CO2 emissions, real output, energy consumption, trade, urbanization and financial development: testing the EKC hypothesis for the USA
This study aims to investigate the relationship between carbon dioxide (CO2) emissions, energy consumption, real output (GDP), the square of real output (GDP2), trade openness, urbanization, and financial development in the USA for the period 1960–2010. The bounds testing for cointegration indicates that the analyzed variables are cointegrated. In the long run, energy consumption and urbanization increase environmental degradation while financial development has no effect on it, and trade leads to environmental improvements. In addition, this study does not support the validity of the environmental Kuznets curve (EKC) hypothesis for the USA because real output leads to environmental improvements while GDP2 increases the levels of gas emissions. The results from the Granger causality test show that there is bidirectional causality between CO2 and GDP, CO2 and energy consumption, CO2 and urbanization, GDP and urbanization, and GDP and trade openness while no causality is determined between CO2 and trade openness, and gas emissions and financial development. In addition, we have enough evidence to support one-way causality running from GDP to energy consumption, from financial development to output, and from urbanization to financial development. In light of the long-run estimates and the Granger causality analysis, the US government should take into account the importance of trade openness, urbanization, and financial development in controlling for the levels of GDP and pollution. Moreover, it should be noted that the development of efficient energy policies likely contributes to lower CO2 emissions without harming real output.
KeywordsCO2 emissions Financial development Urbanization Trade Energy Output
The environmental Kuznets curve (EKC) hypothesizes that the relationship between environmental quality and real output has an inverted U-shaped feature as environmental degradation first increases until a certain level of GDP and then decreases with increases in GDP. Akbostancı et al. (2009), Lee and Lee (2009), Fodha and Zaghdoud (2010), and Saboori et al. (2012) among others analyze the possible presence of the EKC hypothesis in Turkey, panel of 109 countries, Tunisia, and Malaysia, respectively. In addition, a group of study focus on the relationship between GDP per capita (or economic growth) and energy consumption for a variety of countries and regions (Soytas and Sari 2003; Wolde-Rufael 2005; Ozturk 2010; Shahbaz and Lean 2012; Smyth and Narayan 2014; Dogan 2014, 2015a, 2015b; Aslan 2014; Komal and Abbas 2015; Shahbaz et al. 2015). As one of the leading works in the literature, Ang (2007) combines the energy–output nexus and environment–output nexus under a modified EKC framework in which carbon dioxide (CO2) emissions as a proxy for environmental degradation are regressed on energy consumption, GDP per capita, and the square of GDP per capita. Ang (2007) supports the validity of an inverted U-shaped relationship between carbon dioxide emissions and output in France as CO2 is positively impacted by GDP per capita and negatively impacted by the square of GDP per capita. The EKC hypothesis begins to be widely analyzed in the energy literature after energy consumption is inserted as an additional explanatory variable into the conventional EKC model (Soytas et al. 2007; Zhang and Cheng 2009; Soytas and Sari 2009; Ozturk and Acaravci 2010; Wang et al. 2011; Nasir and Rehman 2011; Kanjilal and Ghosh 2013; Salahuddin and Gow 2014; Kasman and Duman 2015; Bastola and Sapkota 2015; Baek 2015).
The existing literature about the energy–environment–output nexus is abundant (Al-Mulali et al. 2015a). Therefore, the linkage between GDP, energy consumption, and environmental degradation should be tested by taking into account particular segments of the economy rather than by testing the validity of the EKC hypothesis using the simple econometric model in which CO2 emissions are regressed on real output, the square of real output, and energy consumption. By means of including additional variables into the simple model, the state-of-the-art attempt to eliminate the omitted-variable bias problem. Trade openness can be considered as a commonly used variable in the literature (Ang. 2009; Halicioglu 2009; Jalil and Mahmud 2009; Nasir and Rehman 2011; Jayanthakumaran et al. 2012). Farhani et al. (2014) decompose the effect of trade on pollution, energy consumption, and output into three components, namely scale, composition, and technique. The scale effect basically implies that the increases in the amount of trade influence output, energy consumption, and thus, CO2 emissions. The composition effect refers to the re-allocation in a country’s traded goods basket. In other words, free trade enables the country to specialize on the production of goods for which it has comparative advantage. Hence, the use of energy and environmental quality may increase or decrease depending on whether or not the sectors that the country specializes need more energy. The technique effect means that trade liberalization leads to environmental improvements since the technology gets better in producing goods and using energy more efficiently.
Another variable that recently begins to be introduced into the simple model is urbanization. Martínez-Zarzoso and Maruotti (2011) argue the possible impact of urbanization on environmental degradation through several channels. At the most basic interpretation, the increase in urban population results in higher industrial output, transportation, and energy consumption and gas emissions. Hossain (2011), Sharma (2011), and Kasman and Duman (2015) add trade and urbanization into the simple EKC model.
Financial development is the recent additional variable used by the recent works in the literature. Financial development may lead to lower financing costs and better and larger financing networks through which enterprises can have higher opportunity to make more investment and buy new machines and equipment, resulting in more energy consumption and CO2 emissions. Because financial development likely links to cheaper personal loan rates, it may trigger consumers to purchase houses, cars, and durable goods (i.e., refrigerator and dish washer), which increases output, energy consumption, and gas emissions. On the other hand, financial development may detracts energy consumption and gas emissions as it can potentially stimulate the efficiency of business performance as well as energy efficiency (Tamazian et al. 2009). Sadorsky (2010) and Aslan et al. (2014) find significantly positive relationship between energy consumption and financial development. Zhang (2011) claims that financial development negatively impacts environmental degradation. Islam et al. (2013) find financial development and economic growth to have impact on energy consumption. Tang and Tan (2014) reveal that financial development affects economic growth and energy consumption influences financial development. Al-Mulali and Lee (2013) indicate that GDP, financial development, urbanization, and trade positively impact energy consumption. In addition, Tamazian et al. (2009), Tamazian and Rao (2010), Jalil and Feridun (2011), Ozturk and Acaravci (2013), Shahbaz et al. (2013a, b), and Omri et al. (2015) include trade and financial development into the simple EKC model.
Al-Mulali et al. (2015b) and Farhani and Ozturk (2015) include trade, urbanization, and financial development as separate sectors of the economy into the simple EKC model so as to attempt to obtain unbiased effect of energy consumption on carbon dioxide emissions and to test the validity of EKC hypothesis. Only these two studies, to the best of our knowledge, simultaneously take into account the possible effects of trade, urbanization, and financial development, and only a few studies consider the impacts of one or two variables among financial development, trade, and urbanization while the influence of energy consumption and GDP on environmental quality is investigated. Furthermore, these studies reach different conclusions in terms of whether trade, urbanization, and financial development negatively or positively impact carbon dioxide emissions and whether or not the EKC hypothesis is present.
In light of the abovementioned arguments, the fundamental contribution of this study is that for the first time in the literature ,this study aims to analyze the relationship between carbon dioxide emissions, energy consumption, real output, trade, urbanization, and financial development for the USA in an econometric model based on the EKC hypothesis. Given that the analyzed variables are connected to each other, this study also tries to eliminate the omitted-variable bias problem. Because this type of investigation is a significant gap in the literature and, thus, the outcome has a high value in terms of policy implications, we focus on a single-country study rather than a panel study. According to the World Development Indicators, the USA is famous for large amount of energy consumption, output, and gas emissions. The USA is also one of the top countries in the world in terms of trade openness, urbanization, and financial development. In addition, the USA is an important country in terms of its impressive regional and world affairs and its role in the energy market since it is among the NAFTA countries, one of the G7 countries, and one of the five permanent members of the United Nations. The rest of the study is as follows: the “Literature review” section provides a literature review, the “Methods and data” section explains the methods and data, the “Empirical results” section reveals the empirical results, and the “Conclusions” section concludes the aims and findings.
A number of existing studies including Ang (2007), Iwata et al. (2010), Hamit-Haggar (2012), Saboori et al. (2012), Tiwari et al. (2013), Lau et al. (2014), Yavuz (2014), and Osabuohien et al. (2014) among others investigate the relationship between income and carbon dioxide emissions and validate the existence of the EKC hypothesis in a variety of countries and regions. In addition, various studies validate the EKC hypothesis using the following panel data: for instance, Skaza and Blais (2013) for 190 developing and developed countries; Al-Mulali and Sheau-Ting (2014) for 189 countries from six different regions, namely Asia Pacific, Eastern Europe, the Americas, Middle East and North Africa (MENA), Sub-Saharan Africa, and Western Europe; Omri et al. (2015) for MENA countries; Ziaei (2015) for European, East Asian, and Oceania countries; and Al-Mulali et al. (2015c) for 93 countries. Moreover, Ang (2007), Jalil and Mahmud (2009), Alam et al. (2012), Ozturk and Acaravci (2013), Shahbaz et al. (2013a), Alkhathlan and Javid (2013), and Boutabba (2014), using time series data, also support the empirical presence of the EKC hypothesis for France, China, Turkey, Bangladesh, South Africa, Saudi Arabia, and India. On the other hand, Al-Mulali et al. (2015a) and Farhani and Ozturk (2015) find controversial results not supporting the validity of EKC for Vietnam and Tunisia.
As seen above, different studies reach conflicting results as to the effect of income on the environment. These differences may occur from omitted-variable bias problem, the choice of specific functional forms (econometric techniques), and sample selection bias. To overcome the omitted-variable bias problem, several studies include different variables ranging from financial development, trade, foreign direct investment, and energy consumption to energy prices, labor, gross fixed capital formation, and urbanization (Farhani and Ozturk 2015; Shahbaz et al. 2015; Al-Mulali et al. 2015a; Komal and Abbas 2015). In the literature, there are some efforts attempting to examine environmental pollution including the impact of trade. Ang (2009) explores the estimation of the Chinese pollution function using CO2 emissions (as an endogenous variable) and GDP, energy use, and trade openness (as exogenous variables) over the annual period of 1953–2006. The findings indicate that more energy use, GDP, and trade openness lead to more CO2 emissions. In the same way, Halicioglu (2009) suggests the dynamic causal relationships between CO2 emissions, GDP, energy consumption, and foreign trade in Turkey over the period of 1960–2005. Jalil and Mahmud (2009) extend the methodology of Halicioglu (2009) for China over the period of 1975–2005. The findings also indicate that CO2 emissions can be determined by GDP and energy consumption, while trade has insignificant impact on CO2 emissions in the long run. Jayanthakumaran et al. (2012) test the long-run and short-run relationships between CO2 emissions, growth, energy use, trade, and endogenously determined structural breaks for both China and India over the period of 1971–2007. Using the autoregressive distributed lag (ARDL) approach to cointegration, the findings indicate that CO2 emissions in China are determined by real GDP, energy consumption, and structural changes while no causal relationship is detected for India.
The financial development factor is recently included in the environmental function through the works of Jalil and Feridun (2011), Ozturk and Acaravci (2013), and Shahbaz (2013). The study of Jalil and Feridun (2011) discusses the impact of economic growth energy consumption, trade openness, and financial development on carbon emissions in China from 1953 to 2006. The findings show that financial development has no significant impact on carbon emissions in the long run, while economic growth, energy consumption, and trade openness present significant impacts on carbon emissions. In addition, Ozturk and Acaravci (2013) investigate the causal relationship between carbon emissions, GDP, energy consumption, trade openness, and financial development in Turkey over the period of 1960–2007. The findings show that an increase in trade openness leads to an increase in carbon emissions, while financial development has no significant effect on carbon emissions in the long run. Finally, Shahbaz (2013) examines the relationship between financial instability and the environmental degradation within the presence of GDP, energy consumption, and trade openness in Pakistan over the period of 1971–2009. The empirical findings indicate that the long-run relationship between variables can be detected and financial instability may increase the environmental degradation.
The inclusion of urbanization in the environmental function presents an intense debate for discussion, especially in terms of environmental and regional development. There are, however, limited works (Hossain 2011; Sharma 2011; Kasman and Duman 2015) that have documented the importance of the inclusion of urbanization in the relationship between CO2 emissions, economic growth, energy consumption, and trade. Hossain (2011) investigates the relationship between gas emissions, energy consumption, real output, energy consumption, trade, and urbanization for newly industrialized countries over the period 1971–2007. The empirical findings show that unidirectional causality runs from real output and trade openness to CO2, from real output to energy consumption, from trade openness to real output, from urbanization to GDP, and from trade openness to urbanization in the short run, although there is no long-run causal relationship between the analyzed variables. Sharma (2011) examines the relationship between environmental quality, energy consumption, GDP, energy consumption, openness, and urbanization for a panel of 69 countries for the years 1985–2005. The results indicate that trade openness, output per capita, and energy consumption lead to environmental degradation while CO2 is negatively impacted by urbanization. Kasman and Duman (2015) analyze the causal linkage between gas emissions, energy consumption, real output, energy consumption, trade, and urbanization for new European Union members and candidate countries over the period 1992–2010. The study supports the evidence of the EKC hypothesis in the analyzed countries; in addition, the fully modified ordinary least squares (FMOLS) regression presents that openness and urbanization have a positive effect on the level of gas emissions. Moreover, one-way causality is detected from energy consumption, trade openness, and urbanization to CO2; from GDP to energy consumption; from GDP, energy consumption, and urbanization to trade openness; from urbanization to GDP; and from urbanization to trade openness.
Al-Mulali et al. (2015b) and Farhani and Ozturk (2015) are the only published studies considering the effects on environmental quality of urbanization, trade, and financial development in conjunction with energy consumption and GDP. Al-Mulali et al. (2015b) analyze the long-run relationship between CO2, energy consumption, real output, urbanization, trade openness, and financial development for a panel of 129 countries from four groups: namely low-income countries, lower middle-income countries, upper middle-income countries, and high-income countries. The results from dynamic ordinary least squares (DOLS) show that energy consumption leads to environmental degradation in all groups while financial development improves the environmental quality in the four groups. In addition, urbanization and GDP have negative and positive effects on CO2 in the three groups, respectively. Finally, trade openness has no significant effect in one group, negative effect in two groups, and positive effect in one group. Farhani and Ozturk (2015) investigate the linkage between gas emissions, GDP, the square of GDP, energy consumption, urbanization, trade, and financial development for Tunisia over the period 1971–2012. According to the results obtained from ARDL approach, all of the analyzed variables lead to environmental degradation. In addition, the EKC hypothesis is not valid in Tunisia since the coefficients on GDP and the square of GDP are positive and statistically significant. The empirical findings of the Granger causality test indicate that long-run causality runs from output, energy consumption, financial development, openness, and urbanization to gas emissions as well as from CO2, output, energy consumption, openness, and urbanization to financial development.
According to the above survey of the literature, the empirical studies fail to achieve unanimous conclusion regarding the effects of urbanization, financial development, and trade openness as well as the validity of the EKC hypothesis. The main reason for the discrepancy in results in the previous research comes from data characteristics, estimation techniques (cointegration methods and causality tests), and development level of the country on which a study is conducted.
Methods and data
The relationship between per capita carbon emissions, per capita real income, and the square of per capita real income, per capita energy consumption, trade openness, urbanization, and financial development in the USA is performed in four steps. First, we test the integration properties of CO2, GDP, GDP2, EC, TR, URB, and FD. Second, in case that they are non-stationary, the long-run relationship among the analyzed variables is investigated using the ARDL bounds testing approach of cointegration. Third, assuming that the variables are cointegrated, the short-run and long-run coefficients on real output, the square of real output, energy consumption, openness, urbanization, and financial development are estimated. Last, we test the causal relationship between the analyzed variables using the error correction-based causality models.
Unit root tests
This study uses the following unit root tests: the augmented Dickey–Fuller test due to Dickey and Fuller (1979) and the Zivot–Andrews test with one structural break due to Zivot and Andrews (2002). The augmented Dickey–Fuller (ADF) unit root test is employed to test the integration level of the variables. A well-known weakness of the ADF unit root test is its potential confusion of structural breaks in the series as evidence of non-stationarity. In other words, it may fail to reject the unit root hypothesis if the series has a structural break. For the series that is found to be I, there may be a possibility that they are, in fact, stationary around the structural break(s), I(0), but are erroneously classified as I(1). To overcome this, the Zivot–Andrews (ZA) unit root test is employed. The ZA unit root test allows for one structural break. In this test, the null hypothesis is that the series has a unit root with structural break against the alternative hypothesis that they are stationary with break. We apply the ZA test with one-time changes in the level and slope of the trend function of the series.
ARDL approach to cointegration
Having established the existence of a long-run relationship based on the F- test, the second step of the ARDL analysis is to estimate the long-run and the associated short-run coefficients. The long-run effects of GDP, GDP2, EC, URB, TR, and FD on CO2 are the estimates of − (γ2\γ1), − (γ3\γ1), − (γ4\γ1), − (γ5\γ1), − (γ6\γ1), and − (γ7\γ1) in Eq. 2. Furthermore, the short-run effects of each explanatory variable on the response variable are posed by the coefficient estimates of the first-differenced series in Eqs. 3–8. For instance, the short-run effects of EC and TR on gas emissions are posed by the estimates of δ4k and δ6k in Eq. 3. The order of the lags in the ARDL model is selected using the Akaike information criterion (AIC) ensuring that there is no evidence of residual serial correlation, functional form misspecification, non-normality, and heteroscedasticity.
Error correction-based Granger causality analysis
Unit root test
Unit root analysis
The conventional augmented Dickey–Fuller (ADF) unit root test indicates that urbanization is stationary at level; however, CO2, GDP, energy consumption, trade openness, and financial development have unit root at levels but become stationary in their first differences. Thereafter, the Zivot–Andrews (ZA) unit root test with one structural break is used to confirm the findings of the ADF unit root test. According to the results obtained from the ZA unit root test, only urbanization does not have a unit root at level; on the other hand, gas emissions, GDP, EC, TR, and FD are not stationary at levels but become stationary in their first differences. Both the ADF unit root test and the ZA unit root test reach the same conclusion. In short, urbanization is determined to be I(0) while CO2, GDP, energy consumption, trade, and financial development are determined to be I(1). Because neither of the analyzed variables I(k) where k > 1, we can proceed to the bounds testing for cointegration.
ARDL approach to cointegration
Cointegration test results
5 % critical values
f(CO2/GDP, GDP2, EC, URB, TR, FD)
Estimated coefficients from ARDL model
(A) Long-run estimates (dependent variable CO2)
(B) Short-run estimates (dependent variable ∆CO2)
ECTt − 1
F - test
Short-run and long-run estimates
The short-run and long-run estimates of GDP, the square of GDP, energy consumption, trade, urbanization, and financial development are reported in Table 3. Moreover, the coefficient estimates of the analyzed variables are economically equal to the elasticity of CO2 with respect to GDP, GDP2, EC, TR, URB, and FD, respectively, because the time series data are transformed into their logarithmic forms. Both the short-run and the long-run elasticity of carbon dioxide emissions with respective to GDP are negative and statistically significant, while the coefficient estimates of the square of GDP are statistically significant and positive both in the short run and long run. In other words, the short-run and long-run estimates of GDP are −1.63 and −2.13, respectively, and the short-run and long-run coefficient estimates of the square of GDP are +0.18 and +0.22, respectively. In the existence of the EKC hypothesis, the effect of GDP and GDP2 on carbon dioxide emissions is expected to be positive and negative, respectively. Because it is not the case for the analyzed variables, the EKC hypothesis is not present in the USA. On the other hand, there is a U-shaped relationship between the level of income and gas emissions. Therefore, the increase in the level of GDP leads to environmental improvements until a certain level but then the increase in the level of income causes environmental degradation. This finding is consistent with Chandran and Tang (2013), Baek (2015), Farhani and Ozturk (2015), and Al-Mulali et al. (2015a).
Granger causality analysis
ECTt − 1
Granger causality test
In light of the evidence of cointegration relationship among the analyzed variables, it is an interest for researchers to perform the Granger causality test so as to pin down appropriate economic policies, environmental policies, and energy strategies by understanding the directions of causality between CO2, GDP (GDP2), EC, TR, URB, and FD. Henceforth, the Granger causality in the vector error correction mechanism is used to exploit the directions of causality between the aforementioned variables as well as to decompose the directions of causality into the short run and long-run effects.
The results obtained from the VECM Granger causality test are reported in Table 4. In the short run, there is bidirectional causality between (1) CO2 and GDP, (2) gas emissions and energy consumption, (3) carbon dioxide emissions and urbanization, (4) GDP and urbanization, (5) trade and urbanization, (6) financial development and trade, and (7) income and trade openness. In addition, we have enough evidence to support one-way causality running (1) from GDP to energy consumption, (2) from financial development to income, and (3) from urbanization to financial development. Lastly, no causality is determined between (1) CO2 and trade openness, (2) carbon dioxide emissions and FD, (3) energy consumption and urbanization, (4) EC and trade, and (5) energy consumption and FD. In the long run, there is an evidence of four causal relationships, namely (1) from GDP, EC, URB, TR, and FD to CO2; (2) from gas emissions, income, trade, urbanization, and financial development to energy consumption; (3) from CO2, GDP, EC, TR, and FD to URB; and (4) from carbon dioxide emissions, income, energy consumption, TR, and URB to financial development.
These findings are consistent with those found in the preceding section. More precisely, energy consumption is found to have impact on environmental degradation both in the short run and long run. In addition, an increase in the level of energy use does not cause income. Thus, the USA may decrease energy consumption without harming the GDP for the sake of environmental quality. Furthermore, the government may encourage and financially support the institutions, universities, and researchers to propose project on increasing the efficiency of energy and on the application of the methods of environmental protection. As expected, income (the square of income) has causal relationship with gas emissions. This also supports the presence of the U-shaped relationship between CO2 and GDP in the USA. Although trade openness and financial development do not cause the Granger environmental quality in the short run, TR and FD have impact on it in the long run. Also, urbanization is found to be a cause of gas emissions and income. Moreover, trade and financial development influence GDP as well. Henceforth, the US government should take into account the importance of trade openness, urbanization, and financial development in controlling for the level of GDP and pollution.
This study examines the relationship between carbon dioxide emissions, energy consumption, real output, the square of real output, trade openness, urbanization, and financial development in the USA for the period 1960–2010. To analyze this relationship, we use unit root tests to find out the stationarity properties of the analyzed variables, the ARDL bounds testing approach to explore the possible cointegration between the variables, and short-run and long-run estimates, Granger causality test based on VECM, to reveal the short-run and long-run causal relationships between analyzed variables.
According to the results obtained from augmented Dickey–Fuller and Zivot–Andrews unit root test, we claim that urbanization is stationary at level and CO2, energy consumption, real output, the square of real output, trade openness, and financial development are stationary at first differences. Then, the ARDL approach to cointegration test indicates that the analyzed variables are cointegrated at 5 % level of significance. The short-run and long-run estimates show that energy consumption is the main cause of CO2 emissions in the USA. In addition, urbanization has positive impact on gas emissions. Furthermore, the estimates of real output and the square of real output suggest that there is strong evidence against the existence of an EKC-type relationship in the USA.
Last, the Granger causality analysis presents that there is strong causal relationship between gas emissions and real output, energy consumption, and urbanization both in the short run and long run. By putting together the results from the short-run and long-run estimates, and Granger causality tests, the US government should take into account the importance of trade openness, urbanization, and financial development in controlling for the levels of GDP and pollution. While the above analysis provides interesting insights, it should be noted that the development of efficient energy policies likely contributes to lower CO2 emissions while preserving real GDP. A promising extension of this work would be to consider the energy supply, rural development, and other environmental variables for the case of the USA.
- Al-mulali U, Tang CF, Ozturk I (2015b) Does financial development reduce environmental degradation? Evidence from a panel study of 129 countries. Environ Sci Pollut Res 1–10Google Scholar
- Dogan E (2014) Energy consumption and economic growth: evidence from low-income countries in Sub-Saharan Africa. Int J Energy Econ Pol 4(2):154–162Google Scholar
- Dogan E (2015b) The relationship between economic growth and electricity consumption from renewable and non-renewable sources: a study of Turkey. Renew Sustain Energy Rev 52:534–546Google Scholar
- Engle RF, Granger CW (1987) Co-integration and error correction: representation, estimation, and testing. Econometrica: J Econo Soc 251–276Google Scholar
- Farhani S, Ozturk I (2015) Causal relationship between CO2 emissions, real GDP, energy consumption, financial development, trade openness, and urbanization in Tunisia. Environ Sci Pollut Res 1–14Google Scholar
- Skaza JS, Blais BS (2013) The relationship between economic growth and environmental degradation: exploring models and questioning the existence of an environmental Kuznets curve. Cent Glob Econ Stud Bryant Univ Work Pap (2013-05)Google Scholar
- Smyth R, Narayan PK (2014) Applied econometrics and implications for energy economics research. Energy EconGoogle Scholar