Tourism’s Influence on Economic Growth and Environment in Saudi: Present and Future

Abstract

Reports from the World Tourism Organization indicate that tourism activity has been increasingly booming; this sector is essential for economic growth and may affect the environment. Tourism is one of the key strategic sectors for planned growth in Saudi Arabia’s Vision 2030. This study is designed to evaluate the long-termning association between tourist arrivals, growth domestic product (GDP), and CO₂ emissions in the Kingdom of Saudi Arabia (KSA). The data related to these variables were assessed for the period 2010 to 2020. The autoregressive distributed lag (ARDL) bounds results revealed that there are long-established relations between tourist arrivals and growth domestic product and tourist arrivals and CO₂ emissions. The dynamic ordinary least squares (DOLS) and fully modified ordinary least squares (FMOLS) model outcomes were compatible with the outcomes of the ARDL model. In reference to the Granger causality test, tourist arrivals cause (affect) the GDP. Such a result confirms the conception that tourism encourages economic growth. No causality runs from tourist arrivals towards CO₂ accumulation. This result may reflect the governmental effort to reduce CO₂ accumulation and/or to perform tourism activities in a sustainable way. The results predicted that the growth rate of tourist arrivals, GDP, and CO₂ accumulation equal 0.0023, 0.048, and 0.0169, respectively, during the forecast period (2021–2030), which appeared to be increasing for tourist arrivals and GDP and decreasing for CO₂ accumulation. The study recommended that, to increase economic growth, tourist arrivals should be increased alongside performing tourism activities in a sustainable way. These findings point to the benefits of governance in ensuring effective policies to decarbonize the environment, and policy proposals are put forward accordingly.

1. Introduction

Worldwide, climate change is a key issue impacting the environment [1]. The level of carbon dioxide (CO₂) emissions drives global warming [2]. The emissions of CO₂ come from different economic and social activities, among which are tourism activities. The tourism sector plays a vital role in job creation, increasing income levels, foreign exchange reserves, and (accordingly) economic growth [3].

This study is consistent with the United Nations Sustainable Development Goals (UN-SDGs 7, 8, 11, 12, and 13) [4]. These goals range from access to cleanliness and responsible energy consumption to sustainable economic growth and climate change mitigation. These goals are consistent with the sustainability goals of the Kingdom of Saudi Arabia.

Due to its vital role in the economy, tourism has been the subject of a huge body of research, highlighting its impact [5,6,7,8]. Correspondingly, sustainability is considered as a strategic part of tourism. The World Trade Organization (WTO) states that sustainable tourism is a “development that meets the needs of present tourists and host regions while protecting and enhancing opportunity for the future” [5,9]. Several writers have considered it as a method to fulfill the requirements of the investors, considering social, environmental, and economic effects [5,10].

Tourism is consideredto be as one of the main strategic sectors for planned growth in Saudi’s Vision 2030. In 2019, Saudi Arabia witnessed a strong boost in the tourism sector, involving openness to entertainment and theatrical events, or events related to international football, golf, boxing, car racing, and tennis [4]. The KSA assumes tourism will form about 10% of the nation’s GDP by 2030. In addition, building an innovative tourism economy will create an abundance of job opportunities (a minimum of one million new jobs) for youth in the KSA [11]. The total contributions from tourism to GDP in the world and in the KSA were equal to 6.1% and 6.5%, respectively, in 2021 [12]. In addition, the total contributions of tourism to job creation in the world and in the KSA were equal to 287 MN (=one in eleven jobs) and 1.30 MN, 10.0% of total jobs, respectively, in 2021 [12]. Due to CO₂ emissions, the tourism industry has a negative impact on the environment [13]. High energy consumption comes from transportation, agricultural production, and food processing.

Tourism and economic growth are closely related and interacting pillars. Referring to the World Tourism Organization (WTO) reports, tourism activity is crucial for job creation and economic growth [14]. Several studies have weighed the association between tourism and economic growth. In reference to the tourism economic literature, previous studies have mostly concentrated on the impact of tourism on economic growth [15], poverty [16], and job creation [17].

Four hypotheses have described the association between economic growth and tourism: the tourism-led growth hypothesis declares a positive effect of tourism on economic growth [18]; there is a hypothesis that states that tourism is motivated by economic growth [19]; there is a hypothesis that states that bidirectional causality exists between economic development and tourism [20]; finally, there is a hypothesis that states that no causal connection exists between economic growth and tourism [18].

With regard to tourism and the environment (CO₂), the influence of tourism significantly differs, ranging from promoting sociocultural values to prompting environmental degradation [21]. The connection between tourism and CO₂ emissions is composite [22].

Among economic sectors, international tourism is characterized by continuous growth; from 2010, it become one of the globe’s leading and quickest-growing sectors [23]. An important factor contributing to economic growth is the availability of energy that supports several economic sectors. Energies such as electricity are frequently generated from burning fossil fuels that contribute to CO₂ emissions (environmental contamination) [24]. Tourism is considered to be an energy-intensive sector. The rapid development of the tourism sector necessitates key investments in other sectors, including transportation and infrastructure [25].

Within the tourism sector, CO₂ emissions are classified into indirect and direct emissions [26]. The production of intermediate goods in this industry and final energy consumption represent indirect and direct CO₂ emissions, respectively. Concerning the consequence of tourism events on the environment, researchers have stated two contrasting viewpoints. Certain studies determined that the tourism sector contributes to CO₂ emissions and, accordingly, global warming [14,27,28,29,30,31,32]. Other studies have illustrated an opposing association between tourism and the environment (CO₂ emissions), concluding that tourism activities must be run in a sustainable manner [14,33,34].

In addition, several studies highlight the relation between tourism and CO₂ emissions in North Africa [35,36] and the Middle East [37]. Other studies at the country level have used energy consumption as a variable in their analyses to identify such a relationship in North Africa [38], Cyprus [39], and Singapore [40]. Simon Kuznets [41] used a U-form curve to describe the relationship between income inequality and income per capita. Grossman and Krueger [42] exploited the idea of the U-form curve in the association between economic development and environmental degradation (the Environmental Kuznets Curve (EKC) hypothesis).

Also, many studies have confirmed that tourism increases CO₂ emission at the country level, e.g., such studies have been conducted for France [43], Turkey [44], and Cyprus [39]. In reference to tourism-related EKC panel analyses, countries in the European Union have been illustrated to have a statistically significant and negative effect on CO₂ accumulation [45].

Many studies have confirmed that tourism has considerably harmful effects on the environment due to CO₂ emissions [28,30]. Another study determined that the size of energy structure, tourist production, and tourist traffic influence CO₂ emissions [46]. Tourism transportation is the most prolific cause of pollution [47,48]. Also, tourism accommodations are playing a vital role in carbon emission accumulation [22,49,50].

The value of this study has emerged from the controversial results of previous studies that examined the impact of tourism on the environment (CO₂ emissions) and economic growth. The paper highlights studies that examine the impact of tourism on CO₂ emissions in various countries; there is a limited focus on the KSA in this context. Also, the ARDL bounds, DOLS, FMOLS, and Granger causality are appropriate for application in addressing this gap. Hence, the study’s objective is to test the effect of tourist arrivals on CO₂ emissions and economic growth in the KSA. Further, the study enriches the existing body of research on this issue.

The main objective of the study is to discover the effect of tourist arrivals on the GDP and CO₂ emissions in the KSA. The specific objectives of this study include the following, accounting for the variables (tourist arrival, GDP, and CO₂ emission) under study:

• Disclosing the relationship between the variables using visual analysis (graph).
• Determine the extent of long- and short-term relationships between the variables.
• Identify the causality direction among the variables.
• Discover the magnitude of the growth rates for the variables in the forecast period.

The study hypothesizes that tourist arrivals have a positive effect on the GDP and a negative effect on CO₂ emissions:

• There is integration between these factors.
• There are existing long- and short-term relationships between these factors.
• There is bidirectional causality between these factors.
• The growth rates of the variables in the forecast period are higher than those in the study period.

The present study is composed of four sections: introduction, research methods, results, and discussion. The final section presents our conclusions and recommendations.

2. Materials and Methods

2.1. Data Description

The data, along with their explanations and information, are presented in

Table 1. Variables description.

Variable	Unit	Sources
Carbon dioxide (CO2)	Ton	“https://countryeconomy.com/energy-and-environment/co2-emissions/saudi-arabia (accessed on 17 March 2023)”
Annual Growth Domestic Product (GDP)	Million USD	“ https://countryeconomy.com/gdp/saudi-arabia (accessed on 18 March 2023)”
Tourism arrival/year (Tou)	Person	“https://countryeconomy.com/trade/international-tourism/saudi-arabia (accessed on 20 March 2023)”

. To illustrate the relationship between the study variables—tourism, GDP, and environment—the data were collected from 2010 to 2020.

2.2. Analysis Methods

2.2.1. Descriptive and Graphical Analysis (for Testing Hypothesis 1)

The data under study were subjected to descriptive and graphical analyses to test their cointegration in a visual analysis.

2.2.2. Cointegration Tests (for Testing Hypothesis 2)

Refs. [51,52] presented the following equation to examine the unit root for the variables of the study.

$$\Delta X_t=C_{t1}+b_1{X}_{t-1}+E_{t1}$$

(1)

$$\Delta X_t=C_{t2}+B_t+b_2{X}_{t-1}+E_{t2}$$

(2)

where the ADF coefficients to be estimated are b₁ and b₂, the trend is B, C is the constant, the error term is E, and the time selected is t. Accepting the null hypothesis (H₀) means that the series is stationary. The chain variables will be stationary if the t-statistics of the ADF coefficients are bigger than the t-critical values.

• The autoregressive distributed lag (ARDL) bounds test was used to assess the long-term connotations among the series. It can be characterized as being effective for minor remarks in time–series data and can be applied regardless of the series order (i.e., it can be applied in I (1) and I (0), by not I (2)). The acceptance of the null hypnosis signifies that there is no long-term connection (null hypothesis: b₃ = b₄ = 0 against alternative hypothesis b₃ ≠ b₄ ≠ 0 in Equation (3)). Likewise, a comparable assessment can be practiced for the Y series (null hypnosis: b₅ = b₆ = 0 against alternative hypothesis b₅ ≠ b₆ ≠ 0 in Equation (4)). This was applied using the following equations [52,53]:

$$\Delta X_t=C_1+{\displaystyle \sum }_{t-1}^p{a}_1\Delta Y_{t-1}+b_3{X}_{t-1}+b_4{Y}_{t-1}+e_1$$

(3)

$$\Delta Y_t=C_2+{\displaystyle \sum }_{t-1}^p{a}_2\Delta X_{t-1}+b_5{Y}_{t-1}+b_6{X}_{t-1}+e_2$$

(4)

where a₁, a₂—coefficients of the difference of lag-independent variables; b₃, b₅—coefficients of lag dependent variable; b₄, b₆—coefficients of lag-independent variable, respectively; e₁, e₂—error terms.

To examine the steadiness of the ARDL model estimates, the cumulative sum (CUSUM) method was used [54]. If the residual is sited inside the 5% critical margins, the valued coefficients steadily signify a 5% level of significance. Equations (3) and (4) have upper and lower bounds with critical F-values; this confirms the joined order 1(1) and 1(0) of the series, respectively [53].

Ref. [55] recorded that, to support the results of ARDL statistics, the dynamic ordinary least squares (DOLS) can be used in addition to fully modified ordinary least squares (FMOLS). The DOLS and FMOLS model outcomes stretch the realistic estimations of long-term relations among the series.

2.

Error correction model (ECM) testing of short-term relations: The ECM test is run to weigh the quickness factor of the short-term connotation among the variables [56]. It applies after the detection of a long-term connotation among the variables. Equations (5) and (6) represent the ECM model [52]:

$$\Delta \left(\mathrm{X}\right)=\Delta b_7{\mathrm{X}}_{-1}+\Delta b_8{\mathrm{Y}}_{-1}+b_9{U}_{t-1}+V_1$$

(5)

$$\Delta \left(\mathrm{Y}\right)=\Delta b_{10}{\mathrm{Y}}_{-1}+\Delta b_{11}{\mathrm{X}}_{-1}+b_{12}{U}_{t-1}+V_2$$

(6)

where b₇ and b₁₀—coefficients of difference of lag dependent variable; b₈ and b₁₁—coefficients of difference of lag-independent variable; b₉ and b₁₂—speed of adjustment (requisite be negative and significant to fixed model instability); U_t−1—error correction term; V₁ and V₂—error terms.

Residual diagnostics tests were used to check ECM viability, as follows: the Breusch–Pagan–Godfrey test was used for heteroskedasticity—the null hypothesis is accepted when the p-value is higher than 0.05, demonstrating that the residual is homoscedasticity. A normality test was conducted—if the Jarque–Bera statistic’s probability is higher than 0.05, then the residual is normally scattered.

2.2.3. Granger Causality (for Testing Hypothesis 3)

Granger causality can be determined using Equations (7) and (8) [57]:

$$X_t={\beta }_0+{\displaystyle \sum }_{j-1}^n{\beta }_{1j}{X}_{t-j}+{\displaystyle \sum }_{h-1}^m{\beta }_{2h}{Y}_{t-h}+e_{1t}$$

(7)

$$Y_t={\alpha }_0+{\displaystyle \sum }_{s-1}^k{\alpha }_{1s}{Y}_{t-s}+{\displaystyle \sum }_{i-1}^m{\alpha }_{2i}{X}_{t-m}+e_{2t}$$

(8)

From Equations (7) and (8), it is supposed that the e_1t and e_2t are unlinked, as e (ε1t, ε2t) = 0 = e(ε2tε2s) ….s ≠ t. Four possible results can be drawn from the equations, as follows:

-

If the coefficient β_2h ≠ 0 and is statistically significant, then Y → Granger causes → X [57,58].

-

If α_2i ≠ 0 and is statistically significant, then X causes a variable for Y [57,59,60].

-

The significance of α_2i and β_2ℎ (≠0) affirms the joint reliance of X and Y.

-

If α_2i and β_2ℎ = 0, then Y and X will be autonomous.

2.2.4. Forecasting Test (for Testing Hypothesis 4)

A forecasting test was run on tourism. The growth rates were calculated for the period 2021 to 2030 (predicting periods). The growth rate was computed using the following equation [61]:

$$W_t=T_t-T_{t-1}÷T_{t-1}$$

(9)

where W_t denotes the growth rate in two successive years. T signifies tourist arrivals and t denotes the year. The growth rate of the variable for the years from 2010 to 2020 and for the years from 2021 to 2030 is calculated by the following equation [61]:

$$R_{t\dots m}=(R_t+R_{t+1}\dots +R_{m)})/m$$

(10)

where R_t…m and m denote the growth rate for a precise period and the number of years, respectively.

3. Results and Discussion

The study aimed to disclose the relationships between tourism, economic growth (GDP), and the environment (CO₂). Accordingly, the study comprises two parts: the cointegration association between tourism and GDP and the cointegration association between tourism and CO₂. Also, the Granger causality test and a forecasting analysis were conducted.

3.1. Descriptive Statistics

The probabilities derived through the Jarque–Bera statistics are higher than 0.05 for all series except the one with the tourism variable, indicating the presence of a normal distribution (

Table 2. Descriptive statistics of the variables.

	CO2	GDP	TOU
Mean	564.99	704,524.6	14,957,983
Median	574.92	703,368.0	16,109,000
Maximum	611.42	816,578.0	18,260,000
Minimum	487.01	528,207.0	4,138,178
Std. Dev.	40.32	81,456.75	4,180,961
Jarque–Bera	1.03	0.73	7.77
Probability	0.60	0.70	0.021
Observations	11	11	11

Source: Data were collected (from the sources mentioned in Table 1) and calculated.

). So, the data were transformed into a logarithmic form. Figure 1 shows a graphical analysis of the cointegration. It appears that the variables under study move up and down together, leading to a conclusion that the variables might be cointegrated.

3.2. Part One: Co-Integration Test Analysis Results (Tourism and GDP)

3.2.1. The Results of the Unit Root Tests

The ADF statistics of GDP and tourism are significant (1% level), concluding the stationarity of the series (

Table 3. Results of unit root test.

	LogCO2	LogGDP	Logtou
Intercept (at level)	−2.76 *	2.83 *	−1.28
Intercept and trend (at level)	−0.91	−2.95	−0.67
Stationarity (at level)	Stationary	Stationary	Non stationary
Intercept (at first difference)	−1.57	−2.18	−1.24
Intercept and trend (first difference)	−2.38	−1.96	−3.90 *
Stationarity (first difference)	Non stationary	Non stationary	Stationary

Source: Data were collected (from sources mentioned in Table 1) and calculated. * 1% level of significance. CO₂—carbon dioxide; GDP—annual GDP; tou—tourism arrival/year; Log—logarithm.

). Since the stationarity order is dissimilar (1(0) and 1(1)), the ARDL model was used to evaluate the relation.

3.2.2. Results of ARDL Tests: Tourism and GDP

The ARDL model viability and residual diagnostics tests results reflect the respective feasibility of the findings. The following tests were used: serial correlation LM test (Breusch–Godfrey); the Breusch–Pagan–Godfrey heteroskedasticity test; the Jarque–Bera test. The findings of these tests indicated no serial correlation, no heteroskedasticity, and residual normal distribution, respectively (

Table 4. Results of ARDL tests (tourism and GDP).

Model
LGDP (as Dependent Variable) Choice ARDL Model (2, 2)
Independent V.	Coefficient
LGDP(−1)	0.53 (0.15)
LGDP(−2)	−0.49 (018)
LTOU	−0.05 (0.30)
LTOU(−1)	−0.79 (0.08)
LTOU(−2)	0.49 (0.04)
C	8.17 (0.04)
R-squared 0.92    Adj. R-squared 0.7
F-statistics 6.85            Prob. 0.07
Jarque–Bera test: 1.24        Prob. 0.54 Serial correlation LM test: Breusch–Godfrey 0.26  Prob. 0.81
Breusch–Pagan–Godfrey heteroskedasticity test: 0.09 Prob. 0.98

Source: Data were collected (from sources mentioned in Table 1) and calculated. LM—Lagrange multiplier. Numbers within ( ) are probabilities.

). Similarly, A CUSUM test was used to assess the steadiness diagnosis [62]. From Figure 2, we can see that the test presented the steadiness of the cumulative sum of the recursive residuals, reflecting the power of the model.

Table 5. The bounds test results: ARDL (GDP as dependent variable).

Dependent V.	Independent V.	F-statistic of bound test
LGDP	LTOU	9.77 *
Significance	Lower Bound	Upper Bound
1%	4.94	5.58
5%	3.62	4.16
10%	3.02	3.51

Source: Data were collected (from sources mentioned in Table 1) and calculated. Note: * means significance at 1% level.

shows the results of the bound test. The bound test was applied to the F-test for the GDP coefficient (one lag period) and tourist arrivals (as the independent variable), deriving a result of 9.77. The F-statistic is bigger than the upper bound of the critical F-statistic (5.58) at 1%, indicating the presence of the long-term relationship among GDP and tourist arrivals during the study period. The results are synchronized with those of a previous study [63]. The outcomes of [63] showed that tourism, financial development, government governance policies, economic growth, and foreign direct investments are positive and significant causes leading to increased CO₂ emissions in Africa.

Long-Term Endorsement

DOLS and FMOLS models were applied to support the outcomes of the ARDL valuations (

Table 6. Results of robustness test.

FMOLS Model		DOLS Model
LGDP (as a Dependent Variable)		LGDP (as a Dependent Variable)
Independent V.	Coefficient	Independent V.	Coefficient
LTOU	0.81 *	LTOU	−0.88 *
C	78.26	C	12.19 *
R-squared 0.92 Adj. R-squared 0.91		R-squared 0.81 Adj. R-squared 0.70

Source: Data were collected (from sources mentioned in Table 1) and calculated. * 1% level of significance.

). The results of DOLS and FMOLS indicate that a long-term relationship was observed between GDP and tourist arrivals. Several former studies run DOLS and FMOLS to reinforce the outcomes of the ARDL model [52,64,65].

The results of DOLS and FMOLS are compatible with those of the ARDL assessment (

Table 7. Long-term relationship confirmation.

Dependent–independent	DOLS	FMOLS	ARDL
LGDP–LTOU	0.88 *	0.81 *	9.77 *

Source: Table 5 and Table 6. * Significance level at 1%; LGDP—GDP; LTOU—tourist arrivals; DOLS—dynamic ordinary least squares; FMOLS—fully modified ordinary least squares; ARDL—autoregressive distributed lag L-logarithm.

). A long-term relationship was noticed among GDP and tourist arrivals.

Results of Short-Term Tests: ECM Tests

To support the outcomes of the ARDL model (long-term connotation), an ECM was conducted to assess short-term association among the series. Lag 1 was selected (

Table 8. Lag selection results.

Lag	LogL	LR	FPE	AIC	SC	HQ
0	19.67	NA *	6.77 × 10−5	−3.93	−3.88	−4.02
1	32.97	7.35	2.75 ×10−5 *	−5.11 *	−4.89 *	−5.58 *

Source: Data were collected (from sources mentioned in Table 1) and calculated. * Significance level at 1%. Nore: lag order was selected by the criterion; logL—log lag variable; LR—sequential modified likelihood ratio (LR) test statistic (each test at 5% level); FPE—final prediction error; AIC—Akaike information criterion; SC—Schwarz information criterion; HQ—Hannan–Quinn information criterion; NA—not available.

), which is essential for performing ECM. From

Table 9. Results of ECM.

Error Correction	D(LGDP)	D(LTOU)
CointEq1	−0.84 [−5.75]	−2.63 [−1.38]
D(LGDP(−1))	0.42 [3.37]	1.72 [1.06]
D(LTOU(−1))	−0.42 [−3.73]	−1.44 [−0.98]
C	0.0035 [0.58]	−0.06 [−0.77]
ECM residual serial correlation LM tests: Lags  LM-Stat  Prob. 1  6.31  0.18 VEC residual heteroskedasticity tests: Chi-sq 18.93  Prob. 0.40
Jarque–Bera statistic    7.44    Prob. = 0.11

Source: Data were collected (from sources mentioned in Table 1) and calculated. Numbers in [] represent critical t-values.

, the coefficient of the alteration factor for GDP (dependent variable) verified the negative sign (−0.84) and the significant value (critical t-value = −5.75), leading to the conclusion that the model has the power to correct the variability of its prior period. Similarly, the outcomes suggest that the adjustment parameter coefficients for tourist arrivals (dependent variable) was negative (−2.63) and statistically insignificant (critical t-value = −1.38). This indicates the powerlessness of the model to accurately predict the changeability that took place in the past. The model suitability tests were applied. The Jarque–Bera statistic equals = 7.44 with Prob. = 0.11, indicating that the residual has a normal distribution. The LM statistic (lag 1) equals 6.31, with Prob. = 0.18, signifying no serial correlation. Chi-sq. equals 18.93 with Prob. = 0.40, demonstrating no heteroskedasticity. These outcomes for the model’s suitability led us to accept the null hypothesis; this is a result of the normal distribution of the residuals, the lack of a serial correlation among the residuals, and no heteroskedasticity.

3.2.3. The Results of Granger Causality Test: (GDP and Tourist Arrivals)

The F-statistics of the Granger causality test equal 7.02, with probabilities equal to 0.04 (

Table 10. Granger causality tests results.

Null Hypothesis	F-Statistic	Prob.
LTOU─LGDP	7.02	0.04
LGDP─LTOU	0.73	0.54

Source: Data were collected (from sources mentioned in Table 1) and calculated. Where; Null Hypothesis means no Granger causality.

), meaning that the causality (unidirectional) goes in the direction of tourist arrivals → GDP. This outcome corroborates the findings of earlier studies [18,65]. Studies have shown that tourism has a casual power on economic growth (unidirectional).

3.3. Part Two: Co-Integration Results (Tourism and CO₂)

3.3.1. Results of ARDL Tests: Tourism and CO₂

The ARDL tests outcomes are found in

Table 11. Results of ARDL tests (tourism and CO₂).

Model 1
L CO2 (as a Dependent Variable) Chosen ARDL Model (1, 1)
Independent V.	Coefficient
L CO2 (−1)	0.35 (0.15)
LTOU	0.02 (0.25)
LTOU(−1)	0.19 (0.12)
C	0.29 (0.51)
R-squared 0.88 Adj. R-squared 0.83 F-statistics 6.85 Prob. 0.07
Jarque–Bera test: 0.34 Prob. 0.84 Serial correlation LM test: Breusch–Godfrey 0.65 Prob. 0.57
Breusch–Pagan–Godfrey heteroskedasticity test: 1.02 Prob. 0.45

Source: Data were collected (from sources mentioned in Table 1) and calculated. LM—Lagrange multiplier. Numbers in ( ) are probabilities.

. Residual diagnostics tests were used to evaluate the ARDL model fitness. The residual diagnostics indicators showed no serial correlation and no heteroskedasticity, and the residual was normally distributed. Similarly, to assess the model stability, a CUSUM test was applied [62]. Figure 3 shows the steadiness of the cumulative sum of the recursive residuals, representing the strength of the model.

Table 12. Results of the bounds test: ARDL.

Dependent V.	Independent V.	F-statistic of bound test
LCO2	–LTOU	5.79 *
Significance	Lower Bound	Upper Bound
1%	4.94	5.58
5%	3.62	4.16
10%	3.02	3.51

Source: Data were collected (from sources mentioned in Table 1) and calculated. Note: * means a 1% level of significance.

documents the outcomes of the bound test. The F-statistic (5.79) is greater than the upper bound of the critical F-statistic (5.58) at 1%, indicating the presence of a long-term relationship between CO₂ and tourist arrivals (independent variable). This result corroborated those of a previous study [66], which revealed that there is a positive relationship between tourism and CO₂ emissions.

Long-Term Relationship Confirmation

Table 13. Results of robustness test.

FMOLS Model		DOLS Model
LCO2 (as Dependent Variable)		LCO2 (as Dependent Variable)
Independent V.	Coefficient	Independent V.	Coefficient
LTOU	0.38 (0.000)	LTOU	0.63 (0.001)
C	78.26 (0.22)	C	−1.78 (0.021)
R-squared 0.92 Adj. R-squared 0.91		R-squared 0.93 Adj. R-squared 0.83

Source: Data were collected (from sources mentioned in Table 1) and calculated. Figures between ( ) are probabilities.

shows the outcomes of DOLS and FMOLS for testing the long-term association. From

Table 14. Long-term relationship confirmation.

Dependent–independent	ARDL	FMOLS	DOLS
LCO2–LTOU	5.79 *	0.38 *	0.63 *

Source: Table 12 and Table 13. * Significance level at 1%; all indicators are illustrated in Table 7; LCO₂—carbon dioxide.

, the results of FMOLS, DOLS, and ARDL are matched, meaning that a long-term association was observed among CO₂ and tourist arrivals.

Results of ECM

Lag 1 was selected (

Table 15. Lag selection results.

Lag	LogL	LR	FPE	AIC	SC	HQ
0	25.96	NA	2.85 × 10−5	−4.79	−4.73	−4.86
1	36.86	15.27 *	7.43 × 10−6 *	−6.17 *	−5.99 *	−6.37 *

Source: Data were collected and calculated. Note: all indicators are illustrated in Table 8; * means 1% level of significance.

). The model suitability checks were run (

Table 16. The results of the ECM test.

Error Correction	D(LCO2)	D(LTOU)
CointEq1	0.15 [1.09]	3.89 [2.44]
D(LCO2(−1))	0.59 [1.12]	14.22 [2.34]
D(LTOU(−1))	0.05 [0.36]	−3.26 [−2.03]
C	−0.0002 [−0.03]	−0.10 [−1.51]
ECM residual serial correlation LM tests: Lags LM-Stat Prob. 1 7.21 0.16
VEC residual heteroskedasticity tests: Chi-sq 21.52 Prob. 0.25
Jarque–Bera statistic 1.03 Prob. = 0.91

Source: Source: Data were collected (from sources mentioned in Table 1) and calculated. Numbers in [ ] represent critical t-values.

); the Jarque–Bera statistic was 1.03, with Prob. = 0.91, indicating that the residual is normally distributed. Also, the LM statistic (lag 1) equals 7.21, with Prob. = 0.16, signifying no serial correlation. Chi-sq. equals 21.52, with Prob. = 0.25, demonstrating no heteroskedasticity. These outcomes of model suitability let us to accept the null hypothesis; there was no normal distribution of the residuals, no serial correlation of the residuals, and no heteroskedasticity. From Table 16, the coefficient of the adjustment parameter for CO₂ (as the dependent variable) was positive and insignificant (critical t-value = 1.09), leading us to identify the inability of the model to correct the variability of the prior period. Equally, the outcomes propose that the coefficients of the adjustment parameter for tourist arrivals (as dependent variable) was positive (3.89) and statistically insignificant (critical t-value = 2.44); this indicates that the model was powerless to accurately identify the changeability of the past. The mentioned results conclude that there is a short-term relationship between tourist arrivals and CO₂.

3.3.2. Results of Granger Causality Test: (CO₂ and Tourist Arrivals)

From

Table 17. Granger causality results.

Null Hypothesis	F-Statistic	Prob.
LTOU─LCO2	2.40	0.17
LCO2─LTOU	0.10	0.76

Source: Data were collected (from sources mentioned in Table 1) and calculated. Where; Null Hypothesis means no Granger causality.

, we can see that the F-statistics of the Granger causality test are equal to 2.40 and 0.10, with probabilities equal to 0.17 and 0.76, respectively, meaning that there is no causality among CO₂ accumulation and tourist arrivals. These results may be due to the adoption of sustainable tourism activities. Previous studies confirm theories surrounding the sustainability of the relationship between tourist arrivals and CO₂ emissions [14,33,34]. Also, the result may reflect governmental efforts to reduce CO₂ accumulation [52].

3.4. Forecasting Results

Figure 4 displays forecasting graphs of tourist arrivals; the blue and red portions show real tourist arrivals for the years from 2010 to 2020 and the predicted years from 2021 to 2030, respectively. The chart shows strength through diverse forecasting indicators: Alpha, mean square error, Gamma, and Beta are equal to 0.9, 0.92, 0.0, and 0.0, respectively. Equally, the tourist arrival growth rates are equivalent to 0.00019 and 0.0023 for the period 2010 to 2020 and for forecast period 2021 to 2030, respectively, leading to the conclusion that there was a rise in tourist arrivals by (0.0023) during the forecast period.

Figure 5 shows the forecasting graphs of GDP; the blue and red portions show the real GDP for the years from 2010 to 2020 and the predicted years from 2021 to 2030, respectively. The chart shows strength through the used of diverse forecasting pointers: Alpha, mean square error, Gamma, and Beta are equal to 0.9, 0.87, 0.0, and 0.0, respectively. Equally, the tourist arrival growth rates are equivalent to 0.016 and 0.0.019 for the period 2010 to 2020 and for forecast period 2021 to 2030, respectively, leading to the conclusion that there was a rise in GDP by (0.019) during the predicted period.

Figure 6 illustrates the forecasting graphs of CO₂; the blue and red portions show real CO₂ for the years from 2010 to 2020 and the predicted years from 2021 to 2030, respectively. The chart shows strength through the used of diverse forecasting pointers: Alpha, mean square error, Gamma, and Beta are equal to 1.00, 0.86, 0.0, and 0.0, respectively. Equally, the CO₂ growth rates are equivalent to 0.0197 and 0.0169 for the period 2010 to 2020 and for forecast period 2021 to 2030, respectively, leading to the conclusion that there was a decrease in CO₂ by (0.0169) during the forecast period.

Table 18. Growth rates of the variables.

Growth Rate	2010–2020	2021–2030 (Forecast Period)
Variables
Tourism arrivals	0.00019	0.0023
GDP	0.040	0.048
CO2	0.0197	0.0169

Source: Data were collected (from sources mentioned in Table 1) and calculated.

shows the growth rate of tourism arrivals, the growth domestic product, and CO₂. The growth rates are increasing for tourism arrivals and growth domestic product from 2010–2020 to 2021–2030; meanwhile, those for CO₂ decrease during the forecasting period, leading to the conclusion that tourism activities should be conducted in a sustainable way. Also, the result of the decreasing growth rate of CO₂ may reflect a continuation in governmental efforts towards reducing CO₂ emissions. Previous studies have illustrated efforts to reduce CO₂ emissions [52].

4. Conclusions and Recommendations

The study assesses the association between tourism, economic growth (GDP), and the environment (CO₂ emission) in the KSA. Data were gathered from numerous sources for the period from 2010 to 2020. The outcomes from the ARDL bounds test indicated that there are long-term relationships between tourist arrivals and GDP and between tourist arrivals and CO₂ emissions. The DOLS and FMOLS model outcomes corroborated the outcomes of the ARDL model. With reference to the Granger causality test, tourist arrivals caused an increase in the GDP; this result supports the hypothesis of “tourism-led economic growth”. No causality was found for the impact of tourist arrivals on CO₂ accumulation. This result may reflect governmental efforts in reducing CO₂ accumulation and/or the conducting of tourism activities in a sustainable way. The present study makes the following recommendations: economic growth that is compatible with the safety of the environment (reduction in CO₂ emissions) should be pursued; tourist arrivals should be increased in combination with performing tourism activities in sustainable way. The results predict that the growth rates of tourist arrivals, GDP, and CO₂ accumulation are 0.0023, 0.048, and 0.0169, respectively, during the forecast period (2021–2030); these growth rates appeared to increase for tourist arrivals and GDP and to decrease for CO₂ accumulation. The study recommended that, to increase economic growth, tourist arrivals should be increased in combination with the sustainable conduction of tourism activities. These findings highlight the benefits of governance in ensuring effective policies for decarbonizing the environment; policy proposals are put forward accordingly.