How important is tourism for the international transmission of cyclical fluctuations? Evidence from the Mediterranean.

Transcription

1 How important is tourism for the international transmission of cyclical fluctuations? Evidence from the Mediterranean. On-line Appendix Fabio Canova Pietro Dallari January 26, 2 This on-line appendix reports additional materials discussed in the paper. Section 1 is dedicated to evidence produced using data on tourist arrivals. Section 2 reports results for the number of nights spent and per-capita expenditures. Contact: Department of Economics, European University Institute, Via della Piazzuola, 43, 53 Florence, Italy. fabio.canova@eui.eu Contact: GPEFM-Departament d Economia i Empresa, Universitat Pompeu Fabra, Carrer de Ramon Trias Fargas, 25-27, 5 Barcelona, Spain. pietro.dallari@upf.edu

2 1 Number of tourist arrivals 1.1 Tourism data in (log) levels Tourist arrivals: Cyprus Figure 1: Number of tourist arrivals: Cyprus AUT tourists BEL tourists FIN tourists FRA tourists

3 Figure 2: Number of tourist arrivals: Cyprus GRC tourists ITA tourists IRL tourists NLD tourists

4 Figure 3: Number of tourist arrivals: Cyprus SWE tourists UK tourists DEU tourists RUS tourists Figure 4: Number of tourist arrivals: Cyprus.5 EU tourists ALL tourists

5 1.1.2 Tourist arrivals: Morocco Figure 5: Number of tourist arrivals: Morocco UK tourists FRA tourists ESP tourists DEU tourists

6 Figure 6: Number of tourist arrivals: Morocco ITA tourists BEL tourists EU tourists ALL tourists

7 1.1.3 Tourist arrivals: Syria Figure 7: Number of tourist arrivals: Syria AUT tourists BEL tourists GRC tourists NLD tourists

8 Figure : Number of tourist arrivals: Syria NLD tourists SWE tourists DEU tourists ITA tourists

9 Figure : Number of tourist arrivals: Syria UK tourists FRA tourists DNK tourists ESP tourists Figure : Number of tourist arrivals: Syria EU tourists RUS tourists

10 1.1.4 Tourist arrivals: Tunisia Figure : Number of tourist arrivals: Tunisia LUX tourists AUT tourists NLD tourists ESP tourists

11 Figure : Number of tourist arrivals: Tunisia SCAND tourists UK tourists BEL tourists FRA tourists

12 Figure : Number of tourist arrivals: Tunisia DEU tourists ITA tourists EU tourists ALL tourists

13 1.1.5 Tourist arrivals: Turkey Figure 14: Number of tourist arrivals: Turkey FIN tourists NLD tourists ESP tourists EU tourists

14 Figure 15: Number of tourist arrivals: Turkey AUT tourists BEL tourists DNK tourists FRA tourists

15 Figure 16: Number of tourist arrivals: Turkey GRC tourists.5 ITA tourists SWE tourists RUS tourists

16 Figure 17: Number of tourist arrivals: Turkey ALL tourists UK tourists DEU tourists

17 1.2 Granger causality test Table 1: Tourist arrivals: granger causality test % c.v. F-test EA output EA tourist arrivals in CY EA tourist arrivals in CY CY output UK output UK tourist arrivals in CY UK tourist arrivals in CY CY output EA output EA tourist arrivals in MA EA tourist arrivals in MA MA output FR output FR tourist arrivals in MA FR tourist arrivals in MA MA output EA output EA tourist arrivals in SY EA tourist arrivals in SY SY output RU output RU tourist arrivals in SY RU tourist arrivals in SY SY output EA output EA tourist arrivals in TN EA tourist arrivals in TN TN output FR output FR tourist arrivals in TN FR tourist arrivals in TN TN output EA output EA tourist arrivals in TR EA tourist arrivals in TR TR output RU output RU tourist arrivals in TR RU tourist arrivals in TR TR output Notes: If F-test>critical value, reject the null hypothesis of no Granger causation. Arrows indicate the direction of causality. The sample length varies across cases: see the paper for details. Country codes: EA is Euro area; UK is United Kingdom; FR is France; RU is Russia; CY is Cyprus; MA is Morocco; SY is Syria; TN is Tunisia; TR is Turkey. 16

18 1.3 Posterior density of the hypervariance parameter Figure 1: Posterior density of the hypervariance parameter Counterfactual output for case studies Figure 1: Tourist arrivals: counterfactual destination country output, case studies.5.4 CY.4.3 TN..6 TR Notes: Left: Cyprus. Middle: Tunisia. Right: Turkey. The source country is the United Kingdom for Cyprus and Russia for Turkey. Continuous line: median posterior IRF. Dotted lines: 6% confidence bands computed from the posterior distribution of IRFs. Starred line: counterfactual dynamic response of the destination country output without the tourism channel. Country codes: CY is Cyprus; TN is Tunisia; TR is Turkey. 17

19 2 Extensions: number of nights spent and per-capita expenditures 2.1 Tourism data in (log) levels Nights spent: Tunisia Figure 2: Number of nights spent: Tunisia. DEU nights AUT nights BEL nights SCAND nights

20 Figure 21: Number of nights spent: Tunisia. NLD nights FRA nights ITA nights UK nights Figure 22: Number of nights spent: Tunisia. EU nights ALL nights

21 2.1.2 Per-capita expenditures: Cyprus Figure 23: Per-capita expenditures: Cyprus. UK ppe DEU ppe GRC ppe NLD ppe

22 Figure 24: Per-capita expenditures: Cyprus. AUT ppe FRA ppe BEL ppe ITA ppe

23 Figure 25: Per-capita expenditures: Cyprus. IRE ppe EU ppetot EU ppeavg RU ppe ALL ppe

24 2.2 Cyclical fluctuations Figure 26: Cyclical fluctuations Notes: Top panel: number of nights spent in Tunisia. Bottom panel: per-capita expenditures in Cyprus. Dashdotted line: annual changes of (log) tourist arrivals. Continuous and dashed lines: annual changes of the source country and destination country (log) output respectively. Shaded regions: recessions. Country codes: CY is Cyprus; TN is Tunisia; EA is Euro area; FR is France; UK is United Kingdom. 23

25 2.3 Unconditional cross-correlations Table 2: Nights spent: unconditional cross-correlations Output in SC & Nights in MED Lags or leads (in years) EA - TN FR - TN Nights in MED & Output in MED EA - TN FR - TN Output in SC & Output in MED EA - TN FR - TN Notes: The numbers in the table represent corr(x t,y t+i), wherei = [ 2, 1,, 1, 2],x t is the country listed first andy t is the country listed second. The sample length varies across cases: see the paper for details. In each table, the top panel computes correlations between output in the source country (SC) and the number of nights spent in the destination country (MED); the middle panel computes correlations between the number of nights spent and output in the destination country (MED); the bottom panel computes correlations between output in the source country (SC) and in the destination country (MED). Starred values mean that the 6% confidence interval does not include zero. Confidence intervals are computed from 5 bootstrapped replications of the sample cross-correlation. Country codes: EA is Euro area; TN is Tunisia; FR is France. Table 3: Per-capita expenditures: unconditional cross-correlations Output in SC & Expenditures in MED Lags or leads (in years) EA - CY UK - CY Expenditures in MED & Output in MED EA - CY UK - CY Output in SC & Output in MED EA - CY UK - CY Notes: The numbers in the table represent corr(x t,y t+i), where i = [ 2, 1,, 1, 2], x t is the country listed first and y t is the country listed second. The sample length varies across cases: see the paper for details. In each table, the top panel computes correlations between output in the source country (SC) and per-capita expenditures in the destination country (MED); the middle panel computes correlations between per-capita expenditures and output in the destination country (MED); the bottom panel computes correlations between output in the source country (SC) and in the destination country (MED). Starred values mean that the 6% confidence interval does not include zero. Confidence intervals are computed from 5 bootstrapped replications of the sample cross-correlation. Country codes: EA is Euro area; UK is United Kingdom; CY is Cyprus. 24

26 2.4 Dynamic correlations Output in SC & Tourism in MED Table 4: Dynamic correlations Frequencies π 2 EA - CY UK - CY EA - TN FR - TN Tourism in MED & Output in MED Frequencies π 2 EA - CY UK - CY EA - TN.47.1 FR - TN Output in SC & Output in MED Frequencies π 2 EA - CY.76.2 UK - CY EA - TN FR - TN Notes: Frequencies centered at zero capture comovement in the long run; frequencies around π/2 coincide with business cycles of about four years. The sample length varies across cases: see the paper for details. The tourism variable identifies number of nights spent for Tunisia and per-capita expenditures for Cyprus. The top panel computes dynamic correlations between output in the source country (SC) and the tourism variable in the destination country (MED); the middle panel computes dynamic correlations between the tourism variable and output in the destination country (MED); the bottom panel computes dynamic correlations between output in the source country (SC) and in the destination country (MED). Country codes: EA is Euro area; UK is United Kingdom; FR is France; CY is Cyprus; TN is Tunisia. 25

27 2.5 Granger causality test Table 5: Nights spent: granger causality test % c.v. F-test EA output EA nights spent in TN EA nights spent in TN TN output FR output FR nights spent in TN FR nights spent in TN TN output Notes: If F-test > critical value, reject the null hypothesis of no Granger causation. Country codes: EA is Euro area; TN is Tunisia; FR is France. Table 6: Per-capita expenditures: granger causality test % c.v. F-test EA output EA expenditures in CY EA expenditures in CY CY output UK output UK expenditures in CY UK expenditures in CY CY output Notes: If F-test > critical value, reject the null hypothesis of no Granger causation. Country codes: EA is Euro area; CY is Cyprus; UK is United Kingdom. 26

28 2.6 Structural analysis Figure 27: IRFs to source country output shocks, sensitivity analysis EA output EA Tourism TN output EA output EA Tourism CY output TN consumption TN investment TN net exports CY consumption CY investment CY net exports Notes: Left: number of nights spent in Tunisia. Right: per-capita expenditures in Cyprus. Continuous line: median posterior IRF. Dotted lines: 6% confidence bands computed from the posterior distribution of IRFs. The order of the plots is the following: source country output, source country tourist variable, MED output, MED consumption, MED investment, MED net exports. Here, MED identifies either Tunisia or Cyprus. Country codes: EA is Euro area; CY is Cyprus; TN is Tunisia. Table 7: Forecast error variance decomposition, sensitivity analysis Tunisia Time horizon (in years) Cyprus Time horizon (in years) 1 4 EA tourism Shock1 5 7 (2,) (3,) (6,16) (7,17) Shock (,) (6,6) (7,) (76,) TN output Shock (,23) (26,41) (27,43) (26,41) Shock (2,) (,22) (16,2) (1,34) TN consumption Shock (3,1) (17,37) (1,37) (1,36) Shock (3,16) (14,32) (2,3) (22,4) TN investment Shock (6,25) (,32) (2,45) (26,43) Shock (21,44) (14,2) (17,33) (2,37) TN net exports Shock (7,33) (,2) (,27) (,2) Shock (,42) (1,43) (27,4) (2,5) 1 4 EA tourism Shock (,2) (15,34) (1,3) (21,3) Shock (71,) (5,7) (44,65) (3,5) CY output Shock (4,5) (36,55) (2,47) (2,47) Shock (1,5) (5,14) (,32) (14,32) CY consumption Shock (57,77) (4,7) (44,65) (41,63) Shock2 2 5 (1,4) (3,) (5,15) (7,1) CY investment Shock (3,66) (42,66) (42,63) (3,5) Shock (1,) (7,21) (,25) (,2) CY net exports Shock (,5) (14,43) (1,42) (2,43) Shock (,42) (1,5) (24,4) (22,44) Notes: The first column indicates the countries considered and the relevant variables in the VAR. Shock1 is output shock in the source country; Shock2 is tourism shock which identifies a shock to the number of nights spent in Tunisia and a shock to per-capita expenditures in Cyprus. The numbers in parenthesis are the lower and upper 6% confidence intervals. Country codes: EA is Euro area; CY is Cyprus; TN is Tunisia. 27