Numerous researchers in various countries or regions have observed the nexus between economic growth and the environment. The results differ based on the size of the sample and the studied period (Koengkan et al., 2019a; Chishti et al., 2021; Qin et al., 2021). A large number of researchers have used the EKC hypothesis to study the nexus between economic growth and environmental quality (Yilanci and Pata, 2020). The validity of the theory is proved in various countries like the US (Atasoy, 2017), Pakistan (Rehman et al., 2021a), Malaysia (Nurgazina et al., 2021), China (Pata and Caglar, 2021), OECD (Cao et al., 2022).

On the other hand, some studies cannot find the nexus between economic growth and environmental degradation. For instance, Zambrano-Monserrate et al. (2018) analyze the nexus in Peru and find that the results do not support the EKC hypothesis. Another research on South Korea by Koc and Bulus (2020) discovers evidence of an N-shaped link between economic growth and environmental degradation that invalidates the EKC theory. The EKC hypothesis is also invalid in Pakistan according to the findings of Ahmed et al. (2020), where an increase in wealth boosts CO 2 emissions by forming a U-shaped nexus.

The EKC theory, on the other hand, has been supported by multiple studies. For example, Katrakilidis et al. (2016) indicated a positive nexus between economic growth and environmental degradation in Greece. Rauf et al. (2018) revealed in research on the Belt and Road Initiative (BRI) countries that the EKC hypothesis fits all regional panels of the BRI countries. Furthermore, Işık et al. (2019) confirmed the EKC theory in the context of ten states across the United States and concluded that economic growth first increased CO 2 emissions but later reduced them. Many scholars (Zhu et al., 2019) examine the nexus between economic growth and CO 2 emissions in China and found that as the economic level rises, the environmental degradation decreases. Murshed et al. (2021) found support for the EKC hypothesis in the long term in Bangladesh. Likewise, Ahmad et al. (2021) confirmed the EKC for developing countries, as well as the positive environmental benefits of both institutional quality and economic complexity. When including green trade in models, Can et al. (2021) claim that the EKC between environmental deterioration and economic growth persists in OECD countries.

A variety of studies have examined the nexus between energy usage and environmental degradation, especially CO 2 emissions (Khan et al., 2021). Bidirectional nexus between these variables has been discovered by certain studies, for example, Pao et al. (2011) identify long-run mutual Granger-causality between energy consumption and environmental degradation in Russia between 1990 and 2007 and suggest that environmental efficiency must be enhanced to decrease pollution. Wasti and Zaidi (2020) determine a mutual causal nexus between energy consumption and environmental degradation in Kuwait. Using the wavelet coherence method, classical Granger, and Toda-Yamamoto causality approaches, Adebayo and Akinsola (2021) discover a bidirectional nexus between environmental degradation and energy consumption in Thailand.

Omri (2013) establishes the presence of positive unidirectional causation between energy consumption and environmental degradation in 14 MENA countries. Furthermore, Ahmed et al. (2017), Aye and Edoja (2017), and Musah et al. (2021) discover energy consumption to be a key booster of CO 2 emissions in five countries of South Asia, 31 emerging nations, and North Africa, respectively. Muhammad (2019) studies the nexus between energy consumption, environmental degradation, and economic growth in the MENA countries by using GMM and SGMM long-term estimations and concludes that an increase in energy consumption causes an increase in environmental degradation in the long run. Using the ARDL bound test approach for Pakistan from 1975 to 2014, Ali et al. (2021) define that fossil energy consumption has a negative influence on environmental degradation. Moreover, Rahman et al. (2021) discover that energy consumption increases the CO 2 emissions of newly industrialized countries between 1979 and 2017 by using Dynamic Ordinary Least Squares (DOLS), Fully Modified Ordinary Least Squares (FMOLS), and Pooled Mean Group (PMG) methods.

On the other hand, Al-Mulali et al. (2015) show that energy usage does not influence CO 2 emissions across Latin America and the Caribbean. Most of the studies, find no impact or positive impact of energy consumption on CO 2 emissions investigate renewable energy. For instance, using disaggregated data for India, Sahoo and Sahoo (2020) evaluate the influence of renewable and nonrenewable energy consumption on environmental degradation and conclude that hydro-energy consumption does not affect the CO 2 emissions, but nuclear electricity consumption reduces CO 2 emissions. Le et al. (2020) reveal that green energy decreases CO₂ emissions in high-income nations using a global panel of 102 economies. Similarly, Ummalla and Goyari (2021) define that using renewable energy reduces CO 2 emissions.

Despite the importance of the topic, there is still a research gap, because numerous studies (Magazzino et al., 2020b; Koengkan and Fuinhas, 2021a; Guo et al., 2022a) in the energy and environmental economics literature apply panel data and time-series analyses to investigate the long-run short-run relation between different variables, however, this research applies the DYNARDL simulations model established by Jordan and Philips (2018) in terms of carbon neutrality in China. This study uses the DYNARDL model to analyze the actual change in the dependent variable in the long and short term by introducing 100% negative shock from explanatory variables. The DYNARDL simulations have the capacity to solve the data’s existence issues and interpret the results of the standard ARDL model. While the remaining variables are kept constant, the DYNARDL simulations will approximate and reflect the predictions of an actual change in the independent variable (Jordan and Philips, 2018).

To study the essential variables’ influence on environmental degradation, the study contributes to the literature by including energy consumption, economic growth, financial development, FDI, and urbanization in the CO₂ emission equation. Furthermore, no other researcher has conducted similar research for the China case by using the same factors and for the same period. Therefore, this study can be evaluated as pioneering and significant. Also, this study contributes to the existing literature by providing a clear route for scholars to understand the nexus between selected factors. Moreover, this study assists policymakers of China in developing and implementing strategies to reduce CO₂ emissions to meet the carbon neutrality objective by 2060.

This study analyses the time series annual data set from 1979 to 2019 for China. Summary statistics are presented in Table 1. CO₂ emissions (metric tons per capita) are used as a reference for the environmental degradation in this study. Furthermore, the study employs GDP per capita (constant 2010 US dollars) for the economic growth (EG), energy consumption (EC) is measured in kilograms of oil equivalent per capita, domestic credit to the private sector (% of GDP) represent financial development (FD), urbanization (UR) is measured in % of the total population, FDI shows the net foreign direct investment inflow (% of GDP). The data for the independent and dependent variables are obtained from the World Bank (WorldBank, 2021).

There are five stages for the evaluation of the impact of economic growth and energy consumption in the context of financial development, urbanization, and FDI inflow for China as indicated in Figure 1.

In the first step, PP and ADF tests are used to identify the order of variable integration. In the second step, the PSS bound test is performed to confirm the presence of long-run cointegration among the variables. In the third step, the ARDL model is employed to identify the short-run and long-run associations between variables. In the fourth step, DYNARDL simulations are estimated. In the fifth step, Kernel-based regularized least squares (KRLS) are utilized to identify the causal association between the variables.

To establish the order of integration and produce comprehensive results, the unit root tests are employed for checking the characteristics of the parameters by utilizing PP and ADF tests (Dickey and Fuller, 1981; Perron, 1989). The findings of the unit root tests are shown in Table 2.

According to the test results, the ARDL and DYNARDL models can be utilized with the studied variables because the dependent variable is integrated at the I (1) and independent parameters are stationary and integrated at the order I (0) and I (1).

The lag range (LR) test, final perdition error (FPE), Akaike Information Criteria (AIC), Hannan-Quinn Information Criteria (HQIC), and Schwarz’s Bayesian information criterion (SBIC) is used to select the optimal lag for further analysis. The results reveal that the suitable lag is lag 2 (Appendix 1).

The results of the ARDL (1, 2, 0, 1, 1, 2) regression are presented in Figure 2 with its empirical results presented in Table 3.

The result of the analysis discloses that economic growth and energy consumption are found to increase environmental degradation in both short-run and long-run analyses. This output aligns with the studies of Fuinhas et al. (2017), Fuinhas et al. (2021), Nawaz et al. (2021), who find the same nexus between energy consumption and CO 2 emissions in 29 European Union Countries, Latin America and BRICS, OECD economies, respectively. Urbanization has a positive impact on CO 2 emissions in the short term, while in the long term increase in the urban population will decrease CO 2 emissions. This is because increasing the compactness of the population will generate economies of scale that improve energy efficiency as well as the efficiency of energy-intensive facilities, thus ultimately reducing CO₂ emissions. Moreover, in addition to economies of scale, population density provides an agglomeration effect that favors technological measures to reduce emissions. This is especially relevant for China, whose population is concentrated in several megapolises, each with more than 20 million people. This result collaborates with the previous studies by Anwar et al. (2020) for Far East Asian Countries. The FDI inflow is only significant in the short-run and negatively affects CO 2 emissions. This result is contrary to the study of Malik et al. (2020), who find that FDI intensifies carbon emissions.

Furthermore, the error correction term (ECT) is negative and significant at less than 5%, indicating that the adjustment speed to the long-run equilibrium will take more than 5 years. Moreover, the R 2 value of 0.8323 indicates that the independent variable can explain 83.23% of the variation in CO 2 emissions.

To examine the long-run cointegration relationship, the ARDL bounds cointegration test (Pesaran et al., 2001) is evaluated by utilizing the novel Kripfganz and Schneider (2020) critical values and approximate p-values. The bounds test’s results are shown in Table 4.

Table 4 illustrates that the F-statistic value of the proposed model is higher than the critical value of the upper bound (4.54) at a 1% level of significance thus there are long-term relationships among variables. Several tests checking autocorrelation, heteroskedasticity normality, and structural breaks are done to ensure the stability of the ARDL model. It can be observed from the result of the Breusch Godfrey LM test that the hypothesis of no serial correlation among variables is accepted at a 5% significance level, which means that residuals are free of serial correlation. Cameron and Trivedi’s decomposition of the IM-test shows that the homoscedasticity null hypothesis is accepted at a 5% level of significance, thus the residuals are free from heteroscedasticity. The Skewness/Kurtosis normality tests disclose that the residuals are normally distributed within the mean.

A standardized normal probability plot (Figure 3A) and quantiles of residuals against quantiles of normal distribution (Figure 3B) are used to further examine the validity of the normality assumption determined by the Skewness/Kurtosis tests.

The residuals based on the ARDL model are normally distributed in both Figures 3A,B. Furthermore, the cumulative sum test is used to analyze potential structural breaks that are presented in Figure 4.

Figure 4 shows that the assessed t-statistic is within the 95% confidence interval which means the calculated coefficients are stable through the years.

Figure 5 shows the variable graph of the DYNARDL, while Table 5 shows its empirical estimation.

The results of the long and short-run simulations illustrate that economic growth will increase CO 2 emissions. This output aligns with the studies of Koengkan et al. (2019b), Malik et al. (2020), Zhang et al. (2021), Rehman et al. (2022). Moreover, growth in energy consumption will worsen the environmental situation. Our result is consistent with numerous studies conducted in various countries such as Khan et al. (2020) in Pakistan, Ummalla and Goyari (2021) in BRICS countries, Adebayo (2021) in Indonesia, Martins et al. (2021) in G7 countries, and Pata and Kumar (2021) in China. Financial development has no impact on environmental degradation. Urbanization will improve the environmental situation in the long run, while in the short-run growth in the urban population will increase CO 2   emissions.

Moreover, the estimated ECT of -0.484, which is significant at a 5% level indicates the long-run cointegration between economic growth, energy consumption, financial development, urbanization, FDI inflow, and environmental degradation. Moreover, the  R 2   value of 0.7622 indicates that the explanatory factors can explain 76.22% of the variation in CO 2   emissions.

In general, both the ARDL and DYNARDL estimates suggest that China’s economic development and energy consumption have a detrimental impact on the environment. The DYNARDL simulation is based on carbon neutrality by 2060 (Mallapaty, 2020; Hepburn et al., 2021; Ren et al., 2021). The simulation results are presented in Figure 6.

The plots presented in Figure 6 expose that -100% of shocks in the estimated economic growth do not affect environmental degradation, while the same shocks in the calculated energy consumption boost CO 2   emissions in the first 5 years and stabilize in the long run.

A machine learning KRLS approach is employed to check and identify the relationships among the variables to additional enhance the results of this study and the results are presented in Table 6.

Table 6 shows that the general model’s predictive power is 0.997 meaning that descriptive factors explain 99.7% of the variation in CO 2   emissions. The average pairwise marginal effect of economic growth, energy consumption, financial development, urbanization, and FDI inflow are 0.12%, 0.28%, -0.02%, 0.25%, and 0.03%, respectively. Except for financial development, the probability value of every parameter at a 1% level of significance indicates that there is evidence of a causal-effect correlation. The long-term impacts of economic growth and energy consumption fluctuation as well as their influence on carbon emissions are also investigated by plotting the pointwise derivative that is presented in Figure 7.

The marginal impact of economic growth and energy consumption on environmental degradation represented in Figure 6 shows that the increasing level of economic growth and energy consumption raise CO 2 emissions until they reach a limit where growing marginal returns occur, but then economic growth and energy consumption decrease when CO 2 emissions increase. Thus, economic growth and energy consumption have declining marginal returns with growing CO 2 emissions.