Back Matter
  • 1 https://isni.org/isni/0000000404811396, International Monetary Fund

VII. References

  • Acevedo, Sebastian, 2014, “Debt, Growth and Natural Disasters: A Caribbean Trilogy”, IMF Working Paper 14/125.

  • Andrews, Donald WK, and Biao Lu. 2001, “Consistent model and moment selection procedures for GMM estimation with application to dynamic panel data models.Journal of Econometrics 101, no. 1 pp. 123164.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Athiyaman, A., 1997, “Knowledge development in tourism: tourism demand research”, Tourism Management, Vol. 18, pp. 221228.

  • Crouch, G., 1994, “The Study of International Tourism Demand: A Survey of Practice,Journal of Travel Research, Vol. 33, pp. 4155.

  • Culiuc, Alexander, 2014, “Determinants of International Tourism,IMF Working Paper No. 14/82.

  • Harvey, Andrew C., and P. H. J. Todd. 1983, “Forecasting economic time series with structural and Box-Jenkins models: A case study.Journal of Business & Economic Statistics 1, no. 4 pp. 299307.

    • Search Google Scholar
    • Export Citation
  • Gustavsson, Patrik, and Jonas Nordström. “The impact of seasonal unit roots and vector ARMA modelling on forecasting monthly tourism flows.Tourism Economics 7, no. 2 (2001): 117133.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kim, Jae H., and Imad A. Moosa. 2005, “Forecasting international tourist flows to Australia: A comparison between the direct and indirect methods.Tourism Management 26, no. 1 pp. 6978.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laframboise, Nicole, Nkunde Mwase, Joonkyu Park, and Yingke Zhou. 2014, “Revisiting Tourism Flows to the Caribbean: What is Driving Arrivals?”. IMF Working Paper 14/229.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Li, Gong, H. Song, and SF Witt., 2005Recent developments in econometric modelling and forecastingJournal of Travel Research 44 (1), pp.8299.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mwase, Nkunde, 2013, “Tourism flows to Caribbean islands: an empirical note,Applied Economics Letters, Volume 20, Issue 10, pp. 957965.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Love, Innessa, 2015, “PVAR: Package of programs to estimate panel VAR”. Available: https://sites.google.com/a/hawaii.edu/inessalove/home/pvar

    • Search Google Scholar
    • Export Citation
  • Song, Haiyan, and Stephen F. Witt., 2006Forecasting international tourist flows to Macau.Tourism management 27, no. 2 pp. 214224.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tsounta, Evridiki, 2008, “What Attracts Tourists to Paradise?IMF Working Paper No. 08/162.

  • Vu, Jo Chau, and Lindsay W. Turner. 2006, “Regional data forecasting accuracy: the case of Thailand.Journal of Travel Research 45, no. 2 pp. 186193.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wolfe, Andy and Rafael Romeu, 2011, “Recession and Policy Transmission to Latin American Tourism: Does Expanded Travel to Cuba Offset Crisis Spillovers?IMF Working Paper No. 11/32.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Witt, Stephen F., and Christine A. Witt. 1992, “Modeling and forecasting demand in tourism”, International Journal of Forecasting, Vol. 8, pp. 643644.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, Xinyan, Haiyan Song, and George Q. Huang. 2009, “Tourism supply chain management: A new research agenda.Tourism management 30, no. 3 pp.345358.

    • Crossref
    • Search Google Scholar
    • Export Citation

VIII. Appendix

Table A1.

Coefficient of Variation (1990-2014) and Significance of Tourist Arrivals Response to Different Airlift Supply Factors

article image
Source: Authors’ calculations based on country specific SVARs.Notes: The numbers in each cell represent the coeficient of variation (standard deviation over mean) statistics for airlift supply variables. The green color indicates the significance of the tourist arrivals response after each airlift supply shock. The dark green color indicates the response is different from zero for more than 4 periods; the lighter green for more than 2 periods.
Figure A1.
Figure A1.

Stochastic Seasonality

Citation: IMF Working Papers 2016, 033; 10.5089/9781498375108.001.A999

Note: The upper figure presents the percentage change of tourist arrivals for Barbados from 1990 to 1993. The bottom figure gives the autocorrelation function of the series. The high correlation with its 12th, 24th and 36th lags indicates a strong seasonality in the series.
Figure A2.
Figure A2.

Unobserved Component Model

Citation: IMF Working Papers 2016, 033; 10.5089/9781498375108.001.A999

Note: Upper left graph gives the seasonal pattern detected by the unobserved component model; Bottom left represents the filtered series (eliminated the seasonal components); Bottom right gives the autocorrelation function of the filtered series. We can see the autocorrelation function looks like a AR process after filtering out the seasonality.
Figure A3.
Figure A3.

Response of Tourist Arrivals to Different Shocks

(specification with vacancy rate, panel VAR)

Citation: IMF Working Papers 2016, 033; 10.5089/9781498375108.001.A999

Notes: The blue dashed represents the percentage deviation from the steady state of the response variable (tourist arrivals) to a one percent positive shock of the impulse variable. The shaded area is the 90% confidence interval and the red dashed line shows the cumulative percentage change of tourist arrivals.
Figure A4.
Figure A4.

Response of Tourist Arrivals to Different Shocks

(specification including jet fuel, panel VAR)

Citation: IMF Working Papers 2016, 033; 10.5089/9781498375108.001.A999

Notes: The blue dashed represents the percentage deviation from the steady state of the response variable (tourist arrivals) to a one percent positive shock of the impulse variable. The shaded area is the 90% confidence interval and the red dashed line shows the cumulative percentage change of tourist arrivals.
Figure A5.
Figure A5.

Impulse Response Functions (benchmark specification, panel SVAR)

Citation: IMF Working Papers 2016, 033; 10.5089/9781498375108.001.A999

Notes: The title of the sub graphs indicates “the impulse variable, the response variable”. The horizontal axis represents percentage deviation from the steady state in response to a 1% positive shock. The shaded area is the 90% confidence interval and the red dashed line shows the cumulative percentage change of tourist arrivals.
Figure A6.
Figure A6.Figure A6.Figure A6.Figure A6.Figure A6.Figure A6.Figure A6.

Response of Tourist Arrivals to Different Shocks (country SVAR Results)

Citation: IMF Working Papers 2016, 033; 10.5089/9781498375108.001.A999

Notes: The blue dashed line represents the percentage deviation from the steady state of the response variable (tourist arrivals) to a one percent positive shock of the impulse variable. The shaded area is the 90% confidence interval and the red dashed line shows the cumulative percentage change of tourist arrivals.
Figure A7.
Figure A7.

Response of Tourist Arrivals to a Natural Disaster Shock in Another Country

(specification with natural disasters affecting other countries, country SVARs)

Citation: IMF Working Papers 2016, 033; 10.5089/9781498375108.001.A999

Source: Authors’ calculations based on country specific SVARs.Notes: The shaded area is the 90% confidence interval and the blue dashed line represents the percentage deviation from the steady state of the response variable (tourist arrivals) after a natural disaster in another Caribbean country.
Figure A8.
Figure A8.

Response of Flights to an Increase in Flights to the Dominican Republic Shock

Citation: IMF Working Papers 2016, 033; 10.5089/9781498375108.001.A999

Source: Authors’ calculations based on the panel SFiVAR.Notes: Impulse: the number of flights to the Dominican Republic. The shaded area is the 90% confidence interval.
1

All the numbers refer to our sample of 14 Caribbean destinations listed in Table 1.

2

Both Antigua and Barbuda and Barbados have historically served as connecting hubs for the Easter Caribbean sub-region. As airlift availability to the rest of the Caribbean has improved over time, with more direct connections and more frequent flights, their hub services have dwindled and their flight traffic has declined accordingly.

3

We define the vacancy rate as empty seats as a percent of total available seats.

4

Since the dataset used only includes direct flights between the U.S. and foreign airports, the data on the number of passengers also captures people in transit to other Caribbean destinations. That is, if a U.S. tourist is traveling to Dominica via a connecting flight in Antigua and Barbuda, it will be recorded as a passenger to Antigua and a U.S. tourist arrival to Dominica, but not as a tourist in Antigua or a passenger to Dominica.

5

Major carriers are defined as those with annual revenue over US$1 billion; national carriers have annual revenue between $100 million and $1 billion; regional carriers have annual revenue below $20 million; chartered certified carriers have a maximum seating capacity of 60 or less seats or a maximum payload of 18,000 pounds or less.

6

The Herfindahl index is used to measure the size of firms in relation to the industry and to indicate concentration in an industry. In general, a measure below 0.01 indicates a “highly competitive market”; below 0.15 an “unconcentrated” market; and above 0.25 “high concentration”.

7

Countries with a Herfindahl index above 0.25 indicating high concentration include: Antigua and Barbuda, Barbados, Belize, Dominica, Grenada, St. Kitts and Nevis, and St. Lucia.

8

A superior gauge would be tourists per number of beaches, but the data is not available.

9

A similar ranking in per capita terms also shows that small countries have good airlift supply from the U.S.

10

Tourist arrivals are the most frequently used measure of tourism demand, followed by tourism expenditure (Li et al, 2005).

11

The data comes from forms T-100 International Segment (All Carriers).

12

It is also possible that larger planes reduce transportation costs and result in lower airfares. Unfortunately, the data from the U.S. DoT does not include information on airfares, an important factor in consumers’ tourism decisions.

13

Anguilla and St. Vincent and the Grenadines were excluded due to data gaps in some of the variables.

14

DF-GLS unit root tests were performed for each variable and country in the sample. No time trends were found for the first differenced variables.

16

The autocorrelation function for each destination and for each variable of interest was plotted. Strong seasonality is found in most cases (see Figure A1 in the Appendix for an example of the seasonality found).

17

No significant seasonal effect is found for the number of airlines or cities. As a robustness check, we apply the same filter as other variables with stochastic seasonal components and find no significant change of our results.

18

We also tested combinations of alternative demand controls, as suggested by Laframboise et al (2014), including real GDP growth in the U.S., relative prices between the U.S. and the destination country, and exogenous supply controls such as the price of jet fuel and the price of oil. These variables are found to be insignificant thus were excluded from the model.

19

The main results do not depend on the number of lags being specified. Models with 4 and 8 lags show consistent results with those presented in the paper.

20

Their proposed MMSC are similar to maximum likelihood-based model selection criteria for time series models, such as the Akaike information criteria (AIC) and the Bayesian information criteria (BIC).

21

These assumptions translate into the following; the number of airlines affect contemporaneously all other supply factors, while the airlines themselves are only affected with a lag. The numbers of U.S. departure cities with direct flights affects contemporaneously the number of seats and flights, but cities are only affected with a lag by those variables. The number of flights has a contemporaneous effect on the number of seats, and lastly the number of seats is affected contemporaneously by all airlift supply factors, but only has an effect on the other factors with a lag.

22

December through Easter is usually the high season for most Caribbean countries.

23

Additionally, in some countries, airlines receive different types of financial incentives to maintain the frequency of flights (minimum seat guarantees, joint marketing with the destination government in source markets, etc) which reduce the incentives for airlines to move their schedules too frequently.

24

The results are robust to alternative aggregate demand control variables including U.S. household real income, and U.S. real GDP. As a robustness check, the seat vacancy rate was substituted for the total number of seats as an alternative control and the results are consistent with those presented here.

25

Based on Panel OLS using annual data, Laframboise et al (2014) found a 1 percent rise in weighted unemployment rates is associated with about 2 percent decline in tourist arrivals.

26

Airlines explain on average 2.5 percent of the variation in tourist arrivals, departure cities 4.4 percent, flights 11.7 percent, and seats 3.5 percent.

27

Dominica is also one of the countries where departure cities appear to have a bigger effect on tourism (Figure 7), but as is shown in Table 4, the impact is negative and insignificant, and the country would benefit mostly from increasing the frequency of flights from the U.S.

28

This is still the case even after controlling for the size of the country (land area).

29

The distance between the destination and the U.S. was also checked, but did not offer any additional insight.

30

For country by country impulse response functions, please see Figure A6 in the Appendix.

31

The heterogeneity in the impulse response functions in reaction to the unemployment rate shock also reveals different demand elasticities of U.S. tourist arrivals across Caribbean countries.

32

Storms (i.e. tropical cyclones) are the most prevalent type of disaster in the Caribbean.

33

This might be due to two reasons: 1) airlines hedge against oil prices shocks and therefore their supply does fluctuate with oil prices, and 2) flights are scheduled and booked several months in advance, therefore by the time of fuel price changes, airlines cannot adjust the number of flights.

34

Some of the robustness results are presented in Figures A3 and A4 in the Appendix; the rest are available upon request.

35

Figure 10 shows that there is already a large number of flights between the U.S. and Cuba. In 2014, they surpassed 4,000 flights per year (more than what each of the smaller islands receives, with the exception of Aruba). In 2014, Cuba received more than 370,000 passengers flying from the U.S.

36

Different specifications were estimated with the number of airlines and number of departure cities to the destination not being affected by the number of flights to Cuba, and the results are robust. The number of flights to all other Caribbean countries is also included as a control variable.

37

Airlines order aircrafts well in advance and cannot increase their fleet suddenly by purchasing from manufacturers. However, if the change is gradual, this might not be a problem.

38

One caveat: the results hold under the assumption that destinations have well developed tourism sectors that attract tourists. Increasing the number of flights by itself will not help a destination without a viable product.

Flying to Paradise: The Role of Airlift in the Caribbean Tourism Industry
Author: Mr. Sebastian Acevedo Mejia, Lu Han, Miss Marie S Kim, and Ms. Nicole Laframboise