Trade liberalization and freer trade are globalization’s drivers and expressions. According to trade theories, trade liberalization increases the efficiency of resource allocation to producers who produce at the lowest cost. However, if all relevant social and environmental costs are included into the cost of natural and environmental resources, global production may be created at the lowest environmental cost owing to freer commerce. Trade liberalization may thereby improve societal welfare and sustainability. Market failure, on the other hand, might manifest itself as unpriced, underpriced, or unaccounted-for externalities, or policy failures in the form of subsidies that are incompatible with environmental norms, thus trade liberalization may result in resource misallocation. However, trade liberalization may still have good consequences (efficiency improvements), but it will also have negative impacts when resources are exhausted, resulting in environmental deterioration. The degree of positive and negative impacts determines the net benefit from trade liberalization (Panayotou, 2000).

Through a variety of ways, globalization has an impact on the natural environment and environmental quality. These channels are covered in great length in this article. Globalization’s scale impact refers to a rise in economic activity and revenue generating. There would be an increase in the aggregate level of natural resources and environmental degradation, and the scale effect’s negative impacts would be more pronounced where market failure and regulatory failure prevail. The income impact has a range of effects on the environment. Increased income causes environmental externalities as well as increased consumption. Furthermore, higher wealth raises environmental consciousness and increases governmental and private environmental spending, resulting in improved environmental quality. Globalization aids in the restructuring of the economy, which is based on a country’s comparative advantage in natural resources. Globalization tends to change the economic structure from resource extraction and processing to manufacturing and eventually to a service-based economy if a country is not facing market or policy failure and if stage of development and scale effects are taken into account; this refers to the structural effects of globalization that improve the country’s environmental quality. Globalization’s benefits to sustainable development, on the other hand, are contingent on whether environmental resources are adequately valued or if globalization-induced structural changes boost unsustainable activity.

The dispersion of goods, technology, ideas, and industrial methods between nations is another way that globalization impacts the environment. Environmentally preferable product trade is a significant contributor to environmental deterioration; the global market is worth around one trillion dollars per year, and it has the potential to grow rapidly (OECD 2012, 2012). On the production side, globalization boosts demand for more efficient capital and greener technology, while on the consuming side, greener items such as low-emission automobiles, organic foods, and recycling expand the potential market for environmentally friendly products. Other aspects of globalization, such as intellectual property rights and investment, also have technical consequences that affect the environment. Demand and supply variables, as well as policy consequences, are all affected by technological changes. On the demand side, corporations are under pressure from authorities, shareholders, consumers, and society to embrace greener manufacturing techniques. In rich nations, Henriques and Sadorsky (1996) suggest that regulatory pressure is a possible driver of greener technology, but in underdeveloped countries, Panayotou (1997) claim that community pressure is a critical determinant of environmental concern.

The EKC hypothesis, which depicts an inverted U-shaped association between per capita and environmental degradation, originates from the work of Grossman and Krueger (1991) on the North American Free Trade Agreement (NAFTA), as there are concerns of researchers, environmentalists and policy-makers about free trade and its effect on the environment. The EKC hypothesis postulates that a rise in income is accompanied by environmental degradation in the early stage of development; however, after a certain level of income, a rise in income is not accompanied by environmental degradation. If one considers the environment to be a normal good, then a rise in income due to trade or economic growth will enhance demand for environmental quality, and a high level of income will increase the ability of governments to meet costly environmental protection investments (Copeland and Taylor, 2004).

Many research studies have been conducted to investigate the EKC hypothesis, both in panel and time series investigations; however, we will focus on time series studies in this study. Acaravci and Ozturk (2010), for example, conducted a research including nineteen European nations to investigate the EKC hypothesis. They did, however, come to the conclusion that the EKC hypothesis was only viable in Denmark and Italy. Similarly, Pao and Tsai (2010) did an empirical analysis for the BRIC countries of Brazil, Russia, India, and China and found that the EKC hypothesis is verified in China, India, and Russia. In China and India, Jayanthakumaran et al. (2012) verified the EKC hypothesis by looking at the link between carbon emissions and per capita income while taking trade liberalisation into account. Shahbaz et al. (2015a) did not find evidence for the EKC theory in China, but did find evidence for it in India. Nasir and Rehman (2011) used a time series investigation to establish the link between carbon emissions and per capita income in Pakistan, confirming the EKC hypothesis. In a similar vein, Saboori and Sulaiman (2013) investigated the EKC hypothesis in five ASEAN nations (ASEAN). They came to the conclusion that Singapore and Thailand had the EKC theory. In a similar research, Shahbaz et al. (2015b) looked at 12 African nations and found that the EKC hypothesis is only true in Congo, Ethiopia, South Africa, and Togo.

In the framework of the EKC hypothesis, Haq et al. (2016) looked at the impact of trade openness on carbon emissions. They discovered that trade openness did not worsen environmental deterioration in Morocco since it has a long-term negative relationship with carbon emissions. Furthermore, the study’s findings in Morocco did not support the EKC theory. Using the ARDL bounds testing technique, Shahbaz et al. (2016) investigated the impact of globalisation on carbon emissions in African nations. Globalization increases carbon emissions in Ghana, Morocco, South Africa, Sudan, and Tanzania, whereas it reduces carbon emissions in Angola, Cameroon, Congo, Kenya, Libya, Tunisia, and Zambia, according to this research. In addition, the study’s findings confirmed the EKC theory in Algeria, Cameroon, Congo, Morocco, Tunisia, and Zambia; however, the EKC hypothesis was not confirmed in Sudan and Tanzania.

Khan and Ullah (2019) investigated the effect of globalization on carbon emissions along with other important factors in Pakistan by applying the dynamic ARDL technique over time series data from 1972 to 2016. The results of the study showed that economic, political and social globalization has a positive effect on carbon emissions in the long run as well as in the short run. Thus, the study concluded that globalization aggravates environmental degradation in Pakistan. The study recommended that the government of Pakistan should encourage foreign investment in green energy projects for the sustainable economic development of Pakistan.

Thus, after discussing the related literature above, this study examines the impact of globalization on environmental degradation in the context of the EKC hypothesis through an empirical model that is depicted in Eq. 1 below: C E P = f ( E C ,   E G , P G ,   S G , P C ,   P C 2 ) (1)

The nonlinear combination of the model is transformed into log-linear as depicted in Eq. 2 l n C E P t = γ 0 + γ 1 l n E C t + γ 2 l n E G t + γ 3 l n P G t + γ 4 l n S G t + γ 5 l n P C t + γ 6 l n P C t 2 + ε t (2) where γ 0 is the intercept term and γ 1 , γ 2 γ 3 , γ 4 , γ 5 and γ 6 are the parameters of the respective variable. Similarly, ε t presets the error term and is presumed to be normally distributed. Moreover, it is expected to have γ 5 significantly positive and γ 6 significantly negative, and in this way, the EKC hypothesis will be confirmed in Kazakhstan. The rest of the coefficients of the explanatory variables may be positive or negative.

Various indicators can be used as proxies for environmental degradation; however, some researchers (Edoja, Aye, & Abu, 2016; Haq, Zhu, & Shafiq, 2016; Naradda Gamage et al., 2017) argue that carbon emissions are the most significant source of pollution, accounting for more than 75% of greenhouse gas emissions (Abbasi & Riaz, 2016). Furthermore, according to Tutulmaz (2015), carbon emissions are the most often used proxy for environmental deterioration, particularly when evaluating the EKC hypothesis. As a result, carbon emissions are used as a proxy for environmental deterioration in this research. The EKC hypothesis is investigated in most research projects, such as Saud et al. (2019) and Gasimli et al. (2019), by looking at carbon emissions per capita. As a result, while evaluating the EKC hypothesis, this research will also utilise carbon emissions per capita. The British Petroleum internet database provided statistics on carbon emissions (British Petroleum, 2021). However, it is divided by the total population obtained from the World Bank to arrive at per capita figures (2021). Similarly, the British Petroleum online database is used to gather statistics on energy use (British Petroleum, 2021). Similarly, based on Dreher’s assessment, statistics on globalization and its components (economic, political, and social) are gathered from KOF institutions (2006). Economic, political, and social globalization are the three aspects of the KOF Globalization Index. Each dimension’s index is calculated on a scale of one to 100, with 100 being the greatest value for each degree of globalization. The data on these dimensions has been standardized, and the data has a low sensitivity to outliers. Trade and financial globalization make up economic globalization, and both have a 50 percent weighting, whereas political globalization categories are equally weighted. Interpersonal, informational, and cultural globalization are all equally weighted components of social globalization. In addition, the World Development Indicators World Bank online database is used to obtain actual per capita statistics in US dollars (World Bank, 2021). This research used data from 1991 to 2018, and to solve the problem of a limited sample size for time series analysis, it used the quadratic sum technique to transform yearly time series data to quarterly data. As a result, this research includes quarterly data from 1991 to 2018, i.e., 1991Q1 to 2018Q4. The Description of Variables are described in Table 1.

Because time series data tend to trend over time and have nonstationarity issues, using conventional least squares may result in misleading regression. The nonstationarity issue may be solved by differencing variables; however, if a long-run link between variables exists, the differencing procedure may prevent it. The first and most important thing a researcher should do is check for nonstationarity in time series data; for this, the researchers will use the Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test Kwiatkowski et al. (1992) and the Elliott–Rothenberg Stock (ERS) test in this study Elliott et al. (1996). In general, the augmented Dickey-Fuller (ADF) test (Dickey & Fuller, 1979) and Phillips-Perron (PP) test (Phillips & Perron, 1988) have extremely low power against I (0) alternatives that are near to being I (0) (1). These tests are known as efficient unit root tests, and they have much more power than ADF or PP unit root tests, particularly when is near to unity. Furthermore, Zivot and Andrew unit root tests will be used to see whether time series data have an unknown structure. This unit root test detects an unidentified structural flaw.

The cointegration technique is used next in a time series analysis to reveal long-term connections between variables. The autoregressive distributed lagged (ARDL) approach, also known as the limits testing strategy developed by Pesaran et al. (1996) and improved by Pesaran et al., was utilised in this study (2001). This ARDL may be used for long-run cointegration over typical processes such as the Johansen and Juselius (1990) cointegration approach due to various features. For starters, it can be employed on a small dataset with significant long-run cointegration that can be trusted (Boutabba, 2014). Second, this technique can accommodate dummy variables as well. For instance, a dummy is introduced for a structural break. Third, time series variables need not be integrated in the same order. The time series may be integrated using I (0), I (1), or a mix of I (0) and I (1), but no higher order than I will be used. Meo et al., 2018; Shahbaz et al., 2013). Fourth, the ARDL limits test separates dependent and independent variables while providing unbiased short- and long-run estimates. Furthermore, there are no endogeneity difficulties in the ARDL bounds test estimations (Narayan, 2005). The empirical model of this study can be presented in an unrestricted ARDL framework as follows: Δ C E P t =    α 0 +   ∑ i = 1 m α i Δ C E P t − i +    ∑ i = 0 m α i Δ E C t − i +    ∑ i = 0 m α i Δ E G L t − i + ∑ i = 0 m α i Δ P G L t − i +   ∑ i = 1 m α i Δ S G L t − i + ∑ i = 1 m α i Δ P C t − i + ∑ i = 1 m α i P C t − i 2 +   β 1 E C t − 1 +    β 2 E G L t − 1 +    β 3 P G L t − 1 +    β 4 S G L t − 1 +    β 5 P C t − 1 + β 6 P C t − 1 2 + β 7 D U M + β 8 E C T + ε t (3)

Equation 3 is an error correction mechanism that gives us short run and long parameters, whereas α i gives the short run dynamic parameter of the respective variables, while β i provides long run parameters of the respective variables. In addition, α 0 represents the drift term, ε t is the white-noise error term, and ECT presents an error correction term that has to be negative and significant so that the model may be dynamically stable. The long-run cointegration hypotheses of the ARDL model are stated as H 0   ;    β 1 = β 2 = β 3 = β 4 =   β 5 = 0  

Against Alternative Hypothesis H 1   ;    β 1 ≠ β 2 ≠ β 3 ≠ β 4 ≠   β 5   ≠ 0

H₀: reveals that variables are not cointegrated in the long run, whereas the alternative hypothesis describes long-run cointegration. This cointegration test is based on two critical values that are lower and upper bounds (Pesaran et al., 2001; Narayan, 2005). One has to accept the alternative hypothesis if the estimated F value is greater than the upper bounds and has to accept a null hypothesis if the estimated F value is lower than the lower bounds. The test is inconclusive if the estimated F value is between the lower and upper bounds.

Table 2 presents descriptive statistics. The descriptive statistics illustrate the mean, minimum, maximum and standard deviation of the data. The carbon emissions per capita mean is 2.72 tons along with its minimum value and maximum value of 1.66 and 3.68, respectively. Likewise, energy consumption per capita has a minimum value and maximum value of 86.16 and 188.64 kg of oil equivalent, respectively, whereas its mean value is 136.20 kg of oil equivalent. The economic globalization index mean is 49.08, and its minimum and maximum values are 23.080 and 61.07, respectively. The mean values of the political and social globalization indices are 56.21 and 53.36, respectively. During the study period, the minimum value of real income per capita was 3,713.98 USD and maximum value of 11,044.59 USD whereas mean income per capita is 6,951.37 USD.

Results of correlation are depicted in Table 3. It can be evidenced from the correlation results that energy consumption is highly correlated with carbon emissions and is followed by per capita income. The correlations of these mentioned variables are 0.99 and 0.64, respectively. Among indices of globalization, economic globalization is not only highly correlated compared to political and social globalization but also negatively correlated. Other than carbon emissions and energy consumption, economic globalization is positively correlated with the rest of the variables. Political and social globalization is positively correlated with carbon emissions. Likewise, political and social globalization are highly positively correlated with income per capita in Kazakhstan.

Before moving on to regression, it is important to check whether the data are stationary or nonstationary and to determine whether variables are integrated of I (0) or I (1). This study applied KPSS and ERS unit root tests. The results of the KPSS and ERS for all variables used in this study are presented in Table 4. The results of both unit root tests, KPSS and ERS, illustrate that carbon emissions per capita and energy consumption per capita are nonstationary when the variables are at level. Both of these variables became stationary by considering them at the first difference. Hence, carbon emissions per capita and energy consumption per capita are integrated of order I (1), whereas the rest of the variables of the study are integrated of order I (0). It can be concluded from unit root tests that the ARDL bounds test is the appropriate cointegration technique that can be applied for cointegration.

To capture unknown structural breaks, the Zivot-Andrew unit root test will be conducted, so the results of the Zivot-Andrew unit root test with structural breaks are presented in Table 5. In the presence of structural breaks, all variables are integrated of order one. This means that all variables are stationary at the first difference in the presence of structural breaks.

Table 6 depicts results of ARDL F-bounds test. The estimated value of the F-bound statistic is greater than the upper bounds critical value at the 1 percent level of significance. Thus, the variables of the study have a long-term relationship. These results confirm that all explanatory variables are determinants of carbon emissions in the long run in Kazakhstan. In addition, after confirmation of the long-run relationship among variables, one can proceed to obtain short-run and long-run estimates based on ARDL.

Table 7 shows the outcomes, whereas Table 7's Panels (A) and (B) show the long and short run results, respectively. Panel (C) of Table 7 displays the diagnostic tests. Let’s start with the long run findings. Energy consumption is observed to have a beneficial influence on carbon emissions over time; hence, Kazakhstan’s energy structure is unsuitable for environmental quality, since the country’s energy demand is met using nonrenewable energy resources. In terms of the influence of globalization on carbon emissions, all aspects of globalization, economic, political, and social, have shown to have positive and substantial long-term impacts on carbon emissions; nonetheless, globalization is accelerating environmental deterioration in Kazakhstan. This study is in line with Shahbaz, Solarin, and Ozturk (2016), who discovered that globalization causes carbon emissions to accumulate in Ghana, Morocco, South Africa, Sudan, and Tanzania. Similarly, Khan et al. (2019) determined that economic, political, and social globalization are increasing environmental degradation in Pakistan, despite the fact that these aspects of globalization have a favorable impact on carbon emissions in Pakistan. The scale impact of globalization, which leads to increased economic activity and output to fulfil global demand, is responsible for the positive effect of economic globalization. Furthermore, the positive impact of globalization on carbon emissions in Kazakhstan refutes Najam et al. (2016)’s argument that globalization plays a proactive role in reducing the severity of environmental problems by enacting global-level legislation. Meyer (1999) claimed, on the other hand, that national policies are designed and implemented according to a global template, since a country’s modernization is a core part of political globalization, which is responsible for environmental damage. The reason for this is that modernization alters the economy’s structure as well as energy and land consumption, placing additional strain on the environment (Malone, 2002). This theory is backed up by research from Kazakhstan, which shows that political globalization increases carbon emissions, worsening environmental deterioration.

The EKC hypothesis is shown to be true in Kazakhstan in the long term, with positive and negative signals for per capita and its squares, respectively. This suggests that carbon emissions rose in tandem with economic activity until they reached a point when they did not, proving an inverted U-shaped link between per capita and environmental deterioration in Kazakhstan. Apergis and Payne (2009), Jalil and Mahmud (2009), Jayanthakumaran et al. (2012), and Khan et al. (2019) in Turkey, China, Bangladesh, and Pakistan, respectively, corroborate this conclusion of the research.

When it comes to short-run outcomes, lags of variables are involved, and the Wald test is used to determine if a variable and its delays are significant or not. Energy consumption has a positive and substantial influence on carbon emissions in the short run, and energy contributes to environmental deterioration in the short run, among the independent variables. Although all aspects of globalization have a negative impact on carbon emissions, economic and political globalization have a substantial impact on whether social globalization has a negligible impact. The EKC hypothesis occurs in Kazakhstan in the short term, since the coefficients of PC and PC2 have the predicted signs and are significant, as shown by the findings. PC has a positive relationship with carbon emissions, but PC2 has a negative relationship. Jalil and Feridun (2011) and Shahbaz et al. (2015a), who tested the EKC hypothesis in China, India, and Ghana, came to the same conclusion. Furthermore, since ECT is significant with a negative sign, the findings demonstrate that the model is stable and in equilibrium. Because the model’s magnitude is 0.30, it may be assumed that it will respond to any external shock in less than a year. The diagnostic test results shown in panel (C) indicated that the overall model is significant (F-test) and stable, as shown by the Ramsey reset test. Similarly, there are no autocorrelation or heteroscedasticity issues in the model.