The spillover of tourism development on CO2 emissions: a spatial econometric analysis.

Climate change and tourism's interaction and vulnerability have been among the most hotly debated topics recently. In this context, the study focuses on how CO2 emissions, the primary cause of global warming and climate change, respond to changes in tourism development. In order to do so, the impact of tourism development on CO2 emissions in the most visited countries is investigated. A panel data from 2000 to 2017 for top 70 tourist countries are analysed using a spatial econometric method to investigate the spatial effect of tourism on environmental pollution. The direct, indirect, and overall impact of tourism on CO2 emissions are estimated using the most appropriate generalized nested spatial econometric (GNS) method. The findings reveal that tourism has a positive direct effect and a negative indirect effect; both are significant at the 1% level. The negative indirect effect of tourism is greater than its direct positive effect, implying an overall significantly negative impact. Further, the outcome of financial development and CO2 emissions have an inverted U-shaped and U-shaped relationship in direct and indirect impacts. Population density, trade openness, and economic growth significantly influence environmental pollution. In addition, education expenditure and infrastructure play a significant moderating role among tourism and environmental pollution. The results have important policy implications as they establish an inverted-U-shaped relationship among tourism and CO2 emissions and indicate that while a country's emissions initially rise with the tourism industry's growth, it begins declining after a limit.

Introduction

Recently, there have been significant changes in the global climate system. The American Meteorologic Society’s state of the climate report states that world surface temperatures were 0.38–0.48 °C higher than the average 1981–2010 and that since 1998, the top 10 years have been all the warmest, with four warmest years occurring since 2014 (Abbas et al., 2020; Fu et al., 2021; Yumei et al., 2021a, b). The analysis stresses that, as a major contributor to global warming, worldwide growth rates of CO2 emissions have almost doubled since the beginning of the 1960s. It also notices observed changes such as ice and snow declines, increases in level of sea, and seasonal durations. This challenge affects economic, political, living, geopolitical, and social growth directly. Climate change and global warming have led to famine, sickness, floods, and water shortages for millions of people (Akbar et al., 2021; Anser et al., 2020a, b; Iqbal et al., 2019a). Scientists today agree that the fundamental explanation for climate change and global warming is the fast increase in CO2 emissions over the past 50 years. Because of its rapid expansion, it is recognized as one of the most energy-intensive industries in the world (Chien et al., 2021a, b; Iqbal et al., 2021; Zhang et al., 2021a). Travelling and staying in hotels, for example, are both known to absorb high frequencies of energy, which has a negative influence on CO2 emissions. Despite its importance of energy consumption prediction and its role in promoting energy consumption patterns, tourism development has received little attention in literature. The same is true for energy consumption, with it being recognized as one of the key factors of energy. This has led to the conclusion that tourism expansion has a variety of effects on natural resource consumption, economic growth, CO2 emissions, and energy consumption through a number of different avenues. Because of this, it cannot be ignored within the empirical framework. The findings of the authors revealed that factors relate positively to economic growth across the board, although its impact on CO2 emissions differs. Tourism, according to the authors, reduces CO2 emissions in the Western EU while increasing in the Eastern EU. The authors ascribe these conclusions to the Western EU’s efficient integration of appropriate tourism rules, which has resulted in a reduction in tourism’s negative environmental impact. Muhammad Khalid Anser et al. (2020a, b) discovered that the tourist sector helps both established and developing economies’ economic development. Surprisingly, although the association among environmental quality and tourism development is considerable. In tourism research, the rapid growth of carbon dioxide emissions has not yet been assessable. Studies focus mostly on the detrimental consequences on the tourism industry of global warming and climate change. Particularly, it is stressed that the climate influences visitor activities, tourist destinations, and overall holiday enjoyment. According to existing research, unfavourable weather conditions are a driving force for tourism development, while favourable weather conditions are seen as a desirable feature. It is also obvious that the influence of climate change and global warming on coastal locations (in particular, as the sea levels rise) has major implications for the tourism development.

In the context of the interaction between climate change and tourism, it is vital to seek answers to the following questions: (i) How does tourism affect CO2 emissions while climate change and global warming have a detrimental effect on the tourism industry? (ii) Does the tourism sector’s development contribute to global carbon emissions? Danish and Wang (2018) investigated the contribution of tourism to global greenhouse gas emissions in theory, emphasizing that the aviation sector contributes significantly to greenhouse gas emissions.

Because of the debate’s multidisciplinary nature, research volumes with new and diverse perspectives consistently increased in the 2000s. However, in the existing research, the impact of tourism on CO2 emissions has not been statistically satisfactory (Carr-Harris and Lang, 2019; Lee and Brahmasrene, 2013; Sigala and Leslie, 2005; Sun, 2016a). Nonetheless highlight the extent to which carbon emissions can be accountable for the tourism sector and contribute significantly to calculating tourism-related carbon flows. In contrast, other research (Arai and Goto, 2017; Simkhada et al., 2016) examines the ecological impact of tourism through qualitative assessments. It can be said that new empirical research has been carried out to take into account the effects of carbon emissions in tourism development. Nations are dependent on tourism to achieve economic progress and extract up to the level of threshold findings in natural environmental deterioration. All economic activities raise CO2 levels, resulting in an increase in waste production.

Tourism is a significant economic industry in both developed and developing countries. Today, tourism’s economic impact matches or exceeds that of oil exports, food products, and automobiles: 9% of GDP; 1/11 employment, direct, indirect, and induced; 6% of global exports; 1.4 trillion in exports; 30% service exports ((UNWTO), 2020). Their economic, social, and political significance is now an irrefutable reality that is through a period of rapid worldwide growth: International tourist arrivals increased by 5% to 1.087 billion in 2013; international tourism earned US$ 1.4 trillion in export profits in 2013; the UNWTO anticipates a growth of between 4% and 4.5% in international tourist arrivals in 2014. One fundamental source of tourism’s ever-increasing carbon emissions is the relentless pursuit of tourism maximization through “boosterism” economic policies, in which continued growth in visitor volume and expenditure is driven by economic imperatives with little regard for the social and environmental impacts of tourism (Hall, 2009). Due to the political and social impossibility of significantly reducing tourist volume, the research has offered an alternate solution through tourism system optimization. Indeed, in a post-COVID-19 environment where visitor numbers are expected to decline dramatically in the medium term, market optimization can optimize the impact of priority market segments to tourism recovery. Carbon reduction methods are an important and long-overdue requirement for the tourism industry. Tourism’s approach to climate change has been to urge businesses to adopt new technologies and to promote more sustainable guest behaviour. These initiatives, however, are insufficient to address tourism’s growing carbon footprint. We propose a novel carbon mitigation strategy based on the concepts of optimization and eco-efficiency. It aims to pro-actively identify, cultivate, and create a long-term tourism market portfolio. This can be accomplished through interfering and reshaping demand with the overarching goal of encouraging low-carbon travel markets. The concept and analytical framework for optimizing the intended market mix quantitatively are presented.

The rest of the paper is organized as follows: Sect. 2 considers two strands of literature on the tourism–environment nexus and the education–environment nexus. Section 3 explains the model, data, and methodology, while Sect. 4 presents the estimation results and a discussion of the findings. Section 5 presents the conclusion.

Literature review

Tourist industry will continue to increase in the future, as per the United Nations World Tourism Organization (UNWTO). Moreover, in terms of new tools and management styles, the nature of development and growth will be different than in past decades. The use of new technology in a labour-intensive industry improved the performance, efficiency, and effectiveness of tourism-related services while also highlighting the need for more skilled workers in this field. One of the most serious issues in the field of tourism is its year-round focus on a few, limited activities.

Tourism activities have also increased in terms of social and environmental consciousness. The observed and develop approaches and technologies that will improve the sector’s performance in the future. The increasing number of consumers (who are more aware and demanding, with higher levels of education and ability) will encourage the tourist industry to produce new products and reimagine established markets (Chien et al., 2021a, b). Economic expansion is frequently related with an increase in emissions. Several studies examine the econometric relationship between economic growth carbon dioxide emissions. Al-mulali et al. (2015) examined Europe, Khan and Ozturk (2021) examined Central America, Iwata et al. (2011) examined 28 nations (OECD and non-OECD), Jalil and Mahmud (2009) examined China, and Ozturk and Acaravci (2010) examined Turkey.

Tourism produces around 5% of global CO2 emissions. Transportation contributes for 75% of the sector’s emissions (of which 50% are from air transport), while accommodation accounts for 22% and rest of the activity-related tourism account for 4%. While tourism is not regarded a particularly polluting activity, estimates for tourism expansion indicate that emissions from tourism activities will more than quadruple by 2035.

Against this backdrop, there were three pertinent concerns to address. To begin, it is critical to understand the variables that contribute to pollution.

Second, are any tourism activities that bear a bigger share of responsibility for climate impact? Tourism activities in Nepal account for around 10% of overall CO2 emissions. Transportation accounts for around 89% of tourism-related CO2 emissions, while accommodation and food and beverages each contribute for approximately 5%. Accommodation services grew at the fastest rate of CO2 emission growth (48.4% between 2000 and 2008), Food, beverage, recreational services, and cultural sports came in second and third, with 34% and 25%, respectively. However, when it comes to CO2 emission intensity (emissions divided by value created), by lowering their performance, all tourist subsectors improved, but this variable declined significantly in transport and travel agencies (33.7% and 30.9%, respectively). Accommodation services, on the other hand, had an increase in intensity of carbon emissions of 5.6%.

Thirdly, are the regulations applicable to each tourism subsector aligned with the primary drivers of emissions in each? Except for the aviation industry, which was included to the Trading System of European Union Emissions (EU ETS) in January 2012 (Directive 2008/101/EC), the European Union Emissions Trading System (EU ETS) excludes tourism subsectors. The maritime industry can only be covered by the first policies aimed at reducing emissions in 2018. Additionally, there is a current policy aimed at reducing emissions from new cars to 95 g/km by 2020 (European Commission, 2012). Only few research study the links among energy consumption, carbon emissions, and tourism utilizing regression technology and analysis of decomposition. The econometric technique is used to analyse and identified the impact of one variable to the other. In particular, Lee and Brahmasrene (2013) found that the impact of international tourism and foreign direct investment on reducing CO2 emissions is significantly higher among European countries, whereas Tiwari et al. (2013) found that international tourism in a small island like Cyprus has a positive statistically significant effect on energy and CO2 emission consumption.

There is not much research on the decomposition analysis methodology to investigate the effects of different influences on CO2 emission variations caused by tourism. However, several studies have broken out the link between carbon emissions and the broader energy–economic system. Ozturk et al. (2021) employed decomposition techniques for assessing CO2 emission intensity and its components across 36 Portugal’s economic sectors and also included decomposition of the forecast error variance and impulse response functions applied to decomposed emission intensity factors. On research specific to tourism, the work of Adedoyin et al. (2021) who apply decomposition analysis in China can be indicated. They concluded that energy intensity, expenditure size, and industry size were the main drivers for emission growth for tourism but that two other parameters had no significant impact on the upsurge of emissions from the tourism industry. Other studies are available which degrade emissions from tourism-related activities such as travel, some services, and accommodation Iqbal et al. (2019b) showed that the impact on scale is important for Ireland when it comes to raising emissions in the transport sector and that energy intensity improvements in the residential or service sector are significant. In another study for China, Fu et al. (2021) concluded that the major factors for reducing CO2 emissions are the effect of transport intensity and transport services. Baloch et al. (2020a, b) compare the numerous decomposition analytical methods globally and advocate the Logarithmic Mean Divisia Index (LMDI) because it is theoretically well founded, easy to adapt, utilize, and interpret its results. In addition to Iqbal et al. (2021), worldwide or in Portugal, applied literature on the LMDI sector and subsectors is scarce.

Zhang et al. (2021b) examine the association between economic advancement and environmental pollution in countries throughout the Europe. Their statistically enhanced findings demonstrate that the application of spatial economic approaches improves model formulation. Anser et al. (2020a, b) investigated the environmental Kuznets curves (EKC) concept in China using a geographic panel data model. Their findings confirm that the spatial panel model outperforms traditional panel approaches, as standard econometric techniques can generate erroneous parameter estimates. Khokhar et al. (2020a, b) examined the regional correlation of carbon intensity in China using a spatial panel data model. The present study’s findings confirm the existence of a spatial association between CO2 intensity in Chinese provinces.

The impact of education on carbon and methane emissions is analysed by Mishra et al. (2020) through the panel methodology for 181 countries, suggesting a negative effect on methane emissions due to education, whereas carbon emissions remain unaffected. Along with a number of control variables, Sovacool et al. (2021) use data from Latin American countries to explore the effect of foreign direct investment and human capital on pollution emission in the environmental Kuznets curves (EKC) framework. Moreover, a negative association between human capital and emissions for high-income countries and a positive association for low-income countries are evident through the panel technique results. The EKC framework is used by You and Lv (2018) to investigate the relationship between education and environmental quality for data collected from Australia; a U-shaped connection between education and emissions is evident through the autoregressive distributed lag (ARDL) results, which suggests the importance of education in reducing emissions after the threshold is reached. Khan et al. (2019) analyse the association of education with carbon emissions by using the ARDL econometric technique through different proxies for Pakistan, reporting a negative association between carbon emissions and education.

However, the in-depth development of the tourism economy is promoted by incorporating a strong economy and improved quality of life in Southeast Asian economies, providing an overall high-quality environment in the long run. Hence, developing and developed countries observe an inverted U-shaped relationship between tourism development, renewable energy, financial development, and CO2 emissions. Similarly, the development of any tourism industry depends on a number of cross-regional cultural and industrial exchanges and factors, and not merely administrative factors.

Tourism and the environmental progress

Several recent studies have explored the connection between tourism and CO2 emissions. Abbas et al. (2020) studied the connection of tourism and economic growth with the emission of CO2 by the Johansen cointegration test for European nations between 1988 and 2009. The empirical data suggest that tourism boosts economic growth as well as CO2 emissions in the area of investigation. In the example of Malaysia, Mohsin et al. (2021) described the long-term ratio between CO2 emissions; moreover, the data revealed a unilateral causal link between tourism and CO2 emission. Similarly, Abbas et al. (2021) have shown that the causality from tourism to CO2 and GDP to tourism is unilateral. They used GMM for data from Asia-Pacific countries between 1995 and 2013. Yumei et al. (2021a, b) observed a considerable contribution to CO2 emissions by tourism. In addition, Iqbal et al. (2020) showed that a single-way causal relationship for the research area is between CO2 emission and tourism. Testing data for Malaysia from 1972 to 2010, Lawal et al. (2018) applied the ARDL model that reveals that the link between tourist arrivals and CO2 emissions has been long-term positive. In a similar study, Nkoro and Uko (2016), Sharif et al. (2020), and Zhang et al., 2021) found the effects of tourist arrivals and the usage of energy on the environment in Tunisia, using panel data from 1995 to 2010. They show that tourist arrivals cut CO2 emissions in the long run. Testing data from 10 Northeast and Southeast Asian countries, Yurtkuran (2021) discovered that improvements in tourist development can help to regenerate the environmental amenities in the longer term, but that tourism has been seen to be a degrading aspect in the region’s ecosystem.

Data and methodology

Spatial model

Spatial impacts are critical when examining the relationship between growth, emissions, and energy. Additionally, many of the concerns in economics, the environment, and energy are intrinsically spatial (Su and Ang, 2010). Despite the fact that several studies have been conducted to examine the relationship with both carbon emissions and growth using spatial econometric techniques, no research has examined the spatial effects of the nexus between economic growth and renewable energy or between income and carbon emissions, while numerous prior researches have emphasized the importance of geographical impacts in growth in the economy, CO2 emissions, and energy studies. For example, Li and Lv (2021) use spatial autoregressive panel data estimate techniques to analyse the effect of adjacent regions’ growth in the economy on a state’s own economy. However, the in-depth development of the tourism economy is promoted by incorporating a strong economy and improved quality of life in Southeast Asian economies, providing an overall high-quality environment in the long run. Hence, developing and developed countries observe an inverted U-shaped relationship between tourism development, renewable energy, financial development, and CO2 emissions. Similarly, the development of any tourism industry depends on a number of cross-regional cultural and industrial exchanges and factors, and not merely administrative factors.

This process generates spatial spillover effects from tourism development through a significant channel. The spatial econometric models used include the spatial lag model (SLM), spatial Durbin model (SDM), and spatial error model (SEM). SLM is used when the dependent variable is spatially correlated. Hence, this study establishes a generalized nested spatial model (GNS), considering the spillover impact of tourism growth, renewable energy, financial development, and other control variables on CO2 emissions.

In Eq. (3), the variables lntourism, lnurban, lnCO2, lnGDP, and lnEI stand for per capita CO2 emissions, urbanization rate, energy intensity, and tourism development, respectively. Tourism contributes to CO2 emissions by increasing demand for transportation, which is exacerbated by the intensity of travel services. Furthermore, tourism development increases food consumption and shopping activities, both of which contribute to carbon emissions.

The logarithm of CO2 emissions in country i at time t is represented as Y, the coefficient of spatial regression as ρ, the control variables as Z, and the error term as ε. The inverse squared distance matrix applied in this study considers neighbouring nexuses as nonlinear—compared with the distance, the decrease is more rapid. The matrix is normalized row-wise and is aligned with previous studies. The matrix representing spatial weights where i and j refer to the element in row i and column j is given as w and the vector of independent variables as x.

Spatial autocorrelation coefficient

As carbon emissions correlate between regions and are considered heterogeneous, the spatial econometric model is built such that emissions of CO2 are considered the main determinant for regional CO2 emissions spatial correlation with spatial spillover and spatial diffusion, which influences the CO2 emissions of neighbouring countries. Moran’s I calculates spatial correlation as follows (Moran, 1948):where i represents spatial units and j is given as N, the concerned variable as y, average of y as , the spatial weights matrix as w, and the sum of the weight entries as W. The statistical significance for each I is evaluated by considering Moran’s I of individual spatial units. Therefore, a positive correlation is represented by a positive I, whereas a negative correlation by a negative value.

Kelejian and Prucha (2010) defined queen contiguity weights. The relations among n units are summarized by the suitable spatial model that is determined through the matrix of spatial weight w. The observations of spatial arrangement for the models in this study are done using spatial contiguity weights, which show the boundaries that are shared by the spatial units. The following equation defines the spatial weighting matrix w, where bnd(i) represents the set of boundary points of unit i:

Data sources

World Bank Development Indicators (WDI) online databases and the international energy agency (IEA) were used to collect the annual data for 2000–2017. Appendix 1 (Table 9) represents the sample used in our study, based on the available data, which was restricted to top 70 tourist countries. The effects of tourism development on environmental degradation were assessed in this study through seven variables. The core independent variable constituted tourism and financial development, whereas CO2 emissions, considered one of the major causes of global warming, was considered the dependent variable. Furthermore, our baseline model considered several control variables to prevent omitted variables, such as GDP per capita (GDP), financial development (FD), renewable energy (RE), population density (Pd), trade openness (trade), and education expenditure (Edu), from causing any bias. Appendix 2 (Table 10) presents the definitions, descriptive statistics of the variables, and data sources.

Table 9

List of countries contributed in this study

Australia	Belgium	Poland	Pakistan
Bangladesh	Bosnia and Herzegovina	Portugal	Qatar
China	Bulgaria	Romania	Saudi Arabia
Nepal	Croatia	Russian Federation	Russia
India	Denmark	Spain	United Arab Emirates
Indonesia	Finland	Sweden	Brazil
Japan	France	Switzerland	Canada
Korea, Rep.	Germany	Turkey	Mexico
Malaysia	Greece	UK	Panama
New Zealand	Hungary	Azerbaijan	Peru
Philippines	Iceland	Bahrain	USA
Singapore	Ireland	Egypt, Arab Rep.	Venezuela, RB
Sri Lanka	Israel	Iran, Islamic Rep.	Czech Republic
Thailand	Italy	Kazakhstan	Luxembourg
Austria	Netherlands	Kuwait	Moldova
Belarus	Norway	Kyrgyz Republic	Slovak Republic
Ukraine	Algeria	Turkmenistan	Chile
Colombia	St. Vincent and the Grenadines

Table 10

Summarized statistics

		Mean	Min	Max	S. dev
lnCO2	Carbon emissions in metric tons	0.829	1.403	−4.059	3.204
lnTourism	Tourism development indicator	1.542	−1.115	5.598	2.675
lnFD	Financial development indicator	6.186E−08	1	−2.919	3.349
lngdp	(Logarithm of) GDP per capita, PPP (constant 2011 international $)	9.254	1.115	6.301	11.355
lnRE	GDP per unit of energy consumption (2011 constant PPP $ per kg of oil equivalent) in log form	2.099	0.47	0.347	3.045
lnEdu	Government expenditure on education, total (% of government expenditure)	4.273	0.699	−1.787	6.081
lnTrade	Trade openness.	4.273	0.699	−1.787	6.081

Notes: Annually 2000–2017. World Development Indicators

A number of studies use proxies for tourism, and this study follows the index of tourism by Khan et al. (2020a, b). With the weights as the focus, three individual variables, the tourist receipts (TR), expenditures on tourism (TEX) both in US$, and number of tourist (TA) are utilized to establish the index in this study. We use the world bank online database to collect data for all the selected variables. Similarly, a single weighted index is constructed by applying the principal component analysis (PCA) on the tourism variables, an approach with internal correlation used for examination and diagnosis. The new variables calculated and categorized as principal components are for this study to reduce the amount of data needed through this method, and the index of tourism development through PCA is given in Table 1. The maximum eigenvalue for the first, second and third are respectively 2.404, 0.4748, and 0.1207, as shown in the first segment of Table 1. Similarly, the highest proportion of variation recorded at 80.14% is given in the first component, at 15.83% for the second component, and the lowest change recorded at 0.403% for the third factor. Furthermore, the eigenvalue loading in three-components in the second segment, including PC1, PC2, and PC3 is shown in Table 1, whereas this study establishes the index of tourism development through the second and third components, indicating smallest negative loadings values. Similarly, correlation between variables is given in the last segment of Table 1, where tourist arrivals positively correlate with tourism expenditures and tourism receipts, whereas the selected countries also observe a correlation between tourism expenditures and tourism receipts.

Table 1

Tourism development index

Component	Eigenvalue	Difference	Proportion	Cumulative
Comp1	3.845	3.693	0.7485	0.7485
Comp2	0.252	0.2487	0.0606	0.899
Comp3	0.0049	–	0.011	1
Eigenvectors
Variable	Comp1	Comp2	Comp3
TEX	0.6905	−0.3018	−0.8814
TA	0.6636	0.8962	0.3202
TR	0.6777	−0.6704	0.6838
Correlation matrix
Variable	TEX	TA	TR
TEX	1
TA	0.622	1
TR	0.5867	0.6579	1

Results and discussion

International tourist arrivals are estimated to grow to 1.8 billion by 2030, suggesting that tourism is the most rapidly growing industry in the world (UNWTO, 2020). Furthermore, the industry also exhibits a geometrical trend in its growth, at the cost of increasing energy consumption, depletion of natural resources, and waste generation. Theoretically, economic processes, energy use, and the environment are significantly impacted by tourism development. Moreover, CO2 emissions increase because of dirty energy consumption in hotels and transportation (Sun et al., 2020a, b) and (Baloch et al., 2020a, b). The role of tourism development as a factor in energy consumption and stimulating energy consumption patterns has hitherto not been considered as a major factor in the literature. The different channels of tourism development influence economic development, resources, energy use patterns, and carbon emissions, and therefore, the empirical framework (Khosravi et al. 2019; Kordej-De Villa and Slijepcevic 2019; Ozoike-Dennis et al. 2019; Sovacool et al., 2021).

The influence of the financial development of neighbouring countries on a country’s carbon dioxide emissions is stressed by this study and aligns with Lv and Li (2021a). Hence, the significantly positive direct effect is taken over by the negative spillover impact of financial growth on CO2 emission presenting a total effect which is significantly negative. The role of financial development in promoting business growth more than promoting technological progress and green projects increases energy consumption and is a possible explanation for this process, and the findings align with Bui (2020) and Charfeddine and Kahia (2019). Conversely, a unit rise in financial development of neighbouring countries suggests 12.5% decline in carbon emissions of the local country through the negatively significant spillover effect.

An open and free policy focusing on the development of the financial system can provide more research and development funds for the development of energy technology.

For GDP per capita and environmental pollution nexus: GDP per capita significantly impacts on CO2 emission through spatial spillover. The direct impact of GDP per capita on CO2 emission is significant, while the direct impact of (PGDP)2 is negative, proving the existence of environmental Kuznets. Similarly, economic growth and carbon dioxide in developed and developing countries share an inversed-U relationship. In both models, the indirect effect of PGDP and (PGDP)2 on carbon dioxide proves the existence of environmental Kuznets. Nevertheless, pollution in the neighbouring countries is increased due to the local economic development and improved quality of life, transferring high-pollution industries to neighbouring countries. Consequently, an increase in environmental pollution through the increased economic development indicates that the income in these countries positively impacts the left side of the inversed-U curve.

Spatio-temporal distribution of selected variables

The alignment between countries with high carbon emissions is given, same as the countries with low carbon emissions clustered with each other. Hence, a spatial dependence is evident based on the distribution of carbon emissions for sample countries. Moran’s I statistic of carbon emissions through 2000 and 2017 is also estimated by this study, and its results are presented in Sect. 4.2.

A positive correlation is seen between the number of tourists and CO2 emissions in Hélde A.D. Hdom and Fuinhas (2020), who analyse the data for Malaysia, incorporating the autoregressive distributed lag (ARDL) model from 1972 to 2010. The adverse effect of tourist arrivals and energy on the environment in Tunisia is evident through the study by Sharif et al. (2020), which utilizes panel data for 1995–2010 suggesting a reduction in CO2 emissions due to tourism. Notwithstanding the decrease in tourism in East Asian countries, the importance of improving tourism development and regenerating environmental.

However, the study sample is divided in two types: one with tourism development on the left side of the axis of symmetry of the inverse U-shape and the other with tourism development on the right side. Our findings are consistent with Katircioglu (2014), Katircioglu et al. (2014), León et al. (2014), and Li and Lv (2021) that conclude tourism has a positive impact on carbon dioxide emissions. Similarly, our results also show the negative impact of tourism on CO2 emissions, which is consistent with the second category of literature (Katircioǧlu, 2014; Lee and Brahmasrene, 2013; Paramati et al., 2017). Hence, overlooking the nonlinear effect of tourism on carbon emissions is possibly the reasons for the different conclusions. Although most studies have confirmed the empirical relationship between carbon emissions and tourism development, according to the above conclusions, the direction of the causal relationship between the two is still unknown. One of the main reasons for drawing conflicting conclusions is to ignore the nonlinear impact of tourism on carbon emissions (Li et al., 2018a, b; Sun, 2016b). In addition, if the spatial interdependence of the regions is not considered, it may lead to erroneous conclusions (Yang et al., 2019). Therefore, this study uses a panel spatial econometric model to estimate the total impact of tourism on carbon emissions, taking into account spatial dependence and nonlinearity. Next, other control variables also produce valuable results, and the detailed explanation is given below: for financial development and environmental pollution nexus. The primary focus of this study is on financial development. The direct effect for financial development is recorded as significantly positive (0.055), the spillover impact is −0.09215, and the total impact is −0.063, as seen in the empirical findings of financial development. On the contrary, an increase of 5.5% in carbon dioxide emissions with a 1% rise in financial development indicates a significantly positive direct effect. The findings also reflect the financial development and carbon dioxide emissions to constitute an inverse U-shaped relationship, which indicates an increase in carbon emissions of a country with growth in its tourism. However, an eventual 12.5% decrease is expected in carbon emissions of the local country after the threshold is reached. When financial development is higher in the neighbouring countries, the quality of the environment in a local country is affected by technological diffusion, better governance, and more sustainable policies, presenting a possible reason for this process. Consequently, lower carbon dioxide emissions are induced due to the external restrictions across nations (Lv and Li, 2021a). Similarly, the carbon dioxide emissions in the local country are reduced by boosting the spillover of technology, and the transfer of knowledge and skills.

Moran’s I spatial dependence test

A number of studies use proxies for tourism, and this study follows the index of tourism by Khan et al. (2020). With the weights as the focus, three individual variables, the tourist receipts (TR), expenditures on tourism (TEX) both in US$, and number of tourist (TA) are utilized to establish the index in this study. We use the world bank online database to collect data for all the selected variables.

The results of Pesaran’s IPS unit root test and the CIPS (cross-sectionally IPS) unit root test, which are the first and second-generation unit root tests, are shown in Table 2. Urbanization is stationary at the level, according to the IPS test results, while the others have unit root. The null hypothesis is not rejected for all variables in the CIPS test, indicating that all variables in Eq. (3) have unit root. When the first differences are taken, however, these variables become stationary.

Table 2

Moran’s I statistical tests

Year	Moran’s I	p value	Year	Moran’s I	p value
2000	0.2136***	0.001	2009	0.1637***	0.03937
2001	0.2168***	0.0029	2010	0.3645***	0.00445
2002	0.3566***	0.004	2011	0.3454***	0.00321
2003	0.3404***	0.003	2012	0.4172***	0.oo35
2004	0.4656***	0.006	2013	0.3564***	0.0027
2005	0.4295***	0.005	2014	0.3639***	0.001875
2006	0.3708***	0.003	2015	0.4544***	0.002023
2007	0.3691***	0.0025	2016	0.4252***	0.002297
2008	0.0316***	0.0078	2017	0.4674***	0.002118

⁎⁎⁎ indicates 1% significance. The null hypothesis. There is no spatial dependence.

Table 2 shows the estimation results of long-run panel cointegration coefficients that can be interpreted as elasticity because each variable in the models was logarithmically transformed. As a result, a 1% increase in any model variable causes the dependent variable to change by x%, where x refers to the variable’s negative or positive coefficient value. The following are the estimation outcomes: (a) The CUP-FM and CUP-BC estimators reveal that lnGDP has a significant and positive impact on lnCO2 emissions. That is, CO2 emissions rise as GDP per capita, a measure of wealth, rises. (b) Urbanization increases CO2 emissions. (c) The results show that energy intensity and CO2 emissions have a negative relationship. We use energy intensity as a technology indicator because efficiency reduces energy intensity.

Panel unit root and cointegration tests

The causality test was used in this study to uncover potential bidirectional causality relationships between tourism development and CO2 emissions. The results of the causality test are shown in Table 3. To put it another way, tourism development has an impact on CO2 emissions, while CO2 emission changes have a significant positive effect on tourism development. These findings, which show a link among tourism and CO2 emissions, are consistent with those of Wen and Tisdell (2001) and Ma et al. (2015). The results of the causality test also show that there is a bidirectional relationship between energy intensity and CO2 emissions, whereas there is a unidirectional relationship between GDP per capita and CO2 emissions.

Table 3

Cross-section dependence of the variables

Variables	Pesaran CD	Pesaran scaled LM	Breusch-Pagan LM
lnTourism	10.6258***	69.1456***	2013.45***
Urb	45.7362***	176.3654***	2028.369***
lnEI	29.3420***	123.258***	2013.425***
lnPGDP	40.7963***	118.2701***	3012.87***
lnRE	7.6647***	60.5645***	952.3214***
lnEC	29.2134***	102.2389***	164.545***
Receipts	24.6542***	100.2134***	1612.657***

Notes: *** denotes significance at the 1% level

Following this process, the CIPS (Pesaran, 2007) and IPS (Im et al., 2003) are used to perform second-generation panel unit root tests. The second-generation unit root test is preferred over the first-generation root test due to the cross-sectional dependence produced by the CIPS test. Table 4 shows the results for IPS and CIPS tests, and according to the results, 1% level of significance shows all the variables as stationary. Hence, panel cointegration is tested.

Table 4

Results of unit root test

Variables	CADF		CIPS
	Level	First difference	Level	First difference
lnTourism	−2.301	−2.121***	−2.254	−3.258***
Urb	−2.231	−4.102***	−2.354	−4.124***
lnEI	−1.926	−3.145***	−1.452	−2.547***
lnPGDP	−2.408	−3.514***	−3.214	−3.254***
lnRE	−1.321	−4.402***	−1.654	−4.789***
lnEC	−2.281	−3.352***	−1.852	−2.145***
Receipts	−1.745	−2.852***	−2.321	−3.457***

*** denotes a significance of 1%

The third step is to use the Padroni cointegration tests to run the panel cointegration tests (Pedroni, 2004). Before performing the panel cointegration test, the mean of the series across panels is calculated and subtracted from the series. This procedure reduces cross-sectional dependence effect (Levin et al., 2002). Similarly, the panel cointegration test is given in Table 4, which uses any statistics to test the null hypothesis of no cointegration. Hence, the following subsection assesses the long-run relationship.

Empirical results

Table 5 provides the model comparison and overall results. Tourist arrivals have a negative impact on CO2 emissions; it means 1% increase in arrival of tourist, and 0.78% increase in carbon emissions, according to the estimation results. These findings demonstrate that tourism receipts, as a measure of wealth, help to reduce CO2 emissions. Our research results on tourism benefits corroborate those of Abou-Shouk et al. (2021), Beladi et al. (2009), Khan et al. (2020b), Li et al. (2018a, b), Sun et al. (2020a, b), and Tourism Tasmania (2018). Surprisingly, whereas overall economic growth increases carbon emissions, tourism receipts have a positive environmental impact by lowering CO2 emissions. This result could be explained by the fact that tourism, as a major subsector of the service sector, uses less energy and is cleaner than agriculture and industry (Ekanayake and Long, 2012; K.C., 2017). Agricultural and industrial sectors, for example, currently contribute 21% and 24% of global CO2 emissions, respectively. The tourism sector contributes about 4.6%, which is significantly less than the other sectors. Global tourism accounts for approximately 8% of global greenhouse gas emissions, according to Maryam Khokhar et al. (2020) carbon footprint calculations.

Table 5

Model comparison and overall results

Variables	GNS	SEM	OLS
lnTourism	2.214***(0.457)	2.205***(0.297)	2.204***(0.533)
Urb	−0.546***(0.163)	−0.547***(0.143)	−0.412***(0.135)
lnEI	0.347***(0.0805)	0.547*** (0.080)	0.214***(0.171)
lnPGDP	2.254***(0.214)	2.243***(0.252)	2.145***(0.171)
lnRE	−1.145***(0.163)	−1.214***(0.163)	−0.914***(0.268)
lnEC	−0.058***(0.014)	−0.069***(0.020)	−0.077***(0.019)
Receipts	0.068***(0.0254)	0.065***(0.205)	0.021***(0.0305)
Cons	1.173***(0.0165)	1.214***(0.0201)	2.342***(0.029)
W* lnTourism	−1.304***(0.324)	−1.742***(0.547)
W*(Urb)2	0.234***(0.067)	0.354***(0.054)
W*lnEI	0.064***(0.145)	1.032***(0.085)
W*lnPGDP	−1.241***(0.452)	−2.054***(1.254)
W*(lnRE)2	0.254***(0.035)	0.095***(0.030)
W*lnEC	0.289***(0.457)	1.754***(0.457)
W*Receipts	1.254***(0.519)	0.201***(0.842)
LM-SEM	35.265
Robust LM-SEM	1.4232
LM-GNS	29.7563
Robust LM-GNS	0.2130
Obs.	1260	1260	1260
R2	0.657	0.7166	0.5326

Eliminating the error of spatial dependence in applied environmental research will produce biassed estimates, because the classical ordinary least square (OLS) model ignores spatial dependence and destroys the scientific basis of research. Both the spatial error model (SEM) and the spatial lag model have widely been used in empirical studies. Moreover, this study investigates the tourism development impact on CO2 emissions by applying the generalized nested spatial (GNS) model. Further, the presence of spatial autocorrelation is tested by applying Lagrange multiplier and robust LM statistics. Next, we evaluate which panel model is most appropriate after the LM and robust LM test rejects the null hypothesis of no spatial autocorrelation. After selecting the most applicable model, we assessed the direct, indirect, and overall impact of tourism development on CO2 emissions. The outcomes of both non-spatial panel ordinary least square (OLS) and spatial panel SEM and GNS models are illustrated in Table 6. Consequently, data profiles with 1260 observations are represented by the developed models, where the application of the spatial econometric model is stressed by LM test rejecting the null hypothesis with p value 1%. Moreover, robust LM-SEM and robust LM-GNS are compared to select tests having the smallest p value. The p value of GNS is smaller, making it the more appropriate option. Similarly, the selected variables are considered to explain the OLS, SAR, and SEM models at 75%, 81%, and 97% of the variation in carbon dioxide emissions, respectively. Hence, the generalized nested spatial (GNS) model is preferred over the OLS or SEM models as suggested by the greater adjusted R2 of the models.

Table 6

The direct, indirect, and total effects of GNS model

	Direct impact		Indirect impact		Total impact
Variable	Coefficient	t values	Coefficient	t values	Coefficient	t values
lnTourism	1.3215***	3.401	−1.7451***	−0.2341	0.0095***	0.089
Urb	−0.356***	−3.478	0.85214***	0.2587	0.14567***	0.355
lnEI	0.0226***	3.456	−0.07412***	−3.4568	−0.04578***	−0.0411
lnPGDP	3.2354***	3.258	5.5471***	1.166	11.2451***	2.255
lnRE	−0.2452***	−3.741	−0.7541***	−1.74156	−0.1425***	−2.3534
lnEC	−0.01245***	−2.4512	−0.004512***	−0.45871	−0.02415***	−0.793
Receipts	0.1450***	2.921	0.4567***	3.2587	1.4521***	5.2478

∗∗∗ p < 0.01; ∗∗ p < 0.05; ∗ p < 0.1

Analysis of GNS model

At this point, various studies use point estimates to examine the existence of spatial spillover effect. However, the marginal impact of the corresponding explanatory variables on the dependent variable is not directly represented by the coefficients of the GNS model, whereas the regression results do not reveal how carbon dioxide emissions are affected marginally by tourism development, renewable energy, and financial development. This study states the importance of explaining the impact of variable changes in spatial models through the partial differential methods. Table 6 provides the direct, indirect, and total effects of the independent variables. Therefore, the impact of changes in explanatory variables on CO2 emissions in a particular country is called a direct impact, whereas the impact of variations in independent variables of neighbouring countries is referred to as indirect effects. Similarly, the sum of indirect and direct effects is called the total effect.

Table 6 represents the results of direct, indirect, and total effect of the GNS model. The findings show that the tourism coefficient (1.25) in the direct effect is considered positive and statistically significant at the 1% significance level. The tourism coefficient (−1.74) in the indirect effect is negatively related to environmental pollution, with a significant level of 1%. Therefore, the total impact of tourism (−0.4899) is recorded as negatively significant at the 1% level. In addition, according to the positive and significant direct effect, every additional unit in the tourism development will increase the carbon dioxide emissions by 1.25%. The positive direct and negative indirect impact of tourism show an inverse U-shaped relationship between tourism development and local country CO2 emissions. This means that CO2 emissions first increase at an increasing rate with the development of the tourism to reach the maximum, and then as the tourism continues to increase, CO2 emissions decrease at a decreasing rate. In contrast, the development of tourism in neighbouring countries has a U-shaped relationship with CO2 emissions, that is, CO2 first decreases with the decrease of tourism, reaching a minimum, and then with the continuous development of tourism, CO2 increases.

However, the study sample is divided in two types, one with tourism development on the left side of the axis of symmetry of the inverse U-shape and the other with tourism development on the right side. Our findings are consistent with Sun (2016a), (Feng et al. 2021; Li et al. 2021a, 2021c) that concludes tourism has a positive impact on carbon dioxide emissions. Similarly, our results also show the negative impact of tourism on CO2 emissions, which is consistent with the second category of literature (Katircioǧlu, 2014; Lee and Brahmasrene, 2013; Paramati et al., 2017). Hence, overlooking the nonlinear effect of tourism on carbon emissions is possibly the reason for the different conclusions. Although most studies have confirmed the empirical association among tourism development and carbon emissions, according to the above conclusions, the direction of the causal relationship between the two is still unknown. One of the main reasons for drawing conflicting conclusions is to ignore the nonlinear impact of tourism on carbon emissions (Li et al., 2018a; Sun, 2016b). In addition, if the spatial interdependence of the regions is not considered, it may lead to erroneous conclusions (Yang et al., 2019). Therefore, this study uses a panel spatial econometric model to estimate the total impact of tourism on carbon emissions, taking into account spatial dependence and nonlinearity.

Next, other control variables also produce valuable results, and the detailed explanation is given below:

For financial development and CO2 emissions nexus: The primary focus of this study is on financial development, and therefore results provide its impact. The direct effect for financial development is recorded as significantly positive (0.055), the spillover impact is −0.09215, and the total impact is −0.063, as seen in the empirical findings of financial development given in Table 6. On the contrary, an increase of 5.5% in carbon dioxide emissions with a 1% rise in financial development indicates a significantly positive direct effect. The findings also reflect the financial development and carbon dioxide emissions to constitute an inverse U-shaped relationship, which indicates an increase in carbon emissions of a country with growth in its tourism. However, an eventual 12.5% decrease is expected in carbon emissions of the local country after the threshold is reached. When financial development is higher in the neighbouring countries, the quality of the environment in a local country is affected by technological diffusion, better governance, and more sustainable policies, presenting a possible reason for this process. Consequently, lower carbon dioxide emissions are induced due to the external restrictions across nations (Li et al. 2021e; Lv and Li, 2021a; Zhao et al. 2021). Similarly, the carbon dioxide emissions in the local country are reduced by boosting the spillover of technology, and the transfer of knowledge and skills. The influence of the financial development of neighbouring countries on a country’s carbon dioxide emissions is stressed by this study and aligns with (Li et al. 2021b; Lv and Li, 2021a; Miao et al. 2019). Hence, the significantly positive direct effect is taken over by the negative spillover impact of financial growth on CO2 emission presenting a total effect which is significantly negative. The role of financial development in promoting business growth more than promoting technological progress and green projects increases energy consumption and is a possible explanation for this process, and the findings align with Bui (2020) and Charfeddine and Kahia (2019). Conversely, a unit rise in financial development of neighbouring countries suggests 12.5% decline in carbon emissions of the local country through the negatively significant spillover effect.

For GDP per capita and CO2 emissions nexus: GDP per capita significantly impacts on CO2 emission through spatial spillover. The direct impact of GDP per capita on CO2 emission is significant, while the direct impact of (PGDP)2 is negative, proving the existence of environmental Kuznet. Similarly, economic growth and carbon dioxide in sample countries share an inversed-U relationship. In both models, the indirect effect of PGDP and (PGDP)2 on carbon dioxide proves the existence of environmental Kuznets. Nevertheless, pollution in the neighbouring countries is increased due to the local economic development and improved quality of life, transferring high-pollution industries to neighbouring countries. Consequently, an increase in environmental pollution through the increased economic development indicates that the income in these countries positively impacts the left side of the inversed-U curve.

For renewable energy and CO2 emissions nexus: The direct effect coefficient (−0.0598) and indirect effect coefficient (−0.0921) are recorded to be negatively significant at 1% significance level, as indicated by the findings for renewable energy. Hence, a country observes 5.98% decrease in carbon dioxide emissions and 9.21% decrease in carbon dioxide emission with 1% increase in renewable energy for a certain country and in its neighbouring countries, as seen in the results. Similarly, in the population density and CO2 emissions nexus, we observed a significantly positive direct effect of population density on CO2 emissions and significantly positive indirect effects of population density on CO2 emissions. Hence, emission of pollution is increased with population density for a certain country along with its neighbouring countries. Also, the results showed a statistically significant positive effect of trade openness on CO2 emissions; total effect and the spillover are also evident. Therefore, the environment of the local country is affected by trade openness in all neighbouring countries.

Moderating role of education and infrastructure

In this section, we find the moderating effect of educating expenditure and transpiration infrastructure on CO2 emissions. The results showed a negative statistically significant direct, indirect, and total effect of education on environmental pollution (Table 7). The direct effect of interaction term is higher than the indirect effect; the 1% increase in education expenditure of the local country helps 5.4% decrease in environmental pollution. At 1% significance level, the spatial coefficient is recorded −0.0399, which suggests strengthening of the negative indirect effect of tourism on carbon emissions along with the strengthening of negative direct effect of tourism on carbon emissions due to sustainable education. The significantly negative effect of the interaction between tourism and education on carbon emissions is shown in Table 7, which allows the negative effect of tourism on CO2 emission to become stronger due to an increase in the sustainable education tends. The carbon emissions induced from tourism are impacted by the environmental protection awareness of tourists, according to many scholars (Zhang and Zhang, 2018), proving empirical evidence for the already presented argument (Rehman et al., 2020; Ahmad et al., 2020; Fatima et al., 2019; Li et al., 2021).

Table 7

Results of moderation role of education infrastructure

Variable	Direct impact		Indirect impact		Total impact
Variable	Coefficient	t values	Coefficient	t values	Coefficient	t values
Tour*Edu	−0.0540***	−0.2382	−0.03998***	−0.453	−0.2680***	−0.4241
Tour*Str	−0.0087**	−0.0139	−0.0067*	−0.038	−0.2680**	−0.4241
Tour	1.4034***	2.0938	−1.1845***	−0.288	0.3826***	0.0902
FD	0.0525***	4.5305	−0.0875***	−3.853	−0.0568***	−2.0938
PGDP	3.7507***	2.8325	4.8725***	1.407	10.625***	2.0577
(PGDP)2	−0.3032***	−3.4656	−0.5295***	−1.775	−0.8330***	−2.6535
RE	−0.0568***	−0.0875	−0.0017***	−0.314	−0.0138***	−0.3353
PopD	0.2193***	2.3284	0.6561***	3.980	0.8754***	6.3825
Trade	0.00992**	0.8125	−0.0162**	−0.4783	−0.0063**	−0.1475

Panel data analysis demonstrates that international tourism transportation expenses have a progressively greater impact on CO2 emissions. Rayamajhi (2013) used a panel data analytic approach to test the links among the energy consumption, GDP, commerce, tourism, and CO2 emissions in OECD nations from 1995 to 2016. The analysis’ findings indicate that tourism development has a growing impact on carbon emissions. Sun et al. (2020) used panel cointegration, FMOLS, and panel causality approaches to examine the effect of tourism revenues on CO2 emissions in Eastern and Western European nations from 1995 to 2013. Their findings indicate that development of tourism has a positive influence in Eastern Europe CO2 emissions but has a negative effect in the Western part of Europe. Arai and Goto (2017) used a panel bootstrap causality test to analyse the link among CO2 emissions and tourist arrivals from 1995 to 2014 in 16 small developing countries. Their findings demonstrate that the association among the tourist arrivals and CO2 emissions is bidirectional. The analysis’ findings confirm that tourism expansion has a carbon-reducing effect. Wondirad et al. (2021) investigated the association between CO2 emissions and economic growth related to tourism in industrialized and developing nations from 2005 to 2013. According to their findings, tourism adds to the increase in CO2 emissions (Ahmad et al., 2021a; Ahmad et al., 2021b; Ahmad and Jabeen, 2020).

Robustness test

We applied different spatial weight matrices to check the robustness of results and the specifications of the spatial weight matrix (Lv and Li, 2021b). For one of the binary matrix of the eight nearest neighbours, if the country j is within the eight nearest neighbours of the country I, the weight wij = 1; otherwise, the weight wij = 0. Furthermore, Table 8 shows the results for direct, indirect, and total effects, giving a similar broad range for different spatial weight matrices of direct, indirect, and total effects. Hence, the spillover and total effects are relatively insignificant due to the two spatial weight matrices showing most of the elements as zero.

Table 8

Robustness check

	Direct effects		Indirect effects		Total effects
Variable	Coefficient	t values	Coefficient	t values	Coefficient	t values
lnTourism	−0.7420***	−2.1425	−0.20015***	−1.251	−0.3021***	−0.3145
Urb	0.03512***	3.251	−0.07521***	−3.2145	−0.04521***	−2.301
lnEI	3.7452***	2.23615	5.0012***	1.166	10.2541***	2.245
lnPGDP	−0.1452***	−3.532	−0.25874***	−1.9852	−0.7932***	−2.654
lnRE	−0.0452***	−0.07412	−0.002130***	−0.2541	−0.02145***	−0.2514
lnEC	0.214***	2.251	1.254115***	3.6215	0.8966***	5.122
Receipts	0.0745***	0.7452	−0.01542***	−0.4021	−0.00442***	−0.4522

Table 8 presents the results of the cross-sectional dependence test. The findings showed that all the selected variables are significant at 1% significant level. These findings suggest that a shock one of the most visited countries may have an impact on other nations’ influencing factors. As a result, in order to obtain reliable results, we applied panel data methodologies of second generation; the interdependence of countries must be taken into account.

Conclusion and policy implication

Tourism has been stressed globally due to its significant contribution to job creation, economic growth, and regional coordination (The World Travel & Tourism Council, 2014). Thus, it was repurposed to revitalize the countryside’s crumbling infrastructure and close the imbalanced development gap between urban and rural areas (Gao & Wu, 2017). Throughout China’s reform and opening-up, much emphasis was placed on urbanization and industry. Meanwhile, substantial economic factors and developing resources gravitated toward urban regions, and the countryside gradually fell further and more behind the city (Liu & Wall, 2006). However, these irrational policies and regulations have acted as a severe impediment to a reasonably affluent, sustainable, and just society (Su and Ang, 2010). The findings support the EKC hypothesis.

The following are three points that summarize the estimation results: (a) the tourist development has a positive significant impact on the carbon emissions. (b) Tourism receipts, on the other hand, reduce CO2 emissions. (c) Tourism development has a bidirectional impact, implying that CO2 emissions and tourism development have a bidirectional causal relationship.

Theoretical foundations of the research results explain why tourist arrivals increase carbon emissions and carbon emissions decreased through tourism receipts. It can be based on two reasons: (a) International tourism transportations are one of the major factors that impact on the natural environment around the world. Tourist arrivals and departures are increased as a result of progress in the sector of tourism, as are transportation services. According to Brida et al. (2020), transportation services account for nearly 95% of tourism-related CO2 emissions, with the aviation sector accounting for the majority of these emissions. As the number of tourists grows, so does the diversity of infrastructure services available, including lodging, restaurants, hotels, ports, airports, telecommunications railways, and roads. The construction of infrastructure and the development of tourist destinations both have important contribution to increase the carbon emissions.