Introduction

Environment helpful economic progress have increased in place of only focusing on growth since the commencement of the industrial era (Teodorescu 2012; Muhyidin et al. 2015; Manzoor et al. 2018). The concept of sustainable development receives more attention from both developed and developing countries due to environmental degradation such as global warming and climate change. However, Uchiyama (2016) emphasized that there is a symphony among scholars that different economic approaches tempt environmental damages. Consequently, an ultimate discussion is about how to mitigate the effect of \({\mathrm{CO}}_{2}\) emissions without disturbing the economic activities. Collins and Zheng (2015) argued that to detect a quick solution for \({\mathrm{CO}}_{2}\) emissions is a difficult task. In the recent Paris agreement (12 December 2015), all countries stand into a universal motive to fight global warming and climate change. The prime aim of this agreement is to retain the global temperature increase of 2 °C till 2100.

Additionally, the objective is to enhance the capability of all nations to confront the effect of environmental degradations. Therefore, 20 nations including the United States, the United Kingdom, China, Australia, and India, were agreed to improve global assistance to overcome the threat of \({\mathrm{CO}}_{2}\) emissions. But a question emerges that how the developing economies make this hypothesis of agreement true. While most of the economy has achieved remarkable economic growth during the last decade, \({\mathrm{CO}}_{2}\) emissions also increase through multiple effects. Therefore, it is crucial to perceive how to reduce \({\mathrm{CO}}_{2}\) emissions by continuing the growth trend. Frankel and Romer (1999) described that we cannot avoid financial development, while the study deals with increasing \({\mathrm{CO}}_{2}\) emissions, since it enhances a country’s national income that ultimately raises \({\mathrm{CO}}_{2}\) emissions. Moreover, Sadorsky (2010) revealed that augmentative and proficient financial institutions appear with compatible consumer lending approaches, increasing the purchasing power of consumers such as houses, refrigerators, automobiles, etc. which produces more \({\mathrm{CO}}_{2}\) emissions. To clarify this solution for South Asia, this current study examines the asymmetric impact of financial development and energy consumption on \({\mathrm{CO}}_{2}\) emissions during 1974–2014 in the presence of economic growth and urbanization.

In recent years, several theoretical and empirical studies for different countries have expressed the conjunction between financial development and \({\mathrm{CO}}_{2}\) emissions. Academic scholars predominantly focus on this connection during the international financial crisis of 2007–2008. For instance: Ayeche et al. (2016) for European Countries; Tamazian and Rao (2010) for transitional economies; Hao et al. (2016) for China; Ozturk and Acaravci (2013) for Turkey; Yuxiang and Chen (2011) for China; Lee, Chen, and Cho (2015) for OECD economies; Mugableh (2015) for Jordan; Farhani and Ozturk (2015) for Tunisia; Tamazian et al. (2009) for BRIC economies; Shahbaz et al. (2013) for Malaysian economy; Boutabba (2014) for Indian economy; Charfeddine and Khediri (2016) for UAE; Dar and Asif (2018) for Turkey; Jalil and Feridun (2011) for China; Sehrawat et al. (2015) for India; Zhang (2011) for China; Dar and Asif (2017) for India; Siddique (2017) for Pakistan; Al-Muali et al. (2015) for 129 economies; Abbasi and Riaz (2016) for emerging countries; Dogan and Seker (2016) for top renewable energy economies; Godli et al. (2020) for Pakistan; Ahmad et al. (2018) for China; and Godli et al. (2020) for Turkey. In the theoretical work, Yuxiang and Chen (2011) explained that financial development has four different effects on environmental performance such as capitalization effect, technology effect, income effect, and finally, the regulation effect. Numerous empirical studies reveal the mixed impact of financial development on \({\mathrm{CO}}_{2}\) emissions. Addressing the positive impact of financial development, Ma and Stern (2008), Cole, Elliott, and Shimamoto (2005), Lundgren (2003), and Yuxiang and Chen (2011) concluded that not only financial development accelerates the economic condition ameliorating excellent manufacturing tools but also reduces environmental damages by declining pollution and loss in production.

Moreover, as mentioned by Lundgren (2003), financial development accounts as an investment effect for the economy by which it produces modern production equipment and update technology to alleviate environmental degradation. Also, financial development imposes several restrictions on product strategy and manages attractive funds for the company to be benefitted economically plus alleviating environmental degradation in production (Cole et al. 2005). In addition, Ma and Stern (2008) identify financial development as a technological benefit for both the economy and environment.Footnote 1 Conversely, several scholars find an inconsistent relationship between financial development and the environment. In general, financial development degrades environmental quality through producing more \({\mathrm{CO}}_{2}\) emissions (Ayeche et al. 2016; Tamaziana and Rao 2010; Ozturk and Acaravci 2013; Lee et al. 2015; Mugableh 2015; Farhani and Ozturk 2015; Basarir and Cakir 2015). As noted by Cole et al. (2005), financial development arises questionable signals for sustainable development by introducing new heavy industries. Additionally, Ozturk and Acaravci (2013) stated that there is no long-term significant impact of financial development on \({\mathrm{CO}}_{2}\) emissions for Turkey. Therefore, the relationship between financial development and \({\mathrm{CO}}_{2}\) emissions is ambiguous.

To the best of our know-how, the asymmetric combined association between financial development, energy consumption, and \({\mathrm{CO}}_{2}\) emissions is not explored for South Asia. Our paper contributes to the relevant body of work in the field by estimating a non-linear autoregressive distributed lag (NARDL) model to examine the impacts of the shocks, positive and negative, in financial development and energy consumption with the existence of urbanization and economic growth. Moreover, this study analyses the linear ARDL approach to explore the symmetric scenario among the variables and to compare with the non-linear model.

The rest of the paper is arranged as follows: “Literature review and hypothesis” reports the literature review and hypothesis. “Data and methodology” discusses the data and methodology of the study. The results and discussion are in “Methodology”, while Sect. 5 concludes the paper.

Literature review and hypothesis

In general, previous researches use several control variables to examine the relationship between financial development and environmental quality. Table 1 reports the causality between financial development–\({\mathrm{CO}}_{2}\) emissions.

Table 1 Literature review on financial development–\({\mathrm{CO}}_{2}\) emissions nexus
Full size table

On the contrary, there are several findings on the relationship between energy consumption and environmental quality. For instance: Siwar et al. (2009) for Malaysia; Akbostanci et al. (2009) for Turkey; Akin (2014) for 85 countries; Sharif et al. (2020a, b) for Turkey; Zafar et al. (2020) for OECD countries; and Sharif et al. (2020a, b) for top-10 polluted countries. Munir and Riaz (2019) showed that there is an asymmetric association between electricity and coal consumption and \({\mathrm{CO}}_{2}\) emissions in the long run for South Asian economies. Following the previous studies, for instance: Ahmad et al. (2018), Lahiani (2019), Dar and Asif (2017), Mohiuddin et al. (2016), Rayhan et al. (2018), Islam et al. (2017), Sarkodie (2018), Munir and Riaz (2019), Sarkodie and Strezov (2019), Destek and Sarkodie (2019) and Bekun et al. (2019a, b), we see that economic growth and urbanization also affect the environmental performance besides financial development and energy consumption.

Numerous studies have investigated the relationship between economic growth and environmental performance. Apart from ambiguous findings, a great number of studies have explored an inverted U-shaped connection between economic growth and environmental degradation. This hypothesis of the U-shaped relationship is known as ‘Environmental Kuznets Curve (EKC)’ which is first theoretically introduced by Grossman and Krueger (1991) and empirically heralded by Shafik and Bandyopadhyay (1992) and Shafik (1994). Thereafter, several scholars of different countries and regions investigate the presence of EKC. For instance: Moomaw and Unruh (1997), Friedl and Getzner (2003), Martinez-Zarzoso and Bengochea-Morancho (2004), Dinda (2004), Dinda and Coondoo (2006), Galeotti et al. (2006), Kanjilal and Ghosh (2013), Managi and Jena (2008), Akbostanci et al. (2009), He and Wang (2012), Ozturk et al. (2016) and Dogan and Turkekul (2016).

Since the asymmetric relationships between financial development, energy consumption, and \({\mathrm{CO}}_{2}\) emissions are not investigated in the context of South Asia. This current study fulfills this research gap by empirically analyzing the association between considered variables whether it is symmetric or asymmetric. Therefore, the proposed hypothesis is as follows:

  • Null \(({H}_{0})\): There is a symmetric association between financial development, energy consumption, and CO2 emissions.

  • Alternative \(({H}_{A})\): There is an asymmetric association between financial development, energy consumption, and CO2 emissions.

Data and methodology

Data

Since the primary sign of climate change and global warming is \({\mathrm{CO}}_{2}\) emissions, this study uses Carbon dioxide emissions (metric tons per capita) as a proxy of the environmental indicator. Besides, the study uses domestic credit to the private sector (% of GDP) for financial development, GDP per capita (constant 2010 US$) for economic growth, the urban population for urbanization, and energy consumption is taken in kg of oil equivalent per capita. All data, over 1974–2014, are collected from the World Development Indicator (WDI).

Methodology

The existing literature investigated symmetric analysis employing Autoregressive Distributive Lag (ARDL) model associated with the error correction model and granger causality test through which it solely presents the existence of sort run and long run relationships. That is why an asymmetric relationship among variables is not possible for the previous studies. The NARDL modeling technique is employed, because it generates an asymmetric and non-linear cointegration among the variables and also captures short run and long run effects. Moreover, the NARDL approach relaxes the integration order restrictions where the order should be the same for the error correction model. This facility is supported by Hoang et al. (2016).

The NARDL model has some other benefits for ascertaining the cointegration association in a small sample (Romilly et al. 2001). Also, it can be applied regardless of whether the regressors are integrated of order one, zero, or both (Pesaran et al. 2001), thus avoiding the prior problems connected with conventional cointegration techniques such as the Engle and Granger (1987) and Johansen and Juselius (1990). Besides, it discovers not only to assess the short run and long run asymmetries but also to unroll concealed cointegration (Shin et al. 2014). Finally, any endogeneity and multicollinearity problems are eschewed with the appropriate interchange of lag lengths in the model (Pesaran et al. 2001; Shin et al. 2014), which makes the approach more flexible than the other techniques.

Shin et al. (2014) developed new modeling called NARDL model, followed by asymmetric error correction model:

$$\Delta {{\mathrm{CO}}_{2}}_{t} = \theta +{ \partial }_{0} {{\mathrm{CO}}_{2}}_{t-1} + {\partial }_{1}{\mathrm{EG}}_{t-1} + {\partial }_{2}{\mathrm{UR}}_{t-1} + {\partial }_{3}{\mathrm{FD}}_{t-1}^{+} + {\partial }_{4}{\mathrm{FD}}_{t-1}^{-} + \sum_{i=1}^{p}{\Psi }_{i}\Delta {{\mathrm{CO}}_{2}}_{t-i} + \sum_{i=1}^{q}{\phi }_{i}{\Delta \mathrm{EG}}_{t-i} + \sum_{i=1}^{r}{\xi }_{i}{\Delta \mathrm{UR}}_{t-i} + \sum_{i=1}^{u}{\lambda }_{i}^{+}{\Delta \mathrm{FD}}_{t-i}^{+} + \sum_{i=1}^{v}{\lambda }_{i}^{-}{\Delta \mathrm{FD}}_{t-i}^{-} + {\varepsilon }_{t}.$$
(1)

Similarly,

$$\Delta {{\mathrm{CO}}_{2}}_{t} = {\theta }^{^\circ }+{{ \partial }_{0}}^{^\circ } {{\mathrm{CO}}_{2}}_{t-1} + {{\partial }_{1}}^{^\circ }{\mathrm{EG}}_{t-1} + {{\partial }_{2}}^{^\circ }{\mathrm{UR}}_{t-1} + {{\partial }_{3}}^{^\circ }{\mathrm{EC}}_{t-1}^{+} + {{\partial }_{4}}^{^\circ }{\mathrm{EC}}_{t-1}^{-} + \sum_{i=1}^{p}{{\Psi }_{i}}^{^\circ }\Delta {{\mathrm{CO}}_{2}}_{t-i} + \sum_{i=1}^{q}{{\phi }_{i}}^{^\circ }{\Delta \mathrm{EG}}_{t-i} + \sum_{i=1}^{r}{{\xi }_{i}}^{^\circ }{\Delta \mathrm{UR}}_{t-i} + \sum_{i=1}^{u}{{\lambda }_{i}^{+}}^{^\circ }{\Delta \mathrm{EC}}_{t-i}^{+} + \sum_{i=1}^{v}{{\lambda }_{i}^{-}}^{^\circ }{\mathrm{EC}}_{t-i}^{-} + {{\varepsilon }_{t}}^{^\circ }.$$
(2)

Following Eqs. (1) and (2), \(\Delta\) reports the first difference term; \(\theta \mathrm{ and }{\theta }^{^\circ }\) are the drift term; p to v is optimum lag orders selected by the Akaike information criterion (AIC); \(\Psi ,\phi ,\xi ,\lambda , \mu ,{\Psi }^{^\circ },{\phi }^{^\circ },{\xi ,}^{^\circ }{\lambda }^{^\circ },\mathrm{ and }{\mu }^{^\circ }\) show the short run effects, while \({\partial }_{i}\mathrm{ and }{{ \partial }_{i}}^{^\circ }\) denote the long-term effects, and \({\varepsilon }_{t}\mathrm{ and }{{\varepsilon }_{t}}^{^\circ }\) are the white noise error term. A long run estimation contains the speed of adjustment and response time towards an equilibrium point, while short term estimation provides the quick reaction of the exogenous variables. To examine the long run asymmetry (\(\partial ={\partial }^{+}={\partial }^{-}\)) and short run asymmetry (\(\lambda ={\lambda }^{+}={\lambda }^{-}\)), this study uses the Wald test.

However, \({\mathrm{FD}}^{+}\) and \({\mathrm{FD}}^{-}\) are attained by a decomposition of financial development into partial sum of positive and negative changes \(({\mathrm{FD}}_{t}= {\mathrm{FD}}_{0}+ {\mathrm{FD}}_{t}^{+}+{\mathrm{FD}}_{t}^{-})\) as follows:

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

Likewise, energy consumption follows the same decomposition of partial positive and negative changes. Shin et al. (2014) proposed a bounds test procedure to explore an asymmetric long-term cointegrating relationship among the variables. From two procedures of the bound test, this study uses F test of Pesaran et al. (2001).

$${\text{The}}\;{\text{null}}\;{\text{hypothesis}}\;{\text{of}}\;F - {\text{statistic}}\;{\text{test}}:\partial^{ + } = \partial^{ - } = \partial = 0\,({\text{no}}\;{\text{cointegration}}).$$

A long run relationship presents among the variables if the null hypothesis is rejected. The estimation of long run asymmetric effect is based on \({L}_{{mi}^{+}}={\partial }_{4}/{\partial }_{0}\mathrm{ and }{L}_{{mi}^{-}}= {\partial }_{5}/{\partial }_{0}\). This study also examines the asymmetric granger causality test (1969) among financial development, energy consumption, and CO2 emissions.

There are some scholars who employed the NARDL modeling approach to exhibit an asymmetric relationship. For instance: Aftab et al. (2017) for emerging financial markets; Bahmani-Oskooee and Aftab (2017); Aftab et al. (2019) for Asian emerging economies; Ahmad et al. (2018) for Pakistan; Godil et al. (2020a, b) for Turkey; and Zafar et al. (2020) for OECD countries.

Results and discussion

Prior to check cointegration analysis for ensuring long run and short run relationships among variables, it is necessary to examine the stationary properties for every single variable. As mentioned by Gujarati and Porter (1999), non-stationary of time series generates spurious outcomes. The ARDL bound test can be carried out if every single time series variable is stationary at \(I(0)\) or \(I(1)\). Additionally, F-statistics (bound test) of Pesaran et al. (2001) becomes ineffective if the integrated order of any single variable is two or more (Ouattara 2004). This study employs the Augmented Dicky–Fuller (ADF) test and Phillips–Perron (PP) test.

Table 2 reports the results of the unit root test. The findings clearly show that \({\mathrm{CO}}_{2}\) emissions, financial development, energy consumption, urbanization, and economic growth are integrated at \(I(1)\) for both ADF and PP test except economic growth is stationary at the level for PP test. Thereby, bound tests may continue. Tables 3 and 4 reports the results of Shin et al. (2014) non-linear ARDL approach for Eqs. (1) and (2). The table contains four segments. Part A and B indicate long run and short run coefficient estimates, while part C and D exhibit ARDL bounds test and Wald test. However, Table 5 reports the findings of the diagnostic test. However, Fig. 1 discloses the positive and negative trend of financial development and energy consumption. Following the results of Table 4, the calculated F-statistic value is 7.614 which lies over the lower and upper bound critical value at a 1% significance level. Therefore, the NARDL bound test explicitly reject the hypothesis of no cointegration relationship among the variables, connoting long run connection among them.

Table 2 Augmented Dicky–Fuller (ADF) and Phillips–Perron (PP) unit root tests
Full size table
Table 3 Asymmetric effect of financial development on environmental degradation
Full size table
Table 4 Asymmetric effect of energy consumption on environmental degradation
Full size table
Table 5 Diagnostic tests
Full size table
Fig. 1
figure 1

Positive and negative trend of financial development and energy consumption

Full size image

The short run results of model (1) reveal that partial positive sum of financial development has significant and negative impact on \({\mathrm{CO}}_{2}\) emissions, while in the long run, it has no significant effect on South Asian economies. Contrariwise, there is a positive and highly significant effect of the partial negative sum of financial development on \({\mathrm{CO}}_{2}\) emissions but a negative and significant impact in the long run. Accordingly, 1% increase in partial positive changes of financial development leads to decrease \({\mathrm{CO}}_{2}\) emissions by 0.79 or 0.92 percent in the short run and 1% decrease in partial negative changes of financial development proves to raise environmental degradation by 1.71% uplifting \({\mathrm{CO}}_{2}\) emissions in the short run, while reduces environmental erosion by 1.77% in the long run for South Asian economy.

However, urbanization only has a negative and significant effect on \({\mathrm{CO}}_{2}\) emissions in the short run, while economic growth has both short and long run impacts. In the short run estimation, economic growth has a significant and negative relationship with \({\mathrm{CO}}_{2}\) emissions where has a positive association with \({\mathrm{CO}}_{2}\) emissions in the long run. The error correction term measures the speed of adjustment that explains how shortly variables respond to the long run equilibrium. Accordingly, the coefficient of \({\mathrm{ECM}}_{\mathrm{t}-1}\) is negative and significant at 1% confidence level which also means that there exists a static long-term relationship (Banerjee et al. 1998). The adjusted \({R}^{2}\) value (0.6895) exhibits the goodness of fit of the models.

However, the results of the Wald test reject the null hypothesis in the long run that financial development asserts a symmetric impact on \({\mathrm{CO}}_{2}\) emissions long run, explaining that positive and negative variation of financial development has a different significant effect on \({\mathrm{CO}}_{2}\) emissions in the long run. Besides, the diagnostic test results of Table 6 validate that the model is free from heteroscedasticity. Also, Skewness and kurtosis sign ensure that the residuals of the model practice normal distribution.

Table 6 Symmetric effect of financial development and energy consumption on environmental degradation
Full size table

Now, according to the estimated results of Table 4, a partial positive sum of energy consumption has a significant and negative effect on \({\mathrm{CO}}_{2}\) emissions in the short period regarding different lag orders, but contrariwise, it has a positive and significant relationship with \({\mathrm{CO}}_{2}\) emissions in the long period for South Asia. Also, partial negative changes in energy consumption has a negative significant impact on \({\mathrm{CO}}_{2}\) emissions in the short run considering various lag order, while as per expectation, it positively affects the \({\mathrm{CO}}_{2}\) emissions in the long run. Accordingly, 1% increase in the partial positive sum of energy consumption minimizes environmental degradation by 0.44% in the short period taking lag order two and over against produces environmental deterioration by 0.37% in the long period for South Asia. However, as expected, 1% decrease in partial negative changes in energy consumption declines environmental wasting by 5.4% in the short run regarding lag order three, while it pollutes the environmental quality by 3.94% in the long run.

Besides, urbanization worsens the environment of South Asia by 1.99% in the long period. Also, economic growth has a significant and negative impact on \({\mathrm{CO}}_{2}\) emissions in the short run and it has no significant effect in the long run. The speed of adjustment (\({\mathrm{ECM}}_{t-1}\)) shows that Eq. (2) is stable as it is negative and highly significant at 1% level of significance. The value of F-statistic is 6.345 which lies over the lower and upper bound critical value at 1% significance level. Thus, the NARDL bound test clearly reject the hypothesis of no cointegration relationship among the variables, indicating long-term association among them.

However, the outcomes of the Wald test reject the null hypothesis mean that energy consumption validates symmetric impact on \({\mathrm{CO}}_{2}\) emissions in the short run and long run, interpreting that positive and negative changes of energy consumption have a different significant impact on \({\mathrm{CO}}_{2}\) emissions (long run: 10.83; short run: 5.43). The adjusted \({\mathrm{R}}^{2}\) value (0.8661) describes the goodness of fit of the models. Additionally, the estimated results of the diagnostic test in Table 5 proves that the model is free from heteroscedasticity. Also, Skewness and kurtosis sign prove that the residuals of the model practice normal distribution.

Now, moving to Table 6, the ARDL model is calculated to match with NARDL model. Following the estimations of the model (1), the long run elasticities of \({\mathrm{CO}}_{2}\) emissions are positive and highly significant for urbanization and economic growth, while negative for financial development. The outcome suggests that 1% increase in financial development reduces \({\mathrm{CO}}_{2}\) emissions by 0.8%. In contrast, relative to economic growth and urbanization, the short run elasticities are negative and significant for the model (1). Turning to the model (2), energy consumption and urbanization badly affect the environment increasing \({\mathrm{CO}}_{2}\) emissions in the long run and economic growth has a negative association. This result explains that 1% rise in energy consumption enhances \({\mathrm{CO}}_{2}\) emissions by 0.49%. The short run results exert that energy consumption has an adverse relationship with \({\mathrm{CO}}_{2}\) emissions, while economic growth positively affects the \({\mathrm{CO}}_{2}\) emissions.

The error correction term (\({\mathrm{ECT}}_{t-1}\)) is statistically significant and negative for models (1) and (2), thus confirming the presence of long-run dynamics in these models. According to the findings, the ARDL bound test rejects the null hypothesis of no cointegration relationship for both models as the F-statistic values lie over the lower and upper bound critical value at 1% significance level. Both models are well defined due to the characteristics of constant variance and homoscedasticity.

Table 7 reports the asymmetric granger causality test. The results exhibit that there is a unidirectional causal relationship from partial positive sum of financial development to \({\mathrm{CO}}_{2}\) emissions, while \({\mathrm{CO}}_{2}\) emissions also have a unidirectional causal connection with partial negative sum of financial development. On the other hand, partial negative sum of energy consumption generates unidirectional causal relationship with \({\mathrm{CO}}_{2}\) emissions, while there is also a unidirectional causality from \({\mathrm{CO}}_{2}\) emissions to partial positive sum of energy consumption.

Table 7 Asymmetric granger causality test
Full size table

Conclusions and remarks

Based on the yearly data from 1972 to 2014, this study explores the relationships among \({\mathrm{CO}}_{2}\) emissions, financial development, energy consumption, economic growth, and urbanization for South Asia. With the help of non-linear ARDL model, the empirical results validate the asymmetric connection between energy consumption, financial development and environment as the \({\mathrm{CO}}_{2}\) emissions are highly affected by both positive and negative shocks in energy consumption and financial development. The outcomes of the ARDL approach express that energy consumption has a positive impact on \({\mathrm{CO}}_{2}\) emissions in the long run, while financial development has an adverse effect. Comparing the analysis of the ARDL and NARDL model, this study confirms that energy consumption vastly contributes to rise the \({\mathrm{CO}}_{2}\) emissions in South Asia than financial development. To prevent the environmental wasting of South Asia, these factors can be used as important techniques for policy makers and governments.

This paper offers some policy recommendation which is consistent with the results. First, the positive connection between \({\mathrm{CO}}_{2}\) emissions and financial development propounds that the policymakers of South Asia should concentrate on financial development while making policy to decline the greenhouse gases. They should adopt different policies for the different order of economic development. For example: when the economic development is at an early period, the ratio of financial development should be exhorted. Contrariwise, when the economy thrives enough, the adverse effects of financial development on the atmosphere should be cautiously governed. To minimize the deleterious effects of financial development on the environment, the banking sector of South Asia should aware of the misallocation of financial funds. The bank authority should be provided useful financial resources to the proficient and productive industry in place of issuing cheap loans to inefficient and consumptive enterprises. Then, there will be environment friendly technology with high production. Second, the positive relationship between \({\mathrm{CO}}_{2}\) emissions and energy consumption in the findings proposes that the government can enhance the environment quality by imposing restrictions on inefficient energy consumption machines and the use of fossil fuels and should grant subsidies on low carbon use technologies such as renewable energy.

This paper suggests further study using other determinants of environmental degradation such as globalization, trade balance, global value chain, and total employment. Moreover, similar econometric tools can be employed considering both renewable and non-renewable energy consumption.