In the global economic outlook, the relationships between economics and environmental issues have been among the most important topics of discussion for the last 50 years among researchers (Lau et al. 2014). In line with economic growth, the excessive consumption of fossil fuels by industrial countries has led to an increase in greenhouse gas (i.e., CO2, CH4, O3, and NO2) emissions in the atmosphere (Xiong et al. 2022), and as a result, many problems such as global warming, environmental pollution, and their result, “climate change,” have emerged (Yousaf et al. 2022). While this situation causes concerns for the future of the world, it is one of the most important affairs that scientists and politicians have to deal with (Balsalobre-Lorente et al. 2021), and with its dynamic structure, the tourism sector has significant impacts on climate change. The fact that the countries’ economic benefits take precedence over environmental degradation, it is of great importance to address the issue including all countries and sectors. To this end, this study provides scientific evidence for possible energy consumption-climate scenarios depending on tourism activities in a growing Turkish economy, which uses non-linear time-series techniques concerning carbon emissions.

Climate change is a global threat that affects many sectors and all segments of society. One of the most essential and affected sectors is the tourism sector, as it is sensitive to the environment (Dogru et al. 2019). Prior to the pandemic in 2019, 1.5 billion people worldwide participated in tourism activities, generating 1.7 trillion dollars. Furthermore, tourism supplies 10.3% of the global gross domestic product (GDP), while it accounts for 10% of employment worldwide (WTTC 2020). The relationship between tourism and environmental issues has piqued interest and grown in importance, particularly since the 1970s. The negative effects of increasing energy consumption (transportation, infrastructure, superstructure construction, and lodging) combined with tourism demands on climate change and its result, “global warming”, have been a critical research area (Liu et al. 2011; Rutty et al. 2021). Tourism is, by definition, a displacement activity accounted for 5% of all emissions produced in 2016 (World Tourism 2019). According to the Intergovernmental Panel on Climate Change (IPCC), global warming of 1.5 °C and 2 °C above preindustrial levels will be exceeded during the twenty-first century and precipitation will decrease until 2100, but there will be frequent and heavy rains depending on these changes (IPCC 2021), which will potentially have an impact on tourism sector.

Contrary to those recent global facts, with proper planning and sustainable policies, tourism’s negative effects on climate change are expected to be mitigated (Gössling 2002). In this regard, it is possible to state that an advanced tourism sector reduces carbon emissions through environmentally friendly practices and transportation (Sghaier et al. 2019). Because increased energy efficiency and use of clean energy sources will benefit tourism’s sustainability, the tourism sector will decrease the use of conventional energy (i.e., fossil fuels). Investments in electricity, wastewater treatment, reuse, and rainwater recovery within the framework of sustainable tourism will raise energy efficiency and reduce CO2 emissions (Ben Jebli et al. 2019). In addition, tourists who are sensitive to environmental deterioration may tend to prefer renewable energy resources and infrastructures providing energy efficiency (Balsalobre-Lorente et al. 2021, Hoogendoorn and Fitchett 2018).

Environmental pollution, energy consumption, and economic growth have all been studied in the literature (see, for example, Ben Jebli and Ben Youssef 2015; Bilgili et al. 2016; Bölük and Mert 2015; Nathaniel et al. 2020). Most of these studies emphasize that economic growth increases energy consumption and has negative environmental consequences. Furthermore, some studies investigating the effects of tourism activities on the environment have also been conducted in Turkey (see, for example, Eyuboglu and Uzar 2020; Godil et al. 2020; Katircioglu 2014; Ozturk et al. 2016; Uzuner et al. 2020; Yorucu 2016). From these studies, Uzuner et al. (2020) investigated the relationships between globalization, tourism, CO2 emissions, and economic growth using the asymmetric causality test method and concluded that tourism has an impact on CO2 emissions. Tourism, globalization, and financial development are positively and significantly related to ecological footprint, according to Godil et al. (2020). These studies looked at the effects of CO2 emissions from tourism on environmental pollution.

However, the studies examining the relationships between climate change and energy consumption in the tourism sector are very limited. For instance, Ohajionu et al. (2022) used econometric time-series approaches to investigate the relationship between income per capita, tourism, foreign direct investment, local credit given to the private sector, and CO2 emissions, and they found that tourism is negatively correlated with CO2 emissions. Nonetheless, while previous research has looked at the effects of tourism on CO2 emissions and the environment, no direct research on climate change has been found. To fill this void, this study investigates the relationships between global climate change and eco-environmental factors for Turkey.

Turkey welcomed approximately 52.5 million tourists in 2019 and ranked sixth among the world’s top tourist destinations. Because mass tourism attracts a large proportion of tourists, developing a tourism policy within the framework of a sustainable tourism has become critical in terms of the resource use. In this regard, the effects of energy consumption, tourism activities, and their impact on climate change have been examined in light of the concept of possible cointegration and policies have been proposed based on the findings. In order for the tourism sector to achieve long-run sustainable development, it is necessary to adapt climate change effects and to reduce the activities that cause climate change. Along with the above-mentioned conceptual framework, the dependent variables are precipitation and temperature, which represent climate change, and the independent variables are tourist arrivals, NREC, REC, and GDP, which represent eco-environmental indicators. Data is collected by the Turkish State Meteorological Service, International Energy Agency, and World Development Indicators, from 1995 to 2020.

This paper contributes to the growing literature with two different models through temperature and precipitation which are considered and conducted with the non-linear autoregressive distributed lag (NARDL) method. Both models show a long-run cointegration relationship between the variables. The trade-off between the economic growth of Turkey and environmental degradation through consumption in the tourism and energy sectors is of great importance in terms of addressing the issue statistically and recommending economic policies to decision-makers. Therefore, with evidence, this research proposes an essential energy use by considering energy diversification including environmentally sensitive tourism activities, and this will shed light on future research on the necessity of cleaner tourism activities and utilization of renewable energy resources for mitigating global warming through climate change. Furthermore, this study could serve as a statistical model for other countries.

According to the NARDL test results, while positive and negative shocks contribute to the decrease in temperature and precipitation in REC in the long-run, they affect the increase in temperature and precipitation in NREC. In addition, a short-run analysis of each variable is included in the analysis, which will be detailly explained in the following sections. The rest of the paper is structured as follows: literature is given in “Literature review” section. “Data and methodology” section presents corresponding data and methodology. “Empirical results and discussion” section shows the results and discussion. Finally, conclusions and policy recommandations are presented in “Conclusions and policy implications” section.

Literature review

International tourism has been identified as one of the most energy-demanding and fossil fuel-dependent industries (Gössling 2013; Nepal 2008), and this information is also valid for Turkey. A tourist in international tourism travels to a destination via transportation (air, land, sea, and railways) and participates in activities such as lodging, food and beverage, entertainment, sports, and various local excursions. These activities necessitate the use of a variety of energy sources (Becken and Simmons 2005). There are many studies investigating the relationships between tourism and energy consumption in literature in recent years (i.e., Işik et al. 2017; Gokmenoglu and Eren 2020; Khan and Hou 2021; Khan et al. 2020b). These studies attempt to cover one- and two-way causal relationships between energy consumption and tourist arrivals, tourism income, and tourist expenditures. Ben Jebli and Hadhri (2018) discovered a bi-directional causality between international tourism and energy consumption. Tourism is a labor-intensive and active industry that operates 24 hours a day, 7 days a week. The use of equipment for heating, cooling, lighting, cleaning, and kitchen services, as well as the special energy consumption of tourists, increases energy consumption in this sector, which employs many people and attracts millions of visitors.

When it is searched in detail, Pablo-Romero et al. (2019) investigated the relationship between electricity consumption and accommodation for the regions on the Mediterranean coast of Spain in the period between 1999 and 2014 using panel data techniques, and a positive relationship is found between these variables. Increasing energy costs also force accommodation businesses to consume more convenient and more efficient energy. In this context, an increasing number of hotels are being designed to benefit more from switching to solar energy systems. Işik et al. (2017) discovered, using the Granger causality test, that there is a one-way causality between tourist arrivals, tourism revenues, and energy consumption in Turkey. Another reference (Aslan et al. 2021), which examined the relationships between energy consumption, tourist arrivals, CO2 emissions, and economic growth for several Mediterranean countries from 1995 to 2014, discovered uni-directional causality between energy consumption and tourist arrivals. Işik et al. (2017) investigated a bi-directional causality relationship between energy consumption and tourist arrivals in the top ten tourism countries around the world. Tang et al. (2016) investigated the relationship between economic development, energy consumption, and the tourism sector in India from 1971 to 2012, and they discovered a strong positive relationship between energy reduction and the tourism sector over the long run.

Countries that prioritize economic growth cause an increase in CO2 emissions, which contribute to global warming. Many studies have found a link between economic growth, CO2 emissions, and global warming, and with the emergence of environmental events, current growth strategies have begun to be debated (Alola and Kirikkaleli 2021). Tourism development causes not only an increase in energy consumption but also significant increases in CO2 emissions (Katircioglu 2014). International tourism causes about 8% of GHG emissions (Lenzen et al. 2018). There are many studies investigating the effects of tourism on CO2 emissions. Some examples include Akadiri et al. (2019), Amin et al. (2020), Ben Jebli et al. (2019), Chen et al. (2018), Dogan et al. (2017), Liu et al. (2019), Paramati et al. (2016), Rahaman et al. (2022), Wei and Ullah (2022), and Zhang and Zhang (2020). Studies examining the impact of tourism on CO2 emissions (Liu et al. 2019), emphasize that the carbon emission elasticity of tourist arrivals in developed countries is higher than in developing countries. Furthermore, a 1% increase in tourist arrivals increased carbon emissions by 0.345% in developed economies and 0.0467% in emerging economies. According to Balli et al. (2019), a 1% increase in tourism revenues per capita is equivalent to an increase in CO2 emissions of 0.09% to 0.14% per capita; thus, they concluded that increases in tourism revenues cause higher CO2 levels. Wei and Ullah (2022) compared the relationship between the international tourism sector and CO2 emissions for Asian economies from 1996 to 2019, using DOLS, FMOLS, and quantile regression techniques, and found a negative correlation between tourist arrivals and CO2 emissions using the FMOLS and DOLS estimators.

There are studies in the literature that investigate the relationship between tourism and CO2 emissions using the environmental Kuznets curve (EKC) hypothesis. According to the EKC hypothesis, because production is increased during the early stages of economic growth, the costs of growth are ignored, thereby, negative environmental externalities are ignored. Ozturk et al. (2016) investigated the EKC hypothesis, using the ecological footprint and tourism as environmental indicators. In this context, they developed a model of environmental degradation for 144 countries from 1988 to 2008. The time-series generalized method of moments (GMM) results revealed a negative relationship between the ecological footprint and its determinants, which was found to be greater in upper-middle-income and high-income countries.

The dynamic relationships between tourism, economic growth, and CO2 emissions in developed and developing economies were empirically examined by Paramati et al. (2016). They discovered that the impact of tourism on CO2 emissions decreases much faster in developed economies than in emerging economies using panel data econometric techniques, and the EKC hypothesis provides a link between tourism, growth, and CO2 emissions. Studies on the relationship between tourism and CO2 emissions in the Turkish economy have also been conducted. One of the studies examining the relationship between tourism and CO2 emissions (Yorucu 2016) stated that rapid tourism development in Turkey threatens sustainability. The conditional error correction model used in their study confirms the hypothesis that the growth of CO2 emissions during the 1960–2010 period had dynamic relationships between electricity consumption and tourist arrivals. Eyuboglu and Uzar (2020) investigated the relationships between CO2 emissions, tourist arrivals, energy consumption, and economic growth in Turkey from 1960 to 2014. They used Bayer and Hanck Fourier Autoregressive Distributive Lag (ADL) and Autoregressive Distributed Lag (ARDL) cointegration tests in the study to examine the long-run relationship between the variables. Their findings indicated that tourism, economic growth, and energy consumption increase CO2 emissions both in the long-run and short-run. Their discovery demonstrates that tourists are taking action by paying attention to the environmental quality of the country to which they travel. Khan et al. (2020a) investigated the causal relationship between tourism, energy consumption, economic growth, and CO2 emissions for Pakistan. Their findings suggest that tourism activities, through economic growth, have a positive impact on energy consumption, which increases CO2 emissions.

On the other hand, the relationship between climate change and tourism is quite sophisticated and quite complex. Tourism and climate are two variables that influence each other in both directions (Ma and Kirilenko 2019). As with all human-induced activities, greenhouse gases are released into the atmosphere because of tourism activities. These gases cause more warming and sometimes cooling in the atmosphere. The warming atmosphere changes the climate by causing the world to warm, causing extreme weather events, and raising sea levels. More importantly, climate change has a large-scale impact on the tourism system and interacts with other large-scale factors (Scott 2021). While tourism is affected by climate change, with effects such as heat waves, unreasonable cold, drought, storms, and heavy rains; tourist comfort and safety (and thus satisfaction) are also affected, as are the products that attract tourists (i.e., snow cover, coral reefs, wildlife) (Scott and Lemieux 2010).

Several studies have been conducted on the potential impacts of climate change on the tourism sector (e.g., winter tourism, coastal tourism, and so on) and on mitigating these effects (see, for example, Arabadzhyan et al. 2021; Azam et al. 2018; Becken et al. 2020; El-Masry et al. 2022; Gössling et al. 2012; Lopes et al. 2022; Neuvonen et al. 2015; Raza et al. 2017; Rutty et al. 2017; Schliephack and Dickinson 2017; Scott 2021); nonetheless, research on the impact of tourism on climate change is very limited (see, for example, Azam et al. 2018; Ben Jebli and Hadhri 2018; Danish and Wang 2018; Katircioglu 2014; Paramati et al. 2016; Zhang and Zhang 2020). Katircioglu (2014) argued in their research in Cyprus that the tourism sector has a statistically significant long-run effect on CO2 emissions, which is evidence that tourism development is a determinant of climate change through escalations in energy consumption. Adebayo et al. (2023) show that globalization, tourist arrivals, economic growth, and energy consumption increase environmental degradation in different quantiles in Thailand. Quantitative causality results also reveal evidence of causality in the mean and variance of most of the quantities, from globalization, tourist arrivals, economic growth, and energy consumption to CO2 emissions. However, these findings were put forward by interpreting the CO2 emissions. This gap in the literature has also been the most important driving force for this research. Table 1 shows the literature review from the perspective of the country and sample, along with the main findings that are thought to be related to this study.

Table 1 Sample of empirical studies of climate change, the tourism sector, CO2 emissions, energy consumption, and economic growth between the years of 2017 and 2022 in the hospitality literature

Through energy consumption, natural resource use, and sustainable production and consumption patterns, tourism has the potential to be effective in climate change mitigation policies. The most significant of these is the reduction of greenhouse gases emissions caused by energy consumption during transportation and lodging activities for tourism purposes, as well as in other sectors. Turkey hosts more than 50 million tourists a year, and with the increasing number of tourists, its energy needs and, accordingly, CO2 emissions are also increasing. Since Turkey is foreign-dependent on energy and the consumption of renewable energy has not yet developed sufficiently, energy utilization is one of the most discussed issues economically, and alternatives should be produced. The consumption of fossil fuels in Turkey can also have negative consequences for the environment. In tourism, energy consumption, especially from transportation vehicles and the accommodation sector (see Khan et al. 2021) for the vulnerability of the tourism sector due to COVID-19 pandemic effects), contributes negatively to carbon emissions. Because climate and the environment are two of the tourism industry’s most important supply sources, it is critical to discuss and expose the negative effects of climate change and the environment, which are the sector’s most important inputs. Studies have specifically revealed the impacts of the tourism sector on CO2 emissions and environmental quality, as well as the effects of climate change on tourism, in the context of the aforementioned literature. However, research on the impact of tourism on climate change is limited and needs to be expanded. As a result, the purpose of this paper is to shed light on the effects of tourism sector on climate change via economic growth and energy consumption, to draw attention to the need for low-carbon and renewable energy consumption, and to make policy recommendations.

Data and methodology


In this study, the relationships between climate change, tourism, renewable and non-renewable energy consumption, and economic growth are investigated using annual data from 1995 to 2020. In the analysis, two variables representing climate change are used as proxies and as dependent variables. The TEMP variable represents Turkey’s maximum temperature in degrees Celsius (°C), and the PREC variable represents Turkey’s total precipitation in millimeters (mm), both of which are obtained from the Turkish State Meteorological Service. Two variables are used in the analysis for energy data, which is one of the independent variables. These are non-renewable energy consumption (NREC) and renewable energy consumption (REC) figures. These two variables are measured in billion-kilowatt hours (b. kWh) and are sourced from the “International Energy Agency.” The gross domestic product (GDP) variable is the current US$ economic growth data variable and from the “World Development Indicators” and is one of the independent variables. On the other hand, the variable used in the analysis, ARRIV, shows the number of tourists traveling for a period not exceeding 12 months, as expressed in the “World Bank” database as the “international tourism number of arrivals.” The calculations include all of the data used in the analysis by taking their logarithms. Descriptive statistics of the used data are presented in Table 2.

Table 2 Descriptive statistics of variables


In the literature of economics, methods such as ordinary least squares, quantile regression, and Granger causality are used to estimate symmetric relationships, and these methods do not show possible asymmetry. For this reason, the NARDL method, that is, non-linear ARDL cointegration approach, has been applied in many studies in recent years, and this method has advantages such as being suitable for small samples and producing short- and long-run coefficient estimates (Jareño et al. 2021). On the other hand, linear models have disadvantages such as not separating positive and negative shocks and not being used to test the effects of asymmetric uncertainty (Liang et al. 2020). For these reasons, the NARDL method is used in this study. The impacts of tourism sector, energy consumption, and economic growth on climate change are investigated with the NARDL model introduced by Shin et al. (2014). As in Brini (2021), but currently using time-series (see Acaroğlu and Güllü 2022 for a similar ARDL approach) technique instead of panel data, the following equations (Eqs. (1) and (2)) are used to measure climate change in two different models: Model 1, where temperature (TEMP) is the dependent variable, and Model 2, where precipitation (PREC) is the dependent variable.

$${{\mathrm{LTEMP}}_{it}= \beta }_{0}+ {\beta }_{1}{\mathrm{LARRIV}}_{it}+{\beta }_{2}{\mathrm{LREC}}_{it}+ {\beta }_{3}{\mathrm{LNREC}}_{it}+ {\beta }_{4}{\mathrm{LGDP}}_{it}+\varepsilon$$
$${{\mathrm{LPREC}}_{it}= \beta }_{0}+ {\beta }_{1}{\mathrm{LARRIV}}_{it}+{\beta }_{2}{\mathrm{LREC}}_{it}+ {\beta }_{3}{\mathrm{LNREC}}_{it}+ {\beta }_{4}{\mathrm{LGDP}}_{it}+\varepsilon$$

where LARRIVit represents tourist arrivals; LNRECit and LRECit represent non-renewable energy consumption and renewable energy consumption, respectively, and LGDPit represents GDP (i.e., all the variables are included in the natural logarithm form). Before examining the asymmetric cointegration relationship between the variables, the order of integration of the variables in the model should be investigated. For this purpose, three different unit root tests are applied. These tests are the Dickey-Fuller GLS test (DF-GLS), the Phillips and Perron unit root test (PP) developed by Peter and Perron (1988), and the Zivot and Andrews structural break unit root test (ZA) developed by Zivot and Andrews (1992). Then, to investigate the short- and long-run relationships of climate change between the tourism sector, NREC, REC, and GDP by breaking it down into negative and positive components, Shin et al.’s (2014) NARDL cointegration approach is used. The asymmetrical relationships between the variables are discovered in the references of Shahbaz et al. (2021), Udemba and Yalçıntaş (2021), and Koondhar et al. (2021).

$$\begin{array}{c}\Delta {\mathrm{TEMP}}_{t}={\beta }_{0}+{\beta }_{1}{\mathrm{TEMP}}_{t-1}+{\beta }_{2}^{+}{\mathrm{REC}}_{t-1}^{+}+{\beta }_{3}^{-}{\mathrm{REC}}_{t-1}^{-}+{\beta }_{4}^{+}N{\mathrm{REC}}_{t-1}^{+}+{\beta }_{5}^{-}{\mathrm{NREC}}_{t-1}^{-}\\ +{\beta }_{6}^{+}{\mathrm{ARRIV}}_{t-1}^{+}+{\beta }_{7}^{-}{\mathrm{ARRIV}}_{t-1}^{-}+{\beta }_{8}^{+}{\mathrm{GDP}}_{t-1}^{+}+{\beta }_{9}^{-}{\mathrm{GDP}}_{t-1}^{-}\\ \begin{array}{c}+\sum_{i=0}^{\rho }{\alpha }_{1}\Delta {\mathrm{TEMP}}_{t-i}+\sum_{i=0}^{\rho }{\alpha }_{2}\Delta {\mathrm{REC}}_{t-i}^{+}+\sum_{i=0}^{\rho }{\alpha }_{3}\Delta {\mathrm{REC}}_{t-i}^{-}+\sum_{i=0}^{\rho }{\alpha }_{4}\Delta {\mathrm{NREC}}_{t-i}^{+}\\ \begin{array}{c}+\sum_{i=0}^{\rho }{\alpha }_{5}\Delta {\mathrm{NREC}}_{t-i}^{-}+\sum_{i=0}^{\rho }{\alpha }_{6}\Delta {\mathrm{ARRIV}}_{t-i}^{+}+\sum_{i=0}^{\rho }{\alpha }_{7}\Delta {\mathrm{ARRIV}}_{t-i}^{-}+\sum_{i=0}^{\rho }{\alpha }_{8}\Delta {\mathrm{GDP}}_{t-i}^{+}\\ +\sum_{i=0}^{\rho }{\alpha }_{9}\Delta {\mathrm{GDP}}_{t-i}^{-}+{\mu }_{t}\end{array}\end{array}\end{array}$$
$$\begin{array}{c}\Delta {\mathrm{PREC}}_{t}={\beta }_{0}+{\beta }_{1}{\mathrm{PREC}}_{t-1}+{\beta }_{2}^{+}{\mathrm{REC}}_{t-1}^{+}+{\beta }_{3}^{-}{\mathrm{REC}}_{t-1}^{-}+{\beta }_{4}^{+}{\mathrm{NREC}}_{t-1}^{+}+{\beta }_{5}^{-}{\mathrm{NREC}}_{t-1}^{-}\\ +{\beta }_{6}^{+}{\mathrm{ARRIV}}_{t-1}^{+}+{\beta }_{7}^{-}{\mathrm{ARRIV}}_{t-1}^{-}+{\beta }_{8}^{+}{\mathrm{GDP}}_{t-1}^{+}+{\beta }_{9}^{-}{\mathrm{GDP}}_{t-1}^{-}\\ \begin{array}{c}+\sum_{i=0}^{\rho }{\alpha }_{1}\Delta {\mathrm{PREC}}_{t-i}+\sum_{i=0}^{\rho }{\alpha }_{2}\Delta {\mathrm{REC}}_{t-i}^{+}+\sum_{i=0}^{\rho }{\alpha }_{3}\Delta {\mathrm{REC}}_{t-i}^{-}+\sum_{i=0}^{\rho }{\alpha }_{4}\Delta {\mathrm{NREC}}_{t-i}^{+}\\ \begin{array}{c}+\sum_{i=0}^{\rho }{\alpha }_{5}\Delta {\mathrm{NREC}}_{t-i}^{-}+\sum_{i=0}^{\rho }{\alpha }_{6}\Delta {\mathrm{ARRIV}}_{t-i}^{+}+\sum_{i=0}^{\rho }{\alpha }_{7}\Delta {\mathrm{ARRIV}}_{t-i}^{-}+\sum_{i=0}^{\rho }{\alpha }_{8}\Delta {\mathrm{GDP}}_{t-i}^{+}\\ +\sum_{i=0}^{\rho }{\alpha }_{9}\Delta {\mathrm{GDP}}_{t-i}^{-}+{\mu }_{t}\end{array}\end{array}\end{array}$$

In Eqs. (3) and (4) (where i = 1 … …9), βi represents the long-run coefficients, and αi represents the short-run coefficients. The effect of the shock of the independent variables, which are divided into two negative and positive categories, on the dependent variable is expressed with ( +) and ( −) signs. In Eq. (5), the decomposition of shocks of independent variables (explanatory variables) (i.e., as xt) represented by ARRIV, REC, NREC, and GDP into positive and negative effects is given.

$${x}_{t}^{+}=\sum_{j=1}^{t}\Delta {x}_{j}^{+}=\sum_{j=1}^{t}\mathrm{max}\left(\Delta {x}_{j},0\right), {x}_{t}^{-}=\sum_{j=1}^{t}\Delta {x}_{j}^{-}=\sum_{j=1}^{t}\mathrm{min}\left(\Delta {x}_{j},0\right),$$

Following this, the Wald test is used to check the variables’ (i.e., \({(\beta }_{LR}:{\beta }^{+}={\beta }^{-})\) and \({(\alpha }_{SR}:{\alpha }^{+}={\alpha }^{-})\)) long- and short-run asymmetries. Here, long-run symmetry is shown with \({\beta }_{LR}\) and short-run symmetric dynamics is shown with \({\alpha }_{SR}\). The bounds test with the F stats and t stats, as well as the cointegration test, are then used by Pesaran et al. (2001). The null hypothesis and alternative hypothesis are expressed as \(\left(\beta ={\beta }^{+}={\beta }^{-}=0 \right)\) and \((\beta ={\beta }^{+}={\beta }^{-}\ne 0)\), respectively, in the bounds cointegration test for the F statistics, while the null hypothesis \((\beta =0)\) (against the alternative \((\beta \ne 0\))) indicates that there is no cointegration. The F statistic is used to test the hypotheses, and the decision is made by comparing the F statistic with the asymptotic table values. If the test results are less than the lower limit of the table values, there is no cointegration; if the test results are greater than the upper limit of the table values, the null hypothesis \((\beta =0)\) is rejected, and it is concluded that there is cointegration. Except for these two cases, if the test statistic is between the lower and upper limits of the table values, then it is in the region of instability, and it cannot be decided whether there is a cointegration relationship (Pesaran et al. 2001). If \(h\to \infty\), \({m}_{h}^{+}\to {\beta }^{+}, {m}_{h}^{-}\to {\beta }^{-}\), the NARDL model employs asymmetric coefficient calculations as a positive coefficient \({\beta }^{+}=-{\theta }^{+}/\partial\) and as a negative coefficient \({\beta }^{-}=-{\theta }^{-}/\partial\). These coefficients assess the relationship between dependent and independent variables in order to decompose the independent variables’ positive and negative shocks. Following determination of the NARDL model’s short- and long-run relationships, the asymmetrical responses of TEMP and PREC to negative and positive shocks in ARRIV, NREC, REC, and GDP are evaluated using the asymmetric dynamic multiplier approach and are calculated by Eqs. (6) and (7).

$$\begin{array}{c}{m}_{h}^{+}=\sum_{j=0}^{h}\frac{\partial {\mathrm{TEMP}}_{t+j}}{\partial {\mathrm{REC}}_{t}^{+}}, {m}_{h}^{-}=\sum_{j=0}^{h}\frac{\partial {\mathrm{TEMP}}_{t+j}}{\partial {\mathrm{REC}}_{t}^{-}} , {m}_{h}^{+}=\sum_{j=0}^{h}\frac{\partial {\mathrm{TEMP}}_{t+j}}{{\partial \mathrm{NREC}}_{t}^{+}},\\ {m}_{h}^{-}=\sum_{j=0}^{h}\frac{\partial {\mathrm{TEMP}}_{t+j}}{{\partial \mathrm{NREC}}_{t}^{-}}, {m}_{h}^{+}=\sum_{j=0}^{h}\frac{\partial {\mathrm{TEMP}}_{t+j}}{{\partial \mathrm{ARRIV}}_{t}^{+}}, {m}_{h}^{-}=\sum_{j=0}^{h}\frac{\partial {\mathrm{TEMP}}_{t+j}}{\partial {\mathrm{ARRIV}}_{t}^{-}},\\ {m}_{h}^{+}=\sum_{j=0}^{h}\frac{\partial {\mathrm{TEMP}}_{t+j}}{{\partial \mathrm{GDP}}_{t}^{+}}, {m}_{h}^{-}=\sum_{j=0}^{h}\frac{\partial {\mathrm{TEMP}}_{t+j}}{\partial {\mathrm{GDP}}_{t}^{-}}, \mathrm{for}\;h=0, 1, 2,\dots \end{array}$$
$$\begin{array}{c}{m}_{h}^{+}=\sum_{j=0}^{h}\frac{\partial {\mathrm{PREC}}_{t+j}}{\partial {\mathrm{REC}}_{t}^{+}}, {m}_{h}^{-}=\sum_{j=0}^{h}\frac{\partial {\mathrm{PREC}}_{t+j}}{\partial {\mathrm{REC}}_{t}^{-}} , {m}_{h}^{+}=\sum_{j=0}^{h}\frac{\partial {\mathrm{PREC}}_{t+j}}{\partial {\mathrm{NREC}}_{t}^{+}},\\ {m}_{h}^{-}=\sum_{j=0}^{h}\frac{\partial {\mathrm{PREC}}_{t+j}}{{\partial \mathrm{NREC}}_{t}^{-}}, {m}_{h}^{+}=\sum_{j=0}^{h}\frac{\partial {\mathrm{PREC}}_{t+j}}{{\partial ARRIV}_{t}^{+}}, {m}_{h}^{-}=\sum_{j=0}^{h}\frac{\partial {\mathrm{PREC}}_{t+j}}{{\partial \mathrm{ARRIV}}_{t}^{-}},\\ {m}_{h}^{+}=\sum_{j=0}^{h}\frac{\partial {\mathrm{PREC}}_{t+j}}{{\partial \mathrm{GDP}}_{t}^{+}}, {m}_{h}^{-}=\sum_{j=0}^{h}\frac{\partial {\mathrm{PREC}}_{t+j}}{{\partial \mathrm{GDP}}_{t}^{-}}, \mathrm{for}\;h=0, 1, 2,\dots \end{array}$$

Diagnostic tests are used in the final stage of the study to assess the robustness of the NARDL model’s short- and long-run relationship analysis. The CUSUM and CUSUM of square tests are used for the stability of the model; the Breusch-Godfrey serial correlation LM test is used for the serial correlation; the Jarque–Bera test is used for the normality test; the Breusch-Pagan-Godfrey test is used for heteroskedasticity; and the Ramsey RESET test is used to examine whether there is a regression specification error.

Empirical results and discussion

Empirical results

Since the methodology of this study is based on the ARDL model and checking whether the variables are I(0) or I(1), and the test is not valid in the case of I(2), it is first needed to check the integration properties of the variables. Three different unit root tests are applied to examine whether the variables are stationary. Thereby, PP, ZA, and DF-GLS unit root test (e.g., see Kocaarslan et al. 2020) results are presented in Table 3 .

Table 3 Unit root test results

Values in parentheses are probability (Prob) values. At 1% level, 5% level, and 10% level, Phillips and Perron test critical values are − 4.39, − 3.61, and − 3.24, respectively. In Zivot and Andrews test, the variables are examined by unit root with a structural break through both the intercept and trend. The Zivot-Andrews test statistic values for 1% critical value is − 5.34; for 5% critical value is − 4.93; and for 10% critical value is − 4.58. Dickey-Fuller GLS detrended (DF-GLS) test critical values at 1% level, 5% level, and 10% level, are − 2.66, − 1.95, and − 1.60, respectively.

At the level as it is shown in Table 3, all the variables are stationary, and none of the variables are I(2). The NARDL model continued to be predicted after the PP, ZA, and DF-GLS unit root tests confirmed that none of the variables were integrated at the second order. First, the existence of a non-linear long-run relationship between the first model (temperature) and then the second model (precipitation) and tourism, energy, and growth was tested using the F test proposed by Pesaran et al. (2001). Table 4 shows the results for Model 1 with TEMP as the dependent variable, and Table 5 shows the results for Model 2 with PREC as the dependent variable.

Table 4 Model 1 NARDL results
Table 5 Model 2 NARDL results

According to the findings of Table 4, positive shocks in the long-run REC aid in temperature reduction. At a 10% significance level, a 1% increase in REC causes a 0.15% decrease in maximum temperature over time (in the long run). A 1% decrease in REC causes a 0.19% decrease in maximum temperature over time (in the long run). A 1% increase in NREC causes a 0.36% increase in maximum temperature in the long-run relationship between NREC and temperature. In the long run, a 1% decrease in NREC causes a 1.08% increase in maximum temperature. A 1% decrease in tourist arrivals causes a 0.13% decrease in maximum temperature in the long-run relationship between the tourism sector and temperature. In the short run, temperature tends to increase against the positive and negative shocks of REC at the 5% significance level. A 1% increase in REC in the short run causes an increase of 0.05% in maximum temperature. A 1% decrease in REC in the short run causes an increase of 0.12% in maximum temperature.

According to Table 5, the long-run negative and positive shocks aid REC in reducing precipitation. A 1% increase in REC results in a 0.57% decrease in precipitation over time (in the long-run). A 1% decrease in REC over time (in the long-run) results in a 0.86% decrease in precipitation. According to the relationship between NREC and precipitation, a 1% increase in NREC results in a 0.77% increase in precipitation over time (in the long-run). In the long-run, a 1% decrease in NREC results in a 3.18% increase in precipitation. In the long-run, according to the tourism-precipitation relationship, a 1% decrease in tourist arrivals results in a 1.17% decrease in precipitation at the 5% significance level.

In the relationship between economic growth and precipitation for the long-run, a 1% decrease in economic growth results in a 1.05% increase in precipitation. In the short-run, a 1% increase in REC results in a 0.19% increase in precipitation, a 1% increase in NREC results in a 2.28% increase in precipitation, and a 1% decrease in tourist arrivals causes a 0.95% decrease in precipitation. At this stage, the F statistic is used to test the null hypothesis of (\({\beta }^{+}={\beta }^{-}=0)\) non-linear cointegration by first applying the bounds testing procedure (Pesaran et al. 2001). Then, the Wald test is used to test the short- and long-run symmetries (Shin et al. 2014). The null hypothesis of the long-run symmetry (\({\beta }^{+}={\beta }^{-}=0\) with \({\beta }^{+}=-{\theta }^{+}/\partial\) and \({\beta }^{-}=-{\theta }^{-}/\partial\)) was tested to investigate the existence of long-run non-linearities, and the results are shown in Table 6.

Table 6 Model 1 and model 2 ARDL bounds test estimation results

For NARDL models, the critical values for the bounds testing procedure are 1.85,–2.11,–2.62 and 2.85,–3.15,–3.77 for the 10%, 5%, and 1% significance levels, respectively. For TEMP, REC, NREC, ARRIV, and GDP variables, TEMP refers to maximum temperature, PREC total precipitation, REC renewable energy consumption, NREC non-renewable energy consumption, ARRIV international tourism arrivals, and GDP economic growth, respectively. In addition, “ + ” superscript denotes positive partial sum and “ − ” superscript denotes negative partial sum.

According to the ARDL bounds test estimates, which are shown in Table 6, the F statistics values are 8.658 for Model 1 and 17.678 for Model 2, which are greater than the upper limit I(1) value. Thereby, it is concluded that there is a long-run cointegration relationship between the variables at the 5% significance level in both models.

Finally, diagnostic test results are included in the study with Table 7 as follows: according to the results of diagnostic test, both models are serially independent; there is no misspecification error; they are normally distributed; and there is no heteroskedasticity problem. Furthermore, according to the results of the CUSUM and CUSUM of square test for both models, the models are found to be stable. These test results can be found in Fig. 1.

Table 7 Diagnostic test results
Fig. 1
figure 1

The cumulative sum of recursive residuals (CUSUM) and CUSUM of square test results


Climate change is one of the serious threats to humanity today, along with global warming. Increasing energy consumption and tourism activities as a result of population growth compel researchers to investigate the impact of environmental factors on climate change from all angles. Society’s production and consumption behaviors are heavily reliant on energy consumption, but the impacts of renewable energy consumption and non-renewable energy consumption can have different environmental consequences. However, depending on the type of energy consumed, the tourism sector can have significant environmental impacts by increasing carbon emissions and contributing to climate change through temperature and precipitation. In this context, by concentrating on climate change in Turkey, the discussion is based on the long- and short-run dynamics by using a non-linear ARDL causality approach for production and consumption policy planning.

Many different long- and short-run results have been obtained using Models 1 and 2. In the long-run, the finding that a decrease in tourist arrivals and an increase in renewable energy consumption will reduce the temperature and precipitation, while an increase in non-renewable energy consumption will raise temperature and precipitation, provides an important conclusion for mitigating global warming. This finding coincides with the findings of Canales et al. (2020) that consuming renewable energy sources has a fundamental role in reducing climate change. Turkey should continue for sustainable economic growth in the cleanest possible energy sector and green environment. On the other hand, it has been observed that consuming renewable energy sources is not yet effective enough to reduce temperature and precipitation in the short-run in developing economies and in Turkey. Anyway, as stated in Kirikkaleli and Sowah (2021), policy-makers should encourage the utilization of renewable energy sources that will hold the increase in global average temperature below 1.5 °C. A limitation of this research is that it concentrates on a country; however, environmental degradation is a global concern and Adebayo et al. (2021b) can be given as an interesting study conducting a panel data causality analysis on the relationship between Latin American economic growth, energy consumption, and CO2 emissions.

In line with this research paper, Adebayo et al. (2023) conducts a quantile-on-quantile regression with time-series data to investigate the non-linear relationship between carbon emissions and globalization, tourist arrivals, economic growth, and energy consumption. They found similar evidence of causality in this paper, but this time for Thailand, and showed that independent variables increase environmental degradation at different quantiles. Another reference, Alqaralleh (2020) investigated the non-linear effect of economic growth and energy consumption on environmental pollution (i.e., CO2 emissions) with a panel smooth regression model for 30 countries, validating the EKC hypothesis and emphasizing the importance of greener energy utilization in combating climate change in the context of sustainable economic growth. On the other hand, the finding that the decrease in the number of tourist arrivals will reduce the temperature and precipitation is also similar to Yorucu’s (2016) finding tourist arrivals in Turkey, which increased CO2 emissions. However, the tourism is a very important sector for the country’s economies, and reducing the number of tourists is not an acceptable solution as it will cause a decrease in the tourism income in many countries. Efforts to limit the number of tourists, together with tourism carrying capacity policies, may have a positive effect on the subject.

However, there are studies showing that carbon emissions from the tourism sector can be reduced by increasing the use of renewable energy in the sector and increasing energy-saving practices in the sector. A time-series analysis using the ARDL model uncovers evidence of one-way causality from energy consumption to GDP as well as the long-run linkage between variables, validating the energy-induced growth hypothesis in South Korea (Adebayo et al. 2021a). Furthermore, this paper discovers a uni-directional causality between CO2 emissions, energy consumption, and economic growth. According to Alola and Kirikkaleli (2021), CO2 emissions caused higher average temperatures using a global approach that employs the wavelet coherence technique and the Toda and Yamamoto test. Finally, Akinsola et al. (2022) examined the relationship between the ecological footprint and the determinants using time-series techniques (i.e., the ARDL and DOLS). Their findings showed that economic growth and energy investment increase environmental degradation in Brazil and there is bi-directional causality between the ecological footprint and economic growth. As a result, decision-makers should encourage green technological innovations in order to achieve long-run economic growth.

Conclusions and policy implications

Energy consumption-climate scenarios depend on socioeconomic factors (i.e., tourism activities in a growing economy), and policy design suggests that this can irreversibly have a non-linear effect on global climate conditions through carbon emissions, leading to global warming. In this context, this paper contributes to the growing literature of evidence on the effects of the tourism sector, non-renewable energy consumption (NREC), renewable energy consumption (REC), and economic growth on climate change in Turkey from 1995 to 2020. Two different models, temperature and precipitation as dependent variables, are considered and examined with the non-linear autoregressive distributed lag method using time-series data that is obtained from various institutions. The trade-off between the economic growth of countries and environmental degradation through consumption in the tourism and energy sectors is of great importance in terms of addressing the issue statistically and recommending economic policies to decision-makers. To this end, a research question is asked as follows: “Is it possible to find evidence of probable causality and asymmetrical relationships both in the short-run and long-run, including the directions for climate change through tourist arrivals and energy consumption in a growing economy?” Both models show a long-run cointegration relationship between the variables, and diagnostic test results suggest that the used econometric model is statistically correct and robust.

With the long-run results, it is first checked for temperature. The findings reveal that positive shocks of REC help to reduce temperature, and positive shocks of NREC have an adverse effect on temperature. On the other hand, a decrease in tourist arrivals causes a decrease in maximum temperature. Then, it is checked for precipitation. The findings reveal that the positive and negative shocks of REC help to reduce precipitation. Nonetheless, the positive and negative shocks of NREC increases precipitation. Moreover, a decrease in tourist arrivals leads to a decrease in precipitation, and a decrease in economic growth leads to an increase in precipitation. Comparably, with the short-run results for temperature, temperature tends to increase against the positive and negative shocks of REC. For precipitation, an increase in REC and NREC leads to an increase in precipitation, and a decrease in tourist arrivals causes a decrease in precipitation.

The findings of the study propose critical policy implications: according to the long-run findings between climate change and REC, it is understood that with the increase of REC, there are decreases in precipitation and temperature, and this is important evidence for policy-makers in mitigating global warming. Furthermore, as NREC increases, temperature and precipitation will rise in the long-run. However, increasing NREC causes an increase in temperature, and decreasing economic growth causes an increase in precipitation in Turkey. According to the short-run findings, a decrease in tourist arrivals causes a decrease in precipitation; an increase in REC causes an increase in temperature and precipitation; and an increase in NREC causes an increase in precipitation. Increased carbon emissions cause severe global climate change and the decrease in tourist arrivals is also interpreted as having a minor impact on carbon emissions and can aid in the return of precipitation rates to normal level. However, the short-run temperature increase associated with an increase in REC has led to the conclusion that REC cannot adequately affect global warming in the short-run. Effective short-run policy proposals can include encouraging and expanding renewable energy investments through feasible government subsidies and trying to reduce the utilization of fossil-based energies in the country. In addition, increasing the use of clean and efficient energy practices instead of reducing the number of tourist arrivals is one of the important steps to be taken to prevent the negative effects of climate change.

Furthermore, this study could serve as a statistical model for other developing countries attempting to mitigate global warming. In the special case of Turkey, reducing the trips that lead to tourism mobility and the emissions caused by these trips, as well as the carbon emissions caused by the food, hotel, and shopping activities of the tourists, will contribute to the reduction of environmental deterioration. It would be beneficial for governments and destination administrations to organize incentives and make inspections on energy saving and renewable energy use (especially carbon-free transport policies). The good news is that Turkey has a promising future with a heating and electricity sector that uses renewable energy sources. On the other hand, given that 75% of visitors to Turkey arrive by air, it is important to remember that tourism transportation causes Turkey to emit more carbon emissions. Because fossil fuels pollute the air and water and increase greenhouse gas emissions, they also contribute global warming. In recent years, the effects of climate change have been felt, particularly in terms of precipitation. There is a decrease in precipitation in some areas, but the most common effects in Turkey are an increase in intensity or prolongation of the rainy season. Nevertheless, the potential asymmetries in the short-run and long-run relationships with both linear and non-linear time-series techniques, the variables used, and the selected locations must be addressed in future works.