The Virus That Knows No Borders? Border Permeability and the Global Diffusion of COVID-19

Ruud Koopmans WZB Berlin & Humboldt Universität Berlin 24 February 2020 • Gemeinsam, geschlossen, koordiniert: So will die Europäische Union eine Ausbreitung der gefährlichen Lungenkrankheit verhindern. Die EU-Staaten müssten so eng wie möglich zusammenarbeiten, um dieser Herausforderung zu begegnen, sagt Gesundheitskommissarin Stella Kyriakides. Anders ginge es nicht, da Viren keine Grenzen kennen würden. … Von nationalen Alleingängen bei Einschränkungen des Reiseverkehrs von und nach China - wie sie etwa Italien verhängt hat-, hält der deutsche Minister nichts. Solche Entscheidungen müssten europaweit getroffen werden. Auch die Forderung nach Fiebermessungen bei der Einreise lehnt Spahn entschieden ab. !13.2.2020) • Der deutsche Gesundheitsminister (CDU) hatte nach einem Treffen mit EU-Ressortkollegen in Rom erklärt: “Wir sind gemeinsam der Meinung, dass zu diesem Zeitpunkt Reisebeschränkungen oder gar das Schließen von Grenzen keine angemessene, verhältnismäßige Maßnahme wäre.” (17.2.2020) • The German government said Wednesday that it saw no need to advise its citizens against travel to Italy, or to restrict Italian air travel into . “We are far from this scenario,” a Foreign Ministry spokesman said during a press briefing in Berlin, as reported by Reuters (26.2.2020) • WHO continues to advise against the application of travel or trade restrictions to countries experiencing COVID-19 outbreaks. (recommendations of 29.2.2020) • Eine Schließung von Grenzen halte er weiter nicht für nötig. Grenzschließungen hätten massive Auswirkungen, sagte Spahn. Auch gegen den Stopp von Direktflügen zwischen China und Deutschland wandte sich Spahn. Ein solcher Schritt könne dazu führen, dass bis zu rund 30.000 Deutsche aus China ausgeflogen werden müssten (2.3.2020). • Spahn sprach sich aber, wie alle übrigen Minister, gegen staatliche Reisebeschränkungen in der EU aus. "Nicht notwendige Reisen sollten unterlassen werden, die Freizügigkeit bleibt erhalten", sagte Spahn. Die Grenzen blieben in der EU offen. Auch Einreiseverbote für Drittstaaten brächten aus Sicht der EU-Minister wenig. "Das Virus ist da. Es ist in Europa", stellte der Gesundheitsminister fest. (6.3.2020) • „Wir sind in Deutschland der Meinung, dass Grenzschließungen keine adäquate Antwort auf die Herausforderungen sind“, sagte die Bundeskanzlerin am Mittwoch in der Bundespressekonferenz in Berlin. Gesundheitsminister Spahn ergänzte: „Die Grenzen pauschal zu schließen verhindert nicht, was da passiert.“ (11.3.2020) • „Abriegelung ist naiv. Das Virus wird trotzdem kommen“, sagte Lothar Wieler, Präsident des Robert-Koch-Instituts. (11.3.2020) • Nicht nur die EU, auch die Außenpolitiker deutscher Parteien wurden von Trumps Maßnahme überrascht. Sie reagierten mit Unverständnis und Empörung. Auch ein gewisser Zynismus spricht aus den Äußerungen. „Ich bin nicht überzeugt, dass pauschale Einreisesperren geeignet sind, die Verbreitung des Coronavirus in den USA zu reduzieren. Immerhin gibt es in den USA bereits eine erhebliche Anzahl von Fällen“, sagte der außenpolitische Sprecher der Unionsfraktion im , Jürgen Hardt (CDU) (12.3.2020) • Der außenpolitische Sprecher der SPD-Fraktion im Bundestag, Nils Schmid, sagte: „Jetzt zählt das Gebot der engen Abstimmung und Koordinierung, um Vorsorgemaßnahmen zwischen den Staaten dieser Welt zu treffen. Denn zur Realität gehört, dass Covid-19 keine Grenzen kennt und längst in den USA angekommen ist.“ (12.3.2020) • “Certain controls may be justified, but general travel bans are not seen as being the most effective by the World Health Organization.“European Commission President said. (13.3.2020) „Border closures are not an adequate answer to the challenges “

• Die EU-Staaten müssten so eng wie möglich zusammenarbeiten, sagt Gesundheitskommissarin Stella Kyriakides. Anders ginge es nicht, da Viren keine Grenzen kennen. (13.2.2020)

• „The WHO continues to advise against the application of travel or trade restrictions to countries experiencing COVID-19 outbreaks“. (WHO recommendations 29.2.2020)

• Spahn sprach sich gegen staatliche Reisebeschränkungen in der EU aus. Auch Einreiseverbote für Drittstaaten brächten aus Sicht der EU-Minister wenig. "Das Virus ist da. Es ist in Europa" (6.3.2020)

• „Wir sind in Deutschland der Meinung, dass Grenzschließungen keine adäquate Antwort auf die Herausforderungen sind“, Bundeskanzlerin (11.3.2020)

• „Abriegelung ist naiv. Das Virus wird trotzdem kommen“. Lothar Wieler, Präsident des Robert-Koch-Instituts. (11.3.2020)

• Außenpolitischer Sprecher der SPD, Nils Schmid: „Jetzt zählt das Gebot der engen Abstimmung und Koordinierung, um Vorsorgemaßnahmen zwischen den Staaten dieser Welt zu treffen. Denn zur Realität gehört, dass Covid-19 keine Grenzen kennt und längst in den USA angekommen ist.“ (12.3.2020)

• “Certain controls may be justified, but general travel bans are not seen as being the most effective by the World Health Organization.“European Commission President Ursula von der Leyen (13.3.2020) …. still they came, albeit at widely diverging points in time

• On 24/2 Mauritius was the first country that closed its borders to and/or imposed quarantines on travellers from Italy. It was followed by, a.o. Israel (26/2), Turkey (29/2), Iceland (2/3), Czechia (5/3), Georgia (6/3), Romania (9/3), Austria, Slovenia, Spain, Moldova, Morocco (10/3), Australia, Malta, Hungary, Portugal (11/3) • On 13/3 the USA announced an entry ban for the entire Schengen Area, expanded a day later with the UK and Ireland • a.o. Israel (12/3), Slovakia (13/3), Denmark (14/3), Poland (15/3), Czechia (16/3) closed their borders entirely • Then, on 16/3, Germany closed its borders to Switzerland, Austria and France, on 18.3. also to air traffic from Italy • On 17/3, the EU closed ist borders to travelers from non-EU/Schengen countries • Almost all entry bans exclude own citizens and permanent residents; for these groups some countries introduced mandatory 14-day quarantines – Germany did so only from 16/4 Research Questions

• Does the degree to which countries are affected by COVID-19 depend on international travel flows? • Have entry restrictions really been futile to limit the spread of the virus? Theoretical framework: diffusion in social networks • Innovations, knowledge, news, etc. spread most rapidly along weak ties that connect socially and geographically distant individuals and groups. Difussion in strong-tie networks, by contrast, is characterized by higher levels of redundancy; • Looking at social networks from a global perspective, international ties are weak ties par excellence. The density of social interaction is much higher within than between nation-states; • The density of International ties determines exposure of a country to contagion: whether or not, and how early or how late a country is exposed to the virus, but also how many independent events of initial contagion („seeds“) occur. • H1: Countries that are more exposed to international travel will see earlier and more independent introductions of the virus and will therefore, ceteris paribus, have higher COVID-19 death rates; • H2: The earlier a country closes its borders, especially to countries with high infection rates, the less independent introductions of the virus will occur, and the lower will be the COVID-19 death rate. Theoretical framework 2: components of contagious growth • In the absence of any constraints, the spread of a contagious item will depend • 1) exponentially on time since first introduction; and • 2) linearly on the number of independent introductions („seeds“) • International travel is irrelevant to 1) but it is the only source of 2) Data

• RQ 1: all independent states with >50000 inhabitants, for which data on COVID-19 and international travel are available (n=177) • RQ 2 (n=102): All European countries (n=41), all non-European members of the OECD (n=10), plus further countries from all continents for which I have been able to code entry restrictions so far (n=51) > sample encompasses 81% of the world population; 99% of all corona deaths Variables • Number of COVID-19 deaths as of 15/5/2020 (per capita or logged) • Date of the 1st registered case of COVID-19, of the 10th case, of the 1st death and of the 10th death • International tourist arrivals, UN World Tourism Organization, 2017 • Date of introduction (if at all) of entry bans for travelers from China, Italy, and all foreign countries as well as of mandatory quarantines for travelers (including own citizens) from China, Italy, and all foreign countries (own coding from multiple sources) • Control variables : GDP per capita, Freedom House political rights and civil liberties score (or Polity), median age of the population (or % above 65), geographical region (Europe, Americas, MENA, Subsahara Africa, Asia-Pacific), population density (not shown because consistently insignificant) Results: deaths per 100000 inhabitants (15/5/20) Baseline Plus tourist Plus Plus entry Football Europe model arrivals tourist restrictions only arrivals GDP per capita +++ +++ ++ + n.s. Median age - (-) n.s. n.s. Democracy n.s. n.s. n.s. n.s. n.s. Region (ref. Asia-Pacific)

Europe +++ +++ ++ + Americas n.s. n.s. n.s. n.s. MENA n.s. n.s. n.s. n.s. Subsahara Africa n.s. n.s. n.s. n.s. Tourist arrivals (millions) +++ +++ ++ ++ B=.32 B=.33 B=.30 B=.34 Average date of six entry -- - restriction types (lower is B= -.17 B=-.34 earlier) R² .34 .45 .42 .47 .46 N= 176 176 100 100 46 Results: logged deaths 15/5/20

Baseline Plus tourist Plus Plus entry Football Europe model arrivals tourist restrictions only arrivals Logged population +++ +++ +++ +++ +++ GDP per capita +++ +++ +++ +++ + Median age + n.s. n.s. n.s. n.s. Democracy (+) (+) n.s. n.s. n.s. Region (ref. Asia-Pacific) Europe +++ +++ +++ ++ Americas +++ +++ +++ n.s. MENA +++ +++ +++ n.s. Subsahara Africa (+) n.s. n.s. n.s. Tourist arrivals (millions) +++ (+) n.s. (+) B=.035 B=.018 B=.010 B=.017 Entry restrictions scale ------B= -1.52 B=-1.66

R² .74 .76 .85 .88 .87 N= 176 176 101 101 47 Date of introduction of Italy travel restrictions and timing relative to the epidemic in selected European countries

80

70

60

50

40

Deaths per 100000 15/5 30 Deaths at introduction Italy restrictions

20

10

0 Sweden Russia (5/3) Russia Israel (26/2) Israel Spain (10/3) Spain France (6/4) France Malta (11/3) Malta Iceland (2/3) Iceland Turkey (29/2) Turkey Poland (15/3) Poland Ireland (13/3) Ireland Austria (10/3) Austria Greece (14/3) Greece Croatia (13/3) Croatia Estonia (17/3) Estonia (19/3) Finland Romania (2/3) Romania Norway (12/3) Norway Belgium (20/3) Belgium Slovakia (13/3) Slovakia Portugal (11/3) Portugal (11/3) Hungary Slovenia (10/3) Slovenia Czech Rep (5/3) Rep Czech Denmark (14/3) Denmark Germany (18/3) Germany Lithuania (16/3) Lithuania United Kingdom United Switzerland (13/3) Switzerland Netherlands (13/3) Netherlands Results: deaths per 100000 inhabitants (15/5/20) All countries that Football Europe ever had a 10th countries that had a death 10th death GDP per capita (+) n.s. Median age n.s. n.s. Democracy n.s. n.s. Region (ref. Asia-Pacific) Europe + Americas n.s. MENA n.s. Subsahara Africa n.s. Tourist arrivals (millions) + n.s. B=.18 B=.10 Timing of entry restrictions (scale from 1 = all before ------10th death to 0 = all after 10th death or never) B= - 16.6 B= - 28.2 Timing of 10th death n.s. n.s. R² .53 .59 N= 77 44 COVID-19 deaths per 100.000 inhabitants in European countries that introduced an entry ban against travelers from Italy before (right in green) oder after (left in red) the 10th domestic death

80 80

70 70

60 60

50 50

40 40

30 30

20 20

10 10

0 0 A CZ TR DK LIT HR ISR B And E UK F S NL IRL CHLux P D ICE FIN EST HEL CYP SVK POL BUL RUS SVN MLT GEO NOR HUN ROM Mold Robustness checks

• Logged deaths as alternative dependent variable: substantively identical results • Controlling additionally for number of deaths on 20 March, assuming that in most cases these cannot yet reflect effects of entry restrictions (e.g., the average date of Italy entry bans was 11 March): entry restrictions effect only slightly reduced (B=-.13 resp. -12.3), i.e. entry restrictions explain differences across countries controlling for pre-intervention differences • Only including countries that were hit early (at least 10 cases by 8 March, the day of the Italy lockdown; n=45): entry restriction effects only become stronger • Restricting the sample to democracies and/or to richer countries (per capita GDP>10000USD): effects only become stronger • Neither the date of school closures nor of lockdowns (from Oxford COVID government response tracker) has any significant impact on cumulative death rates. Entry restrictions remain highly significant in the presence of these other policy measures • Number of tests per capita (never sign.) or test/case ratio (sign. negative effect in logged deaths regression) do not affect the estimates of interest • Alternative measurement of international travel restrictions from Oxford tracker: substantively similar results, both on same sample and on full Oxford global sample More robustness checks

• Controlling additionally for test/case ratio (which may indicate inclusive testing protocols or high test capacity relative to caseload): The test capacity measure is not significant in the per capita regression. It is highly significant in the expected direction in the logged deaths regression but the entry restrictions variable remains highly significant though somewhat reduced in size (partly because low case numbers drive up this measure at constant test resources) • Excluding Italy and Iran: results virtually identical • Excluding smaller states with less than 1 million inhabitants : results nearly identical What about China and Italy?

• On 23 january, China implemented an internal travel ban, precluding all travel in and out of Hubei province. At the time, 17 COVID deaths had been recorded, all of which had occurred within Hubei province. The low per capita death rate in China thus fully conforms to the results (see also Kraemer et al., Science 1 May 2020) • Italy implemented a quarantine of a few municipalities (54,000 inhabitants) on 22 February. At the time, 2 deaths had been recorded. On 8 March Italy banned travel in and out of 14 Northern provinces in 5 regions (with some 16 million inhabitants). By then, 366 people had already died, at least 13 of which outside the quarantine zone. Italy thus implemented travel restrictions much too late and conforms to the results. • But why did Italy see an outbreak at all, even though it had cancelled all flights to and from China early on (on 31 January)? > May just have been bad luck (two Chinese tourists had already tested positive in Italy before the ban) or import of Chinese cases continued in spite of the ban through indirect entry via other Schengen countries. The origin of the infection of „patient zero“ of the Northern Italy outbreak (a manager of a local Unilever branch) remains as yet unknown. …and what about the German „success story“?

COVID-19 death rates along Germany‘s borders with Poland, Czechia and Austria

18

16

14

12

10

8

6

4

2

0 Questions

• Better (dyadic) data on international travel? • How to deal more adequately with spatial correlation? • How to get closer to causal proof? Thank you for your attention!