This study explores the relationships between carbon emissions and their main determinants such as energy consumption, real income, international trade, level of education and level of urbanization in the Russian Federation, employing data for the period 1991–2016. Support for the environmental Kuznets curve hypothesis is found in this study, stating that environment pollution decreases in Russia after income achieves a certain threshold. The ARDL bounds test is employed in order to estimate short-run and long-run relationships in the estimated model. Energy consumption, real income, education and urbanization levels are found to be significant determinants of carbon emissions, while trade openness does not have an impact. The Granger causality test indicates two-way relationships between carbon emissions and energy use, real income and education. Only a single one-way causality runs from carbon emission to trade and no causality was found between carbon emissions and level of urbanization.

CO2 emissions, environment Kuznets curve, cointegration, Russia.

Russia is one of the largest contributors in the world to carbon dioxide (CO₂) emissions after China, the US and India in total kilotons (kt). However, in terms of kilogram (kg) of CO₂ emissions per 2010 dollars of GDP in 2014, Russia left the US far behind with 0.99 kg compared to 0.33 kg in the US. Contrary to the general tendency of decline in CO₂ emissions among developed countries, CO₂ emissions in Russia continue to rise. Since 1998, Russia has increased CO₂ emissions by 14 percent, making an 8 percent overall increase since 2009. The changes are in favor of the environment, however, if GDP is taken into account as CO₂ emissions per 2010 dollars of GDP declined by 45.6 percent in 2014 from 1998 levels and by 7.3 percent from 2009.1 This is explained by the use of more environmentally-friendly equipment in the growing economy.

In 2015, the Russian Federation declared, as its target of intended nationally determined contribution (INDC), a decrease in the level of greenhouse gas (GHG) emissions by 25–30 percent from the 1990 level by 2030. However, this target is criticized in literature on the grounds that, in 2012, emissions in Russia were already 68 percent of the 1990 level (Korppoo and Kokorin, 2017). Even though Russia has reached its target and is working to maintain it for the next decade, it continues to be one of the largest contributors to CO₂ emissions in the world.

New policies need to be designed in order to follow the global continuous declining trend. For example, since 1990, the environmental policy of the European Union (EU) has been focused on the commitment of its countries to decrease greenhouse emissions by 40 percent by 2030. In 2015, 195 countries signed the Paris Agreement, one of the important goals of which is to decrease world greenhouse emissions.

This study investigates factors affecting CO₂ emissions in Russia that contribute to climate policies. At the same time, it examines the environmental Kuznets curve hypothesis in the context of that country. Factors of CO₂ emissions in Russia have not been investigated empirically to date using an expanded framework. Therefore, the aim of this study is to conduct the first test of the environmental Kuznets curve in Russia by eliminating omitted variables bias, investigating the long-run relationships between CO₂ emissions and variables that are responsible for changes, such as energy use, income per capita, trade openness, education and degree of urbanization. The next part reviews empirical findings in literature. The third part introduces the methodology applied in this paper and then empirical results are outlined in the fourth part, followed by a discussion in the conclusion on policy implications.

The problem of worsening environmental quality has been receiving increased attention in literature. Most of the empirical studies have been conducted within the framework of the environmental Kuznets curve (EKC). The EKC hypothesis assumes an inverted U-curve relationship between environmental degradation and income per capita. With an increase in income, pollution initially rises as well; however, after the economic growth reaches a certain threshold, levels of pollution start to decrease. Testing the EKC hypothesis, some studies explain changes in environmental quality purely by economic growth (Panayotou 1993; Grossman and Krueger, 1995; Ekins, 1997; Stern 2004; Narayan and Narayan, 2010). However, recently more studies have begun to identify additional factors that may play important roles in environmental degradation processes. For example, energy consumption is considered a key determinant of environmental quality (Ang, 2007; Acaravci and Özturk, 2010; Pao et al., 2011; Saboori and Sulaiman 2013; Shahbaz et al., 2013; Baek, 2015). Numerous studies agree that energy consumption is mostly responsible for CO₂ emissions (Pablo-Romero and Jesus, 2016) and directly or indirectly plays the principal role in environmental problems (Suri and Chapman, 1998).

However, recently more studies have incorporated the factor of international trade (Suri and Chapman, 1998; Cole, 2004; Halicioğlu, 2009; Jayanthakumaran and Liu 2012; Saboori et al., 2012; Lau et al., 2014; Jebli et al., 2016). The liberalization of international trade has encouraged greater production, leading to economic growth, which in turn has created solutions to environmental degradation (Arrow et al., 1996). International trade leads to the movement of products produced in one country to other countries where they are consumed or involved in further production. Therefore, with the increase in trade, pollutant emissions may move to other countries (Halicioğlu, 2009). Suri and Chapman (1998) found that the export of manufactured goods in industrializing countries is the main determinant of increased energy consumption leading to the rapid growth of pollution, while imports of manufactured goods in industrialized countries are partly responsible for a decline in pollutant emissions.

In order to solve the problem of omitted variables, in literature, different determinants are incorporated in the EKC framework. For example, urbanization has been found to have a dual effect on the environment (Managi and Jena, 2008; Wang et al. 2008; Kasman and Duman, 2015). Generally, it is believed that higher levels of urbanization lead to higher levels of pollution due to greater population. However, improved levels of disposal organization and hygiene provisions in urban areas may create positive relationships between urbanization and environmental quality (Gangadharan and Valenzuela, 2001). The effect of education on environmental quality may vary as well (Hill and Magnani, 2002; Farhani et al., 2014). As discussed in Balaguer and Cantavella (2018), with an increase in the level of education, more knowledge is extracted for the consumption of technology-intensive products, which are relatively more polluting products. However, education can lead to social awareness and motivate people to improve environmental quality by decreasing their consumption of non-renewable products and by improving disposal separation.

Some researchers argue that energy prices play an important role in the EKC (Agras and Chapman, 1999; Lindmark, 2002; Richmond and Kaufmann, 2006; He and Richard, 2010; Al-Mulali and Özturk, 2016; Katircioğlu, 2017). Energy prices are considered a key determinant of pollution although the effect may be negative as well as positive. Higher energy prices encourage producers to substitute energy-intensive technologies for less intensive ones, which leads to a decline in pollution. On the other hand, higher energy prices stimulate higher production of energy, leading to increased levels of pollution (He and Richard, 2010).

To eliminate the possibility of omitted variable bias, researchers include in the EKC framework analysis such variables as inequality (Torras and Boyce, 1998; Magnani, 2000; Heerink et al., 2001), tourism (Katircioğlu, 2014; Arbulu et al., 2015; Zaman et al., 2016), financial development (Tamazian et al., 2009; Jalil and Feridun, 2011; Katircioğlu and Taspinar, 2017) and health (Gangadharan and Valenzuela, 2001; Brajer et al., 2008; and Khan et al., 2016).

The list of major contributors to world CO₂ emissions consists of developed as well as developing countries; therefore, numerous studies on the EKC hypothesis include developed (Millimet et al., 2006; Iwata et al., 2010; Esteve and Tamarit, 2012; Shahbaz et al., 2015; Pablo-Romero et al., 2017) as well as developing and transition countries (Bulte and Soest, 2001; Fodha and Zaghdoud, 2010; Al-Mulali et al., 2015; Jebli and Youssef, 2015; Özturk and Al-Mulali, 2015; Li et al., 2016). While there is no common conclusion for developed countries within the EKC framework, numerous studies on developing countries reveal that there is no “one-fit-for-all” relationship between economic growth and environmental quality, where developing countries do not wait for an increase in income to reach the threshold for environmental improvement, but design their own environment improvement path (He, 2016).

In support of general findings, Halicioğlu and Ketenci (2016) revealed support for the existence of EKC only in three out of fifteen transition countries. Sirag et al. (2017) estimated that the income level in developing countries is below the desired threshold for environmental improvements where the government environmental policies suggested the EKC would be inappropriate. Sayed and Sek (2013) tested the EKC hypothesis in developed and developing countries by decomposing emissions by various environmental measures. Their results, in contrast to previous studies, support the existence of the EKC in most cases of developed and developing countries; however, the income threshold in developed countries is higher than that in developing ones. Alternative methodologies may vary the outcome, thus Stern (2004) criticizes previously employed methodologies of total pollution analysis that may lead to ambiguous results. The author suggests that various decompositions of emissions in analysis may help to enlighten relationships between emissions and economic growth.

Despite the interest in literature in the EKC for polluting countries, empirical studies for Russia are few in number. Among them are Pao et al. (2011), who found in their dynamic study that emissions are “output inelastic” and the EKC is not supported. Halicioğlu and Ketenci (2016) and Yang et al. (2017) found support for the EKC in Russia, where the economy-related greenhouse emissions are presented by energy consumption, emissions from industrial process, from animal husbandry and fugitive emissions. The validity of the environmental Kuznets curve for Russian regions has been tested with mixed results by Mihalischev and Raskina (2015), Rudenko and Skripnuk (2016) and for cities by Ivanova and Vertkina (2017). The results on the EKC existence are controversial as well as in studies where Russia is included in the panel of BRIC countries. For example, Pao and Tsai (2011) found the results in support of the EKC, while the results of estimations in Chang (2015) illustrate that the EKC in BRICS countries is U-shaped, which is opposite to conventional findings.

There is no common conclusion in literature on the long-run relationships between CO₂ emissions and income. There is also no expanded research on EKC for Russia. This study thus seeks to fill the gap in literature.

Taking into account the mixed results in literature on the environmental Kuznets curve to eliminate omitted variables bias, the relationships between CO₂ emissions and economic growth following similar methodologies in previous studies of Hossain (2011), Ferhani et al. (2014), Kasman and Duman (2015) and Balaguer and Cantavella (2018) is expressed in the following three alternative equations:

c_t = β ₀ + β₁e_t + β₂y_t + β₃y_t² + β₄tr_t + β₅ed_t + ε_1t (1)

c_t = β ₀ + β₁e_t + β₂y_t + β₃y_t² + β₄tr_t + β₅u_t + ε_2t (2)

c_t = β ₀ + β₁e_t + β₂y_t + β₃y_t² + β₄tr_t + β₅ed_t + β₆u_t + ε_3t (3)

where c_t is CO₂ emissions per capita; e_t is commercial energy use per capita; y_t represents per capita real income, which is measured as GDP per capita; y_t² is the square of per capita income; tr_t is trade openness and measured in terms of sum of export and import as a share of GDP; ed_t is the level of education measured in the number of secondary education pupils; and u_t is the urban population as a share of total population.

The reason behind employing three alternative models, where education and urbanization are altered as explanation variables, is the robustness check to eliminate the omitted variables bias. It is discussed in literature that energy consumption is the key determinant in emissions changes; therefore, it is expected that β₁ has a positive sign (Suri and Chapman, 1998). The EKC hypothesis assumes that β₂ is positive and β₃ is negative, demonstrating an increase in emission with economic growth and further improvement in environmental quality after income reaches a certain threshold. Coefficients of trade openness, β₄, may be positive as well as negative. With an increase in the reallocation of the international trade production of some products between countries, pollution related to production is being reallocated as well (Halicioğlu, 2009). The coefficients of education and urbanization variables may have either positive or negative signs. Increased levels of urbanization and education provide better access to advanced technologies consumption, which is usually energy-intensive and leads to higher levels of pollution. However, improved levels of disposal organization and of hygiene provision may decrease emissions in urban areas (Gangadharan and Valenzuela, 2001). Increased educational levels expand social awareness of environmental problems and can lead to improvements in environmental quality (Balaguer and Cantavella, 2018).

The main aim of this study is to test the EKC hypothesis in the Russian Federation for the period 1991–2016. In order to avoid the possible problem of omitted variables in addition to trade, education and urbanization, variables are introduced to the model. Three models are estimated in this study. The first model employs energy use per capita, real income per capita, square of real income, trade and education as independent variables. The second model employs the variable of urbanization instead of education. Finally, the third model employs the variables of both education and urbanization in addition to main independent variables. In all three models, carbon dioxide emission per capita is employed as a dependent variable, a measure of environmental quality. The results provide enough evidence to conclude that the EKC hypothesis is valid for Russia, indicating an increase in carbon dioxide emissions with economic growth; however, after income reaches a certain threshold, CO₂ declines. In all three models, the error correction term is significant with an expected negative sign implying that deviations from long-run equilibrium are adjusted by estimated variables in about one year for all models. The results of Granger’s causality test demonstrate bidirectional Granger causality relationships between the pairs of CO₂ and energy use, CO₂ and real income, CO₂ and the square of real income, and CO₂ and education. A one-way causality runs through CO₂ to trade and there is no causality between CO₂ and urbanization. In order to measure the stability of model parameters, the CUSUM and CUSUMSQ techniques and the Chow forecast test were employed. The results provide evidence of parameter stability only for the second model, where the results of tests are inconclusive for the first model and reject the stability of parameters in the third model. Therefore, only the results of the second model may be employed for policy in Russia. From the results of the second model estimations, it can be concluded that economic growth has a short-run impact on CO₂, energy use determines CO₂ in both the short and long runs, while urbanization has significant high long-run impact on emissions.

Based on the results of the estimations, a number of policy implications can be derived for the Russian Federation. The results support the EKC hypothesis, which states that the relationships of CO₂ emissions and economic growth have an inverted U-shape. Therefore, policy-makers should continue to implement policies to sustain economic development that leads to the use of cleaner technologies for lower carbon dioxide emissions. There is strong evidence that energy use has a destructive impact, while the level of urbanization has a refining effect on environmental quality. Therefore, policy-makers should focus on decreasing energy intensity, maintaining environmental policies in urban areas and implementing efficient environmental policies in rural areas to decrease the difference between urban and rural areas. Finally, education in Russia is found to have a damaging effect on environmental quality; therefore, new policies should be directed to increase environmental awareness among the population.

The ARDL framework for Models 2 and 3, respectively is expressed as follows:

(A1)

(A2)

The ECM for Models 2 and 3, respectively is expressed as follows:

(A3)

(A4)