Trade in Tourism Services Explaining Tourism Trade and The

34
This article was downloaded by: [INASP - Pakistan (PERI)] On: 27 March 2014, At: 05:59 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK The Journal of International Trade & Economic Development: An International and Comparative Review Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rjte20 Trade in tourism services: Explaining tourism trade and the impact of the general agreement on trade in services on the gains from trade Camilla Jensen a & Jie Zhang b a School of Economics, Faculty of Arts and Social Sciences, University of Nottingham , Jalan Broga , Semenyih , 43500 , Malaysia b Bornholm Centre for Tourism Research , Denmark Published online: 20 Feb 2012. To cite this article: Camilla Jensen & Jie Zhang (2013) Trade in tourism services: Explaining tourism trade and the impact of the general agreement on trade in services on the gains from trade, The Journal of International Trade & Economic Development: An International and Comparative Review, 22:3, 398-429, DOI: 10.1080/09638199.2011.574723 To link to this article: http://dx.doi.org/10.1080/09638199.2011.574723 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed

Transcript of Trade in Tourism Services Explaining Tourism Trade and The

Page 1: Trade in Tourism Services Explaining Tourism Trade and The

This article was downloaded by: [INASP - Pakistan (PERI)]On: 27 March 2014, At: 05:59Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

The Journal of InternationalTrade & Economic Development:An International andComparative ReviewPublication details, including instructions for authorsand subscription information:http://www.tandfonline.com/loi/rjte20

Trade in tourism services:Explaining tourism trade and theimpact of the general agreementon trade in services on the gainsfrom tradeCamilla Jensen a & Jie Zhang ba School of Economics, Faculty of Arts and SocialSciences, University of Nottingham , Jalan Broga ,Semenyih , 43500 , Malaysiab Bornholm Centre for Tourism Research , DenmarkPublished online: 20 Feb 2012.

To cite this article: Camilla Jensen & Jie Zhang (2013) Trade in tourism services:Explaining tourism trade and the impact of the general agreement on trade inservices on the gains from trade, The Journal of International Trade & EconomicDevelopment: An International and Comparative Review, 22:3, 398-429, DOI:10.1080/09638199.2011.574723

To link to this article: http://dx.doi.org/10.1080/09638199.2011.574723

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, orsuitability for any purpose of the Content. Any opinions and views expressed

Page 2: Trade in Tourism Services Explaining Tourism Trade and The

in this publication are the opinions and views of the authors, and are not theviews of or endorsed by Taylor & Francis. The accuracy of the Content shouldnot be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions,claims, proceedings, demands, costs, expenses, damages, and other liabilitieswhatsoever or howsoever caused arising directly or indirectly in connectionwith, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expresslyforbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 3: Trade in Tourism Services Explaining Tourism Trade and The

© 2013 Taylor & Francis

The Journal of International Trade & Economic Development, 2013Vol. 22, No. 3, 398–429, http://dx.doi.org/10.1080/17517575.2011.574723

Trade in tourism services: Explaining tourism trade and theimpact of the general agreement on trade in services

on the gains from trade

Camilla Jensena* and Jie Zhangb

aSchool of Economics, Faculty of Arts and Social Sciences, University of Nottingham,Jalan Broga, Semenyih 43500, Malaysia; bBornholm Centre for Tourism Research,

Denmark

(Received 17 April 2010; final version received 21 March 2011)

The article addresses two questions related with tourism as a servicetrade. Can tourism be explained as other export activities? Does serviceliberalisation have a positive or negative impact on tourism receipts indestination countries? Previous research has either focused on thedemand side factors (i.e. factors of demand in the origin countries) or ontourism as a long-run factor of economic growth. The research showsthat a complementary perspective such as that offered by trade in asupply side perspective can render additional insights towards under-standing tourism. This approach can explain why countries haveabsolute and comparative advantage. Another finding is that tourismas an export can be explained by some of the same destination factorsthat explain other service exports. Using different panel estimators theimportance of supply side factors that are to some extent exclusive totourism are demonstrated: the general price competitiveness of thedestination, tourism infrastructure and the provision of safety. Theeconometric models also confirm the relevance of other conventionalexplanatory factors of trade in services such as GDP per capita andinternet usage. The last part of the article analyses the welfare gainsfrom trade under the general agreement on trade in services (GATS).The revenue (tourism receipt) effect is decomposed into a volume(arrival) and price effect. Results suggest that liberalisers under theGATS gained especially from a volume effect with average highergrowth rates in the number of arrivals. There is also found to be apositive effect on the average income earned per tourist from being aliberaliser.

Keywords: trade in services; tourism; GATS; liberalisation of trade inservices

JEL Classifications: F10; F12; F14; L83

*Corresponding author. Email: [email protected]

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 4: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 399

Trade in tourism services: Explaining tourism trade and theimpact of the general agreement on trade in services

on the gains from trade

Camilla Jensena* and Jie Zhangb

aSchool of Economics, Faculty of Arts and Social Sciences, University of Nottingham,Jalan Broga, Semenyih 43500, Malaysia; bBornholm Centre for Tourism Research,

Denmark

(Received 17 April 2010; final version received 21 March 2011)

The article addresses two questions related with tourism as a servicetrade. Can tourism be explained as other export activities? Does serviceliberalisation have a positive or negative impact on tourism receipts indestination countries? Previous research has either focused on thedemand side factors (i.e. factors of demand in the origin countries) or ontourism as a long-run factor of economic growth. The research showsthat a complementary perspective such as that offered by trade in asupply side perspective can render additional insights towards under-standing tourism. This approach can explain why countries haveabsolute and comparative advantage. Another finding is that tourismas an export can be explained by some of the same destination factorsthat explain other service exports. Using different panel estimators theimportance of supply side factors that are to some extent exclusive totourism are demonstrated: the general price competitiveness of thedestination, tourism infrastructure and the provision of safety. Theeconometric models also confirm the relevance of other conventionalexplanatory factors of trade in services such as GDP per capita andinternet usage. The last part of the article analyses the welfare gainsfrom trade under the general agreement on trade in services (GATS).The revenue (tourism receipt) effect is decomposed into a volume(arrival) and price effect. Results suggest that liberalisers under theGATS gained especially from a volume effect with average highergrowth rates in the number of arrivals. There is also found to be apositive effect on the average income earned per tourist from being aliberaliser.

Keywords: trade in services; tourism; GATS; liberalisation of trade inservices

JEL Classifications: F10; F12; F14; L83

*Corresponding author. Email: [email protected]

1. Introduction

Trade in tourism services and the benefits of free trade in this type ofservices are topics that have received less attention in internationaleconomics. Tourism is analysed by a specialised group of people or tourismeconomists with relatively little dialogue across sub-disciplines. Theobjective of this article is to fill some of this gap in the literature. Weshow that tourism activity not alone is decided by demand-side factors. Inother words, nations also compete in this activity by directly or indirectlyemploying strategies to attract tourists. As in other industries, ‘natural’comparative advantage based on natural resources (sun, sand and sea intourism) may be challenged by the rising importance of created assets suchas entertainment and culture industries, new technologies, public andprivate infrastructure and institutions. A secondary objective with theresearch is to understand the position of developing countries within anindustry such as tourism. Do they have a ‘natural’ comparative advantage inthis industry or is tourism yet another activity where they are overshadowedby the developed countries?

Within mainstream economics there has evolved two different traditionstowards the analysis of service industries and trade. One approach beingcentral to mainstream or neoclassical economics is to juxtapose services withmanufacturing and use the same models with little adaptation. The mainargument in this stream of literature is that the economic laws are the samefor services and manufacturing industries (see, for example, Deardorf 2001).Some authors have shown that models explaining trade in manufacturesmay be equally relevant towards understanding trade in services (Kimuraand Lee 2006). The other approach is to reject the application of mainstreammodels as adequate to analyse services. This group of researchers stresses thedifferences between services and manufacturing starting from the basicpremise that services are intangibles that must be consumed at their point ofproduction (see, for example, Chesbrough and Spohrer 2006). Hence,services are difficult to store and save for future consumption. Servicesrequire interaction between users and producers rendering servicescomparatively heterogeneous to manufactured goods (Mirza and Nicoletti2004). Output may be more difficult to account for in a standardised wayacross different service industries (Griliches 1992).

This study, being among the first to offer a globally comparative paneldata analysis on tourism as a service trade, shows that developing countrieshave only comparative advantage and little absolute advantage in tourism(as in most other industries). The OECD countries are also often the mainexporters in this industry. Our models suggest in complement to the demandside models (which have demonstrated the importance of factors such astravel time and cost including relative price differences and exchange rates,see Song and Li 2008) that OECD countries are advantaged because of

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 5: Trade in Tourism Services Explaining Tourism Trade and The

400 C. Jensen and J. Zhang

technology factors, including tourism infrastructure and their betterprovision of safety compared to developing countries. It is shown in arobust econometric framework that the general price competitiveness of thedestination only matters to gaining absolute advantage or generating marketshare. Hence, we can conclude that most developing countries are poorperformers in tourism since they are relatively small and mainly holdcomparative advantage. The parameter on which they can compete (price) isovershadowed by a host of other factors that are important when attractingtourists such as technology, public and private infrastructure includingsufficient institutions to provide safety. Maybe for this reason we alsoobserve that a few lucrative islands on the edge of development (forexample, Bermuda, Bahamas, Guam, Macau and Palau) are leading in theindustry as the most comparatively advantaged because they offer the bestof both worlds (superior natural endowment for tourism, island setting,relative proximity to origin countries, better technologies and infrastructureincluding the capacity to provide safety).

Section 2 gives a short review of the existing literature on tourism andservice trade. Section 3 introduces the data which is used subsequently toexplain tourism as an exporting activity for individual countries. The choiceof econometric model in relation to the identification problem is discussed inSection 4. Section 5 gives an account of the statistical results using firstdifferencing for panel data (GMM). The relationship between the growthrates in tourism receipts and liberalisation under the general agreement ontrade in services (GATS) is analysed in Section 6. Section 7 concludes thearticle.

2. Trade in services

A large number of studies in the tourism literature confirm the relevance ofsupply side factors. However, the tourism literature is dominated by demandmodels (see Crouch 1994, and Lim 1997 or for a review of the most recentliterature, see Song and Li 2008). Tourism demand studies have shown theimportance of demand side factors such as: travel distance and time, relativeprice differences between pairs of countries and exchange rates. Some of themost important results for the supply side perspectives in the tourismliterature are discussed by Prideaux (2005) and Cruz and Rolim (2005).Prideaux (2005) gives a broad overview of determinants of tourism flows in asupply side perspective: natural resources, culture and distance, policy,institutions, infrastructure, destination price, external political and healthfactors. General equilibrium models have also been offered to analyse issuesrelated with tourism such as climate change (see, for example, Berrittellaet al. 2006). They incorporate factors of supply related with population,income per capita and climate including fixed factors. The most comparativestudy to date in a supply side perspective is given by Cruz and Rolim (2005)

C. Jensen and J. Zhang

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 6: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 401

technology factors, including tourism infrastructure and their betterprovision of safety compared to developing countries. It is shown in arobust econometric framework that the general price competitiveness of thedestination only matters to gaining absolute advantage or generating marketshare. Hence, we can conclude that most developing countries are poorperformers in tourism since they are relatively small and mainly holdcomparative advantage. The parameter on which they can compete (price) isovershadowed by a host of other factors that are important when attractingtourists such as technology, public and private infrastructure includingsufficient institutions to provide safety. Maybe for this reason we alsoobserve that a few lucrative islands on the edge of development (forexample, Bermuda, Bahamas, Guam, Macau and Palau) are leading in theindustry as the most comparatively advantaged because they offer the bestof both worlds (superior natural endowment for tourism, island setting,relative proximity to origin countries, better technologies and infrastructureincluding the capacity to provide safety).

Section 2 gives a short review of the existing literature on tourism andservice trade. Section 3 introduces the data which is used subsequently toexplain tourism as an exporting activity for individual countries. The choiceof econometric model in relation to the identification problem is discussed inSection 4. Section 5 gives an account of the statistical results using firstdifferencing for panel data (GMM). The relationship between the growthrates in tourism receipts and liberalisation under the general agreement ontrade in services (GATS) is analysed in Section 6. Section 7 concludes thearticle.

2. Trade in services

A large number of studies in the tourism literature confirm the relevance ofsupply side factors. However, the tourism literature is dominated by demandmodels (see Crouch 1994, and Lim 1997 or for a review of the most recentliterature, see Song and Li 2008). Tourism demand studies have shown theimportance of demand side factors such as: travel distance and time, relativeprice differences between pairs of countries and exchange rates. Some of themost important results for the supply side perspectives in the tourismliterature are discussed by Prideaux (2005) and Cruz and Rolim (2005).Prideaux (2005) gives a broad overview of determinants of tourism flows in asupply side perspective: natural resources, culture and distance, policy,institutions, infrastructure, destination price, external political and healthfactors. General equilibrium models have also been offered to analyse issuesrelated with tourism such as climate change (see, for example, Berrittellaet al. 2006). They incorporate factors of supply related with population,income per capita and climate including fixed factors. The most comparativestudy to date in a supply side perspective is given by Cruz and Rolim (2005)

who analysed international tourism flows among the developing economiesof South America, Africa and South Asia. This study confirmed theimportance of supply side determinants for international tourism flows,including national income, tourism attractions, security risks and thegeographical distance to main markets.

Studies of trade in services use general explanatory factors of exportperformance: GDP, exchange rates and internet adoption (Freund andWeinhold 2002). Kimura and Lee (2006) show that the important workhorsein empirical international economics being the gravity equation also workssurprisingly well (even better than mirror samples for manufacturing)towards explaining trade in services. According to Kimura and Lee’s (2006)study, distance is a highly significant negative factor preventing servicetrade. Mirza and Nicoletti (2004) report similar results but give otherexplanations besides physical distance for why gravity type of equationswork particularly well in services related with the o-ring theory ofdevelopment (Kremer 1993).1

Even though some general studies exist in service trade, most availableresearch focuses on single industries among which the most studied arebanking, insurance, advertising and other professional services, e.g.consultancy. With the GATS, new areas are emerging – in particular healthand educational services. In this perspective, tourism is the odd child since itis rarely mentioned in the economics literature. One exception is the studyby Eilat and Einav (2004). However, that study is opposite ours based on thecentral premise in tourism studies being that tourism should be explainedfrom the demand side rather than the supply side.

Studies of banking and insurance emphasise the importance of the cross-border supply mode through foreign direct investment (FDI) towardsgenerating trade (Li, Moshirian, and Sim 2003; Webster and Hardwick2005). Most of the service trade here takes place either as intra-firmtransactions or via arms length transactions among larger and global playersin the industry. Trade in this class of services is best explained by thepresence of FDI and hence indirectly the factors that attract FDI into theseservice industries. Health and educational services share features with bothbanking (cross-border supply) and tourism (flow of people instead of goods)in the way they are traded, where especially the cross-border mode of supplyhas increased in importance across all activities since the conclusion of theUruguay round (Smith 2004).

A number of articles that deal with the relationship between tourism andeconomic growth are also relevant to the present study, mainly due to themethodological challenges of uncovering a robust relationship betweentourism, growth and level of development. For example, Hazari and Sgro(1995) developed a dynamic model showing that tourism can have positiveeffects on the long-term growth in a small economy. They proved that tourismmakes non-traded goods and services exportable, hence promoting domestic

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 7: Trade in Tourism Services Explaining Tourism Trade and The

402 C. Jensen and J. Zhang

welfare and growth. Balaguer andCantavella-Jorda (2002) have examined therole of tourism in the Spanish long-run economic development by usingJohansen’s cointegration methodology and Granger causality tests. Grangercausality tests confirmed the existence of a robust relationship betweeneconomic growth and tourism expansion. Dritsakis (2004) has also examinedthe relationship among the international tourism earning, real exchange rateand economic growth in Greece by using Johansen’s cointegration test andthen Granger causality tests based on a vector error correction model. Theresults of cointegration analysis suggest the existence of a cointegrationrelationship between the three variables. This indicates the presence of acommon trend or a long-run relationship among these variables.

Dritsakis’ findings from causality analysis denote that internationaltourism earnings and real exchange rates cause economic growth in a ‘strongcausal’ relationship, while economic growth and real exchange rate causeinternational tourism earning in a ‘simple causal’ relationship.

Most of the papers focused on explaining tourism as an important factorfor economic growth. However, there are a few papers dealing with bothways of causality. Eugenio-Martin, Moralse and Scarpa (2004) analysed therelationship between tourism and economic growth for Latin Americancountries. The study showed that only in the low and medium incomecountries is tourism growth associated with economic growth. At the sametime has the Latin American study shown that tourist arrivals are positivelyrelated with GDP per capita in all three types of countries (low, medium andhigh income). Tourist arrivals in the high income Latin American countriesrely also on secondary education enrolment, while in the medium incomecountries, tourist arrivals are positively related with higher expectancy oflife. In the low income Latin American countries, tourist arrivals depend onall the three main factors: infrastructure, improvement in education andsafety. In this way, they argue that these are important three areas in whichdeveloped countries have an absolute advantage over developing ones. Thesame kind of tests was made by Oh (2005) by using data for South Korea.Oh (2005) did not confirm the same results as Balaguer and Cantavella-Jorda (2002) in the case of Spain. This means that cointegration betweentourism and economic growth did not exist in South Korea. On the otherhand, did Oh’s study confirm that rapid economic expansion causes a rise intourism receipts in South Korea?

Sequeira and Nunes (2008) apply a dynamic panel data method to testthe relationship between tourism and economic growth in different types ofcountries. They have only confirmed that tourism specialisation is positivelyrelated with economic growth in the poor countries, but not necessarily thesmall countries. This result is indirectly related with the results fromEugenio-Martin, Moralse, and Scarpa (2004).

Lee and Chang (2008) also test two-way causality between tourismdevelopment and economic growth by using fully modified ordinary least

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 8: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 403

welfare and growth. Balaguer andCantavella-Jorda (2002) have examined therole of tourism in the Spanish long-run economic development by usingJohansen’s cointegration methodology and Granger causality tests. Grangercausality tests confirmed the existence of a robust relationship betweeneconomic growth and tourism expansion. Dritsakis (2004) has also examinedthe relationship among the international tourism earning, real exchange rateand economic growth in Greece by using Johansen’s cointegration test andthen Granger causality tests based on a vector error correction model. Theresults of cointegration analysis suggest the existence of a cointegrationrelationship between the three variables. This indicates the presence of acommon trend or a long-run relationship among these variables.

Dritsakis’ findings from causality analysis denote that internationaltourism earnings and real exchange rates cause economic growth in a ‘strongcausal’ relationship, while economic growth and real exchange rate causeinternational tourism earning in a ‘simple causal’ relationship.

Most of the papers focused on explaining tourism as an important factorfor economic growth. However, there are a few papers dealing with bothways of causality. Eugenio-Martin, Moralse and Scarpa (2004) analysed therelationship between tourism and economic growth for Latin Americancountries. The study showed that only in the low and medium incomecountries is tourism growth associated with economic growth. At the sametime has the Latin American study shown that tourist arrivals are positivelyrelated with GDP per capita in all three types of countries (low, medium andhigh income). Tourist arrivals in the high income Latin American countriesrely also on secondary education enrolment, while in the medium incomecountries, tourist arrivals are positively related with higher expectancy oflife. In the low income Latin American countries, tourist arrivals depend onall the three main factors: infrastructure, improvement in education andsafety. In this way, they argue that these are important three areas in whichdeveloped countries have an absolute advantage over developing ones. Thesame kind of tests was made by Oh (2005) by using data for South Korea.Oh (2005) did not confirm the same results as Balaguer and Cantavella-Jorda (2002) in the case of Spain. This means that cointegration betweentourism and economic growth did not exist in South Korea. On the otherhand, did Oh’s study confirm that rapid economic expansion causes a rise intourism receipts in South Korea?

Sequeira and Nunes (2008) apply a dynamic panel data method to testthe relationship between tourism and economic growth in different types ofcountries. They have only confirmed that tourism specialisation is positivelyrelated with economic growth in the poor countries, but not necessarily thesmall countries. This result is indirectly related with the results fromEugenio-Martin, Moralse, and Scarpa (2004).

Lee and Chang (2008) also test two-way causality between tourismdevelopment and economic growth by using fully modified ordinary least

squares (OLS) models. They have 23 OECD countries and 32 non-OECDcountries covering the three continents of Asia, Latin America and Sub-Saharan Africa. The results show that cointegration between GDP andtourism development is substantial. They prove that tourism developmenthas a greater impact on GDP in the non-OECD countries than in OECDcountries. Their causality tests show, however, a unidirectional causalityrelationship from tourism development to economic growth in OECDcountries, and bidirectional relationship in non-OECD countries. Theseresults also give supporting evidence for the less-developed countries,namely that growth in economic development and better infrastructure havea positive impact for tourism development.

Our global panel data study complements this literature in several ways.Firstly, by taking a comparative development perspective on the factors thatgenerate trade in tourism services. Secondly, by adopting a panel dataapproach (GMM) instead of the prevailing time series approach towardsresolving problems of endogeneity (two-way causation) and/or spuriousrelationships (third unobserved factors influence the results). Another maindifference between our study andmost of the available supply side literature isthat instead of analysing their particular influence, we instead control forcountry fixed factors (such as distance, climate and culture) since suchstructural or cross-sectional factors are less relevant to analyse under a paneldata approach (and of course ignoring that even some structural factors dochange over time, e.g. through lowering of transport cost, cultural or climatechange).Atmostwe can treat themas nuisance parameterswith the fixed effectmodel, random parameters with other models or reduce them out usingdynamic models. Hence, we focus in our study on the supply side factors thatcountries can and do affect through their policies and investments over time.Compared to the literature on tourism and economic growth, this articlefocuses only on the single causal relationship running from a nation’s wealthand development towards generating tourism activities.

3. Data

The data used in the study is drawn from two main sources and a number ofcomplimentary sources. Tourism data is obtained from the UN WorldTourism Organization’s (UNWTOs) database (www.unwto.org) and itspublication ‘Yearbook of Tourism Statistics’. General economic data istaken from the World Bank’s World Development Indicators (WDI)published online (www.worldbank.org). In addition, hereto data aboutliberalisation of trade in services under the GATS is taken from the WorldTrade Organization’s (WTOs) online databases on trade in services(www.wto.org). Data on political risk as an approximation to the tourist’sperception of un-safety is taken from Freedom House (www.freedomhouse.org).

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 9: Trade in Tourism Services Explaining Tourism Trade and The

404 C. Jensen and J. Zhang

Individual indicators are explained in further detail now according to theorder they are used in the subsequent sections.

Section 5 concerns the analysis of the determinants of tourism flows inthe supply side perspective of destination attractiveness. The dependentvariable in this part of the study should ideally reflect the absolute andcomparative advantage of the destination country in terms of its absoluteand relative ability to attract tourists and hence generate tourism exports.To calculate absolute and comparative advantage tourist arrivals is usedARRIVAL rather than receipts for two reasons. Firstly, quantities avoidproblems of cross-country comparison due to exchange rate volatility anddifferential exchange rates. Secondly, data availability and reliability withrespect to arrivals is generally better than with respect to tourism receipts.Actual data estimates of trade in tourism services from the WTO couldalso have been used towards calculating absolute and comparativeadvantage. However, the publication and availability of such data are ofa recent nature relative to the tourism data used in the study which goesback to 1982.

As dependent variables are adopted measures of absolute advantage bycalculating annual shares of the market held by each country TOURSH andcomparative advantage by dividing tourism market shares with populationshares TOURSH/POPSH.2 A comparison of these variables for the year2005 is shown in Appendix Table A1. From here it can be verified thatabsolute advantage or market share in international tourism is quitedistinct from comparative advantage. Only a few countries hold bothabsolute and comparative advantage in tourism. The largest economieshave absolute advantage. The top 10 countries in terms of comparativeadvantage are all small and belong often to the category of lucrative islandeconomies.

The variables include data on level of development GDPCAP (capturedwith GDP per capita), investment in tourism infrastructure ROOMS(captured either with total hotel capacity per 1000 capita or the absolutenumber of hotel rooms available), openness of the economy OPEN(captured with the sum of total trade in GDP), adoption of new ICTtechnologies INNET (captured either with the number of registered internetusers per 1000 capita or the absolute number of internet users), the pricecompetitiveness of the destination IPP (captured with GDP expressed inPPP relative to GDP at official exchange rates – in other words how muchspending power that a USD denominated wage translates into in thedestination country) and finally the tourist’s perception of safety UNSAFE(captured with the political rights index published by Freedom HouseRISK), where a higher index means less political freedom or less perceivedsafety. All variables that are not ratios have been log transformed to reducefor the influence of outliers in forth-running data series. This creates moredata consistency by including ratios that already take into account or adjust

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 10: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 405

Individual indicators are explained in further detail now according to theorder they are used in the subsequent sections.

Section 5 concerns the analysis of the determinants of tourism flows inthe supply side perspective of destination attractiveness. The dependentvariable in this part of the study should ideally reflect the absolute andcomparative advantage of the destination country in terms of its absoluteand relative ability to attract tourists and hence generate tourism exports.To calculate absolute and comparative advantage tourist arrivals is usedARRIVAL rather than receipts for two reasons. Firstly, quantities avoidproblems of cross-country comparison due to exchange rate volatility anddifferential exchange rates. Secondly, data availability and reliability withrespect to arrivals is generally better than with respect to tourism receipts.Actual data estimates of trade in tourism services from the WTO couldalso have been used towards calculating absolute and comparativeadvantage. However, the publication and availability of such data are ofa recent nature relative to the tourism data used in the study which goesback to 1982.

As dependent variables are adopted measures of absolute advantage bycalculating annual shares of the market held by each country TOURSH andcomparative advantage by dividing tourism market shares with populationshares TOURSH/POPSH.2 A comparison of these variables for the year2005 is shown in Appendix Table A1. From here it can be verified thatabsolute advantage or market share in international tourism is quitedistinct from comparative advantage. Only a few countries hold bothabsolute and comparative advantage in tourism. The largest economieshave absolute advantage. The top 10 countries in terms of comparativeadvantage are all small and belong often to the category of lucrative islandeconomies.

The variables include data on level of development GDPCAP (capturedwith GDP per capita), investment in tourism infrastructure ROOMS(captured either with total hotel capacity per 1000 capita or the absolutenumber of hotel rooms available), openness of the economy OPEN(captured with the sum of total trade in GDP), adoption of new ICTtechnologies INNET (captured either with the number of registered internetusers per 1000 capita or the absolute number of internet users), the pricecompetitiveness of the destination IPP (captured with GDP expressed inPPP relative to GDP at official exchange rates – in other words how muchspending power that a USD denominated wage translates into in thedestination country) and finally the tourist’s perception of safety UNSAFE(captured with the political rights index published by Freedom HouseRISK), where a higher index means less political freedom or less perceivedsafety. All variables that are not ratios have been log transformed to reducefor the influence of outliers in forth-running data series. This creates moredata consistency by including ratios that already take into account or adjust

for outlier problems with inclusion of forth-running data series in the sameequation.

Section 6 analyses separately the influence of liberalisation on arrivalsand tourism receipts. The indicators hereof are: tourism receipts RECEIPT,tourism arrivals ARRIVAL and the average estimated ex-post priceactually paid by a tourist for a travel experience PRICE. Liberalisation oftourism services is measured by using the number of commitmentsCOMMIT made by each country under GATS in this particular industry.An index of 0 measures zero commitments, whereas an index between 1and 4 reflect countries having made commitments not to introduce newrestrictions in each of the four modes of supply specific to the tourismindustry (hotels and restaurants, travel agencies and tour operator services,tourist guide services and other). This simple indicator assumes that eachmode has equal weight towards liberalisation. Also it is assumed with thisvariable that countries showing commitment to liberalisation are generallyalso those that have lifted most restrictions on their tourism services.However, it should be noted that the GATS system is fundamentallydifferent from other areas of free trade under the WTO since it is based onvolition.

The data variables and their definitions are summarised in Table 1. Thevariables are available in a potential panel of 190 countries and 25 years(1982–2006). Even though the panel is fairly balanced, the data availabilityand the length of the panel is country specific.

4. Model discussion

With the econometric model the aim is to estimate an export supply curve.The main modelling problem concerns the price variable due to the problemof identification. A general problem is that paired observations of price andquantity (in our case tourism arrivals) are the outcome of the theoreticalrelationships inherent in a supply and demand curve system. There existseveral techniques to overcome the identification problem such as theadoption of instrumental variables that can help to pin down or identifyeach curve. For panel data also advanced techniques have been developedand several textbooks deal almost exclusively with this class of problems(see, for example, Arellano 2003).

Another methodological issue is that of choosing the right priceinformation. This problem has generally been less dealt with in theliterature. For example, in tourism studies the ex-post price is often used.3

Using ex-post price information may introduce a tautological problem intothe analysis similar to the tautological problems of analysing centralrelationships in national accounting data that are arrived at using identities.With the choice of price variable, the aim is to overcome both problems atthe same time. If a price variable can be found at the level of countries that is

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 11: Trade in Tourism Services Explaining Tourism Trade and The

406 C. Jensen and J. Zhang

Table 1. Definition of study variables.

ARRIVALit Number of tourist arrivals including all international arrivals nomatter purpose of visit (leisure or business) using UNWTOdata.

COMMITi Number of commitments (from 0 to 4) each country has madeunder the GATS agreement in tourism services using WTOdata (www.wto.org).

COMDUMi A dummy that takes the value of 1 if a country at all is committedto reform of its trading system related with tourism.

DUMMY95 A dummy that takes the value of 1 for all years after 1995 wherethe internet is generally introduced according to the sampleddata from the World Bank.

GDPCAPit Level of development and general state of technology measuredby the income per capita as calculated by the World Bank inUSD in 1995 prices.

INNETit The absolute number of internet users from World Bank WDIdata.

INNETpmcapit Number of per 1000 capita internet users using World Bank WDIdata.

ISLANDi Defined in this study as countries without a land lock and with alocal population less than 15 million (which includes Cuba butexcludes Australia).

OPENit Openness measured as total trade (imports and exports) overGDP as calculated by the World Bank in the WDI series.

POPit Total population measured using World Bank WDI data.POPSHit Population share calculated by dividing country populations with

the global population using World Bank WDI data.IPPit The international price competitiveness of the destination

measured by the ratio of GDP in PPP to GDP at the marketexchange rate (both series in USD using current price data)using World Bank WDI data.

PRICEit The average real price paid per tourist for a travel to the countryin USD, calculated by dividing total real receipts with totalarrivals using UNWTO or World Bank WDI data.

RECEIPTit Total real gross receipts in mio. 2000 USD (using UNWTO orWorld Bank WDI data) that a country is estimated to earnfrom its tourism industry. According to the UNWTO, receiptsinclude an estimate of all expenditures of tourists whilevacating on lodging, food, fuel, transport, entertainment andshopping. It excludes however receipts related with interna-tional transport. For purposes of deflation is used the USdeflator from the WDI series.

ROOMSit The absolute number of hotel rooms according to UNWTO data.ROOMSpmcapit Hotel capacity in number of rooms per 1000 capita using

UNWTO data and population data from the World BankWDI data.

UNSAFEit How safe it is to visit the destination country measured by usingthe political rights index calculated by Freedom House.

TOURSHpmit Market share in promille in tourism calculated by dividingtourism arrivals for each country with the global number ofarrivals using UNWTO data.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 12: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 407

Table 1. Definition of study variables.

ARRIVALit Number of tourist arrivals including all international arrivals nomatter purpose of visit (leisure or business) using UNWTOdata.

COMMITi Number of commitments (from 0 to 4) each country has madeunder the GATS agreement in tourism services using WTOdata (www.wto.org).

COMDUMi A dummy that takes the value of 1 if a country at all is committedto reform of its trading system related with tourism.

DUMMY95 A dummy that takes the value of 1 for all years after 1995 wherethe internet is generally introduced according to the sampleddata from the World Bank.

GDPCAPit Level of development and general state of technology measuredby the income per capita as calculated by the World Bank inUSD in 1995 prices.

INNETit The absolute number of internet users from World Bank WDIdata.

INNETpmcapit Number of per 1000 capita internet users using World Bank WDIdata.

ISLANDi Defined in this study as countries without a land lock and with alocal population less than 15 million (which includes Cuba butexcludes Australia).

OPENit Openness measured as total trade (imports and exports) overGDP as calculated by the World Bank in the WDI series.

POPit Total population measured using World Bank WDI data.POPSHit Population share calculated by dividing country populations with

the global population using World Bank WDI data.IPPit The international price competitiveness of the destination

measured by the ratio of GDP in PPP to GDP at the marketexchange rate (both series in USD using current price data)using World Bank WDI data.

PRICEit The average real price paid per tourist for a travel to the countryin USD, calculated by dividing total real receipts with totalarrivals using UNWTO or World Bank WDI data.

RECEIPTit Total real gross receipts in mio. 2000 USD (using UNWTO orWorld Bank WDI data) that a country is estimated to earnfrom its tourism industry. According to the UNWTO, receiptsinclude an estimate of all expenditures of tourists whilevacating on lodging, food, fuel, transport, entertainment andshopping. It excludes however receipts related with interna-tional transport. For purposes of deflation is used the USdeflator from the WDI series.

ROOMSit The absolute number of hotel rooms according to UNWTO data.ROOMSpmcapit Hotel capacity in number of rooms per 1000 capita using

UNWTO data and population data from the World BankWDI data.

UNSAFEit How safe it is to visit the destination country measured by usingthe political rights index calculated by Freedom House.

TOURSHpmit Market share in promille in tourism calculated by dividingtourism arrivals for each country with the global number ofarrivals using UNWTO data.

both ex-ante and exogenous to the system analysed then both problems havebeen overcome.

The general export supply equation to be estimated takes the followingform, where export supply or quantity of exports is modelled as a linearfunction of price (which takes an expected negative sign) and a number ofother relevant explanatory variables V such as supply side factors includingtourism infrastructure:

XSi ¼ a0 þ b0Pi� þ gnVi þ ei ð1Þ

Fitting the equation to the data, the final equation to be estimated takesthe following form when estimated as a fixed effect model:

TOURSHpmit ¼ ai þ b0IPPþ Pit þ g1LogðGDPCAPÞit þ g2LogðROOMSÞitþ g3LogðINNETÞit þ g4OPENit þ g5UNSAFEit þ lt þ eit

ð1aÞ

Or estimated using first differencing with a lagged dependent variablewhere the fixed effects drop out (but maintaining the fixed year effects l) theequation to be estimated takes the following form:

DTOURSHpmit ¼ a0TOURSHpmit�1 þ b0DIPPitþ � g1DLogðGDPCAPÞitþ g2DLogðROOMSÞit � g3DLogðINNETÞit þ g4DOPENit

þ g5DUNSAFEit þ lt þ eit ð1bÞ

Note that the expected sign for the price variable changes (from negativeto positive) in the applied version of equation (1) because of the way price ismeasured in the study. Price takes outset in the perspective of the foreigncurrency holder, hence when IPP is higher it implies that the price oftourism activities in the destination country is lower.

To explain comparative advantage, a population weighted version ofequation (1b) is adopted (the weights are applied to forthrunningvariables whose size will co-vary with country size – that is ROOMSand INNET):

DTOURSH

POPSH

� �

it

¼ a0TOURSH

POPSH

� �

it�1

þb0DIPPit þ g1DLogðGDPCAPÞit

þ g2DLogðROOMSpmcapÞit þ g3DLogðINNETpmcapÞitþ g4DOPENit þ g5DUNSAFEit þ lt þ eit ð1cÞ

To test the assumption of exogeneity between our price IPP and quantityvariables TOURSH and TOURSH/POPSH a Granger causality test was

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 13: Trade in Tourism Services Explaining Tourism Trade and The

408 C. Jensen and J. Zhang

undertaken (see Appendix Table A2). If the price passes the exogeneityassumption we should find a causal relationship running from the price tothe quantity variables but not the reverse. The Granger shows a strongcausal relationship in particular running from quantity to price.4 This ispuzzling since our price variable is so aggregate that it should not bepossible to be strongly affected by tourism activities except in thosecountries that are highly dependent on tourism. Also we found that on aregional basis this causal relationship disappears. This indicates a spuriousrelationship inherent in the data that is not related with the identificationproblem. Pearson correlation coefficients (see Appendix Table A3) confirmthat there may be a spurious relationship between the price variable IPP, thequantity variable TOURSH and the level of income GDPCAP. Lowerpriced countries (using our price variable IPP) are the less developedcountries because of their stronger incubation from international transmis-sion mechanisms on their internal price levels. (At the same time willdevaluation also lead to improvement in IPP since GDP at official exchangerates diminishes following a devaluation whereas GDP at PPP is the same.)These countries are at the same time generally less competitive in tourismactivities especially when measured on TOURSH. Therefore, the pricevariable passes the exogeneity test but its investigation introduces instead aspurious relationship into the analysis. For this reason, the first differenceapproach is the best approach to the data from a time series perspective. Inpanel data econometrics, this translates into the choice of the generalmethod of moments (GMM) estimator (see, for example, Bond 2002, 15).

5. Econometric results

Results of the panel data analysis are shown with Table 2 (absoluteadvantage or market shares) and Table 3 (comparative advantage orpopulation weighted markets shares), respectively. Sub-sample results byincome groupings are reported in Section 5.2.

5.1. Results for the full sample

First is discussed the performance of the panel data models, then the resultsare interpreted specifically with respect to individually explained andexplanatory variables. The section rounds off with special attention paid tothe internet and its differential role over the period analysed.

The fixed effect model generally performs poorly on the data because ofits emphasis on within country variation. All the between country variationin the data is in this model captured with the fixed effects. The randomeffects model performs slightly better because as a mixed model it gives moreweight to the cross-section dimension in the data. Both models suffer undersevere problems of auto correlation even though both include time fixed

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 14: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 409

undertaken (see Appendix Table A2). If the price passes the exogeneityassumption we should find a causal relationship running from the price tothe quantity variables but not the reverse. The Granger shows a strongcausal relationship in particular running from quantity to price.4 This ispuzzling since our price variable is so aggregate that it should not bepossible to be strongly affected by tourism activities except in thosecountries that are highly dependent on tourism. Also we found that on aregional basis this causal relationship disappears. This indicates a spuriousrelationship inherent in the data that is not related with the identificationproblem. Pearson correlation coefficients (see Appendix Table A3) confirmthat there may be a spurious relationship between the price variable IPP, thequantity variable TOURSH and the level of income GDPCAP. Lowerpriced countries (using our price variable IPP) are the less developedcountries because of their stronger incubation from international transmis-sion mechanisms on their internal price levels. (At the same time willdevaluation also lead to improvement in IPP since GDP at official exchangerates diminishes following a devaluation whereas GDP at PPP is the same.)These countries are at the same time generally less competitive in tourismactivities especially when measured on TOURSH. Therefore, the pricevariable passes the exogeneity test but its investigation introduces instead aspurious relationship into the analysis. For this reason, the first differenceapproach is the best approach to the data from a time series perspective. Inpanel data econometrics, this translates into the choice of the generalmethod of moments (GMM) estimator (see, for example, Bond 2002, 15).

5. Econometric results

Results of the panel data analysis are shown with Table 2 (absoluteadvantage or market shares) and Table 3 (comparative advantage orpopulation weighted markets shares), respectively. Sub-sample results byincome groupings are reported in Section 5.2.

5.1. Results for the full sample

First is discussed the performance of the panel data models, then the resultsare interpreted specifically with respect to individually explained andexplanatory variables. The section rounds off with special attention paid tothe internet and its differential role over the period analysed.

The fixed effect model generally performs poorly on the data because ofits emphasis on within country variation. All the between country variationin the data is in this model captured with the fixed effects. The randomeffects model performs slightly better because as a mixed model it gives moreweight to the cross-section dimension in the data. Both models suffer undersevere problems of auto correlation even though both include time fixed T

able

2.

Panel

data

results,absolute

advantage.

Panel

method

GMM

(FD)

GMM

(FD)

GMM

(FD)

GMM

(FD)

GMM

(FD)

Dependentvariable:

TOURSHpm

TOURSHpm

TOURSHpm

TOURSHpm

TOURSHpm

Explanatory

vars:

equation(1b.1)

equation(1b.2)

equation(1b.3)

equation(1b.4)

equation(1b.5)

LDV

0.87***

0.89***

0.81***

0.80***

0.72***

(40.63)

(43.82)

(29.65)

(32.75)

(24.27)

IPP

1.04***

0.75**

1.37***

1.00**

1.26***

(2.45)

(1.89)

(2.61)

(2.41)

(2.38)

Log(G

DPCAP)

3.48**

–10.82***

8.46***

11.17***

(2.03)

(4.70)

(4.53)

(4.69)

Log(R

OOMS)

1.43***

1.45***

3.22***

70.07

1.18

(2.67)

(2.72)

(4.64)

(70.12)

(1.45)

Log(INNET)

70.01

0.00

0.08*

0.36***

0.23**

(70.29)

(0.04)

(1.74)

(4.99)

(1.98)

(Log(INNET))2

––

–70.03***

(75.94)

OPEN

0.02*

0.03***

0.03**

0.03**

0.06***

(1.73)

(2.54)

(2.00)

(2.09)

(2.98)

UNSAFE

70.78***

70.72***

70.94***

70.84***

70.81***

(72.73)

(72.55)

(72.67)

(72.99)

(72.81)

DUMMY95

––

711.89***

––

(77.84)

Countryeff

ects

FD

FD

FD

FD

FD

Tim

eeff

ects

Yes*

Yes*

Yes**

Yes*

Yes***

Number

ofobs.

2100(1987–2005)

2100(1987–2005)

2100(1987–2005)

2100(1987–2005)

1335(1995–2005)

AR(2)

0.05

0.05

0.04

0.04

70.04

w2(Sargan-test)

(987,183)***

(1003,184)***

(596,182)***

(1006,182)***

(654,147)***

Notes:Thetable

reportsin

parenthesist-statisticsbasedonthe‘difference

specificationinstrumentweightingmatrix’thatcanbecalculatedusingEviews

software

renderingconsistentstandard

errors

inapaneldata

setthatistransform

edusingfirstdifferencing.Theparameter

estimatesare

significantatthe

***1%

level,**5%

level

and*10%

level.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 15: Trade in Tourism Services Explaining Tourism Trade and The

410 C. Jensen and J. Zhang

Table

3.

Panel

data

results,comparativeadvantage.

Panel

method

GMM

(FD)

GMM

(FD)

GMM

(FD)

GMM

(FD)

GMM

(FD)

Dependentvariable:

TOURSH/POPSH

TOURSH/POPSH

TOURSH/POPSH

TOURSH/POPSH

TOURSH/POPSH

Explanatory

vars:

equation(1c.1)

equation(1c.2)

equation(1c.3)

equation(1c.4)

equation(1c.5)

LDV

0.47***

0.48***

0.47***

0.48***

0.50***

(14.22)

(14.63)

(13.82)

(14.37)

(13.88)

IPP

70.36

71.09***

70.36

0.10

71.30***

(70.73)

(72.60)

(70.71)

(0.20)

(72.47)

Log(G

DPCAP)

4.72***

–4.72***

4.40***

4.06***

(2.73)

(3.72)

(2.53)

(2.16)

Log(R

OOMSpmcap)

3.35***

3.68***

3.35***

3.93***

2.48***

(3.93)

(4.27)

(3.79)

(4.46)

(2.96)

Log(INNETpmcap)

70.88***

70.77***

70.88***

71.22***

70.66***

(76.78)

(76.11)

(76.77)

(76.88)

(74.27)

(Log(INNETpmcap))2

––

–0.08***

–(2.80)

OPEN

70.03***

70.02***

70.02***

70.03***

70.02*

(73.73)

(73.41)

(73.72)

(74.28)

(71.94)

UNSAFE

72.07***

72.02***

72.07***

72.08***

72.01***

(79.99)

(79.59)

(79.92)

(79.98)

(78.18)

DUMMY95

––

0.05

––

(0.02)

Countryeff

ects

FD

FD

FD

FD

FD

Tim

eeff

ects

Yes*

Yes**

Yes*

Yes*

Yes*

Number

ofobs.

2099(1987–2005)

2099(1987–2005)

2099(1987–2005)

2099(1987–2005)

1333(1995–2005)

AR(2)

70.06

70.07

70.06

70.06

70.22

w2(Sargan-test)

(517,183)***

(504,184)***

(517,182)***

(504,182)***

(521,147)***

Notes:Thetable

reportsin

parenthesist-statisticsbasedonthe‘difference

specificationinstrumentweightingmatrix’thatcanbecalculatedusingEviews

software

renderingconsistentstandard

errors

inapanel

data

setthatistransform

edusingfirstdifferencing.Theparameter

estimatesare

significantatthe

***1%

level,**5%

level

and*10%

level.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 16: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 411

Table

3.

Panel

data

results,comparativeadvantage.

Panel

method

GMM

(FD)

GMM

(FD)

GMM

(FD)

GMM

(FD)

GMM

(FD)

Dependentvariable:

TOURSH/POPSH

TOURSH/POPSH

TOURSH/POPSH

TOURSH/POPSH

TOURSH/POPSH

Explanatory

vars:

equation(1c.1)

equation(1c.2)

equation(1c.3)

equation(1c.4)

equation(1c.5)

LDV

0.47***

0.48***

0.47***

0.48***

0.50***

(14.22)

(14.63)

(13.82)

(14.37)

(13.88)

IPP

70.36

71.09***

70.36

0.10

71.30***

(70.73)

(72.60)

(70.71)

(0.20)

(72.47)

Log(G

DPCAP)

4.72***

–4.72***

4.40***

4.06***

(2.73)

(3.72)

(2.53)

(2.16)

Log(R

OOMSpmcap)

3.35***

3.68***

3.35***

3.93***

2.48***

(3.93)

(4.27)

(3.79)

(4.46)

(2.96)

Log(INNETpmcap)

70.88***

70.77***

70.88***

71.22***

70.66***

(76.78)

(76.11)

(76.77)

(76.88)

(74.27)

(Log(INNETpmcap))2

––

–0.08***

–(2.80)

OPEN

70.03***

70.02***

70.02***

70.03***

70.02*

(73.73)

(73.41)

(73.72)

(74.28)

(71.94)

UNSAFE

72.07***

72.02***

72.07***

72.08***

72.01***

(79.99)

(79.59)

(79.92)

(79.98)

(78.18)

DUMMY95

––

0.05

––

(0.02)

Countryeff

ects

FD

FD

FD

FD

FD

Tim

eeff

ects

Yes*

Yes**

Yes*

Yes*

Yes*

Number

ofobs.

2099(1987–2005)

2099(1987–2005)

2099(1987–2005)

2099(1987–2005)

1333(1995–2005)

AR(2)

70.06

70.07

70.06

70.06

70.22

w2(Sargan-test)

(517,183)***

(504,184)***

(517,182)***

(504,182)***

(521,147)***

Notes:Thetable

reportsin

parenthesist-statisticsbasedonthe‘difference

specificationinstrumentweightingmatrix’thatcanbecalculatedusingEviews

software

renderingconsistentstandard

errors

inapanel

data

setthatistransform

edusingfirstdifferencing.Theparameter

estimatesare

significantatthe

***1%

level,**5%

level

and*10%

level.

effects. Hence, results are only reported for the GMM model in the finalversion of the article.

The GMM model is as expected much superior compared to the othermodels. First of all, does it allow to analyse the full variation in the datawithout the side effects caused by inclusion of the fixed effect for the cross-sectional dimension. Furthermore, GMM avoids spurious relationshipsbecause of the level effects in the data associated with development. Finally,auto correlation is reduced through the inclusion of lagged dependentvariables. The Sargan test of over identifying restrictions passes at a veryhigh level of significance for all the models estimated. The lagged values ofthe dependent variable by a higher order rather than by its first lag are usedas instruments.

The main difference between columns 1–5 in Tables 2 and 3 concern inparticular the internet variable because it causes a set of particular problemsin this study. The internet variable is censored from the bottom due to itsintroduction only in the second half of the present dataset. Whereas a lot ofattention has been paid to censoring with respect to the dependent variablein applied econometrics, there is very little treatment of censoring problemswith respect to independent or explanatory variables.

Column 1 in Tables 2 and 3 shows the results for estimating Equations(1b) and (1c) in their general form. In column 2, we check for the robustnessof the IPP variable by excluding GDPCAP due to the fact that an element ofvariation in IPP can be seen to be mechanically covariant on the GDPCAPvariable (simply because the GDPCAP variable is the denominator of theIPP variable). However, the reported Pearson correlation coefficients in theAppendix also suggest that multicollinearity is not a major problem becauseof the mechanical co-variation induced. Columns 3–5 show differentvariations on column 1 owing to the censoring problem with the variableINNET. In column 3, a dummy is adopted that takes the value of 1 for allyears after the introduction of the internet (1995 and henceforth). Column 4instead takes a non-linear approach by inclusion of a squared term since itmay give greater weight to the observations in the data for which thevariable takes on positive values. Finally, in column 5 the more traditionalapproach of analysing instead on the reduced time series 1995–2006 isadopted. All three approaches render similar result as discussed furtherbelow.

Towards explaining absolute advantage or market shares, the pricevariable has the expected sign and is significant. For comparative advantage,the results for the price variable take the wrong sign and/or are notsignificant. Especially for the relative or population weighted market sharethere can be a real problem with the exogeneity assumption for the pricevariable. The lucrative islands that register a very high comparativeadvantage may rely strongly on tourism for their exports which also impliesthat their internal price levels are strongly affected by tourism activities.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 17: Trade in Tourism Services Explaining Tourism Trade and The

412 C. Jensen and J. Zhang

Hence, there may be an endogenous dynamic relationship betweencomparative advantage and IPP which gives the opposite sign from whatwas expected. The same results and problem may apply to some extent alsofor the openness variable. Countries with absolute advantage in tourism arealso typically more open, however, the effect is oppositely negative whenexplaining comparative advantage, again this may also be because of overtreliance on tourism among countries that register particularly high on thecomparative advantage indicator (see also Appendix Table A1).

All the other explanatory variables perform with the expected sign andare significant towards explaining both absolute and comparative advantagein tourism. It is generally confirmed that countries with higher levels ofincome generate more tourism activities. Countries that build tourisminfrastructure also attract more tourists. Finally, countries that have thenecessary institutions to provide safety also attract more tourists. Onlythe internet did not render the result we expected. Therefore, we adoptedthe above mentioned strategies to control for the censoring problem. In theinterpretation of these results, it may also be informative to consider theenclosed Figure 1 showing the general evolution in average market shareplotted against the average internet adoption rate.

The analysis confirms the suspicion that the insignificance of the internetvariable is related with the censoring problem. The inclusion of the dummyshows that after 1995 there is a general decline in the average market share

Figure 1. Mean market shares and internet users 1982–2006.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 18: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 413

Hence, there may be an endogenous dynamic relationship betweencomparative advantage and IPP which gives the opposite sign from whatwas expected. The same results and problem may apply to some extent alsofor the openness variable. Countries with absolute advantage in tourism arealso typically more open, however, the effect is oppositely negative whenexplaining comparative advantage, again this may also be because of overtreliance on tourism among countries that register particularly high on thecomparative advantage indicator (see also Appendix Table A1).

All the other explanatory variables perform with the expected sign andare significant towards explaining both absolute and comparative advantagein tourism. It is generally confirmed that countries with higher levels ofincome generate more tourism activities. Countries that build tourisminfrastructure also attract more tourists. Finally, countries that have thenecessary institutions to provide safety also attract more tourists. Onlythe internet did not render the result we expected. Therefore, we adoptedthe above mentioned strategies to control for the censoring problem. In theinterpretation of these results, it may also be informative to consider theenclosed Figure 1 showing the general evolution in average market shareplotted against the average internet adoption rate.

The analysis confirms the suspicion that the insignificance of the internetvariable is related with the censoring problem. The inclusion of the dummyshows that after 1995 there is a general decline in the average market share

Figure 1. Mean market shares and internet users 1982–2006.

and it is very likely related with the greater competition that the internetintroduces in this industry. Once the average decline in market share iscontrolled for the diffusion of the internet is a relevant explanatory factor oftourism exports. The results in column 4 are less straightforward tointerpret. It may not be that this result is obtained because of a non-lineareffect, but perhaps rather because the inclusion of a squared term givesgreater weight in the interpretation of this particular variable for the laterperiod where the variables take on positive values.

The plot of the mean market share (Figure 1) and the mean internetadoption rate also confirmed that mean market shares declined aroundthe date of the introduction of the internet (suggesting strongercompetition among destinations through entry) after which old or newmarket shares were gradually re-established sometime into the 1990s. Theeffect of the internet can also be related with the increase in internationaltrade and openness that is likely to generate more commercial or business-related tourism activities. A complicating factor is however also that inthe same period many ‘new’ countries enter the world economy such asthe European transition countries which could be a competing explana-tion for the decline in the mean market share. This factor is discussedfurther below when discussing the results by estimating the model forabsolute advantage within the more homogenous sub-samples of incomegroupings.

5.2. Results for the income sub-samples

Using the preferred specification from Table 2 (equation [1b]) the results forabsolute advantage are estimated for the sub-samples of high, medium andlow income groups using the World Bank’s income group rankings (as of2010). These results are reported with Table 4 in the first column for the fullsample and in subsequent columns by the income groups high, medium andlow, respectively. For the high income group, the model performs quitesimilar to that for the full sample, however, there are also some differences.The price variable takes on greater importance within the group of highincome countries. Oppositely is infrastructure less important, suggestingthat high income countries are over-capacitated in the industry. The internetis important, whereas neither openness nor safety is important, simplybecause the variation in these variables among the high income countries isvery small. The importance of these variables changes when moving to themiddle income group – e.g. here rooms and openness play a greater role andthe internet a lesser role when comparing countries’ market shares withinthat group. The model performs generally poorly towards understandingcompetition for market share within the group of low income countries,where even price differences do not matter. These changes in results arenecessarily driven by the fact that the full sample results are a composite of

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 19: Trade in Tourism Services Explaining Tourism Trade and The

414 C. Jensen and J. Zhang

Table

4.

Panel

data

results,absolute

advantagebyincomegroupings.

Panel

method

GMM

(FD)

GMM

(FD)

GMM

(FD)

GMM

(FD)

Dependentvariable:

TOURSHpm

TOURSHpm

TOURSHpm

TOURSHpm

equation(1b.3)

equation(1b.3)

equation(1b.3)

equation(1b.3)

Sub-sample:

FULL

SAMPLE

HIG

HIN

COME

MEDIU

MIN

COME

LOW

INCOME

Explanatory

vars:

LDV

0.81***

0.78***

0.81***

0.63***

(29.65)

(21.43)

(33.79)

(15.79)

IPP

1.37***

4.46***

0.41**

70.01

(2.61)

(3.76)

(2.12)

(71.35)

Log(G

DPCAP)

10.82***

13.96***

5.59***

0.19***

(4.70)

(3.75)

(5.24)

(3.25)

Log(R

OOMSpmcap)

3.22***

70.68

0.87**

0.01

(4.64)

(1.18)

(2.40)

(0.65)

Log(INNETpmcap)

0.08*

0.36***

70.15***

70.00

(1.74)

(3.62)

(73.65)

(70.28)

OPEN

0.03**

70.01

0.02***

0.00***

(2.00)

(70.36)

(3.52)

(4.82)

UNSAFE

70.94***

70.22

70.12

70.00

(72.67)

(70.42)

(71.06)

(70.49)

DUMMY95

711.89***

712.44***

71.31

0.02

(77.84)

(77.08)

(70.58)

(0.25)

Countryeff

ects

FD

FD

FD

FD

Tim

eeff

ects

Yes**

Yes**

Yes

Yes

Number

ofobs.

2100(1987–2005)

564(1987–2005)

1156(1987–2005)

380(1987–2005)

AR(2)

0.04

0.07

70.07

70.08

w2(Sargan-test)

(596,182)***

(407,182)***

(757,182)***

(367,177)***

Notes:Thetable

reportsin

parenthesist-statisticsbasedonthe’difference

specificationinstrumentweightingmatrix’thatcanbecalculatedusingEviews

software

renderingconsistentstandard

errors

inapanel

data

setthatistransform

edusingfirstdifferencing.Theparameter

estimatesare

significantatthe

***1%

level,**5%

level

and*10%

level.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 20: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 415

Table

4.

Panel

data

results,absolute

advantagebyincomegroupings.

Panel

method

GMM

(FD)

GMM

(FD)

GMM

(FD)

GMM

(FD)

Dependentvariable:

TOURSHpm

TOURSHpm

TOURSHpm

TOURSHpm

equation(1b.3)

equation(1b.3)

equation(1b.3)

equation(1b.3)

Sub-sample:

FULL

SAMPLE

HIG

HIN

COME

MEDIU

MIN

COME

LOW

INCOME

Explanatory

vars:

LDV

0.81***

0.78***

0.81***

0.63***

(29.65)

(21.43)

(33.79)

(15.79)

IPP

1.37***

4.46***

0.41**

70.01

(2.61)

(3.76)

(2.12)

(71.35)

Log(G

DPCAP)

10.82***

13.96***

5.59***

0.19***

(4.70)

(3.75)

(5.24)

(3.25)

Log(R

OOMSpmcap)

3.22***

70.68

0.87**

0.01

(4.64)

(1.18)

(2.40)

(0.65)

Log(INNETpmcap)

0.08*

0.36***

70.15***

70.00

(1.74)

(3.62)

(73.65)

(70.28)

OPEN

0.03**

70.01

0.02***

0.00***

(2.00)

(70.36)

(3.52)

(4.82)

UNSAFE

70.94***

70.22

70.12

70.00

(72.67)

(70.42)

(71.06)

(70.49)

DUMMY95

711.89***

712.44***

71.31

0.02

(77.84)

(77.08)

(70.58)

(0.25)

Countryeff

ects

FD

FD

FD

FD

Tim

eeff

ects

Yes**

Yes**

Yes

Yes

Number

ofobs.

2100(1987–2005)

564(1987–2005)

1156(1987–2005)

380(1987–2005)

AR(2)

0.04

0.07

70.07

70.08

w2(Sargan-test)

(596,182)***

(407,182)***

(757,182)***

(367,177)***

Notes:Thetable

reportsin

parenthesist-statisticsbasedonthe’difference

specificationinstrumentweightingmatrix’thatcanbecalculatedusingEviews

software

renderingconsistentstandard

errors

inapanel

data

setthatistransform

edusingfirstdifferencing.Theparameter

estimatesare

significantatthe

***1%

level,**5%

level

and*10%

level.

within and between income group differences, whereas the sub-sampleresults alone come about due to within group differences. The variable thatcaptures the capacity to provide safety UNSAFE is only relevant accordingto these results for the full sample.

6. Gains from GATS in tourism

6.1. Expected gains from liberalisation

The expected gains from liberalising trade in services are fundamentally nodifferent from the equivalent welfare gains of liberalising trade inagricultural and manufactured products (Hoekman and Primo Braga1997). The gains from liberalising services are considered to be substantialnot least due to the positive spill-overs it may have on trade in goods(Deardorf 2001) and also for diffusion of technology (Robinson, Wang, andMartin 2002). However, in services particular industries have been singledout which may take national priority such as health, education and thefinancial sector (Wade 2003; Wahba and Mohieldin 1998). Within tourism,traditional arguments against liberalisation are typically invalid (protection-ism makes little sense in its original form, even though some countriesconduct campaigns encouraging their citizens to vacate at home). The mainreason for this is that national and international tourism is considered to beweak substitutes. Generally, liberalisation must be considered a win-wingame since liberalisation is expected to generate more tourism overallthrough lower prices or higher quality of services (and not accounting forthe negative externalities that tourism may have on the environment becauseof the aviation industry).

The main argument against liberalisation of tourism service industrieshas centred on leakages from the income generated in tourism and itsmultiplier effects (Dwyer and Forsyth 1997). Wagner (1997) shows that ifmost of the tourism industries’ inputs (commodities and capital) in theregion are imported, it will generate only a small economic impact in theregion. At stake is therefore the cross-border mode of supply since foreigninvestment may significantly alter input–output structures in destinationcountries and thereby also streams of direct income, profits and last but notleast lead to an increased volatility in tourism flows (Gereffi 1999;Oppermann and Chon 1997).

No larger scale quantitative studies exist to estimate the cost and benefitsfrom liberalisation in tourism services and there has hardly been any focuson the price and income effect that liberalisation may have on tourismreceipts. From a demand-side perspective, it may be anticipated that sincetourism is a luxury good with an income elasticity above one it has a positiveterms-of-trade effect for the exporting countries. However, a positive terms-of-trade effect may be difficult to realise for those destinations that fail toimplement new technologies and thereby realise higher receipts through

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 21: Trade in Tourism Services Explaining Tourism Trade and The

416 C. Jensen and J. Zhang

differentiation strategies. Increasing competition through liberalisation forcountries on the lowest tier on the technology ladder may also imply a termsof trade loss.

6.2. Analysis of changes in tourism receipts over time

Tourismreceipt is ameasure similar toexport revenue since it estimateswhat thedestination country earns fromexporting tourism services.However, comparedto traditional exports there are several problems involved with correctlyestimating tourism service exports. One problem concerns what is includedunder tourism receipts and what is not, where for example earnings from airtransportation will typically not be included. A second related problemconcerns who earns what? Receipts are typically estimated as gross receipts nottaking into account, e.g. leakages due to imports, and the repatriation of profitsand wages earned by foreigners in the destination country. However, eventhough tourism consumption is highly composite, this is also a problem whenestimating export performance in general. The sensitivity of host countryearnings to a multiplicity of factors and the attention tourism researchers havepaid hereto, has paved the way for the elaboration of a highly specific set ofnational accounts. These accounts called tourism satellite accounts give a moreexact picture of tourism in the national economy (UNWTO 2007b; UNWTO2009). Unfortunately, such accounts are in their early development and hencenot useful towardmaking cross-country comparisons on a larger scale such as isthe objective in this study.

For the present purposes, it is therefore necessary to rely on grossreceipts and arrivals towards decomposing gross receipts into a price(receipts over arrivals) and an arrival component.

Revenue from tourism or receipts may be written as the product of priceand quantity. As price is used as receipts over number of tourists to capturethe average aggregate price, a tourist pays for a trip to the destinationcountry (which is sensitive to length of stay but data is not consistentlyavailable here for our panel) . Quantity is taken as the number of arrivals oftourists to the destination country. Hence, a change in gross revenue may bewritten as:

DReceipt ¼ DPrice � Arrivalþ DArrival � Priceþ DPrice � DArrival ð2aÞ

The interaction effect is typically quite small. Focusing on the maineffects, the relative change in net receipts may be expressed as the sum of therelative changes in price and quantity:

DReceiptReceipt

� DPricePrice

þ DArrivalArrival

ð2bÞ

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 22: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 417

differentiation strategies. Increasing competition through liberalisation forcountries on the lowest tier on the technology ladder may also imply a termsof trade loss.

6.2. Analysis of changes in tourism receipts over time

Tourismreceipt is ameasure similar toexport revenue since it estimateswhat thedestination country earns fromexporting tourism services.However, comparedto traditional exports there are several problems involved with correctlyestimating tourism service exports. One problem concerns what is includedunder tourism receipts and what is not, where for example earnings from airtransportation will typically not be included. A second related problemconcerns who earns what? Receipts are typically estimated as gross receipts nottaking into account, e.g. leakages due to imports, and the repatriation of profitsand wages earned by foreigners in the destination country. However, eventhough tourism consumption is highly composite, this is also a problem whenestimating export performance in general. The sensitivity of host countryearnings to a multiplicity of factors and the attention tourism researchers havepaid hereto, has paved the way for the elaboration of a highly specific set ofnational accounts. These accounts called tourism satellite accounts give a moreexact picture of tourism in the national economy (UNWTO 2007b; UNWTO2009). Unfortunately, such accounts are in their early development and hencenot useful towardmaking cross-country comparisons on a larger scale such as isthe objective in this study.

For the present purposes, it is therefore necessary to rely on grossreceipts and arrivals towards decomposing gross receipts into a price(receipts over arrivals) and an arrival component.

Revenue from tourism or receipts may be written as the product of priceand quantity. As price is used as receipts over number of tourists to capturethe average aggregate price, a tourist pays for a trip to the destinationcountry (which is sensitive to length of stay but data is not consistentlyavailable here for our panel) . Quantity is taken as the number of arrivals oftourists to the destination country. Hence, a change in gross revenue may bewritten as:

DReceipt ¼ DPrice � Arrivalþ DArrival � Priceþ DPrice � DArrival ð2aÞ

The interaction effect is typically quite small. Focusing on the maineffects, the relative change in net receipts may be expressed as the sum of therelative changes in price and quantity:

DReceiptReceipt

� DPricePrice

þ DArrivalArrival

ð2bÞ

In other words, the growth rate in tourism receipts can be accounted forby its main components which are growth rates in arrivals and price of thetravel experience, respectively.

6.3. Results of the receipts analysis

Ideally trade liberalisation in services could be measured by the cross-bordermode of supply such as the level of foreign direct investment in the serviceactivities of hotels and restaurants. Unfortunately, such level of detailedFDI data is not available for the large sample of countries covered in thestudy. Instead, trade liberalisation is measured as the political willingness ofcountries to liberalise their tourism service industries. This measure is also insome sense superior to other measures such as FDI. However, it suffers froma major problem in relation to panel data analysis due to its time invariantnature. Hence, it cannot be included in the previous model framework (e.g.asking whether absolute or comparative advantage increases with liberal-isation or whether the comparatively advantaged gain more etc.) and mustbe analysed separately in a reduced model such as the decomposition modeloffered above.

Based on this, a simple set of structural equations using the betweeneffect estimator and with controls for island effects and level of developmentare used. Ideally, a country specific effect should also be included, but sincethe COMMIT variable is perfectly collinear herewith it is not possible tocontrol for the country specific effect.5 This also leads to the choice of thebetween effect estimator whereby both the dependent and independentvariables becomes the average growth rate over the period of study. Similarto a cross-section study a between effect panel data model emphasises onlythe variation between countries and hence excludes the differences within thetime series of the same country.

Table 5 shows the distribution of the 190 countries and the sub-categoryof islands by commitment level. It is necessary to control for the islandeffect, because islands are generally less likely to be committed to free tradein tourism. We also control for level of development using GDPCAP,

Table 5. Number of countries by commitment level.

Commitment level No. of countries (hereof islands)

0 54 (24)1 20 (10)2 43 (5)3 61 (4)4 12 (3)

Total 190 (46)

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 23: Trade in Tourism Services Explaining Tourism Trade and The

418 C. Jensen and J. Zhang

Table

6.

Are

theregainsfrom

liberalisationofservices

tradein

tourism

?

Panel

method

BE

BE

BE

Dependentvariable:

gR

gA

gP

Constant

70.29***

70.31***

74.87

75.76*

70.36***

70.38***

(73.33)

(73.54)

(71.34)

(71.60)

(74.68)

(74.92)

COMMIT

i0.019*

–0.67

–0.019**

–(1.72)

(1.43)

(1.93)

COMDUM

i–

0.031

–2.32***

–0.037

(0.99)

(1.78)

(1.31)

ISLANDi

0.008

70.006

1.33

1.11

0.031

0.018

(0.24)

(70.18)

(0.93)

(0.81)

(1.01)

(0.61)

Log(G

DPCAP)i

0.03***

0.033***

0.34

0.39

0.030***

0.033***

(2.72)

(3.18)

(0.76)

(0.90)

(3.15)

(3.64)

Number

ofobs.

2194

2194

2731

2731

2098

2098

Countries

162

162

163

163

160

160

R2

0.09

0.08

0.02

0.03

0.12

0.10

Notes:Thetable

reportsin

parenthesist-statistics.Theparameter

estimatesare

significantatthe***1%

level,**5%

level

and*10%

level.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 24: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 419

Table

6.

Are

theregainsfrom

liberalisationofservices

tradein

tourism

?

Panel

method

BE

BE

BE

Dependentvariable:

gR

gA

gP

Constant

70.29***

70.31***

74.87

75.76*

70.36***

70.38***

(73.33)

(73.54)

(71.34)

(71.60)

(74.68)

(74.92)

COMMIT

i0.019*

–0.67

–0.019**

–(1.72)

(1.43)

(1.93)

COMDUM

i–

0.031

–2.32***

–0.037

(0.99)

(1.78)

(1.31)

ISLANDi

0.008

70.006

1.33

1.11

0.031

0.018

(0.24)

(70.18)

(0.93)

(0.81)

(1.01)

(0.61)

Log(G

DPCAP)i

0.03***

0.033***

0.34

0.39

0.030***

0.033***

(2.72)

(3.18)

(0.76)

(0.90)

(3.15)

(3.64)

Number

ofobs.

2194

2194

2731

2731

2098

2098

Countries

162

162

163

163

160

160

R2

0.09

0.08

0.02

0.03

0.12

0.10

Notes:Thetable

reportsin

parenthesist-statistics.Theparameter

estimatesare

significantatthe***1%

level,**5%

level

and*10%

level.

because developing countries as a group are less likely to be committed tofree trade. In the subsequent analysis, the commitment variable is eitheradopted as a 0–4 variable where the effect of commitment for tourism isexpected to be linearly increasing in the extent of free trade reform.Alternatively, the commitment variable is adopted as a simple dummyvariable COMDUM which takes the value of 1 if a country at all iscommitted to the reform of its trading system in the area of tourism.

The results of the receipt analysis are shown in Table 6. It is found thatthere is a significant impact of liberalisation on receipts (gR) through bothits main channels of arrival (gA) and average travel price (gP). Non-committers (as captured with the constant) experienced in the study periodsignificantly lower and negative growth rates in both of the maincomponents of receipts. The effect for arrivals is much larger when usingthe variable COMDUM; however, the price effect is equally significant andnegative for the group of non-committers using either of the commitmentvariables.

The World Bank has published a more consistent time series covering theperiod 1996–2005. It was checked if the results of the analysis changed whenusing this slightly shorter but more reliable series.

A note of caution with respect to the results obtained for the receiptanalysis concerns the problem of two-way causation. There may be a self-fulfilling prophecy bias in the observations made, e.g. countries that expectto gain more from liberalisation (e.g. large countries with significantestablished market share) may be more likely to liberalise their service traderelative to less developed countries and regions that expect to gain less fromliberalisation.

7. Discussion and conclusion

The objective of the study is to be one of the first to offer a global study oftourism exports on the supply side of the industry. Whereas many studieshave focused on the economic growth effects of tourism in a globalcomparative perspective, there are fewer global studies available of thefactors that explain exports taking outset in a traditional international tradeor supply oriented approach. In this way, our study complements theexisting literature by taking a comparative development perspective on thefactors that generate trade in tourism services.

Using the general methods of moment estimator as preferred econo-metric strategy in the study, we find that the identified study variables,decided by literature review and data availability, all are relevantexplanatory factors especially towards understanding established marketshares in the industry or absolute advantage in tourism activities. Variablessuch as price levels, income levels and tourism infrastructure play the rolethat we conventionally would expect in a supply side perspective.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 25: Trade in Tourism Services Explaining Tourism Trade and The

420 C. Jensen and J. Zhang

Destinations which offer a comparatively better purchasing power to theinternational currency holder (in USD terms) attract a higher absolutenumber of tourists compared with other destinations. This relationshipfor price and quantity was however not confirmed when trying to analysecomparative advantage. For comparative advantage, reverse causalitymay be a real problem due to high dependence on tourism activitiesamong island communities and small destinations where such activitiesspills over on local price levels (similar to the Samuelson–Balassa effect intraditional traded sectors). Countries with higher income levels and betterdeveloped tourism infrastructure were found to hold both absolute andcomparative advantage in the industry. Countries that indirectly cater totourists by offering better institutions in terms of their capacity toprovide perceptions of safety are those that hold both absolute andcomparative advantage in the industry.

The internet was found to play a different role. Its lacking significance atfirst led to a sample splitting exercise around the time when the diffusion ofthe technology was registered to have taken off (1995). The results suggestthat after 1995 the internet became important towards building marketshare in the industry (whereas established infrastructure or capacitydeclined), even though it has initially been less important for countriesthat hold comparative advantage.

A splitting of the sample by income groupings confirmed the results forthe full sample. The models perform best for the high and middle incomecountries whereas for low income countries it performs poorly. This likelyowes to the lack of dynamics in the data for low income countries.Especially, results for the variable UNSAFE was driven mainly by thebetween country variation in the full sample.

A secondary objective of the research is to investigate with available datawhether liberalisation under the WTO’s Uruguay Round in the area ofservices (GATS) has had any impact on an industry such as tourism. Asimple decomposition exercise of revenue growth into its main componentsof arrival and price growth is undertaken. These very preliminary resultsshow that liberalisation may have had a positive but also weaklydifferentiated effect on both the main components of export revenue intourism. The results come about mainly because non-liberalisers aresignificantly and negatively penalised both in terms of average negativegrowth rates in their arrival and gross prices obtained per tourist arrival.

Several improvements would be necessary to strengthen the conclusionsfrom the receipt analysis and its relationship with trade liberalisation. Itwould be in particular desirable to observe net receipts rather than grossreceipts for purposes of estimating the impact of liberalisation on nationalwelfare. Another problem related especially with the results for the priceeffect is that rather than measuring average price per travel a more validmeasure would be the average price per tourist day.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 26: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 421

Destinations which offer a comparatively better purchasing power to theinternational currency holder (in USD terms) attract a higher absolutenumber of tourists compared with other destinations. This relationshipfor price and quantity was however not confirmed when trying to analysecomparative advantage. For comparative advantage, reverse causalitymay be a real problem due to high dependence on tourism activitiesamong island communities and small destinations where such activitiesspills over on local price levels (similar to the Samuelson–Balassa effect intraditional traded sectors). Countries with higher income levels and betterdeveloped tourism infrastructure were found to hold both absolute andcomparative advantage in the industry. Countries that indirectly cater totourists by offering better institutions in terms of their capacity toprovide perceptions of safety are those that hold both absolute andcomparative advantage in the industry.

The internet was found to play a different role. Its lacking significance atfirst led to a sample splitting exercise around the time when the diffusion ofthe technology was registered to have taken off (1995). The results suggestthat after 1995 the internet became important towards building marketshare in the industry (whereas established infrastructure or capacitydeclined), even though it has initially been less important for countriesthat hold comparative advantage.

A splitting of the sample by income groupings confirmed the results forthe full sample. The models perform best for the high and middle incomecountries whereas for low income countries it performs poorly. This likelyowes to the lack of dynamics in the data for low income countries.Especially, results for the variable UNSAFE was driven mainly by thebetween country variation in the full sample.

A secondary objective of the research is to investigate with available datawhether liberalisation under the WTO’s Uruguay Round in the area ofservices (GATS) has had any impact on an industry such as tourism. Asimple decomposition exercise of revenue growth into its main componentsof arrival and price growth is undertaken. These very preliminary resultsshow that liberalisation may have had a positive but also weaklydifferentiated effect on both the main components of export revenue intourism. The results come about mainly because non-liberalisers aresignificantly and negatively penalised both in terms of average negativegrowth rates in their arrival and gross prices obtained per tourist arrival.

Several improvements would be necessary to strengthen the conclusionsfrom the receipt analysis and its relationship with trade liberalisation. Itwould be in particular desirable to observe net receipts rather than grossreceipts for purposes of estimating the impact of liberalisation on nationalwelfare. Another problem related especially with the results for the priceeffect is that rather than measuring average price per travel a more validmeasure would be the average price per tourist day.

As data availability for an industry such as tourism improves better,more detailed analysis will be possible.

Notes

1. Michael Kremer’s o-ring theory of economic development is based on the idea ofthe need for positive assortative matching in production – e.g. output ismaximised when similarly skilled people work together.

2. With neutral comparative advantage (toursh/popsh¼1), a person from onecountry is as likely to service a person from another country as the next. When acountry has comparative advantage its citizens are more likely to service a personfrom another country.

3. That is the price that can be arrived at by using the accounting identity that theincome from tourism divided with the number of arrivals must be equal to theaverage price paid for a travel (as we use towards analysing the impact of tradeliberalisation in Section 6).

4. If this was a true causal relationship, it would imply something similar to theSamuelson–Balassa effect, e.g. that the upward pressure on local prices andproductivity caused by tourism activities is transmitted to the prices andproductivities in other non-tourism sectors.

5. This has the consequence that the panel structure of the data in this case cannothelp to reduce the problem of unobserved variables. A third factor that is notobserved such as political regime may be the underlying explanatory factor, e.g.of both a low level of commitment towards liberalisation of services andresulting growth rates in receipts RECEIPT and its main components of touristarrivals ARRIVALS and average travel price PRICE.

References

Arellano, Manuel. 2003. Panel data econometrics. New York: Oxford UniversityPress.

Balaguer, J., and M. Cantavella-Jorda. 2002. Tourism as a long-run economicgrowth factor: The Spanish case. Applied Economics 34: 877–84.

Berrittella, M., A. Bigano, R. Roson, and R.S.J. Tol. 2006. A general equilibriumanalysis of climate change impacts on tourism. Tourism Management 27: 913–24.

Bond, S.R. 2002. Dynamic panel data models: A guide to micro data methods andpractice. CEMMAP Working Paper CWP09/02. The Institute for Fiscal Studies,Department of Economics, University College London.

Chesbrough, H., and J. Spohrer. 2006. A research manifesto for services science.Communication of the ACM 49, no. 7: 35–40.

Crouch, G. 1994. The study of international tourism demand: A review of findings.Journal of Travel Research 33, no. 1: 12–23.

Cruz, M.J.V. da, and C.F.C. Rolim. 2005. The determinants of international tourismand the restrictions to the inclusion of developing countries: A comparativeanalysis of South America, Africa and South Asia. Working paper presented atthe 45th Congress of the European Regional Science Association.

Deardorf, A.V. 2001. International provision of trade services, trade, andfragmentation. Review of International Economics 9, no. 2: 233–48.

Dritsakis, N. 2004. Tourism as a long-run economic growth factor: An empiricalinvestigation for Greece using causality analysis. Tourism Economics 10, no. 3:305–16.

Dwyer, L., and P. Forsyth. 1997. Measuring the benefits and yield from foreigntourism. International Journal of Social Economics 24: 1–3.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 27: Trade in Tourism Services Explaining Tourism Trade and The

422 C. Jensen and J. Zhang

Eilat, Y., and L. Einav. 2004. Determinants of international tourism: A three-dimensional panel data analysis. Applied Economics 36: 1315–27.

Eugenio-Martin, J., Moralse, N., and Scarpa, R. 2004. Tourism and economicgrowth in Latin American countries: A panel data approach. Nota de lavoro 26http://papers.ssrn.com

Freund, C., and D. Weinhold. 2002. The Internet and international trade in services.The American Economic Review 92, no. 2: 236–40.

Hazari, R., and M. Sgro. 1995. Tourism and growth in a dynamic model of trade.The Journal of International Trade and Economic Development 2: 243–56.

Gereffi, G. 1999. International trade and industrial upgrading in the apparelcommodity chain. Journal of International Economics 48, no. 1: 37–70.

Griliches, Zvi. 1992. Output measurement in the service sector. Chicago, IL:University of Chicago Press.

Hoekman, B., and C.A. Primo Braga. 1997. Protection and trade in services: Asurvey. Open Economies Review 8, no. 3: 285–308.

Kimura, F., and H.-H. Lee. 2006. The gravity equation in international trade inservices. Review of World Economics 142: 1.

Kremer, Michael. 1993. The O-ring theory of economic development. QuarterlyJournal of Economics 108: 3.

Lee, C.C., and C.P. Chang. 2008. Tourism development and economic growth: Acloser look at panels. Tourism Management 29: 180–92.

Li, D., F. Moshirian, and A.-B. Sim. 2003. The determinants of intra-industry trade in insurance services. Journal of Risk & Insurance 70, no. 2:269–87.

Lim, C. 1997. Review of international tourism-demand models. Annals of TourismResearch 24, no. 835–49

Mirza, D., and G. Nicoletti. 2004. What is so special about trade in services?Research Paper 2004/02. Leverhulme Centre for Research on Globalisation andEconomic Policy, University of Nottingham.

Oh, C. 2005. The contribution of tourism development to economic growth in theKorean economy. Tourism Management 26: 39–44.

Oppermann, Martin, and Kye-Sung Chon. 1997. Tourism in developing countries.London: International Thomson Business Press.

Prideaux, Bruce. 2005. Factors affecting bilateral tourism flows. Annals of TourismResearch 32, no. 3: 780–801.

Robinson, S., Z. Wang, and W. Martin. 2002. Capturing the implications of servicesfor trade liberalization. Economic Systems Research 14, no. 1: 3–33.

Sequeira, T.N., and P.M. Nunes. 2008. Does tourism influence economic growth? Adynamic panel data approach. Applied Economics 40: 2431–41.

Smith, R.D. 2004. Foreign direct investment and trade in health services: A review ofthe literature. Social Science and Medicine 59, no. 11: 2313–23.

Song, H., and G. Li. 2008. Tourism demand modelling and forecasting – A review ofrecent research. Tourism Management 29: 203–20.

UNWTO. 2007a. Yearbook of tourism statistics. Madrid: United Nation’s WorldTourism Organisation.

UNWTO. 2007b. 2008 international recommendations for tourism statistics. Madrid:United Nation’s World Tourism Organisation.

UNWTO. 2009. TSA data around the world – World summary Madrid: UNWTODepartment of Statistics and Tourism Satellite Accounts.

Wade, R.H. 2003. What strategies are viable for developing countries today? TheWorld Trade Organization and the shrinking of ‘development space’. Review ofInternational Political Economy 10, no. 4: 621–44.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 28: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 423

Eilat, Y., and L. Einav. 2004. Determinants of international tourism: A three-dimensional panel data analysis. Applied Economics 36: 1315–27.

Eugenio-Martin, J., Moralse, N., and Scarpa, R. 2004. Tourism and economicgrowth in Latin American countries: A panel data approach. Nota de lavoro 26http://papers.ssrn.com

Freund, C., and D. Weinhold. 2002. The Internet and international trade in services.The American Economic Review 92, no. 2: 236–40.

Hazari, R., and M. Sgro. 1995. Tourism and growth in a dynamic model of trade.The Journal of International Trade and Economic Development 2: 243–56.

Gereffi, G. 1999. International trade and industrial upgrading in the apparelcommodity chain. Journal of International Economics 48, no. 1: 37–70.

Griliches, Zvi. 1992. Output measurement in the service sector. Chicago, IL:University of Chicago Press.

Hoekman, B., and C.A. Primo Braga. 1997. Protection and trade in services: Asurvey. Open Economies Review 8, no. 3: 285–308.

Kimura, F., and H.-H. Lee. 2006. The gravity equation in international trade inservices. Review of World Economics 142: 1.

Kremer, Michael. 1993. The O-ring theory of economic development. QuarterlyJournal of Economics 108: 3.

Lee, C.C., and C.P. Chang. 2008. Tourism development and economic growth: Acloser look at panels. Tourism Management 29: 180–92.

Li, D., F. Moshirian, and A.-B. Sim. 2003. The determinants of intra-industry trade in insurance services. Journal of Risk & Insurance 70, no. 2:269–87.

Lim, C. 1997. Review of international tourism-demand models. Annals of TourismResearch 24, no. 835–49

Mirza, D., and G. Nicoletti. 2004. What is so special about trade in services?Research Paper 2004/02. Leverhulme Centre for Research on Globalisation andEconomic Policy, University of Nottingham.

Oh, C. 2005. The contribution of tourism development to economic growth in theKorean economy. Tourism Management 26: 39–44.

Oppermann, Martin, and Kye-Sung Chon. 1997. Tourism in developing countries.London: International Thomson Business Press.

Prideaux, Bruce. 2005. Factors affecting bilateral tourism flows. Annals of TourismResearch 32, no. 3: 780–801.

Robinson, S., Z. Wang, and W. Martin. 2002. Capturing the implications of servicesfor trade liberalization. Economic Systems Research 14, no. 1: 3–33.

Sequeira, T.N., and P.M. Nunes. 2008. Does tourism influence economic growth? Adynamic panel data approach. Applied Economics 40: 2431–41.

Smith, R.D. 2004. Foreign direct investment and trade in health services: A review ofthe literature. Social Science and Medicine 59, no. 11: 2313–23.

Song, H., and G. Li. 2008. Tourism demand modelling and forecasting – A review ofrecent research. Tourism Management 29: 203–20.

UNWTO. 2007a. Yearbook of tourism statistics. Madrid: United Nation’s WorldTourism Organisation.

UNWTO. 2007b. 2008 international recommendations for tourism statistics. Madrid:United Nation’s World Tourism Organisation.

UNWTO. 2009. TSA data around the world – World summary Madrid: UNWTODepartment of Statistics and Tourism Satellite Accounts.

Wade, R.H. 2003. What strategies are viable for developing countries today? TheWorld Trade Organization and the shrinking of ‘development space’. Review ofInternational Political Economy 10, no. 4: 621–44.

Wagner, J.E. 1997. Estimating the economic impacts of tourism. Annals of TourismResearch 24, no. 3: 592–608.

Wahba, J., and M. Mohieldin. 1998. Liberalizing trade in financial service: TheUruguay round and the Arab countries. World Development 26, no. 7: 1331–48.

Webster, A., and P. Hardwick. 2005. International trade in financial services. ServiceIndustries Journal 25, no. 6: 721–46.

Appendix

Table A1. Comparison of dependent variables in 2005.

Country

TOURSH/POPSH(Number of tourists

served per yearper capita)

TOURSH (Thecountry’s shareof the global

tourism industry)

ARRIVAL(Absolute

number of annualtourist arrivalsin thousands)

Macau 152.5973 0.011173 9014.0Aruba 58.52968 0.000909 733.0Guam 58.34333 0.001522 1228.0N Mariana Islands 49.66917 0.000617 498.0Bahrain 43.24968 0.004851 3914.0US Virgin Islands 42.36229 0.000713 575.0Bahamas 39.83457 0.001993 1608.0Palau 34.26695 0.000107 86.0Bermuda 34.01611 0.000335 270.0Cayman Islands 29.89991 0.000208 168.0Hungary 28.71980 0.044834 36172.0Cyprus 26.10450 0.003061 2470.0Iceland 23.50718 0.001080 871.0Malta 23.24269 0.001451 1171.0St Kitts-Nev 21.19023 0.000157 127.0Austria 19.40822 0.024730 19952.0Hong Kong 17.36563 0.018311 14773.0Luxembourg 16.01046 0.001132 913.0Saint Lucia 15.45492 0.000394 318.0Croatia 15.26132 0.010495 8467.0Barbados 15.03386 0.000679 548.0San Marino 14.20019 6.20E-05 50.0Ireland 14.12067 0.009089 7333.0Singapore 13.29248 0.008775 7080.0Seychelles 12.46259 0.000160 129.0Liechtenstein 11.52360 6.20E-05 50.0Estonia 11.30445 0.002355 1900.0Maldives 10.71300 0.000490 395.0Spain 10.31864 0.069303 55914.0Greece 10.29675 0.017695 14276.0

(continued)

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 29: Trade in Tourism Services Explaining Tourism Trade and The

424 C. Jensen and J. Zhang

Table A1. (Continued).

Country

TOURSH/POPSH(Number of tourists

served per yearper capita)

TOURSH (Thecountry’s shareof the global

tourism industry)

ARRIVAL(Absolute

number of annualtourist arrivalsin thousands)

France 9.999256 0.094201 76001.0Qatar 9.183946 0.001132 913.0Dominica 8.787548 9.79E-05 79.0Switzerland 7.784805 0.008960 7229.0PuertoRico 7.546117 0.004569 3686.0Grenada 7.444897 0.000123 99.0Botswana 7.306845 0.002076 1675.0Denmark 6.948669 0.005824 4699.0Norway 6.684914 0.004783 3859.0French Polynesia 6.516603 0.000258 208.0Belize 6.504833 0.000294 237.0St Vincent G 6.453535 0.000119 96.0Slovenia 6.225367 0.001927 1555.0Swaziland 5.941176 0.001040 839.0Fiji 5.319629 0.000682 550.0Belgium 5.156779 0.008363 6747.0Malaysia 5.129785 0.020366 16431.0Tunisia 5.093309 0.007905 6378.0Bulgaria 5.005048 0.005995 4837.0Italy 4.989658 0.045257 36513.0Czech Republic 4.958370 0.007853 6336.0Netherlands 4.913351 0.012410 10012.0Mauritius 4.902281 0.000943 761.0Finland 4.793648 0.003892 3140.0Lithuania 4.691389 0.002479 2000.0Canada 4.652363 0.023265 18770.0New Zealand 4.583823 0.002933 2366.0Jamaica 4.462297 0.001833 1479.0Samoa 4.443462 0.000126 102.0Jordan 4.420696 0.003702 2987.0Uruguay 4.380312 0.002241 1808.0UK 3.985404 0.037147 29970.0Latvia 3.885215 0.001383 1116.0New Caledonia 3.449766 0.000125 101.0Tonga 3.385372 5.21E-05 42.0Poland 3.189674 0.018840 15200.0Cape Verde 3.128929 0.000245 198.0Costa Rica 3.107520 0.002081 1679.0Namibia 3.085111 0.000964 778.0Suriname 2.832078 0.000198 160.0Trinidad Tbg 2.801285 0.000574 463.0Saudi Arabia 2.784185 0.009962 8037.0Vanuatu 2.305619 7.68E-05 62.0

(continued)

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 30: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 425

Table A1. (Continued).

Country

TOURSH/POPSH(Number of tourists

served per yearper capita)

TOURSH (Thecountry’s shareof the global

tourism industry)

ARRIVAL(Absolute

number of annualtourist arrivalsin thousands)

France 9.999256 0.094201 76001.0Qatar 9.183946 0.001132 913.0Dominica 8.787548 9.79E-05 79.0Switzerland 7.784805 0.008960 7229.0PuertoRico 7.546117 0.004569 3686.0Grenada 7.444897 0.000123 99.0Botswana 7.306845 0.002076 1675.0Denmark 6.948669 0.005824 4699.0Norway 6.684914 0.004783 3859.0French Polynesia 6.516603 0.000258 208.0Belize 6.504833 0.000294 237.0St Vincent G 6.453535 0.000119 96.0Slovenia 6.225367 0.001927 1555.0Swaziland 5.941176 0.001040 839.0Fiji 5.319629 0.000682 550.0Belgium 5.156779 0.008363 6747.0Malaysia 5.129785 0.020366 16431.0Tunisia 5.093309 0.007905 6378.0Bulgaria 5.005048 0.005995 4837.0Italy 4.989658 0.045257 36513.0Czech Republic 4.958370 0.007853 6336.0Netherlands 4.913351 0.012410 10012.0Mauritius 4.902281 0.000943 761.0Finland 4.793648 0.003892 3140.0Lithuania 4.691389 0.002479 2000.0Canada 4.652363 0.023265 18770.0New Zealand 4.583823 0.002933 2366.0Jamaica 4.462297 0.001833 1479.0Samoa 4.443462 0.000126 102.0Jordan 4.420696 0.003702 2987.0Uruguay 4.380312 0.002241 1808.0UK 3.985404 0.037147 29970.0Latvia 3.885215 0.001383 1116.0New Caledonia 3.449766 0.000125 101.0Tonga 3.385372 5.21E-05 42.0Poland 3.189674 0.018840 15200.0Cape Verde 3.128929 0.000245 198.0Costa Rica 3.107520 0.002081 1679.0Namibia 3.085111 0.000964 778.0Suriname 2.832078 0.000198 160.0Trinidad Tbg 2.801285 0.000574 463.0Saudi Arabia 2.784185 0.009962 8037.0Vanuatu 2.305619 7.68E-05 62.0

(continued)

Table A1. (Continued).

Country

TOURSH/POPSH(Number of tourists

served per yearper capita)

TOURSH (Thecountry’s shareof the global

tourism industry)

ARRIVAL(Absolute

number of annualtourist arrivalsin thousands)

Lebanon 2.276425 0.001413 1140.0Turkey 2.253029 0.025128 20273.0Slovakia 2.252365 0.001878 1515.0Israel 2.201303 0.002359 1903.0Romania 2.161562 0.007237 5839.0Germany 2.087943 0.026648 21500.0Australia 1.983398 0.006262 5052.0Panama 1.739826 0.000870 702.0Mexico 1.702557 0.027163 21915.0Kazakhstan 1.661841 0.003896 3143.0Cuba 1.608196 0.002802 2261.0Morocco 1.552483 0.007242 5843.0Thailand 1.470392 0.014337 11567.0Syria 1.427657 0.004175 3368.0El Salvador 1.385990 0.001430 1154.0Micronesia 1.382627 2.35E-05 19.0United States 1.329096 0.060989 49206.0Guyana 1.267177 0.000145 117.0South Africa 1.258575 0.009134 7369.0Lesotho 1.229134 0.000377 304.0Marshall Islands 1.139317 1.12E-05 9.0Azerbaijan 1.123290 0.001459 1177.0Russia 1.115596 0.024715 19940.0Mongolia 1.059910 0.000419 338.0Nicaragua 1.043899 0.000882 712.0Georgia 1.002588 0.000694 560.0Korea South 0.998830 0.007465 6023.0Chile 0.996253 0.002512 2027.0Zimbabwe 0.951691 0.001932 1559.0Lao 0.950224 0.000833 672.0Egypt 0.906322 0.010218 8244.0Armenia 0.846629 0.000395 319.0Guatemala 0.829275 0.001631 1316.0Cambodia 0.816069 0.001763 1422.0Argentina 0.805083 0.004828 3895.0Honduras 0.788690 0.000834 673.0Sao Tome Pm 0.577230 1.36E-05 11.0Gambia 0.549766 0.000138 111.0Ecuador 0.527959 0.001067 861.0Senegal 0.523251 0.000953 769.0Kyrgyzstan 0.490484 0.000390 315.0Zambia 0.466789 0.000829 669.0Paraguay 0.462993 0.000423 341.0

(continued)

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 31: Trade in Tourism Services Explaining Tourism Trade and The

426 C. Jensen and J. Zhang

Table A1. (Continued).

Country

TOURSH/POPSH(Number of tourists

served per yearper capita)

TOURSH (Thecountry’s shareof the global

tourism industry)

ARRIVAL(Absolute

number of annualtourist arrivalsin thousands)

Bosnia Herzg 0.459616 0.000269 217.0Bolivia 0.439608 0.000625 504.0Peru 0.436354 0.001842 1486.0Japan 0.421716 0.008339 6728.0Algeria 0.351762 0.001789 1443.0Kenya 0.345563 0.001904 1536.0Djibouti 0.298763 3.72E-05 30.0China 0.287381 0.058018 46809.0Comoros 0.266746 2.48E-05 20.0Malawi 0.265226 0.000543 438.0Philippines 0.248413 0.003251 2623.0Kiribati 0.242694 3.72E-06 3.0Brazil 0.225396 0.006517 5258.0Sri Lanka 0.223555 0.000680 549.0Venezuela 0.212751 0.000875 706.0Indonesia 0.181633 0.006200 5002.0Bhutan 0.176016 1.74E-05 14.0Solomon Island 0.169530 1.24E-05 10.0Colombia 0.166252 0.001156 933.0Benin 0.166021 0.000218 176.0Ghana 0.152466 0.000532 429.0Burundi 0.150827 0.000183 148.0Eritrea 0.146848 0.000103 83.0Burkina Faso 0.140826 0.000304 245.0Congo 0.135336 7.56E-05 61.0Uganda 0.129483 0.000580 468.0Yemen 0.127561 0.000416 336.0Tanzania 0.122804 0.000731 590.0Madagascar 0.119000 0.000343 277.0Albania 0.116817 5.70E-05 46.0Nepal 0.110850 0.000465 375.0Angola 0.104495 0.000260 210.0Togo 0.103986 0.000100 81.0Mali 0.098636 0.000177 143.0Haiti 0.096490 0.000139 112.0Papua New Guinea 0.091045 8.55E-05 69.0Cameroon 0.079211 0.000218 176.0Belarus 0.074554 0.000113 91.0Sierra Leone 0.057346 4.96E-05 40.0Nigeria 0.057224 0.001252 1010.0Sudan 0.053393 0.000305 246.0Pakistan 0.041029 0.000989 798.0Guinea 0.040033 5.58E-05 45.0

(continued)

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 32: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 427

Table A1. (Continued).

Country

TOURSH/POPSH(Number of tourists

served per yearper capita)

TOURSH (Thecountry’s shareof the global

tourism industry)

ARRIVAL(Absolute

number of annualtourist arrivalsin thousands)

Bosnia Herzg 0.459616 0.000269 217.0Bolivia 0.439608 0.000625 504.0Peru 0.436354 0.001842 1486.0Japan 0.421716 0.008339 6728.0Algeria 0.351762 0.001789 1443.0Kenya 0.345563 0.001904 1536.0Djibouti 0.298763 3.72E-05 30.0China 0.287381 0.058018 46809.0Comoros 0.266746 2.48E-05 20.0Malawi 0.265226 0.000543 438.0Philippines 0.248413 0.003251 2623.0Kiribati 0.242694 3.72E-06 3.0Brazil 0.225396 0.006517 5258.0Sri Lanka 0.223555 0.000680 549.0Venezuela 0.212751 0.000875 706.0Indonesia 0.181633 0.006200 5002.0Bhutan 0.176016 1.74E-05 14.0Solomon Island 0.169530 1.24E-05 10.0Colombia 0.166252 0.001156 933.0Benin 0.166021 0.000218 176.0Ghana 0.152466 0.000532 429.0Burundi 0.150827 0.000183 148.0Eritrea 0.146848 0.000103 83.0Burkina Faso 0.140826 0.000304 245.0Congo 0.135336 7.56E-05 61.0Uganda 0.129483 0.000580 468.0Yemen 0.127561 0.000416 336.0Tanzania 0.122804 0.000731 590.0Madagascar 0.119000 0.000343 277.0Albania 0.116817 5.70E-05 46.0Nepal 0.110850 0.000465 375.0Angola 0.104495 0.000260 210.0Togo 0.103986 0.000100 81.0Mali 0.098636 0.000177 143.0Haiti 0.096490 0.000139 112.0Papua New Guinea 0.091045 8.55E-05 69.0Cameroon 0.079211 0.000218 176.0Belarus 0.074554 0.000113 91.0Sierra Leone 0.057346 4.96E-05 40.0Nigeria 0.057224 0.001252 1010.0Sudan 0.053393 0.000305 246.0Pakistan 0.041029 0.000989 798.0Guinea 0.040033 5.58E-05 45.0

(continued)

Table A1. (Continued).

Country

TOURSH/POPSH(Number of tourists

served per yearper capita)

TOURSH (Thecountry’s shareof the global

tourism industry)

ARRIVAL(Absolute

number of annualtourist arrivalsin thousands)

Myanmar 0.038736 0.000288 232.0Niger 0.038039 7.81E-05 63.0India 0.028675 0.004857 3919.0Ethiopia 0.024184 0.000281 227.0Chad 0.022892 3.59E-05 29.0Turkmenistan 0.019884 1.49E-05 12.0Bangladesh 0.010868 0.000258 208.0Afghanistan NA NA NAAmerican Samoa NA NA NAAntigua NA NA NABrunei NA NA NACent. Afr. Rep. NA 1.49E-05 12.0Coted’Ivoire NA NA NADominican Rep. NA 0.004575 3691.0Gabon NA NA NAIran NA NA NAIraq NA NA NAKoreaNorth NA NA NAKuwait NA NA NALibya NA NA NAMauritania NA NA NAOman NA NA NAPortugal NA NA NARwanda NA NA NASomalia NA NA NASweden NA NA NATajikistan NA NA NAUkraine NA NA NAUzbekistan NA NA NAVietnam NA 0.004298 3468.0

Source: UN World Tourism Organisation, Madrid and the World Bank, Washington D.C.

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 33: Trade in Tourism Services Explaining Tourism Trade and The

428 C. Jensen and J. Zhang

Table

A2.Granger

causality

test

forrelationship

betweenIPPandTOURSH.

H0

Sample

Obs

F-statistic

TOURSHpm

does

notGranger

Cause

IPP

Full

2491

10.72***

IPPdoes

notGranger

Cause

TOURSH1000

Full

2491

1.95

TOURSHpm

does

notGranger

Cause

IPP

DevelopingAfrica

615

3.67**

IPPdoes

notGranger

Cause

TOURSH1000

DevelopingAfrica

615

0.69

TOURSHpm

does

notGranger

Cause

IPP

DevelopingAmericas

531

1,41

IPPdoes

notGranger

Cause

TOURSH1000

DevelopingAmericas

531

0.78

TOURSHpm

does

notGranger

Cause

IPP

DevelopingAsia

381

0.11

IPPdoes

notGranger

Cause

TOURSH1000

DevelopingAsia

381

1.08

TOURSHpm

does

notGranger

Cause

IPP

OECD

countries

461

1.62

IPPdoes

notGranger

Cause

TOURSH1000

OECD

countries

461

3.51**

TOURSHpm

does

notGranger

Cause

IPP

Europeantransitioncountries

247

2.38*

IPPdoes

notGranger

Cause

TOURSH1000

Europeantransitioncountries

247

0.29

TOURSHpm

does

notGranger

Cause

IPP

DevelopingMiddle

East

256

0.21

IPPdoes

notGranger

Cause

TOURSH1000

DevelopingMiddle

East

256

0.27

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014

Page 34: Trade in Tourism Services Explaining Tourism Trade and The

The Journal of International Trade & Economic Development 429

Table

A2.Granger

causality

test

forrelationship

betweenIPPandTOURSH.

H0

Sample

Obs

F-statistic

TOURSHpm

does

notGranger

Cause

IPP

Full

2491

10.72***

IPPdoes

notGranger

Cause

TOURSH1000

Full

2491

1.95

TOURSHpm

does

notGranger

Cause

IPP

DevelopingAfrica

615

3.67**

IPPdoes

notGranger

Cause

TOURSH1000

DevelopingAfrica

615

0.69

TOURSHpm

does

notGranger

Cause

IPP

DevelopingAmericas

531

1,41

IPPdoes

notGranger

Cause

TOURSH1000

DevelopingAmericas

531

0.78

TOURSHpm

does

notGranger

Cause

IPP

DevelopingAsia

381

0.11

IPPdoes

notGranger

Cause

TOURSH1000

DevelopingAsia

381

1.08

TOURSHpm

does

notGranger

Cause

IPP

OECD

countries

461

1.62

IPPdoes

notGranger

Cause

TOURSH1000

OECD

countries

461

3.51**

TOURSHpm

does

notGranger

Cause

IPP

Europeantransitioncountries

247

2.38*

IPPdoes

notGranger

Cause

TOURSH1000

Europeantransitioncountries

247

0.29

TOURSHpm

does

notGranger

Cause

IPP

DevelopingMiddle

East

256

0.21

IPPdoes

notGranger

Cause

TOURSH1000

DevelopingMiddle

East

256

0.27

Table

A3.Pearsoncorrelationcoeffi

cients.

ARRIV

ALSCOMMIT

LGDPCAP

LIN

NET

pmcap

LIN

NET

ISLAND

OPEN

POP

POPSH

IPP

PRIC

EPRIC

EWBRECEIPTSRECEIPTSWB

LROOMS

LROOM

Spmcap

UNSAFE

TOURSHpm

TOURSHPOPSH

ARRIV

ALS

1.000000

0.257155

0.411852

0.339240

0.2815897

0.18283470.118415

0.250947

0.2414187

0.29408970.0179027

0.025996

0.853524

0.791930

0.586801

0.2424707

0.247783

0.963371

0.064427

COMMIT

0.257155

1.000000

0.313823

0.182789

0.1488097

0.34768570.084989

0.062438

0.0627857

0.235824

0.012638

0.044087

0.267359

0.277683

0.415187

0.0209927

0.253417

0.264594

70.119805

LogGDPCAP

0.411852

0.313823

1.000000

0.313076

0.454671

0.035630

0.2820857

0.07722570.0828027

0.494430

0.154091

0.172477

0.393358

0.378137

0.529642

0.7585697

0.493548

0.410456

0.417061

Log(INNETpmcap)

0.339240

0.182789

0.313076

1.000000

0.8565847

0.110255

0.070099

0.121597

0.0923087

0.049781

0.069414

0.077729

0.355640

0.410185

0.453095

0.1935707

0.255093

0.247669

70.015641

Log(INNET)

0.281589

0.148809

0.454671

0.856584

1.000000

0.023849

0.205596

0.00556170.0143517

0.221987

0.093107

0.143465

0.316647

0.311878

0.344487

0.4096217

0.337068

0.185975

0.150489

ISLAND

70.182834

70.347685

0.0356307

0.110255

0.023849

1.000000

0.2932957

0.13489070.1355487

0.171256

0.049227

0.0462357

0.155708

70.149156

70.314621

0.4560567

0.232444

70.182764

0.415970

OPEN

70.118415

70.084989

0.282085

0.070099

0.205596

0.293295

1.0000007

0.21372670.2195337

0.074968

0.0274607

0.0245307

0.122811

70.122915

70.214202

0.3348797

0.122897

70.141448

0.373886

POP

0.250947

0.0624387

0.077225

0.121597

0.0055617

0.13489070.213726

1.000000

0.994076

0.064663

0.010247

0.003397

0.241742

0.260989

0.3010657

0.171908

0.047533

0.228788

70.099112

POPSH

0.241418

0.0627857

0.082802

0.09230870.0143517

0.13554870.219533

0.994076

1.000000

0.057434

0.009426

0.002632

0.230679

0.257891

0.2974027

0.174213

0.048179

0.226027

70.099760

IPP

70.294089

70.2358247

0.4944307

0.04978170.2219877

0.17125670.074968

0.064663

0.057434

1.00000070.1091417

0.1228777

0.312088

70.323027

70.2695857

0.601765

0.415185

70.306021

70.292251

PRIC

E7

0.017902

0.012638

0.154091

0.069414

0.093107

0.049227

0.027460

0.010247

0.0094267

0.109141

1.000000

0.949835

0.046197

0.065832

0.060446

0.1767867

0.051322

70.024273

0.041238

PRIC

EWB

70.025996

0.044087

0.172477

0.077729

0.143465

0.04623570.024530

0.003397

0.0026327

0.122877

0.949835

1.000000

0.085638

0.105681

0.022874

0.1219927

0.074352

70.027036

0.018607

RECEIP

TS

0.853524

0.267359

0.393358

0.355640

0.3166477

0.15570870.122811

0.241742

0.2306797

0.312088

0.046197

0.085638

1.000000

0.987659

0.574279

0.2689467

0.253332

0.794466

0.036628

RECEIP

TSWB

0.791930

0.277683

0.378137

0.410185

0.3118787

0.14915670.122915

0.260989

0.2578917

0.323027

0.065832

0.105681

0.987659

1.000000

0.570625

0.2477907

0.226903

0.784217

0.022124

Log(R

OOMS)

0.586801

0.415187

0.529642

0.453095

0.3444877

0.31462170.214202

0.301065

0.2974027

0.269585

0.060446

0.022874

0.574279

0.570625

1.000000

0.2990497

0.258294

0.582241

0.002974

Log(R

OOMSpmcap)

0.242470

0.020992

0.758569

0.193570

0.409621

0.456056

0.3348797

0.17190870.1742137

0.601765

0.176786

0.121992

0.268946

0.247790

0.299049

1.0000007

0.543831

0.246244

0.640653

UNSAFE

70.247783

70.2534177

0.4935487

0.25509370.3370687

0.23244470.122897

0.047533

0.048179

0.41518570.0513227

0.0743527

0.253332

70.226903

70.2582947

0.543831

1.000000

70.254906

70.287211

TOURSHpm

0.963371

0.264594

0.410456

0.247669

0.1859757

0.18276470.141448

0.228788

0.2260277

0.30602170.0242737

0.027036

0.794466

0.784217

0.582241

0.2462447

0.254906

1.000000

0.068315

TOURSHPOPSH

0.064427

70.119805

0.4170617

0.015641

0.150489

0.415970

0.3738867

0.09911270.0997607

0.292251

0.041238

0.018607

0.036628

0.022124

0.002974

0.6406537

0.287211

0.068315

1.000000

Dow

nloa

ded

by [

INA

SP -

Pak

ista

n (P

ER

I)]

at 0

5:59

27

Mar

ch 2

014