbon emissions, and ${\sigma }_i$ is the carbon emission coefficient of each type of carbon emission source.

Drawing on related studies [18,19,20,21], this paper mainly selects five types of agricultural carbon emission sources. The first type is carbon emissions from agricultural energy consumption, containing coal, crude oil, diesel fuel, gasoline, and fuel oil, which are uniformly converted to standard coal energy; the second type is carbon emissions caused by the consumption of agricultural materials, including chemical fertilizers, pesticides, and farming plastic films; the third type is carbon emissions from agricultural planting activities, which consists of irrigation, plowing, and straw burning, where the stover includes rice, maize, beans, sorghum, and wheat stover; the fourth category mainly refers to the carbon dioxide produced by rice cultivation; and the fifth category is GHG emissions from the intestinal and fecal fermentation during livestock farming, with four species of animals selected: cattle, sheep, pigs, and poultry. The carbon emission coefficients of various emission sources are detailed in

Table 1. Carbon emission sources, emission coefficients, and sources of coefficients.

Categories	Emission Sources	Emission Coefficients	Sources of Coefficients
Energy consumption	Energy consumption (standard coal)	1.47 kg(C)/kg	IPCC
Agricultural materials usage	Fertilizer	0.8956 kg(C)/kg	West T. O., Oak Ridge National Laboratory
	Pesticides	4. 9341 kg(C)/kg	Oak Ridge National Laboratory
	Agricultural film	5. 18 kg(C)/kg	Institute of Resource, Ecosystem and Environment of Agriculture, Nanjing Agricultural University
Agricultural planting	Irrigation	266. 48 kg(C)/hm2	West T. O.
	Plowing	312. 6 kg(C)/hm2	China Agricultural University
	Straw burning	1. 247 kg(C)/kg	Analysis Method on Reducing Emission of SO2 and CO2 by Rural Energy Construction [22]
Rice cultivation	Rice cultivation	3.136 g(C)/(m2·day)	Provincial Agricultural Carbon Emissions in China: Calculation, Performance Change and Influencing Factors [23]
Livestock farming	Cattle	420.11 kg(C)/head·year	IPCC
	Sheep	35.26 kg(C)/head·year	IPCC
	Pig	34.1 kg(C)/head·year	IPCC
	Poultry	0.39 kg(C)/head·year	IPCC

.

(2) Explanatory variable

This study takes the trade of agricultural products as the explanatory variable and specifically measures it using the total import and export volume of agricultural products. In addition, to further verify the relationship between the import and export structure of farm produce as well as the trade substitution effect of agricultural imports and exports and carbon emissions from farming, this study chooses the value of imports from farming goods, the value of exports from farming goods, and the import and export structure of farm produce and exports as the explanatory variables for regression. The structure of farm produce imports and exports is calculated by dividing the value of agricultural imports by the value of exports.

(3) Control variables

Given that there are certain differences among provinces, such as their geography, infrastructure construction, economic development level, agricultural scale, technology level, etc., the total agricultural output value, total mechanical power, area of soil erosion control, per capita years of education, road mileage, total postal and telecommunication business, technology market turnover, and internal R&D expenditures are the control variables. In the specific model estimation process, to obviate the impact of heteroskedasticity, this paper has logarithmically treated all the variables, and the description of the variables is shown in

Table 2. Descriptive statistics of variables.

Variables	Variable Descriptive	Observations	Mean	Standard Error	Minimum	Maximum
lncarbon	Agricultural carbon emissions	660	7.486	1.126	3.635	9.552
lntrade	Total amount of agricultural trade	660	2.850	1.527	0.048	6.779
Lnagr_gdp	Gross agricultural output	660	6.759	1.130	3.388	8.846
lnmach	Total mechanical power	660	7.526	1.089	4.554	9.500
lnsoil	Area of soil erosion control	660	7.509	1.894	0.000	9.723
lnper_edu	Per capita years of education	660	2.270	0.109	1.952	2.662
lnroad	Road mileage	660	2.378	0.756	0.476	3.739
lnpost	Total postal and telecommunication business	660	6.244	1.066	2.654	9.302
Lntech_market	Technology market turnover	660	4.123	1.896	0.058	8.993
lnrd	Internal R&D expenditure	660	14.085	1.657	9.043	17.821

.

4.2. Data Sources and Descriptive Statistics

(1) Data source

This study covers 30 Chinese provinces (including autonomous regions and municipalities directly under the central government) from 2001 to 2022. While most existing research uses data from 31 provinces (excluding Hong Kong, Macao, and Taiwan), this study excludes the Tibet Autonomous Region due to the severe data unavailability after a careful examination.

The data were sourced from the China Statistical Yearbook, China Energy Statistical Yearbook, China Agricultural Yearbook, and provincial statistical yearbooks. Missing values were supplemented using linear interpolation. During the dataset construction, provincial identifiers were used to merge data from different sources, resulting in a balanced panel dataset of 30 Chinese provinces (autonomous regions and municipalities) spanning 2001–2022. The final sample consists of 660 observations (30 provinces × 22 years).

(2) Descriptive statistics

The descriptive statistics of each variable are detailed in Table 2. The average for agricultural imports is 2.85, which is low overall. Its maximum and minimum values are 6.779 and 0.048, with a vast difference between them, indicating a significant difference in the agricultural product trade among provinces. The mean carbon emissions from farming are 7.486, close to its maximum value of 9.552, suggesting that the overall farming carbon emissions in various provinces of China are relatively high. From the perspective of other control variables, such as the total output value of agriculture, the mechanization level, road convenience, the total number of posts and telecommunications services are above the medium level, and the per capita education years are longer. The average turnover of the technical market is 4.123, the trading volume is low, and the technical trading market lacks vitality.

4.3. Model Construction

Drawing on the existing empirical research, this study conducted diagnostic tests on the data [5]. First, a Hausman test was performed, yielding a p-value of 0.000, indicating that the random-effects model would be biased. Subsequently, an F-test for fixed effects was conducted, which also produced a p-value of 0.000, suggesting that the pooled OLS model would be inappropriate. Therefore, a two-way fixed effects model was constructed in this study as follows:

$${lncarbon}_{it}={\alpha }_0+{\alpha }_1{lntrade}_{it}+{\alpha }_2{control}_{it}+{\mu }_i+{\delta }_t+{ϵ}_{it}$$

(2)

In addition, to test the mechanism of the intermediate effect of AT on ACEs, the following models (2) and (3) are constructed:

$$M_{it}={\beta }_0+{\beta }_1{lntrade}_{it}+{\beta }_2{control}_{it}+{\mu }_i+{\delta }_t+{ϵ}_{it}$$

(3)

$${lncarbon}_{it}={\gamma }_0+{\gamma }_1{lntrade}_{it}+{\gamma }_2{M}_{it}+{\gamma }_3{control}_{it}+{\mu }_i+{\delta }_t+{ϵ}_{it}$$

(4)

In the models (1)–(3), $i$ represents different provinces, and $t$ represents different years. ${lncarbon}_{it}$ is carbon emissions; ${lntrade}_{it}$ stands for the total value of imports and exports of farm products; ${control}_{it}$ refers to a set of control variables; $M_{it}$ indicates agricultural technology innovation and the energy consumption intensity; ${\alpha }_0$, ${\beta }_0$, and ${\gamma }_0$ are intercept terms; and ${\alpha }_1$, ${\alpha }_2$, ${\beta }_1$, ${\beta }_2$, ${\gamma }_1$, ${\gamma }_2$, and ${\gamma }_3$ are explained variable estimates of the coefficients. The individual fixed effect is represented by ${\mu }_i$, the time fixed effect by ${\delta }_t$, and ${ϵ}_{it}$ signifies the random error term.

5. Empirical Results and Analysis

5.1. Benchmark Regression Results Analysis

Table 3. Model estimation results for AT and ACEs.

Variables	(1)	(2)
	Fe	Fe
lntrade	−0.106 ***	−0.054 ***
	(−4.62)	(−2.82)
lnagr_gdp		0.447 ***
		(10.85)
lnmach		0.278 ***
		(8.20)
lnsoil		−0.181 ***
		(−7.24)
lnper_edu		−0.419
		(−1.51)
lnroad		−0.040
		(−0.82)
lnpost		−0.284 ***
		(−5.46)
lntech_market		−0.033 ***
		(−3.29)
lnrd		−0.022
		(−0.81)
Time fixed effects	Yes	Yes
Provincial fixed effects	Yes	Yes
Constant	7.527 ***	6.820 ***
	(153.28)	(10.56)
Observations	660	660
R2	0.144	0.536

Note: *** denote statistical significance at the 1% level; values in parentheses are t-values.

shows the return of the benchmark model results, in which model (1) represents the estimated results of the two-way fixed effects model without adding control variables, and model (2) represents the estimation results of the fixed effects model after adding control variables. From model (1), AT has a striking adverse influence on ACEs, after adding control variables. Moreover, the negative impact of AT on ACEs is still significant at the 1% level after adding control variables, and the impact coefficient is 0.054, which indicates that under conditions where other factors are held constant, every 1% increase in the volume of AT products correspondingly reduces the total of ACEs by 0.054%, and appropriately increasing the agricultural products’ foreign trade contributes to the reduction in agricultural carbon emissions. This result is consistent with the research conclusions of Li and Wang [8] and Ike et al. [9]. Thus, hypothesis 1 is verified.

By controlling for the empirical results of the variables, it can be found that the effects of the gross agricultural product, farming machinery total power, soil erosion control area, total postal and telecommunication business, and technology market turnover on agricultural carbon emissions all surpass the 1% significance level test. Among them, the total power of farm machinery and the gross agricultural products significantly increase agricultural carbon emissions. The reason is that along with the enhancement of the agricultural mechanization level and the expansion of agricultural production, the energy consumed and the pollution produced both increase, thus increasing agricultural carbon emissions, which conform with the practical agricultural production principles. On the other hand, the area of soil erosion control, the total volume of postal and telecommunication business, and the technology market turnover all significantly lower farming carbon emissions.

5.2. Robustness Test

The results of the benchmark regression partly confirm hypotheses 1 in the theoretical analysis section. To finely depict agricultural products’ import and export trade relationships with agricultural carbon emissions and ensure the robustness of the benchmark regression, this paper operates three regression means, such as excluding four municipalities directly under the central government, random sampling, and shrinking tails treatments on the data.

5.2.1. The Robustness Test of Removing Municipalities from the Sample

Since the four municipalities in China have a high level of urbanization and a meager share of farming, the degree of farming is quite different from that of the provinces, and the former’s carbon emissions are primarily generated in the secondary and tertiary industries. Thus, the impact of the trade in farming products on the municipalities’ farming carbon emissions is also relatively low. The regression results after excluding the four municipalities are presented in

Table 4. Robustness test results.

Variables	(1)	(2)	(3)	(4)
lntrade	−0.0285 **	−0.060 **	−0.047 **	−0.058 ***
	(−2.04)	(−2.58)	(−2.52)	(−3.58)
Control variables	Control	Control	Control	Control
Time fixed effects	Yes	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes	Yes
Constant	6.603 ***	6.415 ***	6.301 ***	6.326 ***
	(12.96)	(7.32)	(9.78)	(11.92)
Observations	572	396	528	660
R2	0.634	0.572	0.551	0.537

Note: **, and *** denote statistical significance at the 5%, and 1% levels, respectively; values in parentheses are t-values.

and model (1). According to the model, even if the data of municipalities are excluded, the trade in agricultural products still significantly suppresses agricultural carbon emissions at the 5% level. In addition, each control variable has no change in the direction of the coefficient except for a slight change in the size of the coefficient. Therefore, it can be argued that the trade in agricultural products can contribute to carbon reductions in agriculture.

5.2.2. Robustness Test for Random Sampling

The data of 30 provinces from 2001 to 2022 are randomly sampled at 60% and 80% levels, and then the sampling results are regressed, and the specific results are taken on columns (2) and (3) of Table 4. From the random sampling regression results, the trade in agricultural products negatively affects agricultural carbon emissions at a significance level of 5%, implying that the import and export trade of farm products can facilitate carbon emission reductions in farming, proving the robustness of the benchmark regression results.

5.2.3. Robustness Test with Shrinking Tail Treatment

To remove the impact of the abnormal extreme value on the result of the regression, the data are first subjected to a 1% shrinkage treatment. Then, we performed regression processing, and the results are displayed in four columns (4). It is not hard to see that after the 1% shrinkage treatment of the data, the trade in agricultural products still has a striking negative influence on agricultural carbon emissions, and the coefficients of the shrinkage treatment have little differences with those of the benchmark regression results, which proves the robustness of the benchmark regression results once again.

5.3. Endogeneity Test

Although the benchmark regression has controlled for some of the factors that influence farming carbon emissions, there is still the possibility of omitting crucial variables, leading to the endogeneity problem in the model. In this paper, the instrumental variable method was used to deal with the possible endogeneity problems. The selection of instrumental variables must satisfy two conditions: one is the correlation with endogenous variables and the second one is exogenous, i.e., the instrumental variables have no relevance to the residual terms. To guarantee the validity and reliability of regression results, this paper draws on the study of Li and Tang [20] to use the inverse of the distance of provincial capital cities from the coastline multiplied by the nominal exchange rate between the USD and CNY from 2001 to 2022 as one of the instrumental variables for agricultural trade. In addition, one period lag of agricultural trade was added as the second instrumental variable and evaluated by the 2SLS method. Column (1) of

Table 5. Endogeneity test result.

Variables	(1)	(2)
lntrade		−0.045 **
		(−1.96)
L.lntrade	0.920 ***
	(36.98)
Distance	2.306 ***
	(2.73)
Control variables	Control	Control
Time fixed effects	Yes	Yes
Provincial fixed effects	Yes	Yes
Observations	630	630
Under-identification test		444.815 ***
Weak identification test		684.569 ***
Over-identification test		p = 0.2063

Note: **, and *** denote statistical significance at the 5%, and 1% levels, respectively; values in parentheses are t-values.

shows that both agricultural trades lagged one period. The inverse of each province’s distance from the coastline exhibits a remarkable positive influence on agricultural trade, implying that agricultural trade in the previous period promotes the trade volume of agricultural products in the current period. In addition, the closer the distance to the coastline, the higher the trade volume of farming products will be. Upon the further testing of the validity of instrumental variables for the problem of under-identification, the test results rejected the original hypothesis that the instrumental variables are uncorrelated with the explanatory variables at the 1% significance level; the p-value of the over-identification test results was 0.2063, which rejected the hypothesis of endogenous instrumental variables. The above test results indicate that the instrumental variables selected in this paper are relatively appropriate. Model (2) in Table 5 shows that the second stage’s regression results are strikingly negative at the 5% level. The negative coefficient shows that based on weakening and controlling the endogeneity problem, trade in agricultural products still markedly lowers the amount of farming carbon emissions. Thus, the benchmark regression model results can be considered robust.

5.4. Heterogeneity Analysis

5.4.1. The Heterogeneity Analysis of the Three Economic Zones

With reference to the criteria for the division of the three major economic zones in China, 30 provinces (autonomous regions) were divided into eastern, central, and western regions to further study the heterogeneous influence of agricultural trade on agricultural carbon emissions, and the detailed results are displayed in

Table 6. Heterogeneity analysis Ⅰ.

	(1)	(2)	(3)
	Eastern	Central	Western
lntrade	−0.096 **	−0.002	−0.071 ***
	(−2.08)	(−0.23)	(−2.82)
Control Variables	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes
Provincial Fixed Effects	Yes	Yes	Yes
Constant	9.761 ***	8.866 ***	8.159 ***
	(7.70)	(11.64)	(8.02)
Observations	286	132	242
R2	0.728	0.926	0.611

Note: **, and *** denote statistical significance at the 5%, and 1% levels, respectively; values in parentheses are t-values.

. Table 6 reveals that in the eastern and western regions, the impact coefficients of the agricultural product trade are negative and pass at least the 5% significance test, demonstrating that within these two regions, agricultural product trade can produce a notable inhibitory effect on agricultural carbon emissions, i.e., the higher the amount of trade in agricultural products, the better the effect of the emission reduction on agricultural carbon emissions. In the central region, although the coefficient of the impact of trade in agricultural products on agricultural carbon emissions is negative, it does not pass the significance test, indicating that in the central region, trade in agricultural products does not have a clear impact on agricultural carbon emissions. Regarding the coefficient size, the suppression effect of agricultural trade on agricultural carbon emissions is stronger in the eastern region than in the western region. The reason for this may be that compared with the central and western regions, the scale of agricultural trade in the eastern region is generally larger, and the scale of agricultural imports tends to be larger than the scale of exports, which enables the eastern region to obtain more technological spillovers, as well as reducing carbon emissions in the production process. In addition, the eastern region has a large number of colleges and universities and abundant human resources, so the absorption capacity for foreign advanced technology will also be stronger. Therefore, compared with the other two regions, agricultural trade in the eastern region is more capable of curbing agricultural carbon emissions.

5.4.2. Heterogeneity Analysis of Coastal and Non-Coastal Regions

Foreign trade is highly dependent on maritime transport, so coastal areas have a natural transportation advantage over non-coastal areas. This paper further explores the heterogeneity of the impact of agricultural trade on agricultural carbon emissions in coastal and non-coastal areas. The results are exhibited in columns (1) and (2) in

Table 7. Heterogeneity analysis Ⅱ.

	(1)	(2)	(3)	(4)
	Coastal	Non-Coastal	Major-Grain-Producing	Non-Grain-Producing
lntrade	−0.216 ***	−0.021	−0.027 *	−0.098 ***
	(−6.07)	(−0.90)	(−1.83)	(−3.36)
Control Variables	Yes	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes	Yes
Provincial Fixed Effects	Yes	Yes	Yes	Yes
Constant	9.053 ***	6.425 ***	6.326 ***	7.350 ***
	(10.98)	(7.37)	(10.97)	(7.80)
Observations	242	418	286	374
R2	0.563	0.621	0.840	0.590

Note: *, and *** denote statistical significance at the 10%, and 1% levels, respectively; values in parentheses are t-values.

. By column (1), it can be concluded that agricultural trade has a negative impact on agricultural carbon emissions in coastal areas, and it is striking at the level of 1%. While in non-coastal areas, the effect of agricultural trade on agricultural carbon emissions is not remarkable. The reasons for this are similar to those for the heterogeneity of the three major economic zones. Coastal regions are the most economically developed regions in China and have more advanced production technologies for agricultural products, which may enable the reduction in agricultural carbon emissions. In addition, coastal regions usually set tougher standards for agricultural products that are imported and exported, which also pushes for carbon emission reductions in the production process. Thus, agricultural trade in the coastal region is more likely to curb agricultural carbon emissions.

5.4.3. Heterogeneity Analysis of Major Grain-Producing and Non-Grain-Producing Areas

The 30 provinces (municipalities directly under the central government) are divided into main grain-producing areas and non-grain-producing areas according to the division standard of the “Outline of Medium- and Long-Term Plan for National Food Security (2008–2020)”, and then we conducted regressions separately, and the results are shown in columns (3) and (4) of Table 7. From the results of columns (3) and (4), it can be seen that the coefficients of the impact of agricultural trade on agricultural carbon emissions in the main grain-producing areas and non-grain-producing areas are −0.027 and −0.098, respectively, and are significant at the levels of 10% and 1%, respectively. This indicates that trade in agricultural products can suppress agricultural carbon emissions, but the suppression effect is stronger in non-grain-producing areas than in grain-producing areas. The reason for this may be that in China’s main grain-producing areas, due to the huge scale of agricultural production, it tends to be a more large-scale and standardized production, and the technology of agricultural production is also more mature, which makes its carbon emissions less vulnerable to the influence of agricultural trade. In non-food-producing regions, the level of agricultural development is lower, and production technologies and patterns are relatively backward; therefore, their carbon emissions are more susceptible to technology spillovers from trade in agricultural products [24].

5.5. The Mechanism Test of the Impact of AT on ACEs

To deeply analyze the mechanism of the impact of AT on ACEs, this study will test the direct and indirect effects of AT in turn.

5.5.1. The Test of the Direct Effect of AT on ACEs

The theoretical analysis determines that agricultural trade can realize the transfer of carbon emissions through the trade substitution effect. In this paper, the impact of agricultural exports, agricultural imports, and the agricultural trade structure on agricultural carbon emissions was analyzed empirically, in which the agricultural trade structure is measured by the ratio of agricultural imports to agricultural exports. A larger percentage indicates a larger agricultural trade deficit; the specific results are demonstrated in

Table 8. Empirical results of the direct effect of agricultural trade on agricultural carbon emissions.

	(1)	(2)	(3)
lnexport	0.0179
	(0.82)
lnimport		−0.0296 **
		(−2.21)
lnstru			−0.0820 ***
			(−4.14)
Control variables	Control	Control	Control
Time fixed effects	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes
Constant	7.296 ***	7.098 ***	7.154 ***
	(11.19)	(11.17)	(11.39)
Observations	660	660	660
R2	0.531	0.534	0.543

Note: **, and *** denote statistical significance at the 5%, and 1% levels, respectively; values in parentheses are t-values.

. Model (1) in Table 8 is the impact of agricultural exports on agricultural carbon emissions. It can be seen that although the coefficient of the effect of agricultural exports on agricultural carbon emissions is positive, it does not pass the significance test, indicating no clear correlation between agricultural exports and agricultural carbon emissions. Although agricultural exports will theoretically intensify the competition of export enterprises and force domestic agricultural production to reduce their carbon usage under the background of carbon emission control in the international market [17], the expansion of exports of agricultural products will also enlarge the scale of domestic agricultural production, thereby expanding the consumption of energy, such as coal and crude oil, and ultimately leading to an enhancement in carbon emissions [24,25,26]. When the carbon-increasing effect, due to scale-ups, and the carbon-reducing effect, due to technological progress, counterbalance each other, the relationship between agricultural exports and agricultural carbon emissions may be unclear. It also suggests that domestic agribusinesses have made relatively little effort to cope with carbon regulation in international markets.

Model (2) shows the results of the impact of agricultural imports on agricultural carbon emissions. It can be found that the effect of agricultural imports on agricultural carbon emissions is remarkably negative at the 5% level, showing that the larger the scale of agricultural imports, the lower the agricultural carbon emissions. This is because China’s imports of agricultural products are mainly dominated by products such as soybeans, oilseed crops, dairy and eggs, and animal and vegetable fats and oils [23], which generate high carbon emissions in the production process. Transferring the production of high-carbon emission agricultural products to other agricultural exporting countries through imports will reduce the carbon emissions of the importing countries. With the expansion of the import scale, the quality and cost advantages of imported products will make the domestic agricultural business entities improve their own production and operation techniques and management methods, thus contributing to the further reduction in carbon emissions.

Model (3) demonstrates the empirical results of the impact of agricultural products’ import and export structure on agricultural carbon emissions. It can be observed that the import and export structure of agricultural products reduces agricultural carbon emissions at the 1% significance level, revealing that the larger the agricultural trade deficit, the more it can facilitate domestic agricultural carbon emission reductions. The above empirical analysis results prove that agricultural trade can directly reduce agricultural carbon emissions through the carbon substitution effect. It is worth noting that although expanding the trade deficit can decrease agricultural carbon emissions, in the consideration of international trade advantages and domestic food security, the trade deficit should be controlled within a reasonable range and should not be expanded indefinitely for the sake of carbon emission reductions. It should be adjusted to optimize the structure and types of import and export agricultural products and increase the imports of necessary agricultural products and high-carbon agricultural products; meanwhile, to improve the competitive advantages of domestic farm products in the international market, we need to accelerate the research and development of carbon reduction technology, reduce the energy consumption of agricultural production, and increase the added value of exported agricultural products. Therefore, hypothesis 2 is confirmed.

5.5.2. The Test of the Indirect Effect of AT on ACEs

This paper separately tested the impact of AT on agricultural carbon emissions by promoting technological innovation and reducing the energy consumption intensity, and the test results are presented in

Table 9. Empirical results of intermediate effects of agricultural trade on agricultural carbon emissions.

	(1)	(2)	(3)	(4)	(5)
lntrade	−0.0538 ***	2.067 **	−0.0381 **	−0.0214 **	−0.033 *
	(−2.82)	(2.54)	(−2.10)	(−2.55)	(−1.90)
lntech			−0.0076 ***
			(−8.39)
lnenergy					0.972 ***
					(11.56)
Control variables	Control	Control	Control	Control	Control
Time fixed effects	Yes	Yes	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes	Yes	Yes
Constant	6.820 ***	96.81 ***	7.555 ***	1.041 ***	5.808 ***
	(10.56)	(3.51)	(12.23)	(3.67)	(9.83)
Observations	660	660	660	660	660
R2	0.536	0.781	0.585	0.419	0.621

Note: *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively; values in parentheses are t-values.

. What needs to be explained is that this paper uses the number of agricultural patents accepted per 10,000 people to measure agricultural technological innovation, and the energy consumption per unit of agricultural output value measures the energy consumption intensity. The minor consumption per unit of agricultural output value reflects the higher efficiency of energy utilization in the agricultural production process, and the larger number of agricultural patent applications per 10,000 people indicates a higher level of agricultural technological innovation.

Model (1) in Table 9 proves the results of the benchmark regression of the impact of agricultural trade on agricultural carbon emissions, model (2) reveals the results of the impact of agricultural trade on agricultural technological innovation, and model (3) demonstrates the results of the empirical test of the impact of agricultural trade and agricultural technological innovation on agricultural carbon emissions. From the results of model (2), agricultural trade has a positive impact on agricultural technology innovation, and it is significant at the 5% level, illustrating that agricultural trade helps to promote agricultural technology innovation. From the results of model (3), it can be concluded that after adding agricultural technology innovations to the baseline regression model, the impact coefficient of agricultural trade on agricultural carbon emissions is reduced compared with the baseline regression results, but it is still significantly negative; meanwhile, agricultural technology innovation also has a significant suppressive effect on agricultural carbon emissions. This result explains that agricultural trade can inhibit agricultural carbon emissions by promoting agricultural technology innovation. Thus, hypothesis 3 is verified.

Model (4) in Table 9 shows the empirical results of the impact of agricultural trade on the agricultural energy consumption intensity, and model (5) demonstrates the empirical results of the effects of agricultural trade and the agricultural energy consumption intensity on agricultural carbon emissions. From model (4), it can be observed that agricultural trade negatively affects energy consumption efficiency at the 5% significance level, demonstrating that with the expansion of the scale of agricultural trade, the intensity of agricultural energy consumption is gradually decreasing, and the energy utilization rate is increasing. From the results of model (5), it can be concluded that after adding the intensity of agricultural energy consumption into the benchmark model, the coefficient of the impact of agricultural trade on agricultural carbon emissions decreased from 0.0538 to 0.033, but the direction of the effect did not change and passed the 10% significance test; the impact of the intensity of agricultural energy consumption on agricultural carbon emissions is positive at the 1% significance level, demonstrating that the greater the intensity of the energy consumption, the greater the agricultural carbon emissions. The above results suggest that agricultural trade can decrease the intensity of agricultural energy consumption, which in turn suppresses agricultural carbon emissions. Therefore, hypothesis 4 is verified.

In general, with the enlargement of the total trade in agricultural products, the level of agricultural science and technology innovation can be continuously improved, and the intensity of energy consumption is constantly reduced, which in turn promotes the reduction in agricultural carbon emissions. Therefore, the indirect effect of agricultural trade on agricultural carbon emissions can be verified.

6. Conclusions and Policy Recommendations

This study utilizes provincial panel data from 2001 to 2022 to conduct an empirical analysis of the impact of AT on China’s ACEs. The main findings indicate the following: First, AT significantly reduces China’s ACEs, and these results remain robust even after addressing endogeneity issues and conducting rigorous tests. Second, from a regional perspective, compared to central and western regions, non-coastal areas, and major grain-producing regions, the inhibitory effect of AT on ACEs is more pronounced in eastern regions, coastal areas, and non-major grain-producing regions. Third, from the perspective of impact mechanisms, AT can diminish ACEs through both direct and indirect effects. The direct effect manifests as an import substitution (IS) effect, while the indirect effects include agricultural technological innovation (ATI) and the energy consumption intensity effect (ECI).

Considering the above analysis, this study proposes the policy proposals listed below:

First, governments and policymakers must continue to enlarge the size of agricultural product trade and optimize agricultural products’ import and export structure. Meanwhile, the government has a responsibility to strengthen agricultural exchanges and cooperation with countries worldwide, especially by leveraging the development opportunities of the “Belt and Road” initiative. Government priorities should be heavily engaging in farm product trade with countries along the route and promoting technological innovation and resource sharing through collaboration. Simultaneously, the government can establish more efficient trade channels to reduce carbon emissions during transportation. Also, while ensuring food security, we recommend moderately increasing the import of high-carbon agricultural products and the export of low-carbon agricultural products. Through trade substitution, it is advisable to reduce carbon emissions from domestic agricultural production. In addition, we recommend enhancing the added value of domestically exported agricultural products to heighten their competitiveness in the international market.

Additionally, governments need to enhance the communication among regions and set up a regional platform for sharing agricultural technology and information to facilitate the spread of advanced agricultural technology from the eastern and coastal regions to the central and western regions, facilitating a coordinated agricultural development across different areas and enhancing the integral performance of agriculture production. Meanwhile, governments can consider emphasizing the complementary advantages of each region in agricultural resources, rationalizing the utilization of production resources, and achieving an optimal resource allocation. And governments should encourage the integration of agricultural industry chains across regions, forming a more complete and efficient agricultural industry chain to reduce carbon emissions during the transportation process and enhance the overall production efficiency of agriculture.

Furthermore, governments and policymakers should continue to promote the green transformation of energy and technological innovation in agriculture, support the transformation of agricultural energy use, enable the clean and efficient use of energy in agricultural production, and guide agricultural enterprises to adopt renewable energy, such as solar and bio-energy. Of course, governments can gradually shift the energy use in agricultural machinery from fossil fuels to new energy sources. Then, governments ought to increase investments in agricultural science and technology research and development and strengthen collaboration with agricultural colleges and research institutions to promote the development of smart agriculture, popularize advanced production technologies, such as intelligent irrigation and precision fertilization, and support the research and application of new types of fertilizers, pesticides, and seeds, which will improve the production efficiency and reduce carbon emissions.