International trade resilience and the Covid-19 pandemic

Abstract

The Covid-19 pandemic represents a low-probability, high-impact systemic risk that has severely disrupted international trade, reshaping the patterns of globalization. Drawing from the concept of supply chain resilience, which involves both the ability of a system to withstand an impact (robustness) and recover from it (responsiveness), we investigate country-level trade resilience during the 1st wave of the pandemic. By employing Fuzzy-set Qualitative Comparative Analysis (fsQCA), we identify configurations of country-level factors, i.e., country profiles, based on their effectiveness in engendering trade resilience. These factors include social and economic globalization, logistics performance, healthcare preparedness, national government response, and income level. The results show how these factors coalesced to strengthen (or weaken) international trade resilience, contributing to a holistic understanding of the impact of the pandemic on international trade. The findings inform the post-Covid-19 debate on international trade, with implications for managers and policymakers.

1. Introduction

The Covid-19 pandemic is one of the most profound crises of our time. The socio-economic impact of the crisis has been devastating, and repercussions will continue to unfold in years to come. Undoubtedly, the loss of life has been the most poignant consequence of the pandemic, but the economic impact has also been overwhelming. Several articles have suggested that Covid-19 will end globalization altogether (Economist, 2020, Yip, 2021), while others argue that it will at least alter its course (Altman, 2020). What is clear is that while globalization has brought many benefits to the world's economy, it has also exposed nations, firms, and individuals to systemic supply chain risks (Goldin and Mariathasan, 2015, Scheibe and Blackhurst, 2018). This type of risk relates to events that can cause a widespread and sustained shortage of a product or service with no alternatives or substitutes available (Sheffi & Lynn, 2014). The Covid-19 crisis is an extreme example of such a risk, which appears to have forced globalization into retreat.

It is perhaps not surprising that highly globalized countries like the UK, Italy, France, and the USA, were hit fast and hard (Ahluwalia, 2020). The economic and social connections that engendered globalization have also reinforced interdependencies that enabled the spread of the virus (Mas-Coma, Jones, & Marty, 2020). In response, many countries closed their borders, and uncooperative behavior emerged as governments competed to secure access to scarce resources, such as personal protective equipment (PPE), ventilators, and vaccines (Chowdhury et al., 2020, Times, 2020). This jockeying for position has caused additional ripple effects across global supply chains. While the level of globalization might have had a detrimental effect in the early days of the pandemic, many highly globalized countries also have substantial resources, trade links, logistics capabilities, and healthcare infrastructure that may have helped them achieve a swift recovery (New York Times, 2020).

Countries have followed different paths in response to the pandemic. Some, such as China, imposed strict regional restrictions; others, like Sweden, adopted a more laissez-faire approach, yet others, like the UK, changed track several times through the crisis (IMF, 2021, Mayer and Lewis, 2020, Reuters, , 2021). It is still unclear which path will be more effective in the long run. Thus, we need an evidence-based approach to understand the most effective ways to deal with systemic risks, such as the Covid-19 pandemic.

Considerable research has focused on the resilience of firms (e.g., Ambulkar et al., 2015, Parker and Ameen, 2018, Dormady et al., 2019) and supply chains (e.g., Blackhurst et al., 2011, Melnyk et al., 2014, Sheffi, 2005, Zsidisin and Wagner, 2010). Similarly, researchers have investigated economic resilience at a country level (e.g., Rose and Liao, 2005, ROSE, 2007, Xie et al., 2018). However, research into the resilience of the international trading system in which firms, supply chains, and countries operate, has been lacking. This gap is surprising given that in today's globalized economy, international trade resilience appears to be intricately linked to the resilience of countries, supply chains, and firms.

Countries exchange goods and services through imports and exports within the international trading system. In this context, we define international trade resilience as “the ability of a country to both resist disruptions to international trade and recover after disruptions occur” (cf. Melnyk et al., 2014: 36). This definition is helpful because it captures two distinctive aspects of resilience: the ability to resist a disruption; and the ability to recover from it. In this research, we explore the role of different country-level factors in both aspects of resilience: robustness to withstand the initial impact and responsiveness to facilitate the recovery in the context of international trade during the Covid-19 pandemic. Specifically, we aim to address the following research question:

• What configurations of factors led to trade resilience amidst the Covid-19 pandemic?

To address this question, we employ Fuzzy-set Qualitative Comparative Analysis (fsQCA), an analytic technique that uses Boolean algebra and fuzzy set theory to address causal complexity (Beynon et al., 2016, Fiss, 2011, Rihoux and Ragin, 2008, Woodside, 2014). We follow inductive logic to evaluate the role of different country-level factors (e.g., globalization, logistics capabilities, income level, healthcare infrastructure, and national government response), and configurations thereof, in eliciting trade resilience. To this end, we build a unique dataset after combining secondary data from several sources, including the World Bank, World Trade Organization (WTO), the Swiss Economic Institute, and the Johns Hopkins University (JHU). The research focuses on the first wave of the pandemic (March–July 2020). While most countries have suffered multiple waves to date, we argue that the first wave was truly unpredictable and unprecedented, and provides a unique context for examining international trade resilience.

We inductively derive configurations of factors leading to international trade resilience, laying the foundations for a midrange theory (Eisenhardt and Bourgeois, 1988, Hoffman and Ocasio, 2001, Crilly, 2011, Craighead et al., 2016) of this phenomenon in the context of the Covid-19 pandemic. Midrange theorizing involves context-specific conceptualizations and thus provides theoretically grounded insights applicable to a specific empirical context (Craighead et al., 2016). Moreover, midrange theories proved narratives of causal processes and the conditions under which those processes generate outcomes (Russo, Pellathy, & Omar, 2021). These characteristics make midrange theorizing a suitable approach to investigate international trade resilience during the pandemic.

We contribute to the literature in several ways. Firstly, we show that no single factor is necessary or sufficient for trade resilience during a disruption of this scale; rather, it is complex combinations of (high or low levels of) those factors that strengthen or weaken resilience. Secondly, we help explain why some countries remained comparatively trade resilient during the first wave of the pandemic, while others did not. Finally, we complement studies of resilience at a firm, dyad, and network level, by adopting a macro-level perspective, using the country as the unit of analysis. This macro-perspective provides a contextual explanation for the resilience of firms and supply chains operating across borders.

2. Literature review

The Covid-19 pandemic has caused ripple effects across every aspect of human life (Verma & Gustafsson, 2020), upending the business environment and decimating international trade (Verbeke & Yuan, 2021). This Covid-19 phenomenon is characterized by complexity and uncertainty, influencing (and being influenced by) government policies, health systems, firm behavior, and individual behaviors and decisions (Bratianu, 2020, Pappas and Glyptou, 2021). Some authors have used the label ‘chaordic’, emphasizing the chaotically-ordered character of the phenomenon (Pappas, 2021). While the complexity and uncertainty of the Covid-19 pandemic have affected the business environment globally, these effects have not been homogeneous across countries. Some countries suffered dramatically in terms of trade, while others remained relatively resilient to the disruption. It is this resilience, in the face of a highly complex and uncertain environment, that we investigate in this research.

Resilience is a construct that can be applied at multiple levels. From a personal (micro) level (e.g., Luthar, Cicchetti, & Becker, 2000), through to large scale (macro) systems like economies (e.g., ROSE, 2007, Rose and Liao, 2005), and societies (e.g., Cacioppo, Reis, & Zautra, 2011), researchers have tried to understand what makes a system resilient. While there are differences in the conceptualization of resilience at various levels, they all refer to a system's ability to contend with a disruption.

Researchers have distinguished two distinctive sets of capabilities of a resilient system (Melnyk et al., 2014; Välikangas, 2010, Wieland and Wallenburg, 2013). On the one hand, a resilient system can withstand a disruption. Researchers have referred to this capability as robustness, defined as the ability of a system to maintain its function despite internal or external disruptions (Bode et al., 2014). On the other hand, a resilient system also needs to respond after a disruption and return to its original state or move to a new, more desirable state (Christopher & Peck, 2004); this is the system's responsiveness. Robustness and responsiveness are interdependent, but as Melnyk et al. (2014) argue, they can sometimes trade-off against each other, so an increase in robustness could undermine responsiveness and vice versa. For this reason, it is crucial to investigate how different factors affect both robustness and responsiveness.

In this research, we evaluate key factors that can influence both the robustness and responsiveness of the international trade system and the countries that exchange goods and services within the system. In the next subsection, we discuss different factors that might influence a country's ability to participate in international trade, affecting global supply chains and the firms within them.

2.1. Globalization and resilience

Over the past decades, global supply chains have expanded in line with increasing levels of globalization, leading to higher interconnectedness and interdependence among firms (Blackhurst et al., 2005, Christopher and Holweg, 2011). While the interdependence has enhanced supply chain efficiency with practices of lean manufacturing, concurrent engineering, and “just-in-time” deliveries (Soni & Jain, 2011), it has also introduced supply chain vulnerabilities (World Economic Forum, 2019).

The Covid-19 pandemic prompted an unprecedented global stock-out of highly demanded life-saving medical equipment and PPE (Burki, 2020). The crisis has highlighted the vulnerability of interdependent economies and subsequent risks to supply chains. For example, approximately 97% of antibiotics used in the United States are imported from China (MSCI, 2020), while 40–50% of generic drugs come from India, with nearly 70% of those drugs’ active pharmaceutical ingredients (APIs) originating from China (CNBC, 2020). In turn, Chinese manufacturers of ventilators experienced critical production delays due to shut-downs of European sub-suppliers’ production units. While the Covid-19 pandemic highlighted the vulnerability of healthcare supply chains, many other sectors experienced the flip side of globalization, with global demand plummeting.

Globalization was initially driven by relocating subsidiaries or by outsourcing to foreign suppliers. As globalized business models matured, companies gradually offshored more critical business processes through integrated networks of interdependent subsidiaries and suppliers (Contractor et al., 2010). This process of globalization has, in turn, increased the impact of potential disruptions (Christopher, Mena, Khan, & Yurt, 2011) since countries have become increasingly dependent upon each other for goods and services. However, globalization is a multi-dimensional construct. In the DHL Global Connectedness Index, economic indicators of globalization (trade and capital flows) showed steady growth until the 2008–09 global financial crisis and have since fluctuated below their pre-crisis peaks. In contrast, the social aspects of globalization (information and people flows) have been setting new records (Steven & Philip, 2020). In the context of a pandemic involving human-to-human contagion, this broad view of globalization is a key country-level factor for determining vulnerability and exposure. As the WEF Global Risk Report (2006: 4), “the vulnerabilities of our interconnected global system would intensify the human and economic impact” (of a pandemic).

2.2. Logistics performance and resilience

The performance of a country’s logistics and transport system is central to international trade (Martí et al., 2014, Ekici et al., 2016) and the smooth functioning of global supply chains (Closs & Mollenkopf, 2004). Various aspects contribute to logistics performance at a country level, including the quality of trade and transport infrastructure (e.g., ports, roads, airports), the efficiency of customs (ease and speed of clearance), and the level of technology adoption. In turn, these factors can influence the economic outcomes of a country or region (Kurth et al., 2020) and the performance of all supply chains in and out of a country (Arvis et al., 2018, Closs and Mollenkopf, 2004). Thus, in the face of a global disruption, weak logistics infrastructure and competencies can undermine a country’s robustness in terms of international trade.

A country’s logistics infrastructure can be vulnerable to disasters, including a pandemic (Goldin & Mariathasan, 2015). A survey of port authorities and operators worldwide investigating the impact of Covid-19 indicates that many ports have been affected by changes in demand, capacity constraints, labor shortages, and delays caused by changes to procedures (Notteboom & Pallis, 2020). Moreover, bottlenecks can emerge because of the limited flexibility of logistics. For instance, the lack of temperature-controlled infrastructure in some countries has been highlighted as a barrier to vaccine distribution (Wight, 2020). Conversely, a strong logistics system at a country level can also help in recovery, allowing to bring products to the right places at the right time.

The pandemic has caused instability in supply and demand, causing pressure on logistics systems, causing shortages (e.g., PPE, ventilators) (Shih, 2020, Tatelbaum, 2020). As Notteboom and Pallis (2020:3) assert, “port demand is a derived demand,” and thus, changes to demand and supply of products directly impact activity levels at ports. This dependent demand argument can be extended to all elements of a national logistics system, including roads, airports, customs offices, and individual logistics providers. Consequently, a country’s resilience to a crisis will likely be affected by the ability of its logistics system to cope with swings in supply and demand.

2.3. Income level, healthcare preparedness, and resilience

The concept of healthcare preparedness has gained special attention in the context of the Covid-19 pandemic, as governments were forced to lock down economies to reduce case numbers and subsequent hospitalizations, trying to avoid the collapse of overwhelmed healthcare systems. The preventive measures of governments, and the associated economic trade-offs, highlight the linkage between the preparedness of healthcare infrastructure and economic activity (Jovanović et al., 2020). The pandemic effectively stress-tested the capabilities of national healthcare systems, which partly determined the scope and depth of lockdown periods and other preventive measures. In the same vein, a well-prepared healthcare system can support a quicker recovery through, for example, more effective testing and vaccination rollouts (Hale, Webster, Petherick, Phillips, & Kira, 2020b).

Another important factor determining a country’s level of trading activity is its financial ability to withstand extended lockdowns (Chowdhury et al., 2020). This can take the form of national relief packages, access to credit for businesses, or the ability to draw upon higher levels of personal savings. Wealthier countries are also better positioned to secure access to critical resources, such as PPE, medical equipment, and vaccines. For many low-income economies, the crisis revealed that without either political influence or spending power, securing access to scarce global healthcare production capabilities was almost impossible. Due to the lack of equity in access to affordable healthcare products, the pandemic, therefore, had a disproportionate impact across countries (New York Times, 2020). On this basis, a country’s financial strength is considered a key factor in enabling international trade resilience.

2.4. National response and resilience

National governments responded to the Covid-19 pandemic using different approaches, such as containment measures (e.g., lockdowns, workplace closures, and travel bans), health measures (testing, contact tracing, vaccines), economic measures (e.g., income and debt support, workforce retention), and social measures (e.g., strengthening social dialogue) (Hale, Petherick, Phillips, & Webster, 2020a). As the pandemic unfolded in different countries, the scope and stringency of these measures adapted, to balance healthcare, economic, and social outcomes.

The biggest challenge for countries is that trade-offs exist between desirable outcomes. Lockdowns and travel bans, for example, have been effective in containing the spread of the virus and allowed healthcare systems to cope, particularly for nations like New Zealand, Vietnam, and Taiwan (Frieden, 2021). Similarly, in China, strict local lockdowns have kept the virus at bay (Reuters, 2021). However, containment measures can also have a crushing economic impact. The effects of these measures labeled ‘great lockdowns’ have been compared to the great depression (Gopinath, 2020). Undoubtedly this has affected employment, income, and the supply and demand for many products, undermining international trade.

Countries like South Korea have implemented a complex combination of targeted lockdowns, extensive testing, border closures, quarantines, contact tracing, and economic relief packages to balance the various impacts of the pandemic (Frieden, 2021, IMF, 2021). Yet other countries, most notably Sweden, have kept schools and businesses open and even discouraged the use of face masks (IMF, 2021, Vogel, 2020). While this approach has economic benefits and probably buoyed Sweden’s economic outcomes in the early stages of the pandemic, it appears to have significant shortcomings related to healthcare outcomes, which may undermine the country’s recovery (Vogel, 2020).

While it is too early to say which countries and policies have been most effective, ultimately, all countries are connected in the global trading system. Hence, even if some managed to contain the virus and limit its toll on human lives, it is unlikely that their trade volumes have remained unaffected. Hence, it is vital to understand how the stringency and scope of government responses have affected international trade in different countries.

2.5. A neo-configurational approach to country-level trade resilience

Given the complexity of the Covid-19 pandemic, the diversity of impacts on different countries, the varying levels of preparedness for it, and the multitude of approaches followed to counter it, the role of each theorized condition should not be considered in isolation. Instead, they should be thought of as fundamental parts of complex constellations, or country ‘profiles,’ as part of which the role of each dimension will be dependent on the role of others. As such, the Covid-19 pandemic calls for configurational approaches that allow researchers to investigate the complex interactions of various conditions (Pappas & Glyptou, 2021)

In this research, we argue that some of the configurations of the theorized conditions will be more effective in ensuring robustness and fostering the responsiveness of international trade, and our aim is to identify them. A neo-configurational lens (Fiss, 2007, Misangyi et al., 2016, Woodside, 2013), premised on conjunction (i.e., trade resilience results from the interdependence of country-level factors), equifinality (i.e., different configurations might be equally effective in eliciting trade resilience), and asymmetry (i.e., a factor may play a crucially positive role as part of one configuration but might be irrelevant in another) is well-suited to tackle the complexity of the phenomenon. With respect to the set of ‘tenets,’ codified by Woodside (2014) and adopted in several business research publications (e.g., Olya and Altinay, 2016, Pappas, 2021), we expect that:

• •There is no country-level factor that is singlehandedly sufficient for the presence of trade resilience (Tenet 1).
• •Rather, there is (more than one) configuration of consistently sufficient (but not necessary) factors for the presence of trade resilience (Tenet 2 & 3 - the recipe and equifinality principles).
• •The sufficient configurations for the negation of the outcome (i.e., absence of resilience) will not be mirror opposites of the configurations sufficient for the presence of trade resilience (Tenet 4 - the causal asymmetry principle).
• •Both the presence and the absence of a factor can contribute to the presence of trade resilience as part of different recipes, depending on the presence or absence of other factors in those recipes (Tenet 5).
• •No configuration of factors leading to trade resilience will be relevant for all trade-resilient countries (i.e., coverage < 1.00 for any single configuration – Tenet 6), while there might be countries with high membership in a sufficient configuration that do not exhibit trade resilience (i.e., deviant cases – Tenet 7).

3. Methodology

The damage caused by the Covid-19 pandemic on global trade has no precedent, precluding the application of traditional theories of risk and resilience. In this research, we decided to adopt an inductive approach to develop a midrange theory of the phenomenon of trade resilience under exceptional circumstances. Specifically, we follow a bottom-up strategy for midrange theorizing, where the driving force for theory development is the data (Craighead et al., 2016). This strategy is considered appropriate for exploratory research as it relies extensively on induction to identify patterns in the data (Craighead et al., 2016). As a result, we use the extant literature not to formulate hypotheses but to motivate the selection of possible conditions that, conjunctively, might foster trade resilience. We then follow previous inductive research of configurational nature (e.g., Beynon et al., 2016, Crilly, 2011, Pajunen, 2008, Park et al., 2020, Woodside, 2014), by applying fsQCA for data analysis.

FsQCA has rapidly established itself as a valuable and systematic approach to comparative social inquiry due to its ability to address causal complexity. The use of set theory and Boolean algebra enables the conceptualization of a case as a complex combination of theoretical attributes and the identification of attributes (and configurations thereof) that are necessary and/or sufficient for an outcome (Misangyi et al., 2016, Ragin and Rubinson, 2009). Furthermore, the calibration stage allows for the incorporation of substantive knowledge to capture variation across cases that is relevant to the particular research question and defined sets (Ragin, 2008); while the organization of cases (based on their shared set-membership scores) into a Truth Table reveals the commonalities across them and facilitates counterfactual analysis, i.e., a conscious examination of whether an unobserved configuration would lead to the outcome if it was empirically present (Ragin and Sonnett, 2004, Soda and Furnari, 2012). The subsequent algorithmic minimization process leads to a set of three theoretically meaningful solutions (complex, parsimonious, intermediate) that, at the same time, remain ‘true’ to the data (for a general introduction and step-by-step exposition of fsQCA, see Schneider & Wagemann, 2012).

We follow the growing number of fsQCA applications in business research that use the country as the unit of analysis (Estevão et al., 2020, Piñeiro-Chousa et al., 2019, Tekic and Tekic, 2021) and seek to identify configurations of factors that might have helped countries achieve trade resilience over the first wave of the COVID-19 pandemic (March–July 2020).

3.1. Data, sample, and measures

Country-level data from various sources were combined into a unique dataset. First, monthly data (01/2019–11/2020) of total merchandise imports and exports for 74 economies was downloaded from the World Trade Organization’s database inventory (WTO, 2021). Trade resilience of a country in a given month is measured by adding exports and imports (scaled with the country’s 2019 GDP) and taking the Year-on-Year (YoY) monthly percentage (%) change. For example, the measure takes a positive (negative) value for a country that saw an increase (decrease) in trade volume in March 2020 compared to March 2019. The average YoY monthly % change of our sample ranges from −10.15% (June) to −29.14% (April). The largest value observed for an individual country is 21.36% (Ireland in March) and the smallest −62.63% (Bolivia in April).

Logistics performance is measured through the Logistics Performance Index (LPI) developed by the World Bank (Arvis et al., 2018). The index is a weighted average of six indicators (Customs, infrastructure, service quality, international shipments, tracking and tracing, and timeliness), ranging from a minimum of 1 to a maximum of 5. Based on the quintiles of the distribution of the LPI, Arvis et al. (2018) classify countries into ‘logistics-unfriendly,’ ‘partial performers,’ ‘consistent performers,’ and ‘logistics-friendly.’

Healthcare preparedness is measured through the Global Health Security (GHS) index, which was developed in 2019 by the Nuclear Threat Initiative and the Johns Hopkins University (NTI & JHU, 2019), in collaboration with The Economist Intelligence Unit (EIU). The index is designed to assess a country’s capability to prevent and mitigate epidemics and pandemics and takes values from 1 to 100. Each country is assessed using 140 questions organized across six categories, and the overall GHS Index score is a weighted sum of all these. The weights were determined by a panel of international experts. Based on the results, countries were classified into ‘most prepared,’ ‘more prepared,’ and ‘least prepared.’

Economic and Social Globalization are measured through the respective overall indexes developed by the Swiss Economic Institute (KOF) of ETH Zurich (Gygli, Haelg, Potrafke, & Sturm, 2019). Economic Globalization has two sub-domains: trade and financial; while Social Globalization has three: interpersonal contact, information flows, and cultural proximity. Various de facto and de jure indicators for each sub-domain are statistically weighted to form the overall indexes, which take values from 1 to 100.

Government response is measured through the index developed by the University of Oxford (Hale, Webster, Petherick, Phillips, & Kira, 2020b), which systematically tracks the wide range of policies that governments have taken during the pandemic. It takes values from 1 to 100 and is based on 18 standard indicators that capture (predominantly) health, containment, and economic support measures.

Gross National Income (GNI) per capita in 2019 (in current US dollars) is used to measure a country’s income level. Data was downloaded from the World Bank data repository (The World Bank (2019), 2019).

To account for the magnitude of the pandemic-induced disruption and the fact that countries were asymmetrically and asynchronously affected by the pandemic's first wave, we consider the Covid-19 related deaths per 100,000 inhabitants per month as an additional condition. Monthly data of cases and deaths is continuously gathered by JHU and was sourced through Oxford’s government response tracker website. We use deaths rather than cases due to the renowned inconsistencies of testing regimes across countries and to the fact that countries introduced mass testing at different points in time. As such, using case numbers would have introduced intractable biases both across countries and across months.

Of the 74 countries with monthly trade data (i.e., the outcome), four have missing values in at least one of the remaining variables (i.e., conditions), so our final sample size is reduced to 70. The list of countries in the sample is included in Appendix C.

3.2. Measure calibration

FsQCA requires the calibration of the raw measures into fuzzy sets using meaningful thresholds. In this way, every case is assigned a score (ranging from 0 to 1) to denote the degree of its membership in the defined sets (see Ragin, 2008). Descriptive statistics for the original measures are presented in Table 1 , their pairwise correlations are in Table 2 , and all the details specific to the calibration can be found in Table 3 . As seen in columns 4–5 (and endnotes), where available, we use external, theoretically derived thresholds developed by experts and/or used in past research (Greckhamer et al., 2018). It is worth noting that even though our final sample is limited to the 70 countries that have full data availability, to calibrate the conditions and outcome for which no external information exists to assist calibration, we used all available data of their respective measures. In conjunction with the theoretically (based on expert knowledge) and time-sensitive, empirically driven qualitative anchors (see Table 3), this decision increases the study's external validity by introducing information and knowledge ‘external’ to the final, limited sample. The calibration of the time-varying measures (trade resilience, government response, number of deaths) every month, based on their respective distribution, also accounts for possible qualitative changes in the series (e.g., structural breaks) and trends (see Flaherty, 2019).

Table 1.

Descriptive statistics.

Outcome/Conditions	Original measure	Month	Mean	Std. Dev.	Min.	Max.	N
Trade resilience (RESI)	Year-on-year monthly % change in total trade volume	March	−10.21	11.46	−51.14	21.37	74
		April	−29.14	13.81	−62.63	0.19
		May	−28.48	10.18	−57.9	−2.4
		June	−10.15	9.33	−34.44	11.43
		July	−11.39	7.76	−31.57	4.06
Logistics Performance (LPI)	Logistics Performance Index	–	2.88	0.58	1.95	4.2	151
Income Level (GNI)	GNI per capita in current USD (Atlas method)	–	15563.66	21129.32	280	117,730	161
Government response (GOV)	Overall Government Response Index (Uni. Of Oxford)	March	31.74	11.3	3.28	64.24	183
		April	62.07	12.3	14.26	86
		May	61.99	12.22	14.64	83.04
		June	57.42	13.16	13.33	81.17
		July	54.59	13.34	14.55	80.34
Magnitude of disruption (DEATHS)	Number of monthly deaths per 100,000 ppl	March	1.08	6.21	0	76.63	179
		April	3	8.38	0	59.44
		May	1.47	3.17	0	17.02
		June	1.26	2.91	0	24.24
		July	1.76	4.09	0	28.34
Economic Globalization (ECOG)	KOF Economic Globalization Index	–	58.84	16.25	25.64	93.63	170
Social Globalization(SOCG)	KOF Social Globalization Index	–	63.96	17.81	26.7	91.5	176
Healthcare preparedness (GHS)	Global Health Security Index	–	42.01	14.27	16.6	83.5	170

Table 2.

Pairwise correlations of raw measures in the analysis (70 countries – 5 months).

Measures	1	2	3	4	5	6	7	8
1. RESI	1
2. LPI	0.184***	1
3. GNI	0.169***	0.755***	1
4. GOV	−0.465***	−0.11**	−0.15***	1
5. DEATHS	−0.197***	0.2***	0.153***	0.241***	1
6. ECOG	0.172***	0.598***	0.654***	−0.197***	0.084	1
7. SOCG	0.177***	0.686***	0.813***	−0.185***	0.101*	0.867***	1
8. GHS	0.152***	0.767***	0.574***	−0.071	0.233***	0.444***	0.572***	1

Notes: *p < 0.1, **p < 0.05, ***p < 0.01.

Table 3.

Calibration details.

Outcome/Conditions	Original measure	Corresponding fuzzy set	Method of calibration	Anchors1
RESI	Year-on-year monthly % change in total trade volume (March to July)	The set of countries that exhibited above-average trade resilience in the first wave of the pandemic	Direct (based on distributional properties)2	e.g. April: Excl. −42.83% Cross. −28.6% Incl. −11.2%
LPI	Logistics Performance Index	The set of countries with high logistics performance	Direct (using World Bank set thresholds for level of ‘logistics friendliness’ – Section 3.1)	Excl. 2.38 Cross. 2.88 Incl. 3.41
GNI	GNI per capita	The set of high-income countries	Direct (using World Bank set thresholds defining income levels4)	(in USD) Excl. 1036 Cross. 4045 Incl. 12,535
GOV	Overall Government Response Index	The set of country governments that exhibited an above-average strong response	Direct (based on distributional properties)2	e.g. April: Excl. 47.86 Cross. 62.01 Incl. 75.58
DEATHS	Deaths per 100,000 ppl	The set of countries that experienced an above-average disruption impact	Direct (based on distributional properties)2	e.g. April: Excl.: 0 Cross. 3 Incl. 6.39
ECOG	KOF Economic Globalization Index	The set of economically globalized countries	Direct (following precedence)3	Excl. 42.89 Cross. 57.78 Incl. 74.91
SOCG	KOF Social Globalization Index	The set of socially globalized countries	Direct (following precedence)3	Excl. 45.4 Cross. 66.09 Incl. 82.66
GHS	Global Health Security Index	The set of well-prepared countries in the event of a world pandemic	Direct (using ghsindex.org thresholds for level of preparedness – Section 3.1)	Excl. 33.5 Cross. 40.2 Incl. 66.5

¹Incl. denotes the anchor of ‘full membership’ in the target set, Cross the point of ‘maximum ambiguity’ (i.e., 0.5), and Excl. ‘full non-membership’. See Ragin, 2008, Schneider and Wagemann, 2012.

²Thresholds being the mean (cross-over) and 10th and 90th percentiles of the distribution of the measure. Since the measure varies by month, for every month included in the analysis, the measure is re-calibrated based on its distribution in the given month.

^3FollowingGygli et al. (2019)who use the quintiles of the distribution to draw comparisons between countries, we use the top and bottom quintile to denote full and no membership, respectively, and the median as the cross-over point.

⁴https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups.

3.3. Analytical steps

Following convention (Dusa, 2018, Schneider and Wagemann, 2012), we first test whether any of the conditions can be considered necessary for the outcome or its negation. As such, we apply a consistency threshold of 0.9 and a relevance threshold of 0.6. No condition (or any disjunction thereof) in any month passes both thresholds, but the presence of high-income can be considered ‘trivially’ necessary for trade performance in every month, with a consistency score of >0.9 and a relevance score of <0.31. In July only, both social and economic globalization surpass the consistency threshold and come close to be considered relevant (scores of 0.445 and 0.582, respectively). These results suggest that as the first wave dissipates, high trade performance is almost impossible for a country that is not high-income and globalized. For the negation of trade performance, no condition comes close to be considered necessary.

The main analysis consists of identifying configurations that are sufficient for the presence and absence of trade resilience, while no single factor is, by itself, sufficient for either outcome (confirming Woodside’s 1st Tenet). Each month is analyzed separately (i.e., one Truth Table per month), following Aversa, Furnari, and Haefliger (2015) approach to accounting for time in fsQCA. The results are synthesized in Section 4. We apply the Enhanced Standard Analysis (ESA) of Schneider and Wagemann (2012) using the QCA package in R (Dusa, 2019). This means that contradictory simplifying assumptions and simultaneous subset relationships are identified and removed from the Truth Table before generating the parsimonious and intermediate solutions. When it comes to the inclusion/exclusion of empirically observed configurations, as a rule of thumb, we apply a consistency threshold above the conventional 0.80 mark and a frequency threshold of 1. In this way, we seek to include a reasonable number of countries that could be considered to exhibit the outcome (trade resilience or its negation). These details are presented in Table 4, Table 5 in the next section. In the few instances where more than one (parsimonious or intermediate) solution exists, only the one with the highest consistency score is presented. The existence of multiple sufficient configurations (whose consistency scores are always <1.00) for both the presence and the absence of trade resilience for every single month indicates support for Woodside (2014) Tenets 2,3, and 6.

Table 4.

fsQCA results: Analysis of sufficiency for trade resilience.

Month (2020)	Intermediate Solution1,2,3 (consistency, raw/unique coverage)	Solution consistency (coverage)4	Typical Cases5	Consistency & PRI cut-offs6 (Cases covered)
March	1. ∼GOV*GHS*ECOG*SOCG*LPI*GNI (0.879, 0.356/0.174) 2. ∼GOV*∼DEATHS*ECOG*SOCG*∼LPI*GNI (0.864, 0.168/0.025) 3. ∼GOV*GHS*∼DEATHS*∼ECOG*LPI*GNI (0.897, 0.17/0.031)	0.855 (0.412)	MEX SWE THA BGR EST SGP POL CHL BLR URU BRA IDN ZAF	0.84 & 0.52 (13)
April	1a. ∼GOV*GHS*LPI*GNI (0.798, 0.492/0.188) 1b. ∼GOV*GHS*∼DEATHS*ECOG*SOCG*GNI (0.824, 0.327/0.023)	0.8 (0.515)	CHN IDN TUR CAN DNK EST FIN NOR SWE CHE BGR GRC ISL + 3 LVA URU CRI BGR GRE ISL JAP MEX KOR	0.82 & 0.52 (19)
May	1. ∼GOV*GHS*∼DEATHS*ECOG*SOCG*GNI (0.853, 0.391/0.285) 2. ∼GOV*GHS*∼DEATHS*∼SOCG*LPI*GNI (0.837, 0.115/0.009)	0.836 (0.4)	CRI ISL LVA URU BGR CZE EST GRC JAP NZL NOR SLO KOR IDN	0.83 & 0.43 (14)
June	1a. ∼GOV*∼DEATHS*ECOG*SOCG*GNI (0.808, 0.524/0.017) 1b. GHS*∼DEATHS*ECOG*SOCG*GNI (0.781, 0.658/0.151) 2. ∼GOV*GHS*ECOG*SOCG*LPI*GNI (0.803, 0.554/0.064) 3. ∼GOV*GHS*∼DEATHS*∼ECOG*∼SOCG*LPI*GNI (0.89, 0.1/0.011)	0.751 (0.75)	MLT BIH LVA URU AUT AUS HRV DNK EST FIN GRC ISL JAP NZL + 8 ALB CRI CYP DEU HUN ISR NLD MYS SGP THA NOR POL KOR + 19 ESP SWE MEX ROU BGR CHE POL NOR KOR SVK LUX SLO NZL + 11 IDN	0.829 & 0.53 (39)
July	1. ∼GOV*GHS*ECOG*SOCG*LPI*GNI (0.835, 0.59/0.047) 2. ∼GOV*GHS*∼DEATHS*ECOG*SOCG*GNI (0.836, 0.553/0.010)	0.826 (0.6)	SWE BGR MEX ROU NZL NOR POL SVK SLO KOR CHE AUT CZE + 9 LVA URU AUT CZE HRV EST FIN FRA ISL IRL JAP LTU LUX NLD + 7	0.84 & 0.5 (25)

Key: GOV: Government response; GHS: Healthcare preparedness; DEATHS: Death rate in the month (per 100 k); ECOG: Economic globalization; LPI: Logistics performance.

GNI: National Income (per capita); SOCG: Social Globalization; ‘∼’ suggests the negation of the set (e.g., ∼GOV indicates “NOT strong government response”).

Notes: 1. Core conditions in bold, contributing conditions italicized, 2. For each month, the configuration with the highest unique coverage is underlined, 3. Directional expectations: we expect the absence of DEATHS and GOV, and the presence of all other conditions, to be associated with the outcome. Our results are not sensitive to changes in directional expectations, 4. Of the intermediate solution, 5. Cases that belong uniquely to the respective configuration are in bold. 6. Specifically, the consistency and PRI of the most inconsistent empirically observed configuration included in the minimization.

Table 5.

fsQCA results: Analysis of sufficiency for the absence of trade resilience.

Month (2020)	Intermediate Solution1,2,3,4(consistency, unique coverage)	Solution consistency (coverage)	Typical cases	Cases covered
March	GOV*∼GHS*∼LPI (0.781, 0.053) GOV*∼ECOG*∼LPI (0.776, 0.016) GOV*∼SOCG*∼LPI (0.814, 0.021) ∼GHS*∼ECOG*∼SOCG*∼LPI*∼GNI (0.895, 0.018)	0.793 (0.361)	UKR MLT KAZ RUS MAR, PER, SLV EGY	12
April	GOV*DEATHS (0.827, 0.306) GOV*ECOG*∼LPI (0.868, 0.067) GOV*∼GHS*∼SOCG*∼LPI (0.841, 0.011) GOV*∼ECOG*∼SOCG*∼GNI (0.844, 0.030) ∼GHS*∼ECOG*∼SOCG*∼LPI*∼GNI (0.898, 0.017)	0.805 (0.659)	BEL, AUT, FRA, ITA UKR, MLT, ALB GTM, PRY IND, PHL BOL	30
May	GOV*DEATHS*∼LPI (0.885, 0.011) GOV*∼ECOG*LPI (0.796, 0.055) GOV*∼SOCG*∼GNI (0.861, 0.015) GOV*SOCG*∼LPI (0.797, 0.027) ∼GHS*DEATHS*∼LPI (0.933, 0.019)	0.782 (0.429)	RUS, PER COL, ARG, BRA SLV, EGY, VNM UKR, ALB, KAZ BLR, BOL	23
June	GOV*DEATHS*∼ECOG (0.867, 0.103) GOV*DEATHS*∼SOCG*∼LPI (0.893, 0.023) ∼GOV*∼GHS*ECOG*∼SOCG∼LPI*∼GNI (0.892, 0.036)	0.876 (0.364)	ARG, COL, BRA, RUS PER, SLV TUN	12
July	GOV*DEATHS (0.836, 0.185) ∼GHS*∼LPI (0.840, 0.051) DEATHS*∼LPI (0.868, 0.004) ∼GOV*∼ECOG*∼SOCG (0.922, 0.005) ∼ECOG*∼SOCG*∼GNI (0.912, 0.018) GOV*∼SOCG*∼LPI*∼GNI (0.901, 0.003)	0.817 (0.663)	ARG, ZAF, BRA, USA TUN, UKR, MLT BIH IDN IND, PHL MAR	27

Key: GOV: Government response; GHS: Healthcare preparedness; DEATHS: Death rate in the month (per 100 k).

ECOG: Economic globalization; LPI: Logistics performance; GNI: National Income (per capita); SOCG: Social Globalization; ‘∼’ indicates the negation of the set (e.g., ∼GOV indicates “NOT strong government response”).

¹Core conditions in bold, contributing conditions italicized.

²Consistency thresholds ranged from 0.81 to 0.86; PRI threshold set to 0.6.

³For each month, the configuration with the highest unique coverage is underlined.

⁴Directional expectations: we expect the presence of DEATHS and GOV, and the absence of all other conditions, to be associated with the negation of the outcome. Our results are not sensitive to changes in directional expectations.

3.4. Robustness tests

We probe these decisions in a series of robustness tests, and all Truth Tables of the sufficiency analysis for the presence of the outcome are included in Appendix A for transparency purposes. In short, we experimented with alternative consistency thresholds and alternative calibration schemes for the outcome and selected conditions. The results (in Appendix B) reinforce our confidence in the validity of the findings from the baseline analyses, as discussed in the following section.

4. Results and discussion

The main results come from the analysis of sufficiency for the presence and the absence of trade performance. In this section, we present both sets of results.

4.1. Analysis of sufficiency for the presence of the outcome

The parsimonious and intermediate solutions for each month are presented in Table 4. The latter includes additional details such as consistency/coverage scores, typical cases, and directional expectations. At first glance, the stand-out observation is that, even though the absence of ‘strong government response’ (∼GOV) was not formally necessary for the outcome in any month, it emerges as a core condition in all but two sufficient configurations (a ‘false necessary’ condition – see Schneider & Wagemann, 2012). This suggests that maintaining an environment conducive to trade is paramount; countries in which governments did not impose many restrictions and emphasized business continuity (or swift recovery), possibly at the expense of human lives, managed to be relatively more robust and responsive.

Crucially though, the absence of a strong government response (∼GOV) should not be interpreted in isolation since it was, almost invariably, accompanied by both high national income (GNI) and a prepared healthcare system (GHS). This means that countries that followed a lax containment approach (∼GOV) exhibited resilience only due to their relatively high-income levels (and available resources) and their high levels of preparedness for a pandemic. The presence of GHS in all but two configurations is telling; healthcare infrastructure is crucial for maintaining economic activity and international trade during a crisis of this magnitude.

It should also be noted that in most instances, ∼GOV is accompanied by the absence of deaths (∼DEATHS) and never with its presence. This suggests that one of the following mechanisms might be at play. It could be that some countries might have been ‘lucky’ in terms of timing; for example, Brazil and Mexico in March had not yet entered their first wave. Alternatively, governments might have adopted a strict approach in earlier months, mitigating the magnitude of the wave considerably, and only in later periods relaxed many restrictions (i.e., ∼GOV) contributing to recovery (e.g., New Zealand and Australia). Finally, it might be that some countries took selective actions to contain the health crisis (hence ∼DEATHS) but did not restrict international trade (e.g., China). As such, a closer examination of the role of the encompassed indicators of the Oxford government response index warrants further research.

When it comes to logistics performance (LPI) and globalization (ECOG and SOCG), it is their presence that (in conjunction with other factors) leads to trade resilience. In the rare instances where the absence of either of these conditions is part of a sufficient configuration (confirming Tenet 5), the presence of the others seems to compensate for its role. This suggests that high levels of globalization and strong logistics infrastructure might, in instances, act as substitutes for each other in eliciting trade resilience. For example, in March, Uruguay and Belarus were deemed ‘robust’ despite their low LPI score; their high levels of social and economic globalization compensated for it, conjunctively leading to high trade performance. Conversely, Indonesia appears in a couple of configurations, despite its low levels of globalization (but high LPI score).

It is also important to note that ‘High Income’ (GNI) appears in all intermediate solution configurations. Although this is encouraging news for advanced economies, it paints a pessimistic picture regarding the chances of lower-income countries in achieving trade resilience during international disruptions of this magnitude.

4.2. Analysis of sufficiency for the absence of the outcome

The analysis for the negation of the outcome identified various, qualitatively different configurations associated with the absence of trade performance (Table 5) that are not mirror opposites of the configurations in Table 4 (confirming Tenet 4). Overwhelmingly though, not having prepared for a health crisis (∼GHS) and a weak logistics infrastructure (∼LPI) hampered countries’ chances to withstand and recover from the COVID-19 pandemic. This also holds for the two types of globalization, and in the odd case where either is present, solid logistics infrastructure is absent (∼LPI). As expected, a strong government response to prevent the spread of the pandemic or belatedly to mitigate the damage (i.e., GOV*DEATHS) undermines trade resilience.

The fact that there are many different ‘recipes’ for trade underperformance is a cautionary note. However, it might also suggest that influential factors capturing commonalities across countries have not been considered here. Nevertheless, this does not undermine the critical role of preparedness in healthcare and logistics for tackling global health crises.

4.3. Country profiles and trade resilience

The results presented in the previous sub-sections suggest a complex picture of how international trade resilience in different countries was tested by the first wave of the pandemic. In this section, we synthesize the emerging patterns. To do so, we abstract from the month-by-month configurational findings to advance a classification of countries into high-order ‘profiles’ that share similar characteristics over time. The profiles discussed below draw from both sets of analyses (presence and absence of the outcome) and inevitably leave out several countries that do not readily ‘fit’ in them. However, this abstraction comprises an initial attempt to tell a high-level ‘story’ of trade resilience during an unprecedented global disruption.

• •Robust and responsive: This profile includes high-income (GNI), globalized (SOCG and ECOG), well-prepared (GHS) countries with strong logistics infrastructure (LPI) that exhibited trade resilience (especially responsiveness in later months). Most importantly though, they appear to have avoided heavy fatalities (∼DEATHS). While various countries come in and out of this configuration, countries such as New Zealand (NZL), South Korea (KOR), Iceland (ISL), and Norway (NOR) appear consistently from May onwards (i.e., after the global ‘peak’ of the first wave). These countries adopted a strict approach in earlier months (especially March), mitigating the magnitude of the wave considerably. This allowed them to relax many restrictions in later months (∼GOV), contributing (alongside infrastructural factors) towards their recovery in May (IMF, 2021).
• •Stringent but fragile: This profile involves countries that enforced a strong response (GOV), but weak healthcare and logistics infrastructures (∼GHS and ∼LPI) acted as barriers to trade resilience. The exemplar case in this profile is Ukraine, which consistently appears in the non-resilient set (Table 5) across all months except June. Indeed, Ukraine’s response was very stringent, involving bans on international travel and restrictions covering most establishments, including those involving physical interaction with clients (IMF, 2021). This profile reveals the adverse side effects of overly stringent measures in the absence of robust healthcare infrastructure and logistics capabilities.
• •Laissez-faire. In four out of the five months, Sweden (SWE) exhibited trade resilience. Sweden shows high national income (GNI), strong logistics (LPI) and healthcare systems (GHS), and solid globalization, both socially (SOCG) and economically (ECOG). However, the distinctive condition is a relatively lax government response (∼GOV), which allowed businesses to remain open, propping-up the economy and supporting imports and exports. It is important to note that while Sweden maintained trade robustness and responsiveness during the first wave, high mortality rates ensued (JHU, 2021). That is a hefty price to pay for trade resilience.
• •Procrastinators: This profile covers relatively less globalized economies (∼SOCG or ∼ECOG), but with strong healthcare (GHS) and logistics (LPI) infrastructure, which combined with a lax early government response (∼GOV) enabled trade robustness. However, despite signs of robustness in the early months, their international trade performance eventually collapsed. The negative impact was so severe that countries moved from “fully-in” the set of resilient countries in the early months of the analysis to (almost) “fully out” in later months, consequently featuring in both Table 4, Table 5. This profile includes Brazil (BRA) and Indonesia (IDN). In both countries, the lack of stringent government response (∼GOV) in the pandemic's early days helped robustness as business remained open. However, as infections and deaths climbed, governments were forced to adopt a more stringent response (GOV) and suffered economically as a result. In the case of Brazil, they also suffered significantly from fatalities (JHU, 2021).
• •Early victims: This profile includes globalized European countries such as Belgium (BEL), Portugal (PRT), France (FRA), Netherlands (NLD), and Italy (ITA) that were hit fast and hard, both in fatalities and trade. Despite their wealth and solid infrastructure, these countries were severely hit in March and April 2020, at a point where there was little understanding of the virus and limited supplies of protective and testing equipment. This led to a heavy loss of life and very stringent government responses (IMF, 2021), which undermined their trade robustness. Although the trade performance of some countries (e.g., France, Netherlands) recovered in later months (i.e., responsive), of others it did not (e.g., Portugal, Italy).
• •Late victims: This profile includes countries that were lucky to be hit later than the bulk of the European countries; hence they did not suffer the international trade effects of the pandemic in the early days. However, as the virus spread and fatalities rose (DEATHS), they were forced to impose severe restrictions on economic activity (GOV), hampering their trade resilience. This profile includes Argentina (ARG), Colombia (COL), South Africa (ZAF), and to a lesser extent Russia (RUS) and the United States (USA).
• •Never had a chance: This profile includes low-income (∼GNI) countries that exhibited a lack of resilience during most periods. Due to suffering from many weaknesses, such as fragile healthcare (∼GHS), limited globalization (∼ECOG & ∼SOCG), and poor logistics (∼LPI), it would have been extremely hard for them to withstand and/or swiftly recover from a global disruption of this scale. This profile includes countries in Africa (e.g., Morocco, Tunisia, Egypt), Central America (e.g., Guatemala, El Salvador), and South America (e.g., Bolivia, Paraguay, Peru).

5. Conclusions, implications, and further research

Most firms participate directly or indirectly in global supply chains and are embedded in the broader international trade system (Goldin & Mariathasan, 2015). They must abide by international trade rules, use the infrastructure available in the countries where they operate or source from, and rely on suppliers and customers in those countries. As a result, every practitioner, whether in the private or public sector, should be interested in the factors underpinning the resilience of the international trade system (Goldin & Mariathasan, 2015). The macro perspective adopted in this research extends the scope of most supply chain resilience studies, which focus on dyads or networks, and provides a broader explanation for resilience. The findings are relevant to those making sourcing decisions, who are currently reviewing their global supply chain structure and risk exposure in light of the Covid-19 pandemic and preparing for the next crisis. This research can guide the restructuring process by providing insights into the complex relationships between globalization, healthcare infrastructure, logistics capabilities, and international trade. For instance, managers making global sourcing decisions can use this information to create supplier portfolios in different countries with different risk profiles to better manage the overall disruption exposure.

While the Covid-19 pandemic has raised concerns regarding supply chain vulnerability, especially across strategic sectors, and prompted discussions of possible reshoring and nearshoring, this work increases the understanding of specific country-level vulnerabilities (The Economist, 2020, Yip, 2021, Altman, 2020). This allows for a more targeted management of risk and the associated factors that contribute to the resilience of the international trading system. The findings allow for a balanced and informed approach to mitigation and subsequent policy interventions.

This research identifies a set of equifinal configurations of macro-level, country-specific factors that foster (or hamper) international trade resilience. While it is the conjunctive effect of these factors that leads to a positive or negative outcome, some insights with respect to each factor’s role are salient: globalization (economic and social), logistics and healthcare preparedness, and high-income levels, played an overwhelmingly positive role in eliciting trade resilience. On the other hand, a strong government response, and a high number of deaths, had a largely negative effect, demonstrating that balancing and reconciling health-related and economic outcomes is a challenging task.

This study offers a more nuanced understanding of a causally complex, ‘chaordic’ phenomenon, laying the foundation of a midrange theory of international trade resilience. The adopted neo-configurational lens allowed us to explore the relevance of several causal factors (i.e., social and economic globalization, logistics performance, healthcare preparedness, national government response, and income level) and offer several plausible explanations for how these factors can promote (or suppress) international trade resilience. This contribution provides the rudiments of a midrange theory of international trade resilience. We hope our approach motivates future studies considering the extent of globalization and the strength of national logistics systems, and their influence on the robustness and responsiveness of international trade. We invite researchers to continue this theorization process by developing and testing hypotheses based on our findings.

The study also offers significant implications for policymakers, particularly those who focus on international trade policy, since it highlights the importance of key factors that can strengthen the resilience of export economies and hence guide host government investment priorities in support of national economic development strategies. As the Covid-19 crisis has been a major setback for global development targets in most countries, with a particularly negative socio-economic impact on vulnerable groups (New York Times, 2020), the findings emphasize a direct linkage between a resilient international trading system, equitable access to health care, and protection of livelihoods. Thereby, the development of a strong healthcare capability and an effective and efficient logistics infrastructure become integral components of a credible and resilient export-oriented national development strategy.

Furthermore, policymakers involved in designing national preparedness strategies for future pandemics or similar health-related systemic risks might also find this work useful. Drawing from the findings, they can review the factors affecting trade resilience and seek to improve their forecasts of the probability of disruptions emanating from specific sourcing locations. Based on the country profiles encompassing national income, globalization levels, and existing infrastructure capabilities, appropriate mitigation strategies can be developed.

In addition, the insight that lower-income countries lacked the needed trade robustness and responsiveness during an international trade disruption of this scale will need to inform global initiatives to ensure more equitable access to health-related resources in future pandemics. An example is the ongoing COVAX initiative,1 which aims to secure COVID-19 vaccines for 92 low- and middle-income countries at the same time as wealthier nations. Such initiatives will need to be mainstreamed into future global pandemic response frameworks.

This research has some limitations. Firstly, we made an informed decision to limit the analysis to the first wave of the pandemic, and we were constrained by the limited availability of trade data during an unfolding crisis. However, we believe that we have provided a blueprint for research, and as more data becomes available, it will be possible to study subsequent waves of the pandemic. Secondly, the use of secondary data and the chosen analytic technique limited the scope of factors included in the analysis. We acknowledge that other factors, such as national culture, geography, and even regional variations, might have influenced country-level trade resilience during the COVID-19 pandemic. In fact, the relatively low consistency and coverage scores of some solutions point towards this. Thirdly, the cross-country, multi-period lens adopted in this work prevents a detailed examination of each country’s COVID-19 ‘story’, especially of countries that did not achieve trade resilience despite exhibiting the ‘desired’ characteristics (i.e., deviant cases – see Tenet 7 of Woodside, 2014). Fourthly, the absence of external, agreed, qualitative thresholds for some of our measures might have slightly compromised the validity of our calibration scheme. However, we have examined the sensitivity of the results to alternative schemes in our robustness tests. We encourage researchers to build on our findings, using different datasets and both qualitative and quantitative methods, to further our understanding of resilience in the context of large-scale, global disruptions.

The COVID-19 crisis has likely touched the lives of every person on earth. The impact of the loss of life and the changes to human interaction and economic activity will stay with us for generations. It is thus vital to understand the factors in our global economy that have allowed such a devastating impact, the benefits and costs of resilience, and the solutions that can enable a swift recovery. Armed with this understanding, we can start preparing for future crises.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Biographies

Dr. Carlos Mena is the Nike Professor of Supply Chain Management at Portland State University (PSU). The focus of his research is the impact of procurement and supply chain management practices on economic, social and environmental performance. He has more than 30 refereed publications to his name, as well as two books entitled “Leading Procurement Strategy” and “Delivering Performance in Food Supply Chains”. Prior to joining PSU, Carlos worked for Michigan State University and for Cranfield University (UK). He received his Doctorate degree and MSc in Engineering Business Management from the University of Warwick (UK). He also holds a BEng in Industrial Engineering from the Iberoamericana University in Mexico.

Dr. Antonios (Tony) Karatzas is an Assistant Professor at the Norwich Business School, University of East Anglia. He maintains an active research interest in various topics relating to supply networks and their evolution, and likes to apply a multitude of research methods. He has published in several journals, including Production and Operations Management, International Journal of Operations & Production Management and Journal of Supply Chain Management. He received his Doctorate from Cranfield University, and worked as a postdoctoral research fellow at WMG, University of Warwick, before his current post.

Dr. Carsten Hansen is a Visiting Research Fellow, School of Management, Centre for Strategic Procurement and Supply Management at Cranfield University (UK). He has more than 20 years of experience managing international procurement and supply chain operations. He researches in Supply Chain Risk Management, Sustainable Sourcing Strategies and Political Risk Management, mainly focused on the offshore outsourcing industry, public procurement, and global supply chain activities. Publications in the International Journal of Production Research include research investigating the effect of political risk on the offshore service industry and the implications of offshoring entry mode choices and location decisions.

Appendix A. Truth Tables

For transparency purposes, we include herein the Truth Tables of the analysis of sufficiency for the presence of trade resilience, for all months. The column ‘OUT’ denotes the presence of the outcome (and takes the value of 1 for rows that are above the baseline consistency and PRI thresholds), ‘n’ is the frequency and ‘Incl’ is the consistency (or Inclusion) score. It is noteworthy that all truth tables have rows with both high and low consistency values, which suggests that the chosen conditions differentiate the cases well with respect to the outcome (Radaelli & Wagemann, 2019). Hence, although we cannot deny that influential variables might have not been considered here, we can be confident that the chosen variables are indeed relevant (see Table A1, Table A2, Table A3, Table A4, Table A5 ).

Table A1.

Truth Table for March (R Output).

Conf.	GOV	GHS	DEATHS	ECOG	SOCG	LPI	GNI	OUT	n	Incl	PRI
36	0	1	0	0	0	1	1	1	2	0.898	0.756
46	0	1	0	1	1	0	1	1	1	0.893	0.698
40	0	1	0	0	1	1	1	1	1	0.889	0.691
48	0	1	0	1	1	1	1	1	7	0.888	0.749
64	0	1	1	1	1	1	1	1	1	0.88	0.636
14	0	0	0	1	1	0	1	1	1	0.845	0.52
112	1	1	0	1	1	1	1	0	20	0.758	0.596
104	1	1	0	0	1	1	1	0	1	0.744	0.472
100	1	1	0	0	0	1	1	0	4	0.743	0.534
107	1	1	0	1	0	1	0	0	1	0.727	0.492
110	1	1	0	1	1	0	1	0	4	0.716	0.435
1	0	0	0	0	0	0	0	0	1	0.698	0.245
78	1	0	0	1	1	0	1	0	1	0.688	0.328
66	1	0	0	0	0	0	1	0	2	0.683	0.366
128	1	1	1	1	1	1	1	0	13	0.671	0.452
106	1	1	0	1	0	0	1	0	1	0.64	0.243
102	1	1	0	0	1	0	1	0	2	0.63	0.333
77	1	0	0	1	1	0	0	0	1	0.621	0.29
99	1	1	0	0	0	1	0	0	2	0.597	0.333
73	1	0	0	1	0	0	0	0	1	0.594	0.216
105	1	1	0	1	0	0	0	0	2	0.558	0.206
65	1	0	0	0	0	0	0	0	1	0.55	0.204

Table A2.

Truth Table for April (R Output).

Conf.	GOV	GHS	DEATHS	ECOG	SOCG	LPI	GNI	OUT	n	Incl	PRI
52	0	1	1	0	0	1	1	1	1	0.879	0.52
36	0	1	0	0	0	1	1	1	2	0.864	0.708
46	0	1	0	1	1	0	1	1	3	0.864	0.598
64	0	1	1	1	1	1	1	1	7	0.852	0.598
48	0	1	0	1	1	1	1	1	6	0.822	0.596
14	0	0	0	1	1	0	1	0	1	0.812	0.422
116	1	1	1	0	0	1	1	0	1	0.752	0.332
122	1	1	1	1	0	0	1	0	1	0.747	0.074
112	1	1	0	1	1	1	1	0	13	0.736	0.5
104	1	1	0	0	1	1	1	0	2	0.73	0.45
78	1	0	0	1	1	0	1	0	1	0.712	0.314
1	0	0	0	0	0	0	0	0	1	0.704	0.268
107	1	1	0	1	0	1	0	0	1	0.692	0.438
100	1	1	0	0	0	1	1	0	2	0.684	0.402
102	1	1	0	0	1	0	1	0	2	0.681	0.366
110	1	1	0	1	1	0	1	0	2	0.678	0.223
128	1	1	1	1	1	1	1	0	15	0.613	0.257
77	1	0	0	1	1	0	0	0	1	0.6	0.261
66	1	0	0	0	0	0	1	0	2	0.591	0.267
73	1	0	0	1	0	0	0	0	1	0.572	0.188
99	1	1	0	0	0	1	0	0	2	0.556	0.268
105	1	1	0	1	0	0	0	0	2	0.538	0.148
65	1	0	0	0	0	0	0	0	1	0.491	0.181

Table A3.

Truth Table for May (R Output).

Conf.	GOV	GHS	DEATHS	ECOG	SOCG	LPI	GNI	OUT	n	Incl	PRI
46	0	1	0	1	1	0	1	1	3	0.932	0.702
30	0	0	1	1	1	0	1	0	1	0.857	0.124
48	0	1	0	1	1	1	1	1	10	0.856	0.639
36	0	1	0	0	0	1	1	1	1	0.833	0.43
110	1	1	0	1	1	0	1	0	1	0.81	0.294
78	1	0	0	1	1	0	1	0	1	0.8	0.377
107	1	1	0	1	0	1	0	0	1	0.789	0.393
64	0	1	1	1	1	1	1	0	8	0.781	0.44
112	1	1	0	1	1	1	1	0	11	0.781	0.44
104	1	1	0	0	1	1	1	0	2	0.769	0.381
126	1	1	1	1	1	0	1	0	1	0.754	0.144
102	1	1	0	0	1	0	1	0	1	0.752	0.339
66	1	0	0	0	0	0	1	0	2	0.729	0.419
17	0	0	1	0	0	0	0	0	1	0.717	0.102
77	1	0	0	1	1	0	0	0	1	0.695	0.122
100	1	1	0	0	0	1	1	0	2	0.694	0.337
122	1	1	1	1	0	0	1	0	1	0.691	0.032
73	1	0	0	1	0	0	0	0	1	0.69	0.126
118	1	1	1	0	1	0	1	0	1	0.679	0.228
116	1	1	1	0	0	1	1	0	3	0.671	0.253
105	1	1	0	1	0	0	0	0	2	0.659	0.056
128	1	1	1	1	1	1	1	0	12	0.641	0.304
65	1	0	0	0	0	0	0	0	1	0.596	0.235
99	1	1	0	0	0	1	0	0	2	0.576	0.165

Table A4.

Truth Table for June (R Output).

Conf.	GOV	GHS	DEATHS	ECOG	SOCG	LPI	GNI	OUT	n	Incl	PRI
46	0	1	0	1	1	0	1	1	3	0.892	0.728
36	0	1	0	0	0	1	1	1	1	0.892	0.69
30	0	0	1	1	1	0	1	0	1	0.883	0.445
110	1	1	0	1	1	0	1	1	2	0.841	0.658
48	0	1	0	1	1	1	1	1	19	0.839	0.724
14	0	0	0	1	1	0	1	1	1	0.834	0.548
112	1	1	0	1	1	1	1	1	8	0.831	0.671
64	0	1	1	1	1	1	1	1	5	0.829	0.538
9	0	0	0	1	0	0	0	0	1	0.817	0.393
100	1	1	0	0	0	1	1	0	2	0.805	0.642
66	1	0	0	0	0	0	1	0	1	0.764	0.511
107	1	1	0	1	0	1	0	0	1	0.756	0.526
77	1	0	0	1	1	0	0	0	1	0.752	0.398
102	1	1	0	0	1	0	1	0	1	0.729	0.385
105	1	1	0	1	0	0	0	0	1	0.7	0.37
121	1	1	1	1	0	0	0	0	1	0.676	0.284
99	1	1	0	0	0	1	0	0	2	0.645	0.438
128	1	1	1	1	1	1	1	0	9	0.642	0.341
122	1	1	1	1	0	0	1	0	1	0.635	0.217
120	1	1	1	0	1	1	1	0	2	0.61	0.184
116	1	1	1	0	0	1	1	0	3	0.587	0.179
118	1	1	1	0	1	0	1	0	1	0.549	0.145
82	1	0	1	0	0	0	1	0	1	0.544	0.159
81	1	0	1	0	0	0	0	0	2	0.541	0.238

Table A5.

Truth Table for July (R Output).

Conf.	GOV	GHS	DEATHS	ECOG	SOCG	LPI	GNI	OUT	n	Incl	PRI
30	0	0	1	1	1	0	1	0	1	0.876	0.222
64	0	1	1	1	1	1	1	1	4	0.875	0.503
46	0	1	0	1	1	0	1	1	2	0.87	0.595
62	0	1	1	1	1	0	1	0	1	0.861	0.274
48	0	1	0	1	1	1	1	1	19	0.847	0.728
14	0	0	0	1	1	0	1	0	1	0.835	0.312
9	0	0	0	1	0	0	0	0	1	0.797	0.137
112	1	1	0	1	1	1	1	0	15	0.789	0.557
36	0	1	0	0	0	1	1	0	1	0.787	0.26
13	0	0	0	1	1	0	0	0	1	0.749	0.106
100	1	1	0	0	0	1	1	0	2	0.741	0.481
128	1	1	1	1	1	1	1	0	3	0.731	0.28
66	1	0	0	0	0	0	1	0	1	0.721	0.31
107	1	1	0	1	0	1	0	0	1	0.708	0.396
126	1	1	1	1	1	0	1	0	2	0.684	0.203
121	1	1	1	1	0	0	0	0	1	0.682	0.174
122	1	1	1	1	0	0	1	0	1	0.615	0.129
105	1	1	0	1	0	0	0	0	1	0.591	0.069
82	1	0	1	0	0	0	1	0	1	0.578	0.13
120	1	1	1	0	1	1	1	0	2	0.532	0.032
81	1	0	1	0	0	0	0	0	2	0.53	0.102
99	1	1	0	0	0	1	0	0	2	0.507	0.178
116	1	1	1	0	0	1	1	0	3	0.483	0.031
118	1	1	1	0	1	0	1	0	2	0.463	0.011

Appendix B. Robustness tests

B1. Alternative consistency thresholds

To understand whether the results are sensitive to alternative decisions regarding consistency and PRI thresholds, we ran a series of robustness tests for each month. Table B1 includes the results of the most intuitive alternative decision, in the sufficiency analysis for the presence of the outcome. The reader can evaluate these in conjunction with the Truth Tables (see earlier section).

Table B1.

Results from analyses with alternative consistency thresholds.

Month	Alternative Decision1(cases included)	Result
March	Consistency cut: 0.75 (33)	An additional configuration is included in the minimization with 20 extra cases. Substantively, the results remain unchanged.
April	Consistency cut: 0.80 (20)	A more complex solution, including a configuration signifying a non-globalized country (∼FECOG*∼FSOCG) but with minimal government intervention and solid infrastructure (∼FGOV*FGHS*FLPI).
May	Consistency cut: 0.78; PRI: 0.4 (33)	Solution includes two more variations of the predominant configuration, where ∼DEATHS and ∼GOV alternate. Substantive results are even stronger.
June	Consistency cut: 0.80; PRI cut: 0.4 (42)	An almost identical solution is derived.
July	Consistency cut: 0.78; PRI cut: 0.5 (40)	An almost identical solution is derived.

¹All robustness tests reported here apply more liberal (but permissible) thresholds. Tests with more conservative thresholds were also run where it made sense; the results do not challenge the main findings of the baseline analysis.

B2. Alternative calibration scheme for the outcome

In the baseline analyses, for the calibration of the outcome (Year-on-Year monthly % change in total trade volume) we used the 90th and 10th percentiles as thresholds to denote full and no set-membership, respectively, in the set of countries that exhibited above average trade resilience, and the mean as the cross-over point. As a robustness test, we re-run the analysis by using the 80th, 20th and 50th (median) percentiles as respective alternative thresholds. Although the choice of the median instead of the mean may not be very consequential (since for all months the distribution of the original measure is almost normal, with the difference between mean and median being about 1 percentage point), the full inclusion and exclusion thresholds differ by an average of 4 percentage points. As such, we consider this robustness test as quite ‘severe’.

In all 10 analyses, most configurations in the parsimonious and intermediate solutions remain intact. In some instances, alternative variants of existing configurations appear. The results remain substantively the same. Comparatively speaking, we consider our baseline calibration to be superior not only theoretically, but also empirically (due to higher solution consistency and coverage scores for similar cut-off decisions). The results are not presented herein but are available upon request.

B3. Alternative calibration schemes for the conditions

We are fortunate that there exists external, substantive knowledge that aided the calibration of most conditions: LPI, ECOG, SOCG, GHS and GNI. As such, for these conditions we see no reason to experiment with alternative schemes that lack theoretical backing, with one exception: Originally, to calibrate GNI, we used the 2019 World Bank thresholds that classify countries into low-, lower-middle, upper-middle and high-income.2 Arguably, and given that our final sample is dominated by high-income countries, these thresholds might be considered too low. We thus re-calibrate this measure after changing the name of the respective set into ‘the set of very high-income countries’ (for terminological consistency), by applying the following thresholds.

• •‘Fully in’ being the mean GNI per capita of the World Bank designated high-income countries ($45,353)
• •‘Cross-over’ being the World Bank set threshold demarcating the high-income group ($12,535)
• •‘Fully out’ being the World Bank set threshold demarcating the ‘upper-middle’ income countries from ‘lower-middle’ ones ($4,045)

For the two time-variant conditions (GOV and DEATHS) which we originally calibrated using distributional properties, we apply alternative thresholds in the same spirit as in B2 above. Namely, we use the median of the distribution (instead of the mean) as cross-over point, and increase (decrease) the lower (upper) threshold from the 10th (90th) to the 20th (80th) percentile.

To make this robustness test as severe as possible, we run only one set of analyses after substituting all three calibrated measures at the same time. Also, to draw comparisons more easily with the original results, for each month we apply the exact same consistency and PRI thresholds as in the respective baseline analysis (see Table 4, Table 5 of the manuscript).3 The results are in general agreement with the original findings, increasing the confidence in our key insights in Table B3 , we provide more nuance with a month-by-month comparison of the intermediate solutions.

Table B3.

Comparison between the baseline and the alternative intermediate solutions.

Month	Presence of resilience (RESI)	Absence of resilience (∼RESI)
March	A more complex solution emerges, with 3 out of 6 configurations including the absence of very high income (∼GNI). This, however, is accompanied by the absence of strong government response (∼GOV) and deaths (∼DEATHS), or the presence of globalization (ECOG/SOCG), healthcare preparedness (GHS) and logistics infrastructure (LPI).	A more complex solution emerges that resembles very closely the baseline one. One difference is that GHS appears in 2 configurations but accompanied by DEATHS.
April	A more complex solution is generated, which includes variants of the same, baseline, configuration. Substantively speaking, the results are completely consistent with the baseline findings.	A more complex solution emerges, which is consistent with the baseline one. The importance of DEATHS is even more pronounced. But like in March GHS appears in 2 configurations, this time accompanied by ∼ECOG/∼SOCG or ∼LPI.
May	A slightly more complex, but substantively similar, solution is produced. However, one configuration is not in agreement with the main findings since it has GOV, DEATHS and ∼GHS as components. Upon scrutiny though, this configuration represents only Malta, hence we do not consider this a serious threat to the validity of the baseline results.	Like April, but this time the solution shows that even prepared countries (GHS) can fail to achieve trade resilience if they lack the logistics infrastructure (∼LPI) and are constrained by strong government intervention (GOV).
June	A less complex but substantively very similar solution is produced	As above (May).
July	An almost identical solution is produced.	A less complex but substantively very similar solution. Regarding GHS, the same holds as in May and June.

Appendix C. Countries included in the analysis

See Table C1 .

Table C1.

Country names and respective 3-letter codes.

Albania	ALB	Denmark	DNK	Kazakhstan	KAZ	Singapore	SGP
Argentina	ARG	Ecuador	ECU	Latvia	LVA	Slovak Republic	SVK
Australia	AUS	Egypt	EGY	Lithuania	LTU	Slovenia	SVN
Austria	AUT	El Salvador	SLV	Luxembourg	LUX	South Africa	ZAF
Belarus	BLR	Estonia	EST	Malaysia	MYS	South Korea	KOR
Belgium	BEL	Finland	FIN	Malta	MLT	Spain	ESP
Bolivia	BOL	France	FRA	Mexico	MEX	Sweden	SWE
Bosnia and Herzegovina	BIH	Germany	DEU	Morocco	MAR	Switzerland	CHE
Brazil	BRA	Greece	GRC	Netherlands	NLD	Thailand	THA
Bulgaria	BGR	Guatemala	GTM	New Zealand	NZL	Tunisia	TUN
Canada	CAN	Hungary	HUN	Norway	NOR	Turkey	TUR
Chile	CHL	Iceland	ISL	Paraguay	PRY	Ukraine	UKR
China	CHN	India	IND	Peru	PER	United Kingdom	GBR
Colombia	COL	Indonesia	IDN	Philippines	PHL	United States	USA
Costa Rica	CRI	Ireland	IRL	Poland	POL	Uruguay	URY
Croatia	HRV	Israel	ISR	Portugal	PRT	Vietnam	VNM
Cyprus	CYP	Italy	ITA	Romania	ROU
Czech Republic	CZE	Japan	JPN	Russia	RUS