Does trade openness convey a positive impact for the environmental quality? Evidence from a panel of CIS countries

ABSTRACT

The Environmental Kuznets Curve hypothesis suggests that sustainable economic growth can be achieved in transitional countries after the threshold level of per capita income. And trade openness is also one of the critical factors to help transitional economies reduce carbon dioxide emissions and economic growth simultaneously through the combination of scale, composition, and technique effects. This paper is designed to explore the effect of trade openness on the environmental quality and investigate the existence of the environmental Kuznets curve hypothesis for a panel of CIS countries over the period of 2000–2013. The two-equation model is used to estimate the direct effect of trade openness on CO2 emissions and the indirect effect via per capita income. The instrumental variable techniques are employed to consider for endogeneity of per capita income and trade openness to estimate the indirect effect, and GLS analysis is conducted to estimate the direct effect of trade openness on CO2 emissions. The result shows that trade openness increases CO2 emissions directly while indirectly decreasing it due to its negative effect on per capita income. This study provides multiple policy implications for the sample countries to help them achieve sustainable economic growth while improving environmental quality.

KEYWORDS:

• Environmental Kuznets curve(EKC)
• economic growth
• environmental quality
• trade openness
• The Commonwealth of Independent States (CIS)

Introduction

The sustainable economic growth of transitional countries is very important for international society. Global efforts to curb climate change depend not only on the achievements of advanced nations, but also less developed ones. However, this is a huge challenge for transitional economies. The structural transformations occurring in transitional countries leads to rapid economic growth, which requires a substantial amount of energy, much of which is produced from traditional energy sources.

The Environmental Kuznets Curve (hereafter, EKC) hypothesis suggests that sustainable economic growth can be achieved in transitional countries after a certain threshold level of per capita income (Agras and Chapman 1999; Al-Iriani 2006; Apergis and Payne 2009; Arouri et al. 2012; Ozcan 2013; Pao and Tsai 2011; Zhang and Cheng 2009). Trade is also one of the critical factors that can help transitional economies reduce carbon dioxide emissions and maintain economic growth simultaneously through the combination of scale, composition, and technique effects (Grossman and Krueger 1993). Therefore, many studies have been carried out on trade openness and environmental quality based on the EKC framework (Al-Mulali 2012; Sharma 2011; Tamazian and Rao 2010). However, the results of the relationship between trade openness and environmental quality are not consistent. While Kohler (2013) and Shahbaz, Lean, and Shabbir (2012) reported the positive impact of trade openness on environmental quality, Al-Mulali (2012) and Jalil and Mahmud (2009) reported negative impact. In the study of Sharma (2011), trade openness was not significant at any income level in the 69 countries studied.

This inconsistency may be attributed to the difference between the direct and indirect effects of trade openness on environmental quality (Dean 2002). Trade openness enables CO₂ emissions to increase and decrease at the same time through economic growth and technological effect. These multiple simultaneous effects on environmental quality can make efficient estimates difficult. This study estimates the total effect by dividing and estimating the path of the effect of trade openness on environmental quality through both direct and indirect paths.

In order to efficiently estimate the relationship between trade openness and environmental quality of transitional countries, consideration of the institutional factors is needed (Al-Mulali and Ozturk 2015; Torras and Boyce 1998) because the institution provides the rules of the game and its changes affect society as a whole. However, previous studies, except for Tamazian and Rao (2010) and Al-Mulali and Ozturk (2015), have paid little attention to institutional factors. This study uses the Assessment of Transition Qualities Index provided by EBRD and the World Governance Indicator provided by World Bank as institutional variables to estimate the relationship between trade openness and environmental quality.

The analysis target of this study is the Commonwealth of Independent States (hereafter, CIS), which includes countries from the former Soviet Union, and is a prime example of a transitional economy. Except for Tamazian and Rao (2010)’s study of 24 transition economies, the relationship between trade openness and environmental quality of CIS countries is not receiving the attention of researchers even though the CIS has a significant presence in the global energy market as a center of oil and natural gas supplies. According to Apergis and Payne (2010), the CIS countries undertook significant efforts to build their market-based economy. This encourages these countries to engage more in international trade and to develop new trade patterns, however their specific geographical location, availability of natural resources, and ongoing structural changes greatly affects their trade performance. Moreover, the abundance in natural resources allows these countries to meet their growing energy needs via cheap fossil fuels consumption, keeping their economies highly energy-intensive. Hence, it is necessary to pay closer attention to this region in order to deeply understand CO₂ emissions trends in the long term.

The aim of this paper is to empirically investigate the relationship between trade openness, economic growth, and environmental pollution to estimate the direct, indirect, and total effects of trade, and to test the EKC hypothesis on a panel of CIS countries during the period of 2000–2013. Partly following the model suggested by Cole (2007), we first estimate the effect of trade openness on per capita income of these countries and then estimate the overall impact of trade on environmental quality.

This paper can contribute to the previous literature in two ways. First, this research can provide policy implications for the sustainable economic growth of CIS countries in transition. There is a lack of literature that tests the EKC hypothesis for this region, especially literature that includes trade openness as an additional variable. This study can help fill this data gap and draw more attention to the CIS countries. Moreover, to the best of our knowledge, this is the first study that estimates both the direct and indirect effects and the total effect of trade openness on CO₂ emissions for this region. Thus, it shows the necessity of changes in their trade policy and also the necessity of special policy recommendations and regulations that should be implemented for sustainable economic growth and high environmental quality in the CIS countries. Second, some of the existing research does not directly address the problem that trade openness may determine income and environmental outcomes simultaneously. Therefore, this study will help solve the possible bi-directional causality (endogeneity) between the macroeconomic variables. It also more efficiently estimates the relationship between EKC, trade openness, and environmental quality, including institutional variables that were omitted from the existing model. It will contribute to the further research for transitional economies.

The rest of the paper is organized as follows. Section 2 presents a brief introduction of CIS countries and literature review related to this topic. Section 3 includes the data description and methodology explanation. Section 4 provides the results and discussion, and Section 5 introduces the conclusions reached through this research along with some policy implications.

Literature review

The relationship between trade openness and environmental quality has been widely investigated in the Environmental Kuznets Curve (EKC) framework, which is derived from the original Kuznets curve that studies the relationship between per capita income and inequality (Grossman and Krueger 1993, 1995). The EKC hypothesis suggests that carbon dioxide emissions increase with the increase of per capita income until reaching a turning point (threshold level of per capita income), after which carbon dioxide emissions begin to decrease. This is explained by the fact that increased income forms a country’s population appreciation of the environment and, therefore, the demand for better environmental quality. This in turn encourages the government to develop and apply different policies to reduce pollution. The validity of the EKC hypothesis has been checked in numerous studies (Agras and Chapman 1999; Akbostancı, Türüt-Aşık, and Tunç 2009; Al-Iriani 2006; Ang 2007; Apergis and Payne 2009; Arouri et al. 2012; Dinda and Coondoo 2006; Lee and Lee 2009; Narayan and Smyth 2008; Ozcan 2013; Pao and Tsai 2011; Soytas and Sari 2006; Zhang and Cheng 2009). Also, research has begun to apply a multivariate framework to investigate the relationship between economic growth and environment more efficiently by incorporating trade openness, which can affect not only economic growth but also environmental quality.

The impact of trade openness on environmental quality can be divided into the scale, composition, and technique effects (Grossman and Krueger 1993). The scale effect means that economic growth from trade increases pollution emissions by increasing energy consumption and cross-border transportation service. The composition effect is caused by specializing in an industry for competitive advantage. This effect varies depending on the source of competitive advantage. The technique effect is due to the flow of technology toward less developed countries through trade openness. These effects have been investigated by Antweiler, Copeland, and Taylor (2001) and Copeland and Taylor (2004).

The inverted U-curve for the EKC is well explained by the composition effect via structural changes in the economy and by the technique effect through income-induced demand for better environmental quality and strict environmental regulations. According to Grossman and Krueger (1995) the downward sloping area of the EKC might be explained by the fact that “as countries develop, they cease to produce certain pollution intensive goods and begin instead to import these from other countries with less restrictive environmental protection laws.” Suri and Chapman (1998) attempt to investigate these structural changes by including the ratio of manufactured income to domestic manufacturing production, and the same ratio for exports, as variables in the EKC. They determined that both developed and developing countries increase their energy requirements through the export of manufactured goods. However, the growth is observed to be higher in developing countries. Moreover, developed countries reduce their energy requirements via importation of manufactured goods. Thus, for developed countries, export of manufactured products is one of the reasons for the upward sloping area of the EKC, while the importation of manufactured goods coincides with downward slope.

However, the results of previous research on the relationship between trade openness and environmental quality are quite mixed. Though neither were statistically significant, a study by Jalil and Mahmud (2009) in China showed a negative (-) relationship to CO₂ emissions, while a study by Halicioglu (2009) in Turkey showed a positive relationship. In the study by Managi, Hibiki, and Tsurumi (2009), trade openness reduces CO₂ emissions in OECD countries but increases it in non-OECD countries. Even in the study of Sharma (2011), no significant relationship between trade openness and environmental quality appears at any income level. Tamazian and Rao (2010) and Al-Mulali (2012) confirmed in a study of 24 transition economies and 12 Middle Eastern counties, respectively, that trade openness would increase Co2 emissions. A study by Kasman and Duman (2015) of 15 EU member and candidate counties also found that trade openness worsened environmental quality. Table 1 presents a brief summary of the empirical studies on EKC which take into account such variables as trade openness.

Table 1. Brief literature review on EKC and trade openness.

Author	Country/region, Period	Variables used in the study	Results
Jalil and Mahmud (2009)	China, 1971–2005	CO2 emissions, commercial energy use, real income, trade openness	The EKC hypothesis is supported; trade openness decreases CO2 emissions, but statistically insignificant
Halicioglu (2009)	Turkey, 1960–2005	CO2 emissions, GDP, energy consumption, trade openness	The EKC hypothesis is supported; trade openness increases CO2 emissions, but statistically insignificant
Managi, Hibiki, and Tsurumi (2009)	88 countries, 1973–2000	SO2, CO2^,, BOD emissions, GDP, trade openness, capital-labor ratio, relative capital-labor ratio, relative GDP, international environmental treaties	The inverted U-shaped EKC of SO2 and CO2 is verified, trade openness reduces CO2 and SO2 emissions in OECD but increases in non-OECD significantly
Tamazian and Rao (2010)	24 transition economies, 1993–2004	CO2 emissions, GDP, Inflation, FDI, Price liberalization, forex and trade liberalization, trade openness, financial liberalization, institutional quality, energy import, energy consumption,	The EKC hypothesis is supported; trade openness increases CO2 emissions significantly
Sharma (2011)	69 countries, 1985–2005	CO2 emissions, GDP, energy consumption, urbanization, trade openness	The EKC does not apply; trade openness at any level of income does not have a significant effect on co2
Al-Mulali (2012)	12 Middle Eastern countries, 1990–2009	CO2 emissions, GDP, FDI, energy consumption, trade openness	The EKC does not apply; trade openness increases CO2 emissions significantly
Kasman and Duman (2015)	15 EU member and candidate countries, 1992–2010	CO2 emissions, GDP, energy consumption, trade openness and urbanization.	The EKC hypothesis is supported, trade openness increases CO2 emissions significantly

One of the reasons for the inconsistent results is the difference in the subject countries’ level of development. According to most studies, for developed countries the result tends to be negative, international trade reduces carbon dioxide emissions for these countries, while for the developing countries it tends to be positive. This could be explained by the pollution heaven hypothesis (hereafter, PHH) which suggests that low-income developing countries will be made dirtier with trade due to less strict environmental regulations and, therefore, lower pollution abatement costs. Thus, these countries might exploit their “comparative advantage” in pollution. According to the PHH, MNCs will move their production to these countries to avoid the strict environmental policy of developed countries. This causes less-developed countries to specialize in manufacturing the most pollution-intensive products (Cole 2004). But this lacks an explanation for the fact that trade is more affected by factor endowments than by environmental compliance cost, and that foreign costs in developing countries are larger than compliance costs in their own countries (Cole 2004; Grossman and Krueger 1993; Tobey 1990).

Another reason is that trade openness has multiple effects on environmental quality at the same time in the opposite direction. Estimates based on the single-equation model for capturing the multiple effects results in inconsistent. Dean (2002) investigates the effect of trade openness on environmental quality by dividing it into a direct effect through trade and an indirect effect via income. The results suggest trade openness has a negative direct effect on the environment via the terms of trade, but mitigates it with a positive indirect effect via income. Therefore, the composition effect induced by trade openness has a detrimental effect on environmental quality but is outweighed by the beneficial technique effect of income growth.

Moreover, the omitted variable bias, such as the institutional factor, may be the reason for this (Tamazian and Rao 2010; Torras and Boyce 1998). The institutional factor is an important characteristic of transitional countries that must be included in examining the influencing factors on the environmental quality in transitional economies (Chousa et al. 2005). Because the institution provides the rules of the game and its change toward a market-based system affects society as a whole. The institutional variables can be divided into two categories. The first are variables that arise from the transition to a market-based system, such as financial sector development, capital markets, and price liberalization. The second are variables that directly affect environmental quality, such as political stability, regulatory quality, and environmental regulation stringency (Al-Mulali and Ozturk 2015; Bhattacharya, Churchill, and Paramati 2017; Claessens and Feijen 2007; Kumbaroğlu, Karali, and Arıkan 2008; Tamazian and Rao 2010). The transition to a market economy increases trade and FDI, which affects environmental quality through the scale effect, technical effect, and combination effect. Governance-related factors affect environmental quality through the establishment and implementation of policies for environmental regulations.

Research model

Data

This paper focuses on the impact of trade openness on environmental pollution for the 10 CIS countries: 9 member states and 1 associate member, i.e. Armenia, Azerbaijan, Belarus, Kazakhstan, Kyrgyzstan, Moldova, Russian Federation, Tajikistan, Ukraine, and Uzbekistan. Turkmenistan is excluded from the test due to lack of available information. Based on data availability, the yearly sample observations were selected from the time period of 2000 to 2013¹.

Table 2 presents variable descriptions, data definitions, and data sources. GDPpc, CO₂pc, TO, CSPW, IND, POPU, INFL, ENERGY, RECON, REGU, and TRAN represent per capita GDP, per capita carbon dioxide emissions, trade openness, capital stocks per worker, share of industrial output in GDP, total population between the ages 15 to 64, rate of inflation, primary energy consumption, portion of renewable energy of total final energy consumption, regulatory quality of the government to formulate and implement sound policies and regulations, and level of progress in transition toward a market-based system, respectively. All data except TRAN is obtained from the World Development Indicators database (WDI) and World Governance Indicator (WGI) from World Bank. TRAN is obtained from the Assessment of Transition Qualities of European Bank for Reconstruction and Development (EBRD). Trade openness is measured as the sum of exports and imports of a country’s GDP. CSPW represents capital stock per worker calculated by dividing gross fixed capital formation on total population between the ages of 15 to 64. WTO is the dummy variable representing membership in the World Trade Organization. It equals 1 for member countries, and 0 for nonmembers. Landlocked is the dummy variable for a country’s location, 1 for a landlocked position and 0 if a country has access to the sea.

Table 2. Data definition and sources.

Variables	Definition	Unit	Data source
GDPpc	Per capita GDP	Constant 2010 USD	WDI, World Bank
CO2pc	Per capita carbon dioxide emissions	Metric tons	WDI, World Bank
TO	(Export + Import)/GDP	%	WDI, World Bank
CSPW	Capital stock per worker	Constant 2010 USD	WDI, World Bank
IND	Share of industrial output in GDP	%	WDI, World Bank
POPU	Total population between the ages 15 to 64	Number of person	WDI, World Bank
INFL	Rate of inflation, GDP deflator	%	WDI, World Bank
ENERGY	Primary energy consumption	kg of oil equivalent per capita	WDI, World Bank
RENCON	Renewable energy consumption	% of total final energy consumption	WDI, World Bank
REGU	Regulatory Quality	Scale	WGI, World Bank
TRAN	Mean value of six indicators of assessment of progress in a transition economy	Scale	ATQ, EBRD
WTO	Dummy variable for the WTO membership.	0,1	WTO
Landlocked	Dummy variable for the country’s location.	0,1	National statistic agencies

Model and methodology

In order to investigate the relationship among economic growth, environmental quality, and trade openness, to solve the endogeneity problem between income and trade openness, and to estimate the direct, indirect, and total effect of trade openness, this study’s methodology is partly based on the previous research conducted by Cole (2007). According to that research, the model includes three equations. The first equation describes per capita income as a function of trade openness, the second represents per capita carbon dioxide emissions as a function of income, and the last estimates the total effect of trade openness on environmental quality in CIS countries. Equations (1), (2), and (3) are defined as:

(1) $\mathbf{l}\mathbf{n}{\mathbf{Y}}_{\mathbf{i}\mathbf{t}}={\lambda }_{\mathbf{i}}+{\tau }_{\mathbf{t}}+{\alpha }_{\mathbf{1}}\mathbf{T}\mathbf{r}\mathbf{a}\mathbf{d}\mathbf{e}\mathbf{O}\mathbf{p}\mathbf{e}\mathbf{n}\mathbf{n}\mathbf{e}\mathbf{s}{\mathbf{s}}_{\mathbf{i}\mathbf{t}}+{\alpha }_{\mathbf{2}}{\mathbf{K}}_{\mathbf{i}\mathbf{t}}+{\epsilon }_{\mathbf{i}\mathbf{t}}$(1)

(2) $\mathbf{l}\mathbf{n}\mathbf{C}\mathbf{o}{\mathbf{2}}_{\mathbf{i}\mathbf{t}}={\gamma }_{\mathbf{i}}+{\delta }_{\mathbf{t}}+{\beta }_{\mathbf{1}}\mathbf{T}\mathbf{r}\mathbf{a}\mathbf{d}\mathbf{e}\mathbf{O}\mathbf{p}\mathbf{e}\mathbf{n}\mathbf{n}\mathbf{e}\mathbf{s}{\mathbf{s}}_{\mathbf{i}\mathbf{t}}+{\beta }_{\mathbf{2}}\left(\mathbf{l}\mathbf{n}{\stackrel{\wedge }{\mathrm{Y}}}_{\mathbf{i}\mathbf{t}}\right)+{\beta }_{\mathbf{3}}{\left(\mathbf{l}\mathbf{n}{\stackrel{\wedge }{\mathrm{Y}}}_{\mathbf{i}\mathbf{t}}\right)}^{\mathbf{2}}+{\beta }_{\mathbf{n}}{\mathbf{Z}}_{\mathbf{n}\mathbf{t}}+{\omega }_{\mathbf{i}\mathbf{t}}$(2)

(3) $\frac{\mathrm{d}\mathrm{E}}{\mathrm{d}\mathrm{T}\mathrm{O}}=\frac{\delta E}{\delta TO}+\frac{\delta E}{\delta Y}\frac{\delta Y}{\delta TO}$(3)

where the subscripts i and t denote country and year, respectively, per capita GDP, per capita CO₂ emissions, capital stock per worker, population and energy are expressed in natural logarithms.

In Equation (1), Y represents per capita GDP and K is a vector of additional explanatory variables that have commonly been used in research on economic growth. In this study, such additional explanatory variables as capital stock per worker, the portion of industry in GDP, population ages from 15 to 64, and inflation are included (Mankiw, Romer, and Weil 1992; Levine and Zervos 1993). λi and τt represent country and year specific effects. Instrumental variables are used to solve the endogeneity problem of per capita income and trade openness. Per capita GDP is a function of trade openness, yet trade openness is itself likely to be a function of per capita GDP. This is a familiar problem from previous research on trade openness and income.

Regarding the endogeneity problem, which refers to the possibility that trade might be simultaneously determined along with income and environment, Antweiler, Copeland, and Taylor (2001) point out the endogeneity as a potential weakness of their study while many existing researches do not address the endogeneity problem at all. Other researchers do attempt to address some aspects of endogeneity in their studies. Managi, Hibiki, and Tsurumi (2009) treat income and trade as endogenous and use the instrumental variables technique to solve the problem. Dean (2002) treats income as endogenous in her study and utilizes two systems of equations to solve it. Frankel and Rose (2005) address the endogeneity problem of income and trade by using a system of instrumental variables.

This study partially follows Frankel and Rose (2005) to address the endogeneity problem by using a set of instrumental variables and attempts to control the transition to a market economy and governance-related factors using the Assessment of Transition Qualities Index provided by EBRD and the World Governance Indicator provided by World Bank as variables.

According to Hall and Jones (1999), there are two conditions that variables should fit to be used as instrumental variables: 1) They should be positively correlated with trade openness, and 2) There should be a lack of correlation with the disturbance ε. Hence, we adopt two instrumental variables, landlocked status and membership in WTO, based on previous literature (De Groot et al. 2004; Frankel and Rose 2005).

After estimation of the first equation, per capita GDP is adjusted as fitted and applied in Equation (2) in order to obtain more accurate results for the direct effect of trade openness on environmental quality. Equation (2) is based on the environmental Kuznets curve hypothesis. According to it, per capita CO₂ emissions are explained as a function of per capita GDP, square of per capita GDP, trade openness, and variable Z. Z is a vector of other explanatory variables: Primary energy consumption, Portion of renewable energy in total energy consumption, and institutional variables. In addition, γ_i and δ_t represent country and year specific effects. Finally, ε_it and ω_it indicate error terms. Squared GDP assists in finding evidence of the inverted U-shaped EKC.

According to the discussion above, the impact of trade openness on per capita income – environmental quality relationships should be investigated. The sign of β1, which corresponds to trade openness, is expected to be mixed and depends on the level of economic development of the sample countries. According to the EKC hypothesis, it is expected that the sign of β2 is positive, while the sign of β3 is negative. Hence, there is an inverted U-shaped curve showing that after some point an increase in per capita income leads to a decrease in carbon dioxide emissions. The sign corresponding to β4 (primary energy consumption) is expected to be positive because a higher level of energy consumption should result in greater economic activity and stimulate CO₂ emissions. The sign corresponding to β5 (renewable energy consumption %) is expected to be negative if the level of renewable energy which is used is high enough. The sign corresponding to β6 (regulation quality) is expected to be negative because a higher level of regulation on environmental quality leads to a decrease in carbon dioxide emissions. The sign corresponding to β7 (progress in transition) is expected to be positive because a high level of transition to the market-based system leads to an increase in carbon dioxide emissions through economic growth.

In Equation (3) we calculate the total effect of trade openness on CO₂ emissions using the Cole (2007) model for estimating the total effect of corruption on carbon dioxide emissions. According to his model, the total effect of corruption on environmental quality could be decomposed into a direct and an indirect effect, which shows the effect of corruption on income and then the effect of income on environmental quality. Therefore, we propose Equation (3) in the case of trade openness. In Equation (3) dE∕dTO stands for the total effect, which is equal to the sum of the direct and indirect effect; δE/δTO is the direct effect, and δE/δY *δY/δTO reflects the indirect effect of trade openness on carbon dioxide emissions. E, Y, and TO are again used to denote per capita CO_2, per capita GDP, and trade openness.

Results

Descriptive results

The summary statistics of the variables are presented in Table 3. The mean per capita CO₂ emission ranges from 0.36 in Tajikistan to 12.46 in Kazakhstan with Kazakhstan exhibiting the most variation in per capita CO₂ emissions and Tajikistan with the least. In case of per capita GDP the highest mean is observed in the Russian Federation (9462.02), followed by Kazakhstan (7704.88), with the lowest mean is observed in Tajikistan (637.73). Kazakhstan has the highest volatility (1867.16) and Kyrgyzstan has the lowest (105.52). Regarding the mean of trade openness, Kyrgyzstan is observed to have the highest mean (1.35), followed by Belarus (1.27) and Moldova (1.16), whereas Uzbekistan and the Russian Federation have the lowest mean (0.49 and 0.46, respectively). As for the variation, Kyrgyzstan has the highest and the Russian Federation and Tajikistan have the lowest trade openness variation (0.21 and 0.06, 0.06, respectively).

Table 3. Summary statistics of key variables for each country.

	Per capita CO2		Per capita GDP		Trade Openness
Countries	Mean	Std.	Mean	Std.	Mean	Std.
Armenia	1.49	0.32	2751.68	815.73	0.68	0.08
Azerbaijan	3.78	0.31	4001.64	1777.48	0.7	0.1
Belarus	6.16	0.57	4724.72	1382.32	1.27	0.1
Kazakhstan	12.46	2.72	7704.88	1867.16	0.9	0.13
Kyrgyzstan	1.23	0.29	805.89	105.52	1.35	0.21
Moldova	1.27	0.15	1431.97	284.35	1.16	0.19
Russian Federation	11.5	0.69	9462.02	1844.99	0.46	0.06
Tajikistan	0.36	0.05	637.73	136.93	0.63	0.06
Uzbekistan	4.4	0.56	1149.47	279.91	0.49	0.17
Ukraine	6.65	0.47	2746.75	487.92	1.09	0.1
Panel	4.93	4.19	3543.78	3068.95	0.87	0.34

Note: Std. stands for standard deviation; per capita CO₂ is measured in metric tons, per capita GDP is measured in constant 2010 US$.

Estimation of the impact of trade openness on income

Table 4 presents estimation results for Equation (1), which shows the relationship between trade openness and per capita income. According to the results of the Hausman test (chi2(4) = 42.61***), we applied fixed-effects regression for exogenous trade openness. 2SLS Instrumental variable regression with robust standard errors is applied in case of endogenous trade openness. In the first column, trade openness is treated as being exogenous to per capita GDP so instrumental variables are not applied. In this case, trade openness has a negative significant effect on per capita income. If trade openness increases by 1%, per capita GDP decreases by 0.2%. Other variables, such as capital stock per worker and population, have a positive significant effect on per capita GDP. In the second column, trade openness is controlled by the set of the instruments that includes membership in the WTO and the country’s location. According to the table, the effect size of trade openness on income has increased and has a negative significant impact on per capita income. If trade openness increases by 1%, income decreases by 0.3%. Other variables have a positive significant effect on income, as in the previous model. The effects of inflation and industry in both models are found to be insignificant. All variables discussed above account for about 94% of the variation in income (R² = 94%), time trends are controlled. Finally, the test of overidentifying restrictions fails to reject the null hypothesis, confirming that our instruments are valid and our model is specified correctly. In addition, the Olea and Pflueger weak instrument test shows that the instruments are significant.

Table 4. Estimation of the impact of trade openness on income.

LogGDPpc	Exogenous Trade Openness	Endogenous Trade Openness
TO	−0.1957** (0.0853)	−0.2908** (0.1248)
ln CSPW	0.1722*** (0.0431)	0.8216*** (0.0282)
IND	2.12e-13 (5.27e-13)	2.25e-13 (2.59e-13)
ln POPU	0.6016*** (0.1990)	0.0744*** (0.0198)
INFL	0.0003 (0.0006)	0.0009 (0.0009)
Time Trends	Controlled	Controlled
R^2	0.4837	0.9410
Obs.	140	140
F test	42.97***
Wald Chi2		1970.47***
Hausman test	42.61***
Sargan test		1.8113
F test on IV		21.374***

Notes: Significance level in parentheses (probabilities for Sargan and F test). ***, **, * denote statistical significance at 1%, 5% and 10%, respectively.

Estimation of the impact of trade openness on CO₂ emissions

We adjust GDP as a fitted value from Equation (1) and apply it to Equation (2). Before estimating Equation (2), we test the autocorrelation and heteroscedasticity of error terms. The results show that there is heteroscedasticity and autocorrelation of the error terms (LR chi2(9) = 208.79, p = 0.000 and F value = 6.625, p = 0.0300). Therefore, Feasible Generalized Least Square estimation is applied, as it is more efficient than OLS in this case.

Table 5 presents the estimation results for Equation (2). In the first column, a basic equation, (M1) is estimated without trade openness. Explanatory variables (Z) are primary energy consumption, renewable energy consumption %, regulatory quality, and degree of progress in a transition economy. In (M2) model trade openness is added. According to the model (M1), energy use has a positive statistically significant effect on carbon dioxide emissions, while renewable energy consumption reduces it. Regulatory quality has a negative statistically significant effect on carbon dioxide emissions, while degree of progress in a transitional economy increases it. With the addition of trade openness, the results show that it has a positive statistically significant impact on CO₂ emission, when trade openness increases by 1%, the level of CO₂ emissions increases by 0.05% in CIS countries. Therefore, the results show that trade openness has a negative direct effect on the environmental quality in the region. As for other variables, energy use and degree of progress in a transitional economy have a positive statistically significant effect on CO₂ emissions, while renewable energy consumption and regulatory quality have a negative statistically significant effect. The sign of β₂ is positive and the sign of β₃ is negative in all models, which supports the existence of the inverted U-shaped EKC for the CIS countries. Time trends in all models are applied to control unknown effects that may occur between 2000 and 2013 over the entire panel.

Table 5. Estimation of the impact of trade openness on per capita CO₂ emissions.

ln CO2pc	(M1)	(M2)
ln GDPpc	0.9736*** (0.1282)	0.9202*** (0.1759)
ln Square of GDPpc	−0.0545*** (0.0077)	−0.0510*** (0.0111)
ln ENERGY	1.1072*** (0.0174)	1.1273*** (0.0295)
RENCON	−0.0046*** (0.0006)	−0.0037*** (0.0008)
REGU	−0.2772*** (0.0329)	−0.2805*** (0.0425)
TRAN	0.2972*** (0.0321)	0.2989*** (0.0399)
TO		0.0558** (0.0235)
Time Trends	Controlled	Controlled
Obs.	130	130
Wald Chi2	99,105.18***	82,563.44***

Notes: Significance level in parentheses. ***, **, * denote statistical significance at 1%, 5% and 10%, respectively.

As our results support the EKC hypothesis and verify the existence of an inverted U-shaped curve, we decided to estimate the threshold level for CIS countries. Using the results obtained from model (M2), we estimated the turning point of income after determining at what level carbon dioxide emission starts decreasing and per capita GDP continues to increase. The turning point, where emissions reach the maximum level, can be estimated as:

${\mathbf{x}}^{\ast }=\mathbf{e}\mathbf{x}\mathbf{p}\left(-{\beta }_{\mathbf{2}}/\mathbf{2}{\beta }_{\mathbf{3}}\right),$

where β₂ is a coefficient of the linear term, and β₃ is a coefficient of the squared term.

The turning point of CO2 emissions for the CIS countries is estimated to be 8,279 US$ with the presence of trade openness, as after reaching this level of per capita income CO₂ emissions start to decrease. However, not all of the CIS countries reached this level by 2013, some of them perform more poorly than others, especially non-oil exporters. Overall, the verification of the existence of the inverted U-shaped EKC is confirmed for the entire region; however, there is some variance due to the underperformance of some countries.

As mentioned above, the EKC hypothesis that assumes an inverted U-shaped relationship between CO₂ emissions and income is verified in all models. Our result is similar to Apergis and Payne (2010), who investigated the relationship between growth, energy consumption, and pollution for a panel of CIS countries and verified the existence of the inverted U-shaped curve. It also falls in line with the Tamazian and Rao (2010) findings, which support the EKC hypothesis for the 24 transition economies included while taking into account the institutional quality of these countries. However, the result is the opposite of the Pao and Tsai (2011) paper, which found no evidence of the EKC hypothesis in Russia during the period between 1990 and 2007. As for studies that also incorporate a trade openness variable, the result of this research is consistent with Kasman and Duman (2015), Jebli, Youssef, and Ozturk (2016), and Al-Mulali and Ozturk (2016), all of which support the EKC hypothesis.

As our next step, we calculated the total effect of trade openness on CO₂ emissions according to Cole (2007). Table 6 presents the direct, indirect, and total effect of trade openness on per capita CO₂ emissions using the results obtained from the estimation of the impact of trade openness on income in cases where trade openness is treated as being endogenous and in model (M2). The direct effect, as mentioned above, is found to be positive and does not vary with income, increasing trade openness increases CO₂ emissions. However, the indirect effect of trade openness is found to be negative, increasing trade openness reduces CO₂ emissions via its impact on income. Thus, the total effect of trade openness on pollution, which is measured as the direct impact plus indirect impact, is estimated to be negative due to a larger indirect effect.

Table 6. Estimation of the total impact of trade openness on CO₂ emissions.

Direct effect	Indirect effect	Total effect
0.0736	−0.3505	−0.2769

Note: The indirect effect is calculated using the sample median level of GDP per capita ($2572)

Discussion

This paper is designed to explore the relationship between economic growth, environmental quality, and trade openness for CIS countries, which are in transition toward a market-based society. The empirical results are somewhat different from the previous studies. This section discusses these results.

The negative effect of trade openness on economic growth in CIS countries

Trade opening is a representative policy adopted by developing countries for economic growth. Trade can efficiently reallocate resources and accumulate capital for investment. And trade can allow developing economies to acquire advanced technology and increase productivity (Dowrick and Golley 2004; Winters 2004). However, in this research, trade openness has a negative effect on economic growth. This could be explained by the fact that imports into CIS countries continue increasing, and for many non-oil exporters, their import share is larger than their export share (Armenia, Kyrgyzstan, and Moldova). Due to the low production level in these countries, they tend to import more than they produce, especially with regard to high-tech products. Exports from these countries primarily consist of natural resources, in which some CIS countries are abundant (Azerbaijan, Kazakhstan, and the Russian Federation). However, the total balance of trade remains negative for these countries. Therefore, the results show the necessity of change in the trade policy and trade structure of these countries in order to have increasing trade openness benefit the country’s income and not harm it.

Moreover, according to Jenish’s (2013) paper on trade and economic growth in the CIS region, only trade with the Russian Federation was found to be an important determinant of economic growth in these countries (an increase of 1% in the growth rate of trade with the Russian Federation increases the economic growth rate in CIS countries by 0.07%). As for extra-regional trade, it is found to affect growth in the countries which export oil, such as Azerbaijan, Kazakhstan, and the Russian Federation. However, it insignificantly affects economic growth when all countries are taken together due to the lower trade volume of non-oil exporters.

What is more, there are several factors which could explain the low trade volume in CIS countries. A landlocked position is one of the main reasons. According to the literature on gravity model, a country’s location is a powerful determinant of trade. By geographic definition, a landlocked country is a country that does not have open access to the sea. As discussed in Radelet and Sachs (1998), Limao and Venables (2001), and Raballand (2003), countries which have an access to the sea tend to trade more and have lower transportation costs in comparison to the landlocked countries. This is especially important for CIS countries due to a very low level of transport infrastructure development. According to Raballand (2003), developing air freight, which could solve the problem of being landlocked, was not a priority of the Soviet Union. Moreover, overland transport is found to be more expensive than maritime because of border-crossing obstacles. This is especially acute among developing countries or countries in transition due to a high level of corruption and inefficient governance, which lead to excessive documentation, delays, and inadequate border infrastructure. Overall, it increases the costs of border-crossing, which are eventually charged to customers.

Another reason, why trade may negatively affect economic growth is institution quality. According to Biswas, Farzanegan, and Thum (2012) and Leitao (2010), governments in transition countries are less efficient due to the high level of corruption, the large share of the shadow economy, political instability, and rent-seeking behavior. Transparency International and World Governance Indicator indices prove that the level of corruption is very high among the CIS countries as this region remains among the worst scorers every year. As discussed in the previous literature, including Easterly and Levine (2003), Hall and Jones (1999), and Kaufmann, Kraay, and Zoido-Lobatón (1999), low institutional quality negatively affects per capita income. This could be explained by the fact that low institutional quality leads to inefficient resource allocation, innovation inhibition, and investment reduction (Mauro 1995; Mo 2001; Murphy, Shleifer, and Vishny 1993). According to Hakkala, Norbäck, and Svaleryd (2008), Shleifer and Vishny (1993), and Wei (1997), countries with a high level of corruption are less attractive to foreign investors due to high insecurity cost, which reduces opportunities for economic growth. Moreover, according to Biswas, Farzanegan, and Thum (2012), corruption causes an increase in the shadow economy, which affects a country’s GDP because its activity is not reflected in official economic figures and makes for less accurate estimations. According to Dollar and Kraay (2003), countries with better institutions also tend to trade more due to government efficiency, transparent procedures, and lower costs of doing business. Finally, according to Banerjee (1997), corruption might set the social and economic development in a country back due to political inefficiency, dilution of trade patterns, and transaction costs. Overall, the negative effect of trade openness on per capita income indicates the necessity of new trade policy and higher institutional quality in CIS countries.

The negative and direct effect of trade openness on environmental quality in CIS countries

Trade openness is found to have a significant positive effect on CO₂ emissions. This could be explained by several factors. First, all countries in the sample are developing countries whose main export products remain as natural resources. The scale effect of trade openness encourages them to export more natural resources, such as oil and gas, which increases CO₂ emissions due to low energy efficiency extracting practices, aging energy infrastructure, and weak environmental regulations (Apergis and Payne 2010). Despite increasing trade openness, its technique effect is observed to be quite low as CIS countries do not implement clean state-of-the-art production techniques from advanced countries due to less strict environmental regulations. Our results differ from that of Shahbaz et al. (2014) as they show that increasing exports in the UAE reduces CO₂ emissions due to the positive technique effect of trade. However, the negative effect of exports on the environmental quality is partly mitigated by the positive effect of increasing imports. This result is similar to Jebli, Youssef, and Ozturk (2016), as they show that increasing imports has a positive effect on the environmental quality in developed countries. As for the CIS region, it could be explained by the fact that CIS countries import most of their manufactured goods due to the low level of production in these countries. Increasing trade openness has brought structural changes to the composition of these countries’ economies and resulted in the increased importance of the natural resource sector. According to Sari and Soytas (2009), a country’s abundance of natural resources may have a negative impact on other tradable good sectors of the economy, yielding a negative impact on economic growth (“Dutch disease”).

The result is consistent with that of Jebli and Youssef (2015) as they show that increasing trade lead to an increase in CO₂ emissions in the case of Tunisia. It also supports the result of Halicioglu (2009), which showed that increased trade openness increased per capita CO₂ emissions in the case of Turkey. However, in both studies mentioned above, the EKC was not verified. The result of this study is in line with Kasman and Duman (2015), which shows that trade openness has a positive significant impact on pollution, suggesting that increasing trade volumes increase CO₂ emissions in new EU member and candidate countries. This result is contrary to Jebli, Youssef, and Ozturk (2016), which showed that increasing imports or exports reduced CO₂ emissions. It also differs from that of Al-Mulali and Ozturk (2016), which showed that increasing trade reduces pollution in the long run. However, these differences might be due to the fact that most of the countries from their sample are developed countries with strict environmental regulations and high demand for environmental quality.

The total effect of trade openness on environmental quality in CIS countries

The total effect of trade openness on CO₂ emissions is estimated to be negative in this research. Thus, trade openness seems to lead to better environmental quality in CIS countries. However, the economic growth it brings about is far from sustainable. Rather, the results of the study confirm the necessity of changes in the trade policy of CIS countries because the positive effect on environmental quality is caused by the negative effect of trade openness on economic growth. The results of our estimation are found to be slightly different from the results of previous studies. Dean (2002) shows that increased trade openness directly increases CO₂ emissions in China due to its comparative advantage in manufacturing pollution-intensive goods, which is mitigated by the positive indirect effect of raising income. The pattern seems to be similar here. However, for China, increasing trade openness leads to an increase in income, while in CIS countries it decreases per capita GDP. According to Jayanthakumaran, Verma, and Liu (2012), both direct and indirect impacts are found to be positive and lead to an increase in emissions due to increased trade openness (also in the case of China) because of the scale effect dominating over technique effect. As for CIS countries, the composition effect dominates the technique effect. The composition effect could be explained by structural changes in the composition of the countries’ economies due to the trade openness that allows countries to specialize in manufacturing goods in which they have a comparative advantage (Antweiler, Copeland, and Taylor 2001). As some CIS countries have abundant natural resources, they began to intensively export them due to the increased trade openness while the share of imports dramatically increased due to a lower level of production of manufactured goods. Lower technique effect leads to an increase in CO₂ emissions due to the large share of natural resource exports in CIS countries (Ahrend 2005). According to Tamazian and Rao (2010), CIS countries still have a low level of energy efficiency, aging energy infrastructure, and do not implement eco-friendly technology, especially in oil and gas production. Overall, increases in exports lead to increases in CO₂ emissions, but low production levels in these countries further accelerate these increases.

Estimation of the impact of importing/exporting industries of CIS countries on CO₂ emissions

In this section, we separately estimate the impact of the exporting and importing industries on CO₂ emission in the CIS region. The autocorrelation and heteroscedasticity of error term was tested before estimating the impact of importing and exporting on environmental quality. The results show that there are heteroscedasticity and autocorrelation of the error term (LR chi2(9) = 168.22, p = 0.000 and F value = 6.275, p = 0.0336) in the model for the impact of importing and that there are heteroscedasticity and the autocorrelation of the error term (LR chi2(9) = 172.11, p = 0.000 and F value = 7.498, p = 0.0229) in the model for the impact of exporting as well. Therefore, Feasible Generalized Least Square estimation is applied, as it is more efficient than OLS in this case. The estimation results are presented in Table 7.

Table 7. Estimation of the impact of exporting and importing industries on per capita CO₂ emissions.

Log(CO2pc)	IMP	EXP
ln GDPpc	0.0690*** (0.0175)	0.1239*** (0.0265)
ln IMP/GDP	−0.0467*** (0.0116)
ln EXP/GDP		0.2095*** (0.0172)
ln ENERGY	1.0981*** (0.0197)	1.0563*** (0.0294)
RENCON	−0.0071*** (0.0004)	−0.0051*** (0.0006)
REGU	−0.2097*** (0.0327)	−0.3061*** (0.0461)
TRAN	0.2282*** (0.0283)	0.2909*** (0.0449)
Time Trends	Controlled	Controlled
Obs.	130	140
Wald Chi2	73,068.71***	56,345.43***

Notes: Significance level in parentheses. ***, **, * denote statistical significance at 1%, 5% and 10%, respectively.

According to the IMP model in the first column of Table 7, imports have a negative statistically significant effect on per capita CO₂ emissions, which means that a 1% increase in imports reduces CO₂ emissions by 0.04% in the region. As for exports, the EXP model in the second column shows that exports have a positive statistically significant impact on CO₂ emissions, a 1% increase in exports increases CO₂ emissions by 0.20% in the region. Moreover, in both the IMP and EXP models, renewable energy and regulatory quality reduce CO₂ emissions and energy consumption and degree of progress in transitional economies increase CO₂ emissions. Time trends in all models are applied to control unknown effects that may occur between 2000 and 2013 over the entire panel.

The results in Table 7 show that increasing imports reduces CO₂ emissions while increasing exports increases CO₂ emissions. This could be explained by the fact that CIS countries are developing countries, some of the countries in this region are specialized in exporting nonrenewable energy such as oil and gas. The results for imports could be explained by the low production level in these countries. The results for exports are contrary to Shahbaz et al. (2014), who show that increasing exports decreases CO₂ emissions in the case of the UAE. The results for importing industries are similar to those obtained from a panel of 25 OECD countries by Jebli, Youssef, and Ozturk (2016). According to the results, we can make a conclusion that, although importing industries helps to reduce CO₂ emissions, the increase due to exporting industries is higher in absolute values.

Conclusion

This paper is designed to investigate the relationship between CO₂ emissions, economic growth, and trade openness for a panel of CIS countries for the period of 2000–2013. Particularly, the paper attempts to test the EKC hypothesis for the sample countries while considering the endogeneity of income and trade openness. Using a two-equation model helps us to solve this problem and also estimate the direct effect of trade openness on pollution and the indirect effect via income. Moreover, the total effect of trade is estimated as a sum of the direct and indirect effects. Instrumental variable technique and FGLS regression are applied; all necessary post-estimation tests are performed.

The results support the EKC hypothesis and verify the existence of the inverted U-shaped curve between per capita income and carbon dioxide emissions. Therefore, CO₂ emissions increase with per capita income, reach the turning point, and then decline. Trade openness is found to negatively affect economic growth. This could be explained by increasing import share in CIS countries, which is especially high in the case of non-oil exporters. Moreover, due to the composition effect of trade, the production in many CIS countries remains at a very low level, which forces them to import a greater amount of manufactured products. Almost all exports consist of natural resources, as some of CIS countries have them in abundance. Also, the low level of trade in CIS countries, especially extra-regional trade, can be explained by the landlocked position of these countries and low institutional quality. As for the direct effect of trade openness on CO₂ emissions, it is found to be positive, increasing trade increases CO₂ emissions in this region. This happens due to the scale effect of trade openness, which encourages these countries to export more natural resources. The technique effect remains at low levels. CIS countries do not adopt much in the way of clean state-of-the-art techniques for their production and continue using carbon-intensive techniques and an aging energy infrastructure. Overall, the total effect of trade openness is negative due to the larger indirect effect; increasing trade openness reduces CO₂ emissions. However, this result should not be misunderstood. A simple reduction of CO₂ emissions is not enough to conclude that trade openness positively affects CIS countries.

The findings show that CIS countries do not enjoy the positive effects of trade openness on economic growth and environmental quality. To improve this, the following policy efforts are needed: First is an improvement of environmental quality in exports. Exports are very important for economic growth, but involve greenhouse gas emissions. This study found that CIS countries saw a greater increase in CO₂ emissions due to increased exports than a decrease in CO₂ emissions from imports. This is due to inadequate effect practices, engaging energy structures, and weak environmental regulations (Apergis and Payne 2010). Therefore, efforts such as introducing eco-friendly technologies and replacing facilities are urgently needed to overcome them (Yu, Park, and Hwang 2019). The second is active R&D investment to enhance the nation’s absorption capacity. Absorbability is the ability to assess, integrate and apply the value of new knowledge (Cohen and Levinthal 1990). High absorption means that an organization can acquire, transform, and use new knowledge from the outside (Zahra and George 2002). Absorption capacity increases innovation by transferring knowledge acquired from the outside into the organization and spreading it to its members (Nonaka and Takeuchi 1995). Thus, technological advancement in CIS countries through trade openness will be affected by their ability to absorb. While technology transfer through trade openness occurs easily when a country has high absorption capacity, low absorption makes it difficult for technology transfer to occur. The results of this study show the technical effect of trade openness on environmental quality maintains a low level in CIS countries. To overcome this, investment in R&D is needed to enhance absorption capacity. Finally, they require effort to expand the scope of trade openness. According to the WTO (2003), trade openness has a long-term positive effect on a country’s economic growth as it provides an increase in investments and stimulates the spread of technology. However, this study finds that the trade openness of CIS countries has a negative effect on economic growth. The reasons point to a Russia-centric trade openness as well as oil and natural gas-oriented exports (Jenish 2013). The expansion of trade openness include extra-region areas will increase the CIS Counties’ opportunities to take advantage of the scale effect, position effect, and technical effect of trade openness.

In spite of the contribution to the existing EKC literature, this research has several limitations that provide direction for future studies. First, the time period used in this study is relatively short. Second is the endogeneity of energy use in the equation of CO2 emissions. This study controlled the endogeneity between GDP and trade openness, which is mentioned in previous research on trade openness and income. However, it needs to consider the endogeneity of energy consumption with income for more efficient estimation of the determinants of CO₂ emissions. Also, future research could take into account the income level of each of the CIS countries and divide them into low-income countries and oil-exporting countries.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea [NRF-2016S1A5B6925462].

Notes

1. World Governance Indicator from World Bank is not available for 2002.