Energy Related CO 2 Emissions before and after the Financial Crisis

: The 2008–2009 ﬁnancial crisis, often referred to as the Great Recession, presented one of the greatest challenges to economies since the Great Depression of the 1930s. Before the ﬁnancial crisis, and in response to the Kyoto Protocol, many countries were making great strides in increasing energy e ﬃ ciency, reducing carbon dioxide (CO 2 ) emission intensity and reducing their emissions of CO 2 . During the ﬁnancial crisis, CO 2 emissions declined in response to a decrease in economic activity. The focus of this research is to study how energy related CO 2 emissions and their driving factors after the ﬁnancial crisis compare to the period before the ﬁnancial crisis. The logarithmic mean Divisia index (LMDI) method is used to decompose changes in country level CO 2 emissions into contributing factors representing carbon intensity, energy intensity, economic activity, and population. The analysis is conducted for a group of 19 major countries (G19) which form the core of the G20. For the G19, as a group, the increase in CO 2 emissions post-ﬁnancial crisis was less than the increase in CO 2 emissions pre-ﬁnancial crisis. China is the only BRICS (Brazil, Russia, India, China, South Africa) country to record changes in CO 2 emissions, carbon intensity and energy intensity in the post-ﬁnancial crisis period that were lower than their respective values in the pre-ﬁnancial crisis period. Compared to the pre-ﬁnancial crisis period, Germany, France, and Italy also recorded lower CO 2 emissions, carbon intensity and energy intensity in the post-ﬁnancial crisis period. Germany and Great Britain are the only two countries to record negative changes in CO 2 emissions over both periods. Continued improvements in reducing CO 2 emissions, carbon intensity and energy intensity are hard to come by, as only four out of nineteen countries were able to achieve this. Most countries are experiencing weak decoupling between CO 2 emissions and GDP. Germany and France are the two countries that stand out as leaders among the G19.


Introduction
The 2008-2009 financial crisis, often referred to as the Great Recession, presented one of the greatest challenges to economies since the Great Depression of the 1930s. While the origins of the recession were with the collapse of the US subprime mortgage business, the close connection between banks in a global financial network facilitated the quick spread of the recession to other countries [1,2]. Before the financial crisis, and in response to the Kyoto Protocol, many countries were making great strides in increasing energy efficiency, reducing carbon dioxide (CO 2 ) emission intensity and reducing their emissions of CO 2 . Under the Kyoto Protocol, carbon emissions trading was expanding, and the clean development mechanism offered emerging countries an attractive approach to pursue clean energy initiatives [3]. Some countries were pursuing aggressive energy policy to decarbonize their economies [4]. During the financial crisis, consumer spending declined, and firms reduced production. Access to financial capital became difficult and investment in capital-intensive projects declined. The reduction in economic activity resulted in a reduction in energy usage and CO 2 emissions fell.
to evaluate the change in carbon dioxide emissions and its driving factors over the pre-financial crisis and post-financial crisis periods. While there are other decomposition methods, the LMDI is preferred because it is not path dependent, leaves no residual, and can handle zero values [18]. Country level CO 2 emissions are decomposed into factors for carbon intensity, energy intensity, economic activity, and population. The pre-financial period covers the years 2000-2007 and the post-financial period covers the years 2010-2017. This is consistent with the NBER dating of the US recession from December of 2007 to June of 2009 [19].
The results from this analysis reveal several important findings. For the G19, as a group, the increase in CO 2 emissions post-financial crisis was less than the increase in CO 2 emissions pre-financial crisis. Developed countries, as a group, experienced a reduction in carbon dioxide emissions over the post-financial period, while emerging countries experienced an increase. There are, however, differences between countries. For eight countries (two of which are developed), the change in CO 2 emissions after the financial crisis is greater than before the financial crisis. For these countries, improvements in carbon intensity and energy intensity were not enough to overcome the increases in CO 2 emissions resulting from increases in economic activity and population. China is the only BRICS (Brazil, Russia, India, China, South Africa) country to record changes in CO 2 emissions, carbon intensity and energy intensity in the post-financial crisis period that were lower than their respective values in the pre-financial crisis period. China, however, accounts for 72% of post-financial crisis emissions. Compared to the pre-financial crisis period, Germany, France, and Italy also recorded lower CO 2 emissions, carbon intensity and energy intensity in the post-financial crisis period. Germany's success comes from cross-partisan policy consistency, shared goals between political leaders and renewable energy advocates, a strong social movement for renewable energy, and decentralized energy policies [20]. France's reliance on nuclear energy for electricity generation has helped to reduce carbon emissions. Germany and Great Britain are the only two countries to record negative changes in CO 2 emissions over both periods. Continued improvements in reducing CO 2 emissions, carbon intensity and energy intensity are hard to come by, as only four out of nineteen countries were able to achieve this. Most countries are experiencing weak decoupling between CO 2 emissions and GDP. Germany and France are the two countries that stand out as leaders among the G19. Great Britain, the United States, and South Africa have also shown strong performance in reducing carbon dioxide emissions and energy intensity in the post-financial crisis period. A post-financial crisis world of declining CO 2 emissions is essential if the 2 degrees Celsius target of limiting global temperature increases agreed to at the Paris Climate Change Agreement is to be achieved [21]. If the G20 is going to be a global leader on climate change then they need a non-partisan energy policy and the role of environmental ministers must be expanded. This paper is organized as follows. The following sections of the paper set out the methods and data, results, discussion and conclusions.

The Logarithmic Mean Divisia Index (LMDI) Method
Index decomposition analysis (IDA) is a popular approach to analysing the driving factors behind trends in energy consumption and greenhouse gas (GHG) emissions [18,22]. IDA can be implemented using either the Laspeyres index or the Divisia index. Over the past 20 years, the LMDI has been the most preferred approach to IDA. The LMDI is desirable because it does not contain an unexplained residual term, handles zero values, and is consistent in aggregation. The LMDI method can be used to study changes in GHG emissions at both the theoretical and practical level. The analysis in this paper is carried out using the LMDI method. There are several recent research papers that use the LMDI method to study energy trends or CO 2 emissions trends at the country level for different groupings (world, OECD, European Union) of countries. The literature below is organized according to the size (largest to smallest) of the country groups.
Dong et al. [10] use the LMDI to decompose CO 2 emissions into factors for carbon intensity, energy intensity, economic activity, and population. The analysis uses a balanced panel data set of 130 countries for the years 1997-2015. In order to investigate the effect of the financial crisis, analysis was also conducted over two subsamples (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2009)(2010)(2011)(2012)(2013)(2014)(2015). Economic activity has been the biggest driver behind increasing CO 2 emissions, while energy intensity has been the biggest driver behind declining emissions. While emissions from the high-income countries have been stable after the financial crisis, emissions from the lower-middle-income countries have been increasing. Goh et al. [23] use multidimensional index decomposition to analyse the factors behind global aggregate carbon intensity (ACI) from electricity production. Using data from 1990 to 2014, they find that the level of ACI has changed very little over this time period. There has been a shift in electricity production from developed to developing countries and this geographical shift has offset improvements in power generation efficiency. One approach to addressing this issue is to increase the share of renewables used in electricity generation but this is difficult, since many developing countries primarily rely on fossil fuels for electricity generation.
Yao et al. [24] use a LMDI decomposition to analyse the driving forces behind CO 2 emissions in the G20. The data set covers the years 1971-2010. Economic growth was the main factor increasing CO 2 emissions. Decreasing energy intensity and emission intensity were major contributors to reduced CO 2 emissions. The impacts of these factors varied by country.
Chen et al. [25] use the LMDI to study the decomposition of CO 2 emissions in the OECD. The change in CO 2 emissions is decomposed into factors representing CO 2 intensity of fossil fuel energy, energy consumption structure, energy intensity, per capita GDP, population distribution, and size of the population. Per capita GDP and population size are the main factors behind increasing CO 2 emissions. Energy intensity is the main factor behind decreasing CO 2 emissions. The other factors have a relatively small effect on the decomposition of OECD CO 2 emissions. A Tapio decoupling analysis shows that different decoupling states apply to different time periods. In general, technical factors have a greater impact on the decoupling elasticity.
There has been a lot of research on decomposing CO 2 emissions in European countries. Bhattacharyya and Matsumura [26] use the LMDI to analyse the reduction in greenhouse gas emissions in 15 European Union countries over the period 1990-2007. They find that emission intensity has reduced in both energy related activities and other processes at the aggregate level. At the country level, the trend in greenhouse gas emissions varies. Changes in fuel mix, a reduction in energy intensity, and reduction in emission intensity are the main drivers behind the reduction in GHG emissions. Kopidou et al. [27] use the LMDI to study the drivers of CO 2 emissions and employment in the industrial sector of a selection of European Union countries. The time frame for the study is divided into two periods (2000-2007 and 2007-2011) corresponding to the pre-financial crisis and post-financial crisis periods. In both periods, the main drivers of industrial CO 2 emissions and employment were economic growth and resource intensity. Changes in the fuel mix were especially beneficial in reducing carbon dioxide emissions in the post-financial crisis period. Moutinho et al. [28] use the LMDI to identify the factors influencing CO 2 emissions in four groups (eastern, western, northern, southern) of European countries. Carbon dioxide emissions are decomposed into six effects representing carbon intensity, energy mix, energy intensity, renewable energy productivity, renewable energy capacity, and population. Analysis is conducted over the pre-Kyoto period and the post-Kyoto period. The post-Kyoto period shows a reduction in carbon dioxide emissions. This is mainly driven by changes in the energy mix and a reduction in the usage of fossil fuels.
Chapman et al. [29] uses the LMDI to analyse drivers of CO 2 emissions in six Northeast Asian economies. The analysis was conducted for the years 1991-2015. CO 2 emissions are decomposed into factors for carbon intensity, fuel mix, energy intensity, economic activity, and population. CO 2 emissions are driven by economic growth in China and Korea and energy efficiency in Russia and the Democratic People's Republic of Korea. For Japan and Mongolia, no one factor clearly dominates. Regional policy focused on one factor is unlikely to be successful, since the driving factors behind CO 2 emissions varies by country.
Su et al. [30] use index decomposition (Shapley value) to compare the drivers of carbon dioxide emissions between the G7 countries and the BRICS countries. Analysis was conducted over the period 1990-2015. For the G7 countries, energy intensity was the main factor in reducing carbon dioxide emissions. The energy intensity effect was much lower for the BRICS countries and was offset by increases in economic activity. Wang and Jiang [31] use the LMDI to investigate decoupling between economic growth and carbon dioxide emissions in the BRICS countries. They start with the standard Kaya identity that specifies carbon emissions equal the product of carbon intensity, energy intensity, GDP per capita, and population and then specify a Cobb-Douglas production function for GDP. This yields a total of six decomposition factors (carbon intensity, energy intensity, energy structure, technology, labour and capital). The Tapio decoupling method is used to estimate decoupling elasticities. The impact of the decomposition factors and the state of decoupling change across time but there are some major findings. Russia and South Africa show a long period of strong decoupling. Energy intensity improvements are one of the major drivers behind Russia's decoupling. For South Africa, energy structure, energy intensity, and carbon intensity are major drivers behind the decoupling of economic growth and carbon dioxide emissions. China's decoupling is characterized as expansive decoupling, while India's decoupling is characterized as weak decoupling. For China and India, the investment effect and energy intensity were the two largest drivers. Brazil is characterized as negative decoupling. Investment, labour, and technical drivers are the main factors behind Brazil's decoupling.
There are many papers that compare the decomposition of CO 2 emissions among non-standard groupings of countries. Lima et al. [32] look into the main drivers of CO 2 emissions in Portugal, United Kingdom, Brazil, and China using LMDI decomposition. Energy intensity, affluence, and renewable energy sources are the main drivers of CO 2 emission change. For Brazil and Portugal, energy intensity is particularly important, while for China and the UK, increasing renewable fuel mix is most important for reducing CO 2 emissions. Lima et al. [33] use the LMDI to study trends in energy consumption for three developed countries (United Kingdom, Portugal, and Spain) and three emerging countries (Brazil, China, and India). The change in energy consumption is decomposed into components representing activity, structure, and intensity. Intensity and activity are the two major factors driving changes in energy consumption. In the case of Brazil, China, and India, energy intensity has been declining but is not enough to offset the increased activity which results in an increase in energy consumption. Moutinho et al. [34] use the LMDI to study the factors affecting CO 2 emissions in 23 developed and developing countries that rank high in the usage of renewable energy. CO 2 emissions are decomposed into factors for carbon trade intensity, trade of fossil fuels, fossil fuel intensity, renewable energy productivity, the electricity financial power effect, and the financial development effect. The impact of each of these factors changes across time and for countries. For most years and for most countries, the electricity financial power effect was the main contributing factor to the increase in CO 2 emissions, while the renewable source productivity effect was the main contributor towards a decrease in CO 2 emissions. Rustemoglu and Rodriques Andres [35] use decomposition analysis to study the drivers behind CO 2 emissions in Brazil and Russia for the period 1992-2011. They decompose CO 2 emissions into factors for carbon intensity, energy intensity, economic activity, and population. For Brazil, economic activity and population are major drivers in increasing carbon dioxide emissions. For Russia, economic activity increased emissions, while improvements in energy intensity reduced emissions. For both countries, carbon intensity had a small impact on carbon emissions. Streimikiene and Balezentis [36] use index decomposition (Shapley value) to determine the most important factors driving greenhouse gas emissions in the Baltic States. Energy efficiency is the most important factor behind reducing GHG emissions. Although renewable energy usage has increased significantly over the time period under study, it is still a relatively small factor behind reducing GHG emissions. Sun et al. [37] use the LMDI to study the carbon footprint in energy consumption, referred to as ecological pressure, for 56 countries over the years 1994-2014 involved in China's Belt and Road initiative. Growth in ecological pressure has slowed across time. Resource-rich countries have a greater ecological pressure. Energy structure, international trade, and energy intensity can reduce ecological pressure. Zhang et al. [38] use the LMDI to study the decomposition of CO 2 emissions in ten Asian countries (Brunei, Cambodia, China, Indonesia, Malaysia, Myanmar, the Philippines, Singapore, Thailand, and Vietnam). For most countries, the time period covers the years 1990-2014. The main results are that increases in economic activity increase carbon dioxide emissions, while decreases in energy intensity decrease carbon emissions.
In summary, the literature studying the decomposition of CO 2 emissions at the country level finds a reduction in energy intensity to be one of the most important drivers of reducing carbon dioxide emissions, while increases in economic activity are usually the largest factor increasing carbon dioxide emissions.

The Empirical Model
Decomposition analysis is useful for understanding the interactions between carbon dioxide emissions and socio-economic activities. The Kaya identity is an equation that is used to decompose CO 2 emissions into important driving factors [39,40]. Using a Kaya identity, country level CO 2 emissions can be decomposed into four main factors representing carbon intensity, energy intensity, economic activity, and population [38,41].
The symbol E represents energy consumption, GDP represents gross domestic product, and POP represents population. Equation (1) can be expressed as: In Equation (2) CI = CO 2 /E is carbon intensity, EI = E/GDP is energy intensity, EA = GDP/POP is economic activity, and POP is the population. Based on this representation, the LMDI can be used to decompose carbon dioxide emission changes between a base year (0) and a target year (T) using either an additive or multiplicative decomposition. The LMDI additive decomposition for ∆CO 2 = CO 2T − CO 20 in each country is given in Equation (3) [18].
The equation for each component in the additive decomposition is: The LMDI multiplicative decomposition DCO 2 = CO 2T /CO 20 is given in Equation (8) [18].
The equation for each component in the multiplicative decomposition is:

Tapio Decoupling
In the literature on economic growth and the environment, decoupling refers to the relationship between CO 2 emissions and economic growth. Tapio [42] developed a decoupling analysis, based on elasticity calculations, to study decoupling between European economic growth, transportation, and greenhouse gas emissions. The decoupling elasticity between CO 2 emissions and GDP is written as: Decoupling is classified according to eight states: strong decoupling, weak decoupling, expansive decoupling, expansive negative decoupling, strong negative decoupling, weak negative decoupling, recessive coupling, and recessive decoupling. From an environmental perspective, strong decoupling is most desirable because economic growth is positive and carbon dioxide emission growth is negative. Table 1 shows the detailed criteria for determining these states. Tapio decoupling has been widely applied [25,31,[43][44][45][46][47][48][49][50].

Data
Country level data on CO 2 emissions, GDP, population (POP), and energy usage (E) are required for the analysis. Data on GDP (real GDP in millions of 2011 US dollars: gdpna) and population (in millions) come from the Penn World Tables (PWT 9.1) [51]. Data on CO 2 emissions (MMtons) from the consumption of energy and energy consumption (Quad Btus) come from the United States Energy Information Agency (US EIA) [52]. CO 2 emissions from the consumption of energy include emissions that result from the consumption of petroleum, natural gas, and coal and from natural gas flaring. Total energy consumption includes coal, natural gas, petroleum and other liquids, nuclear, renewables and other. The 19 countries included in the study are Argentina, Australia, Brazil, Canada, China, France, Germany, India, Indonesia, Italy, Japan, South Korea, Mexico, Russia, Saudi Arabia, South Africa, Turkey, Great Britain, and the United States. These 19 countries along with the European Union form the group of countries known as the G20. The data set covers the years 2000-2017.
CO 2 emissions for most of the countries show a relatively flat pattern between 2000 and 2017 ( Figure 1). Some countries like India experienced a large increase in emissions over the time period. China experienced a huge increase in emissions up to 2013 (10,800 million tonnes of CO 2 ), after which emissions declined due in part to changes in economic structure ( Figure 2) [53]. US carbon emissions peaked at just about 6000 million tonnes of CO 2 in 2007 and have been experiencing a steady decrease in emissions since then due to fuel switching in power generation (coal to natural gas and renewables) and a decrease in industrial fuel use ( Figure 2).  Summary statistics for CO2, CO2/E, E/GDP, GDP/POP, and POP indicate that CO2 emissions are, on average, higher after the financial crisis (Table 2). In addition, GDP per capita and population are, on average, greater after the financial crisis. On average, there has been a slight reduction in carbon intensity and energy intensity between the pre-financial crisis and post-financial crisis period. In the    Summary statistics for CO2, CO2/E, E/GDP, GDP/POP, and POP indicate that CO2 emissions are, on average, higher after the financial crisis (Table 2). In addition, GDP per capita and population are, on average, greater after the financial crisis. On average, there has been a slight reduction in carbon intensity and energy intensity between the pre-financial crisis and post-financial crisis period. In the  Summary statistics for CO 2 , CO 2 /E, E/GDP, GDP/POP, and POP indicate that CO 2 emissions are, on average, higher after the financial crisis (Table 2). In addition, GDP per capita and population are, on average, greater after the financial crisis. On average, there has been a slight reduction in carbon intensity and energy intensity between the pre-financial crisis and post-financial crisis period. In the case of CO 2 emissions, it seems that the reductions in carbon intensity and energy intensity have not been enough to compensate for the increase in CO 2 emissions coming from increased population and GDP per capita.

The LMDI Analysis
This section presents results from the LMDI analysis. For each country, two decompositions are performed. The first decomposition covers the pre-financial crisis years 2000-2007 and the second decomposition covers the post-financial crisis years 2010-2017. Results from the additive decomposition show that for the G19, as a group, the change in CO 2 emissions post-financial crisis was less than the change in CO 2 emissions pre-financial crisis ( Table 3). The biggest change was for the developed countries (Australia, Canada, France, Germany, Great Britain, Italy, Japan, and the US), where the change in CO 2 emissions in the post-financial crisis period was negative. For both developed countries and emerging countries, carbon intensity and energy intensity showed decreases in the post-financial crisis period. For emerging countries, the population effect was greater in the post-financial crisis period.
Country groupings can, however, mask individual country trends. For Argentina, Australia, Brazil, Canada, India, Indonesia, Saudi Arabia, and Turkey, the change in CO 2 emissions post-financial crisis was greater than the change in CO 2 emissions pre-financial crisis (Table 3). In other words, these countries experienced increases in CO 2 emissions over the period 2010-2017 that were greater than over the period 2000-2007. For example, in the case of Argentina, carbon emissions increased by 20.6 MMtons over the pre-financial crisis period and 21.9 MMtons over the post-financial crisis period. Notice that two of these countries are developed countries and major commodity exporters. Australia and Canada are similar in many ways, since both are small open economies and major commodity exporters (coal in the case of Australia and oil in the case of Canada). Australia has not experienced a recession since 1991 and was relatively unaffected by the financial crisis due to government stimulus spending and commodity (coal) exports. Canada was affected by the financial crisis, although not to the extent that the US was. After the financial crisis, operations in the oil sands ramped up production, which is a major source of carbon dioxide emissions. The oil sands account for approximately 9% of Canada's GHG emissions [54]. Even though China recorded an increase in CO 2 emissions over the post-financial crisis that was less than over the pre-financial crisis, the post-financial crisis value of 1640 MMtons was the largest increase in CO 2 emissions post-financial crisis for any country studied. The next greatest increase in carbon dioxide emissions post-financial crisis was 706 MMtons (India).
Germany and Great Britain experienced decreases in CO 2 emission over both subsamples, which indicates strong continued support for reducing carbon dioxide emissions. In the case of Germany, crisis period. The largest factor for decreasing carbon emissions in Great Britain is a cleaner fuel mix in electricity generation, as coal was switched for natural gas and renewables [55]. Reduced fuel consumption by business and industry also contributed to the reduction in carbon dioxide emissions.
Among the BRICS countries, China, Russia, and South Africa recorded an increase in CO 2 emissions post-financial crisis that was smaller than over the pre-financial crisis period. Brazil and India recorded increases. Brazil experienced a decrease in emissions over the pre-financial period and then recorded a large increase in emissions over the post-financial crisis period. The post-financial crisis period encompassed the 2014 recession, which included the impeachment of the former president, political dissatisfaction, and a softening of commodity exports. Agriculture and activities associated with illegal deforestation contributed to the increase in post-financial crisis carbon dioxide emissions. In addition, the Brazilian government reduced environmental regulations in exchange for political support [56].
As an example of how to interpret the results in Table 3 The carbon intensity effect is smaller in the post-financial crisis period for Australia, China, Germany, France, Great Britain, Italy, Russia, Turkey, and the USA ( Table 3). The energy intensity effect is smaller in the post-financial crisis period for China, German, France, India, Indonesia, Italy, Japan, Mexico, Saudi Arabia, Turkey, and South Africa. China is the only BRICS country to record changes in CO 2 emissions, carbon intensity and energy intensity in the post-financial crisis period that were lower than their respective values in the pre-financial crisis period. Compared to the pre-financial crisis period, Germany, France, and Italy also recorded lower CO 2 emissions, carbon intensity and energy intensity in the post-financial crisis period. In the case of Italy, however, the population effect was negative, and the economic activity effect was positive but very small. The economic activity effect is smaller in the post-financial crisis period for Argentina, Australia, Brazil, Canada, France, Great Britain, Italy, Japan, South Korea, Russia, the United States, and South Africa. Most of these countries are developed countries. Brazil is the only country to record a negative economic effect in the post-financial crisis period, and this was due to the 2014 recession. The population effect is smaller in the post-financial crisis period for Canada, France, Great Britain, Italy, Japan, and the USA. These countries are all developed nations. In the post-financial crisis period, only Italy and Japan recorded negative values for the population effect. For most countries, reductions in energy intensity had a larger impact on reducing carbon dioxide emissions than reductions in carbon intensity, which is consistent with much of the published literature.
The Kaya identity is related to the IPAT model (I = CO 2 emissions, P = population, A = affluence (GDP per capita), T = technology (energy intensity)). As a robustness check a fixed effects regression was used to estimate an IPAT model in natural logarithm form for the pre-financial crisis and post-financial crisis periods. For each variable, in each period, the estimated coefficient is positive, close to unity, and statistically significant. This confirms the importance of population, affluence, and energy intensity in explaining CO 2 emissions. Figure 3 presents a graphical comparison of the difference in post-financial crisis and pre-financial crisis results from Table 3 for Argentina, Australia, Brazil and Canada. For each country, the 2007 values for each decomposition factor are subtracted from their respective 2017 values to determine which factor exhibited the greatest change between the two subperiods. For Argentina, the biggest factor was the decrease in economic activity. For Australia, the biggest factor was an increase in energy intensity. For Brazil and Canada, the biggest factor was an increase in carbon intensity. For China and Germany, the biggest factor was a decrease in energy intensity. For France and Great Britain, the biggest factor was a decrease in carbon intensity (Figure 4). For Indonesia, the biggest factor was a decrease in energy intensity ( Figure 5). For India, the biggest factor was an increase in economic activity. For Italy, the biggest factor was a decrease in carbon intensity. For Japan, the biggest factor was a decrease in energy intensity. For South Korea, the biggest factor was a decrease in economic activity ( Figure 6). For Mexico, the biggest factor was a decrease in energy intensity. For Russia, the biggest factor was an increase in energy intensity and for Saudi Arabia the biggest factor was a decrease in energy intensity. For Turkey, the biggest factor was an increase in economic activity (Figure 7). For the USA, the biggest factor was a decrease in carbon intensity. For South Africa, the biggest factor was a decrease in economic activity. In summary, when looking at the change in decomposition factors between the post-financial crisis and pre-financial crisis periods, carbon intensity was the most important factor affecting changes in carbon dioxide emissions for six countries, energy intensity was the most important factor for seven countries and economic activity was the most important factor for six countries. For China and Germany, the biggest factor was a decrease in energy intensity. For France and Great Britain, the biggest factor was a decrease in carbon intensity (Figure 4). For Indonesia, the biggest factor was a decrease in energy intensity ( Figure 5). For India, the biggest factor was an increase in economic activity. For Italy, the biggest factor was a decrease in carbon intensity. For Japan, the biggest factor was a decrease in energy intensity. For South Korea, the biggest factor was a decrease in economic activity ( Figure 6). For Mexico, the biggest factor was a decrease in energy intensity. For Russia, the biggest factor was an increase in energy intensity and for Saudi Arabia the biggest factor was a decrease in energy intensity. For Turkey, the biggest factor was an increase in economic activity (Figure 7). For the USA, the biggest factor was a decrease in carbon intensity. For South Africa, the biggest factor was a decrease in economic activity. In summary, when looking at the change in decomposition factors between the post-financial crisis and pre-financial crisis periods, carbon intensity was the most important factor affecting changes in carbon dioxide emissions for six countries, energy intensity was the most important factor for seven countries and economic activity was the most important factor for six countries.        LMDI multiplicative decomposition is another way of looking at the decomposition of CO2 emissions (Table 4). Values greater than unity indicate positive increases over the time period under  LMDI multiplicative decomposition is another way of looking at the decomposition of CO2 emissions (Table 4). Values greater than unity indicate positive increases over the time period under LMDI multiplicative decomposition is another way of looking at the decomposition of CO 2 emissions (Table 4). Values greater than unity indicate positive increases over the time period under consideration, while values less than unity indicated reductions over the time period. Multiplicative decomposition gives information on the direction of the changes in decomposition as well as the rates of change over specified time periods. In the case of Argentina, for example, CO 2 emissions increased between 2000 and 2007 by a factor of 1.148 and between 2010 and 2017 by a factor of 1.116. In the post-financial period, the carbon intensity component increased by a factor of 1.031, the energy intensity component increased by a factor of 0.996, the economic activity component increased by a factor of 1.012, and the population component increased by a factor of 1.074. The largest rates of change in CO 2 in the post-recession period were observed in Turkey (1.502), India (1.439), and Saudi Arabia (1.258). The smallest rates of change in CO 2 in the post-recession period were observed in Great Britain (0.795), Italy (0.842), and France (0.915). Great Britain, USA, and Australia had the smallest rates of change in the carbon intensity factor, while Japan, Indonesia, and Turkey had the largest increase. The smallest rates of change in the energy intensity factor occurred in Great Britain, Indonesia, and Japan. The largest rates of change in the energy intensity factor occurred in Brazil, Argentina, and Russia. The smallest rates of change in the economic activity factor occurred in Brazil, Italy, and Argentina, while the largest rates of change occurred in India, China, and Turkey. Japan, Italy, and Russia had the lowest rates of change in the population effect, while Saudi Arabia, Turkey, and Australia had the largest rates of change. In general, the economic activity factor and population factor are most important for large fast-growing developing countries.
For most countries, the largest contributor to CO 2 emission rates of change in the post-recession period was economic activity (13 countries) or population (5). Japan is the outlier, because the largest contrition to CO 2 emission rates of change comes from carbon intensity. This is due to how the 2011 Fukushima nuclear disaster altered the energy supply in Japan [57][58][59]. The Japanese nuclear power sector was shut down in 2012, but some reactors have been allowed to come back online since then. For 16 countries, the smallest contributor to CO 2 emission rates of change was energy intensity.

Decoupling Analysis
The result from Tapio decoupling between CO 2 and GDP are reported in Table 5. Consistent with the decomposition analysis, the changes are calculated over seven-year intervals (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017). Twenty entries are weak decoupling, nine are strong decoupling, five are expansive coupling, three are expansive negative decoupling, and one is recessive decoupling. Weak decoupling is by far the most popular state. Nine countries experienced a change in decoupling state between the pre-financial crisis period and the post-financial crisis period (Argentina, Brazil, China, France, Italy, Japan, Saudi Arabia, USA, and South Africa). Germany, France, Great Britain, Japan, the United States and South Africa experienced strong decoupling in the post-financial crisis period. Two countries that have experienced surprising shifts towards less decoupling are Brazil (strong decoupling pre-financial crisis to expansive negative decoupling post-financial crisis) and Italy (expansive negative decoupling pre-financial crisis to recessive decoupling post-financial crisis). Brazil's 2014 recession affected economic activity and policy related to energy and the environment. Italy has, in some ways, entered a "perma-recession". Massive government debt, a historic debt to GDP ratio, fear of systemic risk, and a cloudy political situation limits the ability of the government to focus on energy policy. Under European Union rules, Italy's debt problems and financial situation prevent government deficit spending to help grow the economy [60]. Notice that in the case of Italy, the 2017 CO 2 elasticity is negative and the GDP elasticity is negative, which is consistent with a slowdown in carbon dioxide emissions associated with a slowdown in economic activity. The green stimulus packages implemented by some countries may be one positive benefit stemming from the financial crisis [8] and may help to explain why some countries transitioned from weak decoupling to strong decoupling.

Discussion
In general, reductions in carbon intensity and energy intensity reduce CO 2 emissions, while increases in economic activity and population increase CO 2 emissions. This is consistent with the existing literature [10,24,25]. For most countries, reductions in energy intensity have had a bigger impact on reducing CO 2 emissions than reductions in carbon intensity, and this is true in both the pre-financial crisis period and post-financial crisis period. For most countries studied, increases in economic activity are the largest contributor to increasing CO 2 emissions, which is consistent with the existing literature [10,24]. Saudi Arabia is an exception, because the population effect was the greatest contributor to CO 2 emissions in both the pre-financial crisis and post-financial crisis periods. For each country, the energy intensity effect is negative in the post-financial crisis period except for Brazil. The 2014 recession in Brazil led to a negative economic effect and a positive energy intensity effect, with the population effect being the largest contributor to CO 2 emissions.
China is the only BRICS country to record changes in CO 2 emissions, carbon intensity and energy intensity in the post-financial crisis period that were lower than their respective values in the pre-financial crisis period. Fuel switching from fossil fuels to renewables, capital equipment modernization, industry sector realignment, and energy related technological improvements are important contributions to reducing carbon dioxide emissions in China [61]. Compared to the pre-financial crisis period, Germany, France, and Italy also recorded lower CO 2 emissions, carbon intensity and energy intensity in the post-financial crisis period, although these reductions were matched with a negative population effect and a negligible economic activity effect in the case of Italy. Continued improvements in reducing CO 2 emissions, carbon intensity and energy intensity are hard to come by, as only four out of nineteen countries were able to achieve this. These results are important, because reductions in carbon intensity and energy intensity are major factors in a country's transition to a low-carbon economy.
The decoupling analysis reveals some important results. Germany, France, Great Britain, Japan, the United States and South Africa experienced strong decoupling in the post-financial crisis period. Strong decoupling is important for environmental sustainability, because this means that carbon dioxide emissions are declining while the economy is growing. Most of the other countries are characterized as weak decoupling.
Germany and France are the two countries that stand out as leaders among the G19. In the post-financial crisis period, they have shown reductions in carbon emissions, carbon intensity and energy intensity and both are characterized as strong decoupling. These results are consistent with the findings from some recent policy studies. While designing an energy policy is one thing, maintaining the intended direction of the policy can be challenging across time. In the case of Germany, cross-partisan policy consistency, shared goals between political leaders and renewable energy advocates, a strong social movement for renewable energy, and decentralized energy policies have been major factors to Germany's continued success on transitioning to a low-carbon economy [20]. Nuclear power is the primary energy source in France's transition to a low-carbon economy [62]. France's almost sole focus on electricity development means that other sectors like transportation have been neglected. In France, oil remains an important fuel for transportation under most future energy consumption scenarios, but this may change if there is widespread adoption of electric vehicles. The results of this present paper are also consistent with those of Midova et al. [63], who study low-carbon scenarios of six north-west European countries (Netherlands, Germany, France, Denmark, the UK, and Belgium). In ranking these countries on ten criteria regarding low-carbon energy scenario design, Germany comes out on top, followed by the UK. France's heavy reliance on nuclear power makes it less receptive to technological innovation for other renewable energy sources. Results from this present study show that Great Britain, the United States, and South Africa have also shown strong performance in reducing carbon dioxide emissions and energy intensity in the post-financial crisis period.
Among the developed countries, Australia and Canada are laggards. These countries are small open-economy natural resource-intensive exporters that have increased CO 2 emissions post-financial crisis. For Australia, the biggest change between the pre-financial period and post-financial period was an increase in energy intensity. For Canada, the biggest change was an increase in carbon intensity. For these countries, a reliance on the natural resource extraction (coal and oil) sector make it difficult to decrease carbon dioxide emissions. Den Elzen et al. [56] find that by comparing projected GHGs with 2030 target emissions, both Australia and Canada are unlikely to meet their nationally determined contributions (NDCs). Further increases in reducing carbon intensity and energy intensity can be obtained by decarbonizing energy supply and increasing energy efficiency in transportation, buildings, and industry. Decarbonizing energy supply is, however, at odds with maintaining fossil fuel exports.
Argentina, Brazil, India, Indonesia, Saudi Arabia, and Turkey are developing countries that experienced increases in CO 2 emissions over the period 2010-2017 that were greater than that over the period 2000-2007. Argentina and Brazil are the two countries experiencing CO 2 emission growth in the post-financial crisis period greater than GDP growth. Both of these countries have unusual economic situations (a deep recession in Brazil and a debt crisis in Argentina), which creates problems for developing and implementing clean energy policy. For these countries, a return to more orderly economic conditions must first precede a commitment to energy policy focusing on reducing carbon intensity and energy intensity. Of the BRICS countries, China, Russia, and South Africa recorded lower increases in CO 2 emissions post-financial crisis. China's reductions in carbon intensity and energy intensity post-financial crisis are impressive but China's carbon dioxide emission increases over this period are the largest of the countries studied and due to the large increases in economic activity and population. Nevertheless, Den Elzen et al. [56] find that China is on track to meet its NDC targets for GHG emissions. China, through a series of five-year plans, has ambitious targets to lower carbon intensity and energy intensity [64]. India, like China, suffers high vulnerability from climate change, but India has no strong numeric targets for their energy policy unlike China [64]. Sarangi et al. [65] find that the Indian government's feed-in tariff policy has been successful in improving electricity accessibility and diversifying the electricity supply.

Conclusions and Policy Implications
The focus of this research is to study how energy related CO 2 emissions and their driving factors after the financial crisis compare to the period before the financial crisis. The logarithmic mean Divisia index (LMDI) method is used to decompose changes in country level CO 2 emissions into contributing factors representing carbon intensity, energy intensity, economic activity, and population. The analysis is conducted for a group of 19 major countries which form the core of the G20. The LMDI decomposition analysis reveals some important results. For the G19, as a group, the increase in CO 2 emissions post-financial crisis was less than the increase in CO 2 emissions pre-financial crisis. A decrease in energy intensity was the biggest driver behind reducing CO 2 emissions, while an increase in economic activity was the biggest factor increasing CO 2 emissions. There are, however, differences between countries. Germany and Great Britain are the only two countries to record negative changes in CO 2 emissions over both periods. Germany and France are the two countries that stand out as leaders among the G19 because, compared to the pre-financial crisis period, they recorded lower CO 2 emissions, carbon intensity and energy intensity in the post-financial crisis period and both are characterized as strong decoupling. Together, these two countries accounted for a reduction of 56.95 million metric tons in CO 2 emissions in the post-financial crisis period. This value was very similar to Brazil's increase in emissions (56.25 MMtons) during the same period. Argentina, Australia, Brazil, Canada, India, Indonesia, Saudi Arabia, and Turkey, however, experienced increases in CO 2 emissions over the period 2010-2017 that were greater than over the period 2000-2007. For Argentina, Brazil, Canada, and Saudi Arabia, the biggest factor behind this increase was an increase in carbon emission intensity. For Australia, the biggest factor was an increase in energy intensity. For India, Indonesia and Turkey, the biggest factor was an increase in economic activity. Reducing CO 2 emissions in Australia, Canada, and Saudi Arabia is challenging because these countries are major exporters of fossil fuels. For these countries, a successful sustainable energy policy requires a decoupling of the fossil fuel export sector from the rest of the economy. An energy policy recommendation for Argentina, Brazil, India, Indonesia and Turkey is to set and maintain stronger NDC goals.
While the main focus of this paper is to compare carbon dioxide emissions post-financial crisis with those pre-financial crises for the G19 countries it is now apparent that a new crisis is unfolding. This time, however, the crisis is a health crisis. It is difficult to compare the 2008-2009 financial crisis with the current COVID-19 pandemic because the pandemic is a health crisis with far greater impacts than what was experienced during the financial crisis. Since COVID-19 has wider ranging impacts than the financial crisis, the impact on reducing CO 2 emissions will be greater and could affect weather patterns.
In summary, 11 of the 19 countries studied showed increases in carbon dioxide emissions over the post-financial crisis period that were lower than over the pre-financial crisis period. This is a step in the right direction for meeting NDCs and global climate change policies but, given the importance of the G19 to the world economy, all G19 countries must act as leaders in the fight against climate change and reduce their emissions. There are several policy implications. The goal should be for all G19 countries to exhibit strong decoupling but only six countries showed strong decoupling in the post-financial crisis period. The chair of the G20 rotates on a yearly basis, which means the G20 agenda for each year depends upon the interests of the chair. An important policy recommendation is for a non-partisan G20 energy policy focused on continuous improvements in reducing carbon intensity and energy intensity through fuel switching and changes in industry structure because this is crucial for future reductions in carbon dioxide emissions. A further policy recommendation is for the G20 environmental ministers to be allowed greater involvement in decision making [66]. The G20 can be a leader in meeting the Paris Climate Change Agreement but all countries must be fully committed to reducing carbon dioxide emissions and cooperate towards this shared goal.
Funding: This research received no external funding.