See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/259891581

Demand for tourism in Greece: A panel data

analysis using the gravity model

Article in International Journal of Tourism Policy · January 2014

DOI: 10.1504/IJTP.2014.063105

CITATIONS READS

2 274

3 authors, including:

Frank A.G den Butter

VU University Amsterdam
245 PUBLICATIONS 588 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Frank A.G den Butter
letting you access and read them immediately. Retrieved on: 11 May 2016
To be published as Chasapopoulos, P., F. A.G. den Butter and E.S. Mihaylov, 2014, Demand for
tourism in Greece: a panel data analysis using the gravity model, International Journal of Tourism
Policy, 6, pp. ??.

Demand for tourism in Greece: A panel data

analysis using the gravity model
Panagiotis Chasapopoulosd, Frank A.G. den Buttera and Emil Mihaylovb,c

Abstract
Tourism is one of the major drivers of the Greek economy. The contribution of
tourism to the Greek economy has proved especially relevant during the period of the
credit and euro crises with a high budgetary and balance of payment deficits. From
that perspective this study examines the impact of the socio-economic and
geographical determinants of foreign tourism demand in Greece. For the empirical
analysis a panel dataset of 31 countries is used over the period 2001-2010. The panel
data estimation indicates that distance and trade have more explanatory power than
relative prices and other determinants such as transport infrastructure. Income is
statistically significant in 3 out of the 8 specifications. Also political stability seems to
play an important role in tourism demand. The results are mixed for the competitive
prices between Greece and its main tourism competitors. An interesting finding is that
the Olympic Games of 2004 seem to have had a negative impact on international
tourist arrivals in Greece in that year.

Keywords: Tourism demand, Greece, gravity model, panel data analysis, system
GMM.

JEL-codes: E21, F14, H62,

a
Corresponding author VU University Amsterdam, Department of Economics and RITM, De
Boelelaan 1105, 1081 HV Amsterdam, The Netherlands; email: [email protected]
b
VU University Amsterdam, Department of Economics and RITM, De Boelelaan 1105, 1081 HV
Amsterdam, The Netherlands; email: [email protected]
c
LEI Wageningen UR (University & Research Centre), Alexanderveld 5, 2585 DB The Hague, The
Netherlands; email: [email protected]
d
VU University Amsterdam, Department of Economics and RITM, De Boelelaan 1105, 1081 HV
Amsterdam, The Netherlands.

Acknowledgements: We are very grateful to the editor and two anonymous reviewers for their
excellent comments and suggestions. Panagiotis Chasapopoulos gratefully acknowledges financial
support from the Flemish Science Foundation's (FWO) Odysseus programme.
Demand for tourism in Greece: A panel data
analysis using the gravity model
Introduction

Tourism is one of the major industries in Greece, which contributes to the economic
welfare of the country. The income generated by direct travel and tourism industries
plus the indirect and induced contributions account for 16.5 per cent of Gross
Domestic Product (GDP) and 18.4 per cent of total employment in Greece (WTTC,
2011). Furthermore, receipts from tourism activity contribute to finance part of the
current account deficit of the balance of payments. In the period 2000-2011 the
surplus on the tourism account amounted to 4 per cent of GDP, on average, whereas
the total deficit of the current account was 9.5 per cent of GDP.

The ongoing global financial crisis of 2007 and the consequent debt crisis had an
important impact on the Greek economy affecting the tourism sector as well. Both the
number of tourist arrivals and tourism receipts declined after 2007, capturing the
consequences of the country’s abnormal economic and social environment. Yet, in
2011 the main tourist indicators showed a substantial improvement allowing the
policy makers to be more optimistic and believe that tourism could play an important
role to alleviate the financial and debt problems and boost Greece’s economic
recovery.

Given the important role of the tourism industry in Greece, and given the perspective
of the credit and debt crisis, this paper examines empirically the factors that influence
the international demand for tourist services in Greece. The analysis is based on a
panel dataset of 31 countries, which constitute nearly 90 per cent of the total tourism
demand in Greece during the period 2001-2010. The demand for tourism is estimated
using gravity equations taking the number of tourist arrivals as a dependent variable.
In comparison with previous empirical studies of international tourist flows to
Greece1, this paper differs in that it utilizes a larger dataset, includes more countries
and uses additional explanatory variables. A second contribution of this study is that
we elaborate on the previous literature by considering not only the relative prices but
also the comparative price levels in the analysis of foreign tourist demand. This is

-2-
done in order to capture differences in the cost of living between destination and
origin countries.

The empirical results show that trade and distance between Greece and the tourist
generating countries are the main factors explaining tourist flows to Greece. The
political stability in the country as well as the relative price levels between Greece and
competing tourist destinations, such as Turkey, Cyprus and Egypt, are also found to
significantly influence tourist arrivals in Greece. Tourists’ personal income,
infrastructure and relative prices between Greece and tourism generating countries are
less important factors in explaining the tourism demand. Interestingly, the 2004
Olympic Games in Athens had a negative impact on tourist flows to Greece.

The rest of the paper is organized as follows. The next section surveys the literature
on international tourism demand and gives a brief discussion of the factors affecting
foreign tourism. The model and the variables used in the empirical analysis are
specified in the third section. The fourth section discusses the estimation method and
presents the results. Finally, the last section concludes and discusses the implications
for tourism policy in Greece.

Demand for tourism: a survey of the literature

Measuring Tourist Demand

Empirical analyses of the demand for tourism generally take the theory of consumer
demand as a starting point. A demand function of the type Q = f(X) is specified,
where Q is the tourism demand and X is a vector of explanatory variables such as
income, distance and price levels that explain Q. Song and Witt (2000) and Proenca
and Soukiazis (2005) consider tourism demand as the aggregate amount of a set of
tourist products and services that the visitors are willing to buy during the period of
their vacation.

The literature distinguishes three main ways to measure foreign demand for tourism.
The first way is to take the total number of arrivals of non-resident tourists at national
borders as measure for tourism demand (see Akis, 1998; Stucka, 2002; Naude and
Saayman, 2005; Phakdisoth and Kim, 2007). Garín-Muñoz (2006) and Dritsakis and

-3-
Athanasiadis (2000) use tourist arrivals per capita to capture the volume of tourism in
Canary Islands and Greece, respectively. Leitao (2010) measures the demand for
tourism using the number of visitors staying in hotels.

As the entrance of travellers from each origination does not take into account their
stay-duration, a second way to measure tourism demand is to consider the number of
nights spent by tourists in the destination country. In a study of international tourist
flows to Spain, Garín-Muñoz and Amaral (2000) use the number of per capita
overnight stays in hotels by each tourist generating country as a measure of tourism
demand. In a similar way Athanasapoulos and Hyndman (2007) model the Australian
domestic tourism demand. A shortcoming of this approach is that it excludes a
considerable number of tourists - those who stay in their own accommodations or are
hosted by family and friends.

Both tourist arrivals and overnight stays define foreign demands for tourist services
without considering the consumption behaviour of visitors. For this reason some
studies apply, as a third way of measurement, the total expenditures made by foreign
tourists as a proxy for tourism demand (see Loeb, 1982; Gonzalez and Moral, 1995;
Tse, 2001). Mello et al. (2002) define tourist demand in the UK as the share of
tourism expenditures of the sending country to alternative destinations. Proenca and
Soukiazis (2005) estimate the demand for tourism in Portugal as the share of the
spending of each origin country to the total tourism expenditures in the destination
country.

Ideally, one would want to measure foreign tourism demand by a combination of the
three approaches mentioned above. In reality, however, this appears unfeasible due to
the great complexity involved in constructing data for such a variable. Furthermore,
data on tourism expenditure are rarely available. Song and Li (2008) and Lim (1997),
who carried out a review of more than 100 empirical studies of international tourism,
conclude that the total number of tourist arrivals remains the most commonly used
method for measuring tourism demand. The present study follows this strand of
literature.

Determinants of Foreign Tourism Demand

-4-
Early empirical studies on tourism demand underline the importance of visitors’
purchasing power for the demand of international tourism services (Gray, 1966;
Kwack, 1972; Loed, 1982). Also in more recent studies (Garín-Muñoz and Amaral,
2000; Song et al., 2003a; Leitao, 2010) income is found to have a strong explanatory
power in the tourism demand function. Higher income leads to more demand for
tourism services. According to Garín-Muñoz (2006), Proenca and Soukiazis (2005)
and Ledesma-Rodríguez et al. (1999) tourism is a luxury good as its share in
consumption spending increases more than proportionally when real income rises.
That is, the income elasticity of demand exceeds unity. On the other hand,
Phakdisoth and Kim (2007) and Habibi et al. (2009) find that tourist flows to Laos
and Malaysia, respectively, are inelastic and hence not regarded by travellers as a
luxury good. This suggests that income elasticity of tourism demand is country
specific and no generalization can be made about its value.

Apart from income, prices are another dominant factor that is found to influence the
international tourism demand (see Gray, 1966; Kwack, 1972; Loed, 1982; Gonzalel
and Moral, 1995; Song et al., 2003b). According to Walsh (1996) the price of tourism
includes three basic components: transportation costs, exchange rate costs and costs of
living in the destination country. While kilometric distance between countries is
usually employed as a proxy for transportation costs (Khadaroov and Seetanah, 2008;
Archibald et al, 2008; Görmüs and Göçer, 2010), the influence of price changes on
international travel is far more complex. The Consumer Price Index (CPI) is widely
used as an indicator for the tourists’ cost of living in the holiday destination (Martin
and Witt, 1987; Morley, 1994). Most empirical studies use the ratio of destination and
tourist generating CPIs adjusted by the exchange rate to measure the differences
between countries’ price levels (see Dritsakis, 2004; Garín-Muñoz and Martin, 2007).
Furthermore, Song et al (2003) and Görmüs and Göçer (2010) employ the ratio of
CPIs in the destination and alternative tourist competing countries in order to capture
substitution price effects.

Many other factors may affect the demand for international tourism, such as the
population of the tourist generating countries and trade relations between the
countries. The population size of the sending country can be an important factor, since
a rise in the number of people living in the origin countries will increase the potential

-5-
consumers of the tourist services. Additionally, strong trade ties between host and
origin countries may translate into advanced transport connections and higher
promotion of the tourist product, facilitating travellers’ flows between trade partners
(see Eilat and Einav, 2004; Phakdisoth and Kim, 2007; Leitao, 2010).

Other factors that influence the choice of a holiday destination are the weather and
climate. Lise and Tol (2002) and Martin (2005) investigate the effect of these factors
on foreign tourism demand and find that their importance is significant. Furthermore,
Seddighi et al. (2001) and Neumayer (2004) emphasize the importance of political
instability and violence for the international demand of tourism.

Additionally, infrastructure in the destination country and other factors related to

transportation and communication, which make the daily life of visitors more
convenient, may also explain a part of the tourism demand (see Khadaroov and
Seetanah, 2008; Archibald et al, 2008; Pulina and Biagi, 2010). Finally, negative
shocks caused by (civil) wars, natural disasters, epidemic diseases and financial crises
will have a substantial influence on tourism demand.

Data Source and Model Specification

The Gravity Model

Our empirical analysis of tourist flows to Greece uses a gravity model. The gravity
model traces its origins to the Newton’s law of gravitation, which states that attraction
between any two objects is proportional to their masses and inversely proportional to
the distance between them. The general specification of the gravity model reads:

mi m j
Fi, j = G (1)
di2, j

where F is the gravitational force between any two objects i and j; mi and mj are the
masses of these two objects, respectively, and d is the distance between them. The
term G refers to a universal gravitational constant.

The gravity model was introduced into international trade by Tinbergen (1962). He
used the model to predict bilateral trade flows between countries on the hand of

-6-
countries’ economic sizes and the distance between them. Since its introduction in
economics, the gravity model has become a common tool for the analysis of trade
flows2.

Rewriting equation (1) in logarithmic form and in terms of tourist flows gives:

lnYi,t = α + µi + β lnX’i,t + εi,t (2)

where Y stands for international demand for tourist services in Greece; X’ is a vector
of variables explaining the Greek tourism demand; µ refers to unobservable country-
specific effects; α is an intercept and ε is the idiosyncratic error term that is
uncorrelated with µ. The terms i and t indicate, respectively, tourism generating
country and year of observation.

The estimation of equation (2) is based on a panel dataset covering 31 countries3,

which constitute nearly 90 per cent of the total tourism demand in Greece. The
analysis uses annual data over the period 2001-2010. Since Greece is mainly a
summer holiday destination, annual data are preferred in order to avoid seasonality
problems.

Model Variables

The Dependent Variable

The dependent variable, Yi,t, is measured by the total non-resident tourist arrivals to
Greece from country i in year t. As discussed before, this analysis uses the number of
tourist arrivals since it appears the best way of to measure tourism demand, given the
available data. The data for tourist arrivals come from the World Tourism
Organization (UNWTO) and cover the period 2001-2010.

The vector X’ includes a number of explanatory variables which are expected to

influence the international demand for Greek tourist services.

Real Personal Income

The most important determinant of tourists’ decision to travel abroad is the level of
their personal income. It is approximated in this analysis by GDP per capita of the

-7-
origin countries4. GDP is measured in constant 2005 Purchasing Power Parities. The
income elasticity of tourism demand is expected to be positive. The GDP data are
taken from the World Bank’s World Development Indicators and cover the years
2001-2010.

Relative and Competitive Prices

The cost of living in the destination country is another factor that may influence the
international tourism demand. For this reason we use a proxy of relative prices
between Greece and the tourist generating countries. It is a general practice in
empirical research to calculate relative prices as the ratio of the CPIs of the
destination and the sending countries, adjusted by the exchange rate (see Martin and
Witt, 1987; Morley, 1994; Proenca and Soukiazis, 2005; Aslan et al., 2009).

Even though this measure of relative prices is commonly accepted and widely used in
the literature, it has one main limitation. The ratio of CPIs captures relative price
levels changes across countries, but it does not say anything about the level of prices.
In order to account for this limitation we apply two alternative approaches in
measuring relative prices.

Firstly, following the previous literature we calculate the relative prices as:

CPI GR ,t
RPi, t = (3)
CPI i ,t EX i ,t

where CPIGR,t and CPIi,t are the consumer price index of Greece and the origin
country, respectively, and EXi,t is the exchange rate between Greece and the origin
country.

Secondly, we use an alternative measure of relative prices, whereby relative prices are
defined as the ratio of comparative price levels (CPLs) between Greece and the tourist
generating countries. The results of this second approach are reported in the
sensitivity analysis.

Additionally, there is also another type of relative prices that should be taken into
account as determinant of foreign tourism demand, namely that between Greece and
its main competitors as holiday destinations. These competitive prices are defined as
the ratio of the Greek CPI to the CPI of the competing destinations. According to the

-8-
Association of Greek Tourism Enterprises, the main competitors of the Greek tourist
product are Spain, Turkey, Croatia, Egypt, and Cyprus due to their geographical and
cultural similarities with Greece. In this analysis we consider also Portugal and Italy
as competing countries. A decrease in the price level of Greece relative to the sending
countries or to the alternative destinations would be an incentive for more tourists to
visit Greece. Therefore, a negative sign for both relative and competitive prices is
expected. These relative price indices are constructed using data on prices and
exchange rates which are extracted from the financial statistics of the International
Monetary Fund (IMF) and the European Central Bank, respectively.

Trade ties

As discussed, bilateral trade relationships may also facilitate the tourist flows between
countries. Trade can serve as an informational knowledge platform between Greece
and the origin countries. Following Phakdisoth and Kim (2007) and Leitao (2010) the
trade value between Greece and each of the origin countries is calculated as follows:

X i ,t  M i ,t
TRVi, t = (4)
GDPGR ,t  GDPi ,t

where Xi,t refers to the annual exports of Greece to each tourist generating country;
Mi,t represents the annual imports of Greece from each origin country; GDPGR,t and
GDPi,t stand, respectively, for GDP per capita of Greece and the tourist generating
country. GDP per capita is measured in PPP and come from the World Bank. The
value of imports and exports is extracted from the trade statistics of IMF.

Distance

The transportation cost is another aspect that can influence the tourists’ decision of
whether or not to visit Greece. The variable to capture the cost of travel to Greece in
this paper is the kilometric distance between Athens and the capitals of all tourist
generating countries. Despite its shortcoming as a measure of travel prices, distance is
widely used in the literature as an effective proxy for transportation costs. We expect
a negative sign for the coefficient of this variable. The data is taken from the CEPII
(French Institute for Research on the World Economy) distance database.

-9-
Political Stability

The global financial crisis that erupted in 2007 further deteriorated the already
problematic macroeconomic fundamentals of the Greek economy and had serious
impacts on the political stability of the country. The general strikes and mass
demonstrations, accompanied often by violent riots, as well as the repeated national
elections are considered to have a negative effect on international tourist arrivals. In
order to control for the effect of political stability on tourist arrivals, we employ the
World Bank’s governance indicator of Political Stability and Absence of Violence.
The indicator measures the perceptions of the likelihood that the government will be
destabilized or overthrown by unconstitutional or violent means, including domestic
violence and terrorism. The decrease in this indicator in the years after 2007 confirms
the political instability of the country. Hence, we expect a positive sign for this
variable (see Naude and Saayman, 2005; Kareem, 2009).

Other Supply Factors

Aside from political stability there are also other supply conditions in the destination
country that affect foreign tourist demand. Examples include infrastructure facilities
such as airports, public transportation, telecommunication and other informational
services. The sufficient provision of infrastructure and travel facilities makes the daily
life of tourists more comfortable and safe, and thus influences the tourists’ holiday
destination choice. In other words, it reduces the transaction costs of tourists when
travelling. In order to measure the effect of supply services on foreign tourism
demand, the analysis employs the Gross Investment Spending in Infrastructure (GISI)
as a proxy. Due to data availability we focus specifically on transport infrastructure,
which is arguably the most important type of infrastructure for tourists. GISI is a
composite indicator. It is calculated as a sum of gross investment spending in airports
infrastructure, roads and rail-roads infrastructure and sea ports infrastructure. This
variable is applied with a one year lag in order to account for time lags associated
with the realization of long-term investments, such as the construction of motorways
etcetera. The data are extracted from OECD Statistics and are measured in million
Euros. The data for this variable cover the period 2001-2008.

- 10 -
Dummy Variable

Finally, we include in the vector X’ a dummy variable for the 2004 Olympic Games
which took place in Athens. The Olympic Games are a popular sport event which
attracts innumerable visitors from all over the world. We expect a positive sign for
this variable. Table 1 summarizes the variables and the data sources used to estimate
equation (2).

Table 1: Overview variables and data sources

Variables Definition Data source

TAi,t Number of tourist arrivals from origin country i to Greece at time t World Tourism Organization

GDPi,t GDP per capita of sending country i at time t World Bank’s WDI

TRVi,t Trade value between Greece and each origin country i at time t Calculated IMF data, DOTS

Disi Distance between Greece and each tourist generating country i CEPII

RPi,t Relative price level between Greece and origin country i at time t Calculated IMF data, IFS

CPSPAt Competitive price level between Greece and Spain at time t Calculated IMF data, IFS

CPPORt Competitive price level between Greece and Portugal at time t Calculated IMF data, IFS

CPTURt Competitive price level between Greece and Turkey at time t Calculated IMF data, IFS

CPCYPt Competitive price level between Greece and Cyprus at time t Calculated IMF data, IFS

CPEGYt Competitive price level between Greece and Egypt at time t Calculated IMF data, IFS

CPCROt Competitive price level between Greece and Croatia at time t Calculated IMF data, IFS

CPITAt Competitive price level between Greece and Italy at time t Calculated IMF data, IFS

PolStt Indicator of political stability in Greece in year t World Bank’s WGI

GISIt-1 Gross investment spending in transport infrastructure (airports, OECD Statistics

roads, rail-roads and sea ports) in t-1

D2004 Dummy variable for the Olympic Games in Athens in 2004 Self-elaborated

COMPRi,t Comparative price levels between Greece and tourist generating OECD/Eurostat/PennWorldTable
country i at time t

Estimation of International Tourism Demand in Greece

Dynamic Estimation

- 11 -
The panel structure of our data allows us to estimate equation (2) using panel data
methods. Panel methods have an advantage over cross-sectional methods, as they
make it possible to account for unobserved country-specific effects which may
influence foreign tourism demand, such as climate, weather, cultural and historic
heritage, hospitality of the destination country, historic ties between origin and
destination countries etcetera. Most of these factors are difficult to include explicitly
in the analysis, as they are not observed in the data. Assuming these factors are
country-specific and time-invariant, one way to account for them is to employ panel
data estimation techniques.

Equation (2) can be estimated using both static and dynamic panel methods. In this
analysis a dynamic estimation is applied in order to account for econometric problems
(associated with static estimation) such as autocorrelation, unit roots and endogeneity5.
Furthermore, the dynamic approach of tourism demand allows us to distinguish
between short term and long term effects, as recognition lags are explicitly included in
the specification by a lagged dependent variable. Rewriting equation (2) in a dynamic
form yields:

where TAi,t-1 is the lagged dependent variable.

The error term is now correlated with the lagged dependent variable, making the
static estimation techniques (i.e. fixed and random effects) inappropriate, since the
estimation results they provide would be biased and inconsistent. A way to overcome
this problem is to find a valid instrument for the lagged value of dependent variable.
The solution provided by Arellano and Bond (1991) is the difference Generalized
Method of Moment (GMM) estimation procedure (see Munoz, 2006; Phakdisoth and
Kim, 2007)6. The authors argue that lagging the lagged dependent variable by two or
more periods provides a valid instrument for this variable. Rewriting equation (5) in
terms of differences yields:

where and similarly for the rest of the variables.

- 12 -
An advantage of this method is that the first-differencing procedure eliminates the
unobserved country-specific effects . Moreover, by first-differencing all variables

the possible presence of non-stationarity is removed. A problem arises with the use of
Arellano and Bond’s (1991) difference estimator when the lagged values of the
explanatory variables turn to be weak instruments for the instrumented endogenous
variables, especially when the variables follow a random walk.

For this reason, the present study employs the system GMM estimator outlined by
Arellano and Bover (1995) and developed by Blundell and Bond (1998) (see Leitao,
2010). Using the system GMM estimator we obtain a system of two equations, the
original equation in levels and a differenced one. In this way additional instruments
can be used to increase efficiency, while the variables in levels are instrumented with
their own first differences7. An assumption required in this case is that first-
differenced instruments are uncorrelated with the unobserved country-specific effects.

The dynamic estimation with a lagged dependent variable can be interpreted as a

distributed lag, representing partial adjustment or adaptive expectations mechanisms.
The log-linear specification allows us to interpret the estimated coefficients as short-
run elasticities. In order to calculate the long-run elasticities each coefficient has to be
multiplied by the recognition lag: .

Estimation results of dynamic model

Table 2 reports the empirical results from the estimation of equation (6) applying the
GMM-system estimator. The eight columns in the table refer to eight different
specifications of the tourism demand function. Model (1) is our basic specification. In
this model foreign income, trade relations, distance and prices are used to explain the
international demand for Greek tourism services. Models (2) to (7) extend the basic
model by including one or more additional determinants of foreign tourism demand.
Moreover, this procedure serves as a robustness check for our baseline results. The
sensitivity of the estimated results is further scrutinized in model (8), where we
employ a different measure of relative prices. Model (8) is separately discussed in the
next-sub-section.

- 13 -
Table 2: Estimation results of GMM-system estimator
(1) (2) (3) (4) (5) (6) (7) (8)

Constant 2.38 1.55 4.33 3.99 3.04 0.80 2.92 0.62

(1.82) (1.22) (2.03) (2.74) (2.38) (0.39) (2.30) (0.29)
lnTAt-1 0.78 0.81 0.75 0.78 0.74 0.78 0.74 0.80
(13.58)*** (14.38)*** (8.25)*** (9.76)*** (14.27)*** (12.71)*** (13.21)*** (16.81)***

lnGDP 0.21 0.17 0.17 0.10 0.22 0.21 0.22 0.33

{0.95} {0.89} {0.68} {0.45} {0.84} {0.95} {0.85} {1.65}
(1.74)* (1.53) (1.50) (0.81) (1.66)* (1.56) (1.59) (1.70)*

lnTRV 0.12 0.10 0.14 0.12 0.14 0.12 0.14 0.12

{0.54} {0.52} {0.56} {0.54} {0.54} {0.54} {0.54} {0.60}
(2.71)*** (2.64)*** (2.45)** (2.02)** (3.01)*** (2.51)** (2.50)** (2.89)***

lnDis -0.16 -0.13 -0.15 -0.12 -0.18 -0.17 -0.18 -0.15

{-0.72} {-0.68} {-0.60} {-0.54} {-0.69} {-0.77} {-0.69} {-0.75}
(-2.09)** (-1.82)* (-1.77)* (-1.47) (-2.03)** (-1.99)** (-2.18)** (-2.20)**

lnRP -0.008 -0.010 0.014 -0.021 -0.011 -0.009 -0.010 -

{-0.04} {-0.05} {-0.06} {-0.09} {-0.04} {-0.04} {-0.04}
(-0.23) (-0.31) (-0.36) (-0.57) (-0.27) (-0.24) (-0.25)

lnPolSt 0.13
{0.68}
(2.43)**
lnGISIt-1 0.16
{-0.64}
(-1.61)
lnCPTUR -0.85
{-3.86}
(-2.39)**

lnCPCYP -0.32
{-1.45}
(-3.24)***

lnCPEGY -0.29
{-1.31}
(-2.18) ***

lnCPSPA 0.79
{3.04}
(0.28)

lnCPOR -0.17
{-0.65}
(-0.13)
lnCPCRO 0.82
{3.72}
(1.02)

lnCPITA 0.52
{2.36}
(0.73)
Dum2004 -0.14
{-0.54}
(-1.73)*
lnCOMPR 0.15
{0.75}
(0.70)
AR(2) 0.35 0.40 0.78 0.80 0.34 0.33 0.68 0.39
[0.726] [0.688] [0.435] [0.423] [0.736] [0.739] [0.498] [0.695]

Hansen test (d.f.) 29.56 29.47 26.60 24.53 29.48 28.86 29.73 30.23
(28) (28) (20) (28) (28) (28) (28) (28)
[0.384] [0.389] [0.147] [0.654] [0.388] [0.420] [0.376] [0.352]
F test (d.f.) 126.09 138.40 74.97 84.47 106.49 140.23 129.72 177.02
(5,30) (6,30) (6,30) (6,30) (7,30) (7,30) (6,30) (5,30)
[0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]

- 14 -
Number of instruments 34 35 27 37 36 36 35 34

Number of groups 31 31 31 31 31 31 31 31

Observations 279 279 217 279 279 279 279 279

Notes: Dependent variable – number of tourist arrivals. All variables are converted in their logarithmic
form (except the dummy variable for Olympic games). T-values are presented in ( ), P-values of the
respective tests reported in [ ], long-run elasticities in { }. Statistical significance at 10, 5 and 1 per cent
indicated by *, ** and ***, respectively. AR(2) is Arellano and Bond test for second-order
autocorrelation which has a hull hypothesis of no second-order serial correlation it the residuals.
Hansen test of overidentifying restrictions tests the null hypothesis of exogeneity of the instrumental
variables. F test of the joint significance of the explanatory variables tests the null hypothesis that the
estimated coefficients are jointly equal to zero. In all eight specifications lnTAt-1 and lnGDP variables
are considered as endogenous.

Model (1) shows that the demand for Greek tourism services is positively affected by
foreign income and trade between Greece and the origin countries, and negatively
affected by the distance between Greece and the sending countries. The coefficient
value of the lagged dependent variable suggests a substantial recognition lag of almost
4.5 years for tourists visiting Greece. Further, the estimation results show that an
increase in the personal income of tourists by 1 per cent will have a positive effect on
tourism demand increasing it by 0.21 and 0.95 per cent in the short-run and long-run,
respectively. Considering that tourism is a consumer product, it is in line with
consumer theory that a rise of income leads to more demand for this product. When
income rises people are willing to acquire more of this product. The fact that both
income elasticities are lower than unity suggests that tourist services in Greece are not
considered as a luxury good. However, the disposable income of tourists does not
appear statistically significant in each model that we estimate. It is interesting to see
how this result compares to other empirical studies that attempt to capture the effect
of tourists’ income. The empirical literature shows mixed results when it comes to the
power of foreign income. Dritsakis and Athanasiadis (2000) estimate the demand
functions for each of the main tourist generating countries to Greece. The empirical
results show that foreign income is not as an important determinant of tourist arrivals
as it would be expected according to the theory. Moreover, another study of Dritsakis
and Gialitaki (2004) found that the real income of travellers has a positive and
profound impact on tourist arrivals from USA to Greece. On the other hand,
examining the German and British tourism demand in Greece Dritsakis (2004)
receives statistically significant coefficients for the income variable although with a

- 15 -
negative sign. The author explains that tourists with higher income may prefer an
alternative holiday destination to Greece.

Also, the trade flows between Greece and tourist generating countries are found to
significantly affect the demand for tourism. The export and consumption of Greek
products abroad is just another way for foreign residents to “discover” Greece as a
potential holiday destination. Trade reduces the cultural distance between countries,
making residents of one country more willing to visit another country. The estimation
results in model (1) show that an increase in trade flows by 1 per cent will result, on
average, in 0.12 per cent additional foreign visitors to Greece in the short-run and
0.54 in the long-run. Distance, on the other hand, works in the opposite direction as
countries further apart send on average fewer tourists to Greece. This is a logical
result, because distance acts in this case as an indicator of transportation costs, or
more broadly, of transaction costs. All else being equal, the higher the transportation
costs are, the less likely people will be to undertake the journey and visit the country
in question.

Finally, the relative prices are found to have a zero effect on tourism. Although the
coefficient has the expected negative sign, it is not statistically different from zero,
indicating that relative prices appear not to have a substantial impact on tourist
arrivals. This can be explained by the fact that the main customers of Greek tourist
product are countries with relatively high standards of living. Therefore, the decision
to visit Greece is determined by the level of personal income rather than by the
relative cost of living.

Model (2) extends the basic specification by including one additional variable
measuring political stability and absence of violence in Greece. The coefficient of this
variable has a positive sign, albeit statistically insignificant at 5 per cent level. This
suggests that the political instability and social unrest in Greece have a negative short-
run impact on the international tourism demand, while this effect is even more harmful
in the long-run.

The estimated results in model (3) show that gross investment spending in transport
infrastructure does not appear to have a beneficial impact on tourist arrivals. One
possible explanation for this result is that the impact of infrastructure on tourism is
captured to some extend by the country specific effects.

- 16 -
Model (4) to (6) consider the effect of the relative prices between Greece and
alternative holiday destinations on the demand for Greek tourism services. We
distinguish three groups of major tourist competitors of Greece. The distinction is
based on geographical and cultural similarities between Greece and these competitors.
The estimated results show that the competitive prices between Greece and Turkey,
Cyprus and Egypt have a negative and significant impact on tourist flows to Greece.
An increase in the price ratio between Greece and these countries will result in a drop
in tourist arrivals to Greece. This effect is even stronger in the long-run than in the
short-run, as it is shown by the estimated coefficients. All else being equal, consumers
(who behave rationally and possess full information) will be more likely to visit the
destination that offers the same services at lower prices.

It seems that the seaside resorts of Cyprus and the low priced tourist services in
Turkey attract a considerable number of tourists during the summer months, affecting
negatively the demand for tourist services in Greece. Moreover, the low tourists’ cost
of living in Egypt in combination with the warm climate in the coastal regions and its
distinct cultural identity, make the country one of the main tourist competitors of
Greece. The competitor price levels seem to play a minor role in the cases of Spain,
Portugal, Italy and Croatia as all four coefficients are not statistically significant.

Model (7) includes a dummy variable for the 2004 Olympic Games in Athens.
Contrary to the expectations, the coefficient of this variable has a negative sign and it
is significant at the 10 per cent level. One possible explanation for this result could be
that tourists who otherwise would have visited Greece decided to visit another
country, fearing higher prices, overcrowded conditions or even lack of adequate
security measures in Greece during the summer of the Olympic Games. It is
interesting to note that a similar negative effect was found for the recent Olympic
Games in London. In the summer of 2012 the UK’s capital experienced a dramatic
drop of nearly 30 per cent in tourist arrivals as compared to previous years (European
Tour Operators Association, 2012). Examining the impact of the Olympic Games on
international tourism flows, Garín-Muñoz and Amaral (2000) found that the 1992
Olympic Games in Barcelona had an insignificant effect on the demand for tourist
services in Spain.

- 17 -
Finally, Table 2 presents Arellano and Bond second-order autocorrelation test
(AR(2))8 as well as Hansen tests of overidentifying restrictions. The tests do not detect
any serial correlation in the residuals and confirm the validity of the instruments,
respectively. These test statistics indicate that our econometric methodology is
justified.

A sensitivity analysis on price effects

This study utilizes the relative prices between Greece and the sending countries to
capture differences in tourists’ cost of living. This is in line with the literature on
international tourism demand. However, a drawback of the relative prices, as defined
here, is that they do not provide any information about how expensive tourist
generating countries are relative to Greece. The relative prices only show whether the
Greek prices change relatively faster as compared to the foreign prices9.

For this reason we consider an alternative measure of relative prices, namely the ratio
of Comparative Price Levels (CPLs) of Greece and the tourist generating countries.
CPLs are calculated as a ratio of purchasing power parities for final consumption
expenditure to exchange rates. In this way they provide a direct measure of price
levels differences between countries. The ratio of Greek to foreign CPLs, therefore,
indicates whether Greece is relatively cheap/expensive compared to the foreign
country.

Model (8) in Table 2 presents the estimation results for this new variable (COMPR).
Similarly to the relative prices, the coefficient of comparative prices is not statistically
significant at any level. Furthermore, model (8) shows that the inclusion of this new
variable has no influence on the estimated effects of the past tourist arrivals, GDP,
trade and distance. The coefficients of these variables remain largely unchanged in
terms of their values and statistical significance throughout the various specifications.
This suggests that the estimated effects are quite robust to the inclusion of CPLs in the
model. Furthermore, the estimated results in model (8) are very similar in terms of
magnitude, sign and significance to the results in model (1).

- 18 -
Concluding Remarks and Policy Recommendations

This paper examines the main determinants of international tourism demand in

Greece. Using data on tourist flows from 31 countries to Greece the analysis shows
that distance and trade ties between Greece and the sending countries are important
factors influencing foreign demand for tourist services. In contrast to empirical studies
of international tourism demand for other destinations, personal income of tourists
does not appear statistically significant in each model. The relative prices and the
proxy of infrastructure and other facilities are not statistically significant at any level,
while the coefficient of political stability exhibits a positive effect on tourism. The
estimated results show a mixed picture for the competitors of Greece in the market for
international tourist services. The relative prices between Greece and Spain, Portugal,
Italy and Croatia appear to be statistically insignificant predictors of tourism demand,
while the corresponding coefficients for Turkey, Cyprus and Egypt do seem to
significantly influence tourism in Greece. The 2004 Olympic Games in Athens had a
negative effect on the international tourist arrivals in that year, a finding which seems
to corroborate with similar findings for the recent London Olympics.

A major finding is that the demand for tourism in Greece does not appear to be a
luxury service for the relatively rich (European) countries in our sample, neither in the
short-run nor in the long-run. A policy recommendation in that respect is that the
Greek tourist industry needs to pay more attention to attract visitors from fast growing
large economies (e.g. Russia, China), where incomes rise rapidly and a holiday to
Greece may still be a luxury service. Moreover, since distance appears to be one of the
most important determinants of tourist demand in Greece, it might be beneficial for
Greece to provide economical tour packages including the transportation costs in order
to attract more visitors. Additionally, as trade ties is another essential determinant of
tourism, Greece has to specifically improve its trade relationships with countries
which are not yet well established clients of the Greek tourist product. Another point is
the competitiveness of the Greek tourism industry. Setting of competitive prices is a
necessary condition to sustain and improve the position of Greek tourism in the global
market. Furthermore, since external shocks such as political instability have a negative
impact on foreign tourist demand, the strategies of policy makers should be focused on

- 19 -
the mitigation of social unrest and massive strikes improving in this way the country’s
image abroad.

Future research needs to further investigate the income elasticity of different groups of
sending countries, providing tourism policy with more focus. The empirical analysis in
this paper can be further extended by using a proxy of marketing expenditures for
tourism, which is not used in this analysis due to data unavailability, in order to
capture the impact of the country’s promotion of Greek tourism services. Finally, the
use of an alternative proxy of supply factors and facilities, apart from investment
spending in transport infrastructure, would be an interesting extension of this study.

References
Apostolakis, A. and Jaffry, S. (2009) ‘Examining expenditure patterns of British
tourists to Greece’, International Journal of Tourism Policy, Vol. 2, No. 3, pp. 187 –
205.
Akis, S. (1998) ‘A Compact econometric model of tourism demand for Turkey’,
Tourism Management, Vol. 19, No. 1, pp. 99–102.

Archibald, X., LaCorbinière, J. and Moorev W. (2008) ‘Analysis of Tourism

Competitiveness in the Caribbean: A Gravity Model Approach’, Presented at the 29th
Annual Review Seminar Research Department Central Bank of Barbados.

Arellano, M. and Bond, S. (1991) ‘Some tests of specification for panel data: Monte
Carlo evidence and an application to employment equations’, Review of Economic
Studies, Vol. 58, pp. 277–297.

Arellano, M. and Bover, O. (1995) ‘Another Look at the Instrumental Variable

Estimation of Error-Components Models’, Journal of Econometrics, Vol. 68, No. 1,
pp. 29-51.

Aslan, A., Kula, F. and Kaplan, M. (2009) ‘International Tourism Demand for
Turkey: A Dynamic Panel Data approach’, Research Journal of International Studies,
Vol. 29, pp. 65–73.

Athanasopoulos, G. and Hyndman, R. J. (2008) ‘Modelling and forecasting Australian

domestic tourism’, Tourism Management, Vol. 29, pp. 19–31.

Baltagi, H. B. (2005) Econometric Analysis of Panel Data, John Wiley & Sons Ltd.
West Sussex, England.

Blundell, R. and Bond, S. (1998) ‘Initial conditions and moment restrictions in

dynamic panel data models’, Journal of Econometrics Review, Vol. 87, No. 1, pp.
115-143.

De Mello, M. M., Pack, A. and Sinclair, M. T. (2002) ‘A system of equations model

of UK tourism demand in neighbouring countries’, Applied Economics, Vol. 34, pp.
509–521.

- 20 -
Dritsakis, N. and Athanasiadis, S. (2000) ‘An econometric model of tourist demand:
The case of Greece’, Journal of Hospitality and Leisure Marketing, Vol. 7, pp. 39–49.

Dritsakis, N. and Gialitaki, A. (2004) ‘Seasonal Tourism Demand Models from USA
to Greece’, Tourism Recreation Research, Vol. 29(3), pp. 7–13.

Dritsakis, N. (2004) ‘Cointegration analysis of German and British tourism demand

for Greece’, Tourism Management, Vol. 25, pp. 111–119.

Eilat, Y. and Einav, L. (2004) ‘The Determinants of International Tourism: A Three -

Dimension Panel Data Analysis’, Applied Economics, Vol. 36, pp. 1315-1327.

European Tour Operators Association (2012) British economy struggles to secure

golden Olympic boost, URL: http://www.etoa.org/news/2012/08/20/british-economy-
struggles-to-secure-golden-olympics-boost, downloaded on 01 October 2012.

Farsari, Y., Butler, R. and Prastacos, P. (2007) ‘Sustainable tourism policy for
Mediterranean destinations: issues and interrelationships’, International Journal of
Tourism Policy, Vol. 1, No. 1, pp. 58-78.

Garín-Muñoz, T. (2006) ‘Inbound international tourism to Canary Islands: A dynamic

panel data model’, Tourism Management, Vol. 27, pp. 281–291.

Garín-Muñoz, T. and Montero-Martin, L. F. (2007) ‘Tourism in the Balearic Islands:

A dynamic model for international demand using panel data’, Tourism Management,
Vol. 28, pp. 1224–1235.

Garín-Muñoz, T. and Amaral, T. (2000) ‘An Econometric Model for International

Tourism Flows to Spain’, Applied Economics, Vol. 7, pp. 525-529.

González, P. and Moral, P. (1995) ‘An Analysis of the International Tourism Demand
in Spain’, International Journal of Forecasting, Vol. 11, pp. 233-251.

Görmüs S. and Göçer, I. (2010) ‘The Socio-Economic Determinant of Tourism

Demand in Turkey: A Panel Data Approach’, International Research Journal of
Finance and Economics, Issue 55, pp. 88-99.

Gray, H. P. (1966) ‘The Demand for International Travel by United States and
Canada’, International Economic Review, Vol. 7, No. 1, pp. 83-92.

Habibi, F., Rahim, K. A, Ramchandrau, S. and Chin L. (2009) ‘Dynamic Model for
International Tourism Demand for Malaysia: Panel Data Evidence’, International
Research Journal of Finance and Economics, Issue 33, pp. 207-217.

Hansen, L. P. (1982), ‘Large Sample Properties of Generalized Method of Moments

Estimators’, Econometrica, Vol. 50, pp. 1029-1054.

Kareem, O. (2009) ‘A Panel Data Analysis of Demand for Tourism in Africa’. Paper
Presented at the 14th African Econometric Society Annual Conference, held at Cape
Town, South Africa.

Khadaroo, J. and Seetanah, B. (2008) ‘The Role of Transport Infrastructure in

International Tourism Development: A Gravity Model Approach’, Tourism
Management, Vol. 29, pp. 831-840.
Kwack, S. Y. (1972), ‘Effects of income and prices on travel spending abroad, 1960
111-1967 IV’, International Economic Review, Vol. 13, No. 2, pp. 245-256.

- 21 -
Lathiras, P. and Siriopoulos, C. (1998) ‘The Demand for Tourism to Greece: a
Cointegration Approach’, Tourism Economics, Vol. 4, No. 2, pp. 171-185.

Ledesma-Rodrıguez, F. J., Navarro-Ibanez, M. and Perez-Rodrıguez, J. V. (2001)

‘Panel data and tourism: A case study of Tenerife’, Tourism Economics, Vol. 7, pp.
75–88.

Leitao, N. C. (2010) ‘Does Trade Help to Explain Tourism Demand? The Case of
Portugal’, Theoretical and Applied Economics, Vol. XVII, No. 3, pp. 63–74.

Lim, C. (1997) ‘Review of international tourism demand models’, Annals of Tourism

Research, Vol. 24, pp. 835–849.

Lise, W. and Tol, R. (2002) ‘Impact of Climate on Tourism Demand’, Climatic

Change, Vol. 55, pp. 429–449.

Loeb, P. (1982) ‘International travel in the United States: An economic evaluation’,

Annals of Tourism Research, Vol. 9, pp. 42–55.

Louvieris, P. (2002) ‘Forecasting international tourism demand for greece: A

contingency approach’, Journal of Travel & Tourism Marketing, Vol. 13, pp. 21–40.

Manuela Pulina, M. and Biagi, B. (2010) ‘The evolution of tourism demand and
supply: a regional policy study’, International Journal of Tourism Policy, Vol. 3, No.
3, pp. 237 – 256.
Martin, B. G. (2005) ‘Weather, Climate and Tourism: A Geographical Perspective’,
Annals of Tourism Research, Vol. 32, No. 3, pp. 571–591.

Martin, C. A. And Witt. S. F. (1987) ‘Tourism demand forecasting models: choice of

appropriate variable to represent tourists’ cost of living’, Tourism Management, Vol.
8, pp. 233-246.

Morley, C. (1994) ‘The Use of CPI for Tourism Prices in Demand Modelling’,
Tourism Management, Vol. 15, No. 5, pp. 342-346.

Naude, W. A. and Saayman, A. (2005) ‘Determinants of tourist arrivals in Africa: A

panel data regression analysis’, Tourism Economics, Vol. 11, pp. 365–391.

Neumayer, E. (2004) ‘The Impact of Political Violence on Tourism’, Journal of

Conflict Resolution, Vol. 48, No. 2, pp. 259-281.

Ouerfelli, C. (2010) ‘Analysis of European tourism demand for Tunisia: a new

approach’, International Journal of Tourism Policy, Vol. 3, No. 3, pp. 223 – 236.

Phakdisoth, L. and Kim, D. (2007) ‘The Determinants of Inbound Tourism in Laos’,

ASEAN Economic Bulletin, Vol. 24, No. 2, pp. 225- 237.

Proenca, S. A. and Soukiazis, E. (2005) ‘Demand for Tourism in Portugal: A Panel

Data Approach’, Document to de Trabalho Rabalho / Discussion Paper (February),
No. 29.

Seddighi, H. R., Nuttall, M. W. and Theocharous, A. L. (2001) ‘Does cultural

background of tourists influence the destination choice? an empirical study with
special reference to political instability’, Tourism Management, Vol. 22, pp. 181- 191.

- 22 -
Song, H. and Li, G. (2008) ‘Tourism Demand Modeling and Forecasting: A Review
of Recent Research’, Tourism Management, Vol. 29, No. 2, pp. 203–220.

Song, H. and Witt, S. F. (2000) Tourism Demand Modelling and Forecasting –

Modern Econometric Approaches, 1st ed., Cambridge: Pergamon.

Song, H., Wong, K. K. F. and Chon, K. K. S. (2003a) ‘Modelling and forecasting the
demand for Hong Kong tourism’, International Journal of Hospitality Management,
Vol. 22, pp.435-451.

Song, H., Witt, S. F. and Li, G. (2003b) ‘Modeling and Forecasting the Demand for
Thai Tourism’, Tourism Economics, Vol. 9, pp. 363–387.

Stucka, T. (2002) A Comparison of Two Econometric Models (OLS and SUR) for
Forecasting Croatian Tourism Arrivals, Croatian National Bank, W-8/2002.

Tse, R. (2001) ‘Estimating the impact of economic factors on tourism: evidence from
Hong Kong’, Tourism Economics, Vol. 7 No. 3, pp. 277–293.

Walsh, M. (1996) ‘Demand Analysis in Irish Tourism’, Journal of the Statistical and
Social Inquiry Society of Ireland, Vol. 27, Part 4.

Wooldridge, J. M. (2009) Introductory Econometrics: A Modern Approach, 4th ed.,

South-Western Cengage Learning.

Endnotes

1. Lathiras and Siriopoulos (1998), Dritsakis and Athanasiadis (2000), Louvieris (2002) and
Dritsakis (2004).
2. For applications of the gravity model in the tourism literature see Archibald et al. (2008),
Leitao (2010) and Görmüs and Göçer (2010).
3. Albania, Australia, Austria, Belgium, Bulgaria, Canada, Cyprus, Czech Republic,
Denmark, Finland, France, Germany, Hungary, Ireland, Israel, Italy, Japan, Netherlands,
Norway, Poland, Portugal, Romania, Russia, Serbia, Slovakia, Spain, Sweden,
Switzerland, Turkey, United Kingdom, United States.
4. Other studies that use real GDP per capita to approximate living standards in tourist
generating countries are Lathiras and Siriopoulos (1998), Song et al. (2003b), Naude and
Saayman (2005) and Garín-Muñoz and Martin (2007).
5. We started first by estimating a static version of the model, applying both fixed and
random effects methods. The Wooldridge test for autocorrelation (Ho: no first-order
autocorrelation) detected, however, the presence of serial autocorrelation. Since
autocorrelation could exist due to a model misspecification, the rest of the paper follows a
dynamic estimation approach. The empirical results of the static model as well as the
associated tests are available upon request. The variables of the static models were tested
also for unit roots (applying the Harris-Tzavalis, Levin-Lin-Chu, Breitung and Im-
Pesaran-Shin tests). Although the different tests yielded different results, in each case
some of the variables were found to be non-stationary. The test results are available upon
request.
6. GMM estimator was developed and formalized by Hansen (1982). Since then it has been
become one of major estimation methods in economics.

- 23 -
7. In order to reduce the bias coming from a large number of instruments we used only the
second lag of the endogenous variables as an instrument.
8. It is more important to look at AR(2) in first differences, since it will detect serial
correlation in levels.
9. Remember that the relative prices were calculated as the ratio of Greek to foreign CPIs
adjusted by the exchange rate. CPIs, however, are used to measure changes in the price
level of consumer goods and services over time. A CPI of 105 means that the prices have
increased by 5 per cent as compared to the base year. We do not know how high or low
these prices are in comparison to other countries.

- 24 -