part of the overall energy consumption. The explanation to this paradox
might be the following: carbon dioxide emissions are directly correlated to
transport, which is a necessary requirement for visitors. At the same time,
the overall energy consumption is broader, being also related to various
heavy industries, which have a detrimental effect over the environment but
are not directly contributing to tourism growth.
Pulido-Fern´
andez et al. [2019] investigated the relationship between envi-
ronmental sustainability and tourism growth, finding that these two factors
are mutually influencing one another. Thus, more tourists visiting a desti-
nation can lead to a worse ecological situation. At the same time, there is a
positive effect of more environmentally sustainable practices over the number
of visitors. The authors therefore demonstrate that the structural transfor-
mation towards more sustainability does not only have ecological benefits,
but is also propitious for the economic performance, including through the
growth of tourism.
More specifically, Snieˇska et al. [2014] studied the elements that have
5
a substantial effect over the development of tourism in rural areas, with
focus on the case of Lithuania. This area, according to the authors, has
a great potential to grow that is currently not used to its full capacity. It
has been found that the economic factors have the biggest effect over the
level of tourism development in this region. These are: the overall economic
development expressed through the GDP, government expenditures and its
revenue, salaries, as well as investments made by foreigners.
Haxton [2015] describes various policies that can beneficial effect over the
growth of tourism, and how they are connected between each other. That is,
the paper emphasizes the importance of green growth because of the envi-
ronmental imbalances brought by the classical economic theory approaches.
Ideally, it should be related to the inclusive growth, which supports a sub-
stantial increase of economic, social, and environmental living standards for
a large part of the population. The paper supports a more holistic approach
to growth because it can help find a perfectly balanced solution, in which the
economic growth is mutually supported by a decreasing income inequality.
The authors suggest that tourism-related policies should not be designed
apart, but rather in a strong connection to other government decisions in
order to avoid over-regulation and inefficient work.
According to Thomas and Long [1999], competitiveness represents one
of the prerequisites for tourism development. That is, small and medium
businesses need to be persuaded about the necessity of competition, and that
they need to be proactive rather than reactive in order to achieve growth.
However, the authors point out that many businesses in the tourism industry
are satisfied with their current state of being and do not express the wish to
economic growth. In my vision though, albeit this such an attitude might
initially seem ideal from both social and economic perspectives, it makes the
SMEs more vulnerable to the activities of big players. It could potentially
lead to the bankruptcy of the latter and to the subsequent monopolization
of the market.
Garrig´
os-Sim´
on et al. [2015] claims that calculating sustainability as a
quantitative indicator is a rather complicated task because of its complexity
and relative subjectivity. Thus, this article presents a mathematical model
designed to measure economic sustainability in the field of tourism through
the leakage in the hotel industry. In this context, leakage means the amount
of money that is made thanks to tourists but which does not stay in the
6
country. The authors point out that sustainable economic growth requires
policies that are suitable for a big variety of stakeholders. Besides that,
the interests of all collaborators need to be taken into account. Indeed, a
multidimensional balance of motivators is often a more important tool to
achieve sustainability than, for instance, quick economic growth.
The paper published by Bianchi [2018] shows the link between vari-
ous points of view in the political economy and tourism development. It
discusses the contrast between the modernization theory and neo-Marxist
tourism critiques. The former claims that developing countries should be
open to tourism and to foreign capital investment because these are drivers
of the economic growth. The latter argues that the development of tourism
in ”Third world” countries would lead to even higher income inequalities
between states. The purpose of our research, though, is not to advocate any
of these theories, but rather to present the ambiguity of ideas. Therefore,
it is important to keep in mind that the effect of various factors on the
level of tourism development does not only depend on objective and quan-
tifiable characteristics, but is also influenced substantially by the current
government policy.
The support of the local community for tourism is another factor that
has potentially huge influence over its development. The study performed
by Woo et al. [2015], based on answers from residents of five distinct destina-
tions, found that the locals perceive tourism as a way of improving ”material
and non-material” life satisfaction, thus being a catalyst that enhances peo-
ple’s well-being in many ways. Moreover, authors claim, the better quality
of life leads to an even more developed tourism industry. Therefore, if the
local residents have initially a positive attitude towards national and inter-
national visitors, if they are willing to share their culture, monuments, and
experience with others, then it would lead to a virtuous circle, in which each
element supports the further development of the other.
In their paper, Ghosh [2019] investigates the impact of the economic and
political uncertainties over tourism. The economic unpredictability has been
estimated using the Economic Policy Index, whereas the terrorism index has
been used as a proxy for the political unpredictability. In my vision, the
usage of the latter as a political indicator seems to be a rather superficial
approach because of the complexity and distinctiveness of various political
regimes. The study, which is based on a time period between 1995 and
7
2016, found that uncertainty has a negative effect over the level of tourism
development. However, the fact that the study was based only on three
countries: France, United States, and Greece, suggests the non-applicability
of the results on other countries, especially on the ones that have a lower
GDP and/or fewer tourists.
Regardless of substantial differences in the approaches that each of the
above-discussed studies have, some of the factors that have substantial effect
over tourism growth have been encountered more regularly. Infrastructure
is the factor that many studies consider crucial in determining the subse-
quent development of tourism - both at the national and at the local level.
Economic factors are also important at both macro- and at micro- levels.
For instance, researchers found that government purchases and business ac-
tivity serve as driving factor for tourism growth. Some studies reveal that
the state can influence our variable of interest not only through increasing
investments, but also through decreasing the taxes, thus motivating en-
trepreneurship in the field. Some qualitative factors, such as sociopolitical
and cultural aspects, as well as the local community involvement, were found
to be significant as well.
8
3
Theory
The purpose of this chapter is to describe the theory that serves as explana-
tion to the subsequently constructed model and regression. In comparison to
the Literature Review (section 2), the ideas described in this chapter do not
have to be directly related to the question of our interest, but are supporting
the explanations behind the outcomes. Some of the discussed theories are
concerning particular relationships between some of the independent vari-
ables, which might be an important issue in the context of multicollinearity.
First, a previous study about the correlation between tourism and crime is
presented. Then, the paper discusses the direct positive relationship between
education and human capital development, as well as between greenhouse
gas emissions and tourism. Human Capital Theory furthermore explains the
benefits of higher education as a driving force for the social and economic
development. Besides that, this chapter emphasizes the importance of eco-
nomic freedom and describes various ways for its estimation. The theory
behind the negative effects of the 2008-09 economic crisis is discussed in
order to explain the introduction of the dummy variable for the year 2010.
Lastly, the law of diminishing returns is extremely useful when discussing
the heterogeneity between countries.
According to Ryan [1993], the relationship between crime and tourism ex-
ists, and it can be identified in various forms. The first one is when tourists
are indirect victims and the main focus of the criminal activity is on the
indigenous population, the destination having no relationship to the illegal
activity. The second form is when the tourist location is popular among law-
breakers because of its nature (higher density of population, higher GDP per
capita, more businesses), but their activity is not oriented towards tourists.
However, criminal activity can be specifically organized against tourists be-
cause of their relative defencelessness due to linguistic, social, or cultural
barriers.
Education, one of the factors included in the empirical model, comes in
a strong positive relationship with the human capital, which is one of the
most important indicators of the country’s multilateral development. From
the entrepreneurship perspective, it represents a resource that increases the
probability that the business will survive in the long term thanks to better
skills and experience that the owner has [Br¨
uderl et al., 1992].
9
Human Capital Theory puts special emphasis on the higher education
because it is double beneficial - both for the people and for the government
[Schultz, 1961]. For every individual, more education will lead to higher
lifetime incomes albeit they start working later due to the studies. It also
brings higher labour flexibility and less time spent while being unemployed.
Besides that, higher levels of education also improve the productivity of fac-
tories and businesses. From the governmental perspective, higher education
leads to a more stable society with generally healthier citizens, which is very
beneficial for the state budget. In consequence, the government can decide
to spend the budget surplus on other areas, such as infrastructure, research
and development, sovereign wealth funds, or also on education. Ideally, it
leads to a virtuous circle, in which more education not only means overall
long-term evolution, but also more education.
The concept of economic freedom, which is one of the factors studied,
represents one of the most important elements of the country’s economic
development. In order to estimate it in a complex way, a combination of
both objective and subjective constituents is required Gwartney and Lawson
[2003]. Its most important elements are: free competition among businesses,
liberalized exchange, personal choice, and property and personal protec-
tion. Under perfect economic freedom, nothing and nobody decides for the
stakeholders regarding their entrepreneurial activities except for themselves,
therefore the personal ownership of goods and intellectual property is par-
ticularly important. However, the author recognizes that the index is too
multiplex and depends on too many factors, thus it cannot be measured
with ideal precision.
It is rather logical that the 2008-2009 economic crisis had a substantial
negative effect over the number of tourists due to the fact that people tend
to keep their savings for more crucial things, such as food and housing. Ac-
cording to Papatheodorou et al. [2010], some travellers definitely decided to
reduce their spending on visiting various destinations, but only in short to
medium term, whereas the long-run effect should be much more moderate.
Thus, the numbers were negative during only a year and a half, whereas
the year 2010 is already considered to have brought recovery. The authors
consider that regional cooperation and collaboration are among the most
important tools that contributed to the convalescence of the tourism sector,
consequently bringing a more sustainable growth and greener ways of trav-
10
elling. Such a conclusion seems to be valuable also in the context of the post
Corona crisis rehabilitation.
Both various studies and empirical outcomes reveal that there is a di-
rect positive relationship between tourism and greenhouse gas emissions
[G¨
ossling, 2013]. The influence of tourism over the national emissions seems
to be statistically significant in all countries studied, but its intensity dif-
fers. Moreover, the study reveals that the effect can be underestimated
because of the inconsistency of the calculation methods used, and in reality
it is higher than the ”official” 5-15%. According to the authors, tourism
is a very energy-intense economic sector that is expanding rapidly, but the
policymakers are not making the necessary decisions in order to redress its
negative outcomes. According to a study based on selected Asia-Pacific
states, for instance, more visitors will lead to higher levels of carbon dioxide
emission in the long run [Shakouri et al., 2017]. The results are, however,
similar in a study that investigates most visited states, as Ko¸cak et al. [2020]
concluded in their paper.
The law of diminishing returns is useful in this study because it explains
why some effects are stronger for particular sets of countries, whereas for
other states they are much less relevant. According to Shephard and F¨
are
[1974], if one applies continuously the same amount of effort, labour, or
capital, on a certain activity, then the output of each additional contribution
will be lower and lower. For instance, if a developing country will invest
x
million euros in education, then this action will have a more observable effect
over its overall development then if a more developed country will invest the
same amount of money. This happens because of a higher productivity of
each item at the initial levels of production. Afterwards, once there are more
items playing the same role, then the influence of each additional one will
be less observable.
11
4
Methodology
The purpose of this section is to describe the methods of finding the data
that has been subsequently processed, as well as the sources used. Since
the question of our interest is rather macroeconomic, it does not imply the
direct usage of surveys and interviews in the process of data collection by
the author of this paper. All the figures have been collected directly from
reliable databases. It is possible, however, that surveys and other types
of qualitative data collection have been used in the process of creating the
above-mentioned databases, which means that, indirectly, they might be
used in this study as well.
The nature of the socioeconomic processes and of the influences between
them is dynamic, and can hardly be analyzed at one certain point in time.
Therefore, this paper deals with panel data that cover a time period of 20
years, between 1995 and 2015, and which is divided into five periods with a
five-year time span between each of them. Unfortunately, more recent infor-
mation regarding some independent variables is not fully available because
some organizations require up to several years to collect and publish the
data.
Also, the five-year gap between distinct time periods was chosen because
of its higher availability. Say, there is more data regarding the year 2005 than
regarding 2003 because some countries and organizations tend to summarize
their results exactly with a five-year interval. Besides that, including every
year in the analysis would have implied a much lower variability because,
obviously, indicators do not change much from one year to another, whereas
five-year alterations are more observable.
The variables included in the
regression are the following:
1. The
growth of tourism
represents the dependent variable of our in-
terest. The information has been extracted from the UNWTO database
[UNWTO, 2021], and represents the number of international arrivals in
a certain year. The number of international arrivals was chosen instead
of a combination of the number of international and domestic tourists
because of the lack of data regarding the latter. That is, the number of
domestic tourists is very often presented by the local tourist organiza-
tion instead of being estimated by UNWTO. Thus, substantially more
information is available regarding developed states, whereas there is
12
little to no data regarding the number of domestic tourists in the coun-
tries that are less economically developed. Thus, using a combination
of international and local arrivals would have led to the exclusion of a
lot of developing states from the dataset because of the lack of data
regarding them. Subsequently, it would have made the study focused
too much on countries with a higher GDP per capita, which would
have decreased the universal applicability of the results because of a
strong selection bias [Heckman, 1990].
2. The first independent variable is the country’s level of economic growth.
It is represented by the
real GDP per capita
growth, and is ex-
tracted from the database provided by the World Bank [WB, 2021b].
The currency used is constant over the whole period studied in or-
der to avoid inflation effects. It is calculated with the value of the
US dollar in 2010. It is generally expected that the level of economic
development and the number of tourists are mutually affecting each
other. Including this variable could lead to the problem of endogeneity
due to the simultaneity [Wintoki et al., 2012]. Therefore, this variable
has been included with a time lag. In this way, only the lagged in-
dependent variable would influence the dependent variable of tourism
growth, and not vice versa.
Figure 1 demonstrates that, at any point in time, there is a positive
relationship between the GDP per capita and the number of tourists.
It implies that these two factors are mutually influencing each other
and gives empirical explanation to the usage the independent variable
with a lag.
13
Figure 1: The positive relationship between GDP per capita and number of
tourists
3. Another important factor that could potentially influence the level
of tourism is related to the environmental sustainability. The
CO2
emissions
(metric tons per capita) were used as a proxy for it, with
the information being extracted from the World Bank database [WB,
2021a]. According to Shahbaz and Sinha [2019], there is a nonlinear
relationship between the economic growth and CO
2
emissions, having
an inverted U shape. This non-linearity implies the necessity to include
this variable in the form:
x
+
x
2
.
4. One could expect that the overall
economic freedom
would give
14
more incentives for businesses and would provide a healthy competi-
tion in the country. There is a positive relationship between the supply
of services in the tourism industry and the demand for them, therefore
a positive effect of the economic freedom over the level of tourism in
the country is expected as well. In other words, the index of economic
freedom has been chosen due to the fact that more economic free-
dom gives more incentives for the people to have economic activities.
The supply usually balances the demand, therefore we would expect
more tourists demanding these services as well. The information is
downloaded from the Index of Economic Freedom database [Heritage
Foundation, 2021].
5.
Safety
represents one of the most important factors that is supposed
to influence people’s decision regarding visiting a certain country. The
level of intentional homicides per 100,000 inhabitants has been used as
a proxy for safety as there is obviously a strong direct inverse relation-
ship between these two factors. This indicator has been chosen because
of its overall high availability both among developed and among de-
veloping countries. The information is provided by the World Bank
[WB, 2021c].
6. One could expect that a higher
level of education
would have a pos-
itive effect over a series of indicators which are harder to quantify but
which could have a strong influence over the number of tourists. This
could be the development of the human capital, the infrastructure ex-
pansion, or the people’s awareness and attitude towards sustainability
and towards new cultures. Thus, education serves as a cumulative
proxy to all the above-mentioned semi-qualitative indicators. The av-
erage years of schooling per person will be used because of its higher
availability [Our World in Data, 2021]. Due to the fact that devel-
oped countries have initially higher levels of education than developing
states, it was decided to include the growth of education instead.
7. Year dummy variables are generally a useful way of taking into account
structural changes in the economies, as well as temporary variations
[Green and Doll, 1974, p.60]. It can be supposed that the 2008-2009
crisis has severely influenced people’s incentives to travel during the
15
next several years, and therefore a
dummy variable for the year
2010
has been included as well.
Overall, information regarding 130 countries has been collected. A higher
number of observations is necessary in order to ensure that the distribution is
standard and normal, according to the Central Limit Theorem [Wooldridge,
2010, p.845]. Besides that, having data regarding more countries is necessary
while dividing them according to regional or socio-cultural indicators.
Some values in the created panel data were missing due to the fact that
not all countries and statistical offices provide with data regarding each
country for the years of our interest.
In order to solve this issue, data
imputation methods were applied Efron [1994]. If an observation was not
available for the year
n
, but was available for one of the years
n
−
1 or
n
+ 1,
then that one was used instead. If the observation was available for both
years
n
−
1 and
n
+ 1, then their average was calculated. Such an approach
does not have a big influence over the variability of independent variables
and over their overall effect. For instance, one would likely expect the index
of economic freedom to be rather stable over a short time span, and that the
observations for three consecutive years are somewhat similar. Thanks to
the fact that the vast part of data was available, imputation was necessary
only for 14 countries, which represents 10.76% of the total number of states.
Overall, 39 numbers were calculated using imputation (in the case of some
countries, finding missing data using imputation has been used more than
once - either for different regressors or for different time periods), which is
1% of the overall amount of data. It means that data imputation has a very
minor effect over the final results.
It is important to mention that the nature of macroeconomic data im-
plies a rather strong relationship between various factors. No variable can
be viewed individually, the context being always of crucial importance be-
cause the vast majority of them are interconnected - either through policies
[M¨
ugge, 2016], or through people’s reactions to them. Therefore, the follow-
ing correlation matrix, which demonstrates a strong positive relationship
between almost all variables, is expected:
16
Figure 2: Correlation Matrix
17
5
Data Processing
The purpose of this section is to describe the overall process of data analysis,
including the choice of the empirical model and of the estimation method.
Besides that, particular attention is dedicated to testing various multiple
linear assumptions and to solving them in the case of potential issues.
The main point of interest of this paper is the following: how do different
factors affect the growth of tourism in the country. Therefore, the equation
behind the main model has the following form:
ln
(
tourists
it
) =
X
it
β
+
a
i
+
u
it
(1)
In this case, every country has a certain index
i
, which can have values
between 1 and 130. The index
t
represents the time period and can have
values between 1 and 5 (for the years 1995, 2000, 2005, 2010, and 2015,
respectively). The vector
X
it
contains all five independent variables that
are expected to influence the variable
tourists
it
. Besides that, the vector
β
includes all the unknown parameters of interest [Wooldridge, 2010, p.83],
a
i
is the fixed effect, and
u
it
is the unobserved error. The model is estimated
using the fixed effects, followed by testing of the MLR assumptions.
5.1
Fixed Effects
The purpose of this section is to briefly present the concept behind the fixed
effects estimation. Equation 1 can be expressed as following:
y
it
=
X
it
β
+
a
i
+
u
it
,
(2)
where
y
it
represents any dependent variable in a generalized way. After-
wards, the average values over time of all variables need to be calculated
and then introduced in the form of equation 2. The last step is to get rid of
the fixed effect through subtracting equation 4 from equation 2:
¯
y
i
=
P
5
t
=1
y
it
,
¯
X
i
=
P
5
t
=1
X
it
,
¯
u
i
=
P
5
t
=1
u
it
(3)
¯
y
i
= ¯
X
i
β
+
a
i
+ ¯
u
i
(4)
18
y
it
−
¯
y
i
= (
X
it
−
¯
X
i
)
β
+
u
it
−
¯
u
i
(5)
5.2
MLR Assumptions
There are several Multiple Linear Regression assumptions that need to be
tested in order to ensure that the estimators obtained are BLUE - Best Lin-
ear Unbiased Estimators, according to the Gauss-Markov theorem [Kariya,
1985].
The first assumption of linearity in parameters is met by the definition
of equation 2. The second assumption of random sampling is not applica-
ble since macroeconomic studies deal with the overall database instead of
working with samples of it.
An important assumption that will be tested is no multicollinearity
among independent variables, which means that they should not be strongly
related to each other. According to Wooldridge [2010],
multicollinearity
becomes an issue when the variance inflation factor (VIF) is higher than 10:
V IF
j
=
V ar
( ˆ
β
j
)
SST
j
σ
2
>
10
(6)
Here,
SST
j
is the total sum of squares of
j
,
V ar
( ˆ
β
j
) is the variation of
the slope
β
j
, and
σ
2
is the variance of the error.
The Variance Inflation Factors for each of the included independent vari-
able are shown in Table 1.
Independent Variable
Variance Inflation Factor
GDP per capita growth
1.0089
Emissions
1.0201
Economic freedom
1.1304
Intentional Homicides
1.0306
Education growth
1.1962
Dummy for the year 2010
1.1143
Table 1: VIF
As noticed, all indicators are substantially lower than 10, which implies
the lack of multicollinearity among the included explanatory variables.
In order for the estimators to be BLUE, the variance of all the errors
should be constant. If this assumption is violated, then
heteroskedasticity
is found. This is verified using the Breusch-Pagan test for heteroskedastic
19
disturbances [Breusch and Pagan, 1979], one for each regression. The re-
sults reveal the presence of heteroskedasticity since the null hypothesis of
homoskedasticity is rejected under a p-value of 1
.
542
∗
10
−
12
.
Besides that, there should be
zero serial correlation
between the dis-
turbances of the same observation, but in different periods of time:
Cov
(
u
ik
, u
il
) = 0
,
∀
k
6
=
l, k, l
= 1
, ...,
5
(7)
This assumption is verified using the Breusch-Godfrey test [Breusch,
1978]. According to it, the null hypothesis of no autocorrelation is rejected
in favor of the alternative hypothesis since the p-value is equal to 2
.
2
∗
10
−
16
.
Heteroskedasticity and autocorrelation were encountered as problems
in the regressions.
Thus, standard errors that are robust to the above-
mentioned issues should be used instead of the default ones. Newey and
West [1987] designed standard errors that can deal with these problems,
and therefore will be presented in the Results section.
20
6
Results
The purpose of this chapter is to present the outcomes of the previously
created model and to comment them using relevant theories and calcula-
tions. Table 2 shows the effect that various factors have over the dependent
variable. Here, the coefficient of determination
R
2
is equal to 0.574, which
means that 57.4% of the variation in the tourism growth is explained by the
included explanatory variables. It is a rather high indicator which demon-
strates the validity of this study. That is because the number of tourist
arrivals is affected by a multitude of details, many of them being qualitative
and therefore hard to quantify, such as the cultural value of destinations or
the popularity of certain nature reserves.
Table 2 shows the estimates related to each independent variable, as
well as its statistical significance. The latter is also demonstrated through
the t-values, which can be found in the parentheses under each estimate.
Thus, we see that all variables, except for the dummy for the year 2010, are
statistically significant at least at the 5% significance level. It proves the
importance of these determinants in the tourism development.
According to the first independent variable, which is statistically signif-
icant at the 1% significance level, GDP per capita growth has the ability to
positively influence the number of tourists arrivals in the future. In this case,
the ”log-log” model should be interpreted in the following way: a 1% increase
in the GDP per capita will lead, after five years, to a 1.1084% increase in the
number of tourists, keeping other factors fixed. Such a big influence comes
from the fact that the economic development is strongly related to a series
of other factors, such as economic openness and the number of operating
businesses. This result demonstrates the fact that the relationship between
tourism and economic growth is bilateral.
21
Tourism Growth
(t-values are in parentheses)
lag (GDP per capita growth)
1
.
1084
∗∗∗
(6
.
9498)
emissions
−
0
.
0018
∗∗∗
(
−
4
.
4722)
emissions
2
−
0
.
0001
∗∗
(
−
2
.
4311)
economic freedom
0
.
009
∗∗
(1
.
9858)
intentional homicides
−
0
.
0076
∗∗
(
−
2
.
1518)
growth of education
1
.
506
∗∗∗
(7
.
3647)
dummy for 2010
0
.
0027
(0
.
0892)
Observations
517
R
2
0.574
F Statistic
85.580
∗∗∗
(df = 6; 381)
Note:
∗
p
<
0.1;
∗∗
p
<
0.05;
∗∗∗
p
<
0.01
Table 2: Overall Regression Results
22
As previously mentioned, the variable
emissions
has a quadratic form:
x
+
x
2
. Because of that, the interpretation of the coefficient that corre-
sponds to this regressor is more complex. What can be concluded here
is that the level of emissions over people’s decisions to travel is negative
and statistically significant. Both independent variables that represent CO
2
have negative values and are statistically significant at the 1% and at the
5% significance levels, respectively. It means that higher emissions have a
detrimental effect over tourism development even at low CO
2
levels, and the
impact gets stronger once this independent variable increases its value. Ce-
teribus paribus, the nonlinear effect of this regressor over the level of tourism
growth can be represented graphically as in Figure 3.
Figure 3: Emissions - Tourism relationship
There is a positive effect of a higher index of economic freedom over
tourism growth, this variable being significant at the 5% level since its t-value
is equal to 1.9858. Such an influence could be expected because economic
liberty is generally beneficial for the country’s well-being. It is favourable
for the state’s legal structure, facilitates the usage of alternative currencies,
and benefits private ownership [Carlsson and Lundstr¨
om, 2002]. As result,
more people have incentives to open businesses, including the enterprises
23
related to tourism. This increases the variety of services provided in the
tourism industry. Consequently, it has a positive effect over the popularity
of the destination and enhances the number of visitors.
The level of security, expressed through the number of intentional homi-
cides per 100,000 inhabitants, is negatively associated with the number of
tourists visiting the country, and this effect is significant at the 5% level.
According to Ryan [1993] and as previously described in Chapter 3, the
relationship between crime and tourism exists, and it can be either direct
or indirect. Regardless of its nature, though, it is clear that more illegal
activity in the region will demotivate tourists from visiting it.
In the long term, education can be of the primary drivers of the country’s
economic development. Besides that, it also can have a positive effect over a
series of factors, such as firm performance [Wang et al., 2008] or productivity
of the workforce [Okoro and Washington, 2012]. The created model demon-
strates that education can also influence positively the growth of tourism in
the country. That is, keeping other factors fixed, a one percent increase in
the average years of schooling per person would lead to a 1.506% increase in
the number of tourists arrivals. This effect is rather indirect due to the fact
that education increases the human capital which, in turn, is expected to
escalate the efficiency of labour and the development of infrastructure. Be-
sides that, higher levels of education are strongly related to people’s cultural
and social tolerance and openness - factors that are crucial in the context of
international travels.
It is important to mention that the dummy variable for the year 2010
is insignificant. It means that, keeping other factors fixed, the post-crisis
year 2010 did not have a huge impact over the growth of tourism across
the globe. That is, the number of international arrivals slightly decreased
only for the year 2009, but its growth after that was rather steady and
continuous, as observable in figure 4. Therefore, the results obtained in this
study are in line with the theory previously presented in chapter 3, which
says that after the year 2009, the tourism sector recovered quickly and then
continued to develop steadily. It can, however, be supposed that the growth
would have been even more considerable if not the crisis. Such a hypothesis
is also supported through the fact that the number of international arrivals
increased even more between the years 2010 and 2015 than it was growing
between any other five-year period before that.
24
Figure 4: Number of international arrivals
25
6.1
Effects by region
Besides presenting and analyzing the overall results, it could be useful to
divide the countries based on certain geographical, economic or social dif-
ferences, thereby getting more specific outcomes and providing with more
distinct policy suggestions. Thus, the initial dataset has been divided into
the following groups:
1. Developed European countries (GDP per capita in 2015
>
15,000
$
)
2. Post-soviet and post-socialist states
3. Africa
4. Latin America
5. Big countries (
>
750
,
000
km
2
)
6. Small countries (
<
50
,
000
km
2
)
The same regression has been used as with the overall dataset, as pre-
sented in equation 1. Besides that, all the MLR assumptions have been ver-
ified according to the description in the Section 5.2. The numerical results
of all the tests are shown in the Appendix. The following figure indicates
the countries included in each dataset:
Figure 5: Regions
26
6.1.1
Developed European Countries
The choice has been made to filter only developed European countries be-
cause they share a series of geographical and economic characteristics. First,
they are easily reachable among each other thanks to shorter distances than
on other continents. Besides that, tourism in these states is encouraged
through a developed transportation infrastructure and, lately, thanks to a
variety of relatively cheap airplane tickets. Moreover, there is an increas-
ing difference between the growth and the maintenance of infrastructure in
developed and in developing countries, respectively. The states with lower
GDP per capita cannot obtain the investments that this field is currently
demanding, whereas more advanced countries encounter fewer problems in
satisfying the infrastructure necessities [Ershova and Posokhov, 2016, p.818].
All these countries have quite high GDP per capita, which implies bigger
personal earnings for each citizen. As result of income surpluses, people are
willing to spend more on recreational activities such as traveling, and es-
pecially to neighboring countries. These common features could potentially
result in interesting regression outcomes and in distinct conclusions regard-
ing them. The regression results are presented in the first column of Table
3.
According to it, the number of international tourists arrivals in developed
countries is not related to their previous economic growth. Such an effect is
generally expected due to the economic convergence [Mankiw et al., 2007,
p.220], which could subsequently be explained through the law of diminish-
ing returns [Shephard and F¨
are, 1974, p.69]. It says that, as the country
continues to develop, then its marginal gains to economic growth will get
lower. Therefore, it can be assumed that after reaching a certain level of
GDP per capita, the country’s level of economic development ceases to be a
driving force for the tourism growth.
The statistically significant negative effect of the emissions on the number
of tourists in developed countries could be explained through the environ-
mental Kuznets curve [Cole et al., 1997]. It assumes that the relationship
between the country’s level of economic development and its emissions is
nonlinear, but rather has an inverted U-shape. That is, if the GDP per
capita is high enough, then more economic growth would be associated with
lower emissions. In our case, the vast majority of developed European coun-
tries have experienced an escalation in the number of tourists, but at the
27
Dependent variable:Tourism Growth
(t-values are in parentheses)
1. Developed
2. Post-Socialist
lag (GDP per capita growth)
0
.
0523
−
0
.
2021
(0
.
9795)
(
−
1
.
4757)
emissions
−
0
.
0003
∗∗∗
0
.
0001
(
−
3
.
9221)
(0
.
0052)
emissions
2
−
0
.
0001
∗∗
0
.
0009
∗∗∗
(
−
2
.
1788)
(4
.
2269)
economic freedom
0
.
0127
∗
0
.
0382
∗∗
(1
.
6896)
(2
.
1305)
intentional homicides
−
0
.
0831
∗∗∗
−
0
.
0421
∗
(
−
4
.
8694)
(
−
1
.
7559)
growth of education
1
.
8447
∗∗∗
3
.
4732
∗∗∗
(6
.
0739)
(3
.
4877)
dummy for 2010
−
0
.
1139
∗∗∗
−
0
.
1106
(
−
3
.
5291)
(
−
0
.
6949)
Observations
121
98
R
2
0.657
0.433
F Statistic
28.695
∗∗∗
(df = 6; 90)
9.162
∗∗∗
(df = 6; 72)
Note:
∗
p
<
0.1;
∗∗
p
<
0.05;
∗∗∗
p
<
0.01
Table 3: Regression Results. Developed European & Post-Soviet and Post-
Socialist
28
same time lower CO
2
emissions. However, there is not only empirical corre-
lation, but also logical causation between these two factors. Lower emissions
are especially important in the context of nature-based and slow tourism,
whose popularity has increased substantially during the last decades. Many
developed countries have started to promote these kinds of tourism, thereby
getting more visitors. However, since the variable
emissions
2
is also neg-
ative and statistically significant, then a very big decrease in the level of
emissions is expected to have a weaker positive effect over the number of
tourists as previously described in Figure 3. One of the reasons might be
the fact that international travelling is strongly related to high emissions
coming from different ways of transportation. This negative effect partly
cancels the above-discussed positive externalities. One can expect that more
ecological transportation methods would contribute to lower CO
2
emissions
related to tourism. Another conclusion consists in the fact that, in the con-
text of sustainable destination development, decreasing the CO
2
emissions
would be beneficial not only from the ecological perspective, but it will also
serve as a primary driver for the subsequent touristic growth.
The level of economic freedom is a less important factor in determining
the growth of number of visitors in developed countries because supposedly
these states have already reached a high enough level of economic liberty,
after which additional improvements become less relevant.
Dostları ilə paylaş: |