Inter‐country distancing, globalisation and the coronavirus pandemic

Abstract Originating in China, the coronavirus has reached the world at different speeds and levels of strength. This paper provides an initial understanding of some driving factors and their consequences. Since transmission requires people, the human factor behind globalisation is essential. Globalisation, a major force behind global well‐being and equality, is highly associated with this factor. The analysis investigates the impact globalisation has on the speed of initial transmission to a country and on the scale of initial infections in the context of other driving factors. Our cross‐country analysis finds that measures of globalisation are positively related to the spread of the virus, both in speed and in scale. However, the study also finds that globalised countries are better equipped to keep fatality rates low. The conclusion is not to reduce globalisation to avoid pandemics, but to better monitor the human factor at the outbreak and mobilise collaboration forces to curtail diseases.

also find that reducing human mobility mitigated the coronavirus transmission in China. Studies on other viruses have shown that the spread is faster during economic booms (Adda, 2016) and with trade growth (Adda, 2016, on influenza;Oster, 2012, on HIV). There may be also long-term growth effects through changes of fertility (Chin & Wilson, 2018).
This research suggests social distancing within countries and more importantly distancing between countries early on, focusing on the human factor are crucial to avoid a pandemic or at least to contain it. Hence, strict monitoring of human mobility across borders (including their closure) may seem appropriate. In the face of the current coronavirus threat, would this require reducing globalisation in the future?
There were also anti-globalist arguments during the more recent 2003 outbreak of SARS (severe acute respiratory syndrome) that started spreading to other countries from Hong Kong. At the time, the speed of transmission was so fast that a future pandemic seemed possible. Fears that originated in the affected countries at that time did not disappear with containment of the virus (Cheng, 2004). While several countries were affected, it was still possible to stop SARS before it became a pandemic (Chan-Yeung & Xu, 2003). But it was the first international epidemic of the 21st century. During that period, the SARS epidemic also triggered an anti-globalism discourse (So & Pun, 2004). Even the World Health Organization (WHO) stated that a new disease with wide-ranging impact might appear soon in a world becoming more and more interconnected with cross-boundary interactions becoming easier and more commonplace (WHO, 2003). However, they also report that globalisation might enable rapid information exchange between countries and a quicker response against a pandemic. With the COVID-19 outbreak becoming a pandemic, similar anti-globalist feelings have started to emerge (Legrain, 2020 andOba, 2020). Many governments have limited the export of medical supplies and medicines (Evenett, 2020). These discussions may result in a more permanent negative effect on the globalisation process since the impact of Coronavirus on the world is much bigger than that of SARS. There was already a lively debate on globalisation underway which this may accelerate (James, 2002).
Since globalisation is not solely a political choice, but a phenomenon related to various factors such as transportation and technology (especially those that affect information flows, see Ozcan, 2018), as well as a matter of the optimal division of work, it seems to be an irreversible process. Countries with globally diversified production are much more resilient to all kinds of shocks. Issues traditionally considered to be of local concern are only now seen as globally relevant and to be addressed through global collaborations. Such collaborations are needed at the beginning of a pandemic in particular to manage human mobility, while capital movements and trade policies can remain liberal (Evenett, 2020).

| METHODOLOGICAL APPROACH AND DATA
We are interested in the initial impact the pandemic has on affected countries in terms of transmission speed and mortality consequences. We neither model the evolution of the epidemic nor attempt to study the impact of health measures to contain the infection. We are only interested in understanding initial forces that drive the spread of the infection around the world. The value of such analysis is that it enables policymakers to better judge their options and the time constraints on action.
The transmission speed (TS) of the pandemic from country of origin (China) to another country is defined as Transmission speed (TS) = duration to reach country (D) times the infection rate (CP) , whereas D is the duration (in days) between the outbreak in China 2 and the first recorded case in a particular country (day gap), and CP is the infection rate defined as the number of confirmed COVID-19 cases C divided by P, the respective population size: As a major outcome variable, we measure the initial impact on mortality captured by the case fatality rate (CFR) defined in the epidemiology literature (Kelly & Cowling, 2013) as the proportion of deaths (M) from the disease divided by the number of confirmed infection cases C: Case fatality ratio 3 (CFR) = number of deaths (M) divided by the confirmed cases C.
Due to the non-linear structure of the data, 4 we analyse the variables linearised as ln TS, ln D, ln CP and ln CFR. 5 We use the COVID-19 data from the Johns Hopkins University Coronavirus Resource Center and will refer to the four variables as Coronavirus Variables in what follows. The data were collected for March 16, which is a few days after the global pandemic declaration on March 11, to avoid effects of government responses that could affect the data due to biological factors about two weeks later. The mortality data (M) are taken from April 6 assuming some delay between infections and deaths. The quality of the infection and mortality data is sometimes debated. However, Jelnov (2020) shows that the cross-country correlation between the log of tests and log of reported cases (per capita) and correlation between log of reported cases and log of reported deaths (per capita) is high, suggesting reliability.
As discussed above, our key hypothesis is that the degree of globalisation reflects important channels that impact the time and size of initial infection across countries. Understanding this relationship is important to enable governments to better design and execute non-pharmaceutical interventions. We measure globalisation using three different indices ('de facto', 'de jure' and 'overall') provided by the Swiss Federal Institute of Technology (KOF). 6 The 'de jure' index concentrates on trade regulations, tax regime, investment restrictions, tourism and capital regulations, international treaties, tariffs and several other legal matters; the 'de facto' index measures actual amounts of trade, foreign investment, international tourism, international students, migration and capital movements; and the 'overall' index combines the two. The alternative measures may provide insights into the nature of the disease's relationship with globalisation and are useful for robustness checks. For instance, the 'de facto' measure of globalisation contains more information related to actual human mobility and should potentially have a larger effect on the transmission of the disease.
The baseline equation is: 2 The disease was first reported on the 31st of December 2019, the global outbreak was reported on the 30th of January 2020, and the pandemic was declared on the 11th of March 2020.

Infection rate (CP) = number of COVID − 19 cases divided by population size P.
3 Also called the case fatality rate. 4 For robustness, we checked the relationships between the non-logarithmic variables. Joint test results and significance of the coefficients of the quadratic versions of the KOF over, KOF de facto, and KOF de jure variants indicate that there are non-linear relationships in most equations. We, therefore, decided to use the logarithmic specifications. X i denotes the vector of controls and i is the error term in the country i. Coronavirus Variables are D, CP, TS or CFR, Economic Globalization is KOF over, KOF de facto or KOF de jure. Control variables are average temperature in March, the median age of the population, population age 65 and above as a percentage of the total population, distance in km between Beijing and the respective country's capital, a democracy index (Institutionalised Democracy Index), a 'Belt Country' dummy variable for the member countries of China's One Belt One Road project, and an index for government ideology with values 1 for right, 2 for moderate and 3 for left. We use the following variables in ln form to model the non-linear relationship in the data and simplify interpretation: Coronavirus Variables, Economic Globalization variables, median age of the population, population with age 65 and above as a percentage of the total population, and distance from Beijing. The available dataset includes the 118 countries listed in the Appendix. Definitions and sources of all variables and their descriptive statistics are provided in Table 1. The data set contains 101 countries for the analysis of the non-zero case fatality ratios.

| EMPIRICAL FINDINGS
An initial illustration of the relationships between the Coronavirus Variables and Economic Globalization (KOF over) is provided in Figures 1-4. In Figure 1, a negative relationship between ln Infections and globalisation are upward-sloping: Figure 2 connects ln KOF over with ln CP, the logged infection rate. Clearly, above the line are Afghanistan, Iraq, Iran and Italy, and below are countries such as Tanzania, Nigeria, Turkey and Ukraine. It is very likely that the challenge is much more marked for developing countries in the longer run as soon as the spread is better measured and had more time to get into effect. Figure 3 deals with the transmission speed (ln T), which combines the previous two pictures confirming basically the relationship revealed by Figure 2. Finally, Figure 4 shows that the case fatality ratio (ln CFR) declines with larger globalisation (ln KOF over) with clear outliers Congo and Turkey above and Kuwait and Qatar below the line. We argue that the fatality statistic we are using is largely reflecting the infection situation before the global pandemic declaration on March 11 and the lockdown decisions of many countries that were executed only step by step thereafter. Therefore, Figure 4 does not inform about country response strategies and their success, which would be much too early to judge at the time when this paper was completed. Nevertheless, there are clear differences among the globalised countries, indicating initial diverse InKOF OVER policy stands: among others, the United Kingdom, the United States, Spain, Belgium and France are above the line, while Sweden is close to the line, and Finland, Norway, the Netherlands, South Korea and the Slovak Republic are below.
These core findings are confirmed by various regressions. Table 2 contains the OLS estimates of Equation (1) in four parts, each with the three alternative measures for globalisation as a robustness check. Globalised countries have consistently received the virus faster (D), with a higher infection rate (CP) and a higher transmission speed (TS), but also with a lower case fatality ratio (CFR). Transmission speed and both of its components D and CP exhibit estimates that all have 1% significance with coefficient sizes for KOF de jure that are somewhat smaller in absolute terms. This is plausible since the KOF de facto measure is more closely related to actual human mobility. The findings for the case fatality ratio confirm this insight: globalised economies seem to be more competitive in managing the infection, and the significance and size of the effect here comes primarily through KOF de facto, stressing the importance of human mobility. The KOF de facto coefficient is significant at 5% and much larger in absolute terms than the KOF de jure coefficient, which is significant only at 10%.
As found by Puhani (2020) and Wang et al. (2020), temperature differences play a role in the transition of the disease (see Table 2). However, the effect is statistically significant only for the duration to reach a country (lnD): warmer countries got the infection earlier. The age variables (age 65 + and median age) do not affect the day gap D at all, but a larger median age increases the infection rate (CP) and the transmission speed (TS), but reduces both with lower significance for the age 65 + variable. This may simply reflect the different exposure the captured age groups have to the virus due to their activities. A higher median age decreases the case fatality ratio (CFR), but a larger portion of age 65 + people increases CFR. These age effects are consistent with prior expectations that COVID-19 is more fatal in elderly people (see also Rothan & Byrareddy, 2020). Distance increases the day gap until infection but is insignificant afterwards. We also have assumed that distance has no effect on the case fatality ratio. Democracy exhibits practically no significant estimates throughout, and countries with more left governments face a smaller day gap for transition (D). Belt and Road partner countries of China are not negatively affected in any way: the infection rate (CP) is even lower for those countries, at least in the short-run period we are studying. The estimates for CP are significant at the 5% level, but the coefficients for day gap for transmission (D) and case fatality ratio (CFR) are not statistically significant at conventional levels.

| CONCLUSIONS
This study provides evidence that globalisation levels of countries affect the transmission speed of the coronavirus, both in terms of first arrival in a country, the infection rate and the fatality ratio. More globalised countries are affected faster and with a larger impact. This has to do with stronger human interactions through travel and migration. The implication is that pandemics can be contained through early measures of temporary inter-country distancing that focuses on human mobility. This is not an argument against globalisation, however, which makes countries wealthier, more competitive, and more able to invest in health infrastructures and through international collaborations (Dreher, 2006;   Potrafke, 2015). The effect can be clearly seen in the lower fatality rates provided in this study. However, the coronavirus crisis should stimulate debates about developing flexible systems to execute appropriate inter-country distancing measures and determining early indicators to trace future pandemic potentials. Trade policies can be designed to strengthen the effective exchange of diseaserelevant goods and services instead of hindering it.