## wpiea2020112-print-pdf

## Source details

**Canonical URL:** [wpiea2020112-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020112-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020112-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020112-print-pdf.pdf.json)

---

### I. Introduction — context and motivation
- COVID-19 described as a global crisis "in a staggeringly short space of time like no other in modern history."
- As of June 30, 2020, "over 10.5 million confirmed cases of COVID-19 in 188 countries, with more than 512,500 deaths."
- No specific vaccines or treatments at the time; mitigation and containment measures (travel restrictions, business and school closings, social distancing) are primary means to prevent and slow spread.
- Tourism sector disruption:
  - Tourism characterized as "a leading engine [of] economic growth and the major source of foreign exchange earnings in many countries."
  - Preliminary data: "a fall of 22 percent in international tourist arrivals in the first quarter of 2020, with a 57 percent drop in March, compared to 2019."
- Historical context: cross-border movement as vector for infectious diseases (examples: avian flu, cholera, Ebola, malaria, SARS, MERS, COVID-19).

### Data, scope, and methodological contributions
- Dataset:
  - Country coverage expanded to "38,184 pairs of countries over the period 1995–2017."
  - Measures actual number of confirmed infectious-disease cases (Ebola, malaria, SARS, yellow fever) scaled by population.
  - Bilateral tourism flows data cover 172 countries of origin and 222 countries of destination, yielding over 261,488 observations.
- Econometric approaches within a gravity framework controlling for macroeconomic, geographic, cultural, and historical factors:
  - Ordinary least squares (OLS)
  - Pseudo poisson maximum likelihood (PPML)
  - Two-stage least squares (2SLS) with instrumental variable (IV)
- Literature gap: models of bilateral tourist movements that capture scale and dynamism of infectious-disease episodes are scarce.

### Data overview (selected descriptive statistics)
- Sample: unbalanced panel of annual observations for 38,184 pairs of countries during 1995–2017.
- Bilateral tourism flows:
  - Observations 261,488; Mean 78,737; Std. Dev. 918,199; Min 1.08; Max 1,100,000
- Real GDP:
  - Origin Observations 410,680; Mean 13,155; Std. Dev. 16,243; Min 184; Max 111,968
  - Destination Observations 381,210; Mean 16,926; Std. Dev. 22,269; Min 184; Max 194,188
- Distance:
  - Observations 370,208; Mean 7,273; Std. Dev. 4,564; Min 271; Max 19,951 (great-circle km between capital cities)
- Population:
  - Origin Observations 411,470; Mean 54,300,000; Std. Dev. 186,000,000; Min 9,298; Max 1,390,000,000
  - Destination Observations 395,038; Mean 45,900,000; Std. Dev. 160,000,000; Min 9,298; Max 1,400,000,000
- Infectious diseases (selected):
  - Ebola Observations 382,651; Mean 2.38; Std. Dev. 90.03; Max 3,811
  - Malaria Observations 382,651; Mean 62,369; Std. Dev. 531,507; Min 0.01; Max 5,000,000
  - SARS Observations 382,651; Mean 312; Std. Dev. 368; Min 0.05; Max 5,327
  - Yellow fever Observations 384,194; Mean 2.03; Std. Dev. 1.01; Min 0; Max 1,192
- Stationarity: variables stationary after logarithmic transformation per Im-Pesaran-Shin (2003) panel unit root test (results available upon request).

### Empirical methodology
- Baseline: gravity model of bilateral tourist flows, log-linear panel with origin, destination, and year fixed effects.
- Controls Xij: log population of origin and destination; binary indicators for common language, colonial history, geographical contiguity.
- Main explanatory variables: number of confirmed cases of Ebola, malaria, SARS, yellow fever scaled by population in destination countries.
- Estimators and identification:
  - OLS (random-effects regression used in some specifications to retain time-invariant gravity factors)
  - PPML (to address zero flows and heteroskedasticity)
  - 2SLS-IV using lagged infectious disease as instrument (to address potential endogeneity)
- Robust standard errors clustered at the country-pair level.

### Estimation results — baseline and robustness (key findings)
- Model fit:
  - Adjusted R-squared around 0.83 in OLS baseline.
  - Pseudo R-squared in PPML around 0.77–0.79.
- Gravity and controls (examples from OLS Table 2 [1]):
  - Real GDP, origin 0.956***; Real GDP, destination 0.833***.
  - Distance example -1.701***.
  - Common language, colonial history, geographical contiguity: positive and generally significant.
  - Population (origin and destination): positive and significant.
- Infectious-disease coefficients:
  - Range between -0.003 and -0.785 depending on disease; negative and statistically significant except yellow fever in baseline OLS (yellow fever coefficient -0.003, not significant).
  - Larger negative impacts for Ebola and SARS (human-to-human transmission, no treatments/vaccines in discussed contexts).
  - Smaller impacts for malaria and yellow fever (vector-borne, treatments/vaccines exist; often endemic).
- Quantitative examples and elasticities (preserve reported magnitudes):
  - OLS baseline SARS coefficient example: -0.785*** (Table 2, column [4]).
  - PPML examples: Ebola -0.011***; SARS -0.032*** (Table 3).
  - Textual elasticity example: "a 10 percent increase in the number of confirmed cases of SARS is associated, on average, with a decline of about 8 percent in bilateral tourism flows."
  - Conclusion/robustness textual statement: "a 10 percent increase in the number of confirmed cases of SARS is found to lead, on average, to a reduction of as much as 9 percent in international tourist arrivals."
  - 2SLS-IV: SARS coefficient magnitude increased to -0.86 (compared with -0.79 baseline).
  - 2SLS-IV income-group heterogeneity: 10 percent increase in confirmed SARS cases → decline of 3.2 percent in bilateral tourism flows in advanced economies, but almost 12 percent in developing countries.
- Robustness checks:
  - Truncated sample at 5th and 95th percentiles; sub-sample 1995-2007; inclusion of life expectancy and hospital beds per 1,000 people; 2SLS-IV with lagged infectious disease as instrument.
  - Results: negative relationship between infectious diseases and international tourism flows robust across checks.
  - Truncated and pre-crisis samples yield higher coefficients on the infectious-disease variable.
  - Health variables (life expectancy; hospital beds per 1,000 people) matter for bilateral tourism flows.
  - 2SLS-IV increases SARS magnitude to -0.86.

### Robustness table highlights (selected coefficients and statistics)
- Truncated sample | Sub-sample (1995-2007) | Additional controls | Instrumental variable (columns)
  - Real GDP, origin: 0.907*** [0.050] | 1.100*** [0.082] | 0.995*** [0.063] | 0.897*** [0.033]
  - Real GDP, destination: 0.750*** [0.045] | 0.823*** [0.069] | 0.989*** [0.067] | 0.786*** [0.032]
  - Distance: -1.621*** [0.022] | -1.704*** [0.026] | -1.719** [0.026] | -1.699*** [0.006]
  - Common language: 1.165*** [0.046] | 1.111*** [0.056] | 1.214*** [0.055] | 1.237*** [0.014]
  - SARS: -0.828*** [0.102] | -1.191*** [0.157] | -0.608*** [0.140] | -0.858*** [0.059]
  - Number of observations: 104,184 | 54,123 | 56,587 | 112,206
  - Adjusted R2: 0.78 | 0.82 | 0.82 | 0.83
- Additional health controls (selected):
  - Life expectancy, destination 0.704* [0.401] (appears in one specification)
  - Hospital beds, destination 0.047** [0.025] (appears in one specification)

### Heterogeneity by income group and region (2SLS-IV highlights)
- Income-group split (Advanced | Developing):
  - Real GDP, origin 0.995*** [0.017] | 0.880*** [0.023]
  - Real GDP, destination 1.264*** [0.035] | 0.638*** [0.022]
  - SARS: -0.317 [0.135] (Advanced) | -1.109*** [0.212] (Developing)
  - Number of observations: 72,859 (Advanced) | 79,744 (Developing)
  - Adjusted R2: 0.88 (Advanced) | 0.81 (Developing)
- Regional differences (selected columns and SARS coefficients):
  - Africa SARS -1.472*** [0.446] (23,016 observations)
  - Asia SARS -3.736*** [0.622] (15,612 observations)
  - Europe SARS -1.544* [0.576] (13,073 observations)
  - Latin America SARS -4.462*** [0.630] (14,102 observations)
  - Middle East SARS -0.615 [0.406] (13,941 observations)
- Interpretation:
  - Infectious-disease impacts are significantly greater in Asia and Latin America and the Caribbean than in other regions.
  - Advanced economies show smaller and sometimes statistically insignificant negative effects compared with developing countries.

### Conclusions and policy implications
- Main conclusions:
  - Strong evidence that international tourism is adversely affected by the risk of infectious diseases measured by confirmed cases.
  - Negative effects are larger and statistically significant for Ebola and SARS (human-to-human transmission, no treatment/vaccine in discussed contexts) compared with malaria and yellow fever.
  - Findings robust across specifications and estimation methods.
  - Health conditions and infrastructure in destination countries (life expectancy; hospital beds per 1,000 people) materially affect bilateral tourism flows.
  - Geographic concentration matters: as of 2019, over 50 percent of international tourist arrivals worldwide were bound for Europe; Africa accounted for 4.8 percent; Asia accounted for 24.5 percent in 2019 (up from 12.8 percent in 1990 and 3.7 percent in 1970).
  - COVID-19 characterized as the first truly global pandemic in modern times, making geographical differentiation impossible for tourism and causing a collapse of international travel.
    - WTO preliminary expectation: international tourist arrivals to decline by 60 to 80 percent in 2020 due to closures, travel bans, and containment measures.
- Policy recommendations:
  - Immediate priority: ensure adequate healthcare resources to protect the population, take care of the sick, and slow the spread of the coronavirus; addressing public health concerns is a prerequisite for economic recovery and the return of international tourists.
  - Economic policy: well-targeted fiscal and financial measures to mitigate the pandemic’s impact, including:
    - Timely and targeted cash and in-kind transfers
    - Wage subsidies
    - Loans and grants
    - Tax relief
  - Special attention to tourism-dependent developing countries with limited fiscal space; targeted support is essential.
  - Broader recovery measures: preserve financial stability; reform labor and product markets; strengthen human and physical capital; build a more conducive environment for investment.
  - Outcome: countries managing public health consequences better and preventing excessive economic disruption will recover faster toward long-term potential.

### Appendices (selected)
- Appendix Table A1: comprehensive List of Countries and Territories (alphabetical list of countries and territories as provided).
- Appendix Table A2: Robustness Checks (dependent variable: bilateral tourism flows in log form) — selected coefficient values and standard errors reported above.
- Appendix Table A3: Income Groups and Regions (2SLS-IV estimations) — selected coefficient values and standard errors reported above.

*Source: wpiea2020112-print-pdf - References .........................................................................................................*

### References .............................................................................................................

### wpiea2020112-print-pdf - References .............................................................................................................

### I. INTRODUCTION — context and motivation
- The COVID-19 pandemic is described as a global crisis unfolding "in a staggeringly short space of time like no other in modern history."
- As of June 30, 2020, there are "over 10.5 million confirmed cases of COVID-19 in 188 countries, with more than 512,500 deaths."
- With no specific vaccines or treatments at this time, mitigation and containment measures (travel restrictions, business and school closings, social distancing) are the primary means to prevent and slow spread, according to WHO.
- The pandemic has caused a collapse in international travel and tourism, a sector described as "a leading engine [of] economic growth and the major source of foreign exchange earnings in many countries" (see Figure 1).
- Preliminary data indicate the COVID-19 pandemic "already caused a fall of 22 percent in international tourist arrivals in the first quarter of 2020, with a 57 percent drop in March, compared to 2019."
- Historical context: cross-border movement of people has long been a potent vector for infectious diseases (McNeill, 1976; Wilson, 1991; Wilson,1995; Richter, 2003; Snowden, 2019). Examples include avian flu, cholera, Ebola, malaria, SARS, MERS, and COVID-19.

### Data, scope, and methodological contributions
- Dataset expansion and measurement:
  - Country coverage and time-series dimension expanded to "38,184 pairs of countries over the period 1995–2017."
  - Measures "the actual number of confirmed infectious-disease cases (Ebola, malaria, SARS, and yellow fever) scaled by population."
- Econometric approaches employed within a gravity framework that controls for macroeconomic, geographic, cultural, and historical factors:
  - Ordinary least squares (OLS)
  - Pseudo poisson maximum likelihood (PPML)
  - Two-stage least squares (2SLS) with instrumental variable (IV)
- Literature gap addressed: scarce research modeling bilateral tourist movements in a gravity framework that accounts for infectious diseases; existing studies often focus on single outbreaks or use dummies that do not capture scale/dynamism.

### Empirical findings and quantitative effects
- General result: "Strong evidence that international tourism is adversely affected by the risk of infectious diseases" with effects that are "statistically and economically significant."
- SARS example:
  - "A 10 percent increase in the number of confirmed cases leads to, on average, a reduction of as much as 9 percent in international tourist arrivals."
- Heterogeneity by income group:
  - Infectious diseases have a "smaller and statistically insignificant negative effect on tourism flows to advanced economies."
  - Larger and statistically significant impacts in developing countries, where diseases are more prevalent and health infrastructure lags.
  - SARS-specific heterogeneity:
    - "10 percent increase in the number of infections leads to a decline of 3.2 percent in bilateral tourism flows in advanced economies, but almost 12 percent in developing countries."
- Robustness: Results "withstand several robustness checks, including alternative specifications and estimation methodologies, to address omitted-variables bias and potential endogeneity."

### Policy implications and recommendations
- Immediate priority: "Ensuring adequate healthcare resources to protect the population, take care of the sick, and slow the spread of the coronavirus."
- Addressing public health concerns is framed as a prerequisite for economic recovery and the return of international tourists.
- Economic policy: "Well-targeted fiscal and financial measures are critical to mitigate the pandemic’s impact," given unequal distribution of consequences across the economy.
- Countries heavily dependent on tourism face severe effects; those "managing the consequences of the coronavirus pandemic better, thus preventing excessive economic disruption, will recover faster toward their long-term potential."

### Organization of the remainder of the study
- Section II: overview of the data used in the empirical analysis.
- Section III: econometric strategy.
- Section IV: empirical results, including robustness checks.

*Source: wpiea2020112-print-pdf - References .........................................................................................................*

### Section V offers concluding remarks with policy implications.

### wpiea2020112-print-pdf - Section V offers concluding remarks with policy implications

### Data overview
- Sample: unbalanced panel of annual observations for 38,184 pairs of countries during the period 1995–2017.
- Bilateral tourism flows: data for 172 countries of origin and 222 countries of destination from the WTO database, yielding a dataset of over 261,488 observations over the sample period.
- Main explanatory variable: number of confirmed infectious-disease cases (Ebola, malaria, SARS, yellow fever) from the WHO database.
- Control variables: real GDP and population (IMF WEO and World Bank WDI); gravity variables (bilateral distance, common official language, colonial history, geographical contiguity) from CEPII; geographic distance measured as great-circle distance in kilometers between capital cities.
- Stationarity: variables are stationary after logarithmic transformation according to the Im-Pesaran-Shin (2003) panel unit root test (results available upon request).
- Descriptive statistics (selected, as presented in Table 1):
  - International tourism flows: Observations 261,488; Mean 78,737; Std. Dev. 918,199; Min 1.08; Max 1,100,000
  - Real GDP, origin: Observations 410,680; Mean 13,155; Std. Dev. 16,243; Min 184; Max 111,968
  - Real GDP, destination: Observations 381,210; Mean 16,926; Std. Dev. 22,269; Min 184; Max 194,188
  - Distance: Observations 370,208; Mean 7,273; Std. Dev. 4,564; Min 271; Max 19,951
  - Population, origin: Observations 411,470; Mean 54,300,000; Std. Dev. 186,000,000; Min 9,298; Max 1,390,000,000
  - Population, destination: Observations 395,038; Mean 45,900,000; Std. Dev. 160,000,000; Min 9,298; Max 1,400,000,000
  - Infectious diseases (selected):
    - Ebola: Observations 382,651; Mean 2.38; Std. Dev. 90.03; Max 3,811
    - Malaria: Observations 382,651; Mean 62,369; Std. Dev. 531,507; Min 0.01; Max 5,000,000
    - SARS: Observations 382,651; Mean 312; Std. Dev. 368; Min 0.05; Max 5,327
    - Yellow fever: Observations 384,194; Mean 2.03; Std. Dev. 1.01; Min 0; Max 1,192

### Empirical methodology
- Baseline framework: gravity model of bilateral tourist flows transformed into log-linear panel specification with origin and destination fixed effects and year fixed effects.
- Baseline equation augmented with:
  - Controls Xij: log population of origin and destination; binary variables for common language, colonial history, geographical contiguity.
  - V_iV_j: number of confirmed cases of Ebola, malaria, SARS, yellow fever scaled by population in destination countries.
- Estimation approaches:
  - OLS (random-effects regression used to retain time-invariant gravity factors)
  - PPML estimator (Santos Silva and Tenreyro, 2006) to address zero flows and heteroskedasticity
  - 2SLS-IV using lagged infectious disease as instrument to account for potential endogeneity
- Robust standard errors clustered at the country-pair level.

### Estimation results — baseline and robustness
- Model fit: Adjusted R-squared around 0.83 in OLS baseline; Pseudo R-squared in PPML around 0.77–0.79.
- Standard gravity and control variable findings (consistent across specifications):
  - Real GDP, origin and destination: positive and highly significant (examples from OLS Table 2: Real GDP, origin 0.956*** in [1]; Real GDP, destination 0.833*** in [1]).
  - Distance: negative and highly significant (example: -1.701*** in [1]).
  - Common language, colonial history, geographical contiguity: positive and generally statistically significant.
  - Population (origin and destination): positive and significant.
- Main variable of interest — infectious diseases:
  - Coefficients on infectious diseases range between -0.003 and -0.785 depending on the disease, always negative and statistically significant except in the case of yellow fever (where coefficient is -0.003 and not statistically significant in baseline OLS).
  - Disease-specific magnitudes and interpretation:
    - Ebola and SARS: both statistically and economically significant with larger negative impacts.
    - Malaria and yellow fever: considerably smaller magnitude.
  - Disease characteristics likely driving differences:
    - Vector of transmission: malaria and yellow fever transmitted by mosquitoes; Ebola and SARS transmitted human-to-human.
    - Existence of treatment or vaccine: vaccine exists against yellow fever; treatments exist against malaria; no treatment or vaccine for Ebola or SARS in the contexts discussed.
    - Temporary outbreak vs endemic presence: endemic diseases (malaria, yellow fever) may not induce travel delays; outbreaks (Ebola, SARS) incentivize delaying travel.
- Quantitative examples:
  - OLS baseline: SARS coefficient example -0.785*** (Table 2, column [4]).
  - PPML estimates produce smaller magnitudes but remain statistically significant (example: Ebola -0.011*** in PPML; SARS -0.032*** in PPML; Table 3).
  - Case elasticity interpretations:
    - In one specification: a 10 percent increase in the number of confirmed cases of SARS is associated, on average, with a decline of about 8 percent in bilateral tourism flows (textual example).
    - In conclusion and robustness discussion: a 10 percent increase in the number of confirmed cases of SARS is found to lead, on average, to a reduction of as much as 9 percent in international tourist arrivals.
    - 2SLS-IV results: increase the magnitude of the coefficient on SARS to -0.86 compared to -0.79 in the baseline estimation.
    - Income-group heterogeneity (baseline 2SLS-IV): 10 percent increase in confirmed SARS cases leads to a decline of 3.2 percent in bilateral tourism flows in advanced economies, but almost 12 percent in developing countries.
- Robustness checks conducted:
  - Truncated sample at the 5th and 95th percentiles to remove outliers.
  - Sub-sample 1995-2007 excluding the post-global financial crisis period.
  - Inclusion of health-related variables: life expectancy and number of hospital beds per 1,000 people.
  - 2SLS-IV with lagged infectious disease to address potential reverse causality.
  - Results: negative relationship between infectious diseases and international tourism flows remains robust across checks; truncated and pre-crisis samples yield higher coefficients on the infectious-disease variable; health variables matter for bilateral tourism flows; 2SLS-IV increases SARS magnitude to -0.86.

### Heterogeneity by income group and region
- Advanced vs developing countries:
  - Infectious diseases (SARS example) have a smaller and statistically insignificant negative effect on tourism flows to advanced economies in some specifications.
  - Much greater magnitude and statistical significance in developing countries, where prevalence is higher and health infrastructure lags.
  - Quantified effect (2SLS-IV baseline): 10 percent increase in SARS cases → decline of 3.2 percent in advanced economies, almost 12 percent in developing countries.
- Regional differences:
  - Disease impact on bilateral tourism flows is significantly greater in Asia and Latin America and the Caribbean than the rest of the world.

### Conclusion and policy implications
- Main conclusions:
  - International tourism is adversely affected by the risk of infectious diseases measured by the number of confirmed cases in past episodes.
  - The negative effect is statistically and economically greater for Ebola and SARS—human-to-human transmitted diseases without treatment or vaccine—compared with malaria and yellow fever.
  - Results are robust to alternative specifications and estimation methodologies.
  - Health conditions and infrastructure in destination countries (life expectancy; number of hospital beds per 1,000 people) matter for bilateral tourism flows.
  - Geographic concentration of tourist arrivals matters for the global impact of outbreaks:
    - As of 2019, over 50 percent of international tourist arrivals worldwide were bound for Europe.
    - Africa accounted for 4.8 percent of international tourist arrivals.
    - Asia accounted for 24.5 percent of international tourist arrivals in 2019, up from 12.8 percent in 1990 and 3.7 percent in 1970.
  - COVID-19 is characterized as the first truly global pandemic in modern times, making geographical differentiation impossible for tourism and leading to a collapse of international travel.
    - WTO preliminary expectation: international tourist arrivals to decline by 60 to 80 percent in 2020 due to closures, travel bans, and containment measures.
- Policy implications and recommendations:
  - Immediate priority: ensure adequate resources for healthcare systems to protect the population, take care of the sick, and slow the spread of the coronavirus. Addressing public health concerns will set the stage for economic recovery and the return of international tourists.
  - Economic policy response: well-targeted fiscal and financial measures are critical to mitigate the pandemic’s impact, including:
    - Timely and targeted cash and in-kind transfers
    - Wage subsidies
    - Loans and grants
    - Tax relief to help households and businesses under strain confront the temporary and sudden stop in economic activity
  - Special attention for tourism-dependent developing countries: many will have limited fiscal space to respond; targeted support is essential.
  - Broader measures to sustain recovery: preserve financial stability; reform labor and product markets; strengthen human and physical capital; build a more conducive environment for investment.
  - Countries that manage public health consequences better and prevent excessive economic disruption will recover faster toward long-term potential.

*Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020112-print-pdf.pdf*

### Appendix Table A1. List of Countries and Territories

### Appendix Table A1. List of Countries and Territories

### List of Countries and Territories
- AfghanistanDenmarkLiberiaRwanda
- AlbaniaDjiboutiLibyaSaba
- AlgeriaDominicaLiechtensteinSaint Eustatius
- American SamoaDominican RepublicLithuaniaSaint Maarten
- AndorraEcuadorLuxembourgSamoa
- AngolaEgyptMacao SARSan Marino
- AnguillaEl SalvadorMadagascarSao Tome And Principe
- Antigua And BarbudaEquatorial GuineaMalawiSaudi Arabia
- ArgentinaEritreaMalaysiaSenegal
- ArmeniaEstoniaMaldivesSerbia
- ArubaEswatiniMaliSeychelles
- AustraliaEthiopiaMaltaSierra Leone
- AustriaFijiMarshall IslandsSingapore
- AzerbaijanFinlandMartiniqueSlovak Republic
- Bahamas, TheFranceMauritaniaSlovenia
- BahrainFrench GuianaMauritiusSolomon Islands
- BangladeshFrench PolynesiaMexicoSomalia
- BarbadosGabonMicronesiaSouth Africa
- BelarusGambia, theMoldovaSouth Sudan
- BelgiumGeorgiaMonacoSpain
- BelizeGermanyMongoliaSri Lanka
- BeninGhanaMontenegroSt. Kitts and Nevis
- BermudaGreeceMontserratSt. Lucia
- BhutanGrenadaMoroccoSt. Vincent and the Grenadines
- BoliviaGuadeloupeMozambiqueSudan
- BonaireGuamMyanmarSuriname
- Bosnia And HerzegovinaGuatemalaNamibiaSweden
- BotswanaGuineaNauruSwitzerland
- BrazilGuinea-BissauNepalSyria
- British Virgin IslandsGuyanaNetherlandsTaiwan Province of China
- Brunei DarussalamHaitiNew CaledoniaTajikistan
- BulgariaHondurasNew ZealandTanzania
- Burkina FasoHong Kong SARNicaraguaThailand
- BurundiHungaryNigerTimor-Leste
- Cabo VerdeIcelandNigeriaTogo
- CambodiaIndiaNiueTonga
- CameroonIndonesiaNorth KoreaTrinidad And Tobago
- CanadaIranNorth Macedonia Tunisia
- Cayman IslandsIraqNorthern Mariana IslandsTurkey
- Central African RepublicIrelandNorwayTurkmenistan
- ChadIsraelOmanTurks And Caicos Islands
- ChileItalyPakistanTuvalu
- ChinaJamaicaPalauUganda
- ColombiaJapanPalestineUkraine
- ComorosJordanPanamaUnited Arab Emirates
- Congo, Republic ofKazakhstanPapua New GuineaUnited Kingdom
- Cook IslandsKenyaParaguayUnited States
- Costa RicaKiribatiPeruUnited States Virgin Islands
- Côte d'IvoireKoreaPhilippinesUruguay
- CroatiaKuwaitPolandUzbekistan
- CubaKyrgyz RepublicPortugalVanuatu
- CuracaoLao P.D.R.Puerto RicoVenezuela
- CyprusLatviaQatarVietnam
- Czech RepublicLebanonReunionYemen
- Democratic Republic Of The CongoLesothoRomaniaZambia
- RussiaZimbabwe

### Appendix Table A2. Robustness Checks (Dependent variable: Bilateral tourism flows)
- Note: The dependent variable is bilateral tourism flows (in log form). Robust standard errors, clustered at the country level, are reported in brackets. A constant is included in each regression, but not shown in the table. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

- Columns: Truncated sample | Sub-sample (1995-2007) | Additional controls | Instrumental variable

- Real GDP, origin
  - 0.907*** [0.050]
  - 1.100*** [0.082]
  - 0.995*** [0.063]
  - 0.897*** [0.033]

- Real GDP, destination
  - 0.750*** [0.045]
  - 0.823*** [0.069]
  - 0.989*** [0.067]
  - 0.786*** [0.032]

- Distance
  - -1.621*** [0.022]
  - -1.704*** [0.026]
  - -1.719** [0.026]
  - -1.699*** [0.006]

- Common language
  - 1.165*** [0.046]
  - 1.111*** [0.056]
  - 1.214*** [0.055]
  - 1.237*** [0.014]

- Colonial history
  - 1.014*** [0.149]
  - 1.060*** [0.142]
  - 1.008*** [0.144]
  - 0.948*** [0.036]

- Geographical contiguity
  - 1.355*** [0.137]
  - 1.324*** [0.136]
  - 1.332** [0.138]
  - 1.432*** [0.025]

- Population, origin
  - 0.551*** [0.88]
  - 1.196*** [0.158]
  - 0.596*** [0.119]
  - 0.377*** [0.059]

- Population, destination
  - 0.254** [0.083]
  - 0.456*** [0.136]
  - 0.444*** [0.118]
  - 0.290*** [0.050]

- Life expectancy, destination
  - 0.704* [0.401] (appears in one specification)

- Hospital beds, destination
  - 0.047** [0.025] (appears in one specification)

- SARS
  - -0.828*** [0.102]
  - -1.191*** [0.157]
  - -0.608*** [0.140]
  - -0.858*** [0.059]

- Number of observations
  - 104,184
  - 54,123
  - 56,587
  - 112,206

- Fixed effects and fit
  - Origin FE: Yes | Yes | Yes | Yes
  - Destination FE: Yes | Yes | Yes | Yes
  - Year FE: Yes | Yes | Yes | Yes
  - Adjusted R2: 0.78 | 0.82 | 0.82 | 0.83

### Appendix Table A3. Income Groups and Regions (2SLS-IV Estimations) (Dependent variable: Bilateral tourism flows)
- Note: The dependent variable is bilateral tourism flows (in log form). Robust standard errors, clustered at the country level, are reported in brackets. A constant is included in each regression, but not shown in the table. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

- Columns: Advanced | Developing | Africa | Asia | Europe | Latin America | Middle East

- Real GDP, origin
  - 0.995*** [0.017]
  - 0.880*** [0.023]
  - 0.886*** [0.043]
  - 0.949*** [0.054]
  - 0.803*** [0.053]
  - 0.762*** [0.049]
  - 1.028*** [0.057

- Real GDP, destination
  - 1.264*** [0.035]
  - 0.638*** [0.022]
  - 0.407*** [0.049]
  - 0.923*** [0.051]
  - 0.335*** [0.080]
  - 0.428*** [0.058]
  - 0.443*** [0.056]

- Distance
  - -1.385*** [0.026]
  - -1.810*** [0.026]
  - -1.533** [0.068]
  - -1.778*** [0.103]
  - -1.563*** [0.120]
  - -1.798*** [0.074]
  - -1.426*** [0.084]

- Common language
  - 0.562*** [0.053]
  - 1.400*** [0.053]
  - 1.199*** [0.078]
  - 0.531*** [0.140]
  - 0.336 [0.501]
  - 1.480*** [0.106]
  - 0.457** [0.147]

- Colonial history
  - 1.174*** [0.083]
  - 0.865*** [0.178]
  - -0.030 [0.574]
  - 1.513** [0.522]
  - -0.183 [0.254]
  - 0.569 [0.941]
  - 0.764* [0.451]

- Geographical contiguity
  - 0.410 [0.106]
  - 1.562*** [0.107]
  - 1.332** [0.138]
  - 1.442*** [0.275]
  - 1.420*** [0.223]
  - 1.106*** [0.263]
  - 1.191*** [0.262]

- Population, origin
  - 0.196*** [0.091]
  - 0.586*** [0.041]
  - 0.690*** [0.075]
  - 1.035*** [0.103]
  - 0.968*** [0.093]
  - 0.967*** [0.089]
  - 0.206 [0.098]

- Population, destination
  - -0.003 [0.055]
  - 0.343*** [0.035]
  - 0.422*** [0.091]
  - 0.680*** [0.172]
  - 0.122 [0.127]
  - 1.271*** [0.179]
  - 0.915*** [0.062]

- SARS
  - -0.317 [0.135]
  - -1.109*** [0.212]
  - -1.472*** [0.446]
  - -3.736*** [0.622]
  - -1.544* [0.576]
  - -4.462*** [0.630]
  - -0.615 [0.406]

- Number of observations
  - 72,859 | 79,744 | 23,016 | 15,612 | 13,073 | 14,102 | 13,941

- Fixed effects and fit
  - Origin FE: Yes for all columns
  - Destination FE: Yes for all columns
  - Year FE: Yes for all columns
  - Adjusted R2: 0.88 | 0.81 | 0.84 | 0.85 | 0.86 | 0.85 | 0.83

*Source: Appendix Tables A1–A3 (content unit provided).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020112-print-pdf.pdf_