Industrial growth and emissions of CO₂ in Ghana: The role of financial development and fossil fuel consumption

Abokyi, Eric; Appiah-Konadu, Paul; Abokyi, Francis; Oteng-Abayie, Eric Fosu Article Industrial growth and emissions of CO₂ in Ghana: The role of financial development and fossil fuel consumption Energy Reports Provided in Cooperation with: Elsevier Suggested Citation: Abokyi, Eric; Appiah-Konadu, Paul; Abokyi, Francis; Oteng-Abayie, Eric Fosu (2019) : Industrial growth and emissions of CO₂ in Ghana: The role of financial development and fossil fuel consumption, Energy Reports, ISSN 2352-4847, Elsevier, Amsterdam, Vol. 5, pp. 1339-1353, https://doi.org/10.1016/j.egyr.2019.09.002 This Version is available at: https://hdl.handle.net/10419/243674 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Energy Reports 5 (2019) 1339–1353 Contents lists available at ScienceDirect Energy Reports journal homepage: www.elsevier.com/locate/egyr Research paper Industrial growth and emissions of CO2 in Ghana: The role of financial development and fossil fuel consumption Eric Abokyia, Paul Appiah-Konadub, Francis Abokyic, Eric Fosu Oteng-Abayied,∗ aDepartment of Economics, Universita‘ Politecnica delle Marche, Italy bDepartment of Economics and Management, University of Brescia, Italy cDepartment of Economics, University of Cape Coast, Ghana dDepartment of Economics, Kwame Nkrumah University of Science and Technology, Ghana article info Article history: Received 4 June 2019 Received in revised form 24 August 2019 Accepted 3 September 2019 Available online xxxx JEL classification: Q43 Q53 Q54 Q56 Keywords: ARDL cointegration EKC hypothesis Fossil fuel Carbon emission Ghana abstract The rapid rise in greenhouse gas emissions have become a global concern catching the attention of policy makers and researchers all over the world. Fossil fuel combustion has been named as the major source of greenhouse gas emissions, meanwhile, studies focusing on fossil fuel impact on CO2 emissions are rare for developing countries including Ghana. This study employed the ARDL procedure with structural breaks and the Bayer–Hanck joint cointegration approach to examine the validity of the EKC hypothesis in the dynamic linkage between industrial growth and emissions of carbon dioxide (CO2) in Ghana, capturing the role of fossil fuel consumption and financial development. The variables are found to be cointegrated and both the short-run and the long-run parameters showed evidence of a U-shaped relationship between industrial growth and CO2 emissions which was further confirmed by the Lind and Mehlum U-test. The short-run causality revealed a uni-directional causality running from fossil fuel consumption to emissions of CO2. For policy purposes, the study advocates for efficient and low carbon emission technologies. ©2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Climate change and environmental degradation are two (2) of the key development-related challenges in the course to achieve sustainable global output growth (United Nations,2016). The need to safeguard environmental quality has therefore taken the center stage in both national and international development discourse in the last three decades in the quest to promote sustainable development globally (United Nations,2013). Global warming and air pollution have been identified as some of the key causes of climate change; and CO2 emissions have been generally considered in literature as a significant contributor to these problems (Ali et al.,2016). Stocker et al. (2013) reported that CO2 emissions is a key contributing factor to the high emissions of green-house gases (GHG) globally. The report further emphasized that 76.7% of GHG emissions emanate from emissions of CO2 mainly coming from developing countries such as Ghana in the effort to accelerate economic growth and increase income levels with little or no recourse to environmental provisions. ∗Corresponding author. E-mail address: [email protected] (E.F. Oteng-Abayie). Several attempts have been made through international conventions and intergovernmental agreements to minimize the destructive effects of global warming by advocating for the reduction of global emissions by both developed and developing countries. Notable among these international environmental agreements are the Paris convention (2015) and the Kyoto protocol, which was adopted in 1997 under the United Nations Framework Convention on Climate Change (UNFCCC) with the underpinning aim of reducing GHG concentration in the atmosphere in order to minimize the pace of climate change (Ali et al., 2016). Ghana is a signatory to the climate change convention (2015) and has also ratified the Kyoto Protocol aimed at minimizing climate change and promoting sustainable development globally. In this regard, the Environmental Protection Agency (EPA), acting on behalf of the government developed a national climate change policy in 2012 to mitigate climate change through the promotion of sustainable smart investments in all sectors of Ghana’s economy. Nonetheless, environmental pollution and degradation of the environment is very rampant and continues to pose a big challenge to the sustainable development of the country. CO2 emissions in Ghana increased by close to 100% from 12.2Mt to 23.9Mt between 2000 and 2010 (Twerefou et al.,2015). https://doi.org/10.1016/j.egyr.2019.09.002 2352-4847/©2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). 1340 E. Abokyi, P. Appiah-Konadu, F. Abokyi et al. / Energy Reports 5 (2019) 1339–1353 List of abbreviations ADF Augmented Dickey–Fuller ARDL Autoregressive distributed lag ECT Error Correction Term EKC Environmental Kuznets Curve EPA Environmental protection agency FDI Foreign direct investment GFEVD Generalized forecast error variance decomposition GHG Greenhouse gas OLS Ordinary least squares PP Phillips Perron SIC Schwarz information criterion VECM Vector error correction model WDI World Development Indicators Fossil fuels supply more than 87% of the energy used globally (Lyman,2016). In Ghana, petroleum and other fossil fuels are the main sources of energy in the transport industry and also in electric power generation. The diagram in Fig. 2 shows that the consumption of fossil fuel is becoming increasingly important than the other types of energy in Ghana. Fossil fuel consumption which accounted for 22% of total energy in 1971 increased significantly to 52% of total energy consumption in 2014 as shown in Fig. 1. Beginning 2009, fossil fuel consumption has taken the first position in Ghana’s energy consumption mix and continue to rise whiles the consumption of other types of energy (Alternative and nuclear energy & Combustible renewals and waste) decline on the average (WDI, 2017). Ghana is acclaimed as one of the fast-growing economies in Sub-Sahara Africa, having achieved an annual average GDP growth rate of above 5% in the last three decades (IMF,2016). Consistent with the sustained output growth, Ghana’s industrial sector has expanded tremendously and accounted for 23.68% of the country’s GDP in 2017 (GSS,2018). A point worth noting is the fact that, an important catalyst required to catapult the rising output levels and to sustain the continuous expansion of the industrial sector is energy. In 2011, Ghana’s oil import bill comprising of crude oil and other petroleum products rose significantly by about 50% from its 2010 level to catapult the historical growth rate of 15% the country achieved (CEPA,2012). Expanding industrial output, which thrives on fossil fuel consumption to power industrial machines and reduced carbon sink resulting from forest depletion have been touted as key contributing factors to the persistent rise in CO2 emission levels in Ghana from 3817.35kt in 1990 to 14466.32kt in 2014 (WDI,2017). Grossman and Krueger (1995) assert that there exists an inverted U-shape relationship between growth in income and environmental degradation. In the literature, this relationship has been termed the Environmental Kuznets Curve (EKC) hypothesis. The fundamental principle of the EKC is that as an economy expands, the benefits of increasing output are so large at the early stages of the growth process such that they dominate the increased demands for environmental quality caused by a rising income and thus, increase in output comes with more environmental degradation. A number of theoretical explanations have been advanced to rationalize the EKC hypothesis. Israel and Levinson (2004, p.2) assert that ‘the inverted U-shaped pollution-income path reflects the natural progression of economic development from clean agrarian economies to dirty industrial economies to clean service economies’. Gangadharan and Valenzuela (2001) point out that clean environment is a typical example of luxurious goods, whose demand rises only after individuals or a country have achieved economic prosperity to a level where they no longer have economic struggles. In line with this idea, Inglehart (1997) in his post-materialism hypothesis also postulate that people pay attention to post materialist values such as environmental quality only after becoming sufficiently affluent to the point that they have no serious material needs. On this premise, Bruneau and Echevarria (2009) argue that the demand for environmental improvement would be low in developing countries with low levels of income, because the poor are ‘too poor to green’. The Pollution Haven Hypothesis also asserts that governments in most developing nations are hesitant to place strict environmental standards on their firms in order to boost the competitiveness of local firms in the global market (Busse, 2004). The Environmental Kuznets Curve and the Pollution Haven Hypothesis are the basis for the assertion in the literature that, increased output to meet the needs of high and rising human population (which requires high consumption of fossil fuels) will most likely lead to high emissions of GHGs such as CO2 and general environmental degradation in a developing country such as Ghana, where environmental regulations are usually not strictly enforced (Antweiler et al.,2001). In the literature, available studies on the EKC have mostly focused on the nexus between energy consumption in general and CO2 emissions (Sarkodie and Strezov,2018;Aboagye,2017; Asumadu-Sarkodie and Owusu,2016;Twerefou et al.,2016;Begum et al.,2015;Gökmenoğlu and Taspinar,2015;Kwakwa et al., 2014;Omisakin,2009). Bilgili et al. (2016) uses a panel data set of 17 OECD countries from 1977–2010, employing the panel FMOLS and panel DOLS estimation techniques to analyze the validity of the EKC hypothesis focusing on renewable energy consumption and environmental quality nexus. The findings of their study give credence to the EKC hypothesis for the countries studied and indicate that the validity of EKC is not dependent on the levels of income of the sample countries. This paper is similar to the present study in the use of CO2 emissions as dependent variable. The key point of divergence between this paper and the present study has to do with the independent variables employed. Whiles they use renewable energy, the present study uses fossil fuels (non-renewable energy) as the key dependent variable of interest. Sarkodie and Strezov (2018,2019) examines the Environmental Sustainability and the Environmental Kuznets curve hypotheses for Australia, China, Ghana and the USA for the period 1971– 2013. The study reveals that the consumption of electric power is the main factor that determines energy intensity in the countries studied. The authors argue that enhancing energy efficiency in Ghana and the other selected countries will enhance energy security and minimize the adverse impacts of economic activities on the environment. Koçak and Şarkgüneşi (2018a) uses the EKC model to analyze the potential effect of FDI on CO2emission in Turkey from 1974–2013. The authors report that there exists a long-run nexus between FDI, economic growth, energy usage, and CO2emission. Increase in FDI inflows thus lead to a rise in CO2emissions in Turkey. More so, the results of this study points to the prevalence of the EKC hypothesis in Turkey. The main difference between this study and the current study is the use of different independent variables in the estimation. Whereas the authors use economic growth as the key independent variable to test for the validity of the EKC, we use industrial growth as a proxy for income growth in the present study. In a recent study, Sarkodie and Strezov (2019) employs the traditional review method to undertake a systematic review of the EKC hypothesis by tracking the historical trends of findings in relation to the EKC hypothesis. The authors report that among the studies that support the EKC hypothesis, the turning point of E. Abokyi, P. Appiah-Konadu, F. Abokyi et al. / Energy Reports 5 (2019) 1339–1353 1341 Fig. 1. Energy consumption. Source: Data Source: WDI (2017). Fig. 2. Energy consumption. Source: Data Source: WDI (2017). average annual income level is US$8910. This finding implies that a developing country like Ghana with an average income level far below the reported turning point is likely to experience increased environmental degradation with increased growth. The present study is different from the study by Sarkodie and Strezov (2019) in its use of empirical data to test the EKC hypothesis rather than undertaking a review of past studies. In another related study, Asumadu-Sarkodie and Owusu (2016) employs the VECM and the ARDL models to analyze the empirical nexus between CO2 emissions, GDP, energy use, and population growth in Ghana from 1971 to 2013. The authors performed Johansen’s multivariate co-integration and a variance decomposition analysis using Cholesky’s technique before testing for Granger causality based on VECM. The variance decomposition revealed that 21% of future shocks in CO2 emissions are caused by fluctuations in energy use, 8% result from fluctuations in GDP, and 6% of shocks result from fluctuations in population. The authors further report the existence of bidirectional causality from energy use to GDP and a unidirectional causality from carbon dioxide emissions to energy use, carbon dioxide emissions to GDP both in the long-run and short-run. Adom et al. (2012) investigates both the short-run and long-run nexus among CO2 emissions, GDP growth, technical efficiency, and industrial structure for Ghana, Senegal and Morocco. Their bounds test cointegration results point to the existence of a multiple long-run equilibrium nexus in the case of Ghana and Senegal but a one-way long-run equilibrium relationship in the case of Morocco. More so, the variance decomposition analysis show that economic growth significantly contributes to variations in CO2 emissions in Senegal and Morocco whereas in the case of Ghana, technical efficiency account for a greater part of the future variations in CO2 emissions. Kwakwa and Alhassan (2018) analyzes the impact of urbanization and energy consumption on CO2 emissions in Ghana in the framework of the EKC model using data for the period 1971–2013. Estimation with the FMOL affirm the existence of the EKC hypothesis in Ghana. The results further show that production of electricity from fossil fuels, industrialization and urbanization result in a rise in CO2 emissions in Ghana. A number of both theoretical and empirical studies in the literature have analyzed the effect of financial development on the quality of the natural environment. Frankel and Romer (1999) theorized that financial sector development enhances growth in output. Birdsall and Wheeler (1993) assert that financial sector development usually precedes the adoption of improved and environmentally-friendly technology through increased supply of financial credit to purchase such technology. On the other hand, Islam et al. (2013) argue that the development of the financial sector could enhance the level of energy consumption in an economy, and as a result lead to a rise in GHG emissions such as CO2 especially in countries where fossil fuels are used in energy production. Financial sector development attracts FDI and advanced technology that have minimal adverse impact on the environment (Ali et al.,2016). Hence the development of a country’s financial sector is expected to cause a reduction in the emissions of GHGs such as CO2 through the use of environmentally-friendly technology. From the foregoing, development of the financial sector could have a favorable or adverse effect on the environment. In theory, financial development impacts the environment via technology, income, improved regulations as well as capitalization (Yuxiang and Chen,2011). Using econometric techniques 1342 E. Abokyi, P. Appiah-Konadu, F. Abokyi et al. / Energy Reports 5 (2019) 1339–1353 and employing a provincial panel data from China, Yuxiang and Chen (2011) report that there is an empirical linkage between financial sector development and industrial pollution discharges. The authors observe that financial development has led to a significant improvement in environmental performance of China in recent times. Jalil and Feridun (2011) uses time series data from 1953 to 2006 on China and the ARDL bounds testing technique to analyze the long-run relationship between financial development and environmental pollution. The results of the analysis point to fact that financial development has resulted in a decline in environmental pollution in China. The authors conclude that CO2 emissions in China in the long term are caused mainly by GDP growth, increased consumption of energy and increased trade openness. More so, their findings give credence to the validity of the Environmental Kuznets Curve in China. Zhang (2011) uses data from China and econometric techniques (cointegration theory, Granger causality test, variance decomposition, etc.), to examine the effect of financial sector development on CO2 emissions. The authors report that financial development is a crucial driver for high CO2 emissions. Their results further show that, FDI has the least influence on variations of carbon emissions among the list of financial development indicators in China owing to its relatively smaller volume in relation to GDP Ghana has seen a drastic development in its financial sector since adopting financial sector deregulation as part of the Structural Adjustment Program (SAP) in 1986 and the establishment of the Ghana Stock Exchange in 1989. The number of commercial banks currently stands at about 25 alongside a large number of savings and loans and microfinance institutions. The number of companies listed on the Ghana stock exchange rose significantly from 11 in 1994 to more than 40 in 2016 with market capitalization of GHc56,120.08 as at 7th July, 2018 (GSE,2018). In response to the development of the country’s financial sector, net FDI inflows increased significantly from $15.6million in 1980 to $3,485.333 million in 2016 (GIPC,2017). There is a general consensus in the literature that, financial sector development as witnessed in Ghana in the last three decades facilitates domestic investment, minimizes financial risk as well as enhances the capital accumulation process; and thereby attract more FDI inflows and modern environmentally-friendly technologies which help to minimize environmental pollution. From the forgoing, understanding the dynamic relationships between fossil fuel consumption, industrial growth, financial development and CO2 emissions in a developing country such as Ghana is of great importance for policy-makers. Based on the review of the earlier studies relating to output growth and the environment within the framework of the EKC hypothesis, this study is the first of its kind covering the relationship between fossil fuel consumption, industrial output growth, financial sector development and CO2 emissions in Ghana. The proposed study is unique in its focus on fossil fuel consumption which is a principal source of carbon dioxide emissions globally as noted by US EPA (2016) instead of energy in general as used by most of the previous studies on Ghana (Sarkodie and Strezov,2018; Aboagye,2017;Asumadu-Sarkodie and Owusu,2016;Twerefou et al.,2016;Kwakwa et al.,2014) with the exception of Kwakwa and Alhassan (2018) who also used fossil fuel in their study. However, instead of Fully Modified OLS used in Kwakwa and Alhassan (2018), we employed the ARDL technique which produces more efficient results in small samples. Another uniqueness of the paper lies in the method of analysis employed, making provision for structural breaks in the time series variables, contrary to previous studies on Ghana. The study used two cointegration techniques to ascertain the robustness of the long-run nexus among the variables. These techniques include the Bayer and Hanck (2013) joint cointegration approach and the ARDL cointegration approach in the presence of structural breaks. Unlike Asumadu-Sarkodie and Owusu (2016), the study goes further to test for the validity of the environmental Kuznets curve hypothesis in the case of Ghana. Whiles the previous studies mostly relied on only the necessary condition to verify the EKC hypothesis, a procedure which has been described as ‘‘flawed’’ by Lind and Mehlum (2010), the present study conducted the Lind and Mehlum (2010) Utest which provides both the necessary and sufficient conditions for the existence of a U-shaped or inverted U-shaped relationship. In order to check the robustness of the causal relationships among the variables, the study conducted both the error correction term (ECT) augmented Granger-causality test and the generalized forecast variance decomposition. Finally, the focus on the nexus between development of the financial sector and environmental pollution is one more interesting dimension of this study. Thus, in a more specific terms, the proposed study makes the following five contributions to literature. First, we examine the role of financial development and fossil fuel consumption in the industrial growth–CO2 emissions nexus within the framework of the EKC hypothesis for Ghana. Second, the method of analysis employed makes provision for structural breaks in the variables as shown in the ARDL approach applied in the presence of structural breaks. Third, the EKC hypothesis is tested on industrial growth and CO2 emissions. Fourth, the sufficient condition for the existence of a U-shaped relationship between industrial growth and CO2 emissions is checked via the Lind and Mehlum U-test. Finally, the causal relationships between the variables are ascertained by conducting both the ECT augmented Granger-causality test and the generalized forecast variance decomposition to check the robustness of the causal relationships between the variables. The remaining part of the study has been structured as follows: section two is used to describe the data and the econometric methodology to be employed in the study. In section three, we present and discuss the empirical results. Section four will cover conclusions and policy implications emanating from the findings. 2. Data source and methodology 2.1. Data and model specification This study used annual time series data covering the period 1971–2014 from WDI (2017) version. This period was used since data on emissions of CO2 in Ghana was available for only the period under study. Data was collected on variables of interest such as CO2 emissions (metric tons per capita), Fossil fuel energy consumption (% of total) and transformed to fossil fuel as a ratio of total energy, Domestic credit provided by financial sector (% of GDP) and transformed into domestic credit provided by financial sector as a ratio of GDP (used to proxy financial development). Data was also collected on Industry value added (current LCU) due to missing values on the real versions of Industry value added in WDI. The Industry value added (current LCU) was transformed into industrial sector real value added using 2010 base year CPI. In this study, the relationship between financial development, industrial growth, fossil fuel consumption and carbon dioxide emission in Ghana was examined. In Ghana, fossil fuel consumption alone accounts for more than half of the total energy consumption. Fossil fuel consumption as a percentage of total energy consumed rose from around 22% in 1971 to about 52% in 2017 (WDI,2017). Meanwhile, according to US EPA (2016), fossil fuel consumption is the primary source of emissions of CO2 globally. The impact of Fossil fuel consumption on emissions of CO2 is also supported empirically (Kwakwa and Alhassan,2018;Nnaji et al., 2013). On financial development and CO2 emissions, a study by Yuxiang and Chen (2011) for China found that there exists a E. Abokyi, P. Appiah-Konadu, F. Abokyi et al. / Energy Reports 5 (2019) 1339–1353 1343 linkage between financial development and industrial pollution. Financial development has been shown in some studies to impact CO2 emissions positively whiles others showed negative relationship (Ali et al.,2016;Başarir and Çakir,2015;Zhang,2011;Jalil and Feridun,2011). This is due to the fact that, it can either lead to the adoption of energy efficient technologies or encourage increased consumption of energy. Financial development may increase the allocation of financial resources to domestic firms, making them able to purchase technologies which are friendly to the environment. Financial development can also lead to degradation of the environment. When financial resources are allocated to domestic firms, they able to increase their scale of production which may also raise the demand for certain types of energy which are not environmentally friendly. Generally, the activities of the industrial sector impact environmental degradation especially in developing countries where output increases and decreases with environmental pollution due to the inefficient technologies used. The impact of industrialization on CO2 emissions is also recorded in literature (Kwakwa and Alhassan,2018;Shahbaz et al.,2014). The proposed study derives the empirical model from the constant returns to scale Cobb–Douglas production function written as follows: Yt=f(At,Kt,Lt) (1) where Yis GDP, Ais technology, Kis capital and Lis labor. In theory, economic activities are the major causes of environmental degradation, therefore CO2 emissions (CE) is written as a function of economic activities (proxied by GDP) as follows: CEt=Yt=f(At,Kt,Lt) (2) Advancement in technology can be achieved through research and development and the latter is aided by financial development. Therefore, technology is replaced with financial development (FD) as done in Ali et al. (2016). The use of capital in production can be put into two categories, namely, polluting capital and non-polluting capital. Where polluting capital is the one that emits CO2 in production whiles non-emitting capital does not emit CO2. Examples of emitting capital include burning of coal and gas (fossil fuels) in production. Therefore, total capital used in production is written as follows: Kt=Kem +Kne (3) where Kem is the emitting capital and Kne is the non-emitting capital. Thus, the environmental degradation arising from the use of capital in production emanates from the polluting capital. Therefore, we replace capital in Eq. (2) with fossil fuel (emitting capital) similar to Ali et al. (2016) and Begum et al. (2015) where emitting capital was replaced with energy consumption for Malaysia. Next, since labor activities can be considered as economic activities but the proposed study focuses on industrial growth, we replaced labor with industrial output (I) similar to Ali et al. (2016) who replaced labor with GDP. From the foregoing, our CO2 emissions function is stated as: CEt=Yt=f(FDt,Ft,It) (4) In order to make interpretation easier by obtaining elasticity values, we applied natural logarithm to transform the variables to their natural log forms. Eq. (4) is written as follows: ln CEt=α0+α1ln Ft+α2ln It+α3ln FDt+ut(5) where ln CE is the emissions of CO2 (in metric tons per capita) expressed in natural log, ln Fis fossil fuel energy consumption as a ratio of total energy expressed in natural log. ln FD is the ratio of domestic credit by financial sector to GDP in natural log, a proxy for financial development. ln Iis the natural log of real industrial value added (a proxy for industrial growth), α0is the intercept and utis the disturbance term. The EKC hypothesis which describes an inverted U-shaped relationship between environmental quality and economic development was introduced and made popular by the writings of Grossman and Krueger (1991) and the World Bank Report (Shafik and Bandyopadhyay,1992;World Bank,1992). Following Grossman and Krueger (1995), we tested the EKC hypothesis on CO2 emissions and industrial growth for Ghana by including the squared term of the natural log of real industrial value added (ln I)2in the regression as shown below: ln CEt=α0+α1ln Ft+α2ln It+α3(ln It)2+α4ln FDt+ut(6) The expected sign of α1is positive since increase in fossil fuel consumption leads to a rise in emissions of CO2. From existing literature, α4can be either positive or negative because if increased credit to the private sector result in expansion and increased output of the industrial sector without change in technology, there will be more pollution. On the other hand, if the increased credit to the private sector leads to the adoption of improved and energy efficient technologies, pollution or CO2 emissions may reduce. Given the non-linear relationship between CO2 emissions and industrial growth in Eq. (1),α2has been suggested in literature to be either negative or positive depending on whether the relationship is U-shape or inverted U-shape. Provided there exists a U-shaped non-linear relationship between carbon and industrial growth, α2is expected to be negative and α3positive. In case of inverted U-shaped relationship, α2is positive and α3is negative. Thus, If the EKC holds (the case of the inverted U-shape), then α2>0 and α3<0. If instead of the EKC hypothesis, we have a U-shape, then α2<0 and α3>0. In each case, α2and α3must be statistically significant. If these conditions hold, then a turning point will exist for the U-shaped or inverted U-shaped relationship between carbon dioxide and industrial growth. By taking the first derivative of the natural logarithm of CO2 emissions (ln CE) with respect to the natural logarithm of industrial growth (ln I) and setting it to zero we are able to obtain the turning point when we solve for ln I. This gives: ln I∗=−α2 2α3 (7) If the EKC hypothesis holds, then this turning point implies that at the initial growth stages, pollution increases but at a diminishing rate up to a threshold beyond which further growth leads to improvement in the environment. In case of a U-shaped relationship, the reverse is expected, where the environment improves in the initial stages of growth up to a turning point or threshold and then begins to deteriorate beyond this point for any further attained growth. Whiles the EKC hypothesis augers well for sustainable development assumptions, the U-shaped relationship does not favor these assumptions as high growth and development also translates into increased pollution to the environment. This study examines the EKC assumptions for Ghana both in the long-run and short-run using econometric techniques outlined. However, it must be stated that the ARDL short-run and long-run results only provide the necessary conditions for the existence of a U-shaped or inverted U-shaped relationship between carbon dioxide emission and industrial growth as implied by Lind and Mehlum (2010). To confirm the existence of a U-shaped or inverted U-shaped relationship between industrial growth and CO2 emissions, the study goes further to conduct the Lind and Mehlum (2010) test for a U-shape or inverted U-shape. 1344 E. Abokyi, P. Appiah-Konadu, F. Abokyi et al. / Energy Reports 5 (2019) 1339–1353 2.2. Bayer–Hanck combined approach to cointegration The unit root properties of the variables were determined by employing both the Augmented Dickey–Fuller (ADF) and the Phillips Perron (PP) tests before proceeding to check for longrun cointegration relationship. However, these traditional unit root testing procedures are not appropriate in the presence of structural break since the break in the series is interpreted as unit root by these conventional unit root tests. In view of this, the study employed the Zivot and Andrews (2002) approach to test for unit root in the presence of structural break. There exist many cointegration approaches in literature employed to examine long-run relationship among variables which are I(1). One of these popular cointegration techniques extensively used in literature is Engle and Granger (1987). Later, Johansen (1991) cointegration approach which was more appropriate and preferred was introduced. Other popular cointegration tests include Phillips and Ouliaris (1990) cointegration test, Boswijk (1994) Structural Error Correction Model (ECM) and then Banerjee et al. (1998) t-test cointegration approach. Despite the popularity and extensive use of these cointegration techniques, they have come under severe criticisms. The explanatory power properties of these traditional cointegration techniques could lead to ambiguous empirical results (Shahbaz et al.,2018). A combined cointegration approach was later developed by Bayer and Hanck (2013) in order to overcome the weakness of these cointegration techniques which tends to cause bias in empirical results rendering it unreliable. In this newly developed cointegration technique, Bayer and Hanck used a combined cointegration approach where the results of Engle and Granger (1987), Johansen (1991), Boswijk (1994) and Banerjee et al. (1998) were combined to increase the power of the test. To employ this newly developed combined cointegration, variables must be I(1). This technique produces the Fisher F-statistics which are compared with the critical values in order to make a more accurate and reliable decision of the cointegration status of the variables. The null hypothesis which suggests no cointegration is rejected if the computed Fstatistic is greater than the critical value. The null is not rejected if the reverse is true. The Fisher’s formulae was employed by Bayer and Hanck (2013) to provide a combined p-values for the four cointegration tests in the equations below. Many studies in literature (Saint Akadiri et al.,2019;Shahbaz et al.,2015a,b; Rafindadi,2015;Farhani et al.,2014) also employed the Bayer and Hanck (2013) combined cointegration in investigating the longrun relationship between variables. The proposed study employs the joint Bayer and Hanck cointegration approach due to its superiority over the traditional cointegration techniques. EG −JOH = −2[ln(PEG)+ln(PJOH )](8) EG −JOH −BO −BDM = −2 ln [(PEG)+(PJOH )+(PBO)+(PBDM )](9) where PEG is the p-value of Engle and Granger (1987), PJOH is the p-value of Johansen (1991), PBO is the p-value of Boswijk (1994) and PBO is the p-value of Banerjee et al. (1998). Depending on the value of the Fisher F-statistic, long-run cointegration relationship will exist if the computed F-statistic is greater than the critical value and no long-run cointegration relationship if computed F-statistic is less than the critical value. 2.3. ARDL approach to cointegration In order to check the robustness of the cointegration result, the study employed the ARDL bound test cointegration approach in the presence of structural break to find the long-run relationship among the variables. This approach was chosen over the traditional cointegration techniques because it produces efficient and consistent result for small sample data. It is also appropriate for a mixture of I(0) and I(1) variables or mutually cointegrated variables. The unrestricted error correction model is stated as follows: ∆ln CEt=δ1+DCE +γ1ln CEt−1+γ2ln Ft−1+γ3ln It−1 +γ4(ln It−1)2+γ5ln FDt−1 + n ∑ i=1 λ1i∆ln CEt−i+ n ∑ i=0 λ2i∆ln Ft−i+ n ∑ i=0 λ3i∆ln It−i + n ∑ i=0 λ4i∆(ln It−i)2 + n ∑ i=0 λ5i∆ln FDt−i+ε1t(10) ∆ln Ft=δ2+DFOS +γ6ln CEt−1+γ7ln Ft−1+γ8ln It−1 +γ9(ln It−1)2+γ10 ln FDt−1 + n ∑ i=0 λ6i∆ln CEt−i+ n ∑ i=1 λ7i∆ln Ft−i+ n ∑ i=0 λ8i∆ln It−i + n ∑ i=0 λ9i∆(ln It−i)2 + n ∑ i=0 λ10i∆ln FDt−i+ε2t(11) ∆ln It=δ3+DIND +γ11 ln CEt−1+γ12 ln Ft−1+γ13 ln It−1 +γ14(ln It−1)2+γ15 ln FDt−1 + n ∑ i=0 λ11i∆ln CEt−i+ n ∑ i=0 λ12i∆ln Ft−i+ n ∑ i=1 λ13i∆ln It−i + n ∑ i=0 λ14i∆(ln It−i)2 + n ∑ i=0 λ15i∆ln FDt−i+ε3t(12) ∆(ln It)2=δ4+DINDSQ +γ16 ln CEt−1+γ17 ln Ft−1+γ18 ln It−1 +γ19(ln It−1)2+γ20 ln FDt−1 + n ∑ i=0 λ16i∆ln CEt−i+ n ∑ i=0 λ17i∆ln Ft−i+ n ∑ i=0 λ18i∆ln It−i + n ∑ i=1 λ19i∆(ln It−i)2 + n ∑ i=0 λ20i∆ln FDt−i+ε4t(13) ∆ln FDt=δ5+DFDT +γ21 ln CEt−1+γ22 ln Ft−1+γ23 ln It−1 +γ24(ln It−1)2+γ25 ln FDt−1 + n ∑ i=0 λ21i∆ln CEt−i+ n ∑ i=0 λ22i∆ln Ft−i+ + n ∑ i=0 λ23i∆ln It−i + n ∑ i=0 λ24i∆(ln It−i)2 E. Abokyi, P. Appiah-Konadu, F. Abokyi et al. / Energy Reports 5 (2019) 1339–1353 1345 + n ∑ i=0 λ25i∆ln FDt−i+ε5t(14) where the DCE ,DFOS ,DIND,DINDSQ and DFDT are dummy variables which capture structural breaks. δ0denotes the intercept and εt is the error term. The long-run and short-run parameters are γi and λirespectively whiles ∆is the change parameter. The study estimated equation (10) to investigate the longrun relationship among the variables. We employed the Ordinary Least Square (OLS) technique in the estimation after which the F-test for joint significance was computed. The null hypothesis, which implies no cointegration, is stated as γ1=γ2=γ3= γ4=γ5=0 whiles the alternative hypothesis implying the existence of cointegration is written as γ1= γ2= γ3= γ4= γ5= 0. The same procedure was followed to test for long-run cointegration among the variables in Eqs. (11)–(14). The study compared the value of the F-statistic computed with the bounds of Pesaran et al. (2001) critical values. If the F-statistic computed exceeds the upper critical value, the null hypothesis of no cointegration is rejected implying the existence of long-run relationship among the variables. On the other hand, we fail to reject the null hypothesis provided the computed F-statistic is lower than the lower critical bound which in this case suggests the variables are not cointegrated. We would have inconclusive result if the F-statistic is within the upper and lower critical values. Since evidence of cointegration is found to exist among the variables, the study proceeded to estimate both the long-run and short-run parameters. 2.4. Lind and Mehlum test for a U-shaped or inverted U-shaped relationship Since the ARDL approach provides only the necessary condition for the existence of a U-shaped or inverted U-shaped relationship but without the sufficient condition, the study proceeds to conduct the Lind and Mehlum (2010) U-shaped test. Lind and Mehlum (2010) built on the work of Sasabuchi (1980) to develop a test that overcomes the problem of misinterpreting the actual non-linear relationship existing between two variables in conventional non-linear econometric models. In these models, U-shaped or inverted U-shaped relationships are examined by looking at the different signs and the statistical significance of the original variable and its squared term. According to Lind and Mehlum (2010), this conventional method of testing for a nonlinear relationship is inappropriate. Lind–Mehlum test ensures that the necessary and sufficient conditions for the existence of a U-shape are provided on some interval values where the relationship decreases at the left side of the interval and increases at the right. To conduct this test, we first estimated the equation below by OLS: ln CEt=α0+DCE +α1ln Ft+α2ln It+α3(ln It)2 +α4ln FDt+ut(15) Next, the following condition was tested: α2+α3ln Imin <0< α2+α3ln Imax This condition examines whether industrial growth is declining at the left side of the interval values and rising at the right side of the interval values to ascertain the exact non-linear relationship (either U-shape or inverted U-shape) between carbon dioxide emission and industrial growth in Ghana. Begum et al. (2015) also employed the Lind and Mehlum (2010) U shaped test to confirm the existence of a U-shaped relationship GDP growth and CO2 emissions in Malaysia. Table 1 Descriptive statistics and correlation matrix. ln CEtln Ftln Itln FDt Mean −1.1658 −1.3705 21.6413 −1.3573 Median −1.1888 −1.5100 21.6591 −1.3190 Maximum −0.5889 −0.6422 23.6655 −0.9340 Minimum −1.5621 −2.1603 19.6369 −1.8089 Std. Dev. 0.2273 0.4082 0.9812 0.2332 Jarque–Bera 2.9600 3.3134 0.2299 1.8397 Probability 0.2276 0.1908 0.8914 0.3986 ln CEt1.0000 ln Ft0.8302 1.0000 ln It0.8327 0.8542 1.0000 ln FDt0.4224 0.5876 0.4560 1.0000 2.5. Granger-causality approach According to Engle and Granger (1987), the existence of cointegration between the variables implies that there exists causality in at least one direction. In order to ascertain the direction of causality between the variables, the proposed study applied the modified Granger-causality test. Engle and Granger (1987) notes that, if the variables are all I(1) and cointegrated, then specifying the Vector Error Correction Model (VECM) is appropriate. Therefore with evidence of cointegration in the I(1) series used in this study, we go further to investigate the direction of causality among the variables using the Error Correction Term (ECT) augmented Granger-causality test. This approach involves setting up a multivariate VECM of pth order as given below: (1 −L) ⎡ ⎢ ⎢ ⎢ ⎣ ln CEt ln Ft ln It (ln It)2 ln FDt ⎤ ⎥ ⎥ ⎥ ⎦ = ⎡ ⎢ ⎢ ⎢ ⎣ θ1 θ2 θ3 θ4 θ5 ⎤ ⎥ ⎥ ⎥ ⎦ + ⎡ ⎢ ⎢ ⎢ ⎣ DCE DFOS DIND DINDSQ DFDT ⎤ ⎥ ⎥ ⎥ ⎦ + p ∑ i=1 (1 −L) ⎡ ⎢ ⎢ ⎢ ⎣ ϖ11iϖ12iϖ13iϖ14iϖ15i ϖ21iϖ22iϖ23iϖ24iϖ25i ϖ31iϖ32iϖ33iϖ34iϖ35i ϖ41iϖ42iϖ43iϖ44iϖ45i ϖ51iϖ52iϖ53iϖ54iϖ55i ⎤ ⎥ ⎥ ⎥ ⎦ ⎡ ⎢ ⎢ ⎢ ⎣ ln CEt−i ln Ft−i ln It−i (ln It−i)2 ln FDt−i ⎤ ⎥ ⎥ ⎥ ⎦ + ⎡ ⎢ ⎢ ⎢ ⎣ ϕ1 0 0 0 0 ⎤ ⎥ ⎥ ⎥ ⎦ [ECTt−1] + ⎡ ⎢ ⎢ ⎢ ⎣ u1t u2t u3t u4t u5t ⎤ ⎥ ⎥ ⎥ ⎦ (16) where (1 −L) in the above VECM is the lag operator, the lagged error correction term is ECTt−1derived from the long-run relationship whiles u1t,u2t,u3t,u4tand u5tare the error terms which are expected to be serially uncorrelated with zero mean. The appropriate lag length pwas selected based on SIC information criteria. The specification of the VECM helps to ascertain causal relationships among the variables in the long-run as well as the short-run. The significance of the t-statistic associated with the term ECTt−1indicates causality in the long-run whiles the significance of the F-statistics associated with the lagged differences of the regressors indicate short-run Granger causality. 3. Results and discussion Information presented in Table 1 include descriptive statistics of all the variables as well as their correlation matrix. The Jarque–Bera statistics show that each of the variables is normally distributed with a mean of zero and a constant variance. The results of the correlation matrix indicate that the variables under study are positively related. Thus, Fossil fuel consumption, 1346 E. Abokyi, P. Appiah-Konadu, F. Abokyi et al. / Energy Reports 5 (2019) 1339–1353 Table 2 ADF and PP unit root tests. Variable ADF PP Level form 1st difference Level form 1st difference t-Statistic t-Statistic Adj. t-Stat t-Statistic ln CEt−0.0008(1) −9.5149(0)*−0.5065[2.96] −9.6387[1.02]* ln Ft0.6880(2) −9.0686(0)*−0.0128[3.13] −9.2627[1.35]* ln It0.2249(2) −4.3750(1)*0.1328[5.31] −3.4813[2.4]** (ln It)20.3408(2) −4.2936(1)*0.2975[5.19] −3.4936[2.3]** ln FDt−2.0135(0) −6.4657(0)*−2.1046[1.15] −6.4657[0.17]* The optimal lags for ADF( ) was selected using SIC. Whiles the band widths for the PP[ ] was selected using Andrews automatic. *Statistically significant at 1%. **Statistically significant at 5%. Table 3 Zivot-Andrews unit root test in the presence of structural breaks. Variable Level form Break year 1st diff Break year t-Stat t-Stat ln CEt−3.745(1) 1984 −6.430(2)*1999 ln Ft−3.311(2) 1983 −8.115(1)*1994 ln It−3.292(2) 1979 −6.675(1)*1984 (ln It)2−3.144(2) 1979 −6.529(1)*1984 ln FDt−3.565(0) 1997 −6.895(0)*2001 *Statistically significant at 1%. industrial growth and financial development have positive association with emissions of CO2. Industrial growth and financial development are also positively related with fossil fuel consumption. Finally, industrial growth and financial development are correlated positively. Unit root test Prior to testing for the existence of cointegration, we first tested for the order of integration of the variables to ensure none of the series used in the study is I(2). This is because the ARDL cointegration approach used in the study is invalid when I(2) time series variables are introduced. Moreover, the Bayer–Hanck cointegration technique also require all variables to be I(1). To investigate the unit root properties of the variables, the study employed the Augmented Dickey–Fuller (ADF) and the Phillip–Perron tests which are conventional approaches for unit root testing. However, as discussed earlier, these tests are inappropriate in the presence of structural breaks, therefore the Zivot and Andrews (2002) unit root test was employed to test for the stationarity properties of the series in the presence of structural breaks. Tables 2 and 3report the stationarity properties of the variables which suggest they are I(1) and therefore suitable for ARDL and the Bayer–Hanck cointegration approaches employed in the study. After determining the maximum lag length using Schwarz information criterion, the study proceeded to employ two cointegration techniques namely, the ARDL bound test and Bayer– Hanck cointegration to ensure robustness. In order to avoid weak cointegration relationship, the significance level was set at 5% for both cointegration procedures. The result of the Bayer–Hanck combined cointegration is reported in Table 4. The result indicates that the Fisher-statistics associated with EG-JOH and EG-JOH-BO-BDM are greater than the critical values at 5% level of significance only when emissions of CO2 is the dependent variable. Thus, using the other variables as the dependent variable failed to give evidence of cointegration suggesting the existence of only one cointegration relationship between the variables. For the ARDL bounds test approach, we tested for cointegration among the variables by computing the F-test to examine the joint significance of the lagged level time series variables. Before computing the F-test, the AIC information criterion was first used to select the optimal lag so as to avoid under fitting the model. From Table 5 where the result of the bound test is reported, we found the existence of cointegration only in one case, when carbon dioxide emissions is the dependent variable. Using the other variables as dependent variable produced no evidence of cointegration and thus confirming the existence of only one cointegration relationship between the variables. The break dates were selected by forming AR(1) process of the series in levels and applying (Bai and Perron,1998) sequential test to select structural break dates which are relatively the most significant. The selected break dates for emissions of CO2, fossil fuel consumption, industrial growth, the squared term of industrial growth and financial development are 1993, 1998, 1983, 1983 and 1996 respectively. The selected breakpoint of 1983 marks the Economic Recovery Program initiated by the government of Ghana to overcome the economic decline that confronted the country at the time. The other estimated break dates are close to 1992 and 1996 which are very important in the political and economic history of Ghana. Ghana took a decisive step to return to democratic rule in 1992 after a long period of military rule. According to Leite et al. (2000), the period 1992 saw significant changes in both monetary and exchange rate policies. Since both 1992 and 1996 were election years in Ghana, Leite et al. (2000) noted that huge cost was imposed on the economy of Ghana due to excess spending in an effort to capture election victory. After finding evidence of cointegration, the study analyzed the long-run impact of the other variables on emissions of CO2. The result, which is illustrated in Table 6, suggests fossil fuel consumption and financial development impact emissions of CO2 positively. The positive sign found on the impact of fossil fuel consumption on CO2 emissions is consistent with other studies on Ghana such as Kwakwa and Alhassan (2018) and Twerefou et al. (2016) though the latter used energy in general. The positive sign of fossil fuel is also in line with other studies (Begum et al., 2015;Ozturk and Acaravci,2010;Park and Lee,2011) which found positive impact of energy consumption on air pollution for Malaysia, Turkey and Republic of Korea respectfully. However, the coefficient is insignificant. Thus, given the sample period under study, there is no enough evidence to suggest that fossil consumption impact CO2 emissions positively in Ghana. This could be due to the different energy variable (fossil fuel) used in this study. The proposed study used fossil fuel consumption which is just a component of total energy consumption, contrary to Twerefou et al. (2016) where energy in general was used. The difference in control variables and time period used in the proposed study could also account for the insignificant impact fossil fuel consumption has on CO2 emissions. Similarly, the insignificant impact of financial development on CO2 emissions could also be attributed to the same reasons such as the difference in variables used to measure financial development, the control variables, the methodology and the difference in time periods used in the analysis. For instance, whiles the proposed study measured financial development as the ratio of domestic credit by financial sector to GDP, Ali et al. (2016) and Başarir and Çakir (2015) measured financial development as broad money supply and the time periods used in their studies are 1985–2012 and 1995–2010 respectively compared to 1971–2014 in the proposed study. The positive sign is consistent with Başarir and Çakir (2015) for Turkey and (Zhang, 2011) for China but contradicts Ali et al. (2016) and Jalil and Feridun (2011) which found financial development to reduce CO2 emissions for Malaysia and China respectively. The signs of the industrial growth and its squared term which are both significant at 1% give evidence of a U-shaped relationship between industrial growth and emissions of CO2 in Ghana. The E. Abokyi, P. Appiah-Konadu, F. Abokyi et al. / Energy Reports 5 (2019) 1339–1353 1353 Pesaran, M.H., Shin, Y., Smith, R.J., 2001. Bounds testing approaches to the analysis of level relationships. J. Appl. Econom. 16 (3), 289–326. Phillips, P.C., Ouliaris, S., 1990. Asymptotic properties of residual based tests for cointegration. Econometrica 165–193. Rafindadi, A.A., 2015. Econometric prediction on the effects of financial development and trade openness on the German energy consumption: A startling revelation from the data set. Int. J. Energy Econ. Policy 5 (1), 182–196. Saint Akadiri, S., Bekun, F.V., Sarkodie, S.A., 2019. Contemporaneous interaction between energy consumption, economic growth and environmental sustainability in South Africa: What drives what? Sci. Total Environ. 686, 468–475. Sarkodie, S.A., Strezov, V., 2018. Empirical study of the environmental Kuznets curve and environmental sustainability curve hypothesis for Australia, China, Ghana and USA. J. Clean. Prod. 201, 98–110. Sarkodie, S.A., Strezov, V., 2019. A review on Environmental Kuznets Curve hypothesis using bibliometric and meta-analysis. Sci. Total Environ.. Sasabuchi, S., 1980. A test of a multivariate normal mean with composite hypotheses determined by linear inequalities. Biometrika 67 (2), 429–439. Shafik, N., Bandyopadhyay, S., 1992. Economic Growth and Environmental Quality: Time-Series and Cross-Country Evidence (Vol. 904). World Bank Publications. Shahbaz, M., Bhattacharya, M., Ahmed, K., 2015. Growth-globalisation-emissions nexus: the role of population in Australia. Monash Business School Discussion Paper, (23/15), p1-33. Shahbaz, M., Khraief, N., Jemaa, M.M.B., 2015b. On the causal nexus of road transport CO2 emissions and macroeconomic variables in Tunisia: Evidence from combined cointegration tests. Renew. Sustain. Energy Rev. 51, 89–100. Shahbaz, M., Naeem, M., Ahad, M., Tahir, I., 2018. Is natural resource abundance a stimulus for financial development in the USA? Resour. Policy 55, 223–232. Shahbaz, M., Uddin, G.S., Rehman, I.U., Imran, K., 2014. Industrialization, electricity consumption and CO2 emissions in Bangladesh. Renew. Sustain. Energy Rev. 31, 575–586. Sims, C.A., 1980. Macroeconomics and reality. Econometrica 1–48. Stocker, T.F., Qin, D., Plattner, G.K., Tignor, M., Allen, S.K., Boschung, J. ..., Midgley, P.M., 2013. Climate change 2013: The physical science basis. Intergovernmental Panel on Climate Change, Working Group I Contribution to the IPCC Fifth Assessment Report (AR5)(Cambridge Univ Press, New York). Twerefou, D.K., Adusah-Poku, F., Bekoe, W., 2016. An empirical examination of the Environmental Kuznets Curve hypothesis for carbon dioxide emissions in Ghana: an ARDL approach. Environ. Soc.-Econ. Stud. 4 (4), 1–12. Twerefou, D.K., Appiah-Konadu, P., Asomanin, K., 2015. Impact of international trade on the quality of the natural environment: A case study of Ghana from 1970 to 2010. Ghana Soc. Sci. J. 12 (1), 1–21. United Nations. Department of Economic and Social Affairs. 2013. World economic and social survey 2013: Sustainable development challenges. UN. United Nations 2016. Identification of Emerging Issues for Sustainable Development. The Global Sustainable Development Report (GSDR), Chapter Five (5) UN, Newyork-USA. US EPA, 2016. Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990 – 2014, EPA 430-R-16-002. Word Bank 2017. World Development Indicators 2017. Washington, DC: World Bank. World Bank. 1992. World Development Report. New York: Oxford University Press. Yuxiang, K., Chen, Z., 2011. Financial development and environmental performance: evidence from China. Environ. Dev. Econ. 16 (1), 93–111. Zhang, Y.J., 2011. The impact of financial development on carbon emissions: An empirical analysis in China. Energy Policy 39 (4), 2197–2203. Zivot, E., Andrews, D.W.K., 2002. Further evidence on the great crash, the oilprice shock, and the unit-root hypothesis. J. Bus. Econom. Statist. 20 (1), 25–44.