**4. Results and Discussion**

In the first step, we use the Augmented Dickey-Fuller and Phillips-Perron unit root tests to check if all variables are stationary. The null hypothesis of the ADF and Perron tests is that the variable contains a unit root, and the alternative is that the variable is generated by a stationary process. The results of the tests with intercept and trend can be found in Table A3. The null hypothesis can be rejected at the 1% level of significance for all variables at the first difference. This implies that all variables used in this study are integrated on the order of one I(1).

After confirming the ordering of the integration, we apply the ARDL and NARDL approaches to examine long-run relationships (cointegration) and estimate the coefficients. To implement these approaches, the selection of appropriate lag length is necessary. We chose one lag based on the results of Akaike's information criterion and Schwarz's Bayesian information criterion. Tables 3 and 4 provide the results of ARDL and NARDL tests for cointegration. In the NARDL test, the null hypothesis of no cointegration between variables was rejected at the 10% level of significance in Bangladesh, India and Pakistan, and at 5% in Sri Lanka. The estimated F-statistics were larger than the critical upper bounds. The results of the NARDL test were more significant (Table 4). The null hypothesis was rejected at the 1% level of significance in India and Pakistan, and at 5% in Bangladesh. We also rejected the null hypothesis for Sri Lanka, accepting that the F-statistic was slightly smaller than the upper bound at the 10% level of significance. In summary, these results show that all equations are co-integrated.


**Table 3.** Results of the ARDL test for cointegration. Model: LnCo2 = f(LnE, LnY, Ln*Y*2, Ln*T*+, Ln*T*−).

Sources: The authors' estimation. Note: \* and \*\* show the significance at 10%, 5% and level respectively.

The differences between coefficients estimated by the ARDL and NARDL approach are highlighted in Table 5. The NARDL estimation captures richer insights into the asymmetric effects of trade openness on CO2 emissions. As specified in Equation (3), trade openness is split into positive and negative shocks in the NARDL model. Table 5 compares the long-run and short-run coefficients.


**Table 4.** Results of the NARDL test for cointegration. Model: LnCo2 = f(LnE, LnY, Ln*Y*2, Ln*T*+, Ln*T*−).

Sources: The authors' estimation. Note: \*, \*\* and \*\*\* show the significance at 10%, 5% and 1% level respectively.


Sources: The authors' estimation. **Notes:** <sup>1</sup> The lag length for CO2 in India is 2; thus, additional coefficients were estimated: 0.825 \* (ΔLNCO2t−1); <sup>−</sup>0.001 (ΔLNEt−1); <sup>−</sup>0.020 \*\* (ΔLNYt−1); 0.004 \*\* (ΔLN*Y*<sup>2</sup> <sup>t</sup>−1); 0.824 \* (ΔLNTt−1). \*, \*\* and \*\*\* show the significance at the 10%, 5% and 1% level, respectively.

As outlined above, the signs of the coefficients associated with GDP per capita can have positive and negative values. Based on the inverted U-shaped EKC hypothesis, the relationship requires that *β*<sup>3</sup> should be negative (and *β*<sup>2</sup> should be positive). We observe similar coefficients in the ARDL and NARDL models. Similarly, Dong et al. [25], Murshed et al. [20], Khan et al. [18], and Sadiq et al. [6] proved the inverted U-shaped EKC hypothesis in that region. Dong et al. [25] pointed out that the turning points lie at \$1181.60 in Pakistan, \$1861.49 in India, and \$1937.23 in Bangladesh, while the turning years were estimated in 2041, 2039, and 2048, respectively. Other studies indicate that using renewable energy is associated with environmental betterment [20], and sustainable development policies can revisit the conflict between globalization and environmental degradation [6,18].

In the NARDL model, the long-run coefficient for squared GDP per capita is negative and significant in India (−1.180 \*) and Pakistan (−1.787 \*\*\*), while it is positive in Sri Lanka (1.663 \*\*). The coefficients for India and Pakistan indicate that we should expect increased environmental quality. Notably, the Indian government has taken many initiatives to reduce environmental degradation in recent years. For example, the International Solar Alliance's launch summit was co-chaired by Prime Minister Narendra Modi and French President Emmanuel Macron in March 2018, demonstrating India's leadership in supporting renewable energy (ISA). In January 2019, the Ministry for Environment introduced the National Clean Air Program (NCAP), which gives the states and union government a framework to tackle air pollution. Since 2018, India's 2019 climate change index (CCPI) performance has improved from 14th to 11th place [86]. Pakistan has recently given serious thought to addressing the world's escalating environmental concerns, according to the United National Development Program 2020. Several Acts have been promulgated along with some policies and public sector initiatives currently in effect. For example, clean and green initiatives have been implemented; environmental protection agencies at the federal and provincial levels have been strengthened; environmental laboratories and courts, national environment quality standards, the National Energy Efficiency and Conservation Authority (NEECA), and national environmental quality standards have all been developed [87].

Another potential environmental problem is that the coefficient associated with GDP per capita is relatively high in Pakistan. Let us recall at this point that the coefficients of GDP per capita indicate the scale effect, which is associated with adverse environmental consequences. It is highly probable that high trade openness causes pollutant emissions due to increased economic activity. Our study corroborates the findings by Ullah et al. [88] and Khan et al. [89], who found that trade liberalization (trade openness) led to increased CO2 emissions in Pakistan. This positive relationship can be explained by scale effects where large-scale manufacturing operations, particularly in fossil-fueled and export-oriented industries, increase emissions of pollutants. This is because in the early stages of the development process, more emphasis is placed on economic growth than on pollution control. At this stage, less developed countries are often "hungry" for rapid economic growth to fight against poverty. The negative sign of *β*<sup>2</sup> in Sri Lanka should definitely be assessed in a positive way.

The results of the long-run coefficients associated with energy consumption, both in the ARDL and NARDL model, surprised us. Usually, energy consumption significantly impacts the dioxide carbon emissions in such a way that there is a positive long-run relationship between these two (cf. Wang et al. [90], Gierałtowska et al. [91] and Verbiˇc et al. [92]). Energy consumption should likely be associated with other factors. This relationship is visible in developed countries. For example, Wang et al. [90] indicate that energy intensity and foreign direct investment and urbanization strongly impact carbon dioxide emissions. In our research, these long-run coefficients are significant only in Pakistan (ARDL 0.889 \* and NARDL 0.945 \*\*\*). By contrast, these coefficients are highly significant in the short run in Bangladesh (37.464 \*\*\*), Sri Lanka (1.976 \*\*\*), and Pakistan (1.153 \*\*\*) in the NARDL model.

The most interesting finding was that the long-run coefficients associated with trade openness shocks, both negative and positive, significantly impacted CO2 emissions only in Sri Lanka (at the significance level of 5%). These research results did not support the hypothesis that trade openness significantly impacts carbon dioxide emissions in South Asian countries. The estimated coefficients of trade openness with positive and negative shocks are 1.887 and 1.594, respectively. Therefore, increasing trade openness by 1% increases carbon dioxide emissions by 1.887%, while reducing trade openness decreases carbon dioxide emissions by 1.594%. These impacts are based on the scale effect. The primary contributors to Sri Lanka's economy are tourism, tea export, textile and garment manufacturing, rice and other agricultural goods, and food products. Gasimli et al. state that domestic investors do not use environmentally friendly technology [93]. Additionally, imported technology in the form of machinery does not have a positive impact on the environment. In the cases of India and Pakistan, trade openness coefficients are significant at 1% only for negative shocks. For example, in India, an increase in trade openness has no significant impact on carbon dioxide emissions, while a reduction by 1% increases carbon dioxide emissions by 2.209%. Otherwise, a recent study by Shahbaz et al. [94] reports that the discussion on the energy-led growth of India necessitates the cross-border movement of resources, which influences the carbon dioxide emissions pattern. As the Indian import portfolio was majorly dependent on crude oil, the import substitution policies have reduced the import of crude oil and other petroleum products and, consequently, the level of carbon dioxide emissions.

The results for short-run trade openness coefficients, for positive and negative shocks, are significant in Bangladesh and India. Moreover, positive and negative shocks perform

considerably differently in Bangladesh. For example, a positive shock (0.179 \*\*\*) impact is greater than a negative shock (−0.106 \*\*), which demonstrates that positive shocks have more profound effects than negative shocks. This proves the significant impact of trade openness on the environment in the short run. But in 2021, Sharma et al. [19] published a paper describing the importance of importing innovative solutions to reduce environmental degradation in the long run. Domestic enterprises will try to import innovative technological solutions to improve their energy efficiency and reduce their carbon footprint.

Finally, we examine the stability of the model. Table 6 presents the diagnostic tests for serial correlation, heteroscedasticity, normality, and Ramsey. The diagnostic tests of the ARDL model indicate problems with serial correlation in all countries, heteroscedasticity in Sri Lanka, and non-linearity in Bangladesh and India. However, we found no serial correlation, heteroscedasticity problem, or normality problems in the NARDL. This diagnostic test confirmed that the NARDL was more appropriate than the ARDL model.


**Table 6.** Diagnostic checks of the ARDL and NARDL tests.

Sources: The authors' estimation. Note: They are *p* values.

#### **5. Conclusions and Recommendations**

In recent years, environmental pollution has become a global threat. In this study, we attempted to establish the short-run and long-run relationships among environmental degradation, economic growth, energy consumption, and trade openness in South Asian countries. Additionally, we verified the hypothesis that trade openness significantly impacts carbon dioxide emissions in South Asian countries. To do so, we used annual data for four South Asian countries (India, Bangladesh, Sri Lanka, and Pakistan) covering the period between 1971 and 2014. Our selection of countries for the study was based on the availability and uniformity of data in that period. We used the linear ARDL and non- linear ARDL (NARDL) model, which allowed us to analyze the impact of positive and negative shocks in trade openness on CO2 emissions. Both methods show the long-run equilibrium relationship between environmental degradation, economic growth, energy consumption, and trade openness. The empirical outcome shows that the environmental Kuznets curve holds for India and Pakistan out of the four analyzed countries.

In the NARDL model, the long-run coefficients for squared GDP per capita are statistically significant and negative for India and Pakistan, while for Sri Lanka they are statistically significant and positive. Bangladesh's squared GDP per capita is negative but not statistically significant. According to the environmental Kuznets curve, the coefficients for India and Pakistan indicate that environmental quality is expected to improve as income increases in the long run. The estimated long-run coefficients associated with energy consumption in the ARDL and NARDL models surprised us. They are statistically significant only in Pakistan. This indicates that energy consumption significantly aggravated environmental degradation only in Pakistan. This may be associated with poor institutional quality due to political instability in Pakistan.

The most interesting finding was that the long-run coefficients associated with trade openness shocks, both negative and positive, significantly impact CO2 emissions only in Sri Lanka. These impacts are based on the scale effect. On the other hand, the results for short-run trade openness coefficients, for positive and negative shocks, are significant in Bangladesh and India. In Bangladesh, positive shock increases carbon dioxide emissions, while negative shock decreases them. However, positive and negative shocks in India reduce environmental pollution. These research results did not support the hypothesis that trade openness significantly impacts carbon dioxide emissions in South Asian countries.

This study has some policy implications. But first, we assumed that if the environmental Kuznets curve is confirmed over a long period in India and Pakistan, there is a high probability that this relationship will exist for a long period. Then we can propose some recommendations. South Asian countries' governments require adequate policy directions to use clean energy while producing output and generating income. Like other low- and middle-income countries, they have limited environmental regulatory capacity. Due to poverty, low-income populations rely on timber wood for food and heating in the winter, causing significant pollution. The region's reliance on fossil fuel energy consumption is not environmentally friendly for long-term development. The consensus believes that developing renewable energies, including wind, solar, and hydroelectric power plants, will replace the infrastructure powered by fossil fuels.

With increased income in this region, governments should prioritize green growth, which is critical for sustainable development. Such actions have been taken in the past. For example, the Pradhan Mantri Ujjwala Yojana (PMUY) is a flagship scheme of India launched on 1 May 2016, by Hon'ble Prime Minister Shri Narendra Modi. The program aims to make clean cooking fuels such as LPG available to rural and deprived households that would otherwise rely on traditional cooking fuels such as firewood, coal, or cow-dung cakes. From a practical point of view, the Indian government should focus on maintaining an affordable price for LPG cylinders, along with taking more steps toward poverty reductions and keeping inflation at a desirable level, especially nowadays when its rate is high. Otherwise, poor people will revert to traditional food preparation methods, which can cause severe health and environmental problems. Consequently, Ujjwala Yojana policy paralysis may occur, leading to increased carbon dioxide emissions.

To combat environmental pollution, the governments in South Asian countries should promote and subsidize green energy by increasing their R&D spending, among others. The fifth-largest economy in the world, India, should take the lead in reducing pollution in the region. Usually, as income levels rise, so does the demand for a cleaner environment, putting pressure on the government to enact stricter environmental regulations. Governments should focus on developing advanced technology, implementing strict environmental policies, and introducing carbon pricing for polluting industries to contribute to sustainable development. Policy-makers should implement some measures to raise environmental standards without lowering income and output levels. Additionally, the financial sector should support companies and households that use environmentally friendly projects to reduce pollution. These findings should be helpful both to policy-makers when developing environmental and trade policies in the South Asian region, and practitioners. There is also a need for more and more awareness to be created among the students at primary, secondary, and tertiary education levels for effective energy utilization and moving toward green energy. All these efforts may provide desirable outcomes. We assume that the success or failure of any policy depends on people's acceptance or rejection of a policy. Therefore, collective efforts are required to reduce pollution.

Although our study has some limitations, it has the scope for further research. The first limitation refers to the sample size. Based on data availability, we examined annual data only for four South Asian countries from 1971–2014. Second, the analysis uses a limited number of factors determining economic growth and environmental degradation. We recommend that other essential variables, such as institutional quality, financial sector development, and urbanization should be considered to understand the relationship between energy use and CO2 emissions in South Asian countries. Moreover, this study did not examine the specific effects of renewable and non-renewable energy sources on

emissions in South Asian countries. Finally, our research can act as a baseline study for other South Asian countries, as the issues discussed pertain to most developing countries. Therefore, the policy recommendations discussed in the study can be generalized.

**Author Contributions:** Conceptualization: B.J., A.K.D., P.K. and A.V.G.; methodology, P.K. and B.J.; software, P.K. and B.J.; formal analysis, P.K., B.J. and A.V.G.; data curation, P.K., B.J. and A.V.G.; writing—original draft preparation, B.J., A.K.D. and A.V.G.; writing—review & editing, B.J.; project administration and funding acquisition, B.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** The APC was funded by the John Paul II Catholic University of Lublin.

**Data Availability Statement:** Publicly available datasets were analyzed in this study. This data can be found here: https://data.worldbank.org (accessed on 28 September 2022).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; collection, analyses, or interpretation of data; writing of the manuscript; or the decision to publish the results.
