A decomposition and decoupling analysis for carbon dioxide emissions: evidence from OECD countries

Abstract

Despite the huge difference in their climatic regimes, the OECD countries are among the world’s largest energy consumers and emitters of greenhouse gases, particularly carbon dioxide. Nonetheless, no studies have been conducted to decompose and decouple the long-term influential primary factors of carbon emissions for these countries. In this research, the Log Mean Divisia Method I is used to inspect the contribution of several influencing factors to fill this knowledge gap. Moreover, Tapio (Transp Policy 12(2):137–151, 2005) decomposition analysis (DA) is performed to investigate the driving forces of CO₂ emissions over the 1990–2019 years. The study provides an in-depth analysis of how to reduce CO₂ emissions and the factors that contribute to their variation, which is crucial for both global and regional climate change policies. DA shows that, up to 2004, the activity effect and the population effect drove the emissions to increase; while, in more recent years, the activity effect was able to curb the emissions. Decoupling analysis show the prevalence of the expansive negative decoupling regime for the 1990–2004 and 2015–2019 periods, while several countries were in the strong decoupling phase over the central period (2005–2009). According to the results, further efforts to increase energy efficiency, political support for digitalization and decentralized energy systems, and setting up a unique emission trading system are recommended for air pollution reduction.

1 Introduction and literature review

Greenhouse Gas (GHG) emissions into the air might provoke climate change, which has been identified as the world’s greatest environmental problem. GHG emissions such as carbon dioxide (CO₂), methane (CH₄), and nitrous oxide (N₂O) are mainly produced by using a high amount of fossil fuels. Among these pollutants, CO₂ emissions account for more than 76% of total GHG emissions, so that global CO₂ emissions increased from 20.64 billion tons in 1990 to 37.5 billion tons in 2022. Therefore, CO₂ emissions are the biggest share of the world’s GHG emissions as well as the main source of global warming. The impacts of CO₂ emissions on human lives and different sectors of society are considerable. Recently, CO₂ emissions have been a potential hazard and threat to public health, particularly respiratory disease, and have reduced life expectancy. It is understandable that CO₂ emissions have not only significantly damaged the environment, but also may have a detrimental role on public health, social welfare, and economic development. Hence, the need for alleviating climate change has prompted the world to pay considerable attention to CO₂ emissions reduction (Magazzino, 2016b, 2019; Mallongi et al., 2021, 2023; Mele et al., 2021; Padilla et al., 2021; Rauf et al., 2021; Song et al., 2018).

Emissions continue to increase in many countries although progress has been registered in decoupling GHG emissions from economic activity. Historically, the Organization for Economic Cooperation and Development (OECD) countries have emitted most of the global GHG; however, there was also a sharp increase in emissions due to Brazil, Russia, India, China, South Africa (BRICS) countries. OECD members were responsible for more than 67% of global CO₂ emissions in 1971, 50% in 1990, and 35% today. Figure 1 shows the historical trend of CO₂ emissions and total energy consumption of OECD members from 1960 up to 2019. According to the projections, CO₂ emissions increased from 5168 Mt in 1960 to 11,335.12 Mt in 1990, then reached their peak in 2007, around 13,294.7 Mt, and since then, international agreements for sustainable development, the Kyoto Protocol (1997), and the Paris Agreement (2015) gradually changed the energy mix among OECD members and have reduced the CO₂ emissions; however, these agreements progress is insufficient and could not stabilize emissions.

Fig. 1

CO₂ emissions and total energy consumption of OECD countries (1960–2019). Source: authors’ elaborations on World Bank data

During the same period, a similar trend for the total energy consumption of OECD members is registered. Total energy consumption doubled, and reached its top, with a value of 66,795.83 TWh in 2007; then suddenly fell to 62,830.5 TWh in 2009 due to the global financial crisis, and since then it has remained almost constant. As shown in Fig. 2, per capita Gross Domestic Product (GDP) exhibits an increasing trend over time, with exception of 1998 and 2009 years. It increased ten times from 263,256.7 LCU in 1960 to 2,639,692.3 LCU in 2019. Recently, emissions have fallen in almost all OECD countries, partly as a result of the reduction in GDP growth rates due first to the sub-prime mortgage and sovereign debt crises, then to the Corona Virus Disease (COVID-19) pandemic, and finally to the conflict between Russia and Ukraine, but also to more daring environmental policies launched especially by the European Union (EU) and the change in energy consumption models, also imposed by the ongoing conflict (IEA, 2015; Song et al., 2018; OECD, 2020; Bersalli et al., 2022; Magazzino & Mele, 2022).

Fig. 2

CO₂ emissions and per capita GDP of OECD countries (1960–2019). Source: authors’ elaborations on World Bank data

CO₂ emissions of OECD countries have been reduced in recent years, but the absolute amount of emissions is still high and unstable. To reduce CO₂ emissions avoiding climate change risks, international cooperation and regional policy-making are needed. CO₂ emissions are significantly dependent on economic activities and energy consumption, such that both variables move along with negative externality problems at a global and local scale. Emissions change over time due to variations in energy demand, economic structure, efficiency improvements, population density, investment in infrastructure, etc. In this regard, OECD members should reconsider their strategy for emissions abatement. To reach the goal, these countries should understand how to reduce emissions and which factors contribute to their dynamics. In addition, it is also important to evaluate the intensity of the relationship between economic growth and CO₂ emissions. Hence, this study aims to identify the driving factors behind CO₂ emissions, as well as which of them has been more significant on CO₂ emissions in the framework of decomposition; besides, decoupling analysis as a supplement emission assessment is used to analyze the conditions registered in each country (Alves & Moutinho, 2014; Tajudeen et al., 2018; Song et al., 2018; Yang et al., 2018; Wang et al., 2018; Wu et al., 2021). A useful way to realize the contribution of each effective factor to change in CO₂ emissions is the Decomposition Analysis (DA). DA provides a reliable measure of emissions changes and gives clarity regarding which factor has been more significant in driving the reduction/enhancement in CO₂ emissions. DA is one of the useful tools in order to investigate the driving factors in CO₂ emissions. DA method has been widely used within the field of energy policy to survey the decarbonization status among economic growth, energy consumption, and CO₂ emissions, measuring the impact of effective factors on CO₂ emissions. Moreover, this technique has been employed to inspect the key factors between carbon emissions and energy consumption (Chong et al., 2017; Pan et al., 2021), between energy and water (Li et al., 2019a), and in the industry sector (Kopidou & Diakoulaki, 2017; Yu et al., 2019). The Structural Decomposition Analysis (SDA), Index Decomposition Analysis (IDA), and Log Mean Divisia Method (LMDI) I and II are the main subgroups of the DA methods. LMDI-I method was proposed by Ang and Liu (2001) to conquer the aggregation and residual terms problems in both SDA and IDA methods. Hence, LMDI methods became the most appropriate and flexible within DA, and a considerable number of studies in energy economics used these methodologies: Zhang et al. (2011), Gonzalez et al. (2014), Cansino et al. (2015), Shahiduzzaman et al. (2015), Torrie et al. (2016), Dong et al. (2020), Lin and Long (2016), Mousavi et al. (2017), Zhao et al. (2017), Li et al. (2018), Wang and Zhou (2018), Cai and Ma (2018), Boqiang and Liu (2017), Song et al. (2018), Li et al. (2019b), Ran et al. (2019), Zhang et al. (2019a), Parker and Bhatti (2020), Ari et al. (2020), Cai et al. (2020), Hasan and Chongbo (2020), Jiang et al. (2020), Nieto et al. (2020), Eskander and Nitschke (2021), Karmellos et al. (2021), Padilla et al. (2021), Tenaw (2021), Zhang et al. (2021), Golas (2022), Gonzalez et al. (2022), Ozawa et al. (2002), Ruiz et al. (2022), Xu et al. (2022), and Wen et al. (2022). As a supplement to CO₂ emissions analysis, decoupling analysis helps regional policymakers to realize the level of green growth and adopt more focused interventions toward sustainable development. On the other hand, few studies have focused on the decoupling between CO₂ emissions and economic growth: Song et al. (2018) and Tenew et al. (2021). Hence, decomposition and decoupling analyses are applied to diverse fields and can reveal interesting outcomes.

In summary, the study aims to conduct DA to explore the driving factors of CO₂ emissions in OECD countries. Then, particular attention to decoupling analysis is paid, to understand the decoupling relationship between CO₂ emissions and economic growth during the years 1990–2019. In addition, an LMDI-I method based on the Kaya factor is developed to decompose and analyze the contributions of the main influencing factors. In seeking to fulfill the decoupling aim, Tapio’s (2005) DA is implemented to map the status and degree of relationship between CO₂ emissions and economic growth over six different sub-periods.

Many research (see Table 4 in the "Appendix") have been conducted on the DA to explore the driving factors behind CO₂ emissions. Some studies analyzed decompositions of both CO₂ emissions and energy intensity, trying to investigate the driving factors of the changes in these variables. As an example, Zhang et al. (2011) analyzed CO₂ emissions in the Chinese transportation sector from 1985 to 2009 using LMDI method. The results proved that per capita economic activity was the key contributor to CO₂ emissions growth. Gonzalez et al. (2014) assessed the contributors behind the change in energy consumption across European countries during the period 2001–2008. The findings highlighted that economic activity is the main driving factor of energy consumption. Cansino et al. (2015) tried to analyze the contribution of CO₂ emissions to Spain’s economy from 1995 to 2009. DA analysis indicated that renewable energy sources are the driving factors of CO₂ emissions. Taking an example from the chemical industry, Lin and Long (2016) stated that output per worker, industrial economic scale, energy intensity, and energy structure were the key driving forces in CO₂ in the Chinese chemical industry. Then, Boqiang and Liu (2017) studied China’s CO₂ emissions from heavy industry for 1991–2015 years. Labour productivity, energy intensity, and industry scale are the contributors to the increase in emissions. Zhao et al. (2017) explored the leading factors of changes in both national and regional CO₂ emissions within China’s provinces from 2000 to 2014. Decomposition results indicated that, at both national and regional scales, the economic activity factor is the main contributor to the increase in CO₂ emissions, and the energy intensity factor is the main key to emissions’ reduction. At the same time, Mousavi et al. (2017) used three variations of LMDI method to identify the driving forces of CO₂ emissions from 2003 to 2014, claiming that the intensity of electricity generation and fossil fuel combustion are two major responsible factors for CO₂ emissions. Li et al. (2017) applied DA method to explore the leading factors of the variations in both national and regional CO₂ emissions, showing that the economic scale factor represents the driving force.

Later, Song et al. (2018) under the framework of Kaya identity, used LMDI decomposition method to understand the effects of CO₂ determinants in OECD countries from 2001 to 2015. Results implied that energy intensity and per capita GDP are the main driving factors of CO₂ emissions. Besides, the decoupling state between the CO₂ emissions, population, energy consumption, and GDP is recessive decoupling. Wang and Zhou (2018) investigated global per capita consumption-based emissions inequality by using the Theil index and IDA analysis from 1995 to 2009, providing that production outsourcing is more responsible than consumption regarding emissions inequality. Wang and Feng (2018) applied LMDI technique to decompose the variation in Chinese industrial CO₂ emissions. They explained that industrial activity is the most significant factor in emissions. Also, when looking at the sectoral subject, Cai and Ma (2018) performed a study to mitigate CO₂ emissions in Chinese commercial buildings from 2001 to 2015. The results indicated that GDP per capita and industry intensity reduce emissions, while the energy intensity effect is the main factor of the increase in CO₂. Ran et al. (2019) applied LMDI method to evaluate CO₂ emissions from the electric power sector in China during 1998–2017. Economic growth is the major factor in emissions increase. Zhang et al. (2019b) studied Chinese provincial-level driving mechanism of CO₂ emissions for the power sector during 2004–2014. The results suggested that the change in CO₂ emissions of the power sector could be mainly attributed to the economic scale, industrial intensity, and also energy intensity within provinces. Li et al., (2019a, 2019b) investigated the leading factors of changes in emissions for the transportation sector within megacities during 1960–2001. Decomposition results introduce rapid urbanization and motorization factors as major driving factors of emissions.

Also, Hasan and Chongbo (2020) investigated the historical CO₂ emissions of the electricity industry in Bangladesh over the period 1979–2018, highlighting that government action, population intensity, and substitution exert a positive impact, while carbon and power intensity exhibit a negative role in the emissions. Jiang et al. (2020) applied LMDI technique across the Chinese non-residential power sector over the period 2007–2016. Empirical results showed that economic growth is the main influencing factor; on the contrary, population growth has an insignificant role in the growth of non-residential power consumption. Parker and Bhatti (2020) documented Asian CO₂ emissions across fourteen countries from 1971 to 2017. They claimed that the per capita indicator is the most important parameter, while carbonization and energy intensity are the least significant in determining the dynamics and fluctuation of emissions. Ari et al. (2020) studied the main contributors to Turkey’s CO₂ emissions for the transportation sector in the 2000–2017 years. LMDI method findings revealed that economic growth is the leading factor of the emissions in the transportation sector; however, population and emissions intensity factors have a positive effect. At the same time, Cai et al. (2020) studied China’s CO₂ emissions by decomposing and analyzing the driving factors. For this aim, LMDI method was applied to data from 1996 to 2016. Findings indicated that economic activity plays a key role in emissions. Nieto et al. (2020) studied the Indian energy transition to a lower-carbon economy using LMDI approach to analyze the contributors behind the CO₂ emissions from 1990 to 2016. The results of the study indicated that the economic growth of India is the leader in CO₂ emissions.

Recently, a few studies have addressed the review of decomposition and decoupling analysis. Tenaw (2021) provided a DA on energy intensity in Ethiopia for the period 1990–2017, revealing that efficiency is the main driving factor; besides, industrialization and Foreign Direct Investments (FDIs) stock exert a positive effect on energy intensity; while, economic growth, renewable energy, and industrial quality show a negative impact on energy intensity factor. For the EU, Karmellos et al. (2021) investigated seven driving factors of CO₂ emissions from electricity generation over the years 2000–2018. The analyzed factors were economic activity, population, electricity intensity, electricity trade, energy intensity, generation structure, and emissions factors. Eskander and Nitschke (2021) examined energy use and CO₂ emissions in UK universities. They noted that the emissions coefficient, intensity, and affluence are the major contributors to total emissions. At the same time, Padilla et al. (2021) found the key leaders of CO₂ emissions and energy intensity by applying LMDI method based on Kaya identity from 1971 to 2017 for Colombia. Population effect is discovered as the main driving force. Zhang et al. (2021) analyzed the influencing factors of CO₂ emissions for the industry sector in China during 2000–2019. The Generalized Divisia Index Method (GDIM) is used for DA. According to the results, the added value of the industry sector is the key contributor to the increase in emissions.

More recently, Wen et al. (2022) studied a GDIM model across the Chinese industrial sub-sector from 2000 to 2017 and claimed that the investment scale is the main driving force for the CO₂ emissions whilst carbon intensity of investment, energy intensity, and investment efficiency assisted in reducing emissions. Ruiz et al. (2022) analyzed and compared the driving factors of the CO₂ emissions for the six largest emitters (China, USA, EU, India, Russia, and Japan). The Kaya-LMDI analysis method and Granger causality technique were used to disentangle the relationship among variables over the period 1990–2018. Results proved that economic growth is the main driving factor. Also, energy intensity is a leading factor in reducing CO₂ emissions. Gonzalez et al. (2022), by using the LMDI method, tried to track the change in Spanish GHG emissions over the period 2008–2018. The results of the DA indicated that energy intensity played a crucial role. Xu et al. (2022) analyzed the decomposition of residential electricity-related CO₂ emissions in China’s provinces over the period 1997–2019. LMDI findings showed that income growth is the main factor behind CO₂ emissions in most provinces.

The current research decomposes the CO₂ emissions into main factors by using LMDI-I method to explore the driving factors of the emissions. Besides, Tapio’s (2005) decoupling analysis is applied to uncover the decoupling degree between CO₂ and economic growth. Compared with previous literature, to the best of our knowledge, no further research—except Song et al. (2018)—performed decomposition and decoupling analysis for OECD countries. In addition, the gap in the literature studies is in different ways; firstly, this study presents a long-term (1990–2019) analysis of decomposition and decoupling. Secondly, the LMDI-I method is improved in the study for OECD members. Thirdly, this study investigates decoupling status to investigate the relationship between economic growth and carbon emissions for OECD members for 6 sub-periods, a topic never addressed in previous research. Therefore, the study is the first in-depth research in the field of decomposition and decoupling analysis for all OECD countries.

The structure of the study is as follows. After the introduction and literature review provided in this Sect. 1, in the next Sect. 2 the data and methodology of decomposition and decoupling analysis are described. In the following Sect. 3, the empirical findings along with a discussion of the results are given. Finally, Sect. 4 contains the main conclusions together with policy implications.

2 Materials and methods

The empirical approach followed in this study uses LMDI-I decomposition method to disentangle the change in CO₂ emissions into a set of possible driving factors and add quantitative analysis to perceive the changes in predefined macroeconomics and energy-related factors. The index illustrates the effects of human activities on the environment, according to the Kaya identity, which was introduced by Kaya (1989), and then re-formalized by Zhang and Ang (2001). It is the most widely used and important analytical technique to detect the driving forces of CO₂ emissions from fossil fuels because of its simple structure and conceptualization. For this purpose, this paper uses LMDI method to identify the factors influencing CO₂ emissions due to its benefits in precisely describing outcomes. Population, GDP, energy intensity, fuel mix, and emission coefficients are the factors to consider. The impact of these variables on emissions has been extensively studied for various countries through different methodologies. As a result, the issue is not whether the factors impact the emissions or not, but rather how much influence they have. After assessing the impact of these factors on CO₂ emissions, the extent of emission intensity and the decoupling of OECD economies from CO₂ emissions are considered to examine future trends. Several studies (Boqiang & Liu, 2017; Cai & Ma, 2018; Cai et al., 2020; Jiang et al., 2020; Hasan & Chongbo, 2020; Nieto et al., 2020; Eskander & Nitschke, 2021; Karmellos et al., 2021) and the Fourth Assessment Report of the IPCC used Kaya based LMDI decomposition method to identify the possible driving factors of CO₂ emissions. Therefore, we applied accordingly LMDI-I method to confirm the contribution of several influencing factors and detect the key factors of CO₂ emissions for OECD countries. The study also aims to provide information for policymakers to point the energy policies in the right direction.

2.1 Decomposition analysis

In this study DA is developed on the Kaya identity, in which CO₂ emissions can be decomposed into a set of possible influencing factors as follows:

$$CO_{{2_{i,t} }} = \frac{{CO_{{2_{i,t} }} }}{{FE_{i,t} }} \times \frac{{FE_{i,t} }}{{TE_{i,t} }} \times \frac{{TE_{i,t} }}{{GDP_{i,t} }} \times \frac{{GDP_{i,t} }}{{POP_{i,t} }} \times POP_{i,t}$$

(1)

where CO₂ denotes carbon emissions, FE denotes fossil energy consumption, TE denotes total energy consumption, GDP denotes Gross Domestic Product, POP denotes population, i represents the specific individual (country), and t is the time identifier. Equation (1) can be re-written as Eq. (2) (Cai & Ma, 2018; Hasan & Chongbo, 2020):

$$C_{it} = CI_{i,t} \times ES_{i,t} \times EI_{i,t} \times AE_{i,t} \times P_{i,t}$$

(2)

where C_it is carbon emissions, CI_it is carbon intensity (CO₂/FE), which implies the amount of emissions emitted per unit of fossil energy consumption; ES_it is energy structure (FE/TE), which means the substitutions of fossil energy consumption; EI_it is energy intensity (TE/GDP), which shows the total energy consumption per unit of GDP; AE_it is the activity effect (GDP/POP), which represents Gross Domestic Product per capita, and P_it is the population (Boqiang & Liu, 2017; Karmellos et al., 2021).

Without residual factors through LMDI method, the arithmetic and cumulative change in carbon emissions for a specific time period can be distributed into the five possible influencing factors as in the following equations:

$$\Delta C = C_{{t_{2} }} - C_{{t_{1} }}$$

(3)

$$\Delta C = \Delta CI + \Delta ES + \Delta EI + \Delta AE + \Delta P$$

(4)

ΔC is the change in CO₂ emissions, ${C}_{{t}_{1}}$ refers to CO₂ emissions at time $t_{1}$, and ${C}_{{t}_{2}}$ represents CO₂ emissions at time $t_{2}$. ΔCI, ΔES, ΔEI, ΔAE, and ΔP are changes in carbon intensity, energy structure, energy intensity, activity effect, and population, respectively.

The individual effect of each component of Eq. (4) can be calculated as Eqs. (5)–(9).

$$\Delta CI = L(C_{{t_{2} }} - C_{{t_{1} }} ) \times \,Ln\,\left( {\frac{{CI_{{t_{2} }} }}{{CI_{{t_{1} }} }}} \right) \Rightarrow \,\left( {\frac{{C_{{t_{2} }} - C_{{t_{1} }} }}{{Ln\left( {\frac{{C_{{t_{2} }} }}{{C_{{t_{1} }} }}} \right)}}} \right) \times Ln\,\left( {\frac{{CI_{{t_{2} }} }}{{CI_{{t_{1} }} }}} \right)$$

(5)

$$\Delta ES = L(C_{{t_{2} }} - C_{{t_{1} }} ) \times \,Ln\,\left( {\frac{{ES_{{t_{2} }} }}{{ES_{{t_{1} }} }}} \right) \Rightarrow \,\left( {\frac{{C_{{t_{2} }} - C_{{t_{1} }} }}{{Ln\left( {\frac{{C_{{t_{2} }} }}{{C_{{t_{1} }} }}} \right)}}} \right) \times Ln\,\left( {\frac{{ES_{{t_{2} }} }}{{ES_{{t_{1} }} }}} \right)$$

(6)

$$\Delta EI = L(C_{{t_{2} }} - C_{{t_{1} }} ) \times \,Ln\,\left( {\frac{{EI_{{t_{2} }} }}{{EI_{{t_{1} }} }}} \right) \Rightarrow \,\left( {\frac{{C_{{t_{2} }} - C_{{t_{1} }} }}{{Ln\left( {\frac{{C_{{t_{2} }} }}{{C_{{t_{1} }} }}} \right)}}} \right) \times Ln\,\left( {\frac{{EI_{{t_{2} }} }}{{EI_{{t_{1} }} }}} \right)$$

(7)

$$\Delta AE = L(C_{{t_{2} }} - C_{{t_{1} }} ) \times \,Ln\,\left( {\frac{{AE_{{t_{2} }} }}{{AE_{{t_{1} }} }}} \right) \Rightarrow \,\left( {\frac{{C_{{t_{2} }} - C_{{t_{1} }} }}{{Ln\left( {\frac{{C_{{t_{2} }} }}{{C_{{t_{1} }} }}} \right)}}} \right) \times Ln\,\left( {\frac{{AE_{{t_{2} }} }}{{AE_{{t_{1} }} }}} \right)$$

(8)

$$\Delta P = L(C_{{t_{2} }} - C_{{t_{1} }} ) \times \,Ln\,\left( {\frac{{P_{{t_{2} }} }}{{P_{{t_{1} }} }}} \right) \Rightarrow \,\left( {\frac{{C_{{t_{2} }} - C_{{t_{1} }} }}{{Ln\left( {\frac{{C_{{t_{2} }} }}{{C_{{t_{1} }} }}} \right)}}} \right) \times Ln\,\left( {\frac{{P_{{t_{2} }} }}{{P_{{t_{1} }} }}} \right)$$

(9)

where the term L(${C}_{{t}_{2}}$-${C}_{{t}_{1}}$) is the logarithmic weight average (Boqiang & Liu, 2017; Hasan & Chongbo, 2020; Karmellos et al., 2021).

2.2 Decoupling analysis

Tapio (2005) proposed the decoupling model to evaluate the state of transition toward a low-carbon economy. According to the model, there are eight states, which show a process of decoupling between economic growth and carbon emissions (Table 1). The ideal state is to achieve negative carbon emissions along with economic growth. The Decoupling Index (DI) can be presented as in Eqs. (10) and (11).

$$DI_{t} = \left( {\frac{{\frac{\Delta C}{{C_{0} }}}}{{\frac{\Delta GDP}{{GDP_{0} }}}}} \right) \Rightarrow \,(\frac{{\left( {\frac{\Delta CI}{{CI_{0} }}} \right) + \left( {\frac{\Delta ES}{{ES_{0} }}} \right) + \left( {\frac{\Delta EI}{{EI_{0} }}} \right) + \left( {\frac{\Delta AE}{{AE_{0} }}} \right) + \left( {\frac{\Delta P}{{AP_{0} }}} \right)}}{{\left( {\frac{\Delta GDP}{{GDP_{0} }}} \right)}}$$

(10)

$$DI_{t} = DI_{CI} + DI_{ES} + DI_{EI} + DI_{AE} + DI_{P}$$

(11)

where DI_t represents the efforts to improve the environment in time t, DI_CI the efforts to optimize fossil energy consumption, DI_ES the efforts to change energy structure, DI_EI the efforts to optimize energy intensity, DI_AE the efforts to improve output per capita, and DI_P the efforts to optimize population scale (Boqiang & Liu, 2017; Karmellos et al., 2021; Tapio, 2005).

Table 1 States of decoupling model analysis

2.3 Data

In this study, time-series data over the period 1990–2019 for the OECD countries have been collected. CO₂ emissions, fossil energy consumption, total energy consumption, GDP, and population were used in the empirical analysis. All data are derived from World Bank (WB) and Our World in Data databases.¹ It is worth noting that the GDP data are on constant price. The basic descriptive statistics on these series are given in Table 2.

Table 2 Description of the study variables

3 Results and discussion

3.1 Results of the decomposition analysis

Following the LMDI-I approach from Eqs. (3–11), the findings of the decomposition analysis for OECD countries are presented in Figs. 1, 2, 3, 4, 5, 6, 7, 8 and 9 for the period 1990–1994, 1995–1199, 2000–2004, 2005–2009, 2010–2014, 2015–2019, and 1990–2019. In the following figures, the contribution of each possible driving factor is shown for all sub-periods. It is worth noting that the USA is excluded from the figures because of its high variations in the results and depicted in an individual figure.

Following the decomposition results in the OECD countries for the period 1990–1994 in Fig. 3, it is evident that Germany (with a total effect of − 79.50 Mt), Czech Republic (− 25.49), and the UK (− 21.30) were the pioneer countries of CO₂ reduction. On the contrary, South Korea (103.98), Japan (70.38), and Mexico (58.28) were responsible for the greater increase in CO₂ emissions. The leading factors that conduct CO₂ emissions to increase were the activity effect (especially in South Korea and Germany), the energy intensity effect (for Germany and Japan), and the population effect (for Germany and Mexico). It seems that the energy intensity effect (ΔEI), had a significant negative impact in most countries (total effect: − 236.01 Mt). Due to the impact of aged technology on production, the activity effect registered high positive scores in South Korea (82.65) and Japan (37.28). Notwithstanding, the energy intensity effect and energy structure had a remarkable contribution to CO₂ reductions in the majority of the countries. Using less energy to produce, CO₂ emissions declined in Germany, Poland, the UK, Italy, and France. Sustainable energy substitution policies in Canada, France, the UK, Japan, and Italy diminished CO₂ emissions. Our findings were mainly in line with Parker and Bhatti (2020), Gonzalez et al. (2014), and Hasan and Chongbo (2020) analysis. According to Parker and Bhatti (2020), the economic activity and in the next step, energy intensity, were the most drivers in explaining of CO₂ emissions in both South Korea and Japan. Gonzalez et al. (2014) found that the activity effect in large economics as Germany was the main factor of CO₂ emissions. Also, Hasan and Chongbo (2020) believed that the population effect was one of the main leaders of CO₂ emissions around the world.

Fig. 3

Decomposition of CO₂ emissions for OECD countries (1990–1994). Source: authors’ elaborations on World Bank data

In Fig. 4, the results of the DA and the driving forces are presented for the period 1995–1999. It can be observed that Mexico, Canada, Australia, and Turkey increased their CO₂ emissions by 53.95, 45.69, 44.03, and 24.16 Mt, respectively. The most significant reduction in emissions can be seen in Germany, while the rest of the countries—with some exceptions—raised their emissions. The main driving factor relating to the CO₂ emissions was the activity effect (total effect: 619.60 Mt). In this respect, all the countries in the sample registered an increase in emissions, with some remarkable cases (Poland, the UK, and South Korea). Also, the energy intensity had the most pronounced impact in Poland (− 76.70 Mt). The second driving force of the CO₂ emissions increase was the population: Mexico, Canada, Australia, and South Korea registered an increase in population. It is evident that the carbon intensity effect drove CO₂ emissions down in most of the countries, especially South Korea (− 48.60) and Poland (− 22.92). In addition, the energy structure effect had a slightly negative contribution in several countries. In comparison with Fig. 1, the activity effect was more pronounced for the period 1994–1999, while the energy intensity effect was smoothed through technological improvement in Colombia, Denmark, Finland, France, Israel, and Portugal. Regarding the carbon intensity effect, most of the countries had a significant contribution to emissions reduction, especially South Korea. The effects of activity and energy intensity factors on CO₂ emissions are in line with the results of Gonzalez et al. (2014) for Germany, Torrie et al. (2016) for Canada, Parker and Bhatti (2020) for South Korea, Eskandader and Nitschke (2021) and Karmellos et al. (2021) for the UK, Golas (2022) for Poland, and Ozawa (2002) for Mexico.

Fig. 4

Decomposition of CO₂ emissions for OECD countries (1995–1999). Source: authors’ elaborations on World Bank data

Fig. 5

Decomposition of CO₂ emissions for OECD countries (2000–2004). Source: authors’ elaborations on World Bank data

Regarding the decomposition and driving forces results for the period 2000–2004, it can be seen that most countries further enhanced their emissions: above all, Spain, Italy, Mexico, South Korea, Japan, Australia, and Canada registered a significant contribution. In addition, in several countries the economic activity effect and the population effect had a notable contribution to emissions. The biggest enhancement in CO₂ emissions due to the activity effect was in South Korea (with 85.03 Mt), while for the rest of the sample this effect was limited between 0.24 (Iceland) and 43.83 Mt (Canada). On the other hand, the population raised almost everywhere, notably in Mexico (22.04 Mt). In relation to the energy structure (total effect: 25.93 Mt), it exerted a relatively small contribution, with the biggest negative value in Germany (− 7.95), Czech Republic (− 7.76), and Turkey (− 7.22 Mt) due to substitution policies. Two main factors, energy intensity and carbon intensity, had been driving CO₂ emissions to decrease. The energy intensity effect had a significant negative impact in most countries (especially in Japan, Canada, the UK, and South Korea) due to technological development, and the biggest positive value in Mexico (12.82 Mt). Regarding the carbon intensity effect, it is easily evident that it had a small negative contribution (total effect: − 22.10 Mt), and CO₂ emissions relatively decreased along with fuel consumption. Moreover, it is worth noticing that CO₂ emissions increased due to development in activity structure and an increase in population more than in the previous periods; while, reduction factors showed a relatively small impact than before. Several previous studies—i.e., Shahiduzzaman et al. (2015), Torrie et al. (2016), Parker and Bhatti (2020), Eskandader and Nitschke (2021), Karmellos et al. (2021), Golas (2022), and Gonzalez et al. (2022)—stated that the energy intensity, activity effect, and population factor were the most driving factors of CO₂ emissions in different countries (South Korea, Japan, Poland, Canada, Australia, and the UK), which are consistent with our results. However, we found a different driving factor for emissions in Spain, which is in contrast with the analysis of Gonzalez et al. (2022).

The results of DA and the contribution of driving forces for the period 2005–2009 are shown in Fig. 6. It is clear that most countries, due to a global financial crisis (although with some exceptions), reduced their CO₂ emissions, so that the whole effect was equal to − 386.99 Mt. Considerable contractions in CO₂ emissions were registered for Japan (− 112.61), the UK (− 75.48), Italy (− 76.77), Germany (− 67.51), and Spain (− 63.01 Mt). In most countries, the main driving factor contributing to the enhancement of CO₂ emissions was the population effect, with a total effect of 177.99 Mt. The remarkable increases in CO₂ emissions due to the demographic factor can be found in Mexico, Australia, and Canada. On the other hand, the energy intensity effect (with a total effect of − 572.59 Mt) sensibly drove emissions’ reductions, especially in Japan (− 88.78), Germany (− 70.50), Canada (− 67.09), the UK (− 57.03), and Poland (− 51.41 Mt). Also, the economic structure effect exerted a negative impact on CO₂ emissions. In South Korea, Turkey, Colombia, and Mexico the economic structure effect had a contribution to the rise of CO₂ emissions. Several factors had been forcing CO₂ emissions to decrease. In all countries—with the exception of Turkey, Mexico, and Iceland—the energy intensity effect led CO₂ emissions to decrease, and the biggest decrease in CO₂ emissions for Japan. On the other hand, the carbon intensity effect (total effect: − 23.31 Mt) provoked a decrease in CO₂ emissions in several countries, especially in Spain (− 19.88), the UK (− 14.65), and Italy (− 12.39 Mt). The results for this sub-period, and particularly the impacts of activity effect, energy intensity, and population effects on CO₂ emissions, are largely comparable with those from the analysis in Gonzalez et al. (2014), Torrie et al. (2016), Hasan and Chongbo (2020), Parker and Bhatti (2020), Eskandader and Nitschke (2021), Karmellos et al. (2021), Golas (2022), and Gonzalez et al. (2022). Notwithstanding, a different driving factor—namely, the energy intensity—has been isolated here, in contrast with Ari et al. (2020) and Karmellos et al. (2021) findings for the case of Turkey and Italy, respectively.

Fig. 6

Decomposition of CO₂ emissions for OECD countries (2005–2009). Source: authors’ elaborations on World Bank data

The DA and relative contribution of the driving factors in CO₂ emissions from 2010 to 2014 for OECD countries are given in Fig. 7. According to the results, almost all the countries reduced their CO₂ emissions. A significant reduction in CO₂ emissions is found for Italy (− 77.78) and the UK (− 66.94 Mt), with the rest of the countries’ reduction was limited between − 43.41 (France) and − 1.06 Mt (Estonia). The highlighted factor in most countries which led CO₂ emissions to increase was the economic activity effect (total effect: 380.82 Mt), remarkably in Turkey (69.10), Germany (59.39), South Korea (57.33), and Japan (50.98 Mt). The second leading factor of CO₂ emissions was the population effect, which was positive in different countries, notably Mexico (24.77), Australia (24.03), Canada (22.63), and Turkey (20.88 Mt). On the other hand, several driving factors led CO₂ emissions to decrease, in which the energy intensity effect (total effect: − 751.70 Mt) was the most effective one in lots of countries, with the exceptions of Portugal (0.95), Greece (0.37), and Norway (0.29 Mt). Over these years, a remarkable reduction in emissions due to this specific effect was found in Japan with a value of − 154.94 Mt. Moreover, in Mexico, Australia, and South Korea, the carbon intensity effect had a significant negative contribution of − 30.58, − 29.66, and − 25.66 Mt, respectively. The energy structure effect exerted a marginally negative impact almost everywhere, except for Japan (160.69 Mt). In comparison with previous literature on this topic, the results by Gonzalez et al. (2014) for Germany, Ari et al. (2020) for Turkey, and Parker and Bhatti (2020) for South Korea are in the line with those presented in this paper. However, we established a different driving factor (the energy intensity), which is different with respect of Parker and Bhatti (2020) and Karmellos et al. (2021) for Japan and the UK, respectively.

Fig. 7

Decomposition of CO2 emissions for OECD countries (2010–2014). Source: authors’ elaborations on World Bank data

In Fig. 8, we report the results of DA and contributions of each driving factor in CO₂ emissions in OECD countries for the period 2015–2019. The results reveal that the emissions decreased in the sample. However, some countries registered an increase, like Turkey (72.82) and South Korea (26.27 Mt). The emissions were essentially driven by the energy intensity and the economic activity effects. The main driving factor contributing to the reduction of emissions in several countries was the energy intensity (total effect: − 446.69 Mt), with a considerable negative impact in Germany (− 65.25) and Japan (− 55.71 Mt). On the other hand, the activity effect drove the CO₂ emissions to rise, especially in South Korea (60.40) and Poland (52.33 Mt). Regarding the impact of the energy structure effect (total effect: -150.43 Mt), it was evident that due to successful substitution and shift in the fuel mix, it had a negative contribution with significant values in Japan (− 62.62), Germany (− 24.73), and Turkey (− 23.88 Mt). The carbon intensity had, generally speaking, a smaller negative effect, with a positive peak in Turkey (48.05) and a negative one in Italy (− 36.46 Mt). These results, above all the impact of the activity effect and energy intensity on CO₂ emissions, are close to ones provided by Torrie et al. (2016), Ari et al. (2020), Parker and Bhatti (2020), Karmellos et al. (2021), and Golas (2022).

Fig. 8

Decomposition of CO₂ emissions for OECD countries (2015–2019). Source: authors’ elaborations on World Bank data

The DA and the relative contribution of the driving factors in CO₂ emissions in all the sub-periods for the USA are depicted in Fig. 9. According to the decomposition results, it is clearly evident that the USA increased its CO₂ emissions for the periods 1990–1994 and 1995–1999. Then, due to the global financial crises, the emissions sharply decreased by approximately -596.53 Mt. For the rest of the period, a relatively small reduction can be seen. Moreover, several driving factors drove emissions to change. The population effect with a value of 261.20 Mt was the main driving force that increases emissions in the USA for the period 1990–94 when the energy intensity effect led emissions to decrease by approximately − 236 Mt. The main driving factor contributing to the enhancement of emissions in the period 1995–1999 was the economic activity effect; however, the energy intensity effect drove emissions to decrease. For the period 2000–2004, it is obvious that the economic activity and population effect had a positive effect on emissions. In addition, most of the contributing factors excluding the population effect led to an emissions decrease for the period 2005–2009. The energy intensity and energy structure effects had a considerable impact. The economic activity along with population growth showed a significant contribution with positive values of 272.64 and 151.44 Mt for the period 2010–2014. At the same time, energy intensity and energy structure effect due to green energy policies and gradually shift to renewable fuel mix had a remarkable contribution to emissions reduction, by approximately − 419.5 and − 60.34 Mt. These findings are in the line with Dong et al. (2020) results for the decomposition of the US CO₂ emissions from 1997 to 2017. Regarding the last period, it is clear that the impacts of the economic activity and energy intensity effects found significant. The energy intensity effect, unlike the economic activity effect, drove emissions to decrease with a value of − 313.7 Mt. The carbon intensity and energy structure effects led emissions to reduce by approximately − 236.7 and − 81.32 Mt, respectively.

Fig. 9

Decomposition of CO₂ emissions for the USA. Source: authors’ elaborations on World Bank data

Finally, the results of the DA and the contribution of each driving factor in emissions for OECD countries for the period 1990–2019 are presented in Fig. 10. Based on the results, it is evident that the emissions increased over this time period (with a global effect of 529.58 Mt). The most significant increase was in South Korea with a value of 383.53 Mt, while Germany (− 237.78), the UK (− 217.88), and Italy (− 103.79 Mt) led emissions to decrease. The economic activity effect played a great role in the increase of emissions in all countries (total effect: 3160.10 Mt), with peaks in South Korea (501.60), Poland (350.11), and Germany (319.77 Mt). The population effect was the other contributor factor that ran emissions to increase, with the highest contribution in Mexico (150.70) and Canada (150.35 Mt). On the contrary, the energy intensity effect reduced the emissions, especially in Germany (− 469.80), Poland (− 354.85), the UK (− 312.03), and Japan (− 292.29 Mt). The energy structure effect played a negative impact in most countries, with a few exceptions. The carbon intensity effect revealed a relatively small negative impact in most countries.

Fig. 10

Decomposition of CO₂ emissions for OECD countries (1990–2019). Source: authors’ elaborations on World Bank data

Furthermore, the results of the DA and the contribution of each driving force in emissions in the USA for the period 1990–2019 are presented in Fig. 11. It is evident that emissions decreased for the period, and the most reduction factor was the energy intensity effect with the value of − 2678.97 Mt, followed by carbon intensity and energy, respectively. Moreover, it seems that the economic activity and population effects had a significant positive contribution. The main driving factor contributing to the growth in emissions was the economic activity effect of 2108.92 Mt, and the population effect drove the emissions to increase by approximately 1321.07 Mt.

Fig. 11

Decomposition of CO₂ emissions for USA (1990–2019). Source: authors’ elaborations on World Bank data

Overall, the DA helps us to depict some general conclusions on how CO₂ emissions are affected. According to the results, the economic activity effect was the significant influencing force in emissions. The effect was most considerable in the USA and South Korea during the years of the analysis. The production structure of all countries revealed significant changes in energy consumption, notably a decrease in fossil energy consumption in the USA, Germany, Poland, the UK, and Canada. Using more efficient generation technologies for the improvement of energy intensity was clear in all countries, specifically in the 2005–2009 period due to the global financial crisis. The ineffectiveness of energy policies for a sustainable fuel mix and CO₂ mitigation were evident, specifically in European countries. The growth of population increased energy demand in countries where a significant population effect was registered, specifically in the USA, Canada, Mexico, and Australia, although there were some exceptions with negative growth rates, such as Estonia and Poland. As expected, the CO₂ intensity effect had a negligible impact in most countries; however, it was an important factor in the USA.

3.2 Decoupling analysis

We calculated the trend of the decoupling index based on Eqs. (10) and (11). Table 3 presents the results of the decoupling analysis, the decoupling degree of each country for CO₂ emissions and economic growth during various time periods.

Table 3 The decoupling state of each country for all study periods

According to the results, it is evident that most countries were in an expansive negative decoupling situation during the period 1990–1994 when they increased their CO₂ emissions at a higher rate than the growth of the economy. Specifically, Germany, Hungary, Italy, and Poland were in a state of strong decoupling, which means that they decreased CO₂ emissions along with an increase in economic growth. Czech Republic was the only country in a state of recessive decoupling, where both CO₂ emissions and economic growth decreased, which is called “green de-growth”. At the same time, the UK experienced a weak decoupling situation, when economic growth increased and CO₂ emissions increased at a rate between 0 and 0.8. It seems that both Finland and Sweden experienced the worst possible condition as they were in strong negative decoupling states; indeed, both of them increased their CO₂ emissions while economic growth went down.

For the period 1995–1999, most countries were in a state of expansive negative decoupling. Colombia, Czech Republic, Estonia, Latvia, and Lithuania were in a state of strong decoupling, meaning that they were successful in fulfilling green policies in that period. Denmark, Germany, Poland, and Slovakia were in a weak decoupling state. For Hungary and Luxembourg, it is evident that both CO₂ emissions and economic growth increased approximately at the same rate, which resembled an expansive coupling state.

The results of the decoupling analysis for the period 2000–2004 suggest that the majority of the countries were driven to an expansive negative decoupling state. However, Germany and Slovakia were in a state of weak decoupling during that period.

Based on the results of the decoupling analysis for the period 2005–2009, it is clear that the global financial crisis shifted the majority of countries to a strong decoupling situation. It means that economic growth increased, but CO₂ emissions decreased, which might be considered a case of “green policies”. Australia, Chile, Colombia, Costa Rica, Iceland, Israel, Japan, South Korea, Mexico, Switzerland, and Turkey were in a state of expansive negative decoupling, which means that economic growth increased and CO₂ emissions increased at different rates of more than 1.2. Denmark, Estonia, Hungary, and Italy were in a state of recessive decoupling or green de-growth situation. On the other hand, Slovenia and Denmark were driven to a state of weak decoupling during the global financial crisis period. Poland was the only country in a state of expansive coupling.

After the years of the global financial crisis, specifically for the period 2010–2014, it is clear that Australia, Canada, Chile, Colombia, Costa Rica, Iceland, South Korea, Mexico, New Zealand, Portugal, Slovenia, Spain, and Turkey were in a state of expansive negative decoupling. Most of the countries (16) were led to a case of green policies, meaning that they managed to fulfill green energy policies in that period. Israel, Japan, Poland, and Estonia were in a state of expansive coupling. It seems that Finland, Greece, and Italy were in a green de-growth situation due to the consequences of the global financial crisis. In addition, the USA was the only country that experienced a state of weak decoupling.

For the last period, between 2015 and 2019, the results of the decoupling analysis present a significant shift towards expansive negative decoupling for the majority of countries. They increased their CO₂ emissions at a higher rate than the growth of the economy. Meanwhile, Denmark, Finland, Germany, Greece, Italy, Japan, Norway, Switzerland, and the UK were led to a state of strong decoupling or green policies. The Czech Republic, Portugal, and Spain were the countries in a state of expansive coupling. At the same time, Estonia, France, the Netherlands, and Sweden registered a weak decoupling.

Finally, the results of the decoupling analysis for the period 1990–2019 show that most countries (21) were in a state of expansive negative decoupling. At the same time, Denmark, Germany, and Italy were in a state of strong decoupling. The countries in a state of weak decoupling were including the Czech Republic, Finland, France, Greece, Hungary, Latvia, Lithuania, Slovakia, Sweden, and the UK. While Estonia, Poland, Slovenia, and Switzerland increased both economic growth and CO₂ emissions at the same rate.

Overall, the considerable point was that the implementation of green energy policies has shifted most of the European countries towards a significant change in the state of decoupling between CO₂ emissions and economic growth. Meanwhile, South Korea, Australia, Chile, Costa Rica, Iceland, Mexico, and Turkey were always in a state of expansive negative decoupling, which is the worst scenario for decoupling. Besides, Austria, Belgium, Canada, Ireland, New Zealand, Slovenia, Luxembourg, the USA, Colombia, Israel, Portugal, Spain, and Switzerland were mostly observed in a state of expansive negative decoupling, with the exceptions of the global financial crisis and the following years. Finally, Estonia, Hungary, and Poland were countries with unstable states of decoupling.

4 Conclusions and policy implications

The study aims to analyze the main leading factors affecting CO₂ emissions for OECD member countries over the period 1990–2019. Through a DA and a decoupling analysis the whole period as well as six sub-periods (1990–1994, 1995–1999, 2000–2004, 2005–2009, 2010–2014, 2015–2019) have been investigated. Various driving factors—energy structure, energy intensity, activity effect, and population effect—have been taken into account. The study presents a comprehensive analysis of CO₂ emissions in the OECD area. To identify factors influencing CO₂ emissions in the sample, the LMDI approach was used. Then, the decoupling state of economic growth and CO₂ emissions was assessed.

The results of the DA show that, in the first part of the time span (the sub-periods 1990–1994, 1995–1999, and 2000–2004), the activity effect and the population effect represent the main driving factors leading to a rise in CO₂ emissions, more than counterbalancing the remaining three effects. A sensible change is observed since the 2005–2009 years, when the total emissions started to decline thanks to a massive impact of the energy intensity effect, which—together with energy structure and carbon intensity effects—were able to more than compensate for the activity and the population effects.

Our empirical findings are not directly comparable with the existing literature, as the effects that decompose the CO₂ emissions are distinct. Nevertheless, similar findings can be detected in several previous papers like: Zhang et al. (2011), Mousavi et al. (2017), Zhao et al. (2017), Li et al. (2018), Song et al. (2018), Parker and Bhatti (2020), Ran et al. (2019), Ari et al. (2020), Cai et al. (2020), Jiang et al. (2020), Nieto et al. (2020), Kamellos et al. (2021), Zhang et al. (2021), Ruiz et al. (2022), and Xu et al. (2022), who stated that economic activity has a considerable strong effect on CO₂ emissions.

Regarding the decoupling analysis, according to our estimates, a great majority of the countries experienced an expansive negative decoupling phase over the 1990–2004 period; afterward, several countries shifted to a strong decoupling regime for the following ten years. Finally, as a result of the economic-financial crisis, the expansive negative decoupling returned to prevailing. In the same line, Song et al. (2018) affirmed that most of the OECD members were in a recessive decoupling state from 2001 to 2015.

DA and decoupling analyses suggest relevant policy recommendations. In fact, the relevance of the effect of the main driving factors capable of lowering CO₂ emissions emerges (as a consequence of the energy intensity effect and the population effect). This is especially urgent in countries that are lagging behind in the decarbonization process to obtain energy and in improving energy efficiency. To the decoupling of CO₂ emissions from economic activity, the energy efficiency effect might represent a significant and beneficial factor; therefore, it is useful to continue efforts to increase energy efficiency in all OECD countries. Moreover, further political support for digitalization and decentralized energy systems, as well as for the creation of energy communities capable of increasing local electricity supply, would be relevant; indeed, these systems and institutions allow various advantages such as the more efficient use of renewable energies, thus managing to favor decarbonization. Along with these recommendations, a unique carbon trading system in a form of carbon pricing can be an effective approach to limit climate change (Magazzino, 2016a).

Although this study tried to present a comprehensive study of CO₂ emissions in the OECD countries, further investigations are necessary to understand the sample’s emissions in a detailed way. The limitations of the study are related to the set of variables used, since a wider selection of series on socio-economic and environmental characteristics may allow more in-depth analyses. Future research may address alternative socio-economic indexes, such as urbanization rate, industrial development, education level, and population structure; evaluate different pollutant agents, such as CH₄ and N₂O; expand the investigation to sub-sectors, i.e. agriculture, industry, service, and transportation sectors.

Appendix

See Table 4.

Table 4 Summary of the studies on decomposition and decoupling analysis