A study conducted by Zaidi et al. (2019) demonstrated that FD affects natural quality, whereas Chen et al. (2020) demonstrate that political globalization reduced environmental pollution. Salahuddin et al. (2018) used the VECM Granger causality test and shows that GDP, FDI, and FD are linked with environmental concerns. Ertugrul et al. (2016) explored that economic growth plays an important role in the growth of CO₂ emissions. Akalpler and Hove, 2019) determined a strong correlation between economic output, capital expenditures, energy consumption, and environmental concerns. Jiang et al. (2019) revealed that FD directly affects environmental pollution. According to GDP and FDI contributed to natural collapse by increasing CO₂ emissions. Meanwhile, Dzator and Acheampong (2020) discovered that GTI has a favorable impact on CO₂ emissions. Usman and Hammer (2020) revealed that the FD and GTI protect environmental quality over the long run. As GTI getting improves, there is a reduction in environmental pollution. (Abdouli and Hammami, 2017).

In addition, Razzaq et al. (2021) stated that uncertainty exists in economic patterns because of time-varying factors. For this reason, it is quite imperative to examine the nexus between environmental and economics. Their study mainly focuses on the Chinese economy to explore long-run and short-run non-linear associations between financial development, carbon emission, globalization, and natural resources for 1980–2017. In doing so, they applied a non-linear autoregressive distributed lag (NARDL) framework and observed that financial development and globalization are positively linked with carbon emissions. However, negative shocks in natural resources also positively determine China’s environmental pollution. Khan et al. (2020) observed the panel heterogeneity among 192 economies for carbon emission, financial development, and energy consumption. The study results confirm that through panel quantile regression estimation, financial development has its increasing influence, whereas clean energy consumption reflects a negative influence on carbon emission. Acheampong et al. (2020) conducted a comparative analysis of CO2 emission intensity and financial market development while considering developed and emerging economies. The empirical findings infer that the development of financial markets increases the intensity of carbon emission among frontier financial economies.

In contrast, there is no linear relationship between financial market development and carbon emission intensity in standalone financial economies (Khan et al., 2022). Shan et al. (2021) examined the impact of green technology innovation, and renewable energy on carbon emission in the Turkish economy through bootstrapping ARDL. The authors confirmed that green technology innovation and renewable energy help in reducing carbon emission, whereas non-renewable energy boosts carbon emission (Sun and Razzaq, 2022a). Furthermore, both population and personal income have their determinantal effect on carbon emissions. Shan et al. (2021) analyzed the effects of GDP, energy sources, and carbon emission in the OECD and BRICS economies in the long and short run through panel data estimation. It has been finalized that there is a negative correlation between GDP and carbon emission, whereas a negative association exists between non-renewable energy and carbon emission, respectively.

According to Haseeb et al. (2016), urbanization harms environmental sustainability, whereas discovered a strong relationship between GDP, energy consumption, FDI, exchange transparency, and CO₂ neutrality. Vélez-Henao (2020) found that population development in cities is a major cause of natural debasement. Amin et al. (2020) discovered that energy usage reduces the natural quality, but trade openness and population improve environmental pollution in the long term. Sadorsky (2014) described that population growth shows its influence on CO₂ emissions. However, the size of the population is a very important factor that directly affects CO₂ emissions (Dagar et al., 2021; Atchike et al., 2022; Cui et al., 2019; Xiang et al., 2022). Dong et al. (2018) observed the trends in CO₂ emission through economic growth, population, and renewable energy sources across different regions. Considering the cross-sectional dependence and slope heterogeneity during 1990–2014, their study mainly observes that at both regional and global levels, and population and economic growth have their significant and direct role toward a higher level of carbon emission. However, energy from renewable sources is causing a decline in CO₂ emissions. Furthermore, it also confirms that panel causality exists along with the variables of interest. Khan et al., 2021 examined renewable and non-renewable energy sources, natural resources, population growth toward CO₂ emission, and ecological footprints. The study findings confirm that population growth and non-renewable energy sources degrade environmental quality. Weber and Sciubba (2019) stated a long-lasting dispute regarding the extent to which population growth is playing its role toward environmental degradation. However, it is stated that regional level analysis may provide some robust output while isolating the effect of population on the carbon dioxide emission during 1990–2006. Their findings provide some considerable outcomes for the regional population growth on the carbon emission. Sulaiman and Abdul-Rahim (2018) explained in its theoretical context that population growth is to be considered as playing its role toward greenhouse gas emission, specifically in the form of CO2 due to a range of human activities. For 1971–2010, their study mainly considers the autoregressive distributed lag model. However, the study findings confirm that population is not a determinant factor toward carbon dioxide emissions. In contrast, economic growth has its long-term determinantal effect on the carbon emission during the study period.

To carry out the statistical analysis, data were collected from 1980 to 2019. Data for the financial development were collected from the official website of the International Monetary Fund (IMF). In contrast, data for the population, employment rate, carbon dioxide emissions, and energy consumption were collected from the official data portal of the World Development Indicator during the study period. Furthermore, the data for green technology innovations have been collected from the official data portal of OECD statistics.

The IPAT is an equation format that integrates sustainability outputs to three major causal factors: population, affluence, and technology. The equation of IPAT was initially proposed during the 1970s to deeply understand the change in population, affluence, and technology toward their environmental impact (I) (Ehrlich and Holdren, 1972). More specifically, in the IPAT application, the term T mainly determines the environmental impact per unit for the economic activity. Such activities are normally determined through economic growth in the form of gross domestic product with the help of I/GDP ratio. Furthermore, technical efficiency is observed as a key factor to reduce the cause of the environmental outcomes caused by anthropogenic activities.

Meanwhile, among IPAT applications, climate change is very popular, specifically among those studies that are energy-emission based (Wang et al., 2017; Zaman and Abd-el Moemen, 2017; Ozcan and Ulucak, 2021). The preceding (Raskin, 1995) discoveries have provided evidence for using the IPAT model to figure out what causes carbon dioxide emissions to occur (Paramati et al., 2020). The IPAT model is extended into a stochastic variation called Stochastic Effects of Relapse on Popular and Technology (STIRPAT) (Dietz, 1997). Although the implication of the IPAT model was reasonable enough, there was a range of limitations linked with it. To overcome this issue, Rosa and Dietz (1998) have proposed the STIRPAT model. A major advantage of the STIRPAT model comes from its ability to test hypotheses empirically and the conditions that will be present during the experiment, C O 2 i t   =   f   ( F D i t , P o P i t , i t , EN C it ,   G T I i t ,   E R i t ) . (1)

The above-expressed model was taken on from the exploration commitments (Paramati et al., 2017). In the above condition (Eq. 1), CO₂ consists of the components of the population, financial development, energy consumption, green technology innovation, and employment rate, respectively.

This research applies the bootstrapping autoregressive distributed lagged (BARDL) bound testing technique as it is unique for two major reasons. First, this model is not complex toward the integration properties of the study variables. Second, this approach is also suitable for dynamic time-series data. However, compared to conventional ARDL models, the BARDL addresses the issues like inconclusive cases, as expressed by McNown et al. (2018). It also helps in reducing indecision cases because critical values are generated. Furthermore, this method is moderate and dynamic with multiple explanatory variables. Goh et al. (2017) have expressed their view that the BARDL approach in its traditional concept can be specified with the help of the following equation: y t = ∑ i = 1 p α i y t − i   + ∑ j = 0 q β j x t − j   + ∑ k = 0 r γ k z t − k   + ∑ i = 1 s τ t D t , l   + μ t , (2) wherein in Eq. 2, the small letters i, j, k, and l show the lags. Meanwhile, the term t shows the time duration, and yt indicates the main dependent variable: carbon dioxide emission. Furthermore, xt and zt show the key predictors or explanatory variables. Additionally, Dt, i reflects the dummy variable in structural breaks (Kim and Perron, 2009). β   a n d   γ demonstrate the coefficients for the dummy variables, whereas μ t is the error correction terms that can be specified with the help of the following equation: △ y t = ϕ y t − 1 + γ x t − 1 + ψ z t − 1 + ∑ i = 1 p − 1 λ i y t − 1 + ∑ j = 1 q − 1 δ j x t − j + ∑ i = 1 r − 1 π k z t − k + ∑ i = 1 s ω i D t , l + μ t . (3)

Here, the terms λ i , δ j , π k ,   a n d   ω i cover the connected functions. However, the transformation of the above equation into error-correction form, an AR vector in the level, and the equation can be presented in the following format: △ y t = c ˜ + ϕ ˜ y t − 1 + γ ˜ x t − 1 + ψ ˜ z t − 1 + ∑ i = 1 p − 1 λ i ˜ y t − i   + ∑ j = 1 q − 1 δ i ˜ x t − j   + ∑ i = 1 p − 1 π t ˜ z t − k   + ∑ i = 1 p − 1 ω i ˜ D t , l + μ t ˜ . (4)

Meanwhile, the above equation indicates the significance of the three null hypotheses to explain the cointegration among the variables. I)

All relevant error-correction terms are tested by the F1 test (   H 0 :   ϕ   =   γ   =   ψ   =   0 whereas H1: At least one ( ϕ ,   γ , ψ ) are not zero.

In addition, the traditional ARDL estimation helps generate the critical values test for the F1 and T tests, respectively. Meanwhile, BARDL helps in generating some critical values.

Table 2 covers the pairwise correlation matrix for the study variables. A positive and significant correlation exists between CO₂ emissions, FD, GTI, and energy consumption, respectively. More specifically, a highly positive and significant correlation exists between carbon emission and financial development during the study period. On the other hand, energy consumption, green technology innovation, GTI, and employment rate are also positively and significantly correlated. Furthermore, this research examines the VIF scores along with the tolerance level (1/VIF) to justify whether the correlation is problematic or not. As shown in Table 2, all the variables of interest reflect a VIF score of less than 5, and similar is the case with Mean VIF. Accordingly, the tolerance level is also above 0.10; hence, it is inferred that the correlation between the variables of interest is acceptable.

The study employs an ADF unit root test, which is very useful in detecting the unit root for each study variable (Tissue et al., 2009). Meanwhile, as stated under Table 3, through different years and relative quantiles, multiple structural breaks have been presented and tested accordingly. Furthermore, the level of integration between the variables also plays an important role in the justification of the relevant technique’s hereafter. Meanwhile, the ADF unit root test has greater explanatory power than some traditional unit root tests, providing some accurate findings. The study findings in Table 3 report that under ADF (level), all the variables have reported unit root problems. For this reason, with the first difference, the variables are found to be statistically significant. Furthermore, the findings under the ZA test confirm that at ZA (Δ), the study variables are statistically significant with the different structural breaks. This would justify the argument that variables have distinctive order of integration; therefore, our study applies cointegration techniques to examine the presence of cointegration.

After analyzing the structural breaks in the data, our study considers the BARDL cointegration analysis through estimated models, lag length, break year, F score, T score, and the rest of the results under Table 4. The findings under Table 4 through F and T values confirm a long-run equilibrium cointegration association between the study variables such as population, energy consumption, economic growth, employment rate, financial development, and carbon emission, respectively. The results of bootstrapped ARDL cointegration test disclosures from the F test and T test, CO₂ emissions are accounted for all of the identifying factors, which has a value of R² which accounts for 80.6% variation in the model. Finally, the results of the JB test show that the variables of interest are normally distributed during the study period. T_DV served as the t-value for the dependent variable, whereas T_IV was the t-value of the major explanatory variables of the study.

To examine the long-run cointegrated association between the outcome and explanatory and outcome variables, findings are presented in Table 5. It confirms that financial development in Chinese economy reduces the carbon emission in a significant manner (i.e., beta = −0.231, t-value = −2.003, p-value = 0.001). It justifies that more development of financial markets and related institutions in the Chinese economy is a productive indication of environmental degradation; hence, such financial development reflects its sustainable outlook. The existing literature has found supporting and contrary evidence for the nexus between financial development and carbon emissions. In this regard, Amin et al. (2020) explored different proxies of financialization toward carbon emission among top emitters with the help of quantile estimation. It is confirmed that financial development increases carbon emission at low quantiles, negatively affecting pollution over higher quantiles when examined through nine different proxies. However, some other studies have contrary findings. For example, Shahzad et al. (2017) showed that financial development increases the carbon emission by 16.5% under long-run estimation, whereas this impact is observed as 8.7% under short-run, respectively. Zhang (2011) also exerted that financial development is an important driver of carbon emission in the past.

Furthermore, the effect of population toward carbon emission is also negatively significant at 1% as shown in Table 6. This means that increasing population is not a problematic sign of environmental pollution. One of the several reasons for this negative nexus between population and carbon emission might be shifting energy sources from non-renewable to renewable ones. Khan et al. (2021) confirmed that population growth and carbon emissions are related. However, Sulaiman and Abdul-Rahim (2018) claimed that population, among other macroeconomic variables, has a direct role toward more emission and vice versa. Qi and Li (2020) focused on estimating the transfer in carbon through population migration and energy use in China. They highlight that migration flows in China are causing differential outcomes in residential carbon emissions. However, China’s population migration has increased the national total carbon emission.

The findings also report that energy consumption and green technology innovations show their positive/negative role toward carbon emission in the Chinese economy as observed through long-run estimation. More specifically, it shows that a 1% change in GTI and ENC is causing an upward/downward shift of 47.1 and 34.1% in the carbon emission in China. The increasing utilization of energy is among the major carbon emission sources, as highlighted in recent and past studies. Zaman and Moemen (2017) focused on various environmental hypotheses to explore the nexus between energy consumption, carbon emission, and economic growth. The study findings confirm that more energy consumption induces carbon emission. Nawaz et al. (2021a) confirmed the empirical association between energy consumption from non-renewable sources and carbon emission, whereas a long-run positive correlation between energy consumption and carbon emission also exists. The literature also supports mixed evidence about the nexus between green technology innovation and carbon emission. For instance, Du and Li (2019) investigated the trends in carbon dioxide emission through green technology innovation for 71 panel economies. Their results show that ecological innovations do not significantly reduce carbon emissions, specifically among economies below a threshold income level. However, contrary to our findings, Du and Li (2019) stated that it is difficult to find significant evidence for the positive effect of green technology innovation toward carbon productivity, specifically among less developed economies.

Furthermore, we observe that the employment rate among the macroeconomic indicators is directly causing environmental pollution. The coefficient of ER is 0.208, with a t-value of 2.96. This would reflect a significant and positive nexus between the employment rate and carbon emission in China. Yu and Li (2021) examined the association between carbon emission and enterprises’ labour demands. It is stated that approximately 11.5% total employment effect of carbon emission trading policy is found in China.

Likewise, the findings under short-run estimation through the BARDL method show that financial development and population are causing a reduction in carbon emission. In contrast, green technology innovation, energy consumption, and employment rate are causing environmental degradation. The findings under short-run estimations are quite identical to long-run estimation except for the change in the coefficient for the green technology innovation. Furthermore, it is observed that the adjusted value of the R² is found to be 36.5% under short-run estimation, reflecting a reasonable variation in the main dependent variable entitled carbon dioxide emission. Using Durbin Watson’s statistics, the autocorrelation is identified and determined to be non-existent in the empirical findings. Besides, stability analysis has provided satisfactory empirical findings both in the long and short run, respectively.

Finally, our study considers the Granger Causality test estimation for the variables of interest. The findings are presented under Table 7 and observe that study variables are Granger cause with each other at 1% level of significance, except for the CO2-FD, which is significant at 10%.