JÖ N K Ö P I N G I N T E R N A T I O N A L BU SI N E SS SC H O O L JÖ N KÖ P IN G U N IVERSITY

International Tourism Demand for Greece: A study of the impact of the Athens Olympic Games 2004

Bachelor Thesis in Economics Author:

Emanuel Raptis EP08,

Tutor:

Lars Pettersson Sofia Wixe

Jönköping: 2011 Spring

Bachelor Thesis in Economics Title:

International Tourism Demand for Greece: A study of the impact of the Athens Olympic Games 2004

Author:

Emanuel Raptis

Tutor(s):

Lars Pettersson and Sofia Wixe

Date:

Spring, 2011

Keywords:

International Tourism Demand, Greece, mega events, Olympic Games, cross-section data analysis

Abstract This paper examines the development of income in the tourist generating countries, the relative prices controlled for the exchange rate, and the distance in kilometers between the capital in the origin countries and Athens as determinants of international tourism demand for Greece. By the deployment of an OLS log-linear regression model coupled with annual cross-section data for the period between 1998 and 2007, the desired effects could be captured. The results from this study indicates that after 2004, both the importance of income in the tourist generating countries and distance between the countries of origin and Athens have experienced a quantum drop in importance as determinants of international tourism demand for Greece. Furthermore, the elasticities of these factors remained at the new level throughout the remaining period studied. This suggests that the respective elasticities have reached a new plateau after 2004 where the impact on international tourism demand is less sensitive to changes in these specific factors. Finally, the investments made in infrastructure supporting the Olympic Games have the possibility to benefit the T&T sector in Greece for an extended period of time going forward.

i

Kandidatuppsats inom Nationalekonomi Titel:

Internationell efterfrågan för turism till Grekland: en studie av de Olympiska Spelen i Atens påverkan före och efter 2004.

Författare: Emanuel Raptis Handledare: Lars Pettersson och Sofia Wixe Datum:

Vårterminen 2011

Sökord:

Internationell turism, Grekland, Olympiska Spelen, analys av tvärsnittsdata

Sammanfattning Denna studie behandlar utvecklingen av variablerna inkomst i de turist genererande länderna, relativa priser kontrollerade för valutakurser mellan länderna samt avstånd mellan de turist genererande ländernas huvudstad och Aten i förhållande till antalet årliga turist ankomster till Grekland. Genom att använda en OLS log-linjär regressions modell tillsammans med årlig tvärsnittsdata för perioden 1998 – 2007 kunde de önskade effekterna observeras. Resultaten av denna studie indikerar att både elasticiteten för inkomst och avstånd minskade i betydelse som förklarande faktorer av turistrelaterade ankomster till Grekland efter 2004. Dessutom har denna studie kunnat påvisa att elasticiteterna för de nämnda variablerna varit stabila på den nya nivån under resten av den studerade perioden. Detta leder till slutsatsen att efter 2004 har de respektive variablernas elasticitet nått en ny nivå där internationella turist ankomster till Grekland är mindre känsligt för förändringar i dessa variabler. Därtill antas de investeringar gjorda i infrastruktur för att stödja de Olympiska Spelen kunna påverka turism sektorn i Grekland positivt under en längre period framöver.

ii

Table of Contents 1 Introduction ............................................................................... 1 1.1 1.2 1.2.1 1.2.2 1.3

Purpose of the study...............................................................................1 Background ............................................................................................2 The T&T sector in Greece .....................................................................2 Athens 2004 Olympic Games .................................................................4 Outline ....................................................................................................4

2 International Tourism Demand ................................................. 4 2.1 2.2 2.2.1 2.2.2

Theoretical Framework ...........................................................................5 Empirical Framework ..............................................................................6 Overview of Empirical Research ............................................................6 Relevant Empirical Studies .....................................................................8

3 Data and Statistical Method.................................................... 10 3.1 3.2

Specification of the model .................................................................... 10 Data and sources ................................................................................. 11

4 Empirical Findings and Discussion ....................................... 12 5 Concluding remarks................................................................ 15 References ................................................................................... 17 Appendices .................................................................................. 19

iii

1

Introduction

The world is becoming more and more interconnected as globalization is fostered by advances in technology and trade in goods and services. As people gain more knowledge about other places and cultures through their social environment, their curiosity about these destinations increase according to their taste. Travel and Tourism (T&T) is in essence a product or service that any country can package and sell according to its endowments as to attract a variety of tourists with different tastes from other countries as well as its domestic citizens. It is thus considered as a positive force driving globalization that promotes international, and interregional, trade. In fact, the T&T sector is one of the largest industries in the world today, accounting for as much as 9.9 % of world GDP in 2009 (Blanke, Chiesa & Herrera, 2009) and an 18 % share of Greece’s GDP in 2009 (Invest in Greece Agency). Furthermore, the T&T sector is assumed to sustain rapid growth in the future as emerging economies continue to experience development at a fast pace. One factor that is widely assumed to increase the attractiveness and perception of a given country’s tourism product is the hosting of mega events. The Olympic Games is the single largest mega event that a country can host and the preparations for such an event are many times extensive and usually stretches over many years before the event takes place. In 2004, Greece hosted the 28:th Summer Olympic Games which was the largest Olympic Games in terms of number of participating nations and number of sports events in history (De Groote, 2005). The possible impacts of the Olympic Games on international tourism demand to Greece is the primary focus of this study. The question is of major importance for a T&T depending country such as Greece and the knowledge gained from this paper can assist in the development of improved strategies requiring an increased understanding about the drivers of international tourism demand for Greece. Furthermore, the study also contributes to an increased knowledge of the impact of mega events on international tourism demand. Essentially, the key question being addressed is this; does it pay off for a country to make the investment of hosting a mega event such as the Olympic Games?

1.1

Purpose of the study

The purpose of the study is to analyze the possible impacts of the Athens 2004 Olympic Games on the factors affecting international tourism demand for Greece. The primary factors in focus are the yearly movements of the elasticities of the following variables; income in the tourist generating countries, the relative prices controlled for the exchange rate, and the distance in kilometers between the capital in the origin countries and Athens as determinants for international tourism demand for Greece. The impact of the Athens 2004 Olympic Games on the elasticities of the explanatory variables is of particular interest. To capture these effects an OLS log-linear regression model has been constructed using annual crosssection data. This method is particularly useful for this purpose since it allows for year-to-year comparisons of the estimated elasticities. The period included in the study is between the years 1998 – 2007 which permits comparisons before and after the Athens 2004 Olympic Games.

1

1.2

Background

1.2.1

The T&T sector in Greece

For many individual economies, the T&T sector is of crucial importance and a key driver for growth. One such country is Greece, where the T&T sector accounted for as much as 18 % of GDP in 2009 (Invest in Greece Agency) and where its government has historically prioritized its policies. Taking into account the indirect benefits from the T&T sector, such as increased production in the upstream industries supporting the T&T sector, the real impact is assumed to be even higher (WTTC, 2010). Furthermore, positive spill-overs of increased and improved infrastructure such as airports, roads, ports, and telecommunications also add to increased productivity and competitiveness of the total economy. These latter effects are of substantial importance for developing economies in particular as it increases the country’s attractiveness of not only tourists but to foreign investors as well. In addition, the T&T sector is a major source of foreign exchange for many developing economies which is also an important factor for improving the country’s national accounts and balance of payments. Figure 1 gives an overview of the trends in international tourist arrivals for the period between 1995 – 2009 for Greece. Tourism Arrivals - Thousands 18,000 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0,000 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Figure 1: International Tourism Arrivals to Greece for the period 1995 – 2009. Source: World Tourism Organization (UNWTO)

In recent years, The World Economic Forum (WEF) has acknowledged the importance and contribution of travel and tourism for the global economy and is therefore conducting a Travel & Tourism Competitiveness Index (TTCI) semi-annualy covering 133 countries in the 2009 Travel & Tourism Competitiveness Report (TTCR). The purpose of the TTCI is to function as a comprehensive tool for countries wishing to improve its competitiveness within the T&T sector. Building upon 14 pillars, the TTCI addresses a total of 71 different indicators for the assessment of each country’s competitiveness of the T&T sector. The top five performers in the 2009 TTCI are; Switzerland, Austria, Germany, France, and Canada, while Greece is found at 24:th place. Furthermore, some of Greece’s main tourism competitors such as Spain and Portugal are found at 6:th and 17:th place respectively, while for example Italy, Turkey, Tunisia, and Egypt are lagging behind Greece at 28:th, 56:th, 44:th, and 64:th place respectively. (World Economic Forum Travel & Tourism Competitiveness Report, 2009)

2

Taking the TTCI as the point of departure for the assessment it is possible to identify a number of strengths and weaknesses in the Greek T&T sector. The relatively high ranking is driven mainly by two positive factors. First, as has already been noted, the government has focused on the T&T sector and has established a well functioning tourism organization. In 2009, the budget for the Ministry of Tourism amounted to a total of EUR 131 million of which about EUR 102 million was allocated to the National Tourism Organization (GNTO). This can be compared to the tourist organizations of other countries in the Mediterranean area Portugal, Italy and Egypt with budgets of EUR 50 million, EUR 76.5 million and USD 40-65 million respectively for the same years (OECD, 2010). The relatively high budget share allocated to the Ministry of Tourism and to the GNTO places Greece on 3:rd place amongs 133 economies in the Prioritization of Travel and Tourism pillar in the TTCI. Second, Greece ranks 5:th in the Tourism infrastructure pillar which includes the following indicators; availability of hotel rooms, presence of major car rental companies, and ATMs accepting VISA cards. A recent example of the government taking actions to maintain a highly competitive tourism infrastructure are measures taken including a reduction in taxes paid by tourism enterprises and a conversion of unemployment benefits to employment benefits for tourism enterprises employing. Furthermore, to increase the quality of the service, quotas and special skills requirements for hotel staff have been issued (OECD, 2010). However, there are also a number of factors holding the development in the T&T sector back and where additional reforms and measures are of high importance for an increase of both T&T and overall competitiveness. Mainly, policy rules not supportive of the sectors development, rules governing FDI and foreign ownership, and time and costs for starting a new company are the factors that strongly constraint the competitiveness of the T&T sector and the Greek economy overall (TTCR, 2009). The International Bank of Reconstruction and Development (IBRD) annually performs the Doing Business Report (DBR) where nine factors and processes involved in doing business are evaluated for 183 economies. Looking at the factors highlighted in the TTCR in the 2011 DBR, where Greece ranks 109 overall, the statements in the TTCR are strongly reinforced. Table 1 gives a comparison between Greece and the OECD average for the three factors restricting the development of the T&T sector in Greece. Table 1: Selected DBR indicators for Greece and the OECD average. *0 indicates low protection of investors and 10 indicates high protection. Country

Starting a business (No. of procedures)

Registering property (No. of procedures)

Protecting investors (Index 0-10*)

Greece

15

11

3,3

OECD average

5,6

4,8

6,0

Source: The World Bank Group, Doing Business 2011 Report

A National Strategic Reference Framework (NSRF) for the period 2007 – 2013 including nine sectoral operational programmes has been developed in cooperation with the European Union (EU) to tackle the most critical issues for increased development in Greece. Enhancing competitiveness and entrepreneurship is one of

3

the sectoral operational programmes in the NSRF addressing many of the weaknesses in the Greek business climate, including those in Table 1. 1.2.2

Athens 2004 Olympic Games

In 2004 Athens hosted the 28:th Olympic Games, over 100 years after the first modern games that was held in Athens in 1896. In terms of the number of participating nations and the number of sports events this was the largest Olympic Games in history and it attracted some 10,500 competitors with accompanying families and staff (De Groote, 2005). Indeed, a mega event like this attracts a large number of people in itself and it is generally the case that the international tourism arrivals increase during the period around the event. Furthermore, the infrastructure developed in order to support the Olympic Games has the potential to benefit the T&T sector in Greece during an extended period of time. For example, the sports facilities in the Olympic complex has been reused to host other large events such as the Eurovision Song Contest final in 2006 and the Olympic village has been converted to residential housing with the capacity of 10,000 inhabitants (Potsiou & Zentelis, 2005). Transport and communication infrastructure was rebuilt and improved as well. Most notably, a new airport was constructed outside of Athens with the capacity of hosting over 16 million visitors annually together with new roads and train connections to the city center. The old airport has now been converted to an exhibition center hosting for instance Posidonia, the largest international shipping exhibition semi-annually, and other facilities is also being used as conference centers or business parks. Moreover, new tram and metro lines were built to connect the visitors to the Olympic facilities, cultural heritages and to the seafront of Athens. In total, public and private investments in infrastructure to support the Olympic Games is estimated to have amounted to EUR 7,2 billion where the government provided about EUR 6 billion (Kasimati & Dawson, 2008). This further strengthens the fact that the government believed hosting this event would benefit the Greek economy and T&T sector in the long-term.

1.3

Outline

The outline of the remaining parts of this paper is organized as follows. Section two gives a presentation of international tourism demand and consults relevant literature, reports, and previous research for this study. In section three, a general explanation of the empirical methodology and data to be used in the study is given. Section four presents the results from the empirical analysis which are then discussed. A summary and concluding remarks are given in the last section.

2

International Tourism Demand

This study will focus on the international tourist which is defined by the United Nations World Tourism Organization (UNWTO) as a person temporary visiting a foreign country with a duration of at least 24 hours but less than one year and where the purpose of the visit fits under one of the following headings; leisure, business, family, mission, and meeting. In general, tourism demand is referred to tourismoriented products such as hotels, restaurants, transportation, and similar tourism

4

services that cannot be purchased within the tourist’s country of residence. Therefore, tourism can be viewed as a regular export product with the distinction that the buyer must purchase the product on the location of the supply. The location of the tourism supply is commonly referred to as the destination and the residential location of the tourist is referred to as the origin. Finally, international tourism demand is generally associated with either the number of tourism arrivals from the origin to the destination, or the tourism receipts in the destination. In this study, international tourism demand is equivalent to the number of tourist arrivals unless stated otherwise.

2.1

Theoretical Framework

In a broad literature review of the tourism phenomena, Papadapoulos (1986) defines the tourism product and the tourism market. He argues that what constitutes the tourism product is on the one hand the country’s set of endowments such as cultural heritage, natural resources, climate as well as other factors including the quality of tourism infrastructure such as hotels and leisure facilities. On the other hand a country’s endowments and factors can be perceived differently by individuals according to their tastes and motives so the possibility to bundle the variety of factors into an attractive augmented product increases in importance. There is thus not a clearcut definition of a single tourism product that drives the demand for a country. Instead, a destination may be perceived as a brand with a reputation and a set of products and services that can be packaged, and promoted, according to a variety of consumer tastes. Due to this complexity of the tourism product, the tourism market is fragmented and consists of many businesses from a large variety of industries, both domestic and foreign, supporting and supplying the tourism market. The buyer of the product is the tourist, which in essence can be either domestic or international and hence one classification of the market can be that of a domestic and an international tourist market. Within these markets it is possible to identify more categories such as for instance markets for holiday and leisure, culture, events, and packaged tours. (Papadopoulos, 1986) A mega event such as the Olympic Games is typically accompanied by heavy publicity for the destination country to the extent that it potentially can influence international tourism demand before, during, and after the the event takes place. These effects are typically referred to as the halo effect of an event. In order to maximize the positive effects and to increase the brand image it is essential that the destination organizes the event with a sustainable approach in mind. Infrastructure developments, inclusion of the citizens in the destination, and safety improvements as a result of the mega event are some of the factors that not only benefit the residents in the destination country, but increases the positive perception of the destination from the international community. If successful in combining these factors, the destination country can achieve a quantum leap in international tourism demand that may result in either accelerated growth, or reaching a new higher plateau, in terms of tourist arrivals after the event. (Getz, 1997)

5

2.2 2.2.1

Empirical Framework Overview of Empirical Research

Lim (1997a), in her review of 100 published studies of empirical international tourism demand models for the period 1961 – 1994, discusses the most common approaches for the study of international tourism demand. By looking at a variety of aspects such as data and sample sizes, dependent and explanatory variables, and model specifications, an overview of the previous research is given and a set of conclusions has been made. First, all studies have used either international tourist arrivals/departures (51/100) or international tourist expenditures (49/100) as the dependent variable. Some studies have used more than one dependent variable, in those cases travel exports and/or imports, length of stay, and nights spent at tourist accommodations have also been used. Another conclusion that can be drawn from the study is that, although the factors influencing variations in the demand for international tourism are many, most studies in the review have focused on economic factors as explanatory variables in their estimations. Primarily, these studies have focused on income, relative prices, tourist prices, and transportation costs as explanatory variables. Income in the country of origin is the most common explanatory variable (84/100) and it is widely believed that income is positively correlated with international tourism demand. Typically, real or nominal per capita income (GDP or GNP) is considered as the measure of income in the country of origin. But proxies such as real per capita consumption, foreign travel budget, and real disposable income less expenditures on necessities has also been used, although much less frequently. The second most frequent explanatory variable used is relative prices between the destination country and the country of origin. Here the consumer price index (CPI) is used as the proxy to reflect the differences in the relatives prices between the countries. Relative prices between the destination and the origin can be expressed as RPij = (CPIi/CPIj)(1/ERij)

(1)

where ERij is the exchange rate between the currency in origin j in terms of the currency in destination i. (Lim, 1997a, 2006). However, one potential problem with the above measure of relative prices is the assumption that the tourist consumption is similar to that of the residents in the destination country. Some studies have tried to construct a price index of a basket of goods that tourists consume (TPI) in the origin and compare this with the same goods in the destination. Since TPI is rarely available, CPI is used as a proxy. Another factor related to prices is the transportation costs between the country of origin and the destination which is usually measured as the cost of round-trip travel. Finally, the relative currency price or the exchange rate is often introduced into international tourism demand models as an important explanatory variable. The exchange rate can either be incorporated to the relative price measure introduced in (1) above by multiplying an index of the relative currency prices to the ratio of CPIs, or as an independent explanatory variable. One argument in favor of including the exchange rate separately is that tourists may be better informed about the relevant exchange rates compared to the relative CPIs, since information about the exchange rates are often easily accessible for the public. On the other hand, since

6

the exchange rate is a type of relative price between currencies, there might be an issue of multicollinearity between the exchange rate and relative CPIs. Table 2 shows the use of explanatory variables and and the frequency of occurrence in the 100 studies reviewed by Lim. (Lim, 1997a, 2006) Table 2: Classification by number of explanatory variables used. Explanatory variable

Frequency

Explanatory variable

Frequency

Income

84

Seasonal factors

14

Relative prices

73

Marketing expenditures

7

Transportation costs

55

Migration

5

Exchange rates

25

Business travel/trade

5

Trend

25

Economic activity indicators

3

Dynamics

26

Qualitative factors

60

Competing destinations/goods

15

Other

27

Source: Lim (1997a)

When it comes to the data used in the 100 studies reviewed by Lim (1997a), the most common data are annual (56/100), cross-section (9/100), and pooled (9/100). For the studies using annual time-series data the average number of observations were 15 years which reflects a common feature, and potential problem, of relatively low number of observations in tourism research studies. To obtain more meaningful regression estimates some studies used monthly, quarterly, cross-section, and pooled annual and cross-section data. One advantage of using time-series data however is that random and deterministic time trends can be included which can be used in forcasting international tourism demand for a specific region or country. In contrast, cross-section data analysis investigates the determinants of international tourism demand for a region or country at a fixed time period. This method increases the number of observations which improves the quality of the results (Lim, 1997c). Finally, Lim (1997c, 2006) has also investigated the most common methods used in international tourism demand models where the most usual econometric approach is the single-equation model, and where the linear or log-linear functional forms dominate. The rationale behind the choice of one of the two models is its computational and interpretational ease where both the short- and long-run marginal effects and elasticities can be studied respectively. During the course of time, the modeling of international tourism demand has shifted more and more from the simple single equation towards more state-of-the-art modeling and econometric approaches. Among these new methods are unit root tests and non-stationary processes such as cointegration methods, and systems of equations such as the Almost Ideal Demand System (AIDS). Method developments of these kinds permit a broader understanding of both short- and long-run model estimations. (Lim, 2006)

7

2.2.2

Relevant Empirical Studies

Several studies on international tourism demand for Greece have been conducted in the past using a variety of methods for estimation. Dritsakis and Athanasiadis (1999) used time-series data for the period between 1960-1993 to construct an OLS multiple log-linear regression model for their estimation. In their study, they used a relative measure of tourism arrivals from a set of origin countries to the destination and total population of the origin countries as dependent variable. The explanatory variables was the disposable national income in the origin, average costs for (i) a ten-day stay in Greece and (ii) a ten-day stay in other competitive Mediterreanean countries, the exchange rate between the destination and origin, gross investments in fixed assets in Greece, promotional expenditures of the GNTO in the origin countries, and a dummy variable for political stability in Greece. As expected, disposable income had a positive relationship with tourist arrivals to Greece in most cases, some countries however continued to generate an increase in tourist flows to Greece even when their disposable incomes fell during times of recession. Furthermore, average total costs and travel costs in and to Greece did only have a small effect on tourism demand. On the other hand, a decrease in costs of a competitor destination had a larger negative impact on tourism arrivals. Surprisingly, the exchange rate movements does not seem to have had a large impact. Although the Greek currency gradually depreciated a rise in foreign exchange earnings did not occur, mainly due to an increase in arrivals of tourists with lower income levels. Gross investments, promotional expenditures and political stability were all found to have an impact on tourism arrivals where an increase of investments in tourism infrastructure and promotional expenditures had a positive effect and where political unrest contributed to a decrease in tourism arrivals. (Dritsakis & Athanasidis, 1999) In a more recent study however, Dritsakis (2004) used annual data between the period 1960-2000 in a cointegration analysis of German and British tourism demand for Greece. In order to explain German and British demand for tourism in Greece, expressed as arrivals to Greece from Germany and the U.K. respectively, the income in the origin, tourism prices in the destination, transportation costs, and exchange rates where used as independent variables. This method permits the testing of long-run relationship and short-run dynamics simultaneously and the results were somewhat different compared to the OLS multiple log-linear regression model above. For instance, higher income levels in Germany and the U.K. resulted in a decrease of tourism arrivals since other, more exotic or distant, tourist destinations became more attractive. However, the signs on the remaining variables were all as consistent with economic demand theory. (Dritsakis, 2004) By looking at previous studies using other countries as destination but with similar methods and variables it is possible to observe if the significance of the variables remain the same or if different destinations yield different results. In a study of the determinats of demand for international tourist flows to Turkey, Uysal and Crompton (1984) identifies and describes a set of variables to explain international tourism arrivals and/or expenditures. Again, the model used was an OLS loglinear multiple regression model with annual time-series data for the period 19601980 for 11 origin countries. The explanatory variables included in their model are; per capita income in the origin, relative prices between the origin and destina-

8

tion, exchange rate, promotional expenditures by the destination, and a dummy variable for special events such as political unrest. As expected, the results from this study is consistent with the results from similar studies with one apparent distinction that promotional expenditures, although significant, seemed to have a minimal impact on tourism flows to Turkey for the period of consideration. (Uysal & Crompton, 1984) Akis (1998) performed a more compact econometric study on the international tourism arrivals to Turkey using annual time-series data, including only two explanatory variables, national income of the tourism generating countries and relative prices, in a double-logarithmic regression model. He argues that this approach gives more satisfactory results since the apparent problem of multicollinearity between the explanatory variables is limited. Furthermore, the relative price variable is in the form of equation (1) above where the exchange rate is included to explain the relative prices. For 15 out of 18 countries the results were significant and satisfactory despite the relatively narrow period, 1980-1993, studied and the R2 was above 0.70 for 14 of the countries. This posts the question of the appropriateness of including more explanatory variables in the model in order to explain international tourism demand. (Akis, 1998). In addition, Aslan, Kaplan and Kula (2008) looked at the impact of some of the supply side factors such as accommodation capacity and infrastructure networks on the international tourism demand to Turkey. Their study consults a dynamic panel data model for the estimation of international tourism demand. Accommodation capacity and the public investment ratio in the destination was added as explanatory variables in addition to per capita income in the origin and relative prices between the destination and the origin. Contrary to the previous studies using Turkey as the destination, income in the origin, although significant, was not found to have a large impact in 9 out of the 10 countries investigated. The interpretation of this is that tourism in Turkey is not seen as a luxury good and thus income is not an important factor in determining the tourism demand to Turkey. Relative prices was significant and had a negative relation with the tourism demand in consistency with demand theory. Interestingly, accommodation capacity was significant even at the 1% level of significance and had an elasticity of 1.72 showing that supply side factors do impact tourism demand. (Aslan, Kaplan & Kula, 2008) Carey (1991) used combined time-series and cross-section data to estimate international tourism demand for the Caribbean islands. By substituting transportation costs with a proxy for distance in addition to income in the origin, prices of tourist services, exchange rates and promotional activities, the apparent problem of multicollinearity between transportation costs and income could be solved without omitting one of the explanatory variables. In most previous studies, this type of problem had been solved by omitting transportation costs from the model, resulting in an upward bias for the income variable. (Carey, 1991) Finally, some attention will be given to previous research assessing the impact of the Athens 2004 Olympic Games on the Greek economy. Kasimati and Dawson (2008) have constructed an aggregated macroeconometric model for this given purpose. In their aggregated model, they used time-series data and an OLS linear regression model to estimate the impact of real income in the OECD countries, real exchange rate (as equation (1) above), and a measure of the quality of Greek tourism services on the exports of tourism services. Their results indicate that, as ex-

9

pected, income in the OECD countries is significant and positively related to the exports of Greek tourism services. Surprisingly however, the real exchange rate was found to be neither significant nor negatively related to exports of tourism services. Although the explanation for the latter results are not entirely clear, it is possible that there exist some endogeneity between the the strengthening of the tourism sector and the strengthening of the exchange rate. Another explanation given by the authors is that the stabilization of the exchange rate due to adhesion to EMU monetary mechanism reflects better quality in tourism services and a stronger economy overall. (Kasimati & Dawson, 2008) In addition to this, Hede (2005) conducted a case study on the impact of the Athens 2004 Olympic Games media cast in Australia on the perceptions and attitudes of Australian citizens towards Greece as a tourist destination. The results from her survey indicated that, across the sample, 38.7% of the respondents had changed their overall perception and attitude towards Greece as a tourist destination in a positive direction. Hence, it is possible that international tourism demand for a destination can increase by the consumption of television mediacasts from a mega event by citizens in the origin countries. (Hede, 2005)

3

Data and Statistical Method

The intention of this research paper is to establish the determinants of tourism inflows to Greece by the use of an empirical model allowing for test of suitable dependent variables assumed to affect tourism inflows. Since the time period included in this study is between the years 1998 – 2007, cross-section data will be used in a OLS multiple regression model and the year-to-year developments of the elasticities of the variables will be mapped out. The stated period examined was chosen mainly due to its permission of comparisons between the period before and after the Athens 2004 Olympic Games.

3.1

Specification of the model

According to Lim (1997c, 2006), the general international tourism demand model may be written as DTij = f(Yj, TCij, RPij, ERij, QFi)

(2)

where δf/δYj > 0, δf/δTCij < 0, δf/δRPij < 0, δf/δERij > 0, and DTij Yj TCij RPij ERij QFi

= number of tourist arrivals in the desitination i from the origin j = income in the country of origin j = transportation cost between the destination i and the origin j = relative prices between the destination i and the origin j = exchange rate between the currency in the origin j in terms of the destination i currency = qualitative factors in the destination i

For the purpose of this study, the general international tourism demand model will be modified to include some explanatory variables found relevant by examining the previous empirical studies. Furthermore, the choice and combination of va-

10

riables will also take into account previous methods used for the avoidance of multicollinearity between the included explanatory variables. No qualitative factors were included in the model, mainly due to the complexity of obtaining relevant data that can be compared between all of the countries included in the study. Except for that, the model is in line with the general model stated by Lim (1997c, 2006). The model is written as DTijt = f(Yjt, RP*ijt, Dij)

(3)

where δf/δYjt > 0, δf/δRP*ijt < 0, δf/δDij < 0, at time t and DTijt Yjt RP*ijt Dij

= number of tourist arrivals in the destination i from the origin j = income in the country of origin j expressed as real GDP per capita in constant (2000) dollars = relative consumer price index (CPI) between the destination i and the origin j multiplied by the exchange rate between the currencies in the origin j in terms of the currency in the destination i = euclidian distance in kilometers between the capitals in destination i and the origin j

where the variable RP*ijt is expressed as RP*ijt = (CPIit/CPIjt)(1/ERijt)

(4)

The following log-linear functional form of the model was chosen for the estimation of the parameters: lnDTijt = β0 + β1lnYjt + β2lnRP*ijt + β3lnDij + εi

(5)

This approach makes possible the estimation of the elasticities of the variables over time which is one of the primary aims of this study. Furthermore, a nonlogarithmic linear functional form was also tested where the data showed nonnormality and very low goodness of fit. The log-linear model solved the problem of non-normality and provided much higher R2 values.

3.2

Data and sources

The annual cross-section data used in the empirical analysis has been obtained from official sources such as the Hellenic Statistical Authority (ELSTAT), the World Bank Group (IBRD), OANDA, and the World Distance Calculator (WDC). The sample consists of annual tourist arrivals from 36 specific countries, which represents between 95 % - 97 % of the total population depending on the year of study, was obtained from ELSTAT. Real GDP per capita in constant (2000) dollars and annual CPI data were collected from the IBRD database. The exchange rates was gathered from OANDA and were calculated as the annual average exchange rates of the relevant year expressed as the origin currency in terms of the destination currency. Furthermore, the actual currencies in use in any given year were used. For example, in the year 1999 the exchange rate between the US and Greece was calculated as USD/GRD, while in 2001 it was calculated as USD/EUR. Countries that had experienced deflation during a single year was omitted from the regression of that year since a negative CPI resulted in a negative RP* which cannot be logarithmically transformed. This is not expected to have disturbed the results in a significant way

11

since it was a rare phenomenon. Finally, the distance in kilometers between the origin capital and the destination capital was calculated as the straight line distance between the capitals based on their latitudes and longitudes. Data on promotional expenditures in the countries of origin by the GNTO was requested but neglected by the GNTO.

4

Empirical Findings and Discussion

Since the analysis is based on the OLS log-linear regression model as given by equation (5), the coefficients in Table 3 are estimates of the elasticities for each explanatory variable. White’s general test for heteroscedasticity and the correlation matrices showed no signs of either heteroscedasticity or multicollinearity in the obtained data. Table 3: Summary of regression results of international tourism demand for Greece. Year

Constant

lnYjt

lnRP*ijt

lnDij

R2

N

1998

12.794*** (2.109)

0.987*** (0.213)

0.003 (0.067)

-1.333*** (0.23)

0.592

35

1999

13.632*** (2.391)

0.859*** (0.258)

0.065 (0.081)

-1.306*** (0.241)

0.613

33

2000

14.197*** (2.199)

0.870*** (0.241)

0.062 (0.083)

-1.390*** (0.248)

0.588

34

2001

15.197*** (2.452)

0.838*** (0.24)

0.057 (0.087)

-1.428*** (0.242)

0.597

34

2002

16.857*** (2.528)

0.777*** (0.25)

0.096 (0.088)

-1.547*** (0.231)

0.647

35

2003

16.410*** (2.368)

0.811*** (0.248)

0.084 (0.083)

-1.566*** (0.235)

0.644

35

2004

15.678*** (2.610)

0.842*** (0.269)

0.076 (0.088)

-1.511*** (0.227)

0.650

34

2005

16.306*** (2.55)

0.547* (0.283)

0.194 (0.119)

-1.208*** (0.254)

0.547

35

2006

16.389*** (2.343)

0.525** (0.258)

0.187* (0.105)

-1.166*** (0.228)

0.557

36

2007

16.779*** (2.489)

0.496* (0.271)

0.160 (0.107)

-1.170*** (0.229)

0.533

36

Notes: Dependent variable is lnDTijy *** Significant at level of significance, α = 0.01 ** Significant at level of significance, α = 0.05 * Significant at level of significance, α = 0.10 Std. errors in parenthesis

The coefficients of the income in the origin countries were all statistically significant at the 5 % level of significance except for the years 2005 and 2007 where they were significant only at the 10 % level. All of the results on the estimates of income elasticities have the expected signs, as income increases in the countries of origin so does the demand for international tourism for Greece aggregated over the 36 countries included in the study. However, the income elasticity is less than one for all years indicating that on an aggregated level, tourism to Greece is considered to be a normal product/service. Interestingly, the impact of a 1 % change in income in the tourist generating country decreased from 0.842 % change in 2004 to 0.547 % in 2005, a decrease of 35 % in importance between the years. This drastic change seems to support the hypothesis that the Olympic Games is likely to have impacted

12

the international tourism demand for Greece in a positive manner. After 2004, the income factor has been less important in generating tourism arrivals to Greece. Moreover, the coefficients of the income elasticities continued to decrease for the years 2006 and 2007 indicating that the relative importance of income as a factor driving tourism demand for Greece is decreasing. Surprisingly, the coefficients of the relative prices were all not statistically significant at the 5 % level of significance but only at the 10% level of significance for the year 2006. The measures of relative prices and the results are in parity with those used and obtained by Kasimati and Dawson (2008). This suggests that there might exist some endogeneity between the strengthening of the tourism sector and the strengthening of the exchange rate. Alternatively, the adhesion to the EMU monetary mechanism reflects a better quality in tourism services and a stronger economy overall as proposed by Kasimati and Dawson (2008). Another explanation might be a high share of tourism flows from neighbouring countries where the purpose of the trip is to visit relatives or friends. In such a case relative prices may be of little concern. As expected, distance in kilometers between the capitals in the countries of origin and Athens was negatively related to tourism arrivals to Greece. All coefficients of the estimated elasticities of distance were statistically significant at the 5 %, and even on the 1 %, level of significance. In general, as the distance to the destination increases so does the transportation costs in terms of more expensive air fares, longer actual travel time, and other costs such as visas if the country of origin is not a member of the EU. However, in accordance to the development of the estimates of the income elasticity, the impact of distance as a factor for tourism demand has also decreased in importance after 2004. In 2004, the elasticity was 1.511 while in 2005 it was -1.208 meaning that a 1 % increase in distance had a lesser negative impact on tourism arrivals to Greece in 2005 compared to the year before. Between the years 2004 and 2005 the coefficient of the elasticity of distance increased by 20 %, further strengthening the hypothesis that the Olympic Games contributed positively to international tourism demand for Greece. In addition, the coefficients of the elasticities of distance increased further in 2006 and remained stable at that level in 2007. Figure 2 plots the graphs of the annual estimates of the elasticities of both income in the countries of origin and distance between the capital cities in the countries of origin and Athens. Annual elasticities of income and distance 1,5 1 0,5 0 -0,5

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

-1 -1,5 -2

Figure 2: Annual estimates of elasticities of income in the countries of origin and distance between the capital cities in the countries of origin and Athens.

13

As made evident by the graphs in figure 2, the elasticities of both income and distance remained relatively stable in the period between 1998 and 2004. After 2004 however, both the elasticity of income and the elasticity of distance experienced a quantum drop in importance as a factor determining international tourism demand for Greece and remained at that level for the rest of the period studied. This suggests that the respective elasticities have reached a new plateau where the impact on international tourism demand is less sensitive to changes in the specific factors studied. To further test the evidence that the Athens 2004 Olympic Games have changed the elasticies of both income and distance significantly, the confidence intervals for the estimated beta coefficients in both 2004 and 2005 was calculated at the 5 %, 10 %, and 20 % level of significance. The key issue here is whether the confidence intervals between the years overlap or not. If they do not overlap, the hypothesis that the elasticities have changed as a result of the Olympic Games is strongly supported. For the year 2004 the confidence interval for income at the 20 % level of significance was [0.4971, 1.1869], and for the year 2005 it was [0.1842, 0.9098]. The confidence intervals for distance at the 20 % level of significance were [-1.802, -1.213] in 2004 and [-1.5336, -0.8824] in 2005. Clearly, the confidence intervals do overlap in both cases when calculated at the 20% level of significance, and hence at the 5 % and 10 % levels as well. Thus, these findings somewhat weakens the evidence that the Athens 2004 Olympic Games has decreased the importance of both income and distance as explanatory variables of international tourism demand for Greece. However, figure 2 shows that there are reasons to suggest that the Athens 2004 Olympic Games has indeed impacted international tourism demand for Greece at least to some degree. In addition to the findings from the empirical analysis, the tourism arrivals to Greece across the sample of the 36 countries included in the study demonstrate an increase in the growth rate immediately after 2004. Between the years 2004 and 2007 the average annual growth rate in international tourism arrivals to Greece was 8.94 % as compared to 3.59 % in the period between 1998 and 2004. The graph in Figure 3 suggests that Greece experienced accelerated growth in international tourism arrivals from the 36 countries included in the study in the years after the Olympic Games in 2004. On the contrary, the graph in Figure 1 showing total international tourism arrivals to Greece in the period between 1995 and 2009 propose that international tourism demand reached a new plateau after 2004 but with a similar growth pattern as the years leading up to the Olympic Games. However, whether Greece has experienced accelerated growth or has reached a new plateau with maintained growth in international tourism demand as described by Getz (1997) after 2004 is not the scope of this paper and is therefore subject to further research.

14

Sample Tourism Arrivals - Thousands 20,000 15,000 10,000 5,000 0,000 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Figure 3: International tourism arrivals to Greece from the 36 countries included in the sample for the period between 1998 – 2007. Source: Hellenic Statistical Authority (ELSTAT)

Apart from the fact that the model fit was relatively high between 0.533 and 0.65, the approach of using cross-sectional data and an OLS log-linear regression model has been proven to yield useful results. The use of cross-section data enables comparisons of the coefficients between years, which is not possible if using for instance time-series data. Hence, the changes in the estimated elasticities of income and distance respectively after the Athens Olympic Games 2004 could be detected, which was one of the primary aims of this paper. To further improve the results, the addition of relevant explanatory variables to the model could be done. Except for increasing the model fit, adding more variables may yield more accurate estimates of the elasticities already included in the model. However, the availability of relevant country specific data is limited and may require substantial resources to obtain. Furthermore, there are a large number of qualitative factors that are assumed to drive international tourism demand for Greece that may be hard to capture in the model. One such example is the development over time of country specific tourist perceptions towards the tourism product/service and the motives for traveling to Greece. Finally, extending the number of years after 2004 is necessary in order to assess if the changes in elasticities of income and distance is indeed enduring or temporary effects of the Athens Olympic Games 2004.

5

Concluding remarks

This paper has focused on the developments of the importance of the following explanatory variables; income in the tourist generating countries, the relative prices controlled for the exchange rate, and the distance in kilometers between the capital in the origin and Athens as determinants of international tourism demand for Greece. A particular point of interest has been the impact of the Athens Olympic Games 2004 on the development of the elasticities after 2004 compared to the period before the Olympic Games. By the deployment of an OLS log-linear regression model coupled with annual cross-section data, the desired effects could be captured. Income in the tourist generating countries and the distance between the capital cities in the origins and Athens were both significant and had the expected relationships with tourist arrivals. On the other hand relative prices, including the exchange rate, between Greece in terms of the countries of origin was not found to be statistically significant. More interestingly however, both the estimated elasticities of income and distance were found to have substantially decreased in impor-

15

tance as determinants of international tourism demand after 2004 and remained stable at the new level for the remaining period studied. This suggests that the respective elasticities have reached a new plateau where the impact on international tourism demand is less sensitive to changes in these specific factors. Further research including more explanatory variables in the model and extending the number of years studied after 2004 is needed in order to assure that these results are robust, and if the effect is of enduring or temporary nature. The findings in this paper contributes with extended knowledge about the possible impacts of mega events, such as the Olympic Games, on international tourism demand for any given country. Furthermore, the awareness of decreasing importance of income and distance as determinants of tourism arrivals, together with an increasing number of tourist arrivals, suggest that there are other factors driving international tourism demand for Greece. Such knowledge contributes to the understanding of the drivers of international tourism demand for Greece and could be useful in the process of developing improved strategies for the development of the T&T sector in Greece as well as in other countries. Finally, the question whether the investment of hosting a mega event pays off is a probably a highly empirical question. For a tourist dependent country like Greece, there seems to be evidence provided in this paper suggesting that hosting a mega event can contribute positively to international tourism demand, at least when it comes to the Olympic Games. Since the T&T sector is one of the largest sectors in Greece, positive contributions to this sector is likely to generate significantly large returns on the investment. However, there may be other factors capable of influencing this effect as well. A country more remotely located than Greece may for instance not experience such effects, despite the fact that it has a strong and well developed T&T sector.

16

References Aczel, A. D., & Sounderpandian, J. (2009). Complete Business Statistics (7th ed.). New York: Irwin/McGraw-Hill Akis, S. (1998). A compact econometric model of tourism demand for Turkey. Tour ism Management. Volume 19, 99-102 Aslan, A., Kaplan, M., & Kula, F. (2008). International Tourism Demand for Turkey: A Dynamic Panel Data Approach. Unpublished Blanke, J., Chiesa, T., & Herrera, E. T. (2009). The Travel and Tourism Competitive ness Index 2009: Measuring sectoral drivers in a downturn. In Blanke, J., & Chiesa, T. (Eds.), The Travel & Tourism Competitiveness Report 2009: Manag ing in a Time of Turbulence (p. 3-26). Geneva: World Economic Forum Carey, K. (1991). Estimation of Caribbean Tourism Demand: Issues in Measure ment and Methodology. Atlantic Economic Journal. Volume 3, 32-40 De Groote, P. (2005). Economic & Tourism Aspects of the Olympic Games. Tourism Review. Volume 60, 20-28 Dritsakis, N., & Agorastos, K. (1999). An Econometric Model of Tourist Demand: The Case of Greece. European Research Studies Journal. Volume 2, 83-90 Dritsakis, N. (2004). Cointegration analysis of German and British tourism demand for Greece. Tourism Management. Volume 25, 111-119 Getz, D. (1997). The impacts of Mega Events on tourism: Strategies for destina tions. In Andersson, T. D., Persson, C., Sahlberg, B., & Ström, L. (Eds.), The Impact of Mega Events (p. 5-32). Östersund: European Tourism Research Institute Gujarati, D. N., & Porter, D. C. (2009). Basic Econometrics (5th ed.). New York: Irwin/McGraw-Hill Hede, M. (2005). Sports-events, tourism and destination marketing strategies: an Australian case study of Athens 2004 and its media telecast. Journal of Sport Tourism. Volume 10, 187-200 Hellenic Statistical Authority. (2008). Arrivals of foreigners to Greece (in Greek). Re trieved February 25, 2011, from http://www.statistics.gr/portal/page/ Invest in Greece Agency. (2011). Tourism in Greece. Retrieved February 19, 2011, from http://www.investingreece.gov.gr/default.asp?pid=36&la=1 Kasimati, E., & Dawson, P. (2008). Assesing the impact of the 2004 Olympic Games on the Greek economy: A small macroeconometric model. Economic Model ling. Volume 26, 139-146 Lim, C. (1997a). Review of international tourism demand models. Annals of Tour ism Research. Volume 24, 835-849

17

Lim, C. (1997c). The functional specification of international tourism demand models. Mathematics and Computers in Simulation. Volume 43, 535-543 Lim, C. (2006). A survey of tourism demand modeling practice: issues and implica tions. In Dwyer, L., & Forsyth, P. (Eds.), International Handbook on the Economics of Tourism (p. 45-72). Cheltenham: Edward Elgar Publishing Limited National Strategic Reference Framework. (2010, March 29). Competitiveness and Entrepreneurship. Retrieved March 1, 2011, from http://www.espa.gr/en/ OANDA. (2011). Currency Converter. Retrieved March http://www.oanda.com/lang/sv/currency/converter/

3,

2011,

from

OECD. (2010). OECD Tourism Trends and Policies 2010. OECD online publishing Papadopoulos, S. I. (1986). The Tourism Phenomenon: an examination of impor tant theories and concepts. Revue de tourisme. Volume 3, 2-11 Potsiou, C. A., & Zentelis, P. (2005). Greece after the Gold Rush: Impact Analysis and Sustainability of the 2004 Olympic Infrastructure. [Presentation]. Cairo: FIG Working Week 2005 Uysal, M., & Crompton J. L. (1984). Determinants of demand for international tour ism flows to Turkey. Tourism Management. Volume 4, 288-297 White, H. (1980). A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity. Econometrica. Volume 48, 817-838 World Bank Group. (2011). Indicators. Retreieved March 2, 2011 from http://data.worldbank.org/indicator World Distance Calculator. (2009). Distance Calculator World. Retrieved March 10, 2011, from http://distancecalculator.globefeed.com/ World Economic Forum. (2009). The Travel & Tourism Competitiveness Report 2009. Geneva: World Economic Forum World Tourism Organization. (2011). World Tourism Data. Retrieved March 2, 2011, from http://data.un.org/DocumentData.aspx?id=245 World Travel & Tourism Council. (2010). Travel & Tourism Economic Impact: Greece 2010. London: World Travel & Tourism Council

18

Appendices Appendix 1 Overall TTCI rankings 2009. Source: World Economic Forum, 2009

19

Appendix 2 Greece individual performance in the TTCR 2009. Source: World Economic Forum, 2009

20

21

Appendix 3 Overall DBR rankings in 2011. Source: World Bank Doing Business Report 2011

22

Appendix 5 Greece individual performance in the DBR 2011 report compared to the DBR 2010 report. Source: World Bank Doing Business Report 2011

23

Appendix 6 Regression results for respective years using equation (5): lnDTijt = β0 + β1lnYjt + β2lnRP*ijt + β3lnDij + εi Year: 1998

Year: 1999

Year: 2000

24

Year: 2001

Year: 2002

Year: 2003

25

Year: 2004

Year: 2005

Year: 2006

26

Year: 2007

27

Appendix 7 Regression results from White’s general test for heteroscedasticity by the use of the auxiliary equation in (6): ε2i =

α0 + α1lnYjt + α2lnRP*ijt + α3lnDij + α4(lnYjt)2 + α5(lnRP*ijt)2 + α6(lnDij)2 + α7(lnYjt)(lnRP*ijt)(lnDij) + vi

Year: 1998, n = 35

Year: 1999, n = 33

Year: 2000, n = 34

Year: 2001, n = 34

Year: 2002, n = 35

Year: 2003, n = 35

28

Year: 2004, n = 34

Year: 2005, n = 35

Year: 2006, n = 36

Year: 2007, n = 36

29

Appendix 8 Countries

Omitted yrs

Countries

Omitted yrs

Albania

None

FYROM

1999

Austria

None

Romania

None

Belgium

None

Russian Federation

None

Bulgaria

None

Sweden

1998

Cyprus

None

Czech Republic

None

France

None

Slovak Republic

None

Germany

None

Finland

None

Denmark

None

Japan

1999-2005

Switzerland

None

Israel

2004

United Kingdom

None

Turkey

None

Ireland

None

Iran, Islamic Republic

None

Spain

None

Egypt, Arab Republic

None

Italy

None

Argentina

1999-2001

Norway

None

Brazil

None

Netherlands

None

Mexico

None

Hungary

None

United States

None

Poland

None

Canada

None

Portugal

None

Australia

None

Individual countries were omitted from the analysis in single years due to deflation in that year.

30

Appendix 9 Diagnostic test for heteroscedasticity Due to the tendency of heteroscedasticity in cross-section data, White’s general test for heteroscedasticity was performed. For this purpose, the following auxiliary regression model was applied (White, 1980): 𝜀2i =

α0 + α1lnYjt + α2lnRP*ijt + α3lnDij + α4(lnYjt)2 + α5(lnRP*ijt)2 + α6(lnDij)2 + α7(lnYjt)(lnRP*ijt)(lnDij) + vi (6)

The obtained R2 value is multiplied with the sample size n, which yields a chisquare value that is to be compared with the critical chi-square value. The critical chi-square value is chosen in accordance with the desired level of significance and the relevant degrees of freedom, which is the same as the number of regressors in (6). nR2 ~ χ27

(7)

If nR2 < χ27 the presence of heteroscedasticity can be ruled out (White, 1980, Gujarati & Porter, 2009). Year

WT

N

1998

4,375

35

1999

4,785

33

2000

6,018

34

2001

5,270

34

2002

4,095

35

2003

3,605

35

2004

2,414

34

2005

7,875

35

2006

5,112

36

2007

5,184

36

31

Appendix 10 Calculation of confidence intervals. Source: Aczel & Sounderpandian (2009), Gujarati & Porter (2009) 𝛽± 𝑡

𝛼 2

𝑠(𝛽 )

32