The link between urbanization and air pollution in Turkey: evidence from dynamic autoregressive distributed lag simulations

This study investigates the relationship between urbanization and air pollution in Turkey. Dynamic ARDL method was used for the period 1960–2015. According to the findings, there is a positive and statistically significant relationship between long-term urbanization and CO2. If urbanization increased by 1%, carbon emissions increased by 0.02%. There is a similar relationship between the shocks that will occur in population growth and CO2 emission in the long term. However, there is a negative and statistically insignificant relationship between the two variables. In the relationship between GDP and CO2, there is a positive relationship in the long term. GDP increase of 1% increases CO2 emissions by 0.11%. There is a similar relationship between long-term GDP shocks and CO2 emissions. According to short-term analysis results, energy consumption increases CO2 emissions by the same rate as GDP. However, the astonishing result of the study emerges here. Empirical results show that a long-term positive shock in energy consumption reduces CO2 emissions and a negative shock increases pollution. According to these results, Turkey has not reached the point of sustainable growth. For this reason, this developing country needs to make regulatory implementations and determine future policies for these impacts affecting air pollution.


Introduction
A 2018 report by the World Health Organization World Health Organization (WHO) (2018) has shared stunning statistics on air pollution. According to this, 9 out of 10 people in the world, or 90% of the world's population, breathe contaminated air. 7 million people die every year in the world due to air pollution. The same report once again highlighted that air pollution is one of the most serious problems facing the world and one of the most threatening factors in human life. Air pollution occurred at the end of a long process. Air pollution has increased as a result of increased human needs in developing economies. The increasing population brings with urbanization, energy demand, and large infrastructure constructions (Rafiq et al. 2016). The United Nations Population Fund United Nations Population Fund (UNFPA) (2019) report states that the total population of world reached 7.8 billion in 2019. From this point of view, it is seen that population growth is an important factor in air pollution. More than 50% of the world population lives in cities. The right of people to move freely to a different place causes an increase in migration. Waves of migration from rural areas are often due to job opportunities in cities. Better health care in cities and higher wage sourcing are other important causes of migration. It is known that the number of vehicles has increased with the increase of urbanization. This causes emissions that cause air pollution. On the other hand, with increasing urbanization, the possibilities of infrastructure services are restricted and other damages such as water pollution and soil pollution are occurring. Increasing the need for fossil fuels, energy sources, soil, food, and water is the main harms of urbanization (Ali et al. 2019).
Turkey is located between developing countries, known as a country where there is intense migration from the countryside to the city. The inevitable consequence of urbanization is environmental problems. It is possible to express this nexus more clearly with the help of Fig. 1. As can be seen from the figure, there has been an increase in both urbanization and CO 2 emissions in Turkey in the last two decades. Although the focus of this figure is on current data, this nexus goes back to earlier times, and it seems that this dynamic process will continue in the future. The importance of this relationship continues to increase today and the nexus with current approaches needs to be addressed concretely. Therefore, the main aim of this study is to investigate the nexus in Turkey by using dynamic ARDL simulations. However, in line with related literature, economic growth and energy consumption are also considered as other main explanatory variables of pollution.
The relationship between environmental pollution and income started with the environmental Kuznets curve (EKC) theory proposed by Grossman and Krueger (1991). According to this theory, there is an inverse-u-shaped relationship between environmental pollution and income. And the other side, the relationship between urbanization and air pollution is named as environmental transition theory in the literature. According to this theory, with urbanization, industrialization increases and therefore energy demand increases. At the same time, with urbanization, the population in the cities will increase with the migration of rural areas. The static structure of the cities is challenged by this change. As a result, with the increasing population, sustainable growth comes to more rather than environmental pollution (Iheonu et al. 2021). There are studies investigating the effect of urbanization on energy demand and other environmental indicators in the literature. For example, Wang et al. (2016) studied for ASEAN countries, Wang et al. (2016b) for BRICS countries, Shahbaz et al. (2014) for the UAE, Dogan and Turkekul (2016) for the USA, and Al-Mulali et al. (2013) for MENA countries and found that urbanization is decisive for energy consumption and environmental pollution. Studies in the literature are usually based on a single-country time series or cross-country panel data studies. Very little work has been done for Turkey on urbanization and its effects. Topcu et al. (2016) is the only study that directly examines the relationship between urbanization and carbon emissions for Turkey by using VECM method. The study covers the period 1960-2011. They determined that the causality relationship is from urbanization to CO 2 emissions.
For Turkey, one of the most important tourism centers of recent years, the fact that the issue of air pollution is not associated with urbanization is a major deficiency in the literature. Filling this gap will be the big contribution of our study to the literature. The majority of the work was carried out for China. As stated in the literature, the models usually cover analysis of STIRPAT, panel data, or single-country time series. Methods such as ARDL or VECM are often used in single-country analysis. For this reason, dynamic ARDL method was used in our study. This current method is very useful in revealing the effect of negative and positive future shocks. We are aware that we will be pioneers in this regard, which is another gap in the literature. In terms of method and example examined, the aim of doing a study to fill the gap in two ways constitutes our motivation. The first part of the study is the introduction, second is the literature review, the third part contains methodology and data, empirical results is in the fourth part, and the final chapter contains the conclusion.

Literature review
CO 2 emissions are generally used as a measure of air pollution or environmental quality in the literature. The relationship between urbanization and air pollution generally turns into urbanization and CO 2 emissions. The focus of this section is  on current studies dealing with the relationship between urbanization and environmental quality. Accordingly, the relevant literature is handled in two parts as studies on the basis of city or region and studies on country basis. Admittedly, the relationship between urbanization and environmental quality in the related literature has been discussed by many studies both at the city and regional level for China. Zhang and Lin (2012) investigated the nexus among urbanization, energy consumption, and CO 2 emissions in China by using regional panel data analysis. Analysis results suggested that urbanization has a positive impact on emissions, but this effect is greater in the central region than in the eastern region. However, Li et al. (2018) studied China and they found a U-shaped relationship between urbanization and CO 2 emissions. Wang et al. (2019) considered geographical features while examining the link between urbanization quality and emissions. They analyzed three different regions of China: central, eastern, and western. According to the results of geographically weighted regression analysis, an increase in urbanization quality causes a decrease in emissions. Bai et al. (2019) studied 64 cities' data from China's urban agglomerations. They performed regression and analysis with the STIRPAT model for the period from 2006 to 2013. They found that urbanization has a positive impact on air pollution. Zhang et al. (2018) analyzed China on a regional level by using the STIRPAT model for the 2005-2014 data period. They focused on both population urbanization and land urbanization. According to their findings, the effect of population urbanization on CO 2 is insignificant, but land urbanization has a positive and significant impact on emissions. Liu and Liu (2019) analyzed the regional disparity, urbanization, and emissions in China. Accordingly, improvement in urbanization causes to decrease in emissions. Unlike the Chinese sample, Kavi Kumar, and S. and Viswanathan, B. (2013) studied two groups in India: urban and rural residents. They examined the impact of carbonintensive energy consumption on environmental pollution for the 2009-2010 data period. In their work, they reached an inverted-U-shaped relationship between income and pollution in the areas where the city is located. Cirilli and Veneri (2014) examined the relationship between transport and urbanization in the case of Italy. They concluded that transportation is one of the major causes of CO 2 emissions in large settlements. They also stated that demographic characteristics are an important factor in CO 2 emissions per capita.
Apart from the studies at the city or regional level, the number of studies dealing with this relationship at the country level is quite more. Poumanyvong and Kaneko (2010) analyzed the effect of urbanization on energy consumption on CO 2 emissions by using the panel dataset of 99 countries from different income groups over the period 1975-2005. They stated that urbanization has a positive impact on emissions in all income groups. Sharma (2011) used dynamic panel data analysis for high-, medium-, and low-income countries for the period from 1985 to 2005 and reached that urbanization harms CO 2 emissions in all country groups. Another study examining the relationship between urbanization and emissions by grouping countries according to their income levels is Martínez-Zarzoso (2008). Results suggested that there is an inverted-U-shaped relationship between them. Chikaraishi et al. (2015) used the STIRPAT model for 140 countries covering the 1980-2008 data period. Results illustrated that an increase in urbanization together with the increase in per capita income and the expansion of the service sector contributes to the environmental quality. Wang et al. (2020) analyzed APEC countries by using the dynamic unrelated seemly regression (DSUR) method for the period from 1990 to 2014. Their results suggested that urbanization, industrial development, and economic development cause air pollution. Zhang et al. (2017) conducted panel data analysis for 141 countries. They reached an inverted-U-shaped relationship between CO 2 and urbanization. Zhang et al. (2020) examined China by using decomposition analysis. They investigated how China can achieve its 2030 CO 2 reduction target. They argued that urbanization can be achieved through the increase in GDP, technological progress, and cultural development, and stated that non-fossil fuel consumption will increase with the increase in urbanization. Another study is Dong et al. (2019). They investigated the effect of urbanization and industrialization on CO 2 emissions in developed countries by using threshold autoregression analysis. Urbanization, GDP, industrial development, GDP per capita, and energy consumption are used as explanatory variables. According to their results, there is an insignificant relationship between CO 2 emissions and urbanization in the low-level urbanization group. However, in the mid-level urbanization group, there is a negative relationship between CO 2 emissions and urbanization. Also, they stated that the income level has critical importance in determining these threshold values. Lin et al. (2017) conducted a panel data study for 53 countries by using the STIRPAT model. Their analysis covered the period from 1991 to 2013. They found that urbanization and economic growth have a trivial effect on emissions in non-high-income countries. Liu and Bae (2018) analyzed the relationship between CO 2 emissions, urbanization, and energy density in China by using ARDL and VECM approaches. According to the results of long-run analysis, energy intensity, economic growth, industrialization, and urbanization have a positive impact on CO 2 emissions. Nathaniel et al. (2020) investigated the factors affecting air pollution in Indonesia by using the ARDL method for the 1971-2014 data period. Findings illustrated that urbanization, growth, and energy use harm environmental quality. Same method used by Joshua et al. (2020). They examined the relationship between FDI, GDP, urbanization, coal consumption, and CO 2 emissions in South Africa for the 1970-2017 data period. They reached that urbanization is an important cause of carbon emissions in this country. Iheonu et al. (2021) analyzed the causes of air pollution in Sub-Saharan African countries with panel quantile and causality methods for the period from 1990 to 2016. Their results demonstrated that GDP and urbanization increase CO 2 emissions, especially in low quantities, in all quantiles. Bashır et al. (2021) investigated the factors affecting air pollution in Indonesia. According to their findings, urbanization and energy consumption cause air pollution in the short run.
Turkey is an economy that must constantly fight air pollution due to its geographical location and is a great tourist country. Accordingly, studies conducted for Turkey in the relevant literature are important in line with the scope of this study. For example, Pata (2018a) analyzed the effect of urbanization and industrialization on CO 2 emissions by using the ARDL approach and reached that these variables have a positive impact on pollution. Also, Pata (2018b) analyzed the nexus among urbanization, financial development, income, renewable energy consumption, and CO 2 emissions with structural breaks. Findings confirmed the validity of the EKC hypothesis and the positive link between urbanization and CO 2 emissions. The same result was obtained by Cetin et al. (2018). They tested the relationship between urbanization and emission by considering structural breaks. Destek (2021) investigated the effect of growth, industrialization, urbanization, and human capital on environmental quality by using nonlinear ARDL approach. Analysis results confirmed that urbanization has a positive impact on environmental degradation.
Based on the summary of the literature review, it is understood that the relationship between urbanization and air pollution is heavily discussed. However, there are a relatively limited number of studies for Turkey. Moreover, it is noticed that this relationship is mostly tested with the ARDL approach. Hence, the dynamic ARDL simulations approach is adopted in the context of methodological innovation in this study. The DARDL econometric method used in this research is superior to classical ARDL and other time series applications due to both short-and long-term estimates and its modification. Therefore, it is aimed to fill an important gap in the literature in terms of the econometric method and country selection. In the light of these innovations, the hypothesis that urbanization is a determinant of pollution is tested.

Data, model specification, and testing procedure
Describes of variables, and model specification This paper focused on Turkey uses annual time series data for the period from 1960 to 2015. Analysis period selection is based on the data availability. Also, since the study is based on a time series analysis, the longest period is obtained. CO 2 emission (kt) is used as the dependent variable. The main independent variable is urban population growth (annual %) and it is a proxy for urbanization. Other explanatory variables are energy use (kg of oil equivalent per capita) and GDP (constant 2010 US$), respectively. All variables are obtained from the World Development Indicators database offered by the World Bank, and these are expressed in logarithms.
The aim of this paper is to explain the effects of urbanization (EC) on carbon dioxide emissions (CO 2 ). First, a basic production function accounting framework yields: where CO 2 denotes carbon dioxide emissions, URB is urban population growth, EC is energy consumption, and GDP displays gross domestic product (constant 2010 US$).
The study's empirical model is borrowed from recent literature (Sharma 2011;Ali et al. 2019;Ali et al. 2019), as follows: where βs are coefficients of dependent variables, μ is the error correction terms, and t denotes time series.

Methodology
In this study, it is used the dynamic ARDL model based on dynamic simulations as seen in Khan where y denotes a change in the dependent variable, α 0 is intercept, and t−1 displays independent variables' maximum level of p and with lags q k in the first differences operator with the error term ( ) in time t. The existence of the cointegration relationship is demonstrated by F-statistic value. With this value, the null hypothesis is tested, indicating that all parameters on the repressors appearing in levels, plus the coefficient on the lagged dependent variable are jointly equal to zero (H 0 = 0 + 1 + … + k = 0). The starting point of Jordan and Phillips (2018) is to eliminate the problems in the investigation of short and long-run relationships between variables that arise in the ARDL model. The dynamic ARDL method can estimate, stimulate, and plot to predict automatically spurious changes in the dependent variable that are due to a regressor while another factors constant Destek et al. (2018). In order to apply the method, the series must be stationary in the first order. This method is used up to 5000 simulations of the vector of parameters (Sarkodie et al. 2019).
Based on empirical specification expressed in Eq. (3), the error correction form of ARDL bounds procedure is demonstrated as:

Empirical findings
In the first step, the series must be proved to be stationary at the first order. Essentially, Jordan and Phillips (2018) suggested that condition I(1) is a necessity for the dependent variable and that at least, the other series should not be integrated higher than stationary at I(1). For this purpose, Phillips-Perron and DF-GLS unit root tests were performed such as Jordan and Phillips (2018). Both test results are presented in Table 1. The test results suggested that the null hypothesis expressing the presence of unit root is rejected in all series, so all variables are integrated in the first order. After providing the most important condition required by dynamic ARDL, the existence of cointegration relationship between series is tested. However, it is investigated whether the model has any econometric problem and autocorrelation problem is determined. Jordan and Phillips (2018) proposed the elimination of the problem in such a situation and the reestimation of the results. Accordingly, all results obtained are presented in two groups as before and after overcoming the problem. Firstly, the results of ARDL cointegration showed that F-statistic value in the case of autocorrelation is significant at 1% statistically significant level, whereas both Fstatistic and t-statistic values are above the critical values after the problem is resolved. Hence, the null hypothesis of no cointegration is rejected. The cointegration results are based on response surface regression with the accurate critical values and p-values approximation proposed by Kripfganz and Schneider (2018). Table 2 also contains dynamic stimulated ARDL coefficient estimation results. Jordan and Phillips (2018) proposed this method to eliminate the complexity of short and long-run coefficient estimation in the other ARDL method. In the case of autocorrelation problem, carbon dioxide emissions increase due to increases in energy consumption and GDP. However, the coefficient of urbanization, which we identified as the main explanatory variable, is positive in both the short and long run, but it is statistically insignificant. The econometric problem faced undermines the reliability of these results. Therefore, we adopted the coefficient estimation results in second column.
The coefficient of urbanization is found positive and statistically significant in long run, while this coefficient is negative but statistically insignificant in the short run. We find that 1% increase in urbanization enhances almost 0.02% carbon dioxide emissions (as seen in Poumanyvong and Kaneko 2010;Poumanyvong and Kaneko 2010;Bekhet and Othman 2017;Pata 2018a;Pata 2018b;Cetine t al., 2018;Ali et al. 2019;Wang et al. 2019;Bashır et al. 2021;Iheonu et al. 2021;Destek 2021). In other words, it is confirmed that urbanization is an important source of emissions in line with the common literature. This means that environmental degradation rises with the increase in the urban population growth in Turkey in the long run. This effect of urbanization was generally associated with the increasing industrialization process, energy consumption, and growth caused by urbanization in previous studies. Thus, this relationship can be attributed to the increase in the weight of industry and service sector in the economy and the relative decrease of the agricultural population. In addition, rapid production and consumption with urbanization lead to pollution. Another result obtained from the analysis findings is that energy consumption causes pollution in the short run. Accordingly, an increase in energy consumption in the short run increases emissions by approximately 0.95% (as seen in Nathaniel 2020; Joshua et al. 2020;Bashır et al. 2021). Although this effect is positive in the long run, it is statistically insignificant. Particularly, environmentally friendly and sustainable production techniques should be encouraged, and renewable energy production and consumption should be emphasized. Also, it is cleared that there is a need for a revision in the energy composition. Another factor that increases carbon emissions in the long run is GDP growth. According to the coefficient estimation results, a 1% increase in GDP increases carbon dioxide emissions by almost 0.11% (as seen in Joshua et al. 2020;Iheonu et al. 2021). Thus, it can be said that Turkey seems not yet reached the stage of sustainable development. This result is critical for the future of the country. Although the shift on production from agriculture to industrial and service sector along with urbanization and increasing production in parallel with this support the GDP increase, environment-friendly policies in production processes should not be ignored. When the short-run results are examined, it is observed that an increase in energy consumption increases pollution at almost the same rate. This result is another indication that environmentally friendly energy sources are not used. All results considered together, the existence of a case against Turkey's environmental performance draws attention. When all variables are considered together, Turkey's policy towards the goal of sustainable growth is important to implement. Since urbanization and consequently rapid production and consumption process cannot be prevented, at least environmental measures should be reviewed in the context of energy in this process. Nowadays, where urbanization, energy consumption, and growth are not considered separately from each other, it has been proven with the results obtained that all three of them should be considered in possible policies. Figures 2, 3, and 4 plot the positive and negative shocks on urbanization, GDP, and energy consumption in carbon dioxide emissions. Figure 1 illustrates how future positive and negative shocks in urbanization will have an impact on carbon dioxide emissions. Accordingly, a positive shock in urban population growth increases emissions, while a negative shock decreases. Although a positive and negative shock likely to emerge in the future in GDP has the same impact on pollution as urbanization, there is a completely different result in energy consumption. That is, as can be seen from Figure 2, a future positive shock in energy consumption reduces pollution, while the negative shock increases the emissions. This unexpected impulse-response relationship signals that the country's future energy consumption composition will be more environmentally friendly. Therefore, each step towards the purpose of sustainability in energy composition has a high long-run return. Considering the increasing energy demand due to urbanization and the increasing economic growth (with energy-intense production), it can be said that the effect of positive future expectation in energy consumption reflects in the other two variables.

Conclusion and policy implications
The aim of this study is to analyze the urbanization and CO 2 emissions relationship in the case of Turkey. The relationship between variables energy consumption, GDP, and urbanization was analyzed by dynamic ARDL method. Short and long-term results were obtained in this study covering the period 1960-2015. An important and positive long-term relationship has been established between urbanization and CO 2 emissions. CO 2 emissions increased by 0.02% if urbanization increased by 1%. This shows that the increase in urbanization in Turkey in the long term has increased carbon emissions. However, when looking at the results of short-term analysis, the relationship between the two variables is negative but statistically meaningless. For Turkey, this positive relationship between urbanization and carbon emissions brings with policy proposals for the future.
Another relationship between GDP and CO 2 has achieved similar results. A similar directional relationship between the two variables has been determined. If GDP increases by 1%, carbon emissions increase by 0.11%. These results show that sustainable growth has not been achieved for Turkey due to GDP increase provides positive effects such as technological innovation, efficiency in energy consumption, efficiency in the consumption of natural resources. On the other hand, with GDP growth, migration movements can take place from the countryside to the cities. In this case, which is an indicator of the transition from agricultural production to industrial production, Turkey should use its energy resources more effectively and reduce fossil fuel consumption. When looking at the relationship between energy consumption and carbon emissions for the short term, it has been observed that energy consumption directly increases emissions. This is an expected result for a developing country. However, unexpected results were observed in the long-term analysis. According to these results, positive shocks in energy consumption reduce pollution. Negative shocks increase carbon emissions.
Consequently, urbanization policies are important in a developing economy like Turkey. Urbanization leads to an increase in industrial production, energy demand, and infrastructure problems. The increase in transportation vehicles used in cities increases the use of fossil fuels. This increases carbon emissions. Using vehicles that minimize emissions  can be an important step to prevent the damage of urbanization. On the other hand, until recently, Turkey did not use fuels such as natural gas that caused less pollution in all its cities. Fossil fuels such as coal and fuel oil are used extensively for heating purposes in rural areas and even in some areas in Turkey. With the increase of urbanization, the narrowing of job opportunities causes people to prefer cheaper heating methods. Another reason is that the necessary infrastructure has not been provided in some regions. As much as possible, the use of fuels that cause less emissions in the production sector and other areas, especially heating, is very important in reducing pollution. Renewable energy investments in Turkey need to be increased and used quickly. The fact that GDP growth is one of the factors that increase pollution in Turkey may be due to increased welfare levels of individuals. Increased welfare indicators, such as individuals who want to live in larger apartments and want to use personal vehicles instead of public transport, increase pollution. These two reasons alone lead to the use of more resources such as heating, electricity, oil, soil, and water. Turkey needs to make arrangements for changes in consumption habits resulting from increased income. For example, to provide thermal insulation in new and old buildings in the country and to set a certain level of carbon emission limits on vehicles. However, the most important thing to do is to ensure efficiency in energy consumption, to create renewable energy sources, and to minimize fossil fuel consumption.
Availability of data and materials Not applicable.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication Not applicable.
Competing interests The authors declare no competing interests.