## 1. Tourism Revenue of Asian and Pacific Countries with Largest Tourism Sectors

## Source details

**Canonical URL:** [1. Tourism Revenue of Asian and Pacific Countries with Largest Tourism Sectors](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021224-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021224-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021224-print-pdf.pdf.json)

---

### I. INTRODUCTION — scope and approach
- Focus:
  - Tourism reliance of Pacific Island Countries (PICs) and vulnerability to local and global shocks, with emphasis on COVID-19 impacts and recovery via potential tourism diversion and travel bubbles.
- Empirical strategy:
  - Use monthly outbound travel data from Australia (1991–2016) to quantify tourism diversion effects on PICs.
  - Exploit variation from external shocks: September 11 attacks, Bali bombings (2002, 2005), SARS epidemic, Indian Ocean tsunami.
- Key model elements:
  - Primary equation: ln TA_it = α + β0 ln income_t + β1 ln price_it + β2 ln exchange_t + β3 ln oil_t + β4 ln TA_t−1 + β5 D_t + μ_t + ε_t
  - ARDL representation used to capture short-run dynamics (β parameters) and long-run relationships (γ parameters).
  - Coefficient of primary interest: β5, captures elasticity of departures to different shocks.
  - Stationarity: Im, Pesaran, and Shin (IPS) tests; all data stationary in first difference.
  - Lag selection: Bayesian Information Criterion (BIC), max lag length = 4.

### II. RECENT DEVELOPMENTS AND CONCEPTUAL FRAMEWORK
- Tourism dependence and trends:
  - Tourism contribution to GDP ranges between 10 and 40 percent in Fiji, Palau, Samoa and Vanuatu, exceeded only by the Maldives (2019 data).
  - Visitor arrivals in PICs grew from 686,000 in 2000 to 1,870,000 in 2019.
  - With COVID-19, arrivals to PICs fell by 84 percent in 2020; (unweighted) average real GDP growth was -4.5 percent in 2020.
- Main source markets:
  - Australia accounted for 37 percent of arrivals to PICs in 2019; New Zealand 21 percent.
  - In 2016, PICs accounted for 6 percent of Australian departures, down from 9 percent in 2003.
- Conceptual distinctions for diversion:
  - Localized shock: reduces willingness to travel to a specific country; diverted tourists may visit competing destinations.
  - Regional/global shock: reduces overall willingness to travel (“tourism destruction”); net effects on less-affected countries depend on magnitude of destruction vs. diversion.

### III. DATA AND METHODOLOGY — identification of shocks and variables
- Data:
  - Monthly Australian outbound tourism by destination country, 1991–2016.
- Controls included:
  - Quarterly Australian real GDP, monthly CPI destination/Australia, exchange rate vs Australian dollar, monthly average Brent spot oil price, lagged departures, monthly seasonal dummies.
- Shock dummies (set = 1 for three months following events unless specified):
  - September 11 attacks: Sep–Dec 2001.
  - Bali bombing 2002: Oct 2002–Jan 2003.
  - Bali bombing 2005: Oct 2005–Jan 2006.
  - Indian Ocean tsunami: Dec 2004–Mar 2005.
  - SARS peak: Apr–Jun 2003.
  - Internal shocks for PIC diversion regressions: Fiji coup 2001, Tropical Cyclone Evan 2012, Tropical Cyclone Pam 2015.
- Model specification:
  - ARDL to control for long-run trends and small sample issues.
  - β5 from Equation (2) reports shock impacts.

### IV. MAIN EMPIRICAL FINDINGS
- Aggregate departures from Australia:
  - Global shocks (September 11 attacks and SARS epidemic) reduced total departures by over 10 percent during affected months.
  - Localized shocks (Bali bombings, Indian Ocean tsunami) had significantly smaller impacts on aggregate departures, consistent with diversion to unaffected destinations.
  - Travel-cost factors (oil price, exchange rate) and local economic conditions significantly affect total departures.
- Impact on affected destinations:
  - All shocks produced significant drops in departures from Australia to the affected destinations.
  - Largest impacts on affected destinations observed for the Bali bombings and SARS epidemic.
  - Example: Departures to the United States fell by about 20 percent after the September 11 attacks.
  - Because the U.S. accounted for ~10 percent of total departures, the ~10 percent fall in total departures implies only a fraction of the aggregate decline was explained mechanically by declines to the U.S.; global risk aversion amplified the fall.
- Diversion effects to PICs and New Zealand:
  - Following the September 11 attacks: number of departures to PICs remained stable while New Zealand fell by around 8 percent.
  - Following the 2002 Bali bombings: departures to PICs increased by 7 percent (absolute increase and share increase), while New Zealand had no gains.
  - During the SARS epidemic: both PICs and New Zealand experienced strong diversion gains; New Zealand gained a larger share compared to PICs.
  - Of all regions, only PICs and New Zealand gained in shares during the SARS epidemic.
- Transience and timing:
  - Diversion gains are temporary; SARS-related gains for PICs peaked shortly after the epidemic wave and then declined as the epidemic waned.
- Heterogeneity across PICs:
  - Most diversion gains among PICs driven by Fiji.
  - Other PICs (Vanuatu, Samoa, Tonga) show expected sign of impacts but not statistically significant in many specifications.
  - Accessibility and developed tourism infrastructure amplify ability to capture diversion flows.

### V. ANNEX I — DETAILED ESTIMATION RESULTS (selected coefficients and statistics)
- Model and estimation notes:
  - Results from Equation (2).
  - ***/**/* indicate statistical significance at the 1%, 5% and 10% level, respectively.
  - Standard errors in parentheses.
  - Exchange rate is the exchange rate of USD against AUD.
  - All shock events are coded as dummies.
- Reported coefficient estimates (Δ ln(departures)t) — three model columns reported:
  - Sep 11 attacks: -0.0438; -0.0643***; -0.104*** (standard errors: (0.0193); (0.0132); (0.0181))
  - 2002 Bali bombings: -0.00946; -0.0493***; -0.0559*** (standard errors: (0.0150); (0.0149); (0.0158))
  - SARS epidemic: -0.0484; -0.142***; -0.171*** (standard errors: (0.0318); (0.0207); (0.0231))
  - Indian Ocean tsunami: 0.00883; -0.0108; 0.0131 (standard errors: (0.0183); (0.0138); (0.0126))
  - 2005 Bali bombings: -0.00655; -0.0304; -0.0306 (standard errors: (0.0296); (0.0260); (0.0206))
  - ln(Aus GDP)t-1: 0.699***; 0.461*** (standard errors: (0.0854); (0.0836))
  - ln(oil prices)t-1: -0.0265**; -0.0149 (standard errors: (0.0119); (0.0120))
  - ln(departures)t–1: -0.352***; -0.235*** (standard errors: (0.0401); (0.0417))
  - ln(exchange rate)t–1: 0.158***; 0.106*** (standard errors: (0.0289); (0.0259))
  - Δ ln(exchange rate)t-1: 0.0527 (standard error: (0.0349))
  - Δ ln(Aus GDP)t-1: -1.735 (standard error: (1.059))
  - Δ ln(oil prices)t-1: -0.0347 (standard error: (0.0361))
  - Δ ln(departures)t-2: -0.567*** (standard error: (0.0601))
  - Δ ln(departures)t-3: -0.347*** (standard error: (0.0556))
  - SARS epidemic t-2: -0.0615** (standard error: (0.0251))
  - SARS epidemic t-3: 0.112*** (standard error: (0.0197))
- Fixed effects and sample statistics:
  - Monthly FE: Yes
  - Observations: 311; 311; 309
  - R-squared: 0.922; 0.924; 0.954
- Notes on destination regressions:
  - All regressions include monthly fixed effects.
  - Controls include exchange rate, relative CPI, Australian GDP, and oil prices.
  - For PICs, exchange rate control uses the USD–AUD ER given many PICs have fixed pegs with significant weight to USD.
  - Outcome variables: change in log of departures to specific destinations and share of departures to these destinations.

### VI. POLICY IMPLICATIONS — recovery from COVID-19 and making diversion gains durable
- Key constraints and differences of COVID-19 shock:
  - COVID-19 is more severe, far-reaching and long-lasting than prior shocks considered.
  - Supply-side constraints matter (airlines, cruise operators, workforce scarring, bankruptcies).
  - Departure restrictions in Australia and New Zealand limit immediate scope for PIC policies to capture diversion.
- Practical recommendations:
  - Monitor diversion dynamics and competitor recoveries to calibrate cyclical policy support.
  - Prioritize timely and comprehensive vaccine rollout and public-health measures to increase traveler confidence.
  - Facilitate travel bubbles with large source markets (Australia, New Zealand) conditional on effective containment and vaccination.
  - Targeted promotions and loyalty programs to attract returning visitors and divert tourists from competitors.
  - Structural reforms to convert temporary gains into permanent gains:
    - Invest in eco-sustainable, lower-density, higher value-added tourism subsectors.
    - Leverage digitilization in tourism services (automation, contact-less payments, real-time information).
    - Support reallocation from cruise to non-cruise tourism.
    - Use public-private partnerships to finance substantial public investments.
  - Regional cooperation among PICs to address supply constraints, achieve economies of scale, improve regional infrastructure and coordinate promotions.
  - Development partners: provide technical assistance, accelerate vaccinations, enable safe travel corridors, and support policy actions for speedy recovery.
  - Risk management improvements: update security and health protocols, adopt insurance instruments, build contingent reserve funds, and diversify away from tourism dependence.

### VII. CONCLUSION AND AVENUES FOR FUTURE RESEARCH
- Empirical summary:
  - PICs have often been insulated from large global shocks and in some cases benefited via strong but temporary tourism diversion gains.
  - Magnitude of diversion depends on shock type, region affected, destination safety, infrastructure, proximity and familiarity.
  - Diversion effects are typically temporary, highlighting need for timely policies.
- Future research directions:
  - Empirically identify determinants of tourism flows and diversion effects as countries reopen and establish travel bubbles.
  - Study pace of recovery and compare COVID-19 with previous shocks.
  - Evaluate effectiveness and optimal mix of policies used to promote tourism recovery.
  - Develop structural demand-supply models calibrated to empirical estimates to predict recovery paths and policy impacts.

*Source: wpiea2021224-print-pdf*

### 1. Tourism Revenue of Asian and Pacific Countries with Largest Tourism Sectors........... 5

### 1. Tourism Revenue of Asian and Pacific Countries with Largest Tourism Sectors

### Figures and tables referenced in this unit
- 2. International Visitor Arrivals in Pacific Island Countries Number of People ................. 5
- 3. Visitor to PICs by Source Country .............................................................................. 6
- 4. Australian Visitors to PICs ......................................................................................... 6
- 5. Schematic: Two Types of Tourism Diversion .............................................................. 7
- 6. Share of Departures from Australia ............................................................................. 9
- 7. Impact of Shocks on Number of Departures from Australia ....................................... 10
- 8. Impact of Shocks on Departures to Affected Countries .............................................. 11
- 9. Impact of External Shocks on Departures to PICs ...................................................... 12
- 10. Change in Shares of Departures during SARS Epidemic .......................................... 12
- 11. Change in Share of Departures to PICs during SARS............................................... 13
- 12. Number of Confirmed COVID-19 Cases ................................................................. 14
- 13. COVID-19 Vaccinations......................................................................................... 14
- 14. Illustrative Path of Tourism Recovery in PICs ......................................................... 15

### Annex
- I. Detailed Estimation Results....................................................................................... 20

*Source: wpiea2021224-print-pdf - 1. Tourism Revenue of Asian and Pacific Countries with Largest Tourism Sectors........... 5*

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

### wpiea2021224-print-pdf - References

### I. INTRODUCTION — scope and approach
- Focus: Tourism reliance of Pacific Island Countries (PICs) and vulnerability to local and global shocks, with emphasis on COVID-19 impacts and recovery via potential tourism diversion and travel bubbles.
- Empirical strategy: Use monthly outbound travel data from Australia (1991–2016) to quantify tourism diversion effects on PICs, exploiting variation from external shocks (September 11 attacks, Bali bombings, SARS epidemic, Indian Ocean tsunami).
- Key model elements:
  - Equation (1): ln TA_it = α + β0 ln income_t + β1 ln price_it + β2 ln exchange_t + β3 ln oil_t + β4 ln TA_t−1 + β5 D_t + μ_t + ε_t
  - ARDL representation (Equation (2)) used to capture short-run dynamics (β parameters) and long-run relationships (γ parameters).
  - Coefficient of primary interest: β5, captures elasticity of departures to different shocks.
  - Stationarity: Im, Pesaran, and Shin (IPS) tests; all data stationary in first difference.
  - Lag selection: Bayesian Information Criterion (BIC), max lag length = 4.

### II. Recent developments and conceptual framework
- Tourism dependence and trends:
  - Tourism contribution to GDP ranges between 10 and 40 percent in Fiji, Palau, Samoa and Vanuatu, exceeded only by the Maldives (2019 data).
  - Visitor arrivals in PICs grew from 686,000 in 2000 to 1,870,000 in 2019.
  - With COVID-19, arrivals to PICs fell by 84 percent in 2020; (unweighted) average real GDP growth was -4.5 percent in 2020.
- Main source markets:
  - Australia accounted for 37 percent of arrivals to PICs in 2019; New Zealand 21 percent.
  - In 2016, PICs accounted for 6 percent of Australian departures, down from 9 percent in 2003.
- Conceptual distinctions for diversion:
  - Localized shock: reduces willingness to travel to a specific country; diverted tourists may visit competing destinations.
  - Regional/global shock: reduces overall willingness to travel (“tourism destruction”); net effects on less-affected countries depend on magnitude of destruction vs. diversion.

### III. Data and methodology — identification of shocks and variables
- Data: Monthly Australian outbound tourism by destination country, 1991–2016.
- Controls included: quarterly Australian real GDP, monthly CPI destination/Australia, exchange rate vs Australian dollar, monthly average Brent spot oil price, lagged departures, monthly seasonal dummies.
- Shock dummies: set to 1 for three months following events (September 11 attacks: Sep–Dec 2001; Bali bombing 2002: Oct 2002–Jan 2003; Bali bombing 2005: Oct 2005–Jan 2006; Indian Ocean tsunami: Dec 2004–Mar 2005; SARS peak: Apr–Jun 2003). Internal shocks (for PIC diversion regressions) include Fiji coup 2001, Tropical Cyclone Evan 2012, Tropical Cyclone Pam 2015.
- Model specification: ARDL to control for long-run trends and small sample issues; β5 from Equation (2) reports shock impacts.

### IV. Main empirical findings
- Aggregate departures from Australia:
  - Global shocks (September 11 attacks and SARS epidemic) reduced total departures by over 10 percent during affected months.
  - Localized shocks (Bali bombings, Indian Ocean tsunami) had significantly smaller impacts on aggregate departures, consistent with diversion to unaffected destinations.
  - Travel-cost factors (oil price, exchange rate) and local economic conditions significantly affect total departures.
- Impact on affected destinations:
  - All shocks produced significant drops in departures from Australia to the affected destinations.
  - Largest impacts on affected destinations observed for the Bali bombings and SARS epidemic (personal-safety-related and regionally concentrated).
  - Example: Departures to the United States fell by about 20 percent after the September 11 attacks.
  - Because the U.S. accounted for ~10 percent of total departures, the ~10 percent fall in total departures implies only a fraction of the aggregate decline was explained mechanically by declines to the U.S.; global risk aversion amplified the fall.
- Diversion effects to PICs and New Zealand:
  - Following the September 11 attacks: number of departures to PICs remained stable while New Zealand fell by around 8 percent.
  - Following the 2002 Bali bombings: departures to PICs increased by 7 percent (absolute increase and share increase), while New Zealand had no gains.
  - During the SARS epidemic: both PICs and New Zealand experienced strong diversion gains; New Zealand gained a larger share compared to PICs.
  - Of all regions, only PICs and New Zealand gained in shares during the SARS epidemic.
- Transience and timing:
  - Diversion gains are temporary; SARS-related gains for PICs peaked shortly after the epidemic wave and then declined as the epidemic waned.
  - Figure evidence: change in share of departures to PICs during SARS shows a short-lived increase (T = 0 at month of shock).
- Heterogeneity across PICs:
  - Most diversion gains among PICs driven by Fiji (reflecting larger and more advanced tourism market).
  - Other PICs (Vanuatu, Samoa, Tonga) show expected sign of impacts but not statistically significant in many specifications.
  - Accessibility and developed tourism infrastructure amplify ability to capture diversion flows.

### V. Policy implications — recovery from COVID-19 and making diversion gains durable
- Key constraints and differences of COVID-19 shock:
  - COVID-19 is more severe, far-reaching and long-lasting than prior shocks considered.
  - Supply-side constraints matter (airlines, cruise operators, workforce scarring, bankruptcies).
  - Departure restrictions in Australia and New Zealand limit immediate scope for PIC policies to capture diversion.
- Practical policy recommendations:
  - Monitor diversion dynamics and competitor recoveries to calibrate cyclical policy support and avoid mistaking temporary diversion gains for structural recovery.
  - Prioritize timely and comprehensive vaccine rollout and public-health measures to increase traveler confidence (vaccine coverage, quarantine protocols, health infrastructure strengthening).
  - Facilitate travel bubbles with large source markets (Australia, New Zealand) conditional on effective containment and vaccination; unequal outbound access could amplify diversion effects if maintained.
  - Targeted promotions and loyalty programs to attract returning visitors and divert tourists from competitors.
  - Implement structural reforms to convert temporary gains into permanent gains:
    - Invest in eco-sustainable, lower-density, higher value-added tourism subsectors to reduce health risks and foster a greener recovery.
    - Leverage digitilization in tourism services (automation, contact-less payments, real-time information) to diversify tourism base.
    - Support reallocation from cruise to non-cruise tourism given slower cruise recovery.
    - Use public-private partnerships to finance substantial public investments given resource constraints in PICs.
  - Regional cooperation among PICs to address supply constraints, achieve economies of scale, improve regional infrastructure and coordinate promotions.
  - Development partners: provide technical assistance, accelerate vaccinations, enable safe travel corridors, and support policy actions for speedy recovery.
  - Risk management improvements: update security and health protocols, adopt insurance instruments, build contingent reserve funds, and diversify away from tourism dependence.

### VI. Conclusion and avenues for future research
- Empirical summary:
  - PICs have often been insulated from large global shocks and in some cases benefited via strong but temporary tourism diversion gains.
  - Magnitude of diversion depends on shock type, region affected, destination safety, infrastructure, proximity and familiarity.
  - Diversion effects are typically temporary, highlighting need for timely policies.
- Future research directions:
  - Empirically identify determinants of tourism flows and diversion effects as countries reopen and establish travel bubbles.
  - Study pace of recovery and compare COVID-19 with previous shocks.
  - Evaluate effectiveness and optimal mix of policies used to promote tourism recovery.
  - Develop structural demand-supply models calibrated to empirical estimates to predict recovery paths and policy impacts.

*Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021224-print-pdf.pdf*

### References

### References (wpiea2021224-print-pdf)

### Literature themes and scope
- Empirical studies on tourism demand shocks from pandemics, terrorism, natural disasters, and policy changes:
  - COVID-19 and pandemics: Gössling, Scott, and Hall, 2020; International Monetary Fund, 2021; Nguyen, 2020.
  - Terrorism and security events: Bonham, Edmonds, and Mak, 2006; Edmonds and Mak, 2005; Misrahi, 2016; Lean and Smyth, 2009.
  - Epidemics and health shocks (SARS, H1N1): Wang, 2009; Russy and Smith, 2013; Lean and Smyth, 2009.
  - Regional and currency effects on tourism flows: Gallego, Ledesma-Rodríguez, and Pérez-Rodríguez, 2015; Seetaram, 2012.
  - Small island and Pacific Island Country (PIC) tourism dynamics and risks: Kumar et al., 2020; Acevedo, Trevor, and Coleridge, 2017; Balasundharam et al., 2021 (forthcoming); Arslanalp, Koepke and Verschuur, 2021 (forthcoming).
  - Policy instruments and fiscal risks: Forsyth et al., 2014; Balasundharam et al., 2021 (forthcoming).
- Data and reporting sources:
  - UNWTO, 2021. Travel restrictions report, 9th edition.

### Annex I — Table 1: Determinants of Australia’s departures in the short run (1991–2016)
- Model and estimation notes:
  - Results from Equation 2.
  - ***/**/* indicate statistical significance at the 1%, 5% and 10% level, respectively.
  - Standard errors in parentheses ().
  - Exchange rate is the exchange rate of USD against AUD.
  - All shock events are coded as dummies.
- Reported coefficient estimates (Δ ln(departures)t):
  - Sep 11 attacks: -0.0438; -0.0643***; -0.104*** (standard errors: (0.0193); (0.0132); (0.0181))
  - 2002 Bali bombings: -0.00946; -0.0493***; -0.0559*** (standard errors: (0.0150); (0.0149); (0.0158))
  - SARS epidemic: -0.0484; -0.142***; -0.171*** (standard errors: (0.0318); (0.0207); (0.0231))
  - Indian Ocean tsunami: 0.00883; -0.0108; 0.0131 (standard errors: (0.0183); (0.0138); (0.0126))
  - 2005 Bali bombings: -0.00655; -0.0304; -0.0306 (standard errors: (0.0296); (0.0260); (0.0206))
  - ln(Aus GDP)t-1: 0.699***; 0.461*** (standard errors: (0.0854); (0.0836))
  - ln(oil prices)t-1: -0.0265**; -0.0149 (standard errors: (0.0119); (0.0120))
  - ln(departures)t–1: -0.352***; -0.235*** (standard errors: (0.0401); (0.0417))
  - ln(exchange rate)t–1: 0.158***; 0.106*** (standard errors: (0.0289); (0.0259))
  - Δ ln(exchange rate)t-1: 0.0527 (standard error: (0.0349))
  - Δ ln(Aus GDP)t-1: -1.735 (standard error: (1.059))
  - Δ ln(oil prices)t-1: -0.0347 (standard error: (0.0361))
  - Δ ln(departures)t-2: -0.567*** (standard error: (0.0601))
  - Δ ln(departures)t-3: -0.347*** (standard error: (0.0556))
  - SARS epidemic t-2: -0.0615** (standard error: (0.0251))
  - SARS epidemic t-3: 0.112*** (standard error: (0.0197))
- Fixed effects and sample statistics:
  - Monthly FE: Yes
  - Observations: 311; 311; 309
  - R-squared: 0.922; 0.924; 0.954

### Annex I — Tables 2–5: Notes on impact regressions across destinations
- Shared model features and controls:
  - Results from Equation 2.
  - All regressions include monthly fixed effects.
  - Controls include exchange rate, relative CPI, Australian GDP, and oil prices (Table 2).
  - For PICs, exchange rate control uses the USD–AUD ER given many PICs have fixed pegs with significant weight to USD (Table 3).
  - Regressions for destination-specific outcomes include controls for Australian GDP and oil prices (Tables 4 and 5).
  - Outcome variables: change in log of departures to specific destinations and share of departures to these destinations.
  - ***/**/* indicate statistical significance at the 1%, 5% and 10% level, respectively.
  - Standard errors reported in parentheses.

*Content extracted from the References and Annex I (Detailed Estimation Results) sections of wpiea2021224-print-pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021224-print-pdf.pdf_