The influence of financial openness, trade openness, and energy intensity on ecological footprint: revisiting the environmental Kuznets curve hypothesis for BRICS countries

Abstract

This study aims to examine the impact of economic growth, financial openness, trade openness, and energy intensity on the ecological footprint of BRICS countries for the period 1996–2016 in the framework of the environmental Kuznets curve (EKC). In the research phases, the effects of financial openness and trade openness on ecological footprint were examined both individually and as a whole using three models. The results indicate that the EKC hypothesis is not valid in all BRICS countries. Specifically, the individual results demonstrate that the EKC model using financial openness is valid only for India, while the EKC model using trade openness is valid both for India and South Africa. Furthermore, financial openness has reduced environmental pollution in India and South Africa. Trade openness has reduced environmental pollution in China and India, while it has increased in South Africa. Lastly, energy intensity has increased environmental pollution in all countries except Russia for both models. Overall, policy-makers should develop policies to reduce energy intensity in BRICS countries.

Introduction

At present, as a result of liberalization policies, borders between countries have disappeared economically, and international capital has been transferred from developed countries to developing countries. As a result, although world economic growth has produced positive results in quantitative terms, the pressure on the environment has increased, and the carrying capacity of nature has been exceeded. According to the World Wild Fund (WWF), despite the 27% increase in world biological capacity in the last 50 years, the ecological footprint (EF) has increased by 190% in the same period (WWF 2018). Due to this gap between ecological footprint and biocapacity, the world is alarmingly undergoing rising temperatures in various regions and extreme weather conditions. Human increased pressure on the world (footprint increase) and lagging biocapacity have caused the emission of carbon not to be absorbed. The resulting global warming from this situation has reduced biological diversity, disrupted the biological organism cycle, and led to natural disasters. As can be seen in Fig. 1, global temperature anomalies have increased in parallel with global GDP increase in the last 50 years. Environmental problems, once ignored due to factors such as increasing economic growth, industrialization, and trade, have grown over time and threatened the whole world.

Fig. 1

Global GDP and global temperature anomalies (1960–2018). Note: a and b show the global GDP and global temperature anomalies, respectively

According to the environmental Kuznets curve (EKC), which explains the relationship between economic growth and environmental degradation, countries focus on production and consumption rather than environmental problems in the first phase of economic growth. In parallel with rising per capita income, environmental degradation problems arise; then, after a point called “threshold value,” environmental quality increases with increasing per capita income. This relationship is represented by an inverted U-shaped curve. Economic growth, which is based on the intensive use of natural resources in its initial stage, reveals environmental problems. Once economic development has been achieved, technology-based projects with high added value are put into operation, and the service sector is developed. This transformation increases environmental awareness, enables the implementation of practices that protect the environment, and contributes to the reduction of environmental degradation (Dinda 2004). Many studies examining the relationship between economic growth and environmental degradation take carbon dioxide emissions as an indicator of environmental degradation (Liu et al. 2017; Dávalos 2016). Recent studies using ecological footprint as an indicator of environmental degradation have gained popularity (Yasin et al. 2020). The ecological footprint is a more comprehensive indicator, as it takes cropland, grazing land, fishing grounds, forestland, carbon footprint, and built-up land into account (Yilanci et al. 2019). EKC appears to have taken a new form as a result of additional variables that had an impact on its slope and shift. It is thought that pollution in a country is not only dependent on income but also on factors such as financial openness, trade openness, urbanization, and energy intensity.

Studies examining the relationship between financial openness and environmental degradation are relatively limited in number. They suggest that financial openness accelerates the inflow of foreign direct investment to a country. Subsequently, R&D activities can be developed and environmental degradation prevented by using technologies that provide high energy efficiency (Tamazian et al. 2009; Aung et al. 2017). According to another view, financial openness increases loan supply and decreases loan costs, which stimulates the investments and consumption made by companies and households as well as economic activity. Therefore, the demand for fossil fuels increases, while environmental deterioration increases (Koengkan et al. 2018). In short, the increase in financial resources is that the country supports current non-environmentally friendly production and energy consumption practices (Bass et al. 2019).

Law (2009) argues that developing countries can prosper by implementing financial openness and trade openness policy together. Because BRICS countries that constitute the research area of this study are also in the status of developing countries, we have added trade openness to our variables. According to Grossman and Krueger (1991), trade openness affects the environment in three ways. The first effect is called the “scale effect,” which indicates a positive relationship between trade openness and environmental degradation. Whenever trade and financial openness cause growth in the economy, energy demand increases, which pollutes the environment. The second effect is the “compositional effect,” which is the result of changes in trade policies. Countries tend to specialize in sectors where they can gain competitive advantages in global markets. When specific internationally traded goods are harmful to the environment, increasing their production results in further pollution. The third effect is the “technique effect,” which is explained by trade liberalization and foreign investment. Foreign investors engage in developing countries by using modern technology. Considering that modern technologies are cleaner than older technologies, positive results follow for the environment. At the same time, investments increase the level of income in these countries, which are turning to more environmentally conscious consumption and production with an increase in welfare. Furthermore, institutions such as the World Trade Organization, which determines the direction of international trade, impose restrictions on the import of environmentally harmful goods and provide control over environmental degradation (Cole 2004).

As it is one of the primary inputs used in the energy production process, regardless of the development level of countries, energy supply needs to be sustained to maintain and improve the current production level and prosperity (Alam et al. 2016). Energy demand in the world is supplied from two sources: renewable energy (hydraulic, solar, wind, geothermal, biomass, wave tide, and hydrogen) and non-renewable energy (fossil and nuclear). Consumption of non-renewable energy sources increases carbon dioxide emissions and causes global warming. In parallel with population and economic growth in developing countries, 90% of energy demand growth will result from developing countries until 2035 (OECD, 2011). On the other hand, considering that countries producing fossil fuels are the minority, diversifying energy consumption and increasing energy efficiency will decrease the energy-based vulnerability of developing economies. Reducing the energy intensity of the economy in any country will thus be beneficial in making energy resources sustainable and minimizing environmental degradation. As seen in Fig. 2, although the energy intensity of the economy is decreasing in the world and BRICS countries, that of BRICS countries is above the world average. The fact that BRICS countries have energy-intensive industries explains Russia’s petrochemicals and why other BRICS countries have heavy industry, mining, and manufacturing industries.

Fig. 2

Energy intensity of GDP

The reason for selecting BRICS countries in this study is that, as can be seen from Fig. 3 in the Appendix, their GDP is approximately 23% of the world’s GDP, with 19.851 trillion dollars and 41.50% of the total world population in 2020. This group stands out as the locomotive of the world economy, as approximately 45% of global economic growth and 20% of global trade are sourced from BRICS economies (New Development Bank 2018). On the other hand, while the total ecological footprint of BRICS countries constituted 30% of the world’s total economic footprint in 1996, this rate increased to 40% in 2016 (ecological footprints in BRICS countries are shown in Fig. 4 in the Appendix). Accordingly, China has the highest amount of ecological footprint in BRICS countries, followed by South Africa, Brazil, Russia, and India. Furthermore, BRICS countries consume approximately 40% of the primary energy resources in the world. The energy consumption of member states is mainly from fossil sources, which explains why BRIC countries account for 38% of the world’s CO₂ emissions following 2014 data (Wang et al. 2020). In a report of the World Bank published in 2019, BRICS countries were expected to increase their economies by 5.1% and 5.3% between 2019 and 2021 (Global Outlook 2019). The mentioned statistical information explains the importance of the variables used in the study (economic growth, financial openness, trade openness, and energy intensity) for BRICS countries. Considering that the economic growth and industrialization of BRICS countries are based on energy-intensive sectors (construction, mining, and manufacturing), the environmental degradation caused by these countries cannot be ignored. However, BRICS members are doing some work to improve environmental quality. At the meeting of BRICS leaders in Xiamen in September 2017, they declared the need for improvement of environmentally sensitive technologies and urban environmental sustainability and result-oriented cooperation of member countries on environmental issues (Xiamen Declaration, 2017). This study explores the ecological footprint determinants of BRICS economies to help policy-makers make better decisions in reducing ecological footprints, improve environmental quality, and take measures to ensure sustainable development. The study thus aims to contribute to the increasing environmental goals of the international community since the Kyoto Protocol was adopted in 1997.

The present investigation sets out to explain the relationship between ecological footprint, economic growth, financial openness, trade openness, and energy intensity within the framework of the EKC hypothesis for BRICS countries. The study is thus expected to contribute to the existing literature at several different points. Firstly, besides the combined effect of trade openness and financial openness on the environment, individual effects are explained with the help of two different models. Consequently, we use three different models to examine the EKC hypothesis. Secondly, according to the best knowledge of the authors, there is no study in the literature where financial openness, trade openness, and energy intensity are analyzed within the framework of the EKC hypothesis for BRICS countries, and this study fills this existing gap. Thirdly, the panel methods used in this study take cross-sectional dependence and heterogeneity into account for reliable results. Furthermore, we use two different estimators for robust results.

The rest of the study is as follows: Literature review section assesses the relevant literature; Data, models, and methodology section presents data, models, and methods; and Empirical results section shows empirical results. Discussions section introduces the discussion, and Conclusion and policy implementations section illustrates the conclusion and policy implementations.

Literature review

In line with the concerns of the real sector, the financial sector, and policy-makers about declining nature and biodiversity and the impact of environmental risks on financial and macroeconomic performance and sustainable development goals, researchers also appear to have an interest in these issues. The validity of the EKC hypothesis, which explains the relationship between economic development and environmental degradation, has been extensively investigated. According to the EKC hypothesis, in the first stage of economic growth, pollution emissions increase; however, after reaching a certain income level, environmental awareness of the society increases and pollution decreases (Dasgupta et al. 2002). Many studies take carbon emissions as an indicator of environmental degradation (Sarkodie and Ozturk 2020; Gorus and Aydin 2019; Shahbaz et al. 2019; Hanif 2018; Moosa 2017; Kang et al. 2016; Shafiei and Salim 2014; Baek and Kim 2013; Ahmed and Long 2013). The ecological footprint is considered as a more comprehensive and essential measure of environmental degradation, as it takes into account the total biocapacity required for the production of resources consumed in a country and the absorption of pollution caused by human activities. In order to provide a complete perspective, this study aims to use the ecological footprint as an indicator of pollution in analyzing the relationship between GDP growth, financial openness, trade openness, energy intensity, and environmental degradation for BRICS countries. The literature is examined in four parts.

EKC hypothesis with carbon emissions

Among the authors investigating EKC using carbon emission, Jalil and Feridun (2011) explained that the EKC hypothesis was valid between 1953 and 2006 in China with the help of the ARDL model. According to the study, while financial development had the effect of reducing carbon emissions in China, energy consumption and trade openness had an increasing effect on carbon emissions. Bölük and Mert (2015) found with the help of the ARDL model that the EKC hypothesis was valid between 1961 and 2010 in Turkey. According to the results, the EKC was not valid in Tunisia between 1971 and 2012 (Farhani and Ozturk 2015). However, Farhani and Ozturk (2015) note that a 1% increase in financial development, trade openness, and energy consumption increased carbon emissions by 0.058%, 0.418%, and 2.355%, respectively, in the long run. Javid and Sharif (2016) explain that the EKC hypothesis was valid for Pakistan and that financial development and energy consumption had an effect on increasing environmental degradation. Still, the effect of trade openness on environmental quality was statistically insignificant. Özokcu and Özdemir (2017) state that the EKC hypothesis was not valid between 1980 and 2010 in 26 high-income OECD countries and 52 developing countries using the panel data. Lin et al. (2016) could not find any evidence for the EKC hypothesis as a result of panel data analysis for the period 1980–2011 in five African countries. Seetanah et al. (2019) found that the EKC hypothesis was valid for the period 2000–2016 in 12 selected small island developing countries. According to the panel results, economic growth, in the long run, reduced environmental degradation; however, they found that the relationship between financial development and environmental degradation was statistically insignificant. Mohammed et al. (2019) found of the ARDL model that the EKC hypothesis was valid in Venezuela between 1971 and 2013 with the help. They also found that financial development reduced environmental degradation and increases energy consumption in the long run. Rahman et al. (2020) examined the validity of the EKC hypothesis in Lithuania between 1971 and 2005. According to the ARDL model results, the EKC hypothesis was confirmed in Lithuania, as there was one-way causality detection from economic growth and trade to carbon emissions.

EKC hypothesis with ecological footprint

Among the authors who examined EKC using EF, Wang et al. (2013) used spatial econometrics techniques involving 150 countries with a population of over 1 million and found that EKC was not valid. Hervieux and Darné (2015) examined the validity of the EKC hypothesis in seven Latin American countries between 1961 and 2007 using the Johansen cointegration test and the error correction model (ECM). They determined that EKC was not valid in the specified period in the mentioned countries. Al-Mulali et al. (2015) examined 93 countries in terms of income levels within the framework of four groups (low-income, low-middle income, upper-middle-income, and high-income countries). The EKC hypothesis was not valid for low-income and low-middle-income countries, yet it was valid for upper-middle-income and high-income countries. Furthermore, while trade openness increased EF, financial development reduced EF in lower-middle-, upper-middle-, and high-income countries. Mrabet and Alsamara (2017) investigated the validity of the EKC hypothesis with the help of the ARDL model in their studies covering Qatar for the period 1980–2011. Although the EKC hypothesis was not valid when carbon emission was used as an indicator of environmental degradation, the EKC hypothesis was valid in the period specified in Qatar when EF was used as an indicator of environmental degradation. Ozcan et al. (2018) examined the validity of the EKC hypothesis in Turkey with a rolling window bootstrap causality. According to the findings, economic growth increased in all sub-periods, and the EKC hypothesis was not valid. Fakher (2019), in a study conducted in selected OPEC countries, found that, although the EKC hypothesis was valid, financial openness increased EF, and trade openness decreased EF. Ahmed and Wang (2019) found with the help of the ARDL model that the EKC hypothesis was valid in India between 1971 and 2014. Furthermore, energy consumption in India increased the ecological footprint in the long run. Dogan et al. (2019) confirmed the validity of the EKC hypothesis in MINT countries between 1971 and 2013 with the help of the ARDL model. Destek and Sarkodie (2019) confirmed the EKC hypothesis between 1977 and 2013 in 11 newly industrialized countries and found a bidirectional causality relationship between economic growth and ecological footprint. Destek and Sinha (2020) examined the validity of the EKC hypothesis. Panel mean group estimator results indicate that it was not valid between 1980 and 2014 in OECD countries; however, the authors found that the use of renewable energy and trade openness reduced ecological footprint, while fossil fuel use increased. Pata and Aydin (2020) tested the EKC hypothesis with ecological footprint for the top six hydropower energy-consuming countries and found that the EKC hypothesis was not valid in these countries. Gormus and Aydin (2020) examined the validity of the EKC hypothesis for the top 10 innovative economies and found mixed results.

EKC hypothesis with financial openness, trade openness, and energy intensity

Among the authors examining the validity of the EKC hypothesis using financial openness, trade openness, and energy intensity, Shahbaz et al. (2015) tested the EKC hypothesis using energy intensity for selected African countries between 1980 and 2012. The results demonstrate that the EKC hypothesis was valid in these countries and that energy intensity had a positive and significant impact on carbon emissions. Cetin et al. (2018) investigated the validity of the EKC hypothesis using trade openness for Turkey between 1960 and 2013. The results indicate that the EKC hypothesis was valid in Turkey and that trade openness had a positive impact on carbon emissions. Rasoulinezhad and Saboori (2018) examined the relationship between energy consumption, economic growth, carbon emissions, financial openness, and trade openness between 1992 and 2015 for the Commonwealth of Independent States (CIS) region. The results indicate that there was a unidirectional causality from economic growth, trade openness, and financial openness to carbon emissions. Koengkan et al. (2018) examined the relationship between financial openness and carbon emissions for MERCOSUR countries using data between 1980 and 2014. The results indicate that financial openness increased carbon emissions. Zafar et al. (2019) investigated the relationship between energy consumption, trade openness, and carbon emissions in the framework of EKC for emerging economics for the 1990–2015 period. According to the results, the EKC hypothesis was valid, and there was a negative relationship between trade openness and carbon emissions. Destek and Sinha (2020) tested the EKC hypothesis using trade openness for 24 OECD countries between 1980 and 2014. According to the results, the EKC hypothesis was not valid in these countries, and trade openness had a negative impact on environmental pollution. Gulistan et al. (2020) tested the validity of the EKC hypothesis using trade openness for the global data between 1995 and 2017. The results confirm the validity of the EKC hypothesis: trade openness had no statistically significant impact on environmental pollution.

EKC hypothesis for BRICS countries

Among the authors examining the validity of the EKC hypothesis in BRICS countries, Haseeb et al. (2018) argue that it was valid following the panel regression results for the period 1995–2014. Furthermore, financial development and energy consumption increased carbon emissions in the long term. Ummalla and Goyari (2020), too, demonstrated that EKC was valid in BRICS countries. Ozturk (2015) examined the EKC hypothesis in those countries between 1980 and 2013. According to the GMM (generalized method of moments) results, the EKC hypothesis was valid in Brazil, India, and South Africa. Khattak et al. (2020) stated that, according to CCEMG estimator results, EKC was valid in BRICS countries except for South Africa and India. Abdouli et al. (2018) analyzed the EKC hypothesis by including Turkey in BRICS countries for the period 1990–2014. Following the OLS conclusion, they found that the EKC hypothesis was valid in other countries except for Russia, whereas the GMM system estimator concluded that EKC was valid in BRICST. Furthermore, results from the dynamic panel estimate show that FDI inflows increased carbon emissions in BRICST countries. Zhang et al. (2019), as a result of augmented mean group (AMG) analysis, stated that the EKC hypothesis was valid between 1990 and 2015 in BRICS economies except for India. Chakravarty and Mandal (2016) examined the validity of the EKC in BRICS economies for the period 1997–2011. While the EKC hypothesis was valid as a result of fixed effect (FE) panel data analysis, the EKC hypothesis was invalid as a result of GMM analysis. Ahmed (2017) investigated the effects of economic growth, financial development, and trade openness on energy consumption in BRICS economies between 1991 and 2013. According to panel OLS results, the EKC hypothesis was valid for BRICS economies. He also found that financial development and trade openness reduced energy consumption after the threshold. Rafique et al. (2020) declared that the EKC hypothesis was valid in BRICS countries in the period 1990–2017. In their study, they found bidirectional causality among economic growth, financial development, energy use, and carbon emissions. Raza et al. (2020) state that EKC was valid in BRICS economies between 1990 and 2015. Furthermore, they found bidirectional causality between financial development and environmental degradation.

Data, models, and methodology

Data

In the course of our research, we examined the impact of financial openness, trade openness, and energy intensity on ecological footprint within the framework of the EKC hypothesis using data from 1996 to 2016. We used the data of the series including energy intensity (energy consumption/GDP) per capita, GDP per capita (constant 2010 US$), the square of real GDP per capita, the Chinn-Ito index as a proxy for financial openness, the trade (import + export)-to-GDP ratio as a proxy for trade openness, and ecological footprint (global hectares per person) as a proxy for environmental pollution. The variables ef, gdp, gdp2, ei, fo, and to denote the ecological footprint, economic growth, the square of economic growth, energy intensity, financial openness, and trade openness, respectively. The data on economic growth, energy consumption, and trade openness were collected from the World Bank database. Moreover, ecological footprint and financial openness data were taken from the Global Footprint Network database and Chinn and Ito’s (2019 study, respectively. All the data used were converted into a logarithmic form.

Models

In this study, we used two different models in the test of the EKC hypothesis to determine the individual effects of financial openness and trade openness. Moreover, we also use model 3, which contains all variables. The first model contains financial openness and is formulated as follows:

$$ \ln {ef}_t={\beta}_0+{\beta}_1\ln {gdp}_{it}+{\beta}_2\ln {gdp}_{it}^2+{\beta}_3\ln {ei}_{it}+{\beta}_4\ln {fo}_{it}+{\varepsilon}_{it} $$

(1)

where β₁, β₂, β₃, and β₄ are the coefficients of lngdp_t, lngdp_t², lnei_it, and lnfo_it, respectively. The second model contains trade openness and is formulated as follows:

$$ \ln {ef}_t={\beta}_5+{\beta}_6\ln {gdp}_{it}+{\beta}_7\ln {gdp}_{it}^2+{\beta}_8\ln {ei}_{it}+{\beta}_9\ln {to}_{it}+{\varepsilon}_{it} $$

(2)

where β₆, β₇, β₈, and β₉ are the coefficients of lngdp_t, lngdp_t², lnei_it, and lnto_it, respectively. The last model contains both trade openness and financial openness and is formulated as follows:

$$ \ln {ef}_t={\beta}_{10}+{\beta}_{11}\ln {gdp}_{it}+{\beta}_{12}\ln {gdp}_{it}^2+{\beta}_{13}\ln {ei}_{it}+{\beta}_{14}\ln {fo}_{it}+{\beta}_{15}\ln {to}_{it}+{\varepsilon}_{it} $$

(3)

where β₁₁, β₁₂, β₁₃, β₁₄, and β₁₅ are the coefficients of lngdp_t, lngdp_t², lnei_it, lnfo_it, and lnto_it, respectively. ε_t is the error term for all models. If the long-run relationship is existing in the models and the coefficients of lngdp_t and lngdp_t² are positive and negative, respectively, the EKC hypothesis is valid. The turning point is calculated using the following formula:

$$ {Y}^{\ast }={e}^{-\frac{\beta_1}{2{\beta}_2}} $$

(4)

Methodology

Cross-sectional dependence and slope homogeneity tests

Dependency between cross-section units causes biased results in panel data analysis. For this reason, second-generation panel tests and estimators (namely, unit root, cointegration, and causality) should be used in case of cross-section dependence. We used Breusch and Pagan (1980) and Pesaran (2004) cross-sectional dependence tests in this study. BP test based on the correlation between errors ${\hat{\rho}}_{ij}$ and the test statistics were calculated using the following model:

$$ CD=T\sum \limits_{i=1}^{N-1}\sum \limits_{j=i+1}^N{\hat{\rho}}_{ij}^2 $$

(5)

Pesaran (2004) criticized the fact that the power of the BP test decreases as the number of cross-section units (N) increases, and even the fact that the test cannot be used in the case of N → ∞. Hence, he suggested the following test statistics to overcome these problems.

$$ {CD}_{LM1}=\sqrt{\frac{1}{N\left(N-1\right)}}\sum \limits_{i=1}^{N-1}\sum \limits_{j=i+1}^N\left(T{\hat{\rho}}_{ij}^2-1\right) $$

(6)

$$ {CD}_{LM2}=\sqrt{\frac{2T}{N\left(N-1\right)}}\left(\sum \limits_{i=1}^{N-1}\sum \limits_{j=i+1}^N{\hat{\rho}}_{ij}\right) $$

(7)

Equation 6 presents powerful results for large N values but cannot be used when N > T. To overcome this problem, Pesaran (2004) developed the test statistics in Eq. 5. The null hypothesis states that there is no cross-section dependence in models for all test statistics. We also used the slope homogeneity tests proposed by Pesaran and Yamagata (2008). These tests are based on the DELTA tests. Pesaran and Yamagata (2008) suggest the following test statistics for slope homogeneities.

$$ \hat{\varDelta}=\sqrt{N}\left(\frac{N^{-1}\tilde{S}-k}{\sqrt{2k}}\right) $$

(8)

$$ {\hat{\varDelta}}_{adj}=\sqrt{N}\left(\frac{N^{-1}\tilde{S}-E\left({\tilde{z}}_{\hat{i}t}\right)}{\sqrt{\operatorname{var}\left({\tilde{z}}_{\hat{i}t}\right)}}\right) $$

(9)

where $E\left({\tilde{z}}_{\hat{i}t}\right)=k,\operatorname{var}\left({\tilde{z}}_{\hat{i}t}\right)=2k\left(T-k-1\right)/T+1$ and $\tilde{S}$ indicate the modified Swamy (1970) statistics. The null hypothesis indicates that there is a homogeneous slope, while the alternative hypothesis states that there is a heterogeneous slope for both test statistics.

Panel unit root and cointegration tests

We used Pesaran’s (2007) second-generation cross-sectional augmented IPS (CIPS) panel unit root test, which takes into account the cross-sectional dependence. CIPS test statistics are based on cross-sectional augmented ADF (CADF) panel unit root test, which is calculated for each section unit.

$$ CIPS\left(N,T\right)={N}^{-1}\sum \limits_{i=1}^N{CADF}_i $$

(10)

The null and alternative hypotheses of the CIPS panel unit root test indicate the unit root and stationarity, respectively. In this study, we also used Westerlund’s (2008) panel cointegration test, which takes the cross-sectional dependence into account. Westerlund (2008) proposes two different test statistics—panel and group statistics—to test cointegration:

$$ {DH}_g=\sum \limits_{i=1}^n{\hat{S}}_i{\left({\tilde{\phi}}_i-{\hat{\phi}}_i\right)}^2\sum \limits_{t=2}^T{\hat{e}}_{it-1}^2 $$

(11)

$$ {DH}_p={\hat{S}}_n{\left(\tilde{\phi}-\hat{\phi}\right)}^2\sum \limits_{i=1}^n\sum \limits_{t=2}^T{\hat{e}}_{it-1}^2 $$

(12)

where DH_g and DH_p show the group and panel statistics, respectively. The null hypothesis indicates that there is no cointegration in the panel. Conversely, the alternative hypothesis suggests that the dependent and independent variables move together in the long-run—namely, there is a cointegration in the panel. Finally, we used two different long-run estimators that took the cross-sectional dependence in the panel into account. These estimators were the augmented mean group (AMG) proposed by Eberhardt and Teal (2010) and the common correlated effects mean group (CCEMG) suggested by Pesaran (2006).

Empirical results

In a globalizing world, cooperation between countries has increased, and as a result, macroeconomic variables have become interdependent. For this reason, the first case to be tested in studies using panel data analysis is the cross-sectional dependence. Table 1 shows the cross-section dependence test results.

Table 1 Cross-section dependence and slope homogeneity tests results

The results of the three different tests indicate that all variables and models have cross-sectional dependence. The slope homogeneity test results demonstrate that all models have heterogeneous slopes. These results suggest that, in the rest of the study, cross-section dependency and heterogeneity should be taken into consideration together. We used the panel unit root test, which takes the cross-section dependence into account when determining the degree of stationary of the variables.

Table 2 indicates the CIPS panel unit root test results of the variables. According to the results, all variables have a unit root at level, while they are stationary at first differences. Accordingly, these variables show the I(1) properties. For this reason, we used the Durbin–Hausman panel cointegration test, which takes into consideration the different integration levels of the variables as well as cross-sectional dependence.

Table 2 CIPS panel unit root test results

According to the results of the panel cointegration test in Table 3, there is a cointegration for all EKC models except model 3. This result indicates that the long-term coefficients should be calculated for models 1 and 2. We used two different estimators estimating the long-run coefficients. Table 4 illustrates these coefficients for model 1. The country-based results are summarized as follows: (i) financial openness is significant and has a reducing impact on environmental pollution in India while it increases environmental pollution in South Africa, and (ii) energy intensity is significant and has a negative impact on the environment in all countries except Russia. On the other hand, the EKC hypothesis is valid only in India for both estimators.

Table 3 Durbin–Hausman panel cointegration test results

Table 4 Robustness tests: Long-run estimates for model 1

Table 5 shows the long-run coefficients of model 2. According to the results, the EKC hypothesis is valid in China, India, and South Africa using the AMG estimator, while it is valid in India and South Africa using the CCEMG estimator. Accordingly, this hypothesis is valid in India and South Africa for both estimators. For the robust results, we took into account the common result of both estimators.

Table 5 Robustness tests: Long-run estimates for model 2

The country-based results are summarized as follows: (i) trade openness is significant and has a reducing effect on environmental pollution in China and India, while it has an increasing effect on the pollution in South Africa, and (ii) for all countries except Russia, energy intensity is significant and increases environmental pollution.

Discussions

In the study, we tested the EKC hypothesis for three different models. All models contain the energy intensity variable. Accordingly, the first one (model 1) is the EKC model, which is expanded with the financial openness variable, and according to the results of this model, the EKC hypothesis is valid only for India. The turning point of the EKC hypothesis was calculated, and it was observed that the economic growth data of India has not yet exceeded the turning point. The second model (model 2) was expanded with trade openness, and the EKC hypothesis is valid for this model in India and South Africa. The turning points of India and South Africa were calculated, and it was observed that the economic growth data of India has not yet exceeded the turning point, while the economic growth data of South Africa has exceeded the turning point. The third and final model (model 3) illustrates the model expanded with both variables. The validity of the EKC hypothesis could not be tested since a long-term relationship between the variables could not be determined for this model. Our country-based results of all models are as follows: (i) financial openness has a significant and positive impact on the environment—namely, it reduces the environmental pollution in India; on the opposite, it increases the pollution in South Africa for model 1; (ii) trade openness reduces environmental pollution in China and India, while it increases pollution in South Africa for model 2; (iii) energy intensity has a statistically significant and negative impact on environment—namely, it increases environmental pollution in all countries except Russia for both models.

Our country-based results are consistent with Shahbaz et al. (2015) for energy intensity, with Cetin et al. (2018), Destek and Sinha (2020), and Gulistan et al. (2020) for trade openness, and with Koengkan et al. (2018) and Rasoulinezhad and Saboori (2018) for financial openness. On the other hand, our panel results are consistent with Destek and Sinha (2020) for model 2. Nonetheless, You et al. (2015) examined the EKC hypothesis with financial openness and found that EKC is valid for global data. However, model 1 results indicate that EKC is not valid in BRICS. According to the best knowledge of the authors, there is no study in the literature with the framework of model 3. The present study thus fills an existing gap.

Conclusion and policy implementation

In this study, we examined the validity of the EKC hypothesis using energy intensity, trade openness, and trade openness data from 1996 to 2016 for BRICS countries. To this aim, we used three different models: EKC with financial openness (model 1), model 2 with trade openness, and model 3 with both variables. According to the results of the panel explained above, the energy intensity appears to be the most crucial variable for environmental degradation in BRICS countries except Russia, which also reveals the dependence of the mentioned countries on energy. As demand for energy is a necessity for development goals, policy-makers will probably contribute to the reduction of environmental degradation by developing energy use policies that do not hinder growth and take into account the environment. Public information and education programs can be designed by taking into consideration the issues of climate change and environmentally friendly energy production and consumption, and awareness can be created in this direction. Furthermore, policy-makers should develop environmental regulatory standards to reduce environmental degradation. For this purpose, taxing the negative environmental externalities caused by economic growth in the style of the Pigou tax may negatively affect the vitality of the economic life and the production sector in BRICS, which have an extensive economic growth model. Therefore, it may be more useful to subsidize businesses and households for sectors that will provide renewable energy solutions instead of removing subsidies from environmentally harmful sectors. Households can receive subsidized loans, tax refunds, and interest rate deductions from the government to supply products that will benefit from renewable energy in daily life. Credit interest rate reduction can be provided to businesses if the sector in which they operate has a low ecological footprint. Thus, cleaner production and consumption in these countries can become the main driving force for sustainable economic growth. Furthermore, the fact that EKC is valid in India and South Africa may not necessarily compromise economic growth, especially in the long term, during the reduction of ecological footprint in these countries.

Trade policies can be redesigned to adapt to changes in energy policies, and green trade policies can contribute to improving environmental quality. The panel results indicate that trade openness increases environmental degradation in South Africa but decreases it in China and India. Further liberalization of trade in South Africa can be dangerous for the environment; therefore, tariffs can be placed on products fabricated with the help of non-environmentally friendly energy. Since the 18th National Congress in 2013, China has attached great importance to ecological civilization construction and ecological and environmental protection to achieve this. Specifically, it has switched to a policy of priority of protecting the environment rather than the priority of growth (Zhenhua 2020). India published the National Action Plan on Climate Change (NAPCC) in 2008 to alleviate climate change issues. In the Paris Agreement in 2015, India committed to reducing emissions by 33–35% by 2030 from 2005 levels and increase non-fossil-derived electricity capacity to 40% (UNFCC 2015). However, India needs at least 2.5 trillion US dollars to carry out the climate change process under this commitment by 2030 (Government of India 2016). This reveals the consistency of the mitigating impact of financial openness on environmental degradation in India, which we have reached as a result of the panel. The financial sector can provide credit opportunities to businesses that want to adopt cleaner and environmentally friendly technologies and support such investments. In contrast, as a result of the fact that financial openness increases environmental degradation in South Africa, public and private financial institutions will contribute to achieving the “technique effect” by providing financial assistance to investors who are launching environmentally friendly initiatives and promoting the purchase of energy-saving products.

Appendix

Fig. 3

GDP and population in BRICS countries in 2020. Note: a and b show the GDP and population, respectively

Fig. 4

Ecological footprint in BRICS countries. Source: Danish et al. (2020)