Sustainability of Tourism and Economic Development in Three Religious Tourism Destinations: The Critical Role of Fossil Fuel Energy on Air Pollution and Human Health

Abstract

The study examined the relations and Granger causality among environmental pollution, air quality, life expectancy, religious tourism, petroleum consumption and economic growth in three countries, Italy, Saudi Arabia and Türkiye, three countries with a prominent role of religious tourism, given the high shares of religious tourism revenues in their economies and due to pilgrimage-type religious tourism activities in total tourism activities. The study employed a yearly sample of 1975–2019 and novel Fourier-augmented vector autoregressive and Fourier Granger causality tests, under the structural breaks in the data. The findings indicate negative effects on environmental pollution and air quality from tourism in addition to such effects on life expectancy in all countries analyzed, and in this relation, fossil fuel consumption in these nations and its acceleration with tourism play crucial roles. These effects are amplified by economic growth coupled with tourism revenues that go in hand with high fossil fuel consumption, which further worsen the impacts on the environment. In the causality testing stage, the results determined unidirectional causality from tourism, fossil fuel energy consumption, and economic growth to both carbon dioxide emissions and to particulate matter 2.5. These effects are also reinforced by feedback effects between air pollution and life expectancy, which enhance the effects on both environment and air quality. These findings are used to suggest important policy recommendations, among which, the reduction in high dependency on fossil fuel in the energy mix is most central. Equally, policies are suggested to encourage sustainable tourism to reverse the adverse effects on health, environmental degradation and worsened air quality in these nations.

1. Introduction

The concept underlying sustainable development is to maintain equitable living standards along with environmental sustainability. A sustainable agenda has been set by the United Nations (UN), with a 2030 completion target of slow-down and reversal of climate change and other growing environmental issues, such as the increasing prevalence of unpredictable weather patterns. Hence, the Sustainable Development Goals (SDGs) must be achieved with fullhearted commitment. The Intergovernmental Panel on Climate Change (IPCC) highlights the alarming increase in average temperature around the globe as the main source: a 1.1 °C increase during 1850–1900 and a similar one in 2011–2020, and their report underlines global warming and its main source as greenhouse gases [1]. As shown by the IPCC, greenhouse gases are at alarming levels as of 2019, as the atmospheric concentrations of carbon dioxide (CO₂) reached 410 parts per million (ppm), and methane and nitrous oxide to 1866 and 332 ppm. These levels are the highest in 800,000 years for the latter two greenhouse gases; however, for CO₂, it is the highest in two million years. Reduction in the concentrations of these gases in the atmosphere is vital; this is especially so for CO₂ [1].

Moreover, particulate matter 2.5 (PM_2.5) also emerges as a type of air pollution that poses important health risks [2]. PM_2.5 penetrates the lungs easily and causes cardiovascular and respiratory diseases. PM_2.5 either results directly from human activities such as transportation, construction and fires, and indirectly from chemical reactions between gases in the atmosphere as an outcome of combustion of solid and liquid fossil fuels [3,4].

Major industries have made important contributions to the overall environmental degradation and air pollution. Among these, tourism, while beneficial for employment and economic growth, has recently faced criticism for its environmental impact. The UN’s World Tourism Organization (WTO) emphasizes tourism’s significant role due to its very high fossil fuel consumption as an “oil-thirsty” industry, resulting from transportation, accommodation and other tourism-associated activities [5]. In addition to exerting environmental stress by utilizing petroleum, tourism contributes to emitting environmental pollutants such as CO₂ and to global climate change [6]. On top of its CO₂ acceleration effects, tourism is criticized for depleting natural resources [7,8].

On the other hand, religion is a significant attractor for tourism activities. The literature on religion-motivated-tourism is collected under “religious tourism”, which is closely associated with pilgrimage-type activities [9]. The studies that explored the diverse aspects of religious tourism gained attention in the 1980s after seminal papers [10,11], and the tourism sector’s historical connection with religion is well established [12]. Travelers motivated by spiritual incentives tend to visit sacred locations [13], and the concept of religious tourism is a conjunction of religion, sociology, psychology and tourism [14]. Religious tourism is predicted to involve 300 to 330 million individuals embarking on religious pilgrimages yearly [15].

It is crucial to acknowledge that tourism could contribute to environmental degradation through various channels, and religious tourism is a strong motivator for tourism [16]. The relationship among tourism, environment, destination and tourist characteristics is depicted in Figure 1. Foremost are the destination’s characteristics, under the influence of its social, environmental and political development; the stage and scale of development; the structure of the sector; resources; the socio-political factors; institutions and values; environmental factors and attitudes.

The second factor in Figure 1 is the characteristics of tourists, followed by international factors: the international investment power of the tourism sector that shapes tourism’s impact. The volume and type of tourists, their attitudes to the environment, their length of stay, seasonality, levels of expenditure and their income affect the ecological footprint [17]. Both channels determine how religious tourism, as an attractor, would affect the environment. The environmental policies of the governments, the lack of environmentally stringent policies, and the way the resources are utilized in the religious tourism destinations have important effects. Combined with the disregard of tourists for the environment and the irrelevancy of the environment, tourists have the foremost influence on sustainable tourism. The high influx of tourists attracted by religious tourism accelerates the depletion of natural resources, amplifies environmental contamination and damages the environment [14].

As the negative environmental externalities of tourism cannot be avoided, sustainable tourism emerges as an important policy. The literature confirms the questionability of sustainable tourism, given the increasing pollution from building infrastructures, particularly hotels, attractions, restaurants and roads, as necessities for attractiveness. Recent empirical findings show that 76% of tourism-related greenhouse gases come from transportation, and the rapid growth of air travel coupled with high levels of pollution per passenger raises concerns about tourism’s role in climate change. In 2025, tourism transportation’s contribution to global CO₂ is predicted to reach 1.5 tons [17].

The effects of the tourism sector on energy consumption and environmental pollution have been discussed by some studies. The author of Gidwani [18] analyzed the impact of fuel crises on tourism in India. Similarly, Hollander et al. [19] estimated the energy demand associated with tourist expenditures in Australia. In another study, Tabatchnaia-Tamirisa et al. [20] investigated the relationship between energy consumption and tourism destinations by revealing that tourists were responsible for an average of 60% of total energy and fuel use. Bach and Gössling [21] accented the relation between tourism and environmental pollution. Additionally, a growing body of research has examined the nexus between tourism and energy consumption [5,22,23,24]. The interrelationship of tourism development, energy demand and environmental degradation has increasingly become a priority on the policy agenda for sustainable tourism. Several studies have addressed sustainable tourism, particularly focusing on its links to energy consumption [25,26] and CO₂ emissions [27,28,29]. Following these studies, some articles have added new variables. In this context, while previous research has largely concentrated on aspects such as economic growth, trade openness, financial development and energy consumption, relatively little attention has been given to other important factors, including air quality, public health, life expectancy and the role of religious tourism. To address this gap, the present study focuses on the distinctive characteristics of religious tourism in Türkiye, Saudi Arabia and Italy.

This study aimed to examine the unexplored empirical links and causality among environmental pollution, air quality, life expectancy, religious tourism, petroleum consumption and economic growth by focusing on three destinations with religious tourism characteristics, Türkiye, Saudi Arabia, and Italy during 1975–2019. The period was terminated in 2019 due to the impact of COVID-19, excluding COVID-19’s effects. Especially, 2020 was subject to a serious structural break due to COVID-19. A dummy variable could be assigned for this year and the one following; however, the results may differ due to the effects of the dummy’s use on other variables. Further, the use of dummies is not an appropriate method, since Fourier terms aim at controlling such breaks; however, the use of additional Fourier terms with the given sample size to control for COVID-19 led to inefficiencies in the estimators. For these reasons, the study’s sample was terminated in this year. The country selection was based on their attractiveness as religious tourism destinations and due to the substantial share of tourism in their gross domestic product (GDP). Notably, these nations witness substantial net tourist inflows, particularly religious pilgrims. In the context of this study, tourism arrivals are a proxy variable of religious tourism. The reason for this is that due to the special location of these countries, the rate of tourism in these countries increases under the influence of holy places. Since religious tourism data could not be found, a proxy was preferred. It was intended to use only the number of tourists coming to the religiously sacred cities of the countries, but the data starting from 1975 could not be found, and the data were found for a very short period. On the other hand, although it was intended to use the tourist reception capacity data, which is considered as an important variable, the data of the variable going back to 1975 could not be found and this variable was excluded from the analysis due to time constraints.

Under this setting, tourism and its relations with economic, environmental and health-related factors, along with petrol consumption, was explored with novel Fourier methods. The modeling and causality testing were undertaken with Fourier-augmented econometric approaches to explore the significance, the size of the effect and the direction of causality. In this study, after the unit root test, if I(1) was determined for the variables, the existence of cointegration was investigated with the Johansen cointegration test. If cointegration was not determined, the Fourier VAR method was applied using the innovations of the variables, and the results obtained can be evaluated as Granger causality.

The study distinguishes itself from the existing literature by concurrently investigating the interconnections among air pollution, environmental pollution, the tourism sector, life expectancy, petroleum consumption and economic growth, and by determining the causality among the variables analyzed. This study is the first to use this method to analyze the effects of religious tourism for these countries. And this paper is the first paper that examines the relations and Granger causality among environmental pollution, air quality, health, religious tourism, petroleum consumption and economic growth in three countries, Italy, Saudi Arabia and Türkiye, three countries with a prominent role of religious tourism, given the high shares of religious tourism revenues in their economies and due to pilgrimage-type religious tourism activities in their total tourism activities.

The literature review is provided in Section 2. The method is presented in Section 3. Section 4 focuses on data and the empirical results. Section 4.4 provides a discussion, and Section 5 covers the conclusion and suggestions.

2. Literature Review

Gössling highlighted, for developing countries, that petroleum consumption associated with the tourism industry adversely affects natural flora, underscoring the significance of environmental pollution [30]. In 2006, Becken and Patterson showed the carbon footprints within the tourism industry [31]. In 2020, Becken proved the “thirst for oil” of tourism, highlighting its dependence on fossil fuel energies, with significant environmental implications [5].

The interconnections among tourism, environmental pollution and energy consumption were further exposed using data from 25 OECD members for 1995–2005, demonstrating causal effects to environmental pollution from tourism [32]. For the EU, the positive impacts of tourism on GDP and negative effects on environmental pollution were shown [33]. In an examination of Cyprus, the findings indicate unidirectional causal links from energy to CO₂ [34]. Divergent effects of tourism on the environment have been shown for various districts in China [35]. In the context of particulate matter 10, the bidirectional causality between environmental degradation and tourism had been shown for five European economies, where findings pointed to degradation of air quality due to tourism in addition to worsened air quality leading to lowered tourism activity [36]. Sherafatian-Jahromi et al. discovered indications of environmental pollution in Southeast Asia from tourism [37]. For China, Kuo et al. [38] found that while increases in tourism receipts led to only modest rises in CO₂ emissions, the number of tourist arrivals had a much more pronounced impact on emissions. Similarly, Solarin [39] identified a unidirectional long-run causality from tourist arrivals to environmental pollution in Malaysia, within a multivariate framework that included real GDP, energy consumption, financial development and urbanization. These findings suggest that growing tourist arrivals contribute directly to pollution in Malaysia. Katircioǧlu [40] reported that tourist arrivals in Singapore had a statistically significant negative effect on CO₂ emissions in both the short and long term.

A set of papers examined tourism and economic relations in the context of the environmental Kuznets curve (EKC). The results for Spain favored the positive effects of tourism on environmental pollution, with links to electricity consumption and evidence against the EKC [41]. The adverse externalities were highlighted for declining air quality [42]. For different EU members, differentiated results were obtained [43]. Sajjad et al. brought attention to the impact of weather anomalies and declining air quality stemming from tourism [44]. Cadarso et al. identified the tourism industry as the primary driver of CO₂ [45]. The long-term interactions among economic development, political instability, tourism industry and energy were shown for MENA, indicating energy-led and tourism-led growth causing environmental degradation [46]. By employing renewable energy instead of fossil fuel, the associations among tourism, CO₂, economic growth, air pollution and more commitment to renewable energy were shown with panel econometric models for the EU [47].

Zhang and Gao [48] tested the relationship among international tourism, economic growth, energy consumption and environmental pollution in China by using panel data analysis from 1995 to 2011. Their results found limited support for the tourism-induced EKC hypothesis in eastern and western China and no evidence in central China. Azam et al. [49] examined the impact of tourist arrivals on CO₂ emissions and energy consumption in Malaysia, Thailand and Singapore from 1990 to 2014, and their empirical results indicated that tourism significantly increased pollution in Malaysia, whereas an inverse relationship was observed in Thailand and Singapore.

Further contributions to this field include the study of Scott et al. [50], which highlighted the risk of tourism becoming a major contributor to GHGs. Zaman et al. [51] investigated the tourism–emissions nexus in both developed and developing countries between 2005 and 2013 by concluding that tourism generally contributes to increased CO₂ emissions. Rico et al. [52] determined a bidirectional causality between tourism and climate change, and found that tourism itself is a major emitter of greenhouse gases due to its reliance on energy consumption. Nepal et al. [53] analyzed the relationships among tourist arrivals, per capita GDP, emissions, energy consumption in Nepal. Accordingly, in their results, they suggested improving energy efficiency and diversifying energy sources. Empirical studies on religious tourism specifically on the countries evaluated in our study are few in number. Among these, the economic causes of religious tourism with a panel dataset for 2008–2017 were examined by employing panel ARDL, and the results revealed positive effects from GDP, exchange rates and international arrivals [54]. For Saudi Arabia, Tabash et al. inspected the impact of religious tourism on environmental pollution, energy consumption, GDP and financial development [55]. Their findings indicate that the influx of pilgrims speeds up several economic processes related to industrial products, which inevitably degrade the environment [55]. For Italy, Nawaz et al. examined the relations among CO₂, geopolitical risk and religious tourism [56]. Their findings point to a CO₂ reduction effect of religious tourism under increased geopolitical risk as well as CO₂ acceleration effects of transportation and foreign direct investment, and they highlight the research gap on the connection between religious tourism and the environment [56].

3. Data and Methods

3.1. Data

The dataset comprises annual data for 1975–2019 on CO₂ emissions, air quality measured as PM_2.5, religious tourism measured as international tourist arrivals, real GDP and petroleum consumption. The data definitions, units, sources and abbreviations are given in

Table 1. Description and sources of the data.

Variable	Abbrev.	Measurement Unit	Source 1
Environmental pollution	CO	CO2 emissions, in kilotons	EDGAR
Air quality	PM	Annual average exposure to particulate matter (PM2.5), micrograms per m3	EDGAR
Life expectancy	LE	Life expectancy at birth, years	WDI
Tourist arrivals	TR	No. of international tourist arrivals	STATISTA and UNWTO
Economic growth	Y	Real GDP, constant 2015 US dollars	WDI
Petroleum consumption	PC	Petroleum consumption, kilotons	BP

¹ EDGAR is the Emissions Database for Global Atmospheric Research database, STATISTA is the Statistics Portal for Market Data, WDI is the World Development Indicators database of World Bank, BP is the British Petrol database, UNWTO is the United Nations’ World Tourism Organization database.

. PM_2.5 is known to cause severe health issues and is an important measure of air quality and pollution. Life expectancy is a widely used health measure in development economics and is also employed in the health pillar of the UN’s human development index.

Descriptive statistics are depicted in

Table 2. Descriptive statistics.

Türkiye	LE	PM	CO	TR	PC	Y
Mean	1.90	1.322	5.592	7.838	7.871	12.18363
Skewness	−0.380	−0.272	−0.811	0.598	−0.346	−0.34282
Kurtosis	1.862	1.799	2.809	2.076	2.70	1.90817
Jarque–Bera	2.498	2.317	3.554	3.050	3.61	2.216270
Italy	LE	PM	CO	TR	PC	Y
Mean	1.864	1.427	5.398	7.285	7.372	11.53657
Skewness	−0.330	0.891	−0.058	−0.082	−0.976	−0.274449
Kurtosis	1.870	2.699	1.734	1.338	3.58	1.921632
Jarque–Bera	2.285	4.3502	2.153	3.715	3. 718	3.723378
Saudi Arabia	LE	PM	CO	TR	PC	Y
Mean	1.907	1.358	5.52	7.013	7.518	11.5030
Skewness	−0.314	−0.942	0.013	−0.349	−0.272	0.09503
Kurtosis	1.722	2.851	1.398	1.749	2.510	2.4736
Jarque–Bera	2.703	4.7621	3.422	2.740	3.354	2.94809

for Italy, Türkiye and Saudi Arabia. Accordingly, PC is subject to negative skewness but is close to the ideal value of zero. For PC, kurtosis approaches the ideal value of 3 for the majority of variables except for LE, PM and TR for Italy, and LE, CO and TR for Türkiye and for Saudi Arabia. Overall testing of normality was conducted with the Jarque–Berra test, which resulted in the non-rejection of normality at the 5% significance level.

3.2. Fourier-Augmented FVAR Method

The study employs novel Fourier Vector Autoregressive (FVAR) models, which are further generalized to Granger causality testing with the Fourier Granger causality (FGC) tests [57]. For simplicity, assume the following bivariate FVAR model with two vectors for the variables A_t and B_t,

$$\Delta A_t={\varphi }_{1,0}+{\displaystyle \sum _{j=1}^p{\phi }_{1,j}\Delta A_{t-j}}+{\displaystyle \sum _{j=1}^q{\theta }_{1,j}\Delta B_{t-j}}+{\lambda }_{1,1}^{}\sin \hspace{0.17em}\left({\displaystyle \frac{2\pi pt}{T}}\right)+\hspace{0.17em}{\lambda }_{1,2}^{}\hspace{0.17em}\cos \hspace{0.17em}\left({\displaystyle \frac{2\pi pt}{T}}\right)+{\epsilon }_{1,t}$$

(1)

$$\Delta B_t={\varphi }_{2,0}+{\displaystyle \sum _{j=1}^p{\phi }_{2,j}\Delta A_{t-j}}+{\displaystyle \sum _{j=1}^q{\theta }_{2,j}\Delta B_{t-j}}+{\lambda }_{2,1}^{}\sin \hspace{0.17em}\left({\displaystyle \frac{2\pi pt}{T}}\right)+\hspace{0.17em}{\lambda }_{2,2}^{}\hspace{0.17em}\cos \hspace{0.17em}\left({\displaystyle \frac{2\pi pt}{T}}\right)+{\epsilon }_{2,t}$$

(2)

where, ${\epsilon }_{1,t}\sim N\left(0,{\delta }_1^2\right),{\epsilon }_{2,t}\sim N\left(0,{\delta }_2^2\right)$ and the FVAR model given in Equations (1) and (2) assumes a single Fourier dimension k = 1 in each vector, with two Fourier parameters for each, λ_1,1 and λ_1,2, respectively. However, it is also possible to examine k for k = 1, 2, …, k_max by testing the overall fit of the model as carried out by [58] for discrete orders of k [59]. In our estimations, k = 1 led to best fit over a candidate range of k = 1,2 by taking k_max = 2, given the tested sample size; for larger samples, higher orders are possible. Lastly, the significance of the Fourier terms should be tested. In Vector 1, the test is carried out by H₀: λ_1,1 = 0, λ_1,2 = 0; similarly, the null hypothesis in Vector 2 is H₀: λ_2,1 = 0, λ_2,2 = 0 in both vectors, with separate F tests following the testing process given by [57].

3.3. Fourier-Augmented Granger Non-Causality Method

The Fourier method in [58] is generalized to FVAR to achieve Fourier Granger causality (FGC) testing, as carried out by [57]. Once the vectors are augmented with Fourier terms, the FGC test is carried out by testing for no Granger causality (GC) from B_t to A_t in Equation (1), the first vector, by testing the null of $H_0:{\theta }_{1,j}=0$ for j = 1, 2, …, q, with a zero restriction on all parameters of B_t. The alternative is $H_1:{\theta }_{1,j}\ne 0$ or at least one ${\theta }_{1,j}$ is not zero. The acceptance of H₁ leads to GC from B_t to A_t. These hypotheses could also be written as H₀: B_t↛A_t and H₁: B_t → A_t. Consequently, the FGC test in Equation (2) assumes no GC from A_t to B_t under $H_0:{\phi }_{2,j}=0$ for all j = 1, 2, …, p, tested against $H_1:{\phi }_{2,j}\ne 0$ or at least one ${\phi }_{2,j}$ is not zero, i.e., GC from A_t to B_t. It should be noted that the two hypotheses could also be written as H₀: A_t↛B_t and H₁: A_t →B_t [57].

3.4. Model of the Study

The model of the study is formed with the following FVAR for 6 variables,

$$\Delta L{E}_t={\sigma }_1+{\displaystyle \sum _{i=1}^p{\alpha }_i\Delta L{E}_{t-i}+{\displaystyle \sum _{i=1}^p{\beta }_i\Delta C{O}_{t-i}}}+{\displaystyle \sum _{i=1}^p{\chi }_i\Delta P{C}_{t-p}}+{\displaystyle \sum _{i=1}^p{\delta }_i\Delta T{R}_{t-i}}+{\displaystyle \sum _{i=1}^p{\varphi }_i\Delta P{M}_{t-p}}+{\displaystyle \sum _{i=1}^p{\phi }_i\Delta Y_{t-i}}+{\gamma }_{0,1}\sin \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\gamma }_{1,1}\cos \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\zeta }_{1}ec{m}_{t-1}+{\epsilon }_{t,1}$$

(3)

$$\Delta C{O}_t={\sigma }_2+{\displaystyle \sum _{i=1}^p{\alpha }_i\Delta C{O}_{t-i}+{\displaystyle \sum _{i=1}^p{\beta }_i\Delta L{E}_{t-i}}}+{\displaystyle \sum _{i=1}^p{\chi }_i\Delta P{C}_{t-i}}+{\displaystyle \sum _{i=1}^p{\delta }_i\Delta T{R}_{t-i}}+{\displaystyle \sum _{i=1}^p{\varphi }_i\Delta P{M}_{t-i}}+{\displaystyle \sum _{i=1}^p{\phi }_i\Delta Y_{t-i}}+{\gamma }_{0,2}\sin \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\gamma }_{1,2}\cos \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\zeta }_{2}ec{m}_{t-1}+{\epsilon }_{t,2}$$

(4)

$$\Delta T{R}_t={\sigma }_3+{\displaystyle \sum _{i=1}^p{\alpha }_i^{}\Delta T{R}_{t-i}+{\displaystyle \sum _{i=1}^p{\beta }_i\Delta L{E}_{t-i}}}+{\displaystyle \sum _{i=1}^p{\chi }_i\Delta P{C}_{t-i}}+{\displaystyle \sum _{i=1}^p{\delta }_i\Delta C{O}_{t-i}}+{\displaystyle \sum _{i=1}^p{\varphi }_i\Delta P{M}_{t-i}}+{\displaystyle \sum _{i=1}^p{\phi }_i\Delta Y_{t-i}}+{\gamma }_{0,3}\sin \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\gamma }_{1,3}\cos \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\zeta }_{3}ec{m}_{t-1}+{\epsilon }_{t,3}$$

(5)

$$\Delta P{C}_t={\sigma }_4+{\displaystyle \sum _{i=1}^p{\alpha }_i\Delta P{C}_{t-i}+{\displaystyle \sum _{i=1}^p{\beta }_i\Delta C{O}_{t-i}}}+{\displaystyle \sum _{i=1}^p{\chi }_i\Delta L{E}_{t-i}}+{\displaystyle \sum _{i=1}^p{\delta }_i\Delta T{R}_{t-i}}+{\displaystyle \sum _{i=1}^p{\varphi }_i\Delta P{M}_{t-i}+{\displaystyle \sum _{i=1}^p{\phi }_i\Delta Y_{t-i}}}+{\gamma }_{0,4}\sin \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\gamma }_{1,4}\cos \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\zeta }_{4}ec{m}_{t-1}+{\epsilon }_{t,4}\hspace{0.17em}$$

(6)

$$\Delta P{M}_t={\sigma }_5+{\displaystyle \sum _{i=1}^p{\alpha }_i\Delta P{M}_{t-i}+{\displaystyle \sum _{i=1}^p{\beta }_i\Delta C{O}_{t-i}}}+{\displaystyle \sum _{i=1}^p{\chi }_j\Delta L{E}_{t-i}}+{\displaystyle \sum _{i=1}^p{\delta }_i\Delta T{R}_{t-i}}+{\displaystyle \sum _{i=1}^p{\varphi }_i\Delta P{C}_{t-i}}+{\displaystyle \sum _{i=1}^p{\phi }_i\Delta Y_{t-i}}+{\gamma }_{0,5}\sin \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\gamma }_{1,5}\cos \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\zeta }_{5}ec{m}_{t-1}+{\epsilon }_{t,5}$$

(7)

$$\Delta Y_t={\sigma }_6+{\displaystyle \sum _{i=1}^p{\alpha }_i\Delta Y_{t-i}+{\displaystyle \sum _{i=1}^p{\beta }_i\Delta C{O}_{t-i}}}+{\displaystyle \sum _{i=1}^p{\chi }_i\Delta L{E}_{t-i}}+{\displaystyle \sum _{i=1}^p{\delta }_i\Delta T{R}_{t-i}}+{\displaystyle \sum _{i=1}^p{\varphi }_i\Delta P{C}_{t-i}}+{\displaystyle \sum _{i=1}^p{\phi }_i\Delta P{M}_{t-i}}+{\gamma }_{0,6}\sin \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\gamma }_{1,6}\cos \left({\displaystyle \frac{2\pi nt}{T}}\right)+{\zeta }_{6}ec{m}_{t-1}+{\epsilon }_{t,6}$$

(8)

As typical, in the ΔLE_t vector in Equation (3), no GC from PM_t to LE_t is tested with $H_0:{\varphi }_i=0$ against the alternative $H_1:{\varphi }_i\ne 0$ or at least one ${\varphi }_i$ is not zero. Alternatively, these hypotheses stand for H₀: PM_t↛LE_t and H₁: PM_t→LE_t. The vectors of the FVAR model in Equations (3)–(8) allow the inclusion of an error correction mechanism, $ec{m}_{t-1}$, with the parameters ${\zeta }_i$ for i = 1, 2, …, 6 under the existence of error correction confirmed by Johansen’s cointegration test. Otherwise, the model could be reduced by ${\zeta }_i=0$ for i = 1, 2, …, 6 in (3)–(8).

4. Results

Empirical analysis proceeds in three stages.

• Fourier augmented Dickey–Fuller (FADF) tests to assess the integration and stationarity of variables.

The unit root test results for the variables are determined as I(1), and the existence of cointegration will be investigated with the Johansen cointegration test. If cointegration among the selected variables is not determined, the Fourier VAR method will be applied by using the innovations of the variables and under these conditions, the results obtained can be evaluated as Granger causality.

• The model incorporates Fourier terms to control for structural breaks and is particularly advantageous for datasets with short time periods.
• The FVAR model provides insights into the statistical significance and magnitude of the effects. The method handles endogeneity and feedback.
• The FGC method is employed for causal relationships. This approach not only determines the presence or absence of causality, but also its direction. The method benefits from Fourier terms and does not require precise knowledge of the exact form, frequency, or date of the structural breaks.

4.1. Fourier Unit Root and Johansen Test Results

Fourier unit root tests were conducted, and the results are presented in

Table 3. Fourier unit root test results.

Variable	Italy	Saudi Arabia	Türkiye
PC	−1.43	−1.25	−1.249
ΔPC	−4.29	−3.482	−5.687
LE	−2.47	−0.89	−0.781
ΔLE	−7.8	−3.16	−3.987
CO	1.152	−0.75	−0.891
ΔCO	−3.9	−3.16	−5.983
PM	−0.42	−1.46	−1.992
ΔPM	−7.64	−10.69	−5.478
TR	0.74	−1.452	−0.274
ΔTR	−4.73	−5.015	−3.786
Y	0.04	−0.926	−1.271
ΔY	−3.02	−4.973	−3.661

, following the Fourier-augmented Dickey–Fuller test of [58]. The H₀ of a unit root cannot be accepted for a level series, whereas all series become stationary after first differencing (denoted Δ), indicating that all series are I(1) processes.

Since all variables are integrated at the same order, the cointegrated relations were tested with the Johansen cointegration test [60] and the results are given in

Table 4. Johansen cointegration test, with Fourier terms included.

Hypothesized No. of Cointegration Vectors Under H0	Critical Value α = 0.05	Italy 1	Saudi Arabia	Türkiye
r = 0	95.753	87.172	92.376	93.979
r ≤ 1	69.81	39.907	42.477	44.119
r ≤ 2	47.856	30.251	31.412	21.147
r ≤ 3	29.797	18.585	19.893	13.952
r ≤ 4	15.494	14.098	12.498	9.4108
r ≤ 5	3.8414	3.1218	3.6250	1.4084

¹ Trace statistics are reported to save space. Maximum eigenvalue statistics led to the same conclusion and are available upon request.

, in which trace statistics are reported. The Johansen cointegration test indicates no cointegration for all variables in all the countries examined. As a result, the analysis continued with FVAR model estimations and FGC testing, as shown in the next section.

4.2. FVAR Model Estimation Results

The FVAR estimation results are presented in

Table 5. FVAR estimation results: Italy.

	COt	LEt	PMt	TRt	PCt	Yt
COt−1	0.588359 ***	−0.02858 **	0.014738	−0.123761	0.937275 **	0.719253
	(2.71541) 1	(−2.09283)	(0.05200)	(−0.64569)	(1.76027)	(0.97084)
COt−2	0.227462	−0.004746	0.135309	0.005117	−0.386831	−0.539467
	(1.02692)	(−0.15081)	(0.46703)	(0.02612)	(−0.71068)	(−1.02661)
LEt−1	0.668721 **	0.350944	0.148440 **	0.727464 *	−0.599830 ***	0.311902 ***
	(2.42232)	(1.55980)	(2.55449)	(1.94718)	(−3.18212)	(4.12543)
LEt−2	−0.460253*	0.678627 **	−0.986059 **	−1.199739 **	1.862938 **	6.420990
	(−1.77063)	(2.52050)	(−2.01172)	(−2.31232)	(2.04433)	(1.43120)
PMt−1	0.005643 **	−0.101574 **	0.398983 *	0.155369	−0.864409 **	−0.027874 ***
	(2.03263)	(−2.06405)	(1.76375)	(0.01556)	(−2.03391)	(−2.59156)
PMt−2	−0.151818	0.001593	−0.007293	−0.425811	0.257822	0.075265 **
	(−0.84616)	(0.06247)	(−0.03108)	(−0.68285)	(0.58475)	(2.33350)
TRt−1	0.236273 **	−0.014328 **	0.380700 **	0.906632 ***	0.514147 **	0.976464 **
	(2.15063)	(−2.49107)	(2.30046)	(4.99116)	(2.01890)	(2.42372)
TRt−2	0.293544 **	−0.016592 **	0.475925 **	−0.286346	−0.665446	0.403993 ***
	(2.30978)	(−2.52101)	(2.25900)	(−1.44432)	(−1.20825)	(2.94217)
PCt−1	0.3252 **	0.003894	−0.012341	−0.090883	0.773537 ***	0.313711 **
	(2.03514)	(0.29610)	(−0.10194)	(−1.10999)	(3.40086)	(2.57402)
PCt−2	0.144597 *	−0.001655 **	0.355039 **	0.105702	−0.023930	−0.892459
	(1.77368)	(−2.13485)	(2.48705)	(1.38306)	(−0.11271)	(−1.57183)
Yt−1	0.057573 **	0.011742 *	−0.262336 *	−0.047623 **	0.013749 **	0.881770 ***
	(2.55095)	(1.79641)	(−1.92659)	(−2.34393)	(2.14697)	(3.40495)
Yt−2	−0.202975	−0.003296 **	−0.256453 **	0.070797 **	0.018259 **	−0.087946 ***
	(−0.16736)	(−2.24944)	(−2.21005)	(2.57050)	(2.21778)	(−3.37893)
F1t	0.00687874 **	0.00193325 **	0.00168419 ***	0.009873 ***	0.000382 ***	0.007901 **
	(2.55044)	(2.57201)	(2.59334)	(2.59205)	(3.29432)	(1.99340)
F2 t	0.008101 **	0.004985 *	0.007116 *	0.032334 *	0.033748 *	0.075265 **
	(2.35568)	(1.85135)	(1.90377)	(1.87143)	(1.65517)	(2.33350)
Diagnostics
A.R2	0.71813	0.75041	0.71503	0.84120	0.7364	0.7265
F	65.50598	74.22558	47.45481	117.7464	28.43398	27.6985
p.	0.0000	0.0000	0.0001	0.0000	0.0002	0.0002
LL	92.57859	151.1173	84.52351	96.25689	65.60525	69.2365
AIC	−5.438573	−9.341155	−4.901567	−5.683793	−3.640350	−3.475
SC	−4.924800	−8.827382	−4.387795	−5.170020	−3.126578	−3.1681

¹ The t statistics are in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1% levels. AIC and SC are the Akaike and Schwarz information criteria; LL is log-likelihood. A.R² is the adjusted R square. F is the F statistic of overall model fit, and p is its p value. F1 and F2 are the two Fourier functions given in Equations (1) and (2), respectively.

,

Table 6. FVAR Estimation Results: Saudi Arabia.

	LEt	PMt	PCt	TRt	COt	Yt
LEt-1	0.041075 **	0.388645 ***	1.911857 ***	1.342475 ***	0.748201 **	0.577323 ***
	(2.24460) 1	(2.90202)	(5.76388)	(7.37709)	(2.02947)	(2.23601)
LEt-2	0.618901	0.436272 ***	1.819140 ***	1.87963 ***	−0.888372 **	−0.687794
	(0.23975)	(2.84444)	(5.64951)	(7.23071)	(−1.98920)	(−0.24700)
PMt-1	−0.031587 **	−0.138098	0.779258	0.877549	0.104593	−0.076562
	(−1.81965)	(−0.67055)	(0.90505)	(0.67620)	(0.72620)	(−0.15483)
PMt-2	0.024568	0.220441	0.426618	0.999983	0.262593*	0.367643
	(1.28625)	(0.97276)	(0.94785)	(0.73589)	(1.75697)	(0.70457)
PCt-1	0.0005151 **	0.002144 **	0.871763 ***	−0.254558	0.108032*	0.291372 **
	(2.53291)	(2.01869)	(3.82737)	(−0.87321)	(1.84706)	(2.14178)
PCt-2	−0.020896 **	−0.065149	−0.271829 **	0.195712	0.052810*	−0.039873
	(−2.33059)	(−0.61245)	(−2.28661)	(0.72376)	(1.70991)	(−0.16693)
TRt-1	−0.001620 **	0.063606 *	0.523827 ***	0.430233 *	0.013429 **	0.094630
	(−2.22110)	(1.73158)	(3.03345)	(1.94663)	(2.22086)	(1.11046)
TRt-2	0.012150 **	0.287392 *	0.183561 **	−0.386888	0.147423 **	0.751894 **
	(2.52154)	(1.92247)	(1.97554)	(−1.60650)	(2.22517)	(2.30352)
COt-1	0.041645	0.873961 **	0.996821	0.691848	0.714217 ***	0.057859
	(1.39551)	(2.46843)	(1.41753)	(0.76870)	(2.88454)	(1.57185)
COt-2	−0.019028 ***	−0.604322 **	−0.657628	−0.802324	−0.279581	0.040395
	(−2.86176)	(−2.30689)	(−1.26393)	(−1.20482)	(−1.52610)	(1.17507)
Yt-1	0.503854 **	0.501982 **	0.401672 **	0.011212 **	0.345567 **	0.632050 **
	(2.36935)	(2.01819)	(1.95887)	(2.13669)	(2.15589)	(2.49770)
Yt-2	−0.010043	−0.093632	−0.218177*	0.109359	−0.220082	0.043134
	(−1.03471)	(−0.92391)	(−1.85189)	(1.43326)	(−0.11660)	(0.18325)
F1	0.17540 ***	0.02152	0.1799 **	0.2147 ***	0.31712 **	0.00473 **
	(2.65221)	(1.43429)	(2.08104)	(3.38514)	(2.01642)	(2.2437)
F2	0.005861 **	0.047935 **	0.00500*	0.012759 **	0.006087 **	0.00372 **
	(2.55694)	(2.43628)	(1.79950)	(2.15424)	(2.12691)	(2.14178)
Diagnostics
A.R2	0.794460	0.876410	0.793019	0.83495	0.7148	0.863028
F	341.0890	13.47344	270.2686	113.2166	664.3842	16.22678
P	0.0000	0.0002	0.0000	0.0000	0.0000	0.0002
LL	163.5328	89.32717	68.74109	61.33801	100.0561	16.22678
AIC	−10.16886	−5.221811	−3.849406	−3.355867	−5.937073	−3.500781
SC	−9.655084	−4.708039	−3.335633	−2.842095	−5.423301	−2.893596

¹ The t statistics are in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1%. AIC and SC are the Akaike and Schwarz information criteria. LL: log-likelihood. A.R²: adjusted R square. F: F statistic testing the model, p is its p value. F1 and F2 are the two Fourier functions.

and

Table 7. FVAR estimation results: Türkiye.

	COt	LEt	PMt	TRt	PCt	Yt
COt−1	0.737544 ***	0.026494	−0.004958	0.441691	−0.456049	−0.259036
	(4.44215) 1	(0.94828)	(−0.05676)	(0.63659)	(−0.53022)	(−0.28672)
COt−2	−0.345673 **	−0.034730 *	−0.087084	−0.904553	−1.027070	−1.475721
	(−1.89595)	(−1.83200)	(−0.90783)	(−1.18721)	(−1.08743)	(−1.43349)
LEt−1	0.925887 ***	0.590444 **	−0.120926 **	4.436249 ***	0.064730 **	0.558219 **
	(2.66082)	(2.50425)	(−2.16404)	(5.75766)	(2.42224)	(2.20466)
LEt−2	0.468322 **	0.364843 **	0.195552 **	1.902335 *	0.926480 **	1.28727
	(2.19786)	(2.50402)	(2.25783)	(2.31578)	(2.19534)	(1.43547)
PMt−1	−0.290496	−0.023681	0.780357	−1.021362	0.262663 **	−0.062483
	(0.39224)	(0.06601)	(0.20637)	(1.33917)	(2.03197)	(−1.24079)
PMt−2	0.085264	−0.069279 **	−0.343464 *	0.264028 **	0.272509 **	−0.040160
	(1.22519)	(−1.9734)	(−1.72413)	(2.16687)	(2.13893)	(−0.77535)
TRt−1	0.183585 ***	−0.008741	0.028564 *	0.584381 **	0.298213 **	0.649593 **
	(3.01162)	(−0.85210)	(1.89062)	(2.29399)	(2.31101)	(2.29927)
TRt−2	0.204643 ***	−0.010341	−0.045205	−0.050103	0.110677 **	0.731614
	(3.14716)	(−0.94510)	(−1.32135)	(−0.18438)	(2.32856)	(0.36349)
PCt−1	0.028614 ***	0.001326	0.009840	−0.025965	0.656657 ***	0.039436
	(2.58487)	(0.16105)	(0.38229)	(−0.12700)	(2.59094)	(0.12084)
PCt−2	0.255022 *	0.017828 **	0.056987 **	0.332153	0.070746	0.126873 **
	(1.7157)	(2.06332)	(2.10945)	(1.4795)	(0.26597)	(2.36314)
Yt−1	0.516720 ***	0.403472 **	0.59756 *	0.200401 **	0.404114 **	0.623329 **
	2.33442)	(2.47882)	(1.78305)	(2.01707)	(2.01918)	(2.39790)
Yt−2	0.019869	0.015890 *	−0.726635	0.091876 ***	0.307020	0.239254
	(0.33533)	(1.84909)	(−0.50024)	(3.29779)	(1.20798)	(0.77661)
F1t	0.00496041 ***	0.0017553 **	0.0511321 ***	0.003627 **	0.00193 **	0.00473 **
	(3.48071)	(2.08075)	(2.75238)	(2.54925)	(2.01329)	(2.2437)
F2 t	0.003275 ***	0.005963 ***	0.00237 ***	0.007736 ***	0.00158 **	0.00036 **
	(3.87459)	(2.63341)	(2.62473)	(2.58428)	(2.44698)	(2.28672)
Diagnostics
A.R2	0.793329	0.85926	0.800241	0.72129	0.857849	0.863028
F	282.9308	133.1024	7.611476	66.27006	43.17657	16.22678
p	0.0000	0.0000	0.0003	0.0002	0.0002	0.0003
LL	91.29428	144.7586	110.5604	48.39249	41.94798	16.22678
AIC	−5.352952	−8.917242	−6.637357	−2.492833	−2.063199	−3.500781
SC	−4.839179	−8.403470	−6.123585	−1.979060	−1.549426	−2.893596

¹ The t statistics are in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1%. AIC and SC are the Akaike and Schwarz information criteria. LL: log-likelihood. A.R²: adjusted R square. F: F statistic testing the model, p is its p value. F1 and F2 are the two Fourier functions.

for the separate countries. The interpretation of parameter estimates provides vital information for the effects among the analyzed variables.

i.

Italy

In Table 5, the impact of tourism and the examined variables on CO variable (CO₂ emissions) is assessed in the first column. Afterwards, Columns 2–6 present the estimation results for the vectors of LE, PM, TR, PC and Y. The overview suggests the positive effects of tourism on CO in the next year and the year after, since the parameters for TR_t−1 and TR_t−2 are 0.236273 and 0.293544, and are significant at the 5% level of significance. Further, one could use the two statistically significant coefficients to calculate the accumulated long-run response of CO_t to tourism as 0.236273 + 0.293544 = 0.529817, where a 1% point increase in TR_t results in a 0.53% point increase in CO_t. Similar calculations yield the finding that both PC and Y are the other factors that have positive effects.

The connection between LE and CO is evaluated in the CO_t vector as showing that an increase in LE positively influences CO the next year and negatively in the second year afterwards, since for LE_t−1 and LE_t−2, the coefficients are 0.668721 and −0.460253. The former is significant at 5%, the latter at the 10% level, and the dominant effect of LE on CO is positive. The accumulated response of CO_t to a 1% increase in LE_t is 0.668721 + (−0.460253) = 0.208468, a 0.21% point increase in CO_t. This effect represents a lead–lag relation and, logically, the causation is expected to be from CO_t to LE_t. To this end, if the LE_t vector is examined, we confirm the negative effects of CO_t on LE_t: a 1% point rise in CO leads to a −0.02858% point decline in LE and, therefore, health in Italy.

Examining the impact of PM on CO, a rise in PM is linked with an upsurge of CO; however, the coefficients are close to zero and nearly negligible. Overall, the results confirm that both TR and PC contribute to higher CO. The elasticities of CO to TR and to PC indicate an inelastic relationship. Similarly, the LE elasticity of CO is also strictly inelastic. PM is an important gauge for air quality. When exploring the connection between LE and PM, we observed a positive coefficient for LE_t−1 and a negative one for LE_t−2, and an increase in TR appears to decrease LE. In the TR_t vector, an increase in LE_t−1 positively affects TR in year t, while the effect next year becomes negative. Considering the sizes, the negative effect of LE dominates. Regarding the effects of PC in the PM vector, we obtained negative and positive coefficients. The negative effect is statistically insignificant, and the results indicate the positive effects of PC on PM, showing that worsened air quality increases health problems. Tourism has positive coefficients for TR_t−1 and TR_t−2; however, tourism deteriorates PM levels.

In the PM_t vector, an upsurge in PC creates positive responses in PM_t; in the PC_t vector, an upsurge in PM is associated with a decrease in PC_t. This outcome may be attributed to strategies such as enhancing energy consumption efficiency and promoting greater utilization of renewable energy. These findings underscore the substantial impact of tourism in religious tourist destinations on environmental pollution and air quality, which also holds for PC and LE in Italy.

ii.

Saudi Arabia

The results are reported in Table 6 for Saudi Arabia.

Accordingly, a reduction in LE decreases PM, and the elasticity is again less than 1, suggesting an inelastic relationship. The increase in PC generates increases in PM, though the coefficients are close to zero and significant for the first lag. Notably, the impacts of TR on PM vary for the first and second lags. For the first, they are statistically insignificant; for the second lag, they are significant and positive, indicating positive effects of TR on PM and suggesting negative effects of tourism on air quality.

For the effects of CO_t on PM_t in the PM_t vector, increases in CO_t−1 amplifies PM_t and the coefficient of the second lag is negative; however, the positive effect dominates. In addition, PC has significant positive effects on both PM_t and CO_t for all lags. Y and TR also have positive impacts on PM_t in Saudi Arabia. In the CO_t vector, increases in LE_t−1 positively influence CO_t; LE_t−2 has a negative influence, the dominant effect is negative, and the result indicates a negative association between LE and CO. The surge in TR has a positive effect on CO_t. Increases in PC_t exert a further positive impact on CO_t. The overall results indicate the positive effects of TR on CO in addition to such effects on PM, with significant implications for health, which also result from the increased level of PC after the upsurge in tourist arrivals due to religious tourism.

iii.

Türkiye

The FVAR estimation results are in Table 7.

In the CO_t vector, we obtained similar findings in terms of the impacts of TR, PC and Y, as well as their relations with PM and LE: both coefficients of TR_t−1 and TR_t−2 are significantly positive, and a 1% point increase in TR leads to an accumulated effect of 0.387% point increase in CO_t. Lower LE is associated with higher CO. In the CO_t vector, increases in LE_t increase CO_t; in the LE_t vector, increases in CO_t reduces LE_t. Compared with others, there is no statistically significant association between PM_t and CO_t in the CO_t and PM_t vectors for Türkiye. However, an increase in TR contributes to heightened CO, in alignment with the other countries.

Similar to the previous countries, increases in TR and PC affect CO positively, consistent with Saudi Arabia and Italy. Elevated levels of CO and PM decrease LE in Türkiye. Specifically, an upsurge in PM reduces LE, and a parallel outcome is observed for CO. As before, the coefficients for both associations are close to zero. However, the effects of tourism are highly positive on PC, PM, Y and CO, but insignificant on LE. Though this is the case, the positive effect of LE cannot be rejected for PM, and the worsened air quality should be expected to worsen health in the long run.

4.3. Fourier-Augmented Granger Causality Test Results

Table 8. Fourier Granger causality test results.

	LE→CO	CO→PC	PM→CO	TR→CO	PC→LE	PM→LE	TR→LE	PM→PC	TR→PC	TR→PM	Y→CO	LE→Y	PM→Y	TR→Y	PC→Y
	CO→LE	PC→CO	CO→PM	CO→TR	LE→PC	LE→PM	LE→TR	PC→PM	PC→TR	PM→TR	CO→Y	Y→LE	Y→PM	Y→TR	Y→PC
	Italy
Lag 1	+ 1	+	+	+	-	-	-	Ø	+	+	+	+	-	+	Ø
Lag 2	-	+	Ø	+	Ø	Ø	-	+	+	+	Ø	Ø	+	+	+
Lag 1	-	+	Ø	Ø	-	+	+	-	Ø	Ø	Ø	+	-	-	+
Lag 2	Ø	Ø	Ø	Ø	+	-	-	Ø	Ø	Ø	Ø	-	-	+	+
Result 2:	↔	↔	PM→CO	TR→CO	↔	↔	↔	↔	TR→PC	TR→PM	Y→CO	↔	↔	↔	↔
	Saudi Arabia
Lag 1	Ø	Ø	Ø	+	+	-	-	Ø	+	+	+	+	Ø	Ø	+
Lag 2	-	Ø	+	+	-	Ø	+	Ø	+	+	Ø	Ø	Ø	+	Ø
Lag 1	+	+	+	Ø	+	+	+	+	Ø	Ø	Ø	+	+	+	+
Lag 2	-	+	-	Ø	+	+	+	Ø	Ø	Ø	Ø	Ø	Ø	Ø	Ø
Result:	↔	PC→CO	↔	TR→CO	↔	↔	↔	PC→PM	TR→PC	TR→PM	Y→CO	↔	Y→PM	↔	↔
	Türkiye
Lag 1	+	Ø	Ø	+	Ø	-	Ø	-	+	Ø	+	+	Ø	+	Ø
Lag 2	+	Ø	Ø	+	+	Ø	Ø	Ø	+	Ø	Ø	Ø	Ø	Ø	+
Lag 1	Ø	+	Ø	Ø	+	+	Ø	+	Ø	Ø	Ø	+	+	+	+
Lag 2	-	+	Ø	Ø	+	+	Ø	+	Ø	+	Ø	+	Ø	+	Ø
Result:	↔	PC→CO	Ø	TR→CO	↔	↔	Ø	↔	TR→PC	TR→PM	Y→CO	↔	Y→PM	↔	↔

¹ + and – denote the sign of the effect for the determined Granger-causality. ² → and ↔ indicate uni- and bi-directional Granger-causality. Ø: No Granger-causality.

presents the outcomes concerning causal relationships and their statistical significance across the examined countries for Italy, Saudi Arabia and Türkiye.

Table 8 also summarizes the sign of the significant coefficients in each FVAR vector to obtain additional information regarding the sign of the effect from one variable to another. In this respect, the sign of the significant coefficients is denoted with + and − signs. Further, Ø indicates statistical insignificance. Lastly, the test results in terms of the direction of causality are stated in the final part. Typically, a two-way arrow (↔) indicates bidirectional causality and a one-way arrow (→) indicates unidirectional causality.

The noteworthy findings are briefly outlined below.

• The results indicate unilateral causal links for all countries from TR to PM and to CO. These results are consistent with previous research on CO which identified unidirectional causality to CO from TR, as demonstrated for the EU [43], for specific areas of China [35] and for Cyprus [34].
• In the case of Saudi Arabia and Türkiye, the results suggest unidirectional causal links to CO from PC. Conversely, for Italy, the causality is bidirectional.
• Among the three countries examined, a mutual causal relationship was identified between LE and CO. Further bidirectional causalities were obtained between LE and PM in these nations with religious tourism destinations.
• There is bidirectional causality between PM and PC for Italy and Türkiye. Regarding the signs of the coefficients, increases in PM have causal and positive effects on PC. However, an increase in PM reduces PC, which is an important point, because an increase in PM leads countries to commit more to greener energy alternatives.
• Unidirectional causality from TR to PM was observed for all countries analyzed, confirming the effects of tourism on air quality and health.

4.4. Discussion

The empirical findings of this study provide compelling evidence of the relationships among religious tourism, energy consumption, life expectancy, air quality, environmental degradation and economic growth in Italy, Saudi Arabia and Türkiye. Utilizing a Fourier-augmented VAR framework complemented by causality tests, the analysis confirms that both tourism activity and petroleum consumption Granger-cause increased CO₂ emissions and PM_2.5 concentrations in all three economies. Additionally, the existence of bidirectional causality between life expectancy and both CO₂ emissions and PM_2.5 underscores the cyclical and mutually reinforcing links between environmental degradation and air quality outcomes.

In the empirical results, both tourism and petroleum consumption are identified as Granger causes of CO₂ and particulate matter 2.5; additionally, bidirectional causality between life expectancy and CO₂ were found across all examined countries, along with a reciprocal relationship between life expectancy and particulate matter 2.5. These findings highlight the significant effects of tourism and petroleum consumption on health and environmental degradation. These interconnected relations emphasize the need for a comprehensive approach to address the sustainability problems of tourism. It is clear that immediate action is required to mitigate the health effects from pollutants such as particulate matter 2.5 and CO₂ to achieve sustainable development. Global CO₂ emissions rose by 0.8% annually and reached 32.4 gigatons in 2015, a level approximately 40% higher than that of the 1800s [58]. Among all industries, tourism is responsible for 8% of global CO₂ [59]; 75% of the CO₂ of tourism results from transportation [60]. Tourism also contributes to global CO₂ through air and sea travel [61]. As a result, among the many factors and industries identified in the literature, the study contributes by empirically showing the effect of tourism in the context of religious tourism destinations and petroleum consumption not only on CO₂ emissions, but also on air quality and human health. Following these findings, policies are advocated to recognize tourism as a significant contributor to environmental degradation and the industry’s role in climate change. These findings also suggest that augmenting air quality should be coupled with enhancing energy efficiency, transitioning to renewable energy and adopting sustainable practices in tourism. Our findings support previous research that highlighted the negative impacts of excessive fuel consumption.

Not only high levels of CO₂ but also particulate matter 2.5 significantly degrade air quality and pose widespread health risks. Specifically, particulate matter 2.5 is shown to penetrate the lungs and bloodstream, contributing to severe health conditions [61]. In 2016, it alone was responsible for seven million premature deaths [62]. Short-term exposure to this type of air pollution leads to respiratory and cardiovascular diseases, while prolonged exposure is linked to chronic conditions, heart disease and lung cancer. In 2016, it was ranked as the sixth leading cause of mortality [63].

Policies should develop regulations on emissions to achieve sustainable tourism and initiatives to reduce petroleum consumption from tourism at the same time to combat factors contributing to particulate matter 2.5. As a result, sustainable tourism emerges as a crucial factor in achieving sustainable development, while the reliance on petroleum worsens sustainable tourism.

In pursuit of the SDGs, states have noted the critical role of the tourism industry [64]. Among the 17 designated SDGs of the UN, achievement of sustainable tourism is linked mainly to SDG12, establishing sustainable consumption and production patterns, and to SDG7, green energy. The UN’s WTO and World Travel and Tourism Council—the two prominent tourism institutions—swiftly embraced sustainable tourism and advocated for environmentally friendly practices [65]. The observers of the WTO closely monitor tourism for its environmental effects [66]. WTO also offers assistance to boost climate action in tourism by reducing its ecological pressures in many areas such as microplastics in the seas and excessive energy consumption [67]. Further, the necessity of collective action from the industry is highlighted to tackle the sustainability challenges of tourism. With this focus, WTO projects focus on providing support for the tourism sector and nations for the conservation of biodiversity, boosting climate action, reducing plastic pollution in the seas, providing clean energy projects for hotels, reducing resource inefficiency and mitigation of the pollution created by transportation, which also reveals the necessary areas to be focused on to achieve sustainable development through the enhancement of sustainability practices in tourism.

For these reasons, sustainable tourism is key for sustainable development. The size and scale of tourism will highlight the criticality of this proposition. Tourism is the second-largest industry after petrochemicals in international trade and is the second-highest industry for its energy consumption after manufacturing [68]. As a result, an established body of literature has emerged on sustainable tourism and its environmental effects, where the primary objective is to reconcile tourism with a theoretical basis to the targets of sustainable development and its principles, aiming to identify areas of convergence or divergence between sustainable development and sustainable tourism [69]. After a decade, studies still conclude on divergence, with sustainable tourism being evolved concurrently but independently from the broader paradigm of sustainable development [70]. Hence, an empirical investigation of the interrelation between these two areas of sustainability to establish the bridges, and to provide understanding about how these concepts align in both policy and practice, holds significance.

Tourism is also described as the “specter haunting our planet” [71,72]. Expanding upon the groundwork laid by early advocates of “alternative tourism” [73], the concept of sustainable tourism has gained growing recognition [74]. Sustainable tourism has been regarded as a purposeful reaction to the manifestation of neoliberal capitalism that worsened the negative consequences of tourism [75]. Murphy observed that the communication regarding sustainable tourism appears to be entangled within a loop involving academia and the government [76].

Briefly, tourism stands out as a highly dynamic sector globally and a major contributor to economic advancement in the early 21st century [77]. The increasing significance of tourism, offering both financial benefits and contributing to humanitarian causes, poses several challenges [77]. Sustainable tourism strives to diminish adverse impacts while amplifying positive social, economic and environmental outcomes [78,79]. Hence, sustainable development should be the inherent goal of tourism [80]. Iftikhar et al. explored the impacts of tourism on sustainable development, and the outcomes revealed that the quality of institutions played a crucial role as a moderating factor for BRI nations, resulting in a notable 4.7% improvement in their progress towards the sustainable development agenda [64].

5. Conclusions

This research aimed at the investigation of the interconnections among air pollution, environmental pollution, tourism, petroleum consumption, economic growth and life expectancy in Italy, Saudi Arabia and Türkiye, spanning the years 1975 to 2019. The country selection is due to the large shares of tourism in their GDP and the prominent role of religious tourism in these countries, where tourists are attracted by the religious destinations. The study employed Fourier-augmented vector autoregressive model and Fourier Granger causality tests to examine the relationships and to determine direction of causality. This study extended the literature by demonstrating how religious tourism and petroleum consumption exacerbate both CO₂ emissions and air pollution.

Our results showed that tourism and petroleum consumption Granger-cause CO₂ and PM_2.5 emissions. These findings emphasize the environmental and health costs of tourism by necessitating integrated policies targeting emission reductions and sustainable tourism practices. Reducing emissions from tourism by limiting petroleum consumption is essential for mitigating health risks and achieving sustainable development.

Beyond CO₂, PM_2.5 significantly deteriorates air quality by penetrating the lungs and bloodstream and causing severe health problems. Both short- and long-term exposure are linked to respiratory, cardiovascular and chronic diseases. Policies must recognize tourism’s environmental footprint by emphasizing energy efficiency, renewable energy transitions and sustainable practices. In Italy, Saudi Arabia and Türkiye, stricter environmental regulations and increased awareness can notably lower emissions in the tourism sector. Industry-specific policies to curb air pollution could offer effective avenues for emission reduction. Continuous efforts are vital to achieving sustainable tourism by emphasizing visitor satisfaction and sustainability awareness. Policies should incentivize tourists to embrace sustainable practices. Governments worldwide should encourage the promotion of green tourism initiatives. Additionally, policies should align sustainable tourism with sustainable development, economic growth and the transition to a better energy mix. In recent years, a new direction of research has identified black carbon [81], and future studies are also encouraged to extend the method and variable set of this study to black carbon as well as to other countries.

As a limitation, the dataset was ended in 2019 in this study, and the reason was data availability. Further, an important structural break occurred with COVID-19 in 2020; however, no adequate number of observations exists for the post-2019 period. As a result, the sample had to be ended at 2019. The results could be extended in future studies to capture the changes in the post-COVID-19 period. Future studies can be expanded to include other countries that are important in the context of religious tourism.