Analysis of the international trade networks of COVID-19 medical products

In this paper, three databases are employed. The first is the source of international trade data related to medical products, while the economic, and geographic variables, as well as the COVID-19-related health data, are obtained from another two data sources.

The dataset compiled using the three data sources (see the Data sources section) is presented in Table 1.

More information about the variables of the Gravity database can be found at http://​www.​cepii.​fr/​DATA_​DOWNLOAD/​gravity/​doc/​Gravity_​documentation.​pdf (accessed: 14 July 2023).

The trading data is represented as a multilevel network (see,e.g., (Hammoud and Kramer 2020)). Multilevel networks include multiple layers that can contain a subset of all available nodes and edges. In our case, the eight product categories form eight layers, nodes are the countries, and directed edges represent their trading activities from the given product category. The weight of the edges within a product category is determined based on the difference between the values exported in 2020 and 2019 as follows:where \(\text {EXP}^{2020}_{i,j}\) is the aggregated exported value from country i to j in 2020, measured in thousand United States dollars (USD), \(i,j\in \{1,2,\dots ,L\}\) and \(L\in {\mathbb {N}}\).

Formally, the trading data is represented by a graph which is a tuple defined by the sets of nodes (\(\text {N}\)), edges (\(\text {E}\)), and eight layers (\(\text {S}\)) as follows:where \(K = 8\). The applied multilevel data structure is illustrated in Fig. 1.

To measure the extent to which the products’ total imported and exported values changed in each country between 2019 and 2020, the directed version of strength (i.e., weighted degree) centrality (see, e.g., (Yook et al. 2001; Barrat et al. 2004)) is applied. For each product category, the in- and out-strength centrality—measured in thousand USD—can be defined as follows:where \(\sum _{j=1}^{L} e_{j,i}\) and \(\sum _{j=1}^{L} e_{i,j}\) alone measure the in- and out-degree centrality of the node i, respectively.

In addition, to provide further details about the examined networks, we have also identified the strong and weak components of each layer. Strong components of a network are subsets of nodes where there is a directed path from any node to any other node within the subset, while weak components are subsets of nodes where there is a path (not necessarily directed) between any two nodes within the subset.

The gravity equation of trade defines trade as a positive function of the attractive “mass” of two economies and a negative function of the distance between them (Lewer and Van den Berg 2008). Similarly to other researchers (see,e.g., (Johnston et al. 2015; Hussain et al. 2021; Fontagné et al. 2022)), we use additional variables to control for demographic, geographic, ethnic, linguistic, and economic conditions. For each product category, the applied regression model is as follows:where \(P = 8\) and \(R = 9\) are the numbers of uni- and bilateral independent variables (apart from \(\log (EXP^{2019}_{i,j})\) and \(dist_{i,j}\)), respectively (see Table 1). The parameters, denoted by \(\alpha\), \(\beta\), \(\gamma\), and \(\delta\), are estimated by ordinary least squares (OLS) regression in which the \(\epsilon _{i,j}\) is the random error term.

The statistical programming language R is applied to compile and analyze the dataset presented in Table 1. To generate the figures and calculate the strength centrality measures, MuxViz (version: 3.1) R package (De Domenico et al. 2015; De Domenico 2022) is applied. This package enables the visualization and analysis of interconnected multilayer networks. More information can be found at https://​github.​com/​manlius/​muxViz (accessed: 14 July 2023).

Import and export growth between 2019 and 2020 is examined using in- and out-strength centrality measures, supplemented with other descriptive statistics that provide further insights into the investigated networks. First, these descriptive statistics are presented in Table 2, and then Fig. 2 illustrates the changes of the trade networks on a map. Figure 3 shows the twenty-five countries with the largest total import and export growth between 2019 and 2020 (i.e., the countries with the highest in- and out-strength centralities). Finally, the top five countries that have the largest import and export growth in each of the examined product categories are presented in Table 3 individually.

According to Fig. 3 and Tables 2 and 3, the traded value of protective garments (category G), medical test kits (category A), and disinfectants and sterilization products (category B) increased the most. China (CHN) alone increased its exports of protective garments (category G) by 71, 362, 502 thousand USD, which is approximately 20.38 times higher than the second-highest export growth in this product category, which was recorded in Malaysia (MYS). For these products, imports increased the most in the United States (USA), United Kingdom (GBR), and Germany (DEU)—22, 224, 368, 9, 287, 062, and 7, 321, 810 thousand USD, respectively; however, as Fig. 2 shows, most countries in the world had significantly increased imports of protective garments by 2020. In categories A and B, China (CHN) and Germany (DEU) achieved the highest export growth (6, 155, 374 and 3, 507, 406 thousand USD, respectively), while the highest import growths were recorded in the United States (USA) and Switzerland (CHE) (7, 060, 897 and 5, 100, 978 thousand USD, respectively).

The international trade of vehicles (category H) typically decreased during the investigated time period, and the related network was much more fragmented than that of other types of products (see Table 2). In this category, the highest export growth was only 67.694 thousand USD, which was reached by the United Arab Emirates (ARE). In the case of categories C to F, China (CHN) achieved the largest increase in exports by 3, 329, 381, 3, 708, 636, 1, 346, 074, and 3, 522, 334 thousand USD, respectively. While the highest import growth was recorded in the United States (USA) from categories C to E (1, 157, 565, 794, 627, 1, 089, 646 thousand USD, respectively), despite its population of only 9.7 million, Hungary (HUN) achieved the largest increase in imports of oxygen therapy equipment and pulse oximeters (category F) by 670, 942 thousand USD. This import growth is related to the Hungarian government’s procurement of more than 16, 000 ventilators from China (CHN).⁶

After aggregating each product category (see Fig. 3), it can be seen that China (CHN) achieved the largest export growth by 90, 833, 770 thousand USD, which is approximately 14.66 times higher than the second-highest cumulative export growth (6, 194, 989 thousand USD), which was recorded in Belgium (BEL). In the case of aggregate import growth, there was no such degree of asymmetry between the top two countries, which were the United States (USA) and Germany (DEU) by 31, 853, 163 and 15, 268, 697 thousand USD, respectively.

The results of the regression analysis are presented in Table 4.

Based on the results, it can be seen that the global F-test was highly significant for all models, while the adjusted \(\text {R}^2\) values varied between 0.199 and 0.805. This difference in adjusted \(\text {R}^2\) mainly reflects the stability of trade relations (network structure) between 2019 and 2020. Thus, e.g., the highest adjusted \(\text {R}^2\) was observed for medical test kits (category A), in which the logarithm of the 2019 and 2020 trading values had the highest Pearson correlation coefficient (\(\rho = 0.873\)). Regardless of product category, the significant values of RTA and the EU membership of the exporter countries increased the traded value, while the effect of the EU membership was shown to be the opposite on the importer’s side. The coefficients of WTO membership were only slightly significant. While they were typically positive on the importer’s side, the sign of the effect varied in the case of exporting countries. The GDP of both the exporter and importer countries increased the traded value, while the coefficient of the exporter countries was approximately 1.5–2.0 times greater than the coefficient of the importer countries.

Cumulative COVID-19 cases were significant in only half of the categories and typically reduced trade value, regardless of whether they were measured for exporting or importing countries. An exception to this was the disinfectants and sterilization products (category C), in which the higher COVID-19 cases in exporter countries typically increased the traded value. Based on these coefficients, the COVID-19 cases recorded in the exporting countries reduced the traded value of protective garments (category G) the most. In addition, the GII score of the exporting countries significantly reduced the trading value only in this product category. This result reflects that the production of protective garments (category G) had significantly lower technological demands than other investigated product categories. Finally, the contiguity of countries increased the traded value the most in the case of disinfectants and sterilization products (category B). After we controlled for this factor, the trade of medical test kits (category A) was the least distance-dependent category, while the distance between trading countries reduced the trade value of other medical consumables (category C) and protective garments (category G) the most.