Content uploaded by Yang Yang
Author content
All content in this area was uploaded by Yang Yang on Nov 06, 2014
Content may be subject to copyright.
1
THE INFLUENCE OF CULTURAL DISTANCE ON
CHINA INBOUND TOURISM FLOWS: A PANEL
DATA GRAVITY MODEL APPROACH
YANG YANG *
Ph.D Student in Geography
Department of Geography, University of Florida, Gainesville, FL, 32611, U.S.A.
E-mail: yang.yang@ufl.edu
KEVIN. K.F. WONG
Associate Professor in Tourism Economics
School of Hotel and Tourism Management, Hong Kong Polytechnic University, Hung
Hom, Hong Kong. E-mail: hmkevinw@polyu.edu.hk
Submitted exclusively to Asian Geographer for publication consideration.
*Corresponding author
Yang Yang, (e-mail:yang.yang@ufl.edu, phone: 1-352-8716187), is a Ph.D student in
Department of Geography, University of Florida. His address is Department of
Geography, University of Florida, Gainesville, FL, 32611, U.S.A..
Kevin K.F. Wong, PhD (e-mail: hmkevinw@inet.polyu.edu.hk, phone:
852-2766-6341), is an associate professor in the School of Hotel and Tourism
Management, the Hong Kong Polytechnic University. His address is School of Hotel
and Tourism Management, the Hong Kong Polytechnic University, Hung Hom,
Kowloon, Hong Kong.
2
THE INFLUENCE OF CULTURAL DISTANCE ON
CHINA INBOUND TOURISM FLOWS: A PANEL
DATA GRAVITY MODEL APPROACH
Abstract: This study focuses on the analysis of the determinants of China inbound
tourism flows and seeks to determine the influence of cultural distance on tourism
flows from a macro perspective. Three models were investigated including a
traditional tourism demand model, a gravity model, and a mixed panel data gravity
model with different sets of relevant variables. The result confirmed that cultural
distance, measured by social axioms at the country level, had a significant negative
effect on inbound tourism flows. In addition, the moderation effect of uncertainty
avoidance on cultural distance is revealed in the findings. Differences in income
elasticities and impacts of cultural and geographical distances between western and
eastern countries of tourist origin are also found in this study.
Keywords: inbound tourism flows; gravity model; China; cultural distance;
Uncertainty Avoidance Index
3
Introduction
Along with the speedy growth of tourism all over the world, the analysis and
understanding of pattern of tourism demand is necessary as it is helpful to increase
our knowledge of the relative importance of diverse economic determinants of
tourism demand (Cooper 2003). During the last two decades, tourism demand
analysis has received considerable attention from scholars, either with economic
models (Song and Li 2008; Lim 1997) or gravity models (Taplin and Qiu 1997; Um
and Lee 1998). However, these past models still left out the consideration important
psychological and emotional factors that influence tourism demand, such as cultural
factors.
To fully understand tourist behavior, cultural issues are very important and have to be
appropriately considered and represented in any demand model construction. Culture
is “the accumulation of shared meaning, rituals, norms, and traditions among
members of a society” (Soloman 1996). Since people’s behavior are shaped by norms
and values, tourists with different cultural backgrounds may act differently before and
during the travel duration (Crotts 2004; McKercher and Chow 2001). Apart from the
cultural background of each individual, tourists’ behavior are also intensively
influenced by the cultural distance (CD) between the origin and destination
(McKercher, Wong, and Lau 2006; Crotts 2004). According to numerous past studies,
CD could influence tourists’ pre-visit decision (Basala and Klenosky 2001; Ng, Lee,
4
and Soutar 2007; San Martin and Rodriguez del Bosque 2008), as well as their
experience during their visitation (Crotts 2004; Crotts and McKercher 2005; Mok and
Armstrong 1998). However, while these past studies explored the influence of CD on
tourist’s individual behavior, very few actually investigated the role of CD in
determining the aggregate tourism demand. In addition, the findings of these past
studies may be subject to selection country bias and may be unreliable due to the fact
that the data were collected from tourists from only a small number of countries.
The present research will contribute to the literature in several specific ways. First, in
this study, CD will be considered in the context of tourism demand analysis, and its
influence on tourism flows will be ascertained by using a panel data gravity model.
Second, by introducing an interaction term consisting of both the Uncertainty
Avoidance Index (UAI) and CD into the model, the findings of this study will also
provide some evidence on the significance of the moderation effect of uncertainty
avoidance on cultural distance. To the best of the authors’ knowledge, this represents a
first attempt to examine the UAI’s moderation effects in tourism geography research.
Finally, in an attempt to avoid country selection bias which existed in earlier studies
involving CD, the sample used in the study covers eighteen origin countries,
representing diverse cultural background of inbound tourists all over the world.
The rest of the paper is organized as follows. Section 2 presents a review of literature
on modeling tourism demand from economic and geographic perspectives. Section 3
5
reviews previous studies on the influence of CD and uncertainty avoidance on
tourists’ behavior. Section 4 discusses the specification of the model and then
describes the data set utilized in this paper. Section 5 gives the empirical results with
interpretations of coefficients. Finally, Section 6 presents the concluding remarks.
Tourism Demand Models
According to Song, Wong and Chon (2003), tourism, especially long-haul tourism, is
typically regarded as a luxury product, exhibiting a non-linear relationship between
the demand for tourism and its determinants which include the price of tourism, the
price of tourism in the substitute destination, and the income level of origin country.
In many empirical studies, apart from these basic variables, other independent
variables have also been introduced to explain tourism demand. These variables
include marketing expenditure, the lagged dependent variables, dummy variables for
special events, time trend variables, exchange rate, transportation cost, and migration
(Lim 1997). To model tourism demand, various modern econometric models have
been utilized to estimate the tourism demand model, such as autoregressive
distributed lag model (Song and Witt 2003), error correction model (Dritsakis 2004),
panel data model (Garín-Muñoz 2007), time varying parameter model (Li et al. 2006),
and Almost Ideal Demand System model (De Mello and Fortuna 2005).
Inasmuch as tourism is not only an economic activity, but also involves significant
emotional experience and other psychological processes, one of the most important
6
limitations of economic models is that they ignore some crucial and essential
variables (Crouch 1994a, 1994b), such as cultural, social, and political factors.
Without these factors, the model is incomplete, and the results are less reliable and
credible. More important, as pointed out by Henrich (2000), economic decisions are
profoundly influenced by cultural difference, and the hypothesis that people share
similar economic decision-making processes is questionable. Therefore, the cultural
factors deserve more consideration in modeling tourism demand.
On the other hand, geographic theories explain tourism demand from a spatial
perspective. According to the distance decay theory in geography, as distance
increases, travelling costs increase and knowledge about the destination decreases. As
a result, tourism flows peak at some distance relatively close to a source market and
then decline exponentially as distance increases (Bull 1991). Moreover, the gravity
model, an extension of distance decay analysis, is used to investigate the distance
decay effect and predict tourism demand by considering the characteristics of the
origin and the destination and their distance. The model assumes the aggregate
tourism demand from one area to another is in proportion to the attractiveness of
destination and the population of origin, and in inverse proportion to the distance
between the two areas (Smith 1983).
Gravity models focus on the spatial interaction between origins and destinations, and
have been heavily used in modeling geographical flows such as migration (Mak and
7
Moncur, 2003), transport flows (Hwang and Shiao, 2011), international trade
(Neumayer, 2011) and commuting flows (McArthur, Kleppe, Thorsen and Ubøe,
2011). As a special case of geographical flows, it is also popular to apply gravity
model to study tourism flows (Gil-Pareja, Llorca-Vivero, and Martínez-Serrano 2007;
Smith 1984; Smith and Xie 2003;Um and Lee 1998). However, as criticized by
Calantone, Di Benedetto, and Bojanic (1987), the traditional gravity model should be
augmented to incorporate more variables to improve statistical performance and
theoretical implication. Therefore, this paper proposes a model which includes
additional relevant variables compared to the limited variables that traditional gravity
model employs.
Cultural Distance, Uncertainty Avoidance and Tourists’ Behavior
1. Cultural distance and tourists’ behavior
As tourism involves emotional experience and other psychological processes, CD
studies are popular in tourists’ behavior research (Crotts and McKercher 2005; Crotts
2004; McKercher and Chow 2001; Ng, Lee, and Soutar 2007). Since differences in
culture, religion, customs, and language are often regarded as barriers for travel
(Basala and Klenosky 2001), different CDs may reflect varied risk perceptions
(Reisinger and Mavondo 2006). As a result, people tend to travel to destinations that
are less culturally distant from their residence to diminish the potential risk and ensure
their safety (Basala and Klenosky 2001; Ng, Lee, and Soutar 2007). According to
8
Reisinger (2009), large CD is regarded as a source of friction between international
tourists and host, while small CD tends to reduce the possibility of cultural conflicts,
and more likely to contribute to the positive experience. Furthermore, since there is a
large number of business travelers from international trade, and large CD tends to
increase transaction costs and reduce the likelihood that international trade (White and
Tadesse, 2008), CD between the origin and destination would also be negatively
associated with business tourist arrivals.
Despite a large number of papers exploring the CD’s influence on tourist’s behavior,
most studies are based on individual tourist from a micro perspective, and hardly any
of them investigated the role of CD in determining the aggregate tourism demand.
However, in micro studies of tourists’ behavior, it may be questionable as to whether
the sampled tourists are representative of the cultural traits of the country, as the
national culture is represented by a large number of individuals, and personal cultural
traits are not evenly distributed among all individuals (Nes, Solberg, & Silkoset,
2007). Therefore, the results and implications drawn by these researchers tend to be
less reliable. Moreover, these past studies on CD covered only a limited number of
countries, and their results may be subject to selection country bias.
This study will analyze the influence of CD on tourism flows within the framework of
tourism demand analysis from a macro perspective. The CD of inbound tourists is
computed based on a country-level cultural measurement since national culture is
9
more likely to be represented by a larger number of individuals. Following the logic
explained earlier, since tourists are inclined to choose a destination with less CD to
minimize potential risks and strangeness, it is hypothesized that the lower the CD
between visitor and destination cultures, the higher the likelihood that the
international visitor will visit a country.
2. Uncertainty avoidance and tourists’ behavior
Uncertainty avoidance is an important dimension of culture, focusing on the level of
tolerance for uncertainty and ambiguity within the society. It reflects the extent to
which members of a culture feel threatened by uncertain or unknown situations
(Hofstede 1980, 2001). Uncertainty Avoidance Index (UAI), a measurement of
uncertainty avoidance, has been frequently adopted in cross-cultural tourist studies
(Money and Crotts 2003; Pizam and Fleischer 2005; Litvin, Crotts, and Hefner 2004).
Since tourism, especially international tourism, inherently involves high levels of
uncertainty, UAI contributes towards explaining the differences in risk perception and
actual tourists’ behavior because of its link to the willingness to accept potential risks.
Previous studies suggest that high UAI tourists are likely to be more concerned about
risk threats (Reisinger and Mavondo 2006; Kozak, Crotts, and Law 2007), and prefer
less dynamic and active activities (Pizam and Fleischer 2005; Money and Crotts 2003;
Litvin, Crotts, and Hefner 2004).
However, there has been no study focusing on uncertainty avoidance’s moderation
10
effect on CD. If UAI’s roles in determining tourists’ behavior are better understood, it
will be of great importance for proposing marketing strategy. According to a
conceptual model proposed by Money and Crotts (2003), tourists from high UAI
cultures tend to minimize the traveling risk by behaving differently in external search,
trip planning, travel party choice, and trip characteristics. Therefore, it is reasonable to
hypothesize that uncertainty avoidance partially offsets the effects of CD by tourists’
adaptation to the environment of destination. This study seeks to verify that
uncertainty avoidance may have a moderation effect on CD’s influence in tourism
flows. The CD to destination, at least to some extent, reflects the degree of
uncertainty and ambiguity within the host society. Therefore, while CD imposes
negative effects on tourism flows, its impact varies among different countries
depending on the degree of uncertainty avoidance; the higher the uncertainty
avoidance is, the larger the negative effect is exerted by CD.
Model Specification and Methodology
This study uses inbound tourism flows to China for empirical study. Although some
studies utilized various time series models to forecast tourist arrivals to China
(Kulendran and Shan 2002; Witt and Turner 2002), only a few identified the
determinants of such arrivals (Song and Fei 2007). Hence, this study represents one of
the first attempts to model the factors influencing tourism flows to China.
1. Traditional tourism demand model
The proposed traditional tourism demand model is of the form:
11
2003619975198943210 lnlnlnln DDDPSYPT itititit
(1)
where i denotes the country of origin, t denotes the year of study.
β
1,
β
2, and
β
3
are price, income, and substitute price elasticities, respectively. According to demand
theory, we expect that
β
1<0, and
β
2 and
β
3>0. Tit is the annual international
tourism arrivals from origin i to China at time t. D1989, D1997, and D2003, are dummy
variables for Tiananmen Square Event, Asian Financial Crisis and SARS outbreak,
respectively. The tourist arrival data can be found in China Tourism Statistic Yearbook
published each year. The income of origin, Yit, is measured by the index of real GDP
(2000=100) and collected from the World Development Index (WDI) database. Since
the tourist arrival data contain a relatively large proportion of business travelers
(about 24% in 2004), GDP, instead of the personal disposable income, is more
appropriate in capturing the influence of income (Song, Wong, and Chon 2003). In
this research, the own price variable is defined as:
ii
chnchn
it EXCPI EXCPI
P
(2)
where CPIchn and CPIi are the consumer price indices (2000=100) for mainland China
and origin country i, respectively; EXchn and EXi are the exchange rate indices
(2000=100) for China and origin country i, respectively. The exchange rate is the
annual average market rate of the local currency against the US dollar. Because of the
diversity of transportation, travel cost is difficult to measure. Therefore, this variable
is excluded from this research. CPI data of China are obtained from the euromonitor
database, while other CPI and exchange rate data are obtained from the WDI database.
The substitute price variable PSit is defined as a weighted index of selected countries.
12
Both geographic and cultural characteristics are considered when selecting the
substitute destinations (Song et al., 2003). In this paper, Singapore, Thailand, Korea,
Japan, Hong Kong, and Malaysia are considered to be substitute destinations for
China. The substitute price index is calculated by weighing the consumer price index
of each of the six substitute destinations according to its share of international tourism
arrivals, and is given as
6
1).(
jj
ii
chnchn
it w
EXCPI EXCPI
PS
(3)
where j represents different substitute destinations; and wj is the share of international
tourism arrivals for country / region j, which is calculated from:
6
1j
jj TTAjTTAw
(4)
where TTAj is the total international tourist arrivals in country / region j, which can be
found in the UNWTO statistics.
2. Gravity model
In this research, traditional gravity model is applied to investigate the distance decay
effect of tourism flows. It is as follows:
iititit DistChnYT lnlnlnln 3210
(5)
where i denotes the country of origin, while t denotes the year of study; Tit is the
number of tourist arrivals. Two variables are included to capture the emissiveness of
tourists from a particular origin country. They are Yit (as defined in Equation 1), which
measures the income of tourists, and Chnit, which is Chinese immigrant population in
13
the origin country. Disti is the geographical distance from the origin i to China. All
these variables are taken into logarithm. Moreover, to demonstrate the annual
attractiveness change of China as a tourism destination, dummy variables of each year
are included in the model. Considering different specifications of distance decay
effects, gravity models can be further specified in other forms, such as the following
(Smith 1983):
DistDistChnYT ititit 43210 lnlnlnln
(6)
2
53210 lnlnlnln DistDistChnYT ititit
(7)
In this study, Chinese immigrant population is regarded as an important determinant
of China inbound tourism flows. First, Chinese immigrants are more likely to travel
back for VFR purposes. Second, Chinese immigrants could influence other citizens’
traveling behavior in the country of residence through word-of-mouth (WOM) effects.
The data on Chinese immigrant population are obtained from the Library for
Documentation and Information on Overseas Chinese Studies, Jinan University, PRC.
Furthermore, we obtained data on the geographical distance of each country to China
from Google Earth software. For most countries, we measured the geographical
distance between its capital and China’s three largest tourist cities/destinations,
namely, Beijing, Shanghai, and Guangzhou. That distance is then weighted by the
share of international tourism arrivals to the city and adopted as the final
measurement variable for empirical use.
14
3. Mixed panel data gravity model
The panel data gravity model specifically aims to estimate the gravity model with
panel data. Based on the traditional tourism demand model and gravity model, the
mixed panel data gravity model to be estimated as:
iiiit
iiitititit DDDCDUAIChn
CDDistPSYPT
200310199791989876
543210 lnln
lnlnlnlnlnln
(8)
where Disti is the geographical distance between the origin country and China; CDi is
the CD from the origin country to China; UAIi is Uncertainty Avoidance Index, and
Chnit is the Chinese immigrant population in the origin country. αi is the origin
country effect. Other variables are the same as in Equation 1.
One of the most important issues is to choose appropriate measurement of
culture-related variables, such as CDi and UAIi. Although Hofstede's five dimensions
for assessing CD have been frequently applied in literature (Crotts and Erdmann 2000;
Pizam and Fleischer 2005; Litvin, Crotts, and Hefner 2004), it suffers from several
disadvantages. For instance, Hofstede’s sampling of countries did not accurately
reflect the full spectrum of national culture (Schwartz 1994), and it is from
country-level study rather than culture-level study (Bond et al. 2004). To measure
cultural distance, Leung et al (2005) suggested the utilization of more recent models
of cultural dimensions. Therefore, we use Bond et al. (2004)’s framework to measure
CD. This framework is based on Leung et al. (2002)’s new concept for CD
measurement – social axioms. According to their research, social axioms are
15
generalized beliefs about oneself, the social and physical environment, or the spiritual
world, and are in the form of an assertion about the relationship between two entities
or concepts (Leung et al. 2002). They are basic premises that people endorse and use
to guide their behavior in daily living (Bond et al. 2004). Therefore, the destination
choices of tourists are also believed to have been made based on their social axioms.
To identify the country score of social axiom, Bond et al. (2004) surveyed 7672
university students and 2252 other adults from more than 40 cultural groups. Based on
ecological factor analysis, two factors were extracted, namely, dynamic externality
and societal cynicism. Dynamic externality was highly correlated with hierarchy,
collectivism, and conservatism. Societal cynicism taped the cognitive component of a
cultural constellation called maleficence, a cultural syndrome associated with a
general mistrust of social systems and other people, which appeared to be a new
dimension. Based on the measurement of social axiom, CD is specified as follows
(Kogut and Singh, 1988):
n
iiicijjnVIICD 1
2])[(
(9)
where Iij is the score of one of cultural dimensions i (such as dynamic externality and
societal cynicism) in the origin country j; Iic is the score of one of cultural dimensions
i in the destination country j; Vi is the variance of cultural dimensions i; and n is the
number of dimensions (n = 2 in this paper). This CD index is the most popular one in
various types of research (Ng, Lee, and Soutar 2007), and has been proved to be valid
in many different situations.
16
In regard to the measurement of uncertainty avoidance, we use Hofstede (1980)’s UAI.
This index is based on a survey data on work-related values obtained between 1967
and 1973 from more than 117,000 IBM employees working in 40 different countries.
Hofstede (2001) evaluated 66 nations, creating index scores and ordinal rankings for
this index. The UAI data is obtained from Hofstede’s homepage
(http://www.geert-hofstede.com/hofstede_dimensions.php).
Since tourists are inclined to choose a destination with less CD to minimize potential
risks and unfamiliarity, it is hypothesized that the lower the CD between visitor and
destination cultures, the higher the likelihood that the international visitor will visit a
destination. Therefore,
φ
5 in Equation 8 is expected to be negative. Considering that
uncertainty avoidance may have a moderation effect on CD, the interaction term of
UAI and lnCD is included in the model. We assume that, although cultural distance
may impose negative effects on tourism flows, the degree of the effect varies among
countries. Since uncertainty avoidance affects the willingness of people to accept
uncertainty, the effects of uncertainty caused by CD could be mitigated by UAI; the
higher the UAI of the country, the larger the negative effect exerted on tourism flows
by CD. Therefore,
φ
7 in Equation 8 is expected to be negative.
4. Data description
In this study, eighteen main origin countries are selected for the empirical study. They
are Australia, Canada, France, Germany, Italy, India, Indonesia, Japan, Korea,
17
Philippines, Malaysia, Netherlands, New Zealand, Singapore, Sweden, Thailand, UK,
and USA. Although Russia and Mongolia are among the top ten source markets for
China inbound tourism, they are excluded from this study because of unstable
economic and political developments during the study period.
Summary descriptive statistics for the panel data are presented in Table 1. The data of
most countries are from 1980 to 2007, with the exception that data of Korea and
Malaysia are from 1992–2007 and while those from Indonesia are from 1982–2007,
with missing values for several years. Judging from the CDs, it is shown that
Germany and France have the smallest CD to China, while Malaysia and Indonesia
are more culturally distant from China. This is somewhat contrary to the physical
situation in terms of their relative geographical locations to China.
<Insert Table 1 here>
The overall descriptive statistics and correlation of the variables after taking their
logarithms are displayed in Table 2. The correlation coefficients show that most of the
variables in the research are weakly correlated, which suggests that multiple
collinearity is not a problem.
<Insert Table 2 here>
5. Econometric methods
In this study, the panel data model is proposed for estimation. A panel data set consists
18
of data on individuals over time, and provides multiple observations on each
individual. Panel data thus allow for the control of individual heterogeneity, reducing
the problem of collinearity and providing more degrees of freedom (Hsiao 2003).
Owing to these advantages, this method is widely used in economics, management,
and sociology. The general panel data model can be specified as follows:
ititit xy
,
itiit c
(10)
where, vit is the composite error term containing two parts, ci and uit. According to the
characteristic of ci, the panel data model can be categorized into the random effect
(RE) model and the fixed effect (FE) model; while uit is the error term, which is white
noise. The key consideration in choosing between the two is whether ci and xit are
correlated (Wooldridge 2002). The RE model can be estimated by the feasible
generalized least squares (FGLS) estimator, while the FE model can be estimated by
the with-in group estimator. To validate which specification is more appropriate in the
panel data, Hausman (1978) proposed a test based on the difference between random
effects and fixed effects estimates to compare the estimation of the two models.
In panel data analysis, serial correlation is fairly common because of the dynamic
nature of numerous economic and sociology phenomena worldwide. Ignoring serial
correlation may result in consistent but inefficient estimates of coefficients with
biased standard errors. In this situation, the result of the coefficient test for
significance is no longer reliable. Baltagi and Wu (1999) proposed a method for
estimation of an unequally spaced panel data regression model with AR(1) remainder
19
disturbances; thus, the composite error term in Equation 10 is specified as:
ittiit
1,
(11)
To estimate this model, FGLS estimator is employed as weighted least squares that
correct parameter estimates and standard errors for first-order autocorrelation with an
unbalanced panel data set.
Results and Findings
1. Traditional tourism demand model
At the first stage of empirical research, as suggested by traditional tourism demand
theory, models including only economic variables were estimated. In this study,
Models 1, 3, and 5 are estimated by the fixed effect (FE) method, while others are
estimated using the random effect (RE) method. Models 1 and 2 include all the origin
countries, while Models 3 and 4 only apply to Western countries and Models 5 and 6
apply to Eastern countries1. Coefficients estimated from the RE and FE models are
nearly the same, indicating the robustness of the specification for panel data model.
Income has a positive effect on tourism movement, with income elasticity greater than
1, indicating that foreign travelers regard traveling to China as a luxury. Own price
has a negative effect on China inbound tourism flows, while the effect of substitute
price is insignificant. Special events are also significant. Tiananmen Square Event in
1989 exerted negative influences on tourism, while the effects of Asian Financial
Crisis in 1997 and SARS outbreak in 2003 were not significant.
20
Moreover, Western and Eastern countries have different elasticities for tourism
demand. In the Western models (Models 3 and 4), the magnitude of the coefficients of
income and own price variables are larger than their counterparts in the Eastern
models (Models 5 and 6). This implies that inbound tourists from Western source
markets tend to be more “economically sensitive”. The two different origin countries
also appeared to respond differently to the occurrence of special events. While the
effect of the Tiananmen Square Events in 1989 seemed to affect the two country
origins equally, the SARS outbreak tended to impose a much stronger influence on
countries with Western origins and resulted in a larger decline in tourist arrivals to
China. Finally, the statistical test for the FE or RE specification is carried out. Since
the Hausman test values in Table 3 for each group of models are negative, we turn to
the BP test (Breusch and Pagan 1980) to verify the appropriateness of RE
specification. From the BP test, all three models are significant at the 0.05 level,
suggesting that RE models are preferred.
<Insert Table 3 here>
2. Gravity model
Unlike economic models, the gravity model focuses on the influence of various
geographical factors, such as distance. Given that “distance” is non-variant in the
study period, only RE specification can be used in panel data model. Furthermore,
dummy variables of 27 years are included to capture the annual changes from 1980 to
21
2007. Three models with different distance specifications are presented,
corresponding to Equations 5, 6, and 7, respectively. The estimated coefficient of each
year dummy variable illustrates the trend of tourism growth but it is not presented
here for simplicity.
The estimation results of the gravity model are shown in Table 4. It is suggested that
all variables, including the distance variables, are significant in all three models. The
estimated coefficient indicates that the increase in income level promotes more
travelers to choose China as a destination. Additionally, lnChn, the variable measuring
the number of Chinese in origin countries, has a positive sign, suggesting that the
more Chinese immigrants the country has, the more travelers will come to China from
that country. Two explanations seem reasonable. First, Chinese immigrants are more
likely to travel back to China for VFR purposes. Moreover, they will also influence
other foreigners’ traveling behavior through WOM recommendation, which is
regarded as very important in tourists’ decision-making process (Oppermann 2000).
<Insert Table 4 here>
In addition to the above, Models 7, 8, and 9 provide information on the proper
specification of distance in the gravity model, which shows the appropriate distance
decay effect on tourism flows. The Wald test indicates that Models 8 and 9 outperform
Model 7. Comparing Models 8 and 9 which are with the same degree of freedom,
Model 8 is favored with larger values, indicating that the distance decay specification
22
from Equation 6 is more appropriate for China inbound tourism flows. Figure 1
demonstrates the distance decay curve of each specification, after controlling for other
variables (in this study, we just assign a value of 0 to other explanatory variables in
the equation for simplicity). This “pure” distance decay curve is more reliable and
accurate than the traditional one because, by holding the influences from other
variables to be constant, the uniform distribution assumption becomes stronger in this
situation (McKercher and Lew 2003). It clearly shows that, in contrast with the
distance decay curve with the specifications of Equations 5 and 7, that of Equation 6
is U-shaped, which suggests a demand decline between 3,500 km–8,000 km away
from China and a subsequent small increase. This result tallies with McKercher and
Lew’s (2003) study of Effective Tourism Exclusion Zones (ETEZ), despite the fact that
their work focused on the supply side, while this study concentrates on the demand
aspect.
<Insert Figure 1 here>
3. Mixed panel data gravity model
The results from a mixed panel data gravity model, which integrated the traditional
tourism demand analysis with the gravity model, are discussed in this section. In this
model, the influence of CD and the moderation effect of uncertainty avoidance are
also taken into consideration. Following the preliminary estimation based on the
FE/RE model, the DW and BW statistics revealed a severe problem of serial
23
correlation. Thus, AR(1) panel data model is applied for estimation (Equations 10 and
11) instead. The estimated results are presented in Table 5. Models 10, 12, and 14
include lnCD as an independent variable, while others include the interaction term of
UAI and lnCD2. Models 10 and 11 use samples from all the origin countries;
Models 12 and 13 focus on Western origins, while Models 14 and 15 include only the
Eastern ones.
In Models 10 and 11, all the variables are significant with reasonable signs. Three
economic variables (income, own price and substitution price) are significant. The
negative coefficient of lnCD indicates that cultural difference impedes international
traveling. The larger the CD is from the origin to China, the more distinct the culture
is from China, and the fewer travelers will choose China as a tourism destination.
Based on the Wald statistics, we find that Model 15, with the interaction term of UAI
and lnCD, performs better than Model 14. Hence, the moderation effect of uncertainty
avoidance on CD is confirmed for Eastern origin countries. This means that the extent
of the negative effect of CD on tourism flows is dependent on UAI; that is, the more
likely tourists tend to avoid uncertainty, the larger the negative effect of CD on the
tourism destination choice.
In addition, the differences between Western and Eastern source markets are
considered. For the Eastern models (Models 14 and 15), substitution price is not
significant. One possible explanation is that the substitution destinations we specified
24
in this study may not be reasonable for Asian tourists, while they are more appropriate
for Western ones. From Models 12 to 15, the coefficients of economic variables also
indicate that Western countries are more economically sensitive. In Western origins,
although CD and geographical distance are estimated to be negative, they are not
significant. Thus, it appears that when long-haul travelers choose destinations, they
have vague and limited knowledge on the exact distance of remote destinations in
both a geographical and a cultural perspective. As such, the influences of these
perspectives on the distant travelers are not significant, although they are negative.
<Insert Table 5 here>
Conclusion
This study attempts to investigate the determinants of China inbound tourism flows
and the influence of CD on tourism flows from a macro perspective. The panel data
gravity model shows that income of the origin is the key determinant of China
inbound tourism flows (Table 5). The income elasticity is estimated to be
approximately 2.80, which suggest that if the income of inbound tourists were to
increase by one percent, tourist arrivals to China will increase by 2.80 percent. This
implies that traveling to China is regarded as a ‘luxury’ consumption item by
foreigners, and any changes in income will be expected to have a considerable impact
on tourism demand for China. Own price elasticity is approximately 0.37, implying
that for a one percent decrease in the tourism price of China, inbound tourist arrivals
to the city will be stimulated to increase by 0.37 percent. The substitute price
25
elasticity is estimated to be 0.25, showing moderate substitution effects due to
changes in the prices of substation destination countries.
Western countries are arguably more economically sensitive, with larger elasticities.
Tiananmen Square Event in 1989 and SARS outbreak in 2003 both had negative
influences (Table 5). Although the effect of Tiananmen Square Event had equal-level
influence on both Western and Eastern origins, SARS outbreak imposed a much larger
influence on Western countries. This study also tests the different distance
specifications in gravity models. The model estimation results show a U-shaped
distance decay effect on China inbound tourism flows (Figure 1), and the ‘pure’
distance decay curve is obtained after controlling for other variables, thereby making
the uniform assumption stronger. The nature of this curve suggests that the distance
decay of tourist movement may not apply for tourists from remote source countries.
One possible explanation for this is that they have rather vague and limited
knowledge regarding the exact distance of remote destinations.
The results also confirm that CD, measured by social axioms in the country level, has
a significant negative effect on tourism flows, which verifies our hypothesis: CD will
be a barrier for international traveling. However, the influence of CD is not significant
in Western countries, and the same is true of the influence of geographical distance.
This may be attributed to the fact that travelers tend to have limited knowledge about
remote destinations. This study also find the moderation effect of uncertainty
26
avoidance on CD; that is, while CD imposes negative effects on tourism flows, its
impact varies among different countries depending on the degree of uncertainty
avoidance; the higher the uncertainty avoidance of the origin country is, the larger the
negative effect exerted by CD on tourism flows is witness.
Based on results of this research, several marketing implications can be drawn for
China’s international tourism development. In the first instance, CD has been found to
be crucial to China’s inbound tourism flows. Some issues ought to be highlighted for
tourism marketing in culturally distant countries, especially for the Eastern ones.
Certain elements about cultural proximity have to be included as significant parts in
the marketing promotion. Moreover, factors that may alleviate the negative effects of
CD, such as security conditions, amity of citizens, and well-developed bilingual
guideposts, should also be communicated to diminish the potential effects caused by
CD. Finally, because the moderation effect of uncertainty avoidance on CD is
detected in our research, the issues mentioned above are especially paramount in high
UAI countries, such as Japan. Furthermore, since tourists from Western countries are
more sensitive to economic variables that those from Eastern ones, a specific
marketing strategy targeted at promoting cheaper flight, accommodation and dining
and shopping alternatives should be put into place. In addition, should any of these
Western countries experience an economic boom, more aggressive marketing efforts
should be implemented to attract the high-spending travelers to Mainland China.
27
In addition, the study found that Chinese immigrants tend to bring about positive
effects on China inbound tourism flows. Consequently, their importance as links
connecting foreigners and Chinese should be emphasized. First, when they visit China
for VFR purposes, their high level of satisfaction should be guaranteed, to increase
their loyalty and WOM effects on other residents when they return to their resident
countries. Second, other strategies can be proposed to make the immigrants more
willing to promote China as a possible travel destination.
Finally, some limitations of this study are highlighted. Although we used social axiom
as a measurement of country-level culture, the accuracy of CD measurement should
also be further considered because we have no knowledge on whether CD as
perceived by potential tourists is the same as that in the average CD level of that
country. Furthermore, our study contains only 18 countries. More countries could be
included in the study to draw a more generalized conclusion to investigate the
influence of CD on tourism flows.
28
Endnote
1. In this paper, Western origins include Australia, Canada, France, Germany, Italy,
Netherlands, New Zealand, Sweden, UK, and USA, and Eastern origins include
India, Indonesia, Japan, Korea, Philippines, Malaysia, Singapore, and Thailand.
2. UAI*lnCD and lnCD are highly correlated. Therefore, if they are included in one
equation, the model suffers collinearity.
29
References
Baltagi, Badi. H., and Ping. X. Wu. 1999. Unequally spaced panel data regressions with AR(1)
disturbances. Econometric Theory, 15 (6), 814-823.
Basala, Sandra L., and David B. Klenosky. 2001. Travel style preferences for visiting a novel
destination: A conjoint investigation across the novelty-familiarity continuum.
Journal of Travel Research, 40 (2), 172-183.
Bond, Michael Harris, Kwok Leung, Al Au, Kwok-Kit Tong, Sharon Reimel de Carrasquel, et
al. 2004. Culture-level dimensions of social axioms and their correlates across 41
cultures. Journal of Cross-Cultural Psychology, 35 (5), 548-570.
Breusch, Trevor, and Adrian Rodney Pagan. 1980. The Lagrange multiplier test and its
applications to model specification in econometrics. Review of Economic Studies, 47
(1), 239-253.
Bull, Adrian. 1991. The economics of travel and tourism. Melbourne: Pitman.
Calantone, Roger J., C. Anthony Di Benedetto, and David Bojanic. 1987. A comprehensive
review of the tourism forecasting literature. Journal of Travel Research, 26 (2),
28-39.
Cooper, Chris. 2003. Classic reviews in tourism. Clevedon: Channel View.
Crotts, John C. 2004. The effect of cultural distance on overseas travel behaviors. Journal of
Travel Research, 43 (1), 83-88.
30
Crotts, John C., and Ron Erdmann. 2000. Does national culture influence consumers'
evaluation of travel services? A test of Hofstede's model of cross-cultural differences.
Managing Service Quality, 10 (6), 410-419.
Crotts, John C., and Bob McKercher. 2005. The adaptation to cultural distance and its
influence on visitor satisfaction: The case of first-time visitors to Hong Kong.
Tourism Analysis, 10 (4), 385-391.
Crouch, Geoffrey I. 1994a. The study of international tourism demand: A review of findings.
Journal of Travel Research, 33 (1), 12-23.
Crouch, Geoffrey I. 1994b. The study of international tourism demand: A survey of practice.
Journal of Travel Research, 32 (4), 41-55.
De Mello, Maria M., and Natércia Fortuna. 2005. Testing alternative dynamic systems for
modelling tourism demand. Tourism Economics, 11 (4), 517-537.
Dritsakis, Nikolaos. 2004. Cointegration analysis of German and British tourism demand for
Greece. Tourism Management, 25 (1), 111-119.
Garín-Muñoz, Teresa. 2007. German demand for tourism in Spain. Tourism Management, 28
(1), 12-22.
Gil-Pareja, Salvador, Rafael Llorca-Vivero, and Jos Antonio Martínez-Serrano. 2007. The
impact of embassies and consulates on tourism. Tourism Management, 28 (2),
355-360.
31
Hausman, Jerry A. 1978. Specification tests in econometrics. Econometrica, 46 (6),
1251-1271.
Henrich, Joseph. 2000. Does culture matter in economic behavior? Ultimatum game
bargaining among the Machiguenga of the Peruvian Amazon. American Economic
Review, 90 (4), 973-979.
Hofstede, Geert. 1980. Culture's consequences: International differences in work-related
values. Beverly Hills: Sage.
Hofstede, Geert. 2001. Cultural consequences. 2nd ed. Thousand Oaks, CA: Sage.
Hsiao, Cheng. 2003. Analysis of panel data 2nd ed. Cambridge: Cambridge University Press.
Hwang, Cherng-Chwan, and Guo-Chou Shiao. 2011. Analyzing air cargo flows of
international routes: an empirical study of Taiwan Taoyuan International Airport.
Journal of Transport Geography, 19(4), 738-744.
Kogut, Bruce and Harbir Singh. 1988. The effect of national culture on the choice of entry
mode. Journal of International Business Studies, 19(3), 411–432.
Kozak, Metin, John C. Crotts, and Rob Law. 2007. The impact of the perception of risk on
international travellers. International Journal of Tourism Research, 9 (4), 233-242.
Kulendran, Nada, and Jordan Shan. 2002. Forecasting China's monthly inbound travel
demand. Journal of Travel & Tourism Marketing, 13 (1/2), 5-19.
32
Leung, Kwok, Rabi S. Bhagat, Nancy R. Buchan, Miriam Erez, and Cristina B. Gibson. 2005.
Culture and international business: Recent advanced and future directions. Journal of
International Business Studies, 36 (4), 357-378.
Leung, Kwok, Michael Harris Bond, Sharon Reimel de Carrasquel, Carlos Munoz, Marisela
Hernandez, et al. 2002. Social axioms: The search for universal dimensions of general
beliefs about how the world functions. Journal of Cross-Cultural Psychology, 33 (3),
286-302.
Li, Gang, Kevin K. F. Wong, Haiyan Song, and Stephen F. Witt. 2006. Tourism demand
forecasting: A time varying parameter error correction model. Journal of Travel
Research, 45 (2), 175-185.
Lim, Christine. 1997. Review of international tourism demand models. Annals of Tourism
Research, 24 (4), 835-849.
Litvin, Stephen W., John C. Crotts, and Frank L. Hefner. 2004. Cross-cultural tourist
behaviour: A replication and extension involving Hofstede's uncertainty avoidance
dimension. International Journal of Tourism Research, 6 (1), 29-37.
Mak, James, and James E.T. Moncur. 2003. Interstate migration of college freshmen. The
Annals of Regional Science, 37(4), 603-612.
McArthur, D. Philip, Gisle Kleppe, Inge Thorsen, and Jan Ubøe. 2011. The spatial
transferability of parameters in a gravity model of commuting flows. Journal of
Transport Geography, 19(4), 596-605.
33
McKercher, Bob, and So-Ming Chow. 2001. Cultural distance and participation in cultural
tourism. Pacific Tourism Review, 5 (1), 23-32.
McKercher, Bob, and Alan A. Lew. 2003. Distance decay and the impact of effective tourism
exclusion zones on international travel flows. Journal of Travel Research, 42 (2),
159-165.
McKercher, Bob, Celia Wong, and Gigi Lau. 2006. How tourists consume a destination.
Journal of Business Research, 59 (5), 647-652.
Mok, Connie, and Robert W. Armstrong. 1998. Expectations for hotel service quality: Do they
differ from culture to culture? Journal of Vacation Marketing, 4 (4), 381-391.
Money, R. Bruce, and John C. Crotts. 2003. The effect of uncertainty avoidance on
information search, planning, and purchases of international travel vacations. Tourism
Management, 24 (2), 191-202.
Neumayer, Eric. 2011. On the detrimental impact of visa restrictions on bilateral trade and
foreign direct investment. Applied Geography, 31(3), 901-907.
Ng, Siew Imm, Julie Anne Lee, and Geoffrey N. Soutar. 2007. Tourists' intention to visit a
country: The impact of cultural distance. Tourism Management, 28 (6), 1497-1506.
Oppermann, Martin. 2000. Tourism destination loyalty. Journal of Travel Research, 39 (1),
78-84.
34
Pizam, Abraham, and Aliza Fleischer. 2005. The relationship between cultural characteristics
and preference for active vs. passive tourist activities. Journal of Hospitality &
Leisure Marketing, 12 (4), 5-25.
Reisinger, Yvette, and Felix Mavondo. 2006. Cultural differences in travel risk perception.
Journal of Travel & Tourism Marketing, 20 (1), 13-31.
Reisinger, Yvette. 2009. International tourism: Cultures and behavior. Oxford: Elsevier.
San Martin, Hector, and Ignacio A. Rodriguez del Bosque. 2008. Exploring the
cognitive-affective nature of destination image and the role of psychological factors
in its formation. Tourism Management, 29 (2), 263-277.
Schwartz, Shalom H. 1994. Beyond individualism-collectivism: New cultural dimensions of
values. In Individualism and collectivism: Theory, method and applications, edited
by U. Kim, H.C.Triandis, C.Kagitcibasi, S.C.Choi and G.Yoon. Newbury Park, CA:
Sage, 85-119.
Smith, Peter B., Mark F. Peterson, and Shalom H. Schwartz. 2002. Cultural values, sources of
guidance, and their relevance to managerial behavior Journal of Cross-Cultural
Psychology, 33 (2), 188-208.
Smith, Stephen L.J. 1983. Recreation geography. New York: Longman.
Smith, Stephen L.J. 1984. A method for estimating the distance equivalence of international
boundaries. Journal of Travel Research, 23 (3), 37-39.
35
Smith, Stephen L.J., and Philip Feifan Xie. 2003. Estimating the distance equivalence of the
Canada-US border on US-to-Canada visitor flows Journal of Travel Research, 42 (2),
191-194.
Soloman, M. 1996. Consumer behavior. 3rd ed. Upper River Saddle, NJ: Prentice Hall.
Song, Haiyan, and Stephen F. Witt. 2003. Tourism forecasting: The general-to-specific
approach. Journal of Travel Research, 42 (1), 65-74.
Song, Haiyan, Kevin K. F. Wong, and Kaye K. S. Chon. 2003. Modelling and forecasting the
demand for Hong Kong tourism. International Journal of Hospitality Management,
22 (4), 435-451.
Song, Haiyan., and Baogang. Fei. 2007. Modelling and forecasting international tourist
arrivals to mainland China. China Tourism Research, 3 (1), 20-40.
Song, Haiyan., and Gang. Li. 2008. Tourism demand modelling and forecasting- A review of
recent research. Tourism Management, 29 (2), 203-220.
Taplin, John H. E., and Min Qiu. 1997. Car trip attraction and route choice in Australia.
Annals of Tourism Research, 24 (3), 624-637.
Um, Seoho, and ChoongKi Lee. 1998. An application of the gravity model in a practical
setting: Estimating the effect of road network improvement in generating foreign
tourists' trips within Bali. Pacific Tourism Review, 2 (1), 21-27.
36
White, Roger and Bedassa Tadesse. 2008. Immigrants, cultural distance and U.S. state-level
exports of cultural products. The North American Journal of Economics and Finance,
19(3), 331-348.
Witt, Stephen F., and Lindsay W. Turner. 2002. Trends and forecasts for inbound tourism to
China. Journal of Travel & Tourism Marketing, 13 (1/2), 99-109.
Wooldridge, Jeffrey. M. 2002. Econometric analysis of cross section and panel data.
Cambridge, Massachusetts: The MIT Press.
37
TABLES AND FIGURES
Table 1. Descriptive statistics for each origin countries
Country
Mean Value
Sampling
period
Tourist
arrival
GDP
index
Price
Sub.
price
Phy.
dist.
CD
Number
of
Chinese
UAI
Australia
171711
82.770
1.003
0.873
8202
1.117
454000
51
1980-2007
Canada
160512
83.277
1.078
0.931
11297
1.573
910000
48
1980-2007
Germany
170045
86.026
1.069
0.901
8316
0.177
111000
65
1980-2007
France
135451
86.418
1.035
0.879
8787
0.219
300000
86
1980-2007
U.K
203049
85.346
1.245
1.062
8613
0.950
250000
35
1980-2007
Indonesia
172217
89.485
0.865
0.722
4381
4.593
7310000
48
1982-1987
1992-2007
India
98229
78.291
0.946
0.864
5889
1.014
135000
40
1980-2007
Italy
62862
87.903
1.043
0.881
8636
1.849
30000
75
1980-2007
Japan
1420434
88.523
1.540
1.285
2145
0.297
840000
92
1980-2007
Korea
1653512
98.971
0.918
0.961
1041
0.401
30000
85
1992-2007
Malaysia
476945
97.791
0.894
0.920
3502
3.893
5280000
36
1992-2007
Netherlands
50615
82.384
1.064
0.901
8412
1.647
80000
53
1980-2007
New Zealand
29087
88.163
1.015
0.860
10102
0.775
35000
49
1980-2007
Philippines
246562
85.933
1.083
0.946
2025
1.591
2200000
44
1980-2007
Singapore
268738
69.159
1.204
1.033
3599
0.297
2291100
8
1980-2007
Sweden
37995
87.329
1.011
0.869
7317
1.340
20000
29
1980-2007
Thailand
189313
79.679
0.996
0.848
2732
1.476
6100000
64
1980-2007
U.S.A
616780
81.744
1.276
1.093
11121
1.675
2000000
46
1980-2007
38
Table 2. Summary of descriptive statistics and correlation of variables
lnT
lnY
lnP
lnPS
lnDist
lnChn
lnCD
Descriptive Statistics
Number of
Observation
474
474
474
474
474
474
474
Mean
11.579
4.396
0.015
-0.085
8.644
12.840
-0.070
Standard.
Deviation
1.517
0.311
0.340
0.205
0.638
1.865
0.882
Minimum
7.883
3.159
-0.734
-1.092
6.948
9.903
-1.730
Maximum
15.379
5.120
1.369
0.680
9.332
15.805
1.524
Correlation
lnT
1
lnY
0.632
1
lnP
-0.231
-0.414
1
lnPS
0.224
-0.007
0.610
1
lnDist
-0.401
0.005
-0.029
-0.130
1
lnChn
0.383
-0.122
0.059
0.071
-0.353
1
lnCD
-0.150
0.040
-0.141
-0.247
0.095
0.246
1
39
Table 3. Panel data estimation for tourism demand model
All countries
Western countries
Eastern countries
Model 1
Model 2
Model 3
Model 4
Model 5
Model 6
lnY
2.984***
2.978***
3.923***
3.926***
2.732***
2.727***
(0.069)
(0.072)
(0.125)
(0.128)
(0.084)
(0.087)
lnP
-0.584***
-0.594***
-0.497***
-0.499***
-0.223*
-0.227*
(0.076)
(0.080)
(0.087)
(0.089)
(0.119)
(0.124)
lnPS
-0.081
-0.019
0.156
0.181
0.047
0.130
(0.130)
(0.139)
(0.159)
(0.163)
(0.191)
(0.200)
D1989
-0.370***
-0.367***
-0.340***
-0.338***
-0.365**
-0.363**
(0.098)
(0.103)
(0.097)
(0.100)
(0.171)
(0.179)
D1997
-0.030
-0.033
0.040
0.037
-0.150
-0.153
(0.089)
(0.094)
(0.097)
(0.100)
(0.136)
(0.142)
D2003
-0.147
-0.142
-0.341***
-0.338***
-0.038
-0.038
(0.090)
(0.095)
(0.099)
(0.100)
(0.137)
(0.143)
Constant
-1.521***
-1.434***
-6.113***
-6.122***
0.253
0.331
(0.310)
(0.356)
(0.552)
(0.608)
(0.373)
(0.431)
Observations
474
474
280
280
194
194
Number of id
18
18
10
10
8
8
R-squared
0.376
0.380
0.365
0.367
0.624
0.632
Model
specification
test
F(6,450)=
558.68***
Wald(6)=
3036.04***
F(6,264)=
454.89***
Wald (6)=
2592.43***
F(6,180=
271.73***
Wald (6)=
1499.88***
Hausman
Test
-2.71 (df=6)
-0.65 (df=6)
-3.71 (df=6)
BP Test for
random
effect
3396.81***
(df=1)
2329.90***
(df=1)
926.81***
(df=1)
Wooldridge
test for
autocorrelati
on
F(1,17)=
194.252***
F(1,9)=
174.567***
F(1,7)=
45.278***
(***significant at the 0.01 level, ** significant at the 0.05 level, * significant at the 0.1 level.)
40
Table 4. Panel data estimation for gravity models
Model 7
Model 8
Model 9
lnChn
0.206*
0.294***
0.248**
(0.110)
(0.107)
(0.104)
lnY
1.016***
1.015***
1.015***
(0.079)
(0.079)
(0.079)
lnDist
-0.712**
-3.122***
-1.938***
(0.315)
(1.053)
(0.628)
Dist
0.001**
(0.000)
Dist*Dist
2.34e-08**
(0.000)
Constant
11.875***
27.978***
20.647***
(3.388)
(7.443)
(5.077)
Observations
474
474
474
Number of id
18
18
18
R-squared
0.715
0.793
0.784
Specification
Test
Wald(30)=
10100.37***
Wald (31)=
10147.28***
Wald (31)
=10146.28***
(***significant at the 0.01 level, ** significant at the 0.05 level, * significant at the 0.1 level.)
41
0
200
400
600
800
1000
1200
1400
1600
1000 2500 4000 5500 7000 8500 10000 11500 13000
Distance to China
Number of Tourist Arrivals
Specification
in Equation 5
Specification
in Equation 6
Specification
in Equation 7
Figure 1. Distance decay curves derived from different gravity models
42
Table 5. Panel data estimation for panel data gravity model
All countries
Western countries
Eastern countries
Model 10
Model 11
Model 12
Model 13
Model 14
Model 15
lnY
2.801***
2.792***
3.997***
3.998***
2.505***
2.498***
(0.130)
(0.128)
(0.220)
(0.220)
(0.144)
(0.144)
lnP
-0.373***
-0.374***
-0.433***
-0.433***
-0.218**
-0.228**
(0.073)
(0.073)
(0.104)
(0.104)
(0.097)
(0.097)
lnPS
0.251**
0.252**
0.482***
0.482***
-0.069
-0.061
(0.110)
(0.110)
(0.148)
(0.148)
(0.160)
(0.160)
Dist
0.0004**
0.0004**
-0.001
-0.001
-0.002
0.001
(0.000)
(0.000)
(0.002)
(0.002)
(0.002)
(0.002)
lnDist
-2.458***
-2.279***
5.228
6.828
3.807
-2.733
(0.784)
(0.806)
(16.761)
(16.427)
(6.426)
(5.226)
lnChn
0.404***
0.397***
0.745***
0.751***
-0.223
0.467
(0.083)
(0.083)
(0.113)
(0.115)
(0.562)
(0.490)
lnCD
-0.436***
-0.043
-0.258
(0.162)
(0.178)
(0.398)
D1989
-0.118***
-0.118***
-0.122***
-0.122***
-0.151***
-0.149***
(0.034)
(0.034)
(0.044)
(0.044)
(0.055)
(0.055)
D1997
-0.020
-0.020
-0.007
-0.007
-0.037
-0.037
(0.028)
(0.028)
(0.039)
(0.039)
(0.042)
(0.042)
D2003
-0.332***
-0.332***
-0.356***
-0.356***
-0.291***
-0.291***
(0.028)
(0.028)
(0.039)
(0.039)
(0.041)
(0.041)
UAI*lnCD
-0.007***
2.2e-6
-0.016**
(0.003)
(0.003)
(0.007)
Constant
12.613**
11.384*
-56.060
-69.050
-20.758
14.120
(5.689)
(5.878)
(136.402)
(133.712)
(36.660)
(29.261)
Observatio
ns
474
474
280
280
194
194
Number of
id
18
18
10
10
8
8
R-squared
0.798
0.789
0.871
0.871
0.866
0.885
Specificatio
n Test
Wald (11)
=840.62***
Wald (11)
=839.19***
Wald (11)
=687.95***
Wald (11)
=687.61***
Wald (11)
=436.14***
Wald (11)
=456.37***
ρ
0.921
0.921
0.862
0.862
0.925
0.925
DW
0.289
0.289
0.365
0.365
0.242
0.242
BW
0.461
0.461
0.509
0.509
0.414
0.414
(*** significant at the 0.01 level, ** significant at the 0.05 level, * significant at the 0.1 level.)
43
APPENDIX. CULTURAL DIMENSION SCORES FOR EACH COUNTRY
Country
UAI
Dynamic Externality
Societal Cynicism
Australia
51
60.3
52.5
Canada
48
65.9
51.0
Germany
65
61.3
61.0
France
86
59.3
58.2
U.K
35
61.9
52.7
Indonesia
48
79.6
51.0
India
40
72.5
60.2
Italy
75
57.3
51.3
Japan
92
60.2
61.4
Korea
85
66.1
62.4
Malaysia
36
80.9
55.4
Netherlands
53
57.7
51.7
New Zealand
49
62.0
53.3
Philippines
44
72.3
53.6
Singapore
8
66.8
56.2
Sweden
29
63.2
51.4
Thailand
64
71.6
64.1
U.S.A
46
65.6
50.7
China
63.5
58.8
(Note: Since the national scores of dynamic externality and societal cynicism for Australia and Sweden
are not provided in Bond et al.,(2004)’s paper, we using missing value analysis to fill in these values
after considering their correlation with other cultural values measured by Hofstede (2001) and Smith
et al.(2002))
