key: cord-0933823-rr9ldv27
authors: Wang, Qiang; Zhang, Fuyu
title: The effects of trade openness on decoupling carbon emissions from economic growth – Evidence from 182 countries
date: 2020-08-22
journal: J Clean Prod
DOI: 10.1016/j.jclepro.2020.123838
sha: eecd6840969537d143fa32fe659fa2911b8a679f
doc_id: 933823
cord_uid: rr9ldv27

The current rise of protectionism has become the main uncertainty associated with global energy, economy, and the environment. Furthermore, the decoupling carbon emissions from economic growth is crucial for implementing Intended Nationally Determined Contributions (INDCs). These INDCs would be discounted if decreasing carbon emissions would require sacrificing economic growth. This study explored the effect of protectionism (by measuring trade openness based on available data) on the decoupling carbon emissions from economic growth. For this, the heterogenous effects of trade openness on carbon emissions were investigated using in data of 182 countries from 1990 to 2015. The results show that trade openness decreased carbon emissions in high-income and upper-middle-income countries, while having no significant impact on carbon emissions of lower-middle-income countries; even worse, for low-income countries, trade openness increased carbon emissions. The heterogeneous effects of trade openness on carbon emissions indicate that trade openness positively impacts the decoupling economic growth from carbon emission in rich countries, but negatively impacts poor countries. In addition, increasing individual incomes and population distort the decoupling economic growth from carbon emissions. Renewable energy and high oil prices contributed to the decoupling economic growth from carbon emissions. These effects are similar in all countries. Targeted policy implications are presented that enable the decoupling economic growth from carbon emissions for countries with different income levels.

to achieve the goals of the Paris Agreement, investigations regarding the impact of trade openness on the decoupling of economic growth and carbon emissions is necessary to develop more effective carbon reduction policies.

In this context, the present study is dedicated to solving three key issues based on panel data between the annual periods of 1990-2015 for 182 selected countries:

(1) Is economic growth globally decoupled from carbon emissions? (2) What is the effect of trade openness on carbon emissions? (3) Is this effect heterogeneous for countries at different income levels?

To address these three issues, first, the Tapio decoupling model was applied to ascertain the decoupling between economic growth and carbon emissions. Then, the effect of trade openness on decoupling of economic growth and carbon emissions was investigated with the carbon functions. Unit root tests, cointegration tests, OLS and FMOLS estimates for panel models were adopted in carbon functions. This study established global panel and four income level panels, the results report the heterogeneous effect of trade openness on the decoupling of economic growth and carbon emissions. Effective policy implications can be drawn toward decoupling economic growth from carbon emissions, especially for countries with different income levels.

This article consists of five sections, which are are organized as follows:

Section 2 reviews and summarizes the relevant literature. Section 3 provides methods and data descriptions. Section 4 shows the results and discussion and J o u r n a l P r e -p r o o f Section 5 summarizes the main results and provides both conclusions and policy implications.

The ideal state of decoupling indicates that economic growth does not depend on the growth of carbon emissions. To accurately understand the relationship between carbon emissions and economic growth, an indicator is required that reflects the relationship between both. The concept of decoupling was first proposed by Von in 1989 and was used to describe the relationship between carbon emissions and the economy. In 2002, the OECD first used the decoupling theory to study the relationship between economic growth and carbon emissions. Thus, the decoupling model gradually emerged. Subsequently, the concepts of primary decoupling and secondary decoupling were developed (Moldan et al., 2012) . In 2005, the Tapio decoupling model began to use decoupling elasticity to describe the decoupling state (Tapio, 2005) . In the following, for research level, the Tapio decoupling model was increasingly used in departmental and national research. At the departmental level, a deep understanding of the decoupling of industrial growth and carbon emissions was achieved . In addition, decoupling process of soil erosion and human activities was investigated for the Loess Plateau of China using the concept of decoupling . In the tourism department, it has been J o u r n a l P r e -p r o o f suggested that China's tourism economy experienced negative and weak decoupling (Tang et al., 2014) . Moreover, many scholars also focused on the national level , and investigated relevant issues in specific countries, e.g., China (Yang et al., 2018) , the United States (Datta, 2019) , OECD (Chen et al., 2018) , Pakistan (Raza and Lin, 2020) , and India . In short, the Tapio decoupling model has been widely used, which indicates the maturity and adaptability of this model. This study thus used the Tapio decoupling model to identify the decoupling status of economic growth and carbon emissions. Moreover, such a comparison between countries with different income levels can help to formulate targeted carbon emission reduction policies. Therefore, further empirical research is needed at the global level.

The literature on factors affecting carbon emissions is quite rich (Al-mulali, 2011; Wang and Zhang, 2020; Zhang and Da, 2015) . The linear econometric model is the most commonly used model to study the the factors affecting carbon emissions (Jalil and Feridun, 2011) . The linear econometric model has been applied to time series data and panel data (Bhattacharya et al., 2017) . As research increases, the existing research in this field can be divided into three categories. The first category investigated the relationship between economic development and carbon emissions (Galeotti et al., 2009; Saboori et al., 2012; Selden and Song, 1994) .

The second category incorporated population and energy into the research J o u r n a l P r e -p r o o f framework of economic development and carbon emissions (Lehmann and Gawel, 2013; O'Neill and Chen, 2002; Weber and Perrels, 2000) . Age structure (Fan et al., 2006), urbanization (Martínez-Zarzoso and Maruotti, 2011) , the size of households (Poumanyvong and Kaneko, 2010) , energy prices (Rout et al., 2008) , and energy consumption (Fortes et al., 2008) were specifically investigated. The third category not only includes population and energy but also control variables such as trade and foreign direct investment (Dasgupta et al., 2001; Tamazian et al., 2009; Zhang, 2011) .

The present research is part of the third category and presents in-depth research on the impact of trade openness on carbon emissions, including individual incomes, population, oil prices, and renewable energy. Free trade helps the global economy to grow faster by increasing the trade volume and income, both in developed and developing countries. However, this growth trend is accompanied by specific environmental consequences (Shahbaz et al., 2017b) .

In general, the impact of trade openness can mainly be divided into two theories in the environmental field. The first theory assumes that the impact of trade openness on pollution is vague and can be divided into scale effect, technology effect and composition effect (Farhani et al., 2014a) . The second theory is the Pollution Haven Hypothesis (Copeland and Taylor, 2004) . Trade openness introduces foreign direct investment. Since different countries set different environmental standards, polluting enterprises will choose to produce in countries J o u r n a l P r e -p r o o f with comparatively low environmental standards, which thus become "pollution haven". Therefore, the impact of trade openness on the environment needs to be considered for specific countries.

Based on these two theories, conclusions from the literature formed four hypotheses: (1) trade openness promotes carbon emissions; (2) carbon emissions promote trade openness; (3) feedback hypothesis: Trade openness and carbon emissions interact; (4) neutral hypothesis: Trade openness is independent of carbon emissions. In the evidence supporting hypothesis (1), at the national level, it has been found that trade openness positively affects carbon emissions in the long run for Pakistan by using the vector error correction model (VECM) (Nasir and Ur Rehman, 2011) . Moreover, it has been observed that increased trade openness will increase pollution. This has been corroborated by applying the panel vector error correction model (PVECM), the fully modified ordinary least squares (PFMOLS) model, and the panel dynamic ordinary least squares (PDOLS) (Farhani et al., 2014b) . With regard to hypothesis (2), China has been studied in the context of globalization using VECM causality as well as the ARDL bounds test (Shahbaz et al., 2017a) . The causal test proved the unidirectional Granger causality of carbon emissions to trade openness. Besides, several international organizations also suggested that environmental regulations exert a serious impact on international trade. Hypothesis (3) refers to the bidirectional causality between trade openness and carbon emissions. At the transnational level, a study of 105 countries identified J o u r n a l P r e -p r o o f bidirectional causality between the global group and the middle-income group by using the panel regression model (Shahbaz et al., 2017b) . While trade openness is affected by carbon emissions, it also affects carbon emissions. Hypothesis (4) does not support the link between trade openness and carbon emissions; however, relatively little literature supporting this hypothesis. At the national level, it has been argued that it is difficult to find a causal relationship between trade and the environment by using a linear econometric model (Frankel and Romer, 1999) . At the transnational level, in the panel regression model, trade openness has been found to be not generally correlated with increased emissions when studying the effects of trade on environmental Kuznets curve (EKC) (Kearsley and Riddel, 2010) . Clearly, the results of different studies support different hypotheses. Consequently, the relationship between trade openness and carbon emissions still merits further investigation.

Although the existing literature covers a similar scope than the present work, this study contributes to previous literature in a number of notable aspects. First, this study extends the literature by incorporating of trade openness into the existing economic growth-carbon emission research framework. Renewable energy and population are used as additional variables, and a systematic study was conducted.

The conclusions also provide comprehensive policy recommendations toward the achieving decoupling of economic growth from carbon emissions. Second, this study not only includes a similarities analysis at the global level, but more J o u r n a l P r e -p r o o f importantly, investigates the differences of four income sub-panels (high-income, upper-middle-income, lower-middle-income, and low-income) on the effect of trade openness in the decoupling of economic growth from carbon emissions. Such an analysis of differences can help more countries to find effective ways to embark on the path of decoupling economic growth from carbon emissions.

This study uses Tapio decoupling model, with the following equation:

Where e(C) represents the decoupling elasticity coefficient between economic activities and carbon emissions, ΔC represents the total carbon emission change from the base period to the end period, C 0 represents the carbon emissions at the base period. ∆G represents the total GDP change from the base period to the end, and G 0 represents the base period GDP. The Tapio 

Based on previous research of ( Dong et al., 2018) (Dogan and Turkekul, 2016) , the following models were established:

In Eq. (2), n (n = 1,2, ..., 182) represents the sample country, t represents the

year, and C represents carbon emissions per capita. The estimation model is converted into a log linear econometric model: lnC nt =α 0 +α 1 lnOPEN nt +βX nt +ε nt

In Eq. (3), X represents control variables, including oil prices, renewable energy consumption, individual incomes and population. Among these, α 0 and ε nt represent the intercept and error terms, respectively, and α 1 and β represent the estimated coefficients of different variables. In Model 1, the only independent J o u r n a l P r e -p r o o f variable is lnOPEN, and in Models 2 and 3, lnOP, lnGDP, lnRE, and lnPOP are added as control variables.

.

The study of the relationship between C, OPEN, OP, GDP, RE, and POP was divided into three steps. First, the panel unit root test was used to test the stability of each variable. Second, the panel cointegration test was used to determine the long-term cointegration relationship between variables. Next, the fixed-effect OLS and FMOLS cointegration estimates were used to analyze the long-term cointegration relationship between variables. The formula of LLC test is as follows (Levin et al., 2002) :

∆Y it-L +c pi d pt +e it ，p=1,2,3

where a i , c pi , d pt , and e it represent the autoregression coefficients of the model, and the corresponding vectors of the regression parameters were p=1,2,3.

The formula of the IPS test (Im et al., 2003) is similar to that of the LLC test. In J o u r n a l P r e -p r o o f addition, the unit root test of the Fisher-PP panel, as developed by Phillips and Perron is a different unit root test (Phillips and Perron, 1988) , the expression of which is as follows:

where m, γ -1 represents the reciprocal of the normal distribution function, and X m represents the P-value of the ADF unit root test. The null hypothesis is a i = 0, which indicates that there is a unit root; if a i < 0, there is no unit root.

The cointegration test can determine whether variables that are stable at a specific level or of the same order have a long-term stable cointegration relationship.

In this study, the panel Kao test (Kao, 1999) and the panel Pedroni test (Pedroni, 2001) were used, both of which are part of the Engle-Granger method. The Pedroni cointegration test includes two alternative hypotheses: panel statistical hypotheses and outlier statistical hypotheses. The specific statistical formula is as follows:

where

The second step of the long-term cointegration relationship is a cointegration estimation. This study used the ordinary least squares (OLS) method to perform regression on Model 2, and fully modified ordinary least squares (FMOLS) to perform regression on Models 1 and 3. FMOLS is widely used for regressions (Liu et al., 2019) . Compared with OLS estimation, FMOLS estimation can correct sequence correlation and prevent pseudo regression, thus, it is a robust panel econometrics technology. In the FMOLS cointegration system, Pedroni (Pedroni, 2000) proposed the following equation.

where

] , δ αβ is the long-term J o u r n a l P r e -p r o o f covariance. In Equation (12), x and y αβ have a cointegration relationship. The long-term covariance can be decomposed into δ α = δ α o = ω α = ω α ' , where ω α represents the automatic covariance and δ α o represents the weighted sum of the covariance and ω α The FMOLS criteria are as follows:

Where y αβ * =y αβ -y α -δ 2,1,α δ 2,2,α ∆x αβ ,γ α =ω 2,1,α +δ 2,1,α o -(ω 2,1,α /ω 2,2,α )(ω 2,2,α +δ 2,2,α )

Based data availability, unbalanced panel data was obtained for 182 countries from 1990 to 2015. Compared with many previous studies, this sample offers greater coverage in terms of country and year. First, a global panel composed of 182 countries was used. Second, the estimated sample was divided into four income sub-panels based on the 2020 World Bank's country classification: low-income (LI), lower-middle-income (LMI), upper-middle-income (UMI) and high-income (HI) (World Bank). Among these, the LI group is composed of 27 countries, the LMI and UMI groups are composed of 45 and 56 countries respectively, and the HI group is composed of 54 countries (see Table A1 in Appendix A). The variable definitions are shown in Table 1 and Table 2 shows the descriptive statistics of variables. LMI (1990 LMI ( -2014 . The fluctuation was obvious compared with UMI over the same period. The reason may be that because of the desire to improve the economy, LMI countries focus more on economic development than environmental quality. However, the dependence of economic growth on energy will further stimulate carbon emissions, which is not conducive to their reduction (Wang and Su, 2020) . In addition, developed countries export industrial production to J o u r n a l P r e -p r o o f 20 LMI. The decrease of carbon emissions in HI countries is thus, in fact, paid for by developing countries (Schaltegger and Csutora, 2012) . Hence, it is difficult for LMI countries to achieve a long-term decoupling of economic growth and carbon emissions.

The decoupling status of LI countries is quite rich, and the following six states were identified: expansive coupling (17%), strong decoupling (25%) 

All variables passed the four panel unit root tests, and the results are shown in Tables 3-7. Independent of the income group, the stationarity of the variables remains stable after the first-order difference. This prompted the next cointegration analysis.

Specifically, since the test results of the four panel unit roots are sometimes inconsistent, the majority of the test results was selected. 140.6720*** 0.0005 729.1150*** 0.0000 Note: ***, **, * represent significant at 1%, 5%, and 10% inspection levels, respectively. 

The results of the panel cointegration test are shown in Table B1 

In light of the evidence of the long-term cointegration relationship among variables, regression estimates were calculated to identify reasonable environmental policies. This identification can be achieved by understanding the linear nexus between carbon emissions, trade openness, oil price, individual incomes, population and renewable energy. The OLS and FMOLS regression results are reported in Table   B6 (see Appendix B). This analysis focuses on the FMOLS estimation results. (Cole and Elliott, 2005; Grossman and Krueger, 1994; Jalil and Feridun, 2011) .

and population positively affects carbon emissions in global group. In Model 3, a 1% increase in individual incomes and population leads to 0.6107% and 0.499% decrease of global carbon emission in the long run respectively. These results match those reported by Mensah et al. (Mensah et al., 2019) , who argued that carbon emissions are likely highly correlated with individual incomes because they are by-products of industrial processes, energy consumption (direct consumption of fossil fuels and electricity), and car use. Although a number of advanced economies have decoupled their economic growth from carbon emissions in recent years (Andreoni and Galmarini, 2012) , this is not common.

The increasing population drives carbon emissions; however, this driving effect is less than that of individual incomes. This is desirable, as certain human behavior customs may directly trigger excessive energy consumption and subsequently influence environmental change. The ensuing increase of the number of private cars J o u r n a l P r e -p r o o f and construction operations has also increased energy consumption, which may be the main reason why increasing population leads to increased carbon emissions (Wong et al., 2015) .

In contrast, higher oil prices and renewable energy hinder global carbon emission regardless of the income group, and the effect of energy generation is stronger.

Regarding renewable energy, a statistical inverse relationship exists with carbon emissions. In Model 3, a 1% increase in the proportion of renewable energy consumption yields a corresponding 0.1568% decrease in global carbon emissions over the considered period. These results confirms that the consumption of renewable energy effectively decreases carbon emissions (Cai et al., 2018) .

A 1% increase in international crude oil prices leads to a 0.0642% decrease of global carbon emissions in the long run, which is supported in by previous research (Winchester and Ledvina, 2017) . Changes in oil prices affect energy consumption since oil is an important component of the global energy consumption structure, and oil is also the main source of carbon dioxide. It should be noted that the turmoil in the financial market caused by the decline in oil prices has negatively impacted the economies of many crude oil producing countries. In the long run, this decline will accelerate the carbon dioxide emissions of these countries, which will undoubtedly bring difficulties to the global carbon emission reduction work.

J o u r n a l P r e -p r o o f 

This study also investigated four income groups. Comparative analysis of the results between different income groups has practical significance for many countries when formulating targeted emission reduction policies. Four carbon functions of different income groups were obtained from Model 3, as shown in Table 8 . Figure 4 shows the distribution of the effects for different income levels. The effects of oil prices, economic growth, population, and renewable energy on carbon emissions are consistent among all four income groups, however, the effect of trade openness is clearly inconsistent. Therefore, this section discusses the heterogeneous effects of trade openness on carbon emissions across four income groups. This is insightful, as it indicates that trade openness has a positive impact on carbon emission reduction in both the HI and UMI countries, but not significant on carbon emissions in LMI countries, and even a negative impact on LI countries. This indicates that with increasing income level, the impact of trade openness on carbon emissions also changes. This supports a previously reported the view (Shahbaz et al., 2017b) , where trade openness contributes to carbon emissions at all income levels but exerts with varying influence on different panels. The heterogeneous effects of trade openness on carbon emission suggest that trade openness improves the environment of rich countries, but aggravates the environmental pollution of poor countries. This is J o u r n a l P r e -p r o o f in line with the recognized phenomenon of carbon transfer in the process of international trade (Essandoh et al., 2020) . Environmental standards in LI countries are generally lower than in other countries with higher income levels, and the environmental management system is deficient. Therefore, with the formation of global supply chains, developed countries either transfer or outsource high-carbon emission industries to LI countries (Baumert et al., 2019) . This supports the views of Grossman and Krueger (Grossman and Krueger, 1994) who pointed out that dirty industries in developing countries tend to cause a heavy share of pollutants. Most developing countries are LI countries. The "Pollution Refuge Hypothesis" was verified (Zhang et al., 2017) and trade implied carbon emissions were also assumed as a key way to transfer pollution (Rafindadi et al., 2018) . Hence, with decreasing income levels, the impact of trade openness on the environment changes from positive to negative. openness has a positive impact on the decoupling of economic growth from carbon emissions in rich countries, but a negative impact in poor countries.

Increasing individual incomes and population distort the decoupling of economic growth from carbon emissions, and the distortion of individual incomes is stronger than population growth. In contrast, renewable energy consumption and high oil prices contributed to the decoupling of economic growth from carbon emission, and the contribution of renewable energy is stronger than that of high oil prices. Moreover, these effects are similar in all countries independent of income levels. J o u r n a l P r e -p r o o f Appendix A. Table A1 . List of sample countries. Note: ***, **, * represent significant at 1%, 5%, and 10% inspection levels, respectively. J o u r n a l P r e -p r o o f Note: ***, **, * represent significant at 1%, 5%, and 10% inspection levels, respectively. Note: ***, **, * represent significant at 1%, 5%, and 10% inspection levels, respectively. Note: ***, **, * represent significant at 1%, 5%, and 10% inspection levels, respectively. Note: ***, **, * represent significant at 1%, 5%, and 10% inspection levels, respectively. 

Oil consumption, CO2 emission and economic growth in MENA countries

Decoupling economic growth from carbon dioxide emissions: A decomposition analysis of Italian energy consumption

Global outsourcing of carbon emissions 1995-2009: A reassessment

The dynamic impact of renewable energy and institutions on economic output and CO2 emissions across regions

Nexus between clean energy consumption, economic growth and CO2 emissions

Decomposition and decoupling analysis of CO2 emissions in OECD

FDI and the Capital Intensity of 'Dirty' Sectors: A Missing Piece of the Pollution Haven Puzzle

Trade, Growth and the Environment

Pollution and Capital Markets in Developing Countries

Decoupling and demand-side management: Evidence from the US electric industry

CO 2 emissions, real output, energy consumption, trade, urbanization and financial development: testing the EKC hypothesis for the USA

Does natural gas consumption mitigate CO2 emissions: Testing the environmental Kuznets curve hypothesis for 14 Asia-Pacific countries

Linking international trade and foreign direct investment to CO2 emissions: Any differences between developed and developing countries?

Analyzing impact factors of CO2

CO2 emissions, output, energy consumption, and trade in Tunisia

What does MENA region initially need: Grow output or mitigate CO2 emissions?

Long term energy scenarios under uncertainty

Does Trade Cause Growth

Economic Growth and the Environment

Testing for unit roots in heterogeneous panels

The impact of growth, energy and financial development on the environment in China: A cointegration analysis

Spurious regression and residual-based tests for cointegration in panel data

A further inquiry into the Pollution Haven Hypothesis and the Environmental Kuznets Curve

Why should support schemes for renewable electricity complement the EU emissions trading scheme?

Unit root tests in panel data: asymptotic and finite-sample properties

Does income inequality facilitate carbon emission reduction in the US?

The impact of urbanization on CO2 emissions: Evidence from developing countries

Analysis on the nexus of economic growth, fossil fuel energy consumption, CO2 emissions and oil price in Africa based on a PMG panel ARDL approach

How to understand and measure environmental sustainability: Indicators and targets

Environmental Kuznets Curve for carbon emissions in Pakistan: An empirical investigation

Demographic Determinants of Household Energy Use in the United States

Fully modified OLS for heterogeneous cointegrated panels

Fully modified OLS for heterogeneous cointegrated panels, Nonstationary panels, panel cointegration, and dynamic panels

Testing for a unit root in time series regression

Does urbanization lead to less energy use and lower CO2 emissions? A cross-country analysis

The effects of FDI and energy consumption on environmental pollution in predominantly resource-based economies of the GCC

Decoupling and mitigation potential analysis of CO2 emissions from Pakistan's transport sector

Impact assessment of the increase in fossil fuel prices on the global energy system, with and without CO2 concentration stabilization

Economic growth and CO2 emissions in Malaysia: A cointegration analysis of the Environmental Kuznets Curve

Carbon accounting for sustainability and management. Status quo and challenges

Environmental Quality and Development: Is There a

Kuznets Curve for Air Pollution Emissions

THE IMPACT OF GLOBALIZATION ON CO2 EMISSIONS IN CHINA

Trade opennesscarbon emissions nexus: The importance of turning points of trade openness for country panels

Does higher economic and financial development lead to environmental degradation: Evidence from BRIC countries

Decoupling indicators of CO2 emissions from the tourism industry in China

Towards a theory of decoupling: degrees of decoupling in the EU and the case of road traffic in Finland between

Is China's economic growth decoupled from carbon emissions?

Is decoupling economic growth from fuel consumption possible in developing countries? -A comparison of China and India

The effects of urbanization and industrialization on decoupling economic growth from carbon emission -A case study of China

Drivers of decoupling economic growth from carbon emission -an empirical analysis of 192 countries using decoupling model and decomposition method

Is energy transition promoting the decoupling economic growth from emission growth? Evidence from the 186 countries

Does increasing investment in research and development promote economic growth decoupling from carbon emission growth? An empirical analysis of BRICS countries

Modelling lifestyle effects on energy demand and related emissions

Decoupling soil erosion and human activities on the Chinese Loess Plateau in the 20th century

The impact of oil prices on bioenergy, emissions and land use

Can energy efficiency rating and carbon accounting foster greener building design decision? An empirical study

Whether China's industrial sectors make efforts to reduce CO2 emissions from production? -A decomposed decoupling analysis

The decomposition of energy-related carbon emission and its decoupling with economic growth in China

The impact of financial development on carbon emissions: An empirical analysis in China

A multi-regional input-output analysis of

The authors would like to thank the editor and these anonymous reviewers for their helpful and constructive comments that greatly contributed to improving the final version of the manuscript. This work is supported by National Natural Science