Introduction

In the current global context, the world economy is under the dual pressure of structural adjustment and cyclical downturn. The international situation is challenging, with the environment becoming increasingly complex, and the global economic system is undergoing profound adjustments and reshaping. The resurgence of trade protectionism, tense geopolitical situations, and the contraction of trade activities have sharply increased the complexity and uncertainty of the global agricultural trade landscape (Anderson, 2022). Against this backdrop, the importance of international agricultural trade has become increasingly prominent. From a demand perspective, the improvement in living standards has led to a trend towards diversification and personalization in agricultural product demand, which is particularly pronounced in populous countries, where the growth rate and structural changes in agricultural product demand are significant. From a supply perspective, agricultural production is not only constrained by uncertain factors such as climate change but is also greatly influenced by policy (Malhi et al., 2021). Adjustments in import and export policies could directly impact the international trade of agricultural products and the domestic market supply (Sun et al., 2021). In this context, gaining a comprehensive understanding of the global agricultural trade structure becomes particularly significant.

With the introduction of social network analysis methods into the field of international trade, an increasing number of scholars have begun to utilize this approach to analyze the composition and evolution of international trade linkages. The structure and characteristics of international trade networks are particularly well-suited for research using social network analysis methods, and it is imperative to focus on the properties of complex networks when studying economic phenomena (Rauch, 2001; Fagiolo et al., 2010). In terms of regional focus, research extends beyond global trade networks to include analyses of RCEP member countries (Zhou et al., 2021), the European Union region (Jang and Yang, 2024), and the Belt and Road Initiative (Li et al., 2021). In terms of research subjects, in addition to agricultural products, various energy minerals are also primary objects of study (Sun et al., 2022; Zhao et al., 2021; Tian et al., 2021). Regarding research content, scholars initially concentrate on the node characteristics and topological features of trade networks, including degree distribution of nodes (Duan et al., 2022; Grassi et al., 2021), the small-world nature of trade networks (Xu et al., 2021; Li et al., 2023), and scale-freeness (Zhang et al., 2021; Ma et al., 2023). Subsequent research focuses on the analysis of trade network structures, with a particular emphasis on “core-periphery” structures (Zhu and Huang, 2023; Zhang and Batinge, 2021) and clustering effects (Wang et al., 2021; Ren et al., 2021; Wang et al., 2021). In the realm of agricultural products trade networks, some scholars have focused their analysis on specific categories of agricultural products, such as wheat (Gutiérrez-Moya et al., 2020), rice (Chen and Zhao, 2023), and soybeans (Ma et al., 2023). Beyond these, there are studies that have analyzed the international agricultural products trade network, with an emphasis on the dynamic evolution of the trade network structure (Dupas et al., 2019; Qiang et al., 2020; Duan et al., 2022; Chen and Zhang, 2022), the vulnerability of the trade network (Xu et al., 2024; Wang and Zhang, 2025), and its stability (Wang et al., 2023). In studying the determinants of agricultural trade networks, scholars commonly use panel regression methods (Sun et al., 2022), QAP model (Fu et al., 2021; Guo et al., 2023), and the Exponential Random Graph Model (ERGM) (Pu et al., 2021; Li et al., 2023). Existing research has laid the foundation for the analysis of agricultural trade networks; however, there are still deficiencies in the research on agricultural trade networks.

In summary, current research on agricultural product trade lacks detailed categorization. Agricultural products constitute a complex category, with different types exhibiting distinct trade characteristics. While existing research has focused on the structural evolution of the overall agricultural products trade network, it has overlooked the horizontal comparison across different types of agricultural products. Furthermore, existing studies have not adequately explored the community structure within the global agricultural product trade network, which hinders a clear representation of the current global trade landscape. When examining the determinants of agricultural product trade networks, the impact of endogenous network factors on these networks has been overlooked, and there is a lack of analysis on the differences between various types of countries. Utilizing trade data from 208 economies worldwide, this paper categorizes agricultural products into six types and delves into the trade network patterns and community distributions for each type. It also analyzes the key factors influencing the formation and evolution of trade networks and discusses how endogenous network factors (reciprocity) affect the networks. The primary issues addressed in this paper include: (1) What roles do key countries play in the trade networks of different agricultural product types? (2) What network patterns and community characteristics are displayed by the trade networks of different agricultural product types? (3) What exogenous factors influence the trade networks of the six types of agricultural products? How does network reciprocity influence trade networks, and what are the differential impacts of these factors on various types of countries?

The remainder of this paper is structured as follows: The next section introduces the data sources and research methods, including the construction of the trade network, the calculation of network indicators, and the ERGM model used for the analysis of influencing factors. Section “Characteristics of the agricultural product trade network” delineates the characteristics of the six agricultural product trade networks from three aspects: node characteristics, network structure features, and community analysis. Section “Analysis of factors influencing agricultural products trade networks” presents the empirical results of the analysis on the factors influencing the trade networks of the six agricultural product types. The final section concludes the paper and offers discussions.

Data and methodology

Data description

This study analyzes the global agricultural product trade flow data from 2013 to 2022, encompassing 208 countries (This paper uses “country” to refer to both sovereign states and non-sovereign entities with distinct economic policies). The data are sourced from the BACI-CEPII database. Utilizing the HS96 coding in the BACI-CEPII database as the primary data source, and referring to the classification method of the WTO’s Agreement on Agriculture, products classified under headings 01–24, 44–46, and 50–53 are defined as agricultural goods. Specifically, headings 01–05 are categorized as live animals and animal products, 06–14 as plant products, chapter 15 as animal and plant fats and oils, 16–24 as processed agricultural products, 44–46 as wood products, and 50–53 as textile materials. A detailed classification of agricultural products along with their commodity codes is provided in the Supplementary Table S1. In the analysis of influencing factors, data on GDP per capita, proportion of agricultural land, and arable land per capita are obtained from the World Bank’s WDI database. Information on geographic proximity, linguistic similarity, and trade agreements is sourced from the CEPII database, while government effectiveness and regulatory quality indices are derived from the World Bank’s WGI index.

Network construction and analysis

Trade network construction

This paper employs social network analysis to construct the global agricultural trade network. In this framework, trading countries are considered as network nodes, and the flow of agricultural products between countries serves as the edges. Together, these nodes and edges form the global agricultural trade network, denoted as G = (V, E, A, W). Here, \(V=\{{v}_{1},{v}_{2},\cdots ,{v}_{n}\}\) represents the set of nodes, which are the participating countries in global agricultural trade; \(S=\{{s}_{ij}\}\) is the set of edges, indicating the trade flow relationships between trading nations; \(A=\{{a}_{ij}\}\) denotes the adjacency matrix of global agricultural trade. W represents the weight of the trade network. In this study, the value of agricultural trade between countries is used as the weight to construct a directed and weighted global agricultural trade network.

Network indicators analysis

After constructing the agricultural trade network using the aforementioned method, the calculation methods and meanings of the relevant indicators are presented in Table 1.

Table 1 Calculation and implications of trade network indicators.
Full size table

Exponential random graph model

The evolution of the global agricultural product trade network is influenced by external factors, the attributes of the actors themselves, and the network’s intrinsic structure. The ERGM can describe the probability of edges occurring in the network, simulate the formation mechanism of the network, and thereby analyze the complex interdependencies within the network. The general form of the ERGM is as follows.

$$P(G=g|\theta )=\frac{\exp (\sum _{i}{\theta }_{i}{\delta }_{i}(g))}{c(\theta )}$$
(1)

In this formula, P represents the probability of the network’s occurrence, \(G=g\) denotes the observed network graph, \(\theta\) is the vector of model parameters, \(c(\theta )\) is the normalization function, which ensures that the sum of probabilities equals 1, \({\theta }_{i}\) represents the estimable coefficients for network statistics, and \({\delta }_{i}(g)\) corresponds to the statistics of the model.

Characteristics of the agricultural product trade network

Node characteristics

Macro-characteristics of global agricultural product trade

Based on the data obtained as shown in Fig. 1, the total value of global agricultural trade reached $2.336 trillion in 2022. Processed agricultural products accounted for the largest share, at 32%, with a trade value of $750 billion, followed by plant products at 29%. Textile materials constituted the smallest share, at only 4%, while both wood products and animal and plant fats and oils each accounted for 8% of the trade value. Examining the trends in the trade values of various agricultural products from 2013 to 2022, the proportions of the six categories remained relatively stable, with processed agricultural products and plant products consistently being the two largest in terms of trade value. The trade values of processed agricultural products, plant products, and live animals and animal products generally exhibited a trend of fluctuating growth, with a more pronounced upward trend emerging since 2020. The trade values of wood products, animal and plant fats and oils, and textile materials were relatively modest, with less distinct fluctuation trends. Notably, wood products and animal and plant fats and oils showed an upward trend in recent years, while the trade value of textile materials slightly declined over the course of the study period.

Fig. 1: Macro-characteristics of global agricultural product trade.
Fig. 1: Macro-characteristics of global agricultural product trade.
Full size image

a shows the proportion of the six categories in the global trade value of agricultural products. b presents the global trade value of the six categories of agricultural products from 2013 to 2022. Among the six categories of agricultural products, the trade value of processed agricultural products is the largest and shows an upward trend.

Analysis of node indegree and outdegree

Examining the indegree distribution of the nodes as shown in Fig. 2, the kernel density curves for all categories of agricultural products during the study period exhibited a right-skewed shape with increasing curve width and a declining trend in the height of the main peak. The peaks shifted to the left to varying degrees, indicating an overall decrease in the indegree levels and a widening gap in the indegree levels among countries. Notably, the main peak heights for “animal and plant fats and oils” and “processed agricultural products” decreased significantly, with a larger curve width, suggesting a widening gap between countries with higher and lower indegrees. This implies an expanding trend in the import disparities of these two product categories. Additionally, the kernel density curves for live animals and animal products, as well as processed agricultural goods, showed a trend from unimodal to bimodal, indicating a potential multi-tiered differentiation in the indegrees of countries. Overall, the import trade activities of agricultural products among countries displayed a trend from dispersion to aggregation, with a widening gap between major and minor importing countries, a phenomenon most pronounced in the case of processed agricultural goods.

Fig. 2: Network node indegree kernel density.
Fig. 2: Network node indegree kernel density.
Full size image

a–f Respectively display the kernel density curves of the indegree of trade network nodes for the six categories of agricultural products. The indegree kernel density curves of the six categories of agricultural products all show a right-skewed tail.

Regarding the outdegree distribution across countries as shown in Fig. 3, the kernel density curves for agricultural products exhibited a pronounced right-skewed tail with a multi-peak trend. Except for textile materials, the main peak values of the remaining agricultural products showed varying degrees of decline. Among the six categories, the kernel density curves for live animals and animal products, as well as plant products, slightly shifted to the right, indicating an overall increase in outdegree and more frequent export activities among countries. The main peak heights for plant products and “animal and plant fats and oils” decreased significantly, with an increase in width, suggesting a more dispersed outdegree distribution and a widening gap in export activities between countries. The longer right-skewed tails for “animal and plant fats and oils” and textile materials reflect that the export of these products is dominated by a few major countries, with most countries having limited export activities. Additionally, the pronounced multi-peak phenomenon for animal and plant fats and oils indicates a multi-tiered differentiation in export activities among countries. The kernel density curve for textile materials remained relatively stable during the study period, suggesting a more stable international trade market for this product. Overall, there were differences in export activities for the six categories of agricultural products among countries, with major countries exerting particularly significant control over the export of animal and plant fats and oils.

Fig. 3: Network node outdegree kernel density.
Fig. 3: Network node outdegree kernel density.
Full size image

a–f Respectively display the kernel density curves of the outdegree of trade network nodes for the six categories of agricultural products. The outdegree kernel density curves of the six categories of agricultural products all exhibit multi-modal phenomena to varying degrees.

Node centrality analysis

In the agricultural product trade network, the position and characteristics of nodes reflect a country’s trade status. Table 2 presents the top five countries in terms of harmonic closeness centrality and betweenness centrality for the six categories of agricultural product trade networks.

Table 2 Harmonious closeness, centrality, and betweenness centrality top five countries.
Full size table

Regarding harmonic closeness centrality, in 2013, the United States ranked first in plant products, animal and plant fats and oils, and processed agricultural products, while China topped the rankings in wood products and textile materials. France, India, and Germany also occupied leading positions in various agricultural products. By 2018, China maintained its dominance in the harmonic closeness centrality of its advantageous products, but experienced a slight decline in the rankings of live animals and animal products, plant products, and animal and plant fats and oils. Simultaneously, the Netherlands and Brazil saw significant rises in the rankings across multiple agricultural products, even surpassing the United States in some categories. By 2022, China remained at the top for wood products and textile materials, and its ranking for processed agricultural products improved, entering the top five overall. Turkey and Malaysia also saw increased rankings in various agricultural products, while the United States remained in the top five for all products except processed agricultural products. These results indicate that during the study period, China maintained close trade ties with other countries in wood products and textile materials, and its trade in processed agricultural products gradually expanded. Moreover, although Malaysia, the Netherlands, and Turkey began to stand out in multiple agricultural product trade networks, the United States still had more extensive and direct trade connections with other countries in most agricultural products.

For betweenness centrality, in 2013, China was among the top ranks for live animals and animal products, wood products, and textile materials, while the United States, France, Germany, and Canada were in the top five for all products except processed agricultural products and textile materials. By 2018, China’s ranking improved for processed agricultural products and wood products, entering the top five overall. The United States and France maintained their trade advantages, ranking in the top three for all six agricultural product categories, while Germany’s ranking slipped in several products. By 2022, China ranked first in wood products and textile materials but experienced a decline in live animals and animal products, as well as processed agricultural products. While the United States and France continued to exert their trade influence, the Netherlands’ ranking rose to the top three in plant products, animal and plant fats and oils, and processed agricultural products. These findings suggest that China played a mediation role primarily in wood products and textile materials in the international trade network. Throughout the study period, the United States and France consistently played significant “middleman” roles in multiple agricultural product trades, with fierce competition between China and the United States in textile materials. In recent years, the Netherlands has increasingly strengthened its role as a trade hub for certain agricultural products.

Overall, China has maintained strong international trade connections in the wood products and textile materials, indicating smooth international trade channels and high trade efficiency in these areas. In recent years, China’s trade status in processed agricultural products has gradually risen, reflecting the development of China’s agricultural processing industry. In terms of plant products, animal and plant fats and oils, and processed agricultural products, the United States, as a global agricultural powerhouse, has long occupied a leading position, maintaining extensive and direct trade connections with other countries. Meanwhile, the Netherlands, Brazil, and Malaysia are increasingly prominent in the agricultural product trade network, with the rise of emerging economies posing challenges to traditional agricultural powers, and the global agricultural trade landscape is evolving.

Network structure features

Basic characteristics of the trade network

From the perspective of network structural characteristics, the number of edges, network density, and average clustering coefficient can reflect the vitality of trade exchanges, the closeness of network node connections, and the frequency of triangular relationships. These data are shown in Table 3. During the study period, the number of trade network edges for all six categories of agricultural products exhibited a trend of initial increase followed by a slight decline. The number of network edges for processed agricultural products and plant products was relatively high, exceeding 15,000, indicating more active trade exchanges and frequent international interactions for these products. In contrast, the number of network edges for textile materials and animal and plant fats was less than 10,000, reflecting the smaller trade scope of these two product categories.

Table 3 Number of edges, network density, and average clustering coefficient of trade networks.
Full size table

Regarding network density for the six categories of agricultural products, processed agricultural products and plant products had higher network densities than the other products and showed significant growth from 2013 to 2018, indicating increased network connectivity. However, in recent years, the network density of these two product categories has slightly declined, suggesting a reduction in the closeness of node connections. The network densities for animal and plant fats and oils and textile materials were lower, around 0.2, with minimal fluctuation, indicating a relatively looser trade network for these products.

As for the average clustering coefficient, all six categories of agricultural products had coefficients exceeding 0.6 and showed an upward trend during the study period. The average clustering coefficient for processed agricultural products was the highest, surpassing 0.73 in 2022, reflecting a tendency among countries to form cluster relationships in agricultural trade, with the trade network of processed agricultural products having more trade clusters compared to other products. The changes in these indicators reflect an overall trend in the international agricultural product market from loose to tight trade connections, with a recent decline in closeness and an enhancement of regional cluster structures. The trade structures of processed agricultural products and plant products have become more concentrated.

Examining the average path length indicator in the trade network (Table 4), all agricultural products except textile materials exhibited a trend of slow decline followed by a slight rebound, whereas the average path length for textile materials fluctuated more significantly, indicating that the trade efficiency of this category is unstable and prone to volatility. The average path lengths for animal and plant fats and oils and textile materials were relatively high, both exceeding 1.8 during the study period, suggesting that more intermediate countries are required on average to establish trade connections for these products. In contrast, the average path lengths for processed agricultural products and plant products were lower, both below 1.65, reflecting higher trade transmission efficiency and contributing to reduced trade costs. Overall, the trade efficiency of agricultural products shows a trend of gradual improvement, with processed agricultural products and plant products having more mature transport paths, leading to more efficient international trade interactions and demonstrating a pronounced small-world phenomenon.

Table 4 Average path length of trade networks.
Full size table

Network transitivity and reciprocity

The transitivity and reciprocity of the trade network reflect the bidirectional and ternary relationships within the network (Fig. 4). Over the study period, both the transitivity and reciprocity of the six agricultural product trade networks fluctuated but generally increased, although subtle differences in network structural characteristics were observed among different agricultural product categories. In terms of transitivity, the change patterns across the six networks were somewhat similar, showing a steady rise from 2013 to 2019 with no significant fluctuations. However, from 2020 to 2022, a trend of decline, rise, and subsequent decline was observed, with particularly pronounced fluctuations in the trade networks of live animals and animal products, wood products, and textile materials. This may have been influenced by the COVID-19 pandemic, which disrupted bilateral trade relations in agricultural products. The most significant decline in transitivity was seen in live animals and animal products due to increased quarantine measures or adjustments in trade policies. In 2021, the transitivity of all agricultural product trade networks except wood products rebounded, surpassing the levels of 2019. By 2022, transitivity across the networks declined again, possibly due to the impact of the Russia–Ukraine conflict on the global agricultural market. Regarding reciprocity, the networks of plant products, animal and plant fats and oils, processed agricultural products, and wood products overall showed an upward trend during the study period. However, the reciprocity of plant products began to decline in 2022. Additionally, the reciprocity of live animals and animal products, as well as textile materials, experienced significant fluctuations, indicating that changes in the international situation and trade policies have a greater impact on the ternary trade relationships of these products.

Fig. 4: Transitivity and reciprocity of trade networks.
Fig. 4: Transitivity and reciprocity of trade networks.
Full size image

a–f Present the numerical variations of six agricultural product categories, respectively. The reciprocity and transitivity of the trade network reflect the bidirectional and triadic relationships within the network. Between 2013 and 2022, both the reciprocity and transitivity of the trade network for the six categories of agricultural products showed a trend of rising with fluctuations.

In summary, the closeness of the global agricultural trade network has undergone a transformation from concentration to dispersion. As the core of the agricultural trade network, processed agricultural products and plant products have higher transmission efficiency in international trade relations. The relative looseness of trade networks for products such as animal and plant fats and oils and textile materials suggests that these sectors face more uncertainty and challenges in the global trade landscape. Recently, the COVID-19 pandemic and geopolitical conflicts have posed obstacles to the development of the global agricultural trade network, particularly evident in the live animals and animal products sector. Nevertheless, the overall reciprocity of the agricultural trade network has shown an upward trend, with an increase in trade clusters and a strengthening of interdependence and complementarity among trading partners.

Network community structure

Utilizing community detection algorithms for modular analysis, the community structure of the global agricultural trade network can be identified. In consideration of the visual clarity and simplicity of the graph, this paper only includes the top 70 countries by degree for the illustration, as shown in Fig. 5. In the figure, the color of the nodes represents different communities, the size of the nodes reflects the outdegree of the nodes, and the thickness of the lines indicates the value of agricultural product trade. Table 5 presents the number and proportion of member countries in the top three communities for each category of agricultural products.

Fig. 5: Agricultural product trade network community division.
Fig. 5: Agricultural product trade network community division.
Full size image

a–l Show the global agricultural trade community changes for six product categories in 2013 and 2022. For the sake of visual clarity and simplicity in graphing, this figure selects only the top 70 countries by degree for illustration. The color of the nodes in the graph represents different communities, the size of the nodes reflects the outdegree of the nodes, and the thickness of the lines indicates the trade value of agricultural products.

Table 5 The top three communities in the agricultural product trade network, the number of member countries, and their respective proportions.
Full size table

Live animals and animal products

In 2013, the trade in live animals and animal products was divided into four communities, with the largest community comprising 96 member countries, primarily led by Asian and American nations such as China, the United States, and Canada. The second-largest community, centered around Germany and the Netherlands, although consisting of only 65 countries, accounted for the majority of the top 70 countries in terms of degree centrality. In 2022, the structure of trade communities underwent adjustments, resulting in a decrease in the number of member countries in the Asia-America community. Concurrently, trade activities among Asian countries, including China, Japan, and Thailand, remained vibrant. From 2013 to 2022, the Asia-America community consistently formed the core of the live animals and animal products trade, although there has been a decline in the number of member countries in recent years.

Plant products

In 2013, the largest community in the plant products trade market was the Asia-America community, led by China and Brazil. This community engaged in significant trade exchanges both internally and with the North American community, which was led by the United States, Mexico, and Canada. Additionally, the European community, headed by Germany, Italy, and the United Kingdom, also had a considerable number of member countries. By 2022, the North American community’s presence in the trade market had contracted, with substantial trade exchanges primarily occurring within the Asia-America community and between China and the United States. Throughout the study period, the Asia-America community, especially the trade exchanges between China and Brazil, maintained stable and significant trade volumes. The European community consistently sustained active trade activities.

Animal and plant fats and oils

In 2013, there were six communities in the trade of animal and plant fats and oils, with the most influential being the Asia-Europe community centered around Indonesia, Spain, and the United Kingdom, and the Asian community led by China, Malaysia, and India. Substantial trade exchanges also took place between these two communities. By 2022, the Europe community, with the United Kingdom, France, and Belgium at its core, expanded its influence, reaching 57 member countries. Simultaneously, due to the pressure from the Asia-Europe community centered on Turkey and Indonesia, the community that China belonged to saw a reduction in size, with only 29 member countries. The overall trade trend indicates that the trade in animal and plant fats and oils is relatively complex, with trade clusters being more dispersed. The Asia-Europe community has consistently been the largest, with Indonesia remaining a prominent export country in this sector.

Processed agricultural products

In 2013, the trade of processed agricultural products was predominantly conducted within the Asia-America community, centered around the United States, China, and Japan, as well as the European community, with Italy and Spain at its core. The significant trade exchanges were mainly concentrated between the United States and Canada. By 2022, the influence gap among the four major communities, excluding the African community, had narrowed. The North American community, led by the United States and Mexico, although having a slightly lower number of member countries, exhibited substantial internal trade volumes and maintained close trade ties with the Asian community, centered around China, Thailand, and South Korea. Between 2013 and 2022, there were significant changes in the communities within the processed agricultural products sector, with Europe and North America remaining the primary trade hubs.

Wood products

In 2013, the largest community in the wood products trade market was the Asian community, centered around China and Thailand, with a total of 85 member countries. Following this was the European community, with Spain, Italy, and France at its core, also encompassing agricultural powerhouses such as Germany, Turkey, and Austria. By 2022, six communities had formed, with some countries from the original Asian community, including Indonesia, Japan, and Malaysia, joining with others like Turkey and Romania to establish a new Asia-Europe community, comprising 71 member countries. The influence of both the Asian and European communities was reduced. The overall trend indicates that the trade in wood products has progressively transcended geographical boundaries, with increasingly close interactions between Asian and European countries.

Textile materials

In 2013, the global textile materials market was dominated by two major communities. One was the Asia-America community, centered around China, India, and the United States, with 115 member countries, including Mexico, Japan, and Brazil, where China was the primary importer. The other was the European community, led by the United Kingdom, Germany, France, and Italy, encompassing 83 member countries. By 2022, the core communities in the global textile materials trade market had evolved into three main groups: the Asian community, led by China and Indonesia; the Asia-America community, led by the United States, India, and Canada; and the European community, led by France, Italy, and Germany. The member country counts of these three communities were relatively close, with significant trade exchanges occurring between the Asian and Asia-America communities. Throughout the study period, the Asia-America community maintained a considerable influence in the international market, while the Asian community, particularly with China at its forefront, was rapidly emerging.

The evolution of the community distribution across various agricultural product categories shows that live animals and animal products, plant products, and animal and plant fats and oils exhibited distinct geographical characteristics in the initial period, with countries in close geographical proximity engaging in more intensive trade. As trade globalization advanced, intercontinental communities such as Asia-Americas or Asia-Europe emerged. Textile materials were predominantly Asian communities, while processed agricultural products were mainly concentrated in European and American communities. The trade in wood products progressively broke through geographical limitations, with Asian and European countries engaging in increasingly close trade ties.

The impacts of COVID-19 and the Russia–Ukraine conflict

The COVID-19 pandemic and the Russia–Ukraine conflict have significantly reshaped international agricultural trade patterns in recent years. To examine the structural changes in global agricultural trade networks induced by these two disruptive events, this study analyzes key network characteristics from 2019 to 2022. Table 6 presents the temporal variations in network density, average clustering coefficient, average path length, reciprocity, and transitivity. Following the COVID-19 outbreak in December 2019, both network density and average clustering coefficient declined in 2020, indicating reduced trade connectivity and more fragmented trade activities. The increased average path length further reflects lower trade efficiency among nations. A slight decrease in reciprocity and transitivity also suggests weakened bilateral and clustered trade interactions. In 2021, the average path length shortened, while other metrics showed modest recovery, signaling a resurgence in trade activity. However, the sudden escalation of the Russia–Ukraine conflict in 2022 triggered another disruption, mirroring the 2020 trend: diminished network clustering, reduced efficiency, weaker reciprocity, fewer trade subgroups, and increased overall fragmentation.

Table 6 Changes in key network metrics.
Full size table

Figure 6 presents the structural evolution of global agricultural trade communities between 2019 and 2022. In 2019, the network exhibited three dominant trading blocs: an American-centric community led by the United States and Canada, a European community centered around France, Germany, and the Netherlands, and a third major community consisting primarily of Asian nations with partial representation from Europe, America, and Africa. The COVID-19 pandemic in 2020 induced significant network fragmentation, increasing the total number of trade communities. Notably, the China-Brazil oriented community experienced reduced membership as nations including Turkey and Egypt disengaged to form a new Afro-Asian trading bloc, while the American and European communities maintained structural stability. By 2021, the Afro-Asian community demonstrated expansionary tendencies, accompanied by revitalized high-volume trade flows between major economies such as the United States and Canada, as well as China and France, indicating preliminary signs of global trade recovery. The geopolitical disruption caused by the Russia–Ukraine conflict in 2022 precipitated a distinct structural shift: the Afro-Asian community underwent contraction, while the European community exhibited enhanced internal cohesion. Concurrently, cross-community high-volume trade transactions diminished significantly, with the notable exception of sustained agricultural trade between the United States and China. This transition reflected an emerging preference for intra-community trade relations among most nations. In summary, the initial pandemic shock generated network dispersion and the emergence of new Afro-Asian trade clusters. The transient recovery observed in 2021 subsequently yielded to a pronounced regionalization trend in 2022, characterized by strengthened European integration and reduced global trade interconnectedness.

Fig. 6: Evolution of global agricultural trade network communities from 2019 to 2022.
Fig. 6: Evolution of global agricultural trade network communities from 2019 to 2022.
Full size image

a–d Present the evolution of global agricultural trade communities from 2019 to 2022. This figure illustrates the impacts of the COVID-19 pandemic and the Russia–Ukraine conflict on global agricultural trade communities. This figure selects only the top 70 countries by degree for illustration. The color of the nodes in the graph represents different communities, the size of the nodes reflects the outdegree of the nodes, and the thickness of the lines indicates the trade value of agricultural products.

Analysis of factors influencing agricultural products trade networks

Variable selection

Dependent variable

Considering the availability and completeness of the data, this study selects the global agricultural trade network in 2021 as the research subject. The raw data include a vast amount of bilateral trade data, forming a complex and extensive network of agricultural trade. To more accurately highlight the significant bilateral trade relationships between countries, this paper employs the Jaccard index to transform the trade network into a network of correlation coefficients. The formula is as follows.

$$Jaccar{d}_{ij}=\frac{{x}_{ij}}{{x}_{i+}+{x}_{+j}-{x}_{ij}}$$
(2)

In the formula, \({x}_{ij}\) represents the trade flow from country i to country j, \({x}_{i+}\) denotes the total export value of country i to all other countries, and \({x}_{+j}\) indicates the total import value of country j from all other countries. The conversion using the Jaccard index serves to eliminate relatively insignificant agricultural trade relationships and to identify the network connections among major agricultural trading nations, which helps to more directly reflect the strength of trade interconnectivity between countries. This paper sets the threshold of the Jaccard index at 0.02 (Li and Liu, 2022), applying a binarization process to obtain a 0–1 matrix.

Reciprocal effects (Reci)

Reciprocal effects are generally more pronounced in directed networks and are used to measure the closeness of interactions among individuals within the network. In the agricultural trade network, reciprocity reflects the tendency of two countries to export products to each other. The emergence of reciprocal effects primarily impacts the evolution of the agricultural trade network through trade balance mechanisms, transportation cost mechanisms, and information mechanisms. First, the trade balance mechanism. Trade imbalances between countries can potentially lead to trade friction, whereas the existence of reciprocal relationships can reduce this possibility to some extent. Additionally, countries may promote imports from others by entering into reciprocal trade agreements or offering tariff preferences (Carrère, 2006; Baier and Bergstrand (2007)), thereby achieving the goal of maintaining trade balance. Second, the transportation cost mechanism. Countries can reduce transportation costs by jointly building transportation infrastructure and sharing logistics service resources (Clark et al., 2004). For example, sea transportation is the main mode for international agricultural product shipping; if there are return shipments after reaching the destination, this can reduce the rate of empty returns and decrease transportation costs. Moreover, reciprocity can effectively increase the volume of product transportation between two countries, achieving economies of scale. Third, the information mechanism. International trade interactions facilitate the exchange and sharing of information between countries, reducing information asymmetry and transaction uncertainty, which can lead to the establishment of more trade relationships.

Univariate exogenous variables

The incorporation of univariate exogenous variables can reveal how a country’s intrinsic attributes influence the agricultural trade network. This paper selects five variables for analysis from three dimensions: economic development, resource endowment, and institutional environment. Firstly, according to the Comparative Advantage Theory, a higher level of a country’s economic development implies better conditions for engaging in international trade, which facilitates product import-export exchanges. This study uses the per capita GDP matrix to represent a country’s level of economic development. Secondly, according to the Resource Endowment Theory, a country’s relative endowment disparities lead to final price variations in products, conferring comparative disadvantages and advantages in different product categories, ultimately shaping international trade and the division of labor (Li, 2012). This paper employs the proportion of agricultural land and per capita cultivated land area to reflect a country’s agricultural resource endowment. Lastly, government and institutional factors influence a country’s international trade to some degree. The government’s strength and speed in policy implementation, along with an effective and fair regulatory system, can maintain domestic market stability and adjust the conduct of foreign trade (Iwanow and Kirkpatrick, 2007). This paper uses the government efficiency index and the regulatory quality index to reflect a country’s institutional factors. The univariate exogenous variables give rise to three basic effects: sender effects, receiver effects, and similarity effects. Sender effects reflect the impact of a country node’s attribute variables on its agricultural product exports; receiver effects reflect the impact of a country node’s attributes on its agricultural product imports; and similarity effects reflect the propensity of two countries with similar node attributes to establish trade relationships. The formulas for these three basic effects are as follows.

$${V}_{sender}={\sum }_{j}{X}_{ij}{V}_{i}$$
(3)
$${V}_{receiver}={\sum }_{j}{X}_{ij}{V}_{j}$$
(4)
$${V}_{similarity}={\sum }_{j}{X}_{ij}(si{m}_{ij}-\overline{sim})$$
(5)

In this formula, \(si{m}_{ij}=\left(1-\frac{|{V}_{i}-{V}_{j}|}{{\max }_{ij}|{V}_{i}-{V}_{j}|}\right)\circ\)

Binary Exogenous Variables. Incorporating binary exogenous variables into the model can reveal whether inter-country relationships impact the agricultural trade network. This paper selects three variables based on geographical distance, trade agreements, and cultural differences. Firstly, the trade distance between countries can influence the frequency of trade transactions to some extent, as transportation costs are directly related to geographical distance, thereby affecting trade relationships (Disdier and Head, 2008). This study uses a geographical proximity matrix to reflect the factor of distance between countries, with adjacent countries coded as 1 and nonadjacent countries as 0. Secondly, the presence of trade agreements between countries can reduce trade barriers and facilitate trade exchanges (Liu, 2007). This paper employs a trade agreement matrix to reflect the status of trade agreements between two countries, with the presence of an agreement coded as 1 and the absence as 0. Lastly, the impact of cultural differences on international trade is increasingly prominent. Language can reflect a country’s culture, and countries sharing a common language have higher communication efficiency, which in turn has a certain degree of influence on the development of trade relationships (Guo, 2004). This paper uses a common language matrix to reflect cultural differences between countries, with countries sharing a common language coded as 1 and those without as 0.

Result

This paper employs the Markov Chain Monte Carlo Maximum Likelihood Estimation (MCMC MLE) method for estimation. The rationale for this choice is that the data under consideration comprises a large-scale agricultural products trade network spanning over two hundred countries. MCMC MLE is particularly adept at handling high-dimensional data and complex network relationships. It constructs a Markov chain to simulate the distribution of parameters and employs stochastic sampling techniques for an extensive search, thereby yielding parameter estimates close to the global optimum. This method is widely used in the parameter estimation of ERGMs, and its efficacy and reliability have been validated (Saul and Filkov, 2007; Xu et al., 2021). The fitting results are presented in Table 7. As the fundamental effect in the ERGM, the edge function is analogous to the intercept term in traditional regression models, with the estimated coefficients typically being negative.

Table 7 Empirical results of ERGM for six types of agricultural product.
Full size table

For the six categories of agricultural products, reciprocity, geographical proximity, cultural differences, and trade agreements all exhibit significant positive impacts. This reflects the effective roles of trade balance mechanisms, transportation cost mechanisms, and information mechanisms in facilitating international trade. Countries that are geographically adjacent, share a common language, or have signed trade agreements are more likely to develop trade relationships. Moreover, the sender effect of GDP per capita exhibits a significant negative impact in Model 3 and Model 5, while its receiver effect is positive at the 1%, 1%, and 5% significance levels in Models 2–4, respectively. This indicates that countries with higher GDP per capita are less likely to export animal and plant fats and oils and wood products but more inclined to import plant products, animal and plant fats and oils, and processed agricultural products. Except for Model 6, the similarity effect of GDP per capita is significantly negative at the 0.1% level, suggesting that countries with similar levels of economic development are less likely to engage in trade for most agricultural products. The fitting results of agricultural land proportion and per capita cultivated land area share some commonalities. Both sender effects are significantly positive in Models 2–4 and Model 6, and their similarity effects are significantly negative in Models 3–6. This suggests that countries with abundant agricultural resources are more likely to export most agricultural products and prefer to establish trade links with countries less endowed with agricultural resources. The impact of agricultural resource endowment on the export of agricultural products is more pronounced for animal and plant fats and oils, processed agricultural products, and textile materials. The sender effect of government efficiency is significantly positive at the 1% level for all six categories of agricultural products, and its receiver effect is only significantly positive in Models 2–6, indicating that countries with higher government efficiency are more inclined to export and import a broad range of agricultural products. Additionally, the sender effect of regulatory quality is significantly negative in Models 1 and 6, and its receiver effect is significantly negative in Model 2 and Model 6. This suggests that countries with higher regulatory quality are less likely to export live animals and animal products, as well as textile materials, and to import plant products and textile materials. This may be due to the fact that higher regulatory quality often implies stricter health and safety standards and more stringent animal protection regulations, which can restrict the import and export of related agricultural products.

Differential analysis

To explore the differences in the impact of various factors on agricultural trade among different types of countries, this paper categorizes the countries studied into three groups: OECD member countries (OECDs), developing countries (DVPIs), and least developed countries (LDCs). The LDCs are identified according to the list published by the United Nations. Countries that are not OECD members or LDCs are classified as DVPIs. Due to space constraints, this paper only analyzes the factors influencing the total agricultural trade of these three types of countries, without further categorizing agricultural products into the six categories. The results are shown in Table 8.

Table 8 Empirical results of ERGM for different types of countries.
Full size table

The results indicate that there are certain similarities as well as distinct differences in the factors influencing agricultural trade among the three types of countries. Geographic proximity exerts a significant positive impact on agricultural trade for all three types, suggesting a preference for trading with neighboring countries. For OECDs and DVPIs, reciprocity and the similarity effect of GDP per capita are significantly positive and negative, respectively. This suggests that reciprocal trade promotes agricultural trade for both types of countries, while OECDs and DVPIs are less likely to engage in agricultural trade with countries of similar economic development levels. This may be due to greater comparative advantages between countries with different economic levels, such as medium-economy countries focusing on labor-intensive agricultural products and higher-economy countries specializing in capital and technology-intensive products. Additionally, differences in income levels and consumption habits between economically diverse countries lead to varying demands for agricultural products. Notably, the sender effect of regulatory quality is significantly positive for OECDs but significantly negative for DVPIs, indicating that OECDs are more inclined to export agricultural products to countries with higher regulatory quality, whereas DVPIs show the opposite tendency. This could be because OECDs typically have higher production standards and quality control systems, meeting the import requirements of countries with high regulatory standards, while DVPIs, constrained by costs and technology, may find it challenging to meet these standards, thus having a lower probability of exporting to such countries. Furthermore, the similarity effect of agricultural land proportion is significantly negative for DVPIs but significantly positive for LDCs, suggesting that LDCs prefer to trade with countries that have similar agricultural resource endowments, whereas DVPIs exhibit the opposite preference. The reason may be that LDCs face more urgent food security issues and thus are more likely to trade with similar countries to ensure basic food supplies, while DVPIs, having already secured domestic food security, are more inclined to use international trade to optimize resource allocation and increase income.

Conclusions and discussion

Conclusions

This paper analyzes the node individual characteristics and network pattern evolution of the global trade networks for six categories of agricultural products—live animals and animal products, plant products, animal and plant fats and oils, processed agricultural products, wood products, and textile materials—based on trade data from 2013 to 2022. Using social network analysis methods and the ERGM, we identify the main factors influencing the trade networks and draw the following conclusions.

  1. (1)

    In terms of node individual characteristics, the core countries in agricultural product trade are relatively stable and predominantly located in European, American, and a few Asian countries, with an expanding gap among countries. States such as the United States, France, Germany, Turkey, and Malaysia occupy central positions in the trade networks for live animals and animal products, plant products, animal and plant fats and oils, playing significant roles as intermediaries. China maintains considerable influence in the wood products and textile materials networks and serves as an important bridge. Moreover, China’s trade influence in processed agricultural products has been increasing annually. The gap in agricultural product import and export activities between major and minor countries is widening, with export activity differences slightly greater than import activity differences. The export of animal and plant fats and oils shows a multi-tiered differentiation phenomenon.

  2. (2)

    In terms of network pattern evolution, all agricultural product trade networks exhibit small-world characteristics, with major communities displaying distinct geographical traits. The transitivity and reciprocity of these trade networks are on the rise, with the networks for live animals and animal products and textile materials showing significant fluctuations in reciprocity. Additionally, the average clustering coefficient of these trade networks has increased, while the average path length has decreased, indicating small-world characteristics. The regional cluster structure and small-world nature are more pronounced in the processed agricultural products network. The major communities in these trade networks have clear geographical features, notably in Asian and European communities, which play a significant role in the trade of animal and plant fats and oils and wood products.

  3. (3)

    Regarding the influencing factors of the trade networks, reciprocity, geographical proximity, cultural differences, and trade agreements all significantly enhance trade connections for various agricultural products. Beyond these, the similarity effect of GDP per capita exhibits a significant negative impact on all agricultural products except textile materials. The sender effects of per capita cultivated land area and government efficiency index show positive impacts on all agricultural products, highlighting the important role of agricultural endowment and governance efficiency in a country’s agricultural product exports.

Discussion

  1. (1)

    Core countries in the agricultural trade network exhibit a certain degree of stability, primarily concentrated in Western and a few Asian countries. This reveals an imbalance in global agricultural trade (Tombe, 2015; Cai and Song, 2016), where a small number of countries dominate, while others occupy more peripheral positions. In particular, countries like the United States, France, and Germany play a brokerage role in multiple agricultural product categories, giving them greater influence in the global agricultural supply chain. This imbalance is also observed in international trade markets for semiconductors (Ren et al., 2023), crude oil (Niu et al., 2023), and fossil fuels (Zhong et al., 2016). The trade imbalance is mainly due to the stronger economic foundation, industrial capacity, and more mature political and economic systems of core countries, enabling them to maintain central positions in international trade consistently.

  2. (2)

    Global trade exhibits varying degrees of clustering, with countries tending to form trade communities of different sizes (Ji et al., 2014). The formation of these communities demonstrates certain geographical characteristics (Maluck and Donner, 2015). While the establishment of trade communities can help countries enhance trade efficiency and increase trade interactions (Lee and Shin, 2006), the strengthening of regional cluster structures may also lead to the emergence of trade protectionism and regional barriers, making the global trade landscape more insular and imbalanced (Fan et al., 2014). particularly under the impact of the recent COVID-19 pandemic and the Russia–Ukraine conflict, international trade has shown signs of contraction and turbulence (Hassani and Dost, 2020; Zhou et al., 2023). Therefore, how to optimize the structure of international agricultural trade and break down regional barriers is a question that warrants attention in the future.

  3. (3)

    The geographical proximity, cultural differences, and trade agreements between countries all have a significant impact on agricultural trade, which aligns with findings in the existing literature (Ghazalian, 2015; Vollrath et al., 2009). However, the differences between various types of agricultural products and countries at different levels of development have been overlooked. For instance, agricultural endowments have a substantial influence on the international trade of textile materials but a lesser impact on processed agricultural products, often determined by the distinct manufacturing processes of agricultural goods. The production of textile materials primarily relies on natural resource endowments such as land and climate, whereas the manufacturing process of processed agricultural products is more dependent on processing and manufacturing stages, thus being less influenced by agricultural endowments. Moreover, differences in national development levels also lead to variations in trade behavior. For more developed countries, the stability and effectiveness of governance are often key factors they consider in trade interactions (Gani, 2011; Ribeiro et al., 2025). In contrast, for less developed countries, ensuring domestic food security is the primary concern (Sun and Zhang, 2021).